diff options
Diffstat (limited to 'src/mesa/drivers/dri/i915')
80 files changed, 17282 insertions, 11085 deletions
diff --git a/src/mesa/drivers/dri/i915/Makefile b/src/mesa/drivers/dri/i915/Makefile index 805abf75e0..b9328a48f3 100644 --- a/src/mesa/drivers/dri/i915/Makefile +++ b/src/mesa/drivers/dri/i915/Makefile @@ -7,16 +7,6 @@ LIBNAME = i915_dri.so MINIGLX_SOURCES = server/intel_dri.c DRIVER_SOURCES = \ - i915_context.c \ - i915_debug.c \ - i915_fragprog.c \ - i915_metaops.c \ - i915_program.c \ - i915_state.c \ - i915_tex.c \ - i915_texprog.c \ - i915_texstate.c \ - i915_vtbl.c \ i830_context.c \ i830_metaops.c \ i830_state.c \ @@ -24,27 +14,60 @@ DRIVER_SOURCES = \ i830_tex.c \ i830_texstate.c \ i830_vtbl.c \ + intel_render.c \ + intel_regions.c \ + intel_buffer_objects.c \ intel_batchbuffer.c \ + intel_mipmap_tree.c \ + i915_tex_layout.c \ + intel_tex_layout.c \ + intel_tex_image.c \ + intel_tex_subimage.c \ + intel_tex_copy.c \ + intel_tex_validate.c \ + intel_tex_format.c \ + intel_tex.c \ + intel_pixel.c \ + intel_pixel_copy.c \ + intel_pixel_read.c \ + intel_pixel_draw.c \ + intel_buffers.c \ + intel_blit.c \ + i915_tex.c \ + i915_texstate.c \ + i915_context.c \ + i915_debug.c \ + i915_debug_fp.c \ + i915_fragprog.c \ + i915_metaops.c \ + i915_program.c \ + i915_state.c \ + i915_vtbl.c \ intel_context.c \ + intel_decode.c \ intel_ioctl.c \ - intel_pixel.c \ - intel_render.c \ intel_rotate.c \ intel_screen.c \ intel_span.c \ intel_state.c \ - intel_tex.c \ - intel_texmem.c \ - intel_tris.c + intel_tris.c \ + intel_fbo.c \ + intel_depthstencil.c \ + intel_bufmgr_ttm.c C_SOURCES = \ $(COMMON_SOURCES) \ + $(COMMON_BM_SOURCES) \ $(DRIVER_SOURCES) ASM_SOURCES = - +DRIVER_DEFINES = -I../intel $(shell pkg-config libdrm --atleast-version=2.3.1 \ + && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP") include ../Makefile.template +intel_decode.o: ../intel/intel_decode.c +intel_tex_layout.o: ../intel/intel_tex_layout.c + symlinks: diff --git a/src/mesa/drivers/dri/i915/i830_context.c b/src/mesa/drivers/dri/i915/i830_context.c index 7ca601e1b5..2ff8621c42 100644 --- a/src/mesa/drivers/dri/i915/i830_context.c +++ b/src/mesa/drivers/dri/i915/i830_context.c @@ -38,37 +38,38 @@ * Mesa's Driver Functions ***************************************/ -static const struct dri_extension i830_extensions[] = -{ - { "GL_ARB_texture_env_crossbar", NULL }, - { NULL, NULL } +static const struct dri_extension i830_extensions[] = { + {"GL_ARB_texture_env_crossbar", NULL}, + {NULL, NULL} }; -static void i830InitDriverFunctions( struct dd_function_table *functions ) +static void +i830InitDriverFunctions(struct dd_function_table *functions) { - intelInitDriverFunctions( functions ); - i830InitStateFuncs( functions ); - i830InitTextureFuncs( functions ); + intelInitDriverFunctions(functions); + i830InitStateFuncs(functions); + i830InitTextureFuncs(functions); } -GLboolean i830CreateContext( const __GLcontextModes *mesaVis, - __DRIcontextPrivate *driContextPriv, - void *sharedContextPrivate) +GLboolean +i830CreateContext(const __GLcontextModes * mesaVis, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate) { struct dd_function_table functions; - i830ContextPtr i830 = (i830ContextPtr) CALLOC_STRUCT(i830_context); - intelContextPtr intel = &i830->intel; + struct i830_context *i830 = CALLOC_STRUCT(i830_context); + struct intel_context *intel = &i830->intel; GLcontext *ctx = &intel->ctx; - GLuint i; - if (!i830) return GL_FALSE; + if (!i830) + return GL_FALSE; - i830InitVtbl( i830 ); - i830InitDriverFunctions( &functions ); + i830InitVtbl(i830); + i830InitDriverFunctions(&functions); - if (!intelInitContext( intel, mesaVis, driContextPriv, - sharedContextPrivate, &functions )) { + if (!intelInitContext(intel, mesaVis, driContextPriv, + sharedContextPrivate, &functions)) { FREE(i830); return GL_FALSE; } @@ -77,48 +78,27 @@ GLboolean i830CreateContext( const __GLcontextModes *mesaVis, intel->ctx.Const.MaxTextureImageUnits = I830_TEX_UNITS; intel->ctx.Const.MaxTextureCoordUnits = I830_TEX_UNITS; - intel->nr_heaps = 1; - intel->texture_heaps[0] = - driCreateTextureHeap( 0, intel, - intel->intelScreen->tex.size, - 12, - I830_NR_TEX_REGIONS, - intel->sarea->texList, - (unsigned *) & intel->sarea->texAge, - & intel->swapped, - sizeof( struct i830_texture_object ), - (destroy_texture_object_t *)intelDestroyTexObj ); - - /* FIXME: driCalculateMaxTextureLevels assumes that mipmaps are tightly - * FIXME: packed, but they're not in Intel graphics hardware. + /* Advertise the full hardware capabilities. The new memory + * manager should cope much better with overload situations: */ - intel->ctx.Const.MaxTextureUnits = I830_TEX_UNITS; - i = driQueryOptioni( &intel->optionCache, "allow_large_textures"); - driCalculateMaxTextureLevels( intel->texture_heaps, - intel->nr_heaps, - &intel->ctx.Const, - 4, - 11, /* max 2D texture size is 2048x2048 */ - 8, /* max 3D texture size is 256^3 */ - 10, /* max CUBE texture size is 1024x1024 */ - 11, /* max RECT. supported */ - 12, - GL_FALSE, - i ); - - _tnl_init_vertices( ctx, ctx->Const.MaxArrayLockSize + 12, - 18 * sizeof(GLfloat) ); + ctx->Const.MaxTextureLevels = 12; + ctx->Const.Max3DTextureLevels = 9; + ctx->Const.MaxCubeTextureLevels = 11; + ctx->Const.MaxTextureRectSize = (1 << 11); + ctx->Const.MaxTextureUnits = I830_TEX_UNITS; - intel->verts = TNL_CONTEXT(ctx)->clipspace.vertex_buf; + _tnl_init_vertices(ctx, ctx->Const.MaxArrayLockSize + 12, + 18 * sizeof(GLfloat)); - driInitExtensions( ctx, i830_extensions, GL_FALSE ); + intel->verts = TNL_CONTEXT(ctx)->clipspace.vertex_buf; - i830InitState( i830 ); + driInitExtensions(ctx, i830_extensions, GL_FALSE); + i830InitState(i830); + i830InitMetaFuncs(i830); - _tnl_allow_vertex_fog( ctx, 1 ); - _tnl_allow_pixel_fog( ctx, 0 ); + _tnl_allow_vertex_fog(ctx, 1); + _tnl_allow_pixel_fog(ctx, 0); return GL_TRUE; } - diff --git a/src/mesa/drivers/dri/i915/i830_context.h b/src/mesa/drivers/dri/i915/i830_context.h index bae777dd5a..9397fa45b5 100644 --- a/src/mesa/drivers/dri/i915/i830_context.h +++ b/src/mesa/drivers/dri/i915/i830_context.h @@ -49,17 +49,15 @@ */ #define I830_DESTREG_CBUFADDR0 0 #define I830_DESTREG_CBUFADDR1 1 -#define I830_DESTREG_CBUFADDR2 2 -#define I830_DESTREG_DBUFADDR0 3 -#define I830_DESTREG_DBUFADDR1 4 -#define I830_DESTREG_DBUFADDR2 5 -#define I830_DESTREG_DV0 6 -#define I830_DESTREG_DV1 7 -#define I830_DESTREG_SENABLE 8 -#define I830_DESTREG_SR0 9 -#define I830_DESTREG_SR1 10 -#define I830_DESTREG_SR2 11 -#define I830_DEST_SETUP_SIZE 12 +#define I830_DESTREG_DBUFADDR0 2 +#define I830_DESTREG_DBUFADDR1 3 +#define I830_DESTREG_DV0 4 +#define I830_DESTREG_DV1 5 +#define I830_DESTREG_SENABLE 6 +#define I830_DESTREG_SR0 7 +#define I830_DESTREG_SR1 8 +#define I830_DESTREG_SR2 9 +#define I830_DEST_SETUP_SIZE 10 #define I830_CTXREG_STATE1 0 #define I830_CTXREG_STATE2 1 @@ -73,7 +71,7 @@ #define I830_CTXREG_AA 9 #define I830_CTXREG_FOGCOLOR 10 #define I830_CTXREG_BLENDCOLOR0 11 -#define I830_CTXREG_BLENDCOLOR1 12 +#define I830_CTXREG_BLENDCOLOR1 12 #define I830_CTXREG_VF 13 #define I830_CTXREG_VF2 14 #define I830_CTXREG_MCSB0 15 @@ -84,17 +82,16 @@ #define I830_STPREG_ST1 1 #define I830_STP_SETUP_SIZE 2 -#define I830_TEXREG_TM0LI 0 /* load immediate 2 texture map n */ -#define I830_TEXREG_TM0S0 1 -#define I830_TEXREG_TM0S1 2 -#define I830_TEXREG_TM0S2 3 -#define I830_TEXREG_TM0S3 4 -#define I830_TEXREG_TM0S4 5 -#define I830_TEXREG_MCS 6 /* _3DSTATE_MAP_COORD_SETS */ -#define I830_TEXREG_CUBE 7 /* _3DSTATE_MAP_SUBE */ -#define I830_TEX_SETUP_SIZE 8 +#define I830_TEXREG_TM0LI 0 /* load immediate 2 texture map n */ +#define I830_TEXREG_TM0S1 1 +#define I830_TEXREG_TM0S2 2 +#define I830_TEXREG_TM0S3 3 +#define I830_TEXREG_TM0S4 4 +#define I830_TEXREG_MCS 5 /* _3DSTATE_MAP_COORD_SETS */ +#define I830_TEXREG_CUBE 6 /* _3DSTATE_MAP_SUBE */ +#define I830_TEX_SETUP_SIZE 7 -#define I830_TEXBLEND_SIZE 12 /* (4 args + op) * 2 + COLOR_FACTOR */ +#define I830_TEXBLEND_SIZE 12 /* (4 args + op) * 2 + COLOR_FACTOR */ struct i830_texture_object { @@ -104,30 +101,39 @@ struct i830_texture_object #define I830_TEX_UNITS 4 -struct i830_hw_state { +struct i830_hw_state +{ GLuint Ctx[I830_CTX_SETUP_SIZE]; GLuint Buffer[I830_DEST_SETUP_SIZE]; GLuint Stipple[I830_STP_SETUP_SIZE]; GLuint Tex[I830_TEX_UNITS][I830_TEX_SETUP_SIZE]; GLuint TexBlend[I830_TEX_UNITS][I830_TEXBLEND_SIZE]; GLuint TexBlendWordsUsed[I830_TEX_UNITS]; - GLuint emitted; /* I810_UPLOAD_* */ + + struct intel_region *draw_region; + struct intel_region *depth_region; + + /* Regions aren't actually that appropriate here as the memory may + * be from a PBO or FBO. Will have to do this for draw and depth for + * FBO's... + */ + dri_bo *tex_buffer[I830_TEX_UNITS]; + GLuint tex_offset[I830_TEX_UNITS]; + + GLuint emitted; /* I810_UPLOAD_* */ GLuint active; }; -struct i830_context +struct i830_context { struct intel_context intel; - - DECLARE_RENDERINPUTS(last_index_bitset); + + GLuint lodbias_tm0s3[MAX_TEXTURE_UNITS]; + DECLARE_RENDERINPUTS(last_index_bitset); struct i830_hw_state meta, initial, state, *current; }; -typedef struct i830_context *i830ContextPtr; -typedef struct i830_texture_object *i830TextureObjectPtr; - -#define I830_CONTEXT(ctx) ((i830ContextPtr)(ctx)) @@ -148,71 +154,60 @@ do { \ /* i830_vtbl.c */ -extern void -i830InitVtbl( i830ContextPtr i830 ); +extern void i830InitVtbl(struct i830_context *i830); +extern void +i830_state_draw_region(struct intel_context *intel, + struct i830_hw_state *state, + struct intel_region *color_region, + struct intel_region *depth_region); /* i830_context.c */ -extern GLboolean -i830CreateContext( const __GLcontextModes *mesaVis, - __DRIcontextPrivate *driContextPriv, - void *sharedContextPrivate); +extern GLboolean +i830CreateContext(const __GLcontextModes * mesaVis, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate); /* i830_tex.c, i830_texstate.c */ -extern void -i830UpdateTextureState( intelContextPtr intel ); - -extern void -i830InitTextureFuncs( struct dd_function_table *functions ); +extern void i830UpdateTextureState(struct intel_context *intel); -extern intelTextureObjectPtr -i830AllocTexObj( struct gl_texture_object *tObj ); +extern void i830InitTextureFuncs(struct dd_function_table *functions); /* i830_texblend.c */ -extern GLuint i830SetTexEnvCombine(i830ContextPtr i830, - const struct gl_tex_env_combine_state * combine, GLint blendUnit, - GLuint texel_op, GLuint *state, const GLfloat *factor ); +extern GLuint i830SetTexEnvCombine(struct i830_context *i830, + const struct gl_tex_env_combine_state + *combine, GLint blendUnit, GLuint texel_op, + GLuint * state, const GLfloat * factor); -extern void -i830EmitTextureBlend( i830ContextPtr i830 ); +extern void i830EmitTextureBlend(struct i830_context *i830); /* i830_state.c */ -extern void -i830InitStateFuncs( struct dd_function_table *functions ); +extern void i830InitStateFuncs(struct dd_function_table *functions); -extern void -i830EmitState( i830ContextPtr i830 ); +extern void i830EmitState(struct i830_context *i830); -extern void -i830InitState( i830ContextPtr i830 ); +extern void i830InitState(struct i830_context *i830); /* i830_metaops.c */ -extern GLboolean -i830TryTextureReadPixels( GLcontext *ctx, - GLint x, GLint y, GLsizei width, GLsizei height, - GLenum format, GLenum type, - const struct gl_pixelstore_attrib *pack, - GLvoid *pixels ); - -extern GLboolean -i830TryTextureDrawPixels( GLcontext *ctx, - GLint x, GLint y, GLsizei width, GLsizei height, - GLenum format, GLenum type, - const struct gl_pixelstore_attrib *unpack, - const GLvoid *pixels ); - -extern void -i830ClearWithTris( intelContextPtr intel, GLbitfield mask, - GLboolean all, GLint cx, GLint cy, GLint cw, GLint ch); +extern void i830InitMetaFuncs(struct i830_context *i830); extern void -i830RotateWindow(intelContextPtr intel, __DRIdrawablePrivate *dPriv, +i830RotateWindow(struct intel_context *intel, __DRIdrawablePrivate * dPriv, GLuint srcBuf); -#endif +/*====================================================================== + * Inline conversion functions. These are better-typed than the + * macros used previously: + */ +static INLINE struct i830_context * +i830_context(GLcontext * ctx) +{ + return (struct i830_context *) ctx; +} +#endif diff --git a/src/mesa/drivers/dri/i915/i830_metaops.c b/src/mesa/drivers/dri/i915/i830_metaops.c index c1d7fe349c..13e4ab3aac 100644 --- a/src/mesa/drivers/dri/i915/i830_metaops.c +++ b/src/mesa/drivers/dri/i915/i830_metaops.c @@ -34,6 +34,7 @@ #include "intel_screen.h" #include "intel_batchbuffer.h" #include "intel_ioctl.h" +#include "intel_regions.h" #include "i830_context.h" #include "i830_reg.h" @@ -41,34 +42,26 @@ /* A large amount of state doesn't need to be uploaded. */ #define ACTIVE (I830_UPLOAD_INVARIENT | \ - I830_UPLOAD_TEXBLEND(0) | \ - I830_UPLOAD_STIPPLE | \ I830_UPLOAD_CTX | \ I830_UPLOAD_BUFFERS | \ - I830_UPLOAD_TEX(0)) + I830_UPLOAD_STIPPLE | \ + I830_UPLOAD_TEXBLEND(0) | \ + I830_UPLOAD_TEX(0)) #define SET_STATE( i830, STATE ) \ do { \ - i830->current->emitted = 0; \ + i830->current->emitted &= ~ACTIVE; \ i830->current = &i830->STATE; \ - i830->current->emitted = 0; \ + i830->current->emitted &= ~ACTIVE; \ } while (0) -/* Operations where the 3D engine is decoupled temporarily from the - * current GL state and used for other purposes than simply rendering - * incoming triangles. - */ -static void set_initial_state( i830ContextPtr i830 ) -{ - memcpy(&i830->meta, &i830->initial, sizeof(i830->meta) ); - i830->meta.active = ACTIVE; - i830->meta.emitted = 0; -} - -static void set_no_depth_stencil_write( i830ContextPtr i830 ) +static void +set_no_stencil_write(struct intel_context *intel) { + struct i830_context *i830 = i830_context(&intel->ctx); + /* ctx->Driver.Enable( ctx, GL_STENCIL_TEST, GL_FALSE ) */ i830->meta.Ctx[I830_CTXREG_ENABLES_1] &= ~ENABLE_STENCIL_TEST; @@ -76,6 +69,13 @@ static void set_no_depth_stencil_write( i830ContextPtr i830 ) i830->meta.Ctx[I830_CTXREG_ENABLES_1] |= DISABLE_STENCIL_TEST; i830->meta.Ctx[I830_CTXREG_ENABLES_2] |= DISABLE_STENCIL_WRITE; + i830->meta.emitted &= ~I830_UPLOAD_CTX; +} + +static void +set_no_depth_write(struct intel_context *intel) +{ + struct i830_context *i830 = i830_context(&intel->ctx); /* ctx->Driver.Enable( ctx, GL_DEPTH_TEST, GL_FALSE ) */ @@ -87,35 +87,56 @@ static void set_no_depth_stencil_write( i830ContextPtr i830 ) i830->meta.emitted &= ~I830_UPLOAD_CTX; } -/* Set stencil unit to replace always with the reference value. +/* Set depth unit to replace. */ -static void set_stencil_replace( i830ContextPtr i830, - GLuint s_mask, - GLuint s_clear) +static void +set_depth_replace(struct intel_context *intel) { - /* ctx->Driver.Enable( ctx, GL_STENCIL_TEST, GL_TRUE ) - */ - i830->meta.Ctx[I830_CTXREG_ENABLES_1] |= ENABLE_STENCIL_TEST; - i830->meta.Ctx[I830_CTXREG_ENABLES_2] |= ENABLE_STENCIL_WRITE; - + struct i830_context *i830 = i830_context(&intel->ctx); /* ctx->Driver.Enable( ctx, GL_DEPTH_TEST, GL_FALSE ) + * ctx->Driver.DepthMask( ctx, GL_TRUE ) */ i830->meta.Ctx[I830_CTXREG_ENABLES_1] &= ~ENABLE_DIS_DEPTH_TEST_MASK; i830->meta.Ctx[I830_CTXREG_ENABLES_2] &= ~ENABLE_DIS_DEPTH_WRITE_MASK; - i830->meta.Ctx[I830_CTXREG_ENABLES_1] |= DISABLE_DEPTH_TEST; - i830->meta.Ctx[I830_CTXREG_ENABLES_2] |= DISABLE_DEPTH_WRITE; + i830->meta.Ctx[I830_CTXREG_ENABLES_1] |= ENABLE_DEPTH_TEST; + i830->meta.Ctx[I830_CTXREG_ENABLES_2] |= ENABLE_DEPTH_WRITE; + + /* ctx->Driver.DepthFunc( ctx, GL_ALWAYS ) + */ + i830->meta.Ctx[I830_CTXREG_STATE3] &= ~DEPTH_TEST_FUNC_MASK; + i830->meta.Ctx[I830_CTXREG_STATE3] |= (ENABLE_DEPTH_TEST_FUNC | + DEPTH_TEST_FUNC + (COMPAREFUNC_ALWAYS)); + + i830->meta.emitted &= ~I830_UPLOAD_CTX; +} + + +/* Set stencil unit to replace always with the reference value. + */ +static void +set_stencil_replace(struct intel_context *intel, + GLuint s_mask, GLuint s_clear) +{ + struct i830_context *i830 = i830_context(&intel->ctx); + + /* ctx->Driver.Enable( ctx, GL_STENCIL_TEST, GL_TRUE ) + */ + i830->meta.Ctx[I830_CTXREG_ENABLES_1] |= ENABLE_STENCIL_TEST; + i830->meta.Ctx[I830_CTXREG_ENABLES_2] |= ENABLE_STENCIL_WRITE; /* ctx->Driver.StencilMask( ctx, s_mask ) */ i830->meta.Ctx[I830_CTXREG_STATE4] &= ~MODE4_ENABLE_STENCIL_WRITE_MASK; i830->meta.Ctx[I830_CTXREG_STATE4] |= (ENABLE_STENCIL_WRITE_MASK | - STENCIL_WRITE_MASK((s_mask&0xff))); + STENCIL_WRITE_MASK((s_mask & + 0xff))); /* ctx->Driver.StencilOp( ctx, GL_REPLACE, GL_REPLACE, GL_REPLACE ) */ i830->meta.Ctx[I830_CTXREG_STENCILTST] &= ~(STENCIL_OPS_MASK); - i830->meta.Ctx[I830_CTXREG_STENCILTST] |= + i830->meta.Ctx[I830_CTXREG_STENCILTST] |= (ENABLE_STENCIL_PARMS | STENCIL_FAIL_OP(STENCILOP_REPLACE) | STENCIL_PASS_DEPTH_FAIL_OP(STENCILOP_REPLACE) | @@ -125,14 +146,14 @@ static void set_stencil_replace( i830ContextPtr i830, */ i830->meta.Ctx[I830_CTXREG_STATE4] &= ~MODE4_ENABLE_STENCIL_TEST_MASK; i830->meta.Ctx[I830_CTXREG_STATE4] |= (ENABLE_STENCIL_TEST_MASK | - STENCIL_TEST_MASK(0xff)); + STENCIL_TEST_MASK(0xff)); i830->meta.Ctx[I830_CTXREG_STENCILTST] &= ~(STENCIL_REF_VALUE_MASK | - ENABLE_STENCIL_TEST_FUNC_MASK); - i830->meta.Ctx[I830_CTXREG_STENCILTST] |= + ENABLE_STENCIL_TEST_FUNC_MASK); + i830->meta.Ctx[I830_CTXREG_STENCILTST] |= (ENABLE_STENCIL_REF_VALUE | ENABLE_STENCIL_TEST_FUNC | - STENCIL_REF_VALUE((s_clear&0xff)) | + STENCIL_REF_VALUE((s_clear & 0xff)) | STENCIL_TEST_FUNC(COMPAREFUNC_ALWAYS)); @@ -141,38 +162,43 @@ static void set_stencil_replace( i830ContextPtr i830, } -static void set_color_mask( i830ContextPtr i830, GLboolean state ) +static void +set_color_mask(struct intel_context *intel, GLboolean state) { + struct i830_context *i830 = i830_context(&intel->ctx); + const GLuint mask = ((1 << WRITEMASK_RED_SHIFT) | - (1 << WRITEMASK_GREEN_SHIFT) | - (1 << WRITEMASK_BLUE_SHIFT) | - (1 << WRITEMASK_ALPHA_SHIFT)); + (1 << WRITEMASK_GREEN_SHIFT) | + (1 << WRITEMASK_BLUE_SHIFT) | + (1 << WRITEMASK_ALPHA_SHIFT)); i830->meta.Ctx[I830_CTXREG_ENABLES_2] &= ~mask; if (state) { - i830->meta.Ctx[I830_CTXREG_ENABLES_2] |= - (i830->state.Ctx[I830_CTXREG_ENABLES_2] & mask); + i830->meta.Ctx[I830_CTXREG_ENABLES_2] |= + (i830->state.Ctx[I830_CTXREG_ENABLES_2] & mask); } - + i830->meta.emitted &= ~I830_UPLOAD_CTX; } /* Installs a one-stage passthrough texture blend pipeline. Is there * more that can be done to turn off texturing? */ -static void set_no_texture( i830ContextPtr i830 ) +static void +set_no_texture(struct intel_context *intel) { + struct i830_context *i830 = i830_context(&intel->ctx); static const struct gl_tex_env_combine_state comb = { GL_NONE, GL_NONE, - { GL_TEXTURE, 0, 0, }, { GL_TEXTURE, 0, 0, }, - { GL_SRC_COLOR, 0, 0 }, { GL_SRC_ALPHA, 0, 0 }, + {GL_TEXTURE, 0, 0,}, {GL_TEXTURE, 0, 0,}, + {GL_SRC_COLOR, 0, 0}, {GL_SRC_ALPHA, 0, 0}, 0, 0, 0, 0 }; i830->meta.TexBlendWordsUsed[0] = - i830SetTexEnvCombine( i830, & comb, 0, TEXBLENDARG_TEXEL0, - i830->meta.TexBlend[0], NULL); + i830SetTexEnvCombine(i830, &comb, 0, TEXBLENDARG_TEXEL0, + i830->meta.TexBlend[0], NULL); i830->meta.TexBlend[0][0] |= TEXOP_LAST_STAGE; i830->meta.emitted &= ~I830_UPLOAD_TEXBLEND(0); @@ -181,18 +207,22 @@ static void set_no_texture( i830ContextPtr i830 ) /* Set up a single element blend stage for 'replace' texturing with no * funny ops. */ -static void enable_texture_blend_replace( i830ContextPtr i830 ) +static void +set_texture_blend_replace(struct intel_context *intel) { + struct i830_context *i830 = i830_context(&intel->ctx); static const struct gl_tex_env_combine_state comb = { GL_REPLACE, GL_REPLACE, - { GL_TEXTURE, GL_TEXTURE, GL_TEXTURE }, { GL_TEXTURE, GL_TEXTURE, GL_TEXTURE, }, - { GL_SRC_COLOR, GL_SRC_COLOR, GL_SRC_COLOR }, { GL_SRC_ALPHA, GL_SRC_ALPHA, GL_SRC_ALPHA }, + {GL_TEXTURE, GL_TEXTURE, GL_TEXTURE,}, {GL_TEXTURE, GL_TEXTURE, + GL_TEXTURE,}, + {GL_SRC_COLOR, GL_SRC_COLOR, GL_SRC_COLOR}, {GL_SRC_ALPHA, GL_SRC_ALPHA, + GL_SRC_ALPHA}, 0, 0, 1, 1 }; i830->meta.TexBlendWordsUsed[0] = - i830SetTexEnvCombine( i830, & comb, 0, TEXBLENDARG_TEXEL0, - i830->meta.TexBlend[0], NULL); + i830SetTexEnvCombine(i830, &comb, 0, TEXBLENDARG_TEXEL0, + i830->meta.TexBlend[0], NULL); i830->meta.TexBlend[0][0] |= TEXOP_LAST_STAGE; i830->meta.emitted &= ~I830_UPLOAD_TEXBLEND(0); @@ -206,717 +236,222 @@ static void enable_texture_blend_replace( i830ContextPtr i830 ) /* Set up an arbitary piece of memory as a rectangular texture * (including the front or back buffer). */ -static void set_tex_rect_source( i830ContextPtr i830, - GLuint offset, - GLuint width, - GLuint height, - GLuint pitch, /* in bytes */ - GLuint textureFormat ) +static GLboolean +set_tex_rect_source(struct intel_context *intel, + dri_bo *buffer, + GLuint offset, + GLuint pitch, GLuint height, GLenum format, GLenum type) { - GLint numLevels = 1; + struct i830_context *i830 = i830_context(&intel->ctx); GLuint *setup = i830->meta.Tex[0]; + GLint numLevels = 1; + GLuint textureFormat; + GLuint cpp; -/* fprintf(stderr, "%s: offset: %x w: %d h: %d pitch %d format %x\n", */ -/* __FUNCTION__, offset, width, height, pitch, textureFormat ); */ + /* A full implementation of this would do the upload through + * glTexImage2d, and get all the conversion operations at that + * point. We are restricted, but still at least have access to the + * fragment program swizzle. + */ + switch (format) { + case GL_BGRA: + switch (type) { + case GL_UNSIGNED_INT_8_8_8_8_REV: + case GL_UNSIGNED_BYTE: + textureFormat = (MAPSURF_32BIT | MT_32BIT_ARGB8888); + cpp = 4; + break; + default: + return GL_FALSE; + } + break; + case GL_RGBA: + switch (type) { + case GL_UNSIGNED_INT_8_8_8_8_REV: + case GL_UNSIGNED_BYTE: + textureFormat = (MAPSURF_32BIT | MT_32BIT_ABGR8888); + cpp = 4; + break; + default: + return GL_FALSE; + } + break; + case GL_BGR: + switch (type) { + case GL_UNSIGNED_SHORT_5_6_5_REV: + textureFormat = (MAPSURF_16BIT | MT_16BIT_RGB565); + cpp = 2; + break; + default: + return GL_FALSE; + } + break; + case GL_RGB: + switch (type) { + case GL_UNSIGNED_SHORT_5_6_5: + textureFormat = (MAPSURF_16BIT | MT_16BIT_RGB565); + cpp = 2; + break; + default: + return GL_FALSE; + } + break; - setup[I830_TEXREG_TM0LI] = (_3DSTATE_LOAD_STATE_IMMEDIATE_2 | - (LOAD_TEXTURE_MAP0 << 0) | 4); - setup[I830_TEXREG_TM0S0] = (TM0S0_USE_FENCE | offset); + default: + return GL_FALSE; + } + + i830->meta.tex_buffer[0] = buffer; + i830->meta.tex_offset[0] = offset; + + setup[I830_TEXREG_TM0LI] = (_3DSTATE_LOAD_STATE_IMMEDIATE_2 | + (LOAD_TEXTURE_MAP0 << 0) | 4); setup[I830_TEXREG_TM0S1] = (((height - 1) << TM0S1_HEIGHT_SHIFT) | - ((width - 1) << TM0S1_WIDTH_SHIFT) | - textureFormat); - setup[I830_TEXREG_TM0S2] = ((((pitch / 4) - 1) << TM0S2_PITCH_SHIFT)); - setup[I830_TEXREG_TM0S3] &= ~TM0S3_MAX_MIP_MASK; - setup[I830_TEXREG_TM0S3] &= ~TM0S3_MIN_MIP_MASK; - setup[I830_TEXREG_TM0S3] |= ((numLevels - 1)*4) << TM0S3_MIN_MIP_SHIFT; + ((pitch - 1) << TM0S1_WIDTH_SHIFT) | + textureFormat); + setup[I830_TEXREG_TM0S2] = + (((((pitch * cpp) / 4) - + 1) << TM0S2_PITCH_SHIFT) | TM0S2_CUBE_FACE_ENA_MASK); + + setup[I830_TEXREG_TM0S3] = + ((((numLevels - + 1) * + 4) << TM0S3_MIN_MIP_SHIFT) | (FILTER_NEAREST << + TM0S3_MIN_FILTER_SHIFT) | + (MIPFILTER_NONE << TM0S3_MIP_FILTER_SHIFT) | (FILTER_NEAREST << + TM0S3_MAG_FILTER_SHIFT)); + + setup[I830_TEXREG_CUBE] = (_3DSTATE_MAP_CUBE | MAP_UNIT(0)); setup[I830_TEXREG_MCS] = (_3DSTATE_MAP_COORD_SET_CMD | - MAP_UNIT(0) | - ENABLE_TEXCOORD_PARAMS | - TEXCOORDS_ARE_IN_TEXELUNITS | - TEXCOORDTYPE_CARTESIAN | - ENABLE_ADDR_V_CNTL | - TEXCOORD_ADDR_V_MODE(TEXCOORDMODE_WRAP) | - ENABLE_ADDR_U_CNTL | - TEXCOORD_ADDR_U_MODE(TEXCOORDMODE_WRAP)); + MAP_UNIT(0) | + ENABLE_TEXCOORD_PARAMS | + TEXCOORDS_ARE_IN_TEXELUNITS | + TEXCOORDTYPE_CARTESIAN | + ENABLE_ADDR_V_CNTL | + TEXCOORD_ADDR_V_MODE(TEXCOORDMODE_WRAP) | + ENABLE_ADDR_U_CNTL | + TEXCOORD_ADDR_U_MODE(TEXCOORDMODE_WRAP)); i830->meta.emitted &= ~I830_UPLOAD_TEX(0); + return GL_TRUE; } -/* Select between front and back draw buffers. - */ -static void set_draw_region( i830ContextPtr i830, - const intelRegion *region ) -{ - i830->meta.Buffer[I830_DESTREG_CBUFADDR1] = - (BUF_3D_ID_COLOR_BACK | BUF_3D_PITCH(region->pitch) | BUF_3D_USE_FENCE); - i830->meta.Buffer[I830_DESTREG_CBUFADDR2] = region->offset; - i830->meta.emitted &= ~I830_UPLOAD_BUFFERS; -} - -/* Setup an arbitary draw format, useful for targeting - * texture or agp memory. - */ -#if 0 -static void set_draw_format( i830ContextPtr i830, - GLuint format, - GLuint depth_format) -{ - i830->meta.Buffer[I830_DESTREG_DV1] = (DSTORG_HORT_BIAS(0x8) | /* .5 */ - DSTORG_VERT_BIAS(0x8) | /* .5 */ - format | - DEPTH_IS_Z | - depth_format); -} -#endif - - -static void set_vertex_format( i830ContextPtr i830 ) +static void +set_vertex_format(struct intel_context *intel) { - i830->meta.Ctx[I830_CTXREG_VF] = (_3DSTATE_VFT0_CMD | - VFT0_TEX_COUNT(1) | - VFT0_DIFFUSE | - VFT0_SPEC | - VFT0_XYZW); + struct i830_context *i830 = i830_context(&intel->ctx); + i830->meta.Ctx[I830_CTXREG_VF] = (_3DSTATE_VFT0_CMD | + VFT0_TEX_COUNT(1) | + VFT0_DIFFUSE | VFT0_XYZ); i830->meta.Ctx[I830_CTXREG_VF2] = (_3DSTATE_VFT1_CMD | - VFT1_TEX0_FMT(TEXCOORDFMT_2D) | - VFT1_TEX1_FMT(TEXCOORDFMT_2D) | - VFT1_TEX2_FMT(TEXCOORDFMT_2D) | - VFT1_TEX3_FMT(TEXCOORDFMT_2D)); + VFT1_TEX0_FMT(TEXCOORDFMT_2D) | + VFT1_TEX1_FMT(TEXCOORDFMT_2D) | + VFT1_TEX2_FMT(TEXCOORDFMT_2D) | + VFT1_TEX3_FMT(TEXCOORDFMT_2D)); i830->meta.emitted &= ~I830_UPLOAD_CTX; } -static void draw_quad(i830ContextPtr i830, - GLfloat x0, GLfloat x1, - GLfloat y0, GLfloat y1, - GLubyte red, GLubyte green, - GLubyte blue, GLubyte alpha, - GLfloat s0, GLfloat s1, - GLfloat t0, GLfloat t1 ) -{ - GLuint vertex_size = 8; - GLuint *vb = intelEmitInlinePrimitiveLocked( &i830->intel, - PRIM3D_TRIFAN, - 4*vertex_size, - vertex_size ); - intelVertex tmp; - int i; - - -/* fprintf(stderr, "%s: %f,%f-%f,%f 0x%x%x%x%x %f,%f-%f,%f\n", */ -/* __FUNCTION__, */ -/* x0,y0,x1,y1,red,green,blue,alpha,s0,t0,s1,t1); */ - - - /* initial vertex, left bottom */ - tmp.v.x = x0; - tmp.v.y = y0; - tmp.v.z = 1.0; - tmp.v.w = 1.0; - tmp.v.color.red = red; - tmp.v.color.green = green; - tmp.v.color.blue = blue; - tmp.v.color.alpha = alpha; - tmp.v.specular.red = 0; - tmp.v.specular.green = 0; - tmp.v.specular.blue = 0; - tmp.v.specular.alpha = 0; - tmp.v.u0 = s0; - tmp.v.v0 = t0; - for (i = 0 ; i < 8 ; i++) - vb[i] = tmp.ui[i]; - - /* right bottom */ - vb += 8; - tmp.v.x = x1; - tmp.v.u0 = s1; - for (i = 0 ; i < 8 ; i++) - vb[i] = tmp.ui[i]; - - /* right top */ - vb += 8; - tmp.v.y = y1; - tmp.v.v0 = t1; - for (i = 0 ; i < 8 ; i++) - vb[i] = tmp.ui[i]; - - /* left top */ - vb += 8; - tmp.v.x = x0; - tmp.v.u0 = s0; - for (i = 0 ; i < 8 ; i++) - vb[i] = tmp.ui[i]; - -/* fprintf(stderr, "%s: DV1: %x\n", */ -/* __FUNCTION__, i830->meta.Buffer[I830_DESTREG_DV1]); */ -} - -static void draw_poly(i830ContextPtr i830, - GLubyte red, GLubyte green, GLubyte blue, GLubyte alpha, - GLuint numVerts, - GLfloat verts[][2], - GLfloat texcoords[][2]) +static void +meta_import_pixel_state(struct intel_context *intel) { - GLuint vertex_size = 8; - GLuint *vb = intelEmitInlinePrimitiveLocked( &i830->intel, - PRIM3D_TRIFAN, - numVerts * vertex_size, - vertex_size ); - intelVertex tmp; - int i, k; - - /* initial constant vertex fields */ - tmp.v.z = 1.0; - tmp.v.w = 1.0; - tmp.v.color.red = red; - tmp.v.color.green = green; - tmp.v.color.blue = blue; - tmp.v.color.alpha = alpha; - tmp.v.specular.red = 0; - tmp.v.specular.green = 0; - tmp.v.specular.blue = 0; - tmp.v.specular.alpha = 0; - - for (k = 0; k < numVerts; k++) { - tmp.v.x = verts[k][0]; - tmp.v.y = verts[k][1]; - tmp.v.u0 = texcoords[k][0]; - tmp.v.v0 = texcoords[k][1]; - - for (i = 0 ; i < vertex_size ; i++) - vb[i] = tmp.ui[i]; - - vb += vertex_size; - } -} - -void -i830ClearWithTris(intelContextPtr intel, GLbitfield mask, - GLboolean allFoo, - GLint cxFoo, GLint cyFoo, GLint cwFoo, GLint chFoo) -{ - i830ContextPtr i830 = I830_CONTEXT( intel ); - __DRIdrawablePrivate *dPriv = intel->driDrawable; - intelScreenPrivate *screen = intel->intelScreen; - int x0, y0, x1, y1; - GLint cx, cy, cw, ch; - GLboolean all; - - INTEL_FIREVERTICES(intel); - SET_STATE( i830, meta ); - set_initial_state( i830 ); -/* set_no_texture( i830 ); */ - set_vertex_format( i830 ); - - LOCK_HARDWARE(intel); - - /* get clear bounds after locking */ - cx = intel->ctx.DrawBuffer->_Xmin; - cy = intel->ctx.DrawBuffer->_Ymin; - cw = intel->ctx.DrawBuffer->_Xmax - cx; - ch = intel->ctx.DrawBuffer->_Ymax - cy; - all = (cw == intel->ctx.DrawBuffer->Width && - ch == intel->ctx.DrawBuffer->Height); - - if(!all) { - x0 = cx; - y0 = cy; - x1 = x0 + cw; - y1 = y0 + ch; - } else { - x0 = 0; - y0 = 0; - x1 = x0 + dPriv->w; - y1 = y0 + dPriv->h; - } - - /* Don't do any clipping to screen - these are window coordinates. - * The active cliprects will be applied as for any other geometry. - */ - - if(mask & BUFFER_BIT_FRONT_LEFT) { - set_no_depth_stencil_write( i830 ); - set_color_mask( i830, GL_TRUE ); - set_draw_region( i830, &screen->front ); - draw_quad(i830, x0, x1, y0, y1, - intel->clear_red, intel->clear_green, - intel->clear_blue, intel->clear_alpha, - 0, 0, 0, 0); - } - - if(mask & BUFFER_BIT_BACK_LEFT) { - set_no_depth_stencil_write( i830 ); - set_color_mask( i830, GL_TRUE ); - set_draw_region( i830, &screen->back ); - - draw_quad(i830, x0, x1, y0, y1, - intel->clear_red, intel->clear_green, - intel->clear_blue, intel->clear_alpha, - 0, 0, 0, 0); - } - - if(mask & BUFFER_BIT_STENCIL) { - set_stencil_replace( i830, - intel->ctx.Stencil.WriteMask[0], - intel->ctx.Stencil.Clear); - - set_color_mask( i830, GL_FALSE ); - set_draw_region( i830, &screen->front ); - draw_quad( i830, x0, x1, y0, y1, 0, 0, 0, 0, 0, 0, 0, 0 ); - } + struct i830_context *i830 = i830_context(&intel->ctx); + + i830->meta.Ctx[I830_CTXREG_STATE1] = i830->state.Ctx[I830_CTXREG_STATE1]; + i830->meta.Ctx[I830_CTXREG_STATE2] = i830->state.Ctx[I830_CTXREG_STATE2]; + i830->meta.Ctx[I830_CTXREG_STATE3] = i830->state.Ctx[I830_CTXREG_STATE3]; + i830->meta.Ctx[I830_CTXREG_STATE4] = i830->state.Ctx[I830_CTXREG_STATE4]; + i830->meta.Ctx[I830_CTXREG_STATE5] = i830->state.Ctx[I830_CTXREG_STATE5]; + i830->meta.Ctx[I830_CTXREG_IALPHAB] = i830->state.Ctx[I830_CTXREG_IALPHAB]; + i830->meta.Ctx[I830_CTXREG_STENCILTST] = + i830->state.Ctx[I830_CTXREG_STENCILTST]; + i830->meta.Ctx[I830_CTXREG_ENABLES_1] = + i830->state.Ctx[I830_CTXREG_ENABLES_1]; + i830->meta.Ctx[I830_CTXREG_ENABLES_2] = + i830->state.Ctx[I830_CTXREG_ENABLES_2]; + i830->meta.Ctx[I830_CTXREG_AA] = i830->state.Ctx[I830_CTXREG_AA]; + i830->meta.Ctx[I830_CTXREG_FOGCOLOR] = + i830->state.Ctx[I830_CTXREG_FOGCOLOR]; + i830->meta.Ctx[I830_CTXREG_BLENDCOLOR0] = + i830->state.Ctx[I830_CTXREG_BLENDCOLOR0]; + i830->meta.Ctx[I830_CTXREG_BLENDCOLOR1] = + i830->state.Ctx[I830_CTXREG_BLENDCOLOR1]; + i830->meta.Ctx[I830_CTXREG_MCSB0] = i830->state.Ctx[I830_CTXREG_MCSB0]; + i830->meta.Ctx[I830_CTXREG_MCSB1] = i830->state.Ctx[I830_CTXREG_MCSB1]; + + + i830->meta.Ctx[I830_CTXREG_STATE3] &= ~CULLMODE_MASK; + i830->meta.Stipple[I830_STPREG_ST1] &= ~ST1_ENABLE; + i830->meta.emitted &= ~I830_UPLOAD_CTX; - UNLOCK_HARDWARE(intel); - INTEL_FIREVERTICES(intel); - SET_STATE( i830, state ); + i830->meta.Buffer[I830_DESTREG_SENABLE] = + i830->state.Buffer[I830_DESTREG_SENABLE]; + i830->meta.Buffer[I830_DESTREG_SR1] = i830->state.Buffer[I830_DESTREG_SR1]; + i830->meta.Buffer[I830_DESTREG_SR2] = i830->state.Buffer[I830_DESTREG_SR2]; + i830->meta.emitted &= ~I830_UPLOAD_BUFFERS; } -#if 0 -GLboolean -i830TryTextureReadPixels( GLcontext *ctx, - GLint x, GLint y, GLsizei width, GLsizei height, - GLenum format, GLenum type, - const struct gl_pixelstore_attrib *pack, - GLvoid *pixels ) +/* Select between front and back draw buffers. + */ +static void +meta_draw_region(struct intel_context *intel, + struct intel_region *color_region, + struct intel_region *depth_region) { - i830ContextPtr i830 = I830_CONTEXT(ctx); - intelContextPtr intel = INTEL_CONTEXT(ctx); - intelScreenPrivate *screen = i830->intel.intelScreen; - GLint pitch = pack->RowLength ? pack->RowLength : width; - __DRIdrawablePrivate *dPriv = i830->intel.driDrawable; - int textureFormat; - GLenum glTextureFormat; - int src_offset = i830->meta.Buffer[I830_DESTREG_CBUFADDR2]; - int destOffset = intelAgpOffsetFromVirtual( &i830->intel, pixels); - int destFormat, depthFormat, destPitch; - drm_clip_rect_t tmp; - - if (INTEL_DEBUG & DEBUG_PIXEL) - fprintf(stderr, "%s\n", __FUNCTION__); - - - if ( ctx->_ImageTransferState || - pack->SwapBytes || - pack->LsbFirst || - !pack->Invert) { - fprintf(stderr, "%s: check_color failed\n", __FUNCTION__); - return GL_FALSE; - } - - switch (screen->fbFormat) { - case DV_PF_565: - textureFormat = MAPSURF_16BIT | MT_16BIT_RGB565; - glTextureFormat = GL_RGB; - break; - case DV_PF_555: - textureFormat = MAPSURF_16BIT | MT_16BIT_ARGB1555; - glTextureFormat = GL_RGBA; - break; - case DV_PF_8888: - textureFormat = MAPSURF_32BIT | MT_32BIT_ARGB8888; - glTextureFormat = GL_RGBA; - break; - default: - fprintf(stderr, "%s: textureFormat failed %x\n", __FUNCTION__, - screen->fbFormat); - return GL_FALSE; - } - - - switch (type) { - case GL_UNSIGNED_SHORT_5_6_5: - if (format != GL_RGB) return GL_FALSE; - destFormat = COLR_BUF_RGB565; - depthFormat = DEPTH_FRMT_16_FIXED; - destPitch = pitch * 2; - break; - case GL_UNSIGNED_INT_8_8_8_8_REV: - if (format != GL_BGRA) return GL_FALSE; - destFormat = COLR_BUF_ARGB8888; - depthFormat = DEPTH_FRMT_24_FIXED_8_OTHER; - destPitch = pitch * 4; - break; - default: - fprintf(stderr, "%s: destFormat failed %s\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(type)); - return GL_FALSE; - } - - destFormat |= (0x02<<24); - -/* fprintf(stderr, "type: %s destFormat: %x\n", */ -/* _mesa_lookup_enum_by_nr(type), */ -/* destFormat); */ - - intelFlush( ctx ); - - SET_STATE( i830, meta ); - set_initial_state( i830 ); - set_no_depth_stencil_write( i830 ); - - LOCK_HARDWARE( intel ); - { - intelWaitForIdle( intel ); /* required by GL */ - - if (!driClipRectToFramebuffer(ctx->ReadBuffer, &x, &y, &width, &height)) { - UNLOCK_HARDWARE( intel ); - SET_STATE(i830, state); - fprintf(stderr, "%s: cliprect failed\n", __FUNCTION__); - return GL_TRUE; - } - -#if 0 - /* FIXME -- Just emit the correct state - */ - if (i830SetParam(i830->driFd, I830_SETPARAM_CBUFFER_PITCH, - destPitch) != 0) { - UNLOCK_HARDWARE( intel ); - SET_STATE(i830, state); - fprintf(stderr, "%s: setparam failed\n", __FUNCTION__); - return GL_FALSE; - } -#endif - - - y = dPriv->h - y - height; - x += dPriv->x; - y += dPriv->y; - - - /* Set the frontbuffer up as a large rectangular texture. - */ - set_tex_rect_source( i830, - src_offset, - screen->width, - screen->height, - screen->front.pitch, - textureFormat ); - - - enable_texture_blend_replace( i830 ); - - - /* Set the 3d engine to draw into the agp memory - */ + struct i830_context *i830 = i830_context(&intel->ctx); - set_draw_region( i830, destOffset ); - set_draw_format( i830, destFormat, depthFormat ); - - - /* Draw a single quad, no cliprects: - */ - i830->intel.numClipRects = 1; - i830->intel.pClipRects = &tmp; - i830->intel.pClipRects[0].x1 = 0; - i830->intel.pClipRects[0].y1 = 0; - i830->intel.pClipRects[0].x2 = width; - i830->intel.pClipRects[0].y2 = height; - - draw_quad( i830, - 0, width, 0, height, - 0, 255, 0, 0, - x, x+width, y, y+height ); - - intelWindowMoved( intel ); - } - UNLOCK_HARDWARE( intel ); - intelFinish( ctx ); /* required by GL */ - - SET_STATE( i830, state ); - return GL_TRUE; + i830_state_draw_region(intel, &i830->meta, color_region, depth_region); } -GLboolean -i830TryTextureDrawPixels( GLcontext *ctx, - GLint x, GLint y, GLsizei width, GLsizei height, - GLenum format, GLenum type, - const struct gl_pixelstore_attrib *unpack, - const GLvoid *pixels ) +/* Operations where the 3D engine is decoupled temporarily from the + * current GL state and used for other purposes than simply rendering + * incoming triangles. + */ +static void +install_meta_state(struct intel_context *intel) { - intelContextPtr intel = INTEL_CONTEXT(ctx); - i830ContextPtr i830 = I830_CONTEXT(ctx); - GLint pitch = unpack->RowLength ? unpack->RowLength : width; - __DRIdrawablePrivate *dPriv = intel->driDrawable; - int textureFormat; - GLenum glTextureFormat; - int dst_offset = i830->meta.Buffer[I830_DESTREG_CBUFADDR2]; - int src_offset = intelAgpOffsetFromVirtual( intel, pixels ); - - if (INTEL_DEBUG & DEBUG_PIXEL) - fprintf(stderr, "%s\n", __FUNCTION__); - - /* Todo -- upload images that aren't in agp space, then texture - * from them. - */ + struct i830_context *i830 = i830_context(&intel->ctx); + memcpy(&i830->meta, &i830->initial, sizeof(i830->meta)); - if ( !intelIsAgpMemory( intel, pixels, pitch*height ) ) { - fprintf(stderr, "%s: intelIsAgpMemory failed\n", __FUNCTION__); - return GL_FALSE; - } - - /* Todo -- don't want to clobber all the drawing state like we do - * for readpixels -- most of this state can be handled just fine. - */ - if ( ctx->_ImageTransferState || - unpack->SwapBytes || - unpack->LsbFirst || - ctx->Color.AlphaEnabled || - ctx->Depth.Test || - ctx->Fog.Enabled || - ctx->Scissor.Enabled || - ctx->Stencil.Enabled || - !ctx->Color.ColorMask[0] || - !ctx->Color.ColorMask[1] || - !ctx->Color.ColorMask[2] || - !ctx->Color.ColorMask[3] || - ctx->Color.ColorLogicOpEnabled || - ctx->Texture._EnabledUnits || - ctx->Depth.OcclusionTest) { - fprintf(stderr, "%s: other tests failed\n", __FUNCTION__); - return GL_FALSE; - } - - /* Todo -- remove these restrictions: - */ - if (ctx->Pixel.ZoomX != 1.0F || - ctx->Pixel.ZoomY != -1.0F) - return GL_FALSE; - - - - switch (type) { - case GL_UNSIGNED_SHORT_1_5_5_5_REV: - if (format != GL_BGRA) return GL_FALSE; - textureFormat = MAPSURF_16BIT | MT_16BIT_ARGB1555; - glTextureFormat = GL_RGBA; - break; - case GL_UNSIGNED_SHORT_5_6_5: - if (format != GL_RGB) return GL_FALSE; - textureFormat = MAPSURF_16BIT | MT_16BIT_RGB565; - glTextureFormat = GL_RGB; - break; - case GL_UNSIGNED_SHORT_8_8_MESA: - if (format != GL_YCBCR_MESA) return GL_FALSE; - textureFormat = (MAPSURF_422 | MT_422_YCRCB_SWAPY -/* | TM0S1_COLORSPACE_CONVERSION */ - ); - glTextureFormat = GL_YCBCR_MESA; - break; - case GL_UNSIGNED_SHORT_8_8_REV_MESA: - if (format != GL_YCBCR_MESA) return GL_FALSE; - textureFormat = (MAPSURF_422 | MT_422_YCRCB_NORMAL -/* | TM0S1_COLORSPACE_CONVERSION */ - ); - glTextureFormat = GL_YCBCR_MESA; - break; - case GL_UNSIGNED_INT_8_8_8_8_REV: - if (format != GL_BGRA) return GL_FALSE; - textureFormat = MAPSURF_32BIT | MT_32BIT_ARGB8888; - glTextureFormat = GL_RGBA; - break; - default: - fprintf(stderr, "%s: destFormat failed\n", __FUNCTION__); - return GL_FALSE; - } - - intelFlush( ctx ); - - SET_STATE( i830, meta ); - - LOCK_HARDWARE( intel ); - { - intelWaitForIdle( intel ); /* required by GL */ - - y -= height; /* cope with pixel zoom */ - - if (!driClipRectToFramebuffer(ctx->ReadBuffer, &x, &y, &width, &height)) { - UNLOCK_HARDWARE( intel ); - SET_STATE(i830, state); - fprintf(stderr, "%s: cliprect failed\n", __FUNCTION__); - return GL_TRUE; - } - - - y = dPriv->h - y - height; - - set_initial_state( i830 ); - - /* Set the pixel image up as a rectangular texture. - */ - set_tex_rect_source( i830, - src_offset, - width, - height, - pitch, /* XXXX!!!! -- /2 sometimes */ - textureFormat ); - - - enable_texture_blend_replace( i830 ); - - - /* Draw to the current draw buffer: - */ - set_draw_offset( i830, dst_offset ); - - /* Draw a quad, use regular cliprects - */ -/* fprintf(stderr, "x: %d y: %d width %d height %d\n", x, y, width, height); */ + i830->meta.active = ACTIVE; + i830->meta.emitted = 0; - draw_quad( i830, - x, x+width, y, y+height, - 0, 255, 0, 0, - 0, width, 0, height ); + SET_STATE(i830, meta); + set_vertex_format(intel); + set_no_texture(intel); +} - intelWindowMoved( intel ); - } - UNLOCK_HARDWARE( intel ); - intelFinish( ctx ); /* required by GL */ - +static void +leave_meta_state(struct intel_context *intel) +{ + struct i830_context *i830 = i830_context(&intel->ctx); + intel_region_release(&i830->meta.draw_region); + intel_region_release(&i830->meta.depth_region); +/* intel_region_release(intel, &i830->meta.tex_region[0]); */ SET_STATE(i830, state); - - return GL_TRUE; } -#endif -/** - * Copy the window contents named by dPriv to the rotated (or reflected) - * color buffer. - * srcBuf is BUFFER_BIT_FRONT_LEFT or BUFFER_BIT_BACK_LEFT to indicate the source. - */ + void -i830RotateWindow(intelContextPtr intel, __DRIdrawablePrivate *dPriv, - GLuint srcBuf) +i830InitMetaFuncs(struct i830_context *i830) { - i830ContextPtr i830 = I830_CONTEXT( intel ); - intelScreenPrivate *screen = intel->intelScreen; - const GLuint cpp = screen->cpp; - drm_clip_rect_t fullRect; - GLuint textureFormat, srcOffset, srcPitch; - const drm_clip_rect_t *clipRects; - int numClipRects; - int i; - - int xOrig, yOrig; - int origNumClipRects; - drm_clip_rect_t *origRects; - - /* - * set up hardware state - */ - intelFlush( &intel->ctx ); - - SET_STATE( i830, meta ); - set_initial_state( i830 ); - set_no_texture( i830 ); - set_vertex_format( i830 ); - set_no_depth_stencil_write( i830 ); - set_color_mask( i830, GL_FALSE ); - - LOCK_HARDWARE(intel); - - /* save current drawing origin and cliprects (restored at end) */ - xOrig = intel->drawX; - yOrig = intel->drawY; - origNumClipRects = intel->numClipRects; - origRects = intel->pClipRects; - - if (!intel->numClipRects) - goto done; - - /* - * set drawing origin, cliprects for full-screen access to rotated screen - */ - fullRect.x1 = 0; - fullRect.y1 = 0; - fullRect.x2 = screen->rotatedWidth; - fullRect.y2 = screen->rotatedHeight; - intel->drawX = 0; - intel->drawY = 0; - intel->numClipRects = 1; - intel->pClipRects = &fullRect; - - set_draw_region( i830, &screen->rotated ); - - if (cpp == 4) - textureFormat = MAPSURF_32BIT | MT_32BIT_ARGB8888; - else - textureFormat = MAPSURF_16BIT | MT_16BIT_RGB565; - - if (srcBuf == BUFFER_BIT_FRONT_LEFT) { - srcPitch = screen->front.pitch; /* in bytes */ - srcOffset = screen->front.offset; /* bytes */ - clipRects = dPriv->pClipRects; - numClipRects = dPriv->numClipRects; - } - else { - srcPitch = screen->back.pitch; /* in bytes */ - srcOffset = screen->back.offset; /* bytes */ - clipRects = dPriv->pBackClipRects; - numClipRects = dPriv->numBackClipRects; - } - - /* set the whole screen up as a texture to avoid alignment issues */ - set_tex_rect_source(i830, - srcOffset, - screen->width, - screen->height, - srcPitch, - textureFormat); - - enable_texture_blend_replace(i830); - - /* - * loop over the source window's cliprects - */ - for (i = 0; i < numClipRects; i++) { - int srcX0 = clipRects[i].x1; - int srcY0 = clipRects[i].y1; - int srcX1 = clipRects[i].x2; - int srcY1 = clipRects[i].y2; - GLfloat verts[4][2], tex[4][2]; - int j; - - /* build vertices for four corners of clip rect */ - verts[0][0] = srcX0; verts[0][1] = srcY0; - verts[1][0] = srcX1; verts[1][1] = srcY0; - verts[2][0] = srcX1; verts[2][1] = srcY1; - verts[3][0] = srcX0; verts[3][1] = srcY1; - - /* .. and texcoords */ - tex[0][0] = srcX0; tex[0][1] = srcY0; - tex[1][0] = srcX1; tex[1][1] = srcY0; - tex[2][0] = srcX1; tex[2][1] = srcY1; - tex[3][0] = srcX0; tex[3][1] = srcY1; - - /* transform coords to rotated screen coords */ - - for (j = 0; j < 4; j++) { - matrix23TransformCoordf(&screen->rotMatrix, - &verts[j][0], &verts[j][1]); - } - - /* draw polygon to map source image to dest region */ - draw_poly(i830, 255, 255, 255, 255, 4, verts, tex); - - } /* cliprect loop */ - - intelFlushBatchLocked( intel, GL_FALSE, GL_FALSE, GL_FALSE ); - - done: - /* restore original drawing origin and cliprects */ - intel->drawX = xOrig; - intel->drawY = yOrig; - intel->numClipRects = origNumClipRects; - intel->pClipRects = origRects; - - UNLOCK_HARDWARE(intel); - - SET_STATE( i830, state ); + i830->intel.vtbl.install_meta_state = install_meta_state; + i830->intel.vtbl.leave_meta_state = leave_meta_state; + i830->intel.vtbl.meta_no_depth_write = set_no_depth_write; + i830->intel.vtbl.meta_no_stencil_write = set_no_stencil_write; + i830->intel.vtbl.meta_stencil_replace = set_stencil_replace; + i830->intel.vtbl.meta_depth_replace = set_depth_replace; + i830->intel.vtbl.meta_color_mask = set_color_mask; + i830->intel.vtbl.meta_no_texture = set_no_texture; + i830->intel.vtbl.meta_texture_blend_replace = set_texture_blend_replace; + i830->intel.vtbl.meta_tex_rect_source = set_tex_rect_source; + i830->intel.vtbl.meta_draw_region = meta_draw_region; + i830->intel.vtbl.meta_import_pixel_state = meta_import_pixel_state; } - diff --git a/src/mesa/drivers/dri/i915/i830_reg.h b/src/mesa/drivers/dri/i915/i830_reg.h index 98cee2f214..41280bca7c 100644 --- a/src/mesa/drivers/dri/i915/i830_reg.h +++ b/src/mesa/drivers/dri/i915/i830_reg.h @@ -407,7 +407,7 @@ #define LOGICOP_SET 0xf #define MODE4_ENABLE_STENCIL_TEST_MASK ((1<<17)|(0xff00)) #define ENABLE_STENCIL_TEST_MASK (1<<17) -#define STENCIL_TEST_MASK(x) ((x)<<8) +#define STENCIL_TEST_MASK(x) (((x)&0xff)<<8) #define MODE4_ENABLE_STENCIL_WRITE_MASK ((1<<16)|(0x00ff)) #define ENABLE_STENCIL_WRITE_MASK (1<<16) #define STENCIL_WRITE_MASK(x) ((x)&0xff) @@ -554,8 +554,8 @@ #define MAPSURF_4BIT_INDEXED (7<<6) #define TM0S1_MT_FORMAT_MASK (0x7 << 3) #define TM0S1_MT_FORMAT_SHIFT 3 -#define MT_4BIT_IDX_ARGB8888 (7<<3) /* SURFACE_4BIT_INDEXED */ -#define MT_8BIT_IDX_RGB565 (0<<3) /* SURFACE_8BIT_INDEXED */ +#define MT_4BIT_IDX_ARGB8888 (7<<3) /* SURFACE_4BIT_INDEXED */ +#define MT_8BIT_IDX_RGB565 (0<<3) /* SURFACE_8BIT_INDEXED */ #define MT_8BIT_IDX_ARGB1555 (1<<3) #define MT_8BIT_IDX_ARGB4444 (2<<3) #define MT_8BIT_IDX_AY88 (3<<3) @@ -563,9 +563,9 @@ #define MT_8BIT_IDX_BUMP_88DVDU (5<<3) #define MT_8BIT_IDX_BUMP_655LDVDU (6<<3) #define MT_8BIT_IDX_ARGB8888 (7<<3) -#define MT_8BIT_I8 (0<<3) /* SURFACE_8BIT */ +#define MT_8BIT_I8 (0<<3) /* SURFACE_8BIT */ #define MT_8BIT_L8 (1<<3) -#define MT_16BIT_RGB565 (0<<3) /* SURFACE_16BIT */ +#define MT_16BIT_RGB565 (0<<3) /* SURFACE_16BIT */ #define MT_16BIT_ARGB1555 (1<<3) #define MT_16BIT_ARGB4444 (2<<3) #define MT_16BIT_AY88 (3<<3) @@ -573,16 +573,17 @@ #define MT_16BIT_BUMP_88DVDU (5<<3) #define MT_16BIT_BUMP_655LDVDU (6<<3) #define MT_16BIT_DIB_RGB565_8888 (7<<3) -#define MT_32BIT_ARGB8888 (0<<3) /* SURFACE_32BIT */ +#define MT_32BIT_ARGB8888 (0<<3) /* SURFACE_32BIT */ #define MT_32BIT_ABGR8888 (1<<3) +#define MT_32BIT_XRGB8888 (2<<3) /* XXX: Guess from i915_reg.h */ #define MT_32BIT_BUMP_XLDVDU_8888 (6<<3) #define MT_32BIT_DIB_8888 (7<<3) -#define MT_411_YUV411 (0<<3) /* SURFACE_411 */ -#define MT_422_YCRCB_SWAPY (0<<3) /* SURFACE_422 */ +#define MT_411_YUV411 (0<<3) /* SURFACE_411 */ +#define MT_422_YCRCB_SWAPY (0<<3) /* SURFACE_422 */ #define MT_422_YCRCB_NORMAL (1<<3) #define MT_422_YCRCB_SWAPUV (2<<3) #define MT_422_YCRCB_SWAPUVY (3<<3) -#define MT_COMPRESS_DXT1 (0<<3) /* SURFACE_COMPRESSED */ +#define MT_COMPRESS_DXT1 (0<<3) /* SURFACE_COMPRESSED */ #define MT_COMPRESS_DXT2_3 (1<<3) #define MT_COMPRESS_DXT4_5 (2<<3) #define MT_COMPRESS_FXT1 (3<<3) diff --git a/src/mesa/drivers/dri/i915/i830_state.c b/src/mesa/drivers/dri/i915/i830_state.c index f7980201f9..3c149e6905 100644 --- a/src/mesa/drivers/dri/i915/i830_state.c +++ b/src/mesa/drivers/dri/i915/i830_state.c @@ -38,149 +38,151 @@ #include "intel_screen.h" #include "intel_batchbuffer.h" +#include "intel_fbo.h" #include "i830_context.h" #include "i830_reg.h" +#define FILE_DEBUG_FLAG DEBUG_STATE + static void -i830StencilFuncSeparate(GLcontext *ctx, GLenum face, GLenum func, GLint ref, +i830StencilFuncSeparate(GLcontext * ctx, GLenum face, GLenum func, GLint ref, GLuint mask) { - i830ContextPtr i830 = I830_CONTEXT(ctx); + struct i830_context *i830 = i830_context(ctx); int test = intel_translate_compare_func(func); mask = mask & 0xff; - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s : func: %s, ref : 0x%x, mask: 0x%x\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(func), ref, mask); + DBG("%s : func: %s, ref : 0x%x, mask: 0x%x\n", __FUNCTION__, + _mesa_lookup_enum_by_nr(func), ref, mask); I830_STATECHANGE(i830, I830_UPLOAD_CTX); i830->state.Ctx[I830_CTXREG_STATE4] &= ~MODE4_ENABLE_STENCIL_TEST_MASK; i830->state.Ctx[I830_CTXREG_STATE4] |= (ENABLE_STENCIL_TEST_MASK | - STENCIL_TEST_MASK(mask)); + STENCIL_TEST_MASK(mask)); i830->state.Ctx[I830_CTXREG_STENCILTST] &= ~(STENCIL_REF_VALUE_MASK | - ENABLE_STENCIL_TEST_FUNC_MASK); + ENABLE_STENCIL_TEST_FUNC_MASK); i830->state.Ctx[I830_CTXREG_STENCILTST] |= (ENABLE_STENCIL_REF_VALUE | - ENABLE_STENCIL_TEST_FUNC | - STENCIL_REF_VALUE(ref) | - STENCIL_TEST_FUNC(test)); + ENABLE_STENCIL_TEST_FUNC | + STENCIL_REF_VALUE(ref) | + STENCIL_TEST_FUNC(test)); } static void -i830StencilMaskSeparate(GLcontext *ctx, GLenum face, GLuint mask) +i830StencilMaskSeparate(GLcontext * ctx, GLenum face, GLuint mask) { - i830ContextPtr i830 = I830_CONTEXT(ctx); - - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s : mask 0x%x\n", __FUNCTION__, mask); + struct i830_context *i830 = i830_context(ctx); + DBG("%s : mask 0x%x\n", __FUNCTION__, mask); + mask = mask & 0xff; I830_STATECHANGE(i830, I830_UPLOAD_CTX); i830->state.Ctx[I830_CTXREG_STATE4] &= ~MODE4_ENABLE_STENCIL_WRITE_MASK; i830->state.Ctx[I830_CTXREG_STATE4] |= (ENABLE_STENCIL_WRITE_MASK | - STENCIL_WRITE_MASK(mask)); + STENCIL_WRITE_MASK(mask)); } static void -i830StencilOpSeparate(GLcontext *ctx, GLenum face, GLenum fail, GLenum zfail, +i830StencilOpSeparate(GLcontext * ctx, GLenum face, GLenum fail, GLenum zfail, GLenum zpass) { - i830ContextPtr i830 = I830_CONTEXT(ctx); + struct i830_context *i830 = i830_context(ctx); int fop, dfop, dpop; - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s: fail : %s, zfail: %s, zpass : %s\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(fail), - _mesa_lookup_enum_by_nr(zfail), - _mesa_lookup_enum_by_nr(zpass)); + DBG("%s: fail : %s, zfail: %s, zpass : %s\n", __FUNCTION__, + _mesa_lookup_enum_by_nr(fail), + _mesa_lookup_enum_by_nr(zfail), + _mesa_lookup_enum_by_nr(zpass)); - fop = 0; dfop = 0; dpop = 0; + fop = 0; + dfop = 0; + dpop = 0; - switch(fail) { - case GL_KEEP: - fop = STENCILOP_KEEP; + switch (fail) { + case GL_KEEP: + fop = STENCILOP_KEEP; break; - case GL_ZERO: - fop = STENCILOP_ZERO; + case GL_ZERO: + fop = STENCILOP_ZERO; break; - case GL_REPLACE: - fop = STENCILOP_REPLACE; + case GL_REPLACE: + fop = STENCILOP_REPLACE; break; - case GL_INCR: + case GL_INCR: fop = STENCILOP_INCRSAT; break; - case GL_DECR: + case GL_DECR: fop = STENCILOP_DECRSAT; break; case GL_INCR_WRAP: - fop = STENCILOP_INCR; + fop = STENCILOP_INCR; break; case GL_DECR_WRAP: - fop = STENCILOP_DECR; + fop = STENCILOP_DECR; break; - case GL_INVERT: - fop = STENCILOP_INVERT; + case GL_INVERT: + fop = STENCILOP_INVERT; break; - default: + default: break; } - switch(zfail) { - case GL_KEEP: - dfop = STENCILOP_KEEP; + switch (zfail) { + case GL_KEEP: + dfop = STENCILOP_KEEP; break; - case GL_ZERO: - dfop = STENCILOP_ZERO; + case GL_ZERO: + dfop = STENCILOP_ZERO; break; - case GL_REPLACE: - dfop = STENCILOP_REPLACE; + case GL_REPLACE: + dfop = STENCILOP_REPLACE; break; - case GL_INCR: + case GL_INCR: dfop = STENCILOP_INCRSAT; break; - case GL_DECR: + case GL_DECR: dfop = STENCILOP_DECRSAT; break; case GL_INCR_WRAP: - dfop = STENCILOP_INCR; + dfop = STENCILOP_INCR; break; case GL_DECR_WRAP: - dfop = STENCILOP_DECR; + dfop = STENCILOP_DECR; break; - case GL_INVERT: - dfop = STENCILOP_INVERT; + case GL_INVERT: + dfop = STENCILOP_INVERT; break; - default: + default: break; } - switch(zpass) { - case GL_KEEP: - dpop = STENCILOP_KEEP; + switch (zpass) { + case GL_KEEP: + dpop = STENCILOP_KEEP; break; - case GL_ZERO: - dpop = STENCILOP_ZERO; + case GL_ZERO: + dpop = STENCILOP_ZERO; break; - case GL_REPLACE: - dpop = STENCILOP_REPLACE; + case GL_REPLACE: + dpop = STENCILOP_REPLACE; break; - case GL_INCR: + case GL_INCR: dpop = STENCILOP_INCRSAT; break; - case GL_DECR: + case GL_DECR: dpop = STENCILOP_DECRSAT; break; case GL_INCR_WRAP: - dpop = STENCILOP_INCR; + dpop = STENCILOP_INCR; break; case GL_DECR_WRAP: - dpop = STENCILOP_DECR; + dpop = STENCILOP_DECR; break; - case GL_INVERT: - dpop = STENCILOP_INVERT; + case GL_INVERT: + dpop = STENCILOP_INVERT; break; - default: + default: break; } @@ -188,27 +190,30 @@ i830StencilOpSeparate(GLcontext *ctx, GLenum face, GLenum fail, GLenum zfail, I830_STATECHANGE(i830, I830_UPLOAD_CTX); i830->state.Ctx[I830_CTXREG_STENCILTST] &= ~(STENCIL_OPS_MASK); i830->state.Ctx[I830_CTXREG_STENCILTST] |= (ENABLE_STENCIL_PARMS | - STENCIL_FAIL_OP(fop) | - STENCIL_PASS_DEPTH_FAIL_OP(dfop) | - STENCIL_PASS_DEPTH_PASS_OP(dpop)); + STENCIL_FAIL_OP(fop) | + STENCIL_PASS_DEPTH_FAIL_OP + (dfop) | + STENCIL_PASS_DEPTH_PASS_OP + (dpop)); } -static void i830AlphaFunc(GLcontext *ctx, GLenum func, GLfloat ref) +static void +i830AlphaFunc(GLcontext * ctx, GLenum func, GLfloat ref) { - i830ContextPtr i830 = I830_CONTEXT(ctx); + struct i830_context *i830 = i830_context(ctx); int test = intel_translate_compare_func(func); GLubyte refByte; GLuint refInt; UNCLAMPED_FLOAT_TO_UBYTE(refByte, ref); - refInt = (GLuint)refByte; + refInt = (GLuint) refByte; I830_STATECHANGE(i830, I830_UPLOAD_CTX); i830->state.Ctx[I830_CTXREG_STATE2] &= ~ALPHA_TEST_REF_MASK; i830->state.Ctx[I830_CTXREG_STATE2] |= (ENABLE_ALPHA_TEST_FUNC | - ENABLE_ALPHA_REF_VALUE | - ALPHA_TEST_FUNC(test) | - ALPHA_REF_VALUE(refInt)); + ENABLE_ALPHA_REF_VALUE | + ALPHA_TEST_FUNC(test) | + ALPHA_REF_VALUE(refInt)); } /** @@ -221,45 +226,49 @@ static void i830AlphaFunc(GLcontext *ctx, GLenum func, GLfloat ref) * This function is substantially different from the old i830-specific driver. * I'm not sure which is correct. */ -static void i830EvalLogicOpBlendState(GLcontext *ctx) +static void +i830EvalLogicOpBlendState(GLcontext * ctx) { - i830ContextPtr i830 = I830_CONTEXT(ctx); + struct i830_context *i830 = i830_context(ctx); I830_STATECHANGE(i830, I830_UPLOAD_CTX); if (RGBA_LOGICOP_ENABLED(ctx)) { i830->state.Ctx[I830_CTXREG_ENABLES_1] &= ~(ENABLE_COLOR_BLEND | - ENABLE_LOGIC_OP_MASK); + ENABLE_LOGIC_OP_MASK); i830->state.Ctx[I830_CTXREG_ENABLES_1] |= (DISABLE_COLOR_BLEND | - ENABLE_LOGIC_OP); - } else if (ctx->Color.BlendEnabled) { + ENABLE_LOGIC_OP); + } + else if (ctx->Color.BlendEnabled) { i830->state.Ctx[I830_CTXREG_ENABLES_1] &= ~(ENABLE_COLOR_BLEND | - ENABLE_LOGIC_OP_MASK); + ENABLE_LOGIC_OP_MASK); i830->state.Ctx[I830_CTXREG_ENABLES_1] |= (ENABLE_COLOR_BLEND | - DISABLE_LOGIC_OP); - } else { + DISABLE_LOGIC_OP); + } + else { i830->state.Ctx[I830_CTXREG_ENABLES_1] &= ~(ENABLE_COLOR_BLEND | - ENABLE_LOGIC_OP_MASK); + ENABLE_LOGIC_OP_MASK); i830->state.Ctx[I830_CTXREG_ENABLES_1] |= (DISABLE_COLOR_BLEND | - DISABLE_LOGIC_OP); + DISABLE_LOGIC_OP); } } -static void i830BlendColor(GLcontext *ctx, const GLfloat color[4]) +static void +i830BlendColor(GLcontext * ctx, const GLfloat color[4]) { - i830ContextPtr i830 = I830_CONTEXT(ctx); + struct i830_context *i830 = i830_context(ctx); GLubyte r, g, b, a; - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s\n", __FUNCTION__); - + DBG("%s\n", __FUNCTION__); + UNCLAMPED_FLOAT_TO_UBYTE(r, color[RCOMP]); UNCLAMPED_FLOAT_TO_UBYTE(g, color[GCOMP]); UNCLAMPED_FLOAT_TO_UBYTE(b, color[BCOMP]); UNCLAMPED_FLOAT_TO_UBYTE(a, color[ACOMP]); I830_STATECHANGE(i830, I830_UPLOAD_CTX); - i830->state.Ctx[I830_CTXREG_BLENDCOLOR1] = (a<<24) | (r<<16) | (g<<8) | b; + i830->state.Ctx[I830_CTXREG_BLENDCOLOR1] = + (a << 24) | (r << 16) | (g << 8) | b; } /** @@ -268,9 +277,10 @@ static void i830BlendColor(GLcontext *ctx, const GLfloat color[4]) * function because some blend equations (i.e., \c GL_MIN and \c GL_MAX) * change the interpretation of the blend function. */ -static void i830_set_blend_state( GLcontext * ctx ) +static void +i830_set_blend_state(GLcontext * ctx) { - i830ContextPtr i830 = I830_CONTEXT(ctx); + struct i830_context *i830 = i830_context(ctx); int funcA; int funcRGB; int eqnA; @@ -279,71 +289,72 @@ static void i830_set_blend_state( GLcontext * ctx ) int s1; - funcRGB = SRC_BLND_FACT( intel_translate_blend_factor( ctx->Color.BlendSrcRGB ) ) - | DST_BLND_FACT( intel_translate_blend_factor( ctx->Color.BlendDstRGB ) ); + funcRGB = + SRC_BLND_FACT(intel_translate_blend_factor(ctx->Color.BlendSrcRGB)) + | DST_BLND_FACT(intel_translate_blend_factor(ctx->Color.BlendDstRGB)); - switch(ctx->Color.BlendEquationRGB) { + switch (ctx->Color.BlendEquationRGB) { case GL_FUNC_ADD: - eqnRGB = BLENDFUNC_ADD; + eqnRGB = BLENDFUNC_ADD; break; case GL_MIN: eqnRGB = BLENDFUNC_MIN; funcRGB = SRC_BLND_FACT(BLENDFACT_ONE) | DST_BLND_FACT(BLENDFACT_ONE); break; - case GL_MAX: + case GL_MAX: eqnRGB = BLENDFUNC_MAX; funcRGB = SRC_BLND_FACT(BLENDFACT_ONE) | DST_BLND_FACT(BLENDFACT_ONE); break; - case GL_FUNC_SUBTRACT: - eqnRGB = BLENDFUNC_SUB; + case GL_FUNC_SUBTRACT: + eqnRGB = BLENDFUNC_SUB; break; case GL_FUNC_REVERSE_SUBTRACT: - eqnRGB = BLENDFUNC_RVRSE_SUB; + eqnRGB = BLENDFUNC_RVRSE_SUB; break; default: - fprintf( stderr, "[%s:%u] Invalid RGB blend equation (0x%04x).\n", - __FUNCTION__, __LINE__, ctx->Color.BlendEquationRGB ); + fprintf(stderr, "[%s:%u] Invalid RGB blend equation (0x%04x).\n", + __FUNCTION__, __LINE__, ctx->Color.BlendEquationRGB); return; } - funcA = SRC_ABLEND_FACT( intel_translate_blend_factor( ctx->Color.BlendSrcA ) ) - | DST_ABLEND_FACT( intel_translate_blend_factor( ctx->Color.BlendDstA ) ); + funcA = SRC_ABLEND_FACT(intel_translate_blend_factor(ctx->Color.BlendSrcA)) + | DST_ABLEND_FACT(intel_translate_blend_factor(ctx->Color.BlendDstA)); - switch(ctx->Color.BlendEquationA) { + switch (ctx->Color.BlendEquationA) { case GL_FUNC_ADD: - eqnA = BLENDFUNC_ADD; + eqnA = BLENDFUNC_ADD; break; - case GL_MIN: + case GL_MIN: eqnA = BLENDFUNC_MIN; funcA = SRC_BLND_FACT(BLENDFACT_ONE) | DST_BLND_FACT(BLENDFACT_ONE); break; - case GL_MAX: + case GL_MAX: eqnA = BLENDFUNC_MAX; funcA = SRC_BLND_FACT(BLENDFACT_ONE) | DST_BLND_FACT(BLENDFACT_ONE); break; - case GL_FUNC_SUBTRACT: - eqnA = BLENDFUNC_SUB; + case GL_FUNC_SUBTRACT: + eqnA = BLENDFUNC_SUB; break; case GL_FUNC_REVERSE_SUBTRACT: - eqnA = BLENDFUNC_RVRSE_SUB; + eqnA = BLENDFUNC_RVRSE_SUB; break; default: - fprintf( stderr, "[%s:%u] Invalid alpha blend equation (0x%04x).\n", - __FUNCTION__, __LINE__, ctx->Color.BlendEquationA ); + fprintf(stderr, "[%s:%u] Invalid alpha blend equation (0x%04x).\n", + __FUNCTION__, __LINE__, ctx->Color.BlendEquationA); return; } iab = eqnA | funcA - | _3DSTATE_INDPT_ALPHA_BLEND_CMD - | ENABLE_SRC_ABLEND_FACTOR | ENABLE_DST_ABLEND_FACTOR - | ENABLE_ALPHA_BLENDFUNC; + | _3DSTATE_INDPT_ALPHA_BLEND_CMD + | ENABLE_SRC_ABLEND_FACTOR | ENABLE_DST_ABLEND_FACTOR + | ENABLE_ALPHA_BLENDFUNC; s1 = eqnRGB | funcRGB - | _3DSTATE_MODES_1_CMD - | ENABLE_SRC_BLND_FACTOR | ENABLE_DST_BLND_FACTOR - | ENABLE_COLR_BLND_FUNC; + | _3DSTATE_MODES_1_CMD + | ENABLE_SRC_BLND_FACTOR | ENABLE_DST_BLND_FACTOR + | ENABLE_COLR_BLND_FUNC; - if ( (eqnA | funcA) != (eqnRGB | funcRGB) ) + if ((eqnA | funcA) != (eqnRGB | funcRGB)) iab |= ENABLE_INDPT_ALPHA_BLEND; else iab |= DISABLE_INDPT_ALPHA_BLEND; @@ -363,70 +374,68 @@ static void i830_set_blend_state( GLcontext * ctx ) i830EvalLogicOpBlendState(ctx); if (0) { - fprintf(stderr, "[%s:%u] STATE1: 0x%08x IALPHAB: 0x%08x blend is %sabled\n", - __FUNCTION__, __LINE__, - i830->state.Ctx[I830_CTXREG_STATE1], - i830->state.Ctx[I830_CTXREG_IALPHAB], - (ctx->Color.BlendEnabled) ? "en" : "dis"); + fprintf(stderr, + "[%s:%u] STATE1: 0x%08x IALPHAB: 0x%08x blend is %sabled\n", + __FUNCTION__, __LINE__, i830->state.Ctx[I830_CTXREG_STATE1], + i830->state.Ctx[I830_CTXREG_IALPHAB], + (ctx->Color.BlendEnabled) ? "en" : "dis"); } } -static void i830BlendEquationSeparate(GLcontext *ctx, GLenum modeRGB, - GLenum modeA) +static void +i830BlendEquationSeparate(GLcontext * ctx, GLenum modeRGB, GLenum modeA) { - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s -> %s, %s\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(modeRGB), - _mesa_lookup_enum_by_nr(modeA)); + DBG("%s -> %s, %s\n", __FUNCTION__, + _mesa_lookup_enum_by_nr(modeRGB), + _mesa_lookup_enum_by_nr(modeA)); (void) modeRGB; (void) modeA; - i830_set_blend_state( ctx ); + i830_set_blend_state(ctx); } -static void i830BlendFuncSeparate(GLcontext *ctx, GLenum sfactorRGB, - GLenum dfactorRGB, GLenum sfactorA, - GLenum dfactorA ) +static void +i830BlendFuncSeparate(GLcontext * ctx, GLenum sfactorRGB, + GLenum dfactorRGB, GLenum sfactorA, GLenum dfactorA) { - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s -> RGB(%s, %s) A(%s, %s)\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(sfactorRGB), - _mesa_lookup_enum_by_nr(dfactorRGB), - _mesa_lookup_enum_by_nr(sfactorA), - _mesa_lookup_enum_by_nr(dfactorA)); + DBG("%s -> RGB(%s, %s) A(%s, %s)\n", __FUNCTION__, + _mesa_lookup_enum_by_nr(sfactorRGB), + _mesa_lookup_enum_by_nr(dfactorRGB), + _mesa_lookup_enum_by_nr(sfactorA), + _mesa_lookup_enum_by_nr(dfactorA)); (void) sfactorRGB; (void) dfactorRGB; (void) sfactorA; (void) dfactorA; - i830_set_blend_state( ctx ); + i830_set_blend_state(ctx); } -static void i830DepthFunc(GLcontext *ctx, GLenum func) +static void +i830DepthFunc(GLcontext * ctx, GLenum func) { - i830ContextPtr i830 = I830_CONTEXT(ctx); + struct i830_context *i830 = i830_context(ctx); int test = intel_translate_compare_func(func); - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s\n", __FUNCTION__); - + DBG("%s\n", __FUNCTION__); + I830_STATECHANGE(i830, I830_UPLOAD_CTX); i830->state.Ctx[I830_CTXREG_STATE3] &= ~DEPTH_TEST_FUNC_MASK; i830->state.Ctx[I830_CTXREG_STATE3] |= (ENABLE_DEPTH_TEST_FUNC | - DEPTH_TEST_FUNC(test)); + DEPTH_TEST_FUNC(test)); } -static void i830DepthMask(GLcontext *ctx, GLboolean flag) +static void +i830DepthMask(GLcontext * ctx, GLboolean flag) { - i830ContextPtr i830 = I830_CONTEXT(ctx); - - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s flag (%d)\n", __FUNCTION__, flag); + struct i830_context *i830 = i830_context(ctx); + DBG("%s flag (%d)\n", __FUNCTION__, flag); + I830_STATECHANGE(i830, I830_UPLOAD_CTX); i830->state.Ctx[I830_CTXREG_ENABLES_2] &= ~ENABLE_DIS_DEPTH_WRITE_MASK; @@ -443,14 +452,15 @@ static void i830DepthMask(GLcontext *ctx, GLboolean flag) * The i830 supports a 4x4 stipple natively, GL wants 32x32. * Fortunately stipple is usually a repeating pattern. */ -static void i830PolygonStipple( GLcontext *ctx, const GLubyte *mask ) +static void +i830PolygonStipple(GLcontext * ctx, const GLubyte * mask) { - i830ContextPtr i830 = I830_CONTEXT(ctx); + struct i830_context *i830 = i830_context(ctx); const GLubyte *m = mask; GLubyte p[4]; - int i,j,k; + int i, j, k; int active = (ctx->Polygon.StippleFlag && - i830->intel.reduced_primitive == GL_TRIANGLES); + i830->intel.reduced_primitive == GL_TRIANGLES); GLuint newMask; if (active) { @@ -458,23 +468,26 @@ static void i830PolygonStipple( GLcontext *ctx, const GLubyte *mask ) i830->state.Stipple[I830_STPREG_ST1] &= ~ST1_ENABLE; } - p[0] = mask[12] & 0xf; p[0] |= p[0] << 4; - p[1] = mask[8] & 0xf; p[1] |= p[1] << 4; - p[2] = mask[4] & 0xf; p[2] |= p[2] << 4; - p[3] = mask[0] & 0xf; p[3] |= p[3] << 4; - - for (k = 0 ; k < 8 ; k++) - for (j = 3 ; j >= 0; j--) - for (i = 0 ; i < 4 ; i++, m++) - if (*m != p[j]) { - i830->intel.hw_stipple = 0; - return; - } + p[0] = mask[12] & 0xf; + p[0] |= p[0] << 4; + p[1] = mask[8] & 0xf; + p[1] |= p[1] << 4; + p[2] = mask[4] & 0xf; + p[2] |= p[2] << 4; + p[3] = mask[0] & 0xf; + p[3] |= p[3] << 4; + + for (k = 0; k < 8; k++) + for (j = 3; j >= 0; j--) + for (i = 0; i < 4; i++, m++) + if (*m != p[j]) { + i830->intel.hw_stipple = 0; + return; + } newMask = (((p[0] & 0xf) << 0) | - ((p[1] & 0xf) << 4) | - ((p[2] & 0xf) << 8) | - ((p[3] & 0xf) << 12)); + ((p[1] & 0xf) << 4) | + ((p[2] & 0xf) << 8) | ((p[3] & 0xf) << 12)); if (newMask == 0xffff || newMask == 0x0) { @@ -495,49 +508,54 @@ static void i830PolygonStipple( GLcontext *ctx, const GLubyte *mask ) /* ============================================================= * Hardware clipping */ -static void i830Scissor(GLcontext *ctx, GLint x, GLint y, - GLsizei w, GLsizei h) +static void +i830Scissor(GLcontext * ctx, GLint x, GLint y, GLsizei w, GLsizei h) { - i830ContextPtr i830 = I830_CONTEXT(ctx); - intelScreenPrivate *screen = i830->intel.intelScreen; + struct i830_context *i830 = i830_context(ctx); int x1, y1, x2, y2; - if (!i830->intel.driDrawable) + if (!ctx->DrawBuffer) return; - x1 = x; - y1 = i830->intel.driDrawable->h - (y + h); - x2 = x + w - 1; - y2 = y1 + h - 1; + DBG("%s %d,%d %dx%d\n", __FUNCTION__, x, y, w, h); - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "[%s] x(%d) y(%d) w(%d) h(%d)\n", __FUNCTION__, - x, y, w, h); - - if (x1 < 0) x1 = 0; - if (y1 < 0) y1 = 0; - if (x2 < 0) x2 = 0; - if (y2 < 0) y2 = 0; - - if (x2 >= screen->width) x2 = screen->width-1; - if (y2 >= screen->height) y2 = screen->height-1; - if (x1 >= screen->width) x1 = screen->width-1; - if (y1 >= screen->height) y1 = screen->height-1; + if (ctx->DrawBuffer->Name == 0) { + x1 = x; + y1 = ctx->DrawBuffer->Height - (y + h); + x2 = x + w - 1; + y2 = y1 + h - 1; + DBG("%s %d..%d,%d..%d (inverted)\n", __FUNCTION__, x1, x2, y1, y2); + } + else { + /* FBO - not inverted + */ + x1 = x; + y1 = y; + x2 = x + w - 1; + y2 = y + h - 1; + DBG("%s %d..%d,%d..%d (not inverted)\n", __FUNCTION__, x1, x2, y1, y2); + } + x1 = CLAMP(x1, 0, ctx->DrawBuffer->Width - 1); + y1 = CLAMP(y1, 0, ctx->DrawBuffer->Height - 1); + x2 = CLAMP(x2, 0, ctx->DrawBuffer->Width - 1); + y2 = CLAMP(y2, 0, ctx->DrawBuffer->Height - 1); + + DBG("%s %d..%d,%d..%d (clamped)\n", __FUNCTION__, x1, x2, y1, y2); I830_STATECHANGE(i830, I830_UPLOAD_BUFFERS); i830->state.Buffer[I830_DESTREG_SR1] = (y1 << 16) | (x1 & 0xffff); i830->state.Buffer[I830_DESTREG_SR2] = (y2 << 16) | (x2 & 0xffff); } -static void i830LogicOp(GLcontext *ctx, GLenum opcode) +static void +i830LogicOp(GLcontext * ctx, GLenum opcode) { - i830ContextPtr i830 = I830_CONTEXT(ctx); - int tmp = intel_translate_logic_op( opcode ); - - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s\n", __FUNCTION__); + struct i830_context *i830 = i830_context(ctx); + int tmp = intel_translate_logic_op(opcode); + DBG("%s\n", __FUNCTION__); + I830_STATECHANGE(i830, I830_UPLOAD_CTX); i830->state.Ctx[I830_CTXREG_STATE4] &= ~LOGICOP_MASK; i830->state.Ctx[I830_CTXREG_STATE4] |= LOGIC_OP_FUNC(tmp); @@ -545,14 +563,14 @@ static void i830LogicOp(GLcontext *ctx, GLenum opcode) -static void i830CullFaceFrontFace(GLcontext *ctx, GLenum unused) +static void +i830CullFaceFrontFace(GLcontext * ctx, GLenum unused) { - i830ContextPtr i830 = I830_CONTEXT(ctx); + struct i830_context *i830 = i830_context(ctx); GLuint mode; - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s\n", __FUNCTION__); - + DBG("%s\n", __FUNCTION__); + if (!ctx->Polygon.CullFlag) { mode = CULLMODE_NONE; } @@ -560,9 +578,9 @@ static void i830CullFaceFrontFace(GLcontext *ctx, GLenum unused) mode = CULLMODE_CW; if (ctx->Polygon.CullFaceMode == GL_FRONT) - mode ^= (CULLMODE_CW ^ CULLMODE_CCW); + mode ^= (CULLMODE_CW ^ CULLMODE_CCW); if (ctx->Polygon.FrontFace != GL_CCW) - mode ^= (CULLMODE_CW ^ CULLMODE_CCW); + mode ^= (CULLMODE_CW ^ CULLMODE_CCW); } else { mode = CULLMODE_BOTH; @@ -573,18 +591,18 @@ static void i830CullFaceFrontFace(GLcontext *ctx, GLenum unused) i830->state.Ctx[I830_CTXREG_STATE3] |= ENABLE_CULL_MODE | mode; } -static void i830LineWidth( GLcontext *ctx, GLfloat widthf ) +static void +i830LineWidth(GLcontext * ctx, GLfloat widthf) { - i830ContextPtr i830 = I830_CONTEXT( ctx ); + struct i830_context *i830 = i830_context(ctx); int width; int state5; - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s\n", __FUNCTION__); - - width = (int)(widthf * 2); - CLAMP_SELF(width, 1, 15); + DBG("%s\n", __FUNCTION__); + width = (int) (widthf * 2); + CLAMP_SELF(width, 1, 15); + state5 = i830->state.Ctx[I830_CTXREG_STATE5] & ~FIXED_LINE_WIDTH_MASK; state5 |= (ENABLE_FIXED_LINE_WIDTH | FIXED_LINE_WIDTH(width)); @@ -594,19 +612,19 @@ static void i830LineWidth( GLcontext *ctx, GLfloat widthf ) } } -static void i830PointSize(GLcontext *ctx, GLfloat size) +static void +i830PointSize(GLcontext * ctx, GLfloat size) { - i830ContextPtr i830 = I830_CONTEXT(ctx); - GLint point_size = (int)size; - - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s\n", __FUNCTION__); + struct i830_context *i830 = i830_context(ctx); + GLint point_size = (int) size; + DBG("%s\n", __FUNCTION__); + CLAMP_SELF(point_size, 1, 256); I830_STATECHANGE(i830, I830_UPLOAD_CTX); i830->state.Ctx[I830_CTXREG_STATE5] &= ~FIXED_POINT_WIDTH_MASK; i830->state.Ctx[I830_CTXREG_STATE5] |= (ENABLE_FIXED_POINT_WIDTH | - FIXED_POINT_WIDTH(point_size)); + FIXED_POINT_WIDTH(point_size)); } @@ -614,23 +632,21 @@ static void i830PointSize(GLcontext *ctx, GLfloat size) * Color masks */ -static void i830ColorMask(GLcontext *ctx, - GLboolean r, GLboolean g, - GLboolean b, GLboolean a) +static void +i830ColorMask(GLcontext * ctx, + GLboolean r, GLboolean g, GLboolean b, GLboolean a) { - i830ContextPtr i830 = I830_CONTEXT( ctx ); + struct i830_context *i830 = i830_context(ctx); GLuint tmp = 0; - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s r(%d) g(%d) b(%d) a(%d)\n", __FUNCTION__, r, g, b, a); + DBG("%s r(%d) g(%d) b(%d) a(%d)\n", __FUNCTION__, r, g, b, a); tmp = ((i830->state.Ctx[I830_CTXREG_ENABLES_2] & ~WRITEMASK_MASK) | - ENABLE_COLOR_MASK | - ENABLE_COLOR_WRITE | - ((!r) << WRITEMASK_RED_SHIFT) | - ((!g) << WRITEMASK_GREEN_SHIFT) | - ((!b) << WRITEMASK_BLUE_SHIFT) | - ((!a) << WRITEMASK_ALPHA_SHIFT)); + ENABLE_COLOR_MASK | + ENABLE_COLOR_WRITE | + ((!r) << WRITEMASK_RED_SHIFT) | + ((!g) << WRITEMASK_GREEN_SHIFT) | + ((!b) << WRITEMASK_BLUE_SHIFT) | ((!a) << WRITEMASK_ALPHA_SHIFT)); if (tmp != i830->state.Ctx[I830_CTXREG_ENABLES_2]) { I830_STATECHANGE(i830, I830_UPLOAD_CTX); @@ -638,9 +654,10 @@ static void i830ColorMask(GLcontext *ctx, } } -static void update_specular( GLcontext *ctx ) +static void +update_specular(GLcontext * ctx) { - i830ContextPtr i830 = I830_CONTEXT( ctx ); + struct i830_context *i830 = i830_context(ctx); I830_STATECHANGE(i830, I830_UPLOAD_CTX); i830->state.Ctx[I830_CTXREG_ENABLES_1] &= ~ENABLE_SPEC_ADD_MASK; @@ -651,22 +668,22 @@ static void update_specular( GLcontext *ctx ) i830->state.Ctx[I830_CTXREG_ENABLES_1] |= DISABLE_SPEC_ADD; } -static void i830LightModelfv(GLcontext *ctx, GLenum pname, - const GLfloat *param) +static void +i830LightModelfv(GLcontext * ctx, GLenum pname, const GLfloat * param) { - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s\n", __FUNCTION__); - + DBG("%s\n", __FUNCTION__); + if (pname == GL_LIGHT_MODEL_COLOR_CONTROL) { - update_specular( ctx ); + update_specular(ctx); } } /* In Mesa 3.5 we can reliably do native flatshading. */ -static void i830ShadeModel(GLcontext *ctx, GLenum mode) +static void +i830ShadeModel(GLcontext * ctx, GLenum mode) { - i830ContextPtr i830 = I830_CONTEXT(ctx); + struct i830_context *i830 = i830_context(ctx); I830_STATECHANGE(i830, I830_UPLOAD_CTX); @@ -675,58 +692,62 @@ static void i830ShadeModel(GLcontext *ctx, GLenum mode) i830->state.Ctx[I830_CTXREG_STATE3] &= ~SHADE_MODE_MASK; if (mode == GL_FLAT) { - i830->state.Ctx[I830_CTXREG_STATE3] |= (ALPHA_SHADE_MODE(SHADE_MODE_FLAT) | - FOG_SHADE_MODE(SHADE_MODE_FLAT) | - SPEC_SHADE_MODE(SHADE_MODE_FLAT) | - COLOR_SHADE_MODE(SHADE_MODE_FLAT)); - } else { - i830->state.Ctx[I830_CTXREG_STATE3] |= (ALPHA_SHADE_MODE(SHADE_MODE_LINEAR) | - FOG_SHADE_MODE(SHADE_MODE_LINEAR) | - SPEC_SHADE_MODE(SHADE_MODE_LINEAR) | - COLOR_SHADE_MODE(SHADE_MODE_LINEAR)); + i830->state.Ctx[I830_CTXREG_STATE3] |= + (ALPHA_SHADE_MODE(SHADE_MODE_FLAT) | FOG_SHADE_MODE(SHADE_MODE_FLAT) + | SPEC_SHADE_MODE(SHADE_MODE_FLAT) | + COLOR_SHADE_MODE(SHADE_MODE_FLAT)); + } + else { + i830->state.Ctx[I830_CTXREG_STATE3] |= + (ALPHA_SHADE_MODE(SHADE_MODE_LINEAR) | + FOG_SHADE_MODE(SHADE_MODE_LINEAR) | + SPEC_SHADE_MODE(SHADE_MODE_LINEAR) | + COLOR_SHADE_MODE(SHADE_MODE_LINEAR)); } } /* ============================================================= * Fog */ -static void i830Fogfv(GLcontext *ctx, GLenum pname, const GLfloat *param) +static void +i830Fogfv(GLcontext * ctx, GLenum pname, const GLfloat * param) { - i830ContextPtr i830 = I830_CONTEXT(ctx); + struct i830_context *i830 = i830_context(ctx); - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s\n", __FUNCTION__); - - if (pname == GL_FOG_COLOR) { - GLuint color = (((GLubyte)(ctx->Fog.Color[0]*255.0F) << 16) | - ((GLubyte)(ctx->Fog.Color[1]*255.0F) << 8) | - ((GLubyte)(ctx->Fog.Color[2]*255.0F) << 0)); + DBG("%s\n", __FUNCTION__); + + if (pname == GL_FOG_COLOR) { + GLuint color = (((GLubyte) (ctx->Fog.Color[0] * 255.0F) << 16) | + ((GLubyte) (ctx->Fog.Color[1] * 255.0F) << 8) | + ((GLubyte) (ctx->Fog.Color[2] * 255.0F) << 0)); I830_STATECHANGE(i830, I830_UPLOAD_CTX); - i830->state.Ctx[I830_CTXREG_FOGCOLOR] = (_3DSTATE_FOG_COLOR_CMD | color); + i830->state.Ctx[I830_CTXREG_FOGCOLOR] = + (_3DSTATE_FOG_COLOR_CMD | color); } } /* ============================================================= */ -static void i830Enable(GLcontext *ctx, GLenum cap, GLboolean state) +static void +i830Enable(GLcontext * ctx, GLenum cap, GLboolean state) { - i830ContextPtr i830 = I830_CONTEXT(ctx); + struct i830_context *i830 = i830_context(ctx); - switch(cap) { + switch (cap) { case GL_LIGHTING: case GL_COLOR_SUM: - update_specular( ctx ); + update_specular(ctx); break; case GL_ALPHA_TEST: I830_STATECHANGE(i830, I830_UPLOAD_CTX); i830->state.Ctx[I830_CTXREG_ENABLES_1] &= ~ENABLE_DIS_ALPHA_TEST_MASK; if (state) - i830->state.Ctx[I830_CTXREG_ENABLES_1] |= ENABLE_ALPHA_TEST; + i830->state.Ctx[I830_CTXREG_ENABLES_1] |= ENABLE_ALPHA_TEST; else - i830->state.Ctx[I830_CTXREG_ENABLES_1] |= DISABLE_ALPHA_TEST; + i830->state.Ctx[I830_CTXREG_ENABLES_1] |= DISABLE_ALPHA_TEST; break; @@ -740,17 +761,17 @@ static void i830Enable(GLcontext *ctx, GLenum cap, GLboolean state) /* Logicop doesn't seem to work at 16bpp: */ if (i830->intel.intelScreen->cpp == 2) - FALLBACK( &i830->intel, I830_FALLBACK_LOGICOP, state ); + FALLBACK(&i830->intel, I830_FALLBACK_LOGICOP, state); break; - + case GL_DITHER: I830_STATECHANGE(i830, I830_UPLOAD_CTX); i830->state.Ctx[I830_CTXREG_ENABLES_2] &= ~ENABLE_DITHER; if (state) - i830->state.Ctx[I830_CTXREG_ENABLES_2] |= ENABLE_DITHER; + i830->state.Ctx[I830_CTXREG_ENABLES_2] |= ENABLE_DITHER; else - i830->state.Ctx[I830_CTXREG_ENABLES_2] |= DISABLE_DITHER; + i830->state.Ctx[I830_CTXREG_ENABLES_2] |= DISABLE_DITHER; break; case GL_DEPTH_TEST: @@ -758,46 +779,44 @@ static void i830Enable(GLcontext *ctx, GLenum cap, GLboolean state) i830->state.Ctx[I830_CTXREG_ENABLES_1] &= ~ENABLE_DIS_DEPTH_TEST_MASK; if (state) - i830->state.Ctx[I830_CTXREG_ENABLES_1] |= ENABLE_DEPTH_TEST; + i830->state.Ctx[I830_CTXREG_ENABLES_1] |= ENABLE_DEPTH_TEST; else - i830->state.Ctx[I830_CTXREG_ENABLES_1] |= DISABLE_DEPTH_TEST; + i830->state.Ctx[I830_CTXREG_ENABLES_1] |= DISABLE_DEPTH_TEST; /* Also turn off depth writes when GL_DEPTH_TEST is disabled: */ - i830DepthMask( ctx, ctx->Depth.Mask ); + i830DepthMask(ctx, ctx->Depth.Mask); break; case GL_SCISSOR_TEST: I830_STATECHANGE(i830, I830_UPLOAD_BUFFERS); - + if (state) - i830->state.Buffer[I830_DESTREG_SENABLE] = - (_3DSTATE_SCISSOR_ENABLE_CMD | - ENABLE_SCISSOR_RECT); + i830->state.Buffer[I830_DESTREG_SENABLE] = + (_3DSTATE_SCISSOR_ENABLE_CMD | ENABLE_SCISSOR_RECT); else - i830->state.Buffer[I830_DESTREG_SENABLE] = - (_3DSTATE_SCISSOR_ENABLE_CMD | - DISABLE_SCISSOR_RECT); + i830->state.Buffer[I830_DESTREG_SENABLE] = + (_3DSTATE_SCISSOR_ENABLE_CMD | DISABLE_SCISSOR_RECT); break; case GL_LINE_SMOOTH: I830_STATECHANGE(i830, I830_UPLOAD_CTX); - + i830->state.Ctx[I830_CTXREG_AA] &= ~AA_LINE_ENABLE; if (state) - i830->state.Ctx[I830_CTXREG_AA] |= AA_LINE_ENABLE; + i830->state.Ctx[I830_CTXREG_AA] |= AA_LINE_ENABLE; else - i830->state.Ctx[I830_CTXREG_AA] |= AA_LINE_DISABLE; + i830->state.Ctx[I830_CTXREG_AA] |= AA_LINE_DISABLE; break; case GL_FOG: I830_STATECHANGE(i830, I830_UPLOAD_CTX); i830->state.Ctx[I830_CTXREG_ENABLES_1] &= ~ENABLE_DIS_FOG_MASK; if (state) - i830->state.Ctx[I830_CTXREG_ENABLES_1] |= ENABLE_FOG; + i830->state.Ctx[I830_CTXREG_ENABLES_1] |= ENABLE_FOG; else - i830->state.Ctx[I830_CTXREG_ENABLES_1] |= DISABLE_FOG; + i830->state.Ctx[I830_CTXREG_ENABLES_1] |= DISABLE_FOG; break; case GL_CULL_FACE: @@ -808,20 +827,32 @@ static void i830Enable(GLcontext *ctx, GLenum cap, GLboolean state) break; case GL_STENCIL_TEST: - if (i830->intel.hw_stencil) { - I830_STATECHANGE(i830, I830_UPLOAD_CTX); - - if (state) { - i830->state.Ctx[I830_CTXREG_ENABLES_1] |= ENABLE_STENCIL_TEST; - i830->state.Ctx[I830_CTXREG_ENABLES_2] |= ENABLE_STENCIL_WRITE; - } else { - i830->state.Ctx[I830_CTXREG_ENABLES_1] &= ~ENABLE_STENCIL_TEST; - i830->state.Ctx[I830_CTXREG_ENABLES_2] &= ~ENABLE_STENCIL_WRITE; - i830->state.Ctx[I830_CTXREG_ENABLES_1] |= DISABLE_STENCIL_TEST; - i830->state.Ctx[I830_CTXREG_ENABLES_2] |= DISABLE_STENCIL_WRITE; - } - } else { - FALLBACK( &i830->intel, I830_FALLBACK_STENCIL, state ); + { + GLboolean hw_stencil = GL_FALSE; + if (ctx->DrawBuffer) { + struct intel_renderbuffer *irbStencil + = intel_get_renderbuffer(ctx->DrawBuffer, BUFFER_STENCIL); + hw_stencil = (irbStencil && irbStencil->region); + } + if (hw_stencil) { + I830_STATECHANGE(i830, I830_UPLOAD_CTX); + + if (state) { + i830->state.Ctx[I830_CTXREG_ENABLES_1] |= ENABLE_STENCIL_TEST; + i830->state.Ctx[I830_CTXREG_ENABLES_2] |= ENABLE_STENCIL_WRITE; + } + else { + i830->state.Ctx[I830_CTXREG_ENABLES_1] &= ~ENABLE_STENCIL_TEST; + i830->state.Ctx[I830_CTXREG_ENABLES_2] &= + ~ENABLE_STENCIL_WRITE; + i830->state.Ctx[I830_CTXREG_ENABLES_1] |= DISABLE_STENCIL_TEST; + i830->state.Ctx[I830_CTXREG_ENABLES_2] |= + DISABLE_STENCIL_WRITE; + } + } + else { + FALLBACK(&i830->intel, I830_FALLBACK_STENCIL, state); + } } break; @@ -830,13 +861,12 @@ static void i830Enable(GLcontext *ctx, GLenum cap, GLboolean state) * I'll do more testing later to find out exactly which hardware * supports it. Disabled for now. */ - if (i830->intel.hw_stipple && - i830->intel.reduced_primitive == GL_TRIANGLES) - { - I830_STATECHANGE(i830, I830_UPLOAD_STIPPLE); - i830->state.Stipple[I830_STPREG_ST1] &= ~ST1_ENABLE; - if (state) - i830->state.Stipple[I830_STPREG_ST1] |= ST1_ENABLE; + if (i830->intel.hw_stipple && + i830->intel.reduced_primitive == GL_TRIANGLES) { + I830_STATECHANGE(i830, I830_UPLOAD_STIPPLE); + i830->state.Stipple[I830_STPREG_ST1] &= ~ST1_ENABLE; + if (state) + i830->state.Stipple[I830_STPREG_ST1] |= ST1_ENABLE; } break; @@ -846,7 +876,8 @@ static void i830Enable(GLcontext *ctx, GLenum cap, GLboolean state) } -static void i830_init_packets( i830ContextPtr i830 ) +static void +i830_init_packets(struct i830_context *i830) { intelScreenPrivate *screen = i830->intel.intelScreen; @@ -855,197 +886,192 @@ static void i830_init_packets( i830ContextPtr i830 ) /* Set default blend state */ i830->state.TexBlend[0][0] = (_3DSTATE_MAP_BLEND_OP_CMD(0) | - TEXPIPE_COLOR | - ENABLE_TEXOUTPUT_WRT_SEL | - TEXOP_OUTPUT_CURRENT | - DISABLE_TEX_CNTRL_STAGE | - TEXOP_SCALE_1X | - TEXOP_MODIFY_PARMS | - TEXOP_LAST_STAGE | - TEXBLENDOP_ARG1); + TEXPIPE_COLOR | + ENABLE_TEXOUTPUT_WRT_SEL | + TEXOP_OUTPUT_CURRENT | + DISABLE_TEX_CNTRL_STAGE | + TEXOP_SCALE_1X | + TEXOP_MODIFY_PARMS | + TEXOP_LAST_STAGE | TEXBLENDOP_ARG1); i830->state.TexBlend[0][1] = (_3DSTATE_MAP_BLEND_OP_CMD(0) | - TEXPIPE_ALPHA | - ENABLE_TEXOUTPUT_WRT_SEL | - TEXOP_OUTPUT_CURRENT | - TEXOP_SCALE_1X | - TEXOP_MODIFY_PARMS | - TEXBLENDOP_ARG1); + TEXPIPE_ALPHA | + ENABLE_TEXOUTPUT_WRT_SEL | + TEXOP_OUTPUT_CURRENT | + TEXOP_SCALE_1X | + TEXOP_MODIFY_PARMS | TEXBLENDOP_ARG1); i830->state.TexBlend[0][2] = (_3DSTATE_MAP_BLEND_ARG_CMD(0) | - TEXPIPE_COLOR | - TEXBLEND_ARG1 | - TEXBLENDARG_MODIFY_PARMS | - TEXBLENDARG_DIFFUSE); + TEXPIPE_COLOR | + TEXBLEND_ARG1 | + TEXBLENDARG_MODIFY_PARMS | + TEXBLENDARG_DIFFUSE); i830->state.TexBlend[0][3] = (_3DSTATE_MAP_BLEND_ARG_CMD(0) | - TEXPIPE_ALPHA | - TEXBLEND_ARG1 | - TEXBLENDARG_MODIFY_PARMS | - TEXBLENDARG_DIFFUSE); + TEXPIPE_ALPHA | + TEXBLEND_ARG1 | + TEXBLENDARG_MODIFY_PARMS | + TEXBLENDARG_DIFFUSE); i830->state.TexBlendWordsUsed[0] = 4; - i830->state.Ctx[I830_CTXREG_VF] = 0; + i830->state.Ctx[I830_CTXREG_VF] = 0; i830->state.Ctx[I830_CTXREG_VF2] = 0; i830->state.Ctx[I830_CTXREG_AA] = (_3DSTATE_AA_CMD | - AA_LINE_ECAAR_WIDTH_ENABLE | - AA_LINE_ECAAR_WIDTH_1_0 | - AA_LINE_REGION_WIDTH_ENABLE | - AA_LINE_REGION_WIDTH_1_0 | - AA_LINE_DISABLE); + AA_LINE_ECAAR_WIDTH_ENABLE | + AA_LINE_ECAAR_WIDTH_1_0 | + AA_LINE_REGION_WIDTH_ENABLE | + AA_LINE_REGION_WIDTH_1_0 | + AA_LINE_DISABLE); i830->state.Ctx[I830_CTXREG_ENABLES_1] = (_3DSTATE_ENABLES_1_CMD | - DISABLE_LOGIC_OP | - DISABLE_STENCIL_TEST | - DISABLE_DEPTH_BIAS | - DISABLE_SPEC_ADD | - DISABLE_FOG | - DISABLE_ALPHA_TEST | - DISABLE_COLOR_BLEND | - DISABLE_DEPTH_TEST); - + DISABLE_LOGIC_OP | + DISABLE_STENCIL_TEST | + DISABLE_DEPTH_BIAS | + DISABLE_SPEC_ADD | + DISABLE_FOG | + DISABLE_ALPHA_TEST | + DISABLE_COLOR_BLEND | + DISABLE_DEPTH_TEST); + +#if 000 /* XXX all the stencil enable state is set in i830Enable(), right? */ if (i830->intel.hw_stencil) { i830->state.Ctx[I830_CTXREG_ENABLES_2] = (_3DSTATE_ENABLES_2_CMD | - ENABLE_STENCIL_WRITE | - ENABLE_TEX_CACHE | - ENABLE_DITHER | - ENABLE_COLOR_MASK | - /* set no color comps disabled */ - ENABLE_COLOR_WRITE | - ENABLE_DEPTH_WRITE); - } else { + ENABLE_STENCIL_WRITE | + ENABLE_TEX_CACHE | + ENABLE_DITHER | + ENABLE_COLOR_MASK | + /* set no color comps disabled */ + ENABLE_COLOR_WRITE | + ENABLE_DEPTH_WRITE); + } + else +#endif + { i830->state.Ctx[I830_CTXREG_ENABLES_2] = (_3DSTATE_ENABLES_2_CMD | - DISABLE_STENCIL_WRITE | - ENABLE_TEX_CACHE | - ENABLE_DITHER | - ENABLE_COLOR_MASK | - /* set no color comps disabled */ - ENABLE_COLOR_WRITE | - ENABLE_DEPTH_WRITE); + DISABLE_STENCIL_WRITE | + ENABLE_TEX_CACHE | + ENABLE_DITHER | + ENABLE_COLOR_MASK | + /* set no color comps disabled */ + ENABLE_COLOR_WRITE | + ENABLE_DEPTH_WRITE); } i830->state.Ctx[I830_CTXREG_STATE1] = (_3DSTATE_MODES_1_CMD | - ENABLE_COLR_BLND_FUNC | - BLENDFUNC_ADD | - ENABLE_SRC_BLND_FACTOR | - SRC_BLND_FACT(BLENDFACT_ONE) | - ENABLE_DST_BLND_FACTOR | - DST_BLND_FACT(BLENDFACT_ZERO) ); + ENABLE_COLR_BLND_FUNC | + BLENDFUNC_ADD | + ENABLE_SRC_BLND_FACTOR | + SRC_BLND_FACT(BLENDFACT_ONE) | + ENABLE_DST_BLND_FACTOR | + DST_BLND_FACT(BLENDFACT_ZERO)); i830->state.Ctx[I830_CTXREG_STATE2] = (_3DSTATE_MODES_2_CMD | - ENABLE_GLOBAL_DEPTH_BIAS | - GLOBAL_DEPTH_BIAS(0) | - ENABLE_ALPHA_TEST_FUNC | - ALPHA_TEST_FUNC(COMPAREFUNC_ALWAYS) | - ALPHA_REF_VALUE(0) ); + ENABLE_GLOBAL_DEPTH_BIAS | + GLOBAL_DEPTH_BIAS(0) | + ENABLE_ALPHA_TEST_FUNC | + ALPHA_TEST_FUNC(COMPAREFUNC_ALWAYS) + | ALPHA_REF_VALUE(0)); i830->state.Ctx[I830_CTXREG_STATE3] = (_3DSTATE_MODES_3_CMD | - ENABLE_DEPTH_TEST_FUNC | - DEPTH_TEST_FUNC(COMPAREFUNC_LESS) | - ENABLE_ALPHA_SHADE_MODE | - ALPHA_SHADE_MODE(SHADE_MODE_LINEAR) | - ENABLE_FOG_SHADE_MODE | - FOG_SHADE_MODE(SHADE_MODE_LINEAR) | - ENABLE_SPEC_SHADE_MODE | - SPEC_SHADE_MODE(SHADE_MODE_LINEAR) | - ENABLE_COLOR_SHADE_MODE | - COLOR_SHADE_MODE(SHADE_MODE_LINEAR) | - ENABLE_CULL_MODE | - CULLMODE_NONE); + ENABLE_DEPTH_TEST_FUNC | + DEPTH_TEST_FUNC(COMPAREFUNC_LESS) | + ENABLE_ALPHA_SHADE_MODE | + ALPHA_SHADE_MODE(SHADE_MODE_LINEAR) + | ENABLE_FOG_SHADE_MODE | + FOG_SHADE_MODE(SHADE_MODE_LINEAR) | + ENABLE_SPEC_SHADE_MODE | + SPEC_SHADE_MODE(SHADE_MODE_LINEAR) | + ENABLE_COLOR_SHADE_MODE | + COLOR_SHADE_MODE(SHADE_MODE_LINEAR) + | ENABLE_CULL_MODE | CULLMODE_NONE); i830->state.Ctx[I830_CTXREG_STATE4] = (_3DSTATE_MODES_4_CMD | - ENABLE_LOGIC_OP_FUNC | - LOGIC_OP_FUNC(LOGICOP_COPY) | - ENABLE_STENCIL_TEST_MASK | - STENCIL_TEST_MASK(0xff) | - ENABLE_STENCIL_WRITE_MASK | - STENCIL_WRITE_MASK(0xff)); + ENABLE_LOGIC_OP_FUNC | + LOGIC_OP_FUNC(LOGICOP_COPY) | + ENABLE_STENCIL_TEST_MASK | + STENCIL_TEST_MASK(0xff) | + ENABLE_STENCIL_WRITE_MASK | + STENCIL_WRITE_MASK(0xff)); i830->state.Ctx[I830_CTXREG_STENCILTST] = (_3DSTATE_STENCIL_TEST_CMD | - ENABLE_STENCIL_PARMS | - STENCIL_FAIL_OP(STENCILOP_KEEP) | - STENCIL_PASS_DEPTH_FAIL_OP(STENCILOP_KEEP) | - STENCIL_PASS_DEPTH_PASS_OP(STENCILOP_KEEP) | - ENABLE_STENCIL_TEST_FUNC | - STENCIL_TEST_FUNC(COMPAREFUNC_ALWAYS) | - ENABLE_STENCIL_REF_VALUE | - STENCIL_REF_VALUE(0) ); - - i830->state.Ctx[I830_CTXREG_STATE5] = (_3DSTATE_MODES_5_CMD | - FLUSH_TEXTURE_CACHE | - ENABLE_SPRITE_POINT_TEX | - SPRITE_POINT_TEX_OFF | - ENABLE_FIXED_LINE_WIDTH | - FIXED_LINE_WIDTH(0x2) | /* 1.0 */ - ENABLE_FIXED_POINT_WIDTH | - FIXED_POINT_WIDTH(1) ); + ENABLE_STENCIL_PARMS | + STENCIL_FAIL_OP(STENCILOP_KEEP) + | + STENCIL_PASS_DEPTH_FAIL_OP + (STENCILOP_KEEP) | + STENCIL_PASS_DEPTH_PASS_OP + (STENCILOP_KEEP) | + ENABLE_STENCIL_TEST_FUNC | + STENCIL_TEST_FUNC + (COMPAREFUNC_ALWAYS) | + ENABLE_STENCIL_REF_VALUE | + STENCIL_REF_VALUE(0)); + + i830->state.Ctx[I830_CTXREG_STATE5] = (_3DSTATE_MODES_5_CMD | FLUSH_TEXTURE_CACHE | ENABLE_SPRITE_POINT_TEX | SPRITE_POINT_TEX_OFF | ENABLE_FIXED_LINE_WIDTH | FIXED_LINE_WIDTH(0x2) | /* 1.0 */ + ENABLE_FIXED_POINT_WIDTH | + FIXED_POINT_WIDTH(1)); i830->state.Ctx[I830_CTXREG_IALPHAB] = (_3DSTATE_INDPT_ALPHA_BLEND_CMD | - DISABLE_INDPT_ALPHA_BLEND | - ENABLE_ALPHA_BLENDFUNC | - ABLENDFUNC_ADD); + DISABLE_INDPT_ALPHA_BLEND | + ENABLE_ALPHA_BLENDFUNC | + ABLENDFUNC_ADD); i830->state.Ctx[I830_CTXREG_FOGCOLOR] = (_3DSTATE_FOG_COLOR_CMD | - FOG_COLOR_RED(0) | - FOG_COLOR_GREEN(0) | - FOG_COLOR_BLUE(0)); + FOG_COLOR_RED(0) | + FOG_COLOR_GREEN(0) | + FOG_COLOR_BLUE(0)); i830->state.Ctx[I830_CTXREG_BLENDCOLOR0] = _3DSTATE_CONST_BLEND_COLOR_CMD; i830->state.Ctx[I830_CTXREG_BLENDCOLOR1] = 0; i830->state.Ctx[I830_CTXREG_MCSB0] = _3DSTATE_MAP_COORD_SETBIND_CMD; i830->state.Ctx[I830_CTXREG_MCSB1] = (TEXBIND_SET3(TEXCOORDSRC_VTXSET_3) | - TEXBIND_SET2(TEXCOORDSRC_VTXSET_2) | - TEXBIND_SET1(TEXCOORDSRC_VTXSET_1) | - TEXBIND_SET0(TEXCOORDSRC_VTXSET_0)); - + TEXBIND_SET2(TEXCOORDSRC_VTXSET_2) | + TEXBIND_SET1(TEXCOORDSRC_VTXSET_1) | + TEXBIND_SET0(TEXCOORDSRC_VTXSET_0)); + i830->state.Stipple[I830_STPREG_ST0] = _3DSTATE_STIPPLE; i830->state.Buffer[I830_DESTREG_CBUFADDR0] = _3DSTATE_BUF_INFO_CMD; - i830->state.Buffer[I830_DESTREG_CBUFADDR1] = - (BUF_3D_ID_COLOR_BACK | - BUF_3D_PITCH(screen->front.pitch) | /* pitch in bytes */ - BUF_3D_USE_FENCE); + i830->state.Buffer[I830_DESTREG_CBUFADDR1] = (BUF_3D_ID_COLOR_BACK | BUF_3D_PITCH(screen->front.pitch) | /* pitch in bytes */ + BUF_3D_USE_FENCE); i830->state.Buffer[I830_DESTREG_DBUFADDR0] = _3DSTATE_BUF_INFO_CMD; - i830->state.Buffer[I830_DESTREG_DBUFADDR1] = - (BUF_3D_ID_DEPTH | - BUF_3D_PITCH(screen->depth.pitch) | /* pitch in bytes */ - BUF_3D_USE_FENCE); - i830->state.Buffer[I830_DESTREG_DBUFADDR2] = screen->depth.offset; - + i830->state.Buffer[I830_DESTREG_DBUFADDR1] = (BUF_3D_ID_DEPTH | BUF_3D_PITCH(screen->depth.pitch) | /* pitch in bytes */ + BUF_3D_USE_FENCE); i830->state.Buffer[I830_DESTREG_DV0] = _3DSTATE_DST_BUF_VARS_CMD; +#if 0 switch (screen->fbFormat) { - case DV_PF_555: case DV_PF_565: - i830->state.Buffer[I830_DESTREG_DV1] = (DSTORG_HORT_BIAS(0x8) | /* .5 */ - DSTORG_VERT_BIAS(0x8) | /* .5 */ - screen->fbFormat | - DEPTH_IS_Z | - DEPTH_FRMT_16_FIXED); + i830->state.Buffer[I830_DESTREG_DV1] = (DSTORG_HORT_BIAS(0x8) | /* .5 */ + DSTORG_VERT_BIAS(0x8) | /* .5 */ + screen->fbFormat | + DEPTH_IS_Z | + DEPTH_FRMT_16_FIXED); break; case DV_PF_8888: - i830->state.Buffer[I830_DESTREG_DV1] = (DSTORG_HORT_BIAS(0x8) | /* .5 */ - DSTORG_VERT_BIAS(0x8) | /* .5 */ - screen->fbFormat | - DEPTH_IS_Z | - DEPTH_FRMT_24_FIXED_8_OTHER); + i830->state.Buffer[I830_DESTREG_DV1] = (DSTORG_HORT_BIAS(0x8) | /* .5 */ + DSTORG_VERT_BIAS(0x8) | /* .5 */ + screen->fbFormat | + DEPTH_IS_Z | + DEPTH_FRMT_24_FIXED_8_OTHER); break; } - +#endif i830->state.Buffer[I830_DESTREG_SENABLE] = (_3DSTATE_SCISSOR_ENABLE_CMD | - DISABLE_SCISSOR_RECT); + DISABLE_SCISSOR_RECT); i830->state.Buffer[I830_DESTREG_SR0] = _3DSTATE_SCISSOR_RECT_0_CMD; i830->state.Buffer[I830_DESTREG_SR1] = 0; i830->state.Buffer[I830_DESTREG_SR2] = 0; } -void i830InitStateFuncs( struct dd_function_table *functions ) +void +i830InitStateFuncs(struct dd_function_table *functions) { functions->AlphaFunc = i830AlphaFunc; functions->BlendColor = i830BlendColor; @@ -1070,20 +1096,21 @@ void i830InitStateFuncs( struct dd_function_table *functions ) functions->StencilOpSeparate = i830StencilOpSeparate; } -void i830InitState( i830ContextPtr i830 ) +void +i830InitState(struct i830_context *i830) { GLcontext *ctx = &i830->intel.ctx; - i830_init_packets( i830 ); + i830_init_packets(i830); _mesa_init_driver_state(ctx); - memcpy( &i830->initial, &i830->state, sizeof(i830->state) ); + memcpy(&i830->initial, &i830->state, sizeof(i830->state)); i830->current = &i830->state; i830->state.emitted = 0; - i830->state.active = (I830_UPLOAD_TEXBLEND(0) | - I830_UPLOAD_STIPPLE | - I830_UPLOAD_CTX | - I830_UPLOAD_BUFFERS); + i830->state.active = (I830_UPLOAD_INVARIENT | + I830_UPLOAD_TEXBLEND(0) | + I830_UPLOAD_STIPPLE | + I830_UPLOAD_CTX | I830_UPLOAD_BUFFERS); } diff --git a/src/mesa/drivers/dri/i915/i830_tex.c b/src/mesa/drivers/dri/i915/i830_tex.c index 3c4aedb35c..fed464d1aa 100644 --- a/src/mesa/drivers/dri/i915/i830_tex.c +++ b/src/mesa/drivers/dri/i915/i830_tex.c @@ -45,261 +45,13 @@ - -/** - * Set the texture wrap modes. - * - * The i830M (and related graphics cores) do not support GL_CLAMP. The Intel - * drivers for "other operating systems" implement GL_CLAMP as - * GL_CLAMP_TO_EDGE, so the same is done here. - * - * \param t Texture object whose wrap modes are to be set - * \param swrap Wrap mode for the \a s texture coordinate - * \param twrap Wrap mode for the \a t texture coordinate - */ -static void i830SetTexWrapping(i830TextureObjectPtr tex, - GLenum swrap, - GLenum twrap) -{ - tex->Setup[I830_TEXREG_MCS] &= ~(TEXCOORD_ADDR_U_MASK|TEXCOORD_ADDR_V_MASK); - - switch( swrap ) { - case GL_REPEAT: - tex->Setup[I830_TEXREG_MCS] |= TEXCOORD_ADDR_U_MODE(TEXCOORDMODE_WRAP); - break; - case GL_CLAMP: - case GL_CLAMP_TO_EDGE: - tex->Setup[I830_TEXREG_MCS] |= TEXCOORD_ADDR_U_MODE(TEXCOORDMODE_CLAMP); - break; - case GL_CLAMP_TO_BORDER: - tex->Setup[I830_TEXREG_MCS] |= - TEXCOORD_ADDR_U_MODE(TEXCOORDMODE_CLAMP_BORDER); - break; - case GL_MIRRORED_REPEAT: - tex->Setup[I830_TEXREG_MCS] |= - TEXCOORD_ADDR_U_MODE(TEXCOORDMODE_MIRROR); - break; - default: - break; - } - - switch( twrap ) { - case GL_REPEAT: - tex->Setup[I830_TEXREG_MCS] |= TEXCOORD_ADDR_V_MODE(TEXCOORDMODE_WRAP); - break; - case GL_CLAMP: - case GL_CLAMP_TO_EDGE: - tex->Setup[I830_TEXREG_MCS] |= TEXCOORD_ADDR_V_MODE(TEXCOORDMODE_CLAMP); - break; - case GL_CLAMP_TO_BORDER: - tex->Setup[I830_TEXREG_MCS] |= - TEXCOORD_ADDR_V_MODE(TEXCOORDMODE_CLAMP_BORDER); - break; - case GL_MIRRORED_REPEAT: - tex->Setup[I830_TEXREG_MCS] |= - TEXCOORD_ADDR_V_MODE(TEXCOORDMODE_MIRROR); - break; - default: - break; - } -} - - -/** - * Set the texture magnification and minification modes. - * - * \param t Texture whose filter modes are to be set - * \param minf Texture minification mode - * \param magf Texture magnification mode - * \param bias LOD bias for this texture unit. - */ - -static void i830SetTexFilter( i830TextureObjectPtr t, GLenum minf, GLenum magf, - GLfloat maxanisotropy ) -{ - int minFilt = 0, mipFilt = 0, magFilt = 0; - - if(INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s\n", __FUNCTION__); - - if ( maxanisotropy > 1.0 ) { - minFilt = FILTER_ANISOTROPIC; - magFilt = FILTER_ANISOTROPIC; - } - else { - switch (minf) { - case GL_NEAREST: - minFilt = FILTER_NEAREST; - mipFilt = MIPFILTER_NONE; - break; - case GL_LINEAR: - minFilt = FILTER_LINEAR; - mipFilt = MIPFILTER_NONE; - break; - case GL_NEAREST_MIPMAP_NEAREST: - minFilt = FILTER_NEAREST; - mipFilt = MIPFILTER_NEAREST; - break; - case GL_LINEAR_MIPMAP_NEAREST: - minFilt = FILTER_LINEAR; - mipFilt = MIPFILTER_NEAREST; - break; - case GL_NEAREST_MIPMAP_LINEAR: - minFilt = FILTER_NEAREST; - mipFilt = MIPFILTER_LINEAR; - break; - case GL_LINEAR_MIPMAP_LINEAR: - minFilt = FILTER_LINEAR; - mipFilt = MIPFILTER_LINEAR; - break; - default: - break; - } - - switch (magf) { - case GL_NEAREST: - magFilt = FILTER_NEAREST; - break; - case GL_LINEAR: - magFilt = FILTER_LINEAR; - break; - default: - break; - } - } - - t->Setup[I830_TEXREG_TM0S3] &= ~TM0S3_MIN_FILTER_MASK; - t->Setup[I830_TEXREG_TM0S3] &= ~TM0S3_MIP_FILTER_MASK; - t->Setup[I830_TEXREG_TM0S3] &= ~TM0S3_MAG_FILTER_MASK; - t->Setup[I830_TEXREG_TM0S3] |= ((minFilt << TM0S3_MIN_FILTER_SHIFT) | - (mipFilt << TM0S3_MIP_FILTER_SHIFT) | - (magFilt << TM0S3_MAG_FILTER_SHIFT)); -} - -static void i830SetTexBorderColor(i830TextureObjectPtr t, GLubyte color[4]) -{ - if(INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s\n", __FUNCTION__); - - t->Setup[I830_TEXREG_TM0S4] = - INTEL_PACKCOLOR8888(color[0],color[1],color[2],color[3]); -} - - -/** - * Allocate space for and load the mesa images into the texture memory block. - * This will happen before drawing with a new texture, or drawing with a - * texture after it was swapped out or teximaged again. - */ - -intelTextureObjectPtr i830AllocTexObj( struct gl_texture_object *texObj ) -{ - i830TextureObjectPtr t = CALLOC_STRUCT( i830_texture_object ); - if ( !t ) - return NULL; - - texObj->DriverData = t; - t->intel.base.tObj = texObj; - t->intel.dirty = I830_UPLOAD_TEX_ALL; - make_empty_list( &t->intel.base ); - - t->Setup[I830_TEXREG_TM0LI] = 0; /* not used */ - t->Setup[I830_TEXREG_TM0S0] = 0; - t->Setup[I830_TEXREG_TM0S1] = 0; - t->Setup[I830_TEXREG_TM0S2] = 0; - t->Setup[I830_TEXREG_TM0S3] = 0; - t->Setup[I830_TEXREG_MCS] = (_3DSTATE_MAP_COORD_SET_CMD | - MAP_UNIT(0) | - ENABLE_TEXCOORD_PARAMS | - TEXCOORDS_ARE_NORMAL | - TEXCOORDTYPE_CARTESIAN | - ENABLE_ADDR_V_CNTL | - TEXCOORD_ADDR_V_MODE(TEXCOORDMODE_WRAP) | - ENABLE_ADDR_U_CNTL | - TEXCOORD_ADDR_U_MODE(TEXCOORDMODE_WRAP)); - - - i830SetTexWrapping( t, texObj->WrapS, texObj->WrapT ); - i830SetTexFilter( t, texObj->MinFilter, texObj->MagFilter, - texObj->MaxAnisotropy ); - i830SetTexBorderColor( t, texObj->_BorderChan ); - - return &t->intel; -} - - -static void i830TexParameter( GLcontext *ctx, GLenum target, - struct gl_texture_object *tObj, - GLenum pname, const GLfloat *params ) -{ - i830TextureObjectPtr t = (i830TextureObjectPtr) tObj->DriverData; - if (!t) - return; - - switch (pname) { - case GL_TEXTURE_MIN_FILTER: - case GL_TEXTURE_MAG_FILTER: - case GL_TEXTURE_MAX_ANISOTROPY_EXT: - i830SetTexFilter( t, tObj->MinFilter, tObj->MagFilter, - tObj->MaxAnisotropy); - break; - - case GL_TEXTURE_WRAP_S: - case GL_TEXTURE_WRAP_T: - i830SetTexWrapping( t, tObj->WrapS, tObj->WrapT ); - break; - - case GL_TEXTURE_BORDER_COLOR: - i830SetTexBorderColor( t, tObj->_BorderChan ); - break; - - case GL_TEXTURE_BASE_LEVEL: - case GL_TEXTURE_MAX_LEVEL: - case GL_TEXTURE_MIN_LOD: - case GL_TEXTURE_MAX_LOD: - /* The i830 and its successors can do a lot of this without - * reloading the textures. A project for someone? - */ - intelFlush( ctx ); - driSwapOutTextureObject( (driTextureObject *) t ); - break; - - default: - return; - } - - t->intel.dirty = I830_UPLOAD_TEX_ALL; -} - - -static void i830TexEnv( GLcontext *ctx, GLenum target, - GLenum pname, const GLfloat *param ) +static void +i830TexEnv(GLcontext * ctx, GLenum target, + GLenum pname, const GLfloat * param) { - i830ContextPtr i830 = I830_CONTEXT( ctx ); - GLuint unit = ctx->Texture.CurrentUnit; switch (pname) { - case GL_TEXTURE_ENV_COLOR: -#if 0 - { - GLubyte r, g, b, a; - GLuint col; - - UNCLAMPED_FLOAT_TO_UBYTE(r, param[RCOMP]); - UNCLAMPED_FLOAT_TO_UBYTE(g, param[GCOMP]); - UNCLAMPED_FLOAT_TO_UBYTE(b, param[BCOMP]); - UNCLAMPED_FLOAT_TO_UBYTE(a, param[ACOMP]); - - col = ((a << 24) | (r << 16) | (g << 8) | b); - - if (col != i830->state.TexEnv[unit][I830_TEXENVREG_COL1]) { - I830_STATECHANGE(i830, I830_UPLOAD_TEXENV); - i830->state.TexEnv[unit][I830_TEXENVREG_COL1] = col; - } - - break; - } -#endif + case GL_TEXTURE_ENV_COLOR: case GL_TEXTURE_ENV_MODE: case GL_COMBINE_RGB: case GL_COMBINE_ALPHA: @@ -319,38 +71,30 @@ static void i830TexEnv( GLcontext *ctx, GLenum target, case GL_ALPHA_SCALE: break; - case GL_TEXTURE_LOD_BIAS: { - int b = (int) ((*param) * 16.0); - if (b > 63) b = 63; - if (b < -64) b = -64; - I830_STATECHANGE(i830, I830_UPLOAD_TEX(unit)); - i830->state.Tex[unit][I830_TEXREG_TM0S3] &= ~TM0S3_LOD_BIAS_MASK; - i830->state.Tex[unit][I830_TEXREG_TM0S3] |= - ((b << TM0S3_LOD_BIAS_SHIFT) & TM0S3_LOD_BIAS_MASK); - break; - } + case GL_TEXTURE_LOD_BIAS:{ + struct i830_context *i830 = i830_context(ctx); + GLuint unit = ctx->Texture.CurrentUnit; + int b = (int) ((*param) * 16.0); + if (b > 63) + b = 63; + if (b < -64) + b = -64; + I830_STATECHANGE(i830, I830_UPLOAD_TEX(unit)); + i830->lodbias_tm0s3[unit] = + ((b << TM0S3_LOD_BIAS_SHIFT) & TM0S3_LOD_BIAS_MASK); + break; + } default: break; } } -static void i830BindTexture( GLcontext *ctx, GLenum target, - struct gl_texture_object *texObj ) -{ - i830TextureObjectPtr tex; - - if (!texObj->DriverData) - i830AllocTexObj( texObj ); - - tex = (i830TextureObjectPtr)texObj->DriverData; -} -void i830InitTextureFuncs( struct dd_function_table *functions ) +void +i830InitTextureFuncs(struct dd_function_table *functions) { - functions->BindTexture = i830BindTexture; - functions->TexEnv = i830TexEnv; - functions->TexParameter = i830TexParameter; + functions->TexEnv = i830TexEnv; } diff --git a/src/mesa/drivers/dri/i915/i830_texblend.c b/src/mesa/drivers/dri/i915/i830_texblend.c index 49e0347643..58f220eb7c 100644 --- a/src/mesa/drivers/dri/i915/i830_texblend.c +++ b/src/mesa/drivers/dri/i915/i830_texblend.c @@ -46,46 +46,42 @@ /* ================================================================ * Texture combine functions */ -static GLuint pass_through( GLuint *state, GLuint blendUnit ) +static GLuint +pass_through(GLuint * state, GLuint blendUnit) { state[0] = (_3DSTATE_MAP_BLEND_OP_CMD(blendUnit) | - TEXPIPE_COLOR | - ENABLE_TEXOUTPUT_WRT_SEL | - TEXOP_OUTPUT_CURRENT | - DISABLE_TEX_CNTRL_STAGE | - TEXOP_SCALE_1X | - TEXOP_MODIFY_PARMS | - TEXBLENDOP_ARG1); + TEXPIPE_COLOR | + ENABLE_TEXOUTPUT_WRT_SEL | + TEXOP_OUTPUT_CURRENT | + DISABLE_TEX_CNTRL_STAGE | + TEXOP_SCALE_1X | TEXOP_MODIFY_PARMS | TEXBLENDOP_ARG1); state[1] = (_3DSTATE_MAP_BLEND_OP_CMD(blendUnit) | - TEXPIPE_ALPHA | - ENABLE_TEXOUTPUT_WRT_SEL | - TEXOP_OUTPUT_CURRENT | - TEXOP_SCALE_1X | - TEXOP_MODIFY_PARMS | - TEXBLENDOP_ARG1); + TEXPIPE_ALPHA | + ENABLE_TEXOUTPUT_WRT_SEL | + TEXOP_OUTPUT_CURRENT | + TEXOP_SCALE_1X | TEXOP_MODIFY_PARMS | TEXBLENDOP_ARG1); state[2] = (_3DSTATE_MAP_BLEND_ARG_CMD(blendUnit) | - TEXPIPE_COLOR | - TEXBLEND_ARG1 | - TEXBLENDARG_MODIFY_PARMS | - TEXBLENDARG_CURRENT); + TEXPIPE_COLOR | + TEXBLEND_ARG1 | + TEXBLENDARG_MODIFY_PARMS | TEXBLENDARG_CURRENT); state[3] = (_3DSTATE_MAP_BLEND_ARG_CMD(blendUnit) | - TEXPIPE_ALPHA | - TEXBLEND_ARG1 | - TEXBLENDARG_MODIFY_PARMS | - TEXBLENDARG_CURRENT); + TEXPIPE_ALPHA | + TEXBLEND_ARG1 | + TEXBLENDARG_MODIFY_PARMS | TEXBLENDARG_CURRENT); return 4; } -static GLuint emit_factor( GLuint blendUnit, GLuint *state, GLuint count, - const GLfloat *factor ) +static GLuint +emit_factor(GLuint blendUnit, GLuint * state, GLuint count, + const GLfloat * factor) { GLubyte r, g, b, a; GLuint col; - + if (0) fprintf(stderr, "emit constant %d: %.2f %.2f %.2f %.2f\n", - blendUnit, factor[0], factor[1], factor[2], factor[3]); + blendUnit, factor[0], factor[1], factor[2], factor[3]); UNCLAMPED_FLOAT_TO_UBYTE(r, factor[0]); UNCLAMPED_FLOAT_TO_UBYTE(g, factor[1]); @@ -94,21 +90,27 @@ static GLuint emit_factor( GLuint blendUnit, GLuint *state, GLuint count, col = ((a << 24) | (r << 16) | (g << 8) | b); - state[count++] = _3DSTATE_COLOR_FACTOR_N_CMD(blendUnit); + state[count++] = _3DSTATE_COLOR_FACTOR_N_CMD(blendUnit); state[count++] = col; return count; } -static __inline__ GLuint GetTexelOp(GLint unit) +static INLINE GLuint +GetTexelOp(GLint unit) { - switch(unit) { - case 0: return TEXBLENDARG_TEXEL0; - case 1: return TEXBLENDARG_TEXEL1; - case 2: return TEXBLENDARG_TEXEL2; - case 3: return TEXBLENDARG_TEXEL3; - default: return TEXBLENDARG_TEXEL0; + switch (unit) { + case 0: + return TEXBLENDARG_TEXEL0; + case 1: + return TEXBLENDARG_TEXEL1; + case 2: + return TEXBLENDARG_TEXEL2; + case 3: + return TEXBLENDARG_TEXEL3; + default: + return TEXBLENDARG_TEXEL0; } } @@ -132,12 +134,10 @@ static __inline__ GLuint GetTexelOp(GLint unit) * partial support for the extension? */ GLuint -i830SetTexEnvCombine(i830ContextPtr i830, - const struct gl_tex_env_combine_state * combine, - GLint blendUnit, - GLuint texel_op, - GLuint *state, - const GLfloat *factor ) +i830SetTexEnvCombine(struct i830_context * i830, + const struct gl_tex_env_combine_state * combine, + GLint blendUnit, + GLuint texel_op, GLuint * state, const GLfloat * factor) { const GLuint numColorArgs = combine->_NumArgsRGB; const GLuint numAlphaArgs = combine->_NumArgsA; @@ -162,7 +162,7 @@ i830SetTexEnvCombine(i830ContextPtr i830, TEXPIPE_ALPHA | TEXBLEND_ARG0 | TEXBLENDARG_MODIFY_PARMS, }; - if(INTEL_DEBUG&DEBUG_TEXTURE) + if (INTEL_DEBUG & DEBUG_TEXTURE) fprintf(stderr, "%s\n", __FUNCTION__); @@ -188,23 +188,23 @@ i830SetTexEnvCombine(i830ContextPtr i830, } - switch(combine->ModeRGB) { - case GL_REPLACE: + switch (combine->ModeRGB) { + case GL_REPLACE: blendop = TEXBLENDOP_ARG1; break; - case GL_MODULATE: + case GL_MODULATE: blendop = TEXBLENDOP_MODULATE; break; - case GL_ADD: + case GL_ADD: blendop = TEXBLENDOP_ADD; break; case GL_ADD_SIGNED: - blendop = TEXBLENDOP_ADDSIGNED; + blendop = TEXBLENDOP_ADDSIGNED; break; case GL_INTERPOLATE: - blendop = TEXBLENDOP_BLEND; + blendop = TEXBLENDOP_BLEND; break; - case GL_SUBTRACT: + case GL_SUBTRACT: blendop = TEXBLENDOP_SUBTRACT; break; case GL_DOT3_RGB_EXT: @@ -215,55 +215,54 @@ i830SetTexEnvCombine(i830ContextPtr i830, case GL_DOT3_RGBA: blendop = TEXBLENDOP_DOT3; break; - default: - return pass_through( state, blendUnit ); + default: + return pass_through(state, blendUnit); } blendop |= (rgb_shift << TEXOP_SCALE_SHIFT); /* Handle RGB args */ - for(i = 0; i < 3; i++) { - switch(combine->SourceRGB[i]) { - case GL_TEXTURE: - args_RGB[i] = texel_op; - break; + for (i = 0; i < 3; i++) { + switch (combine->SourceRGB[i]) { + case GL_TEXTURE: + args_RGB[i] = texel_op; + break; case GL_TEXTURE0: case GL_TEXTURE1: case GL_TEXTURE2: case GL_TEXTURE3: - args_RGB[i] = GetTexelOp( combine->SourceRGB[i] - GL_TEXTURE0 ); - break; + args_RGB[i] = GetTexelOp(combine->SourceRGB[i] - GL_TEXTURE0); + break; case GL_CONSTANT: - args_RGB[i] = TEXBLENDARG_FACTOR_N; - need_factor = 1; - break; + args_RGB[i] = TEXBLENDARG_FACTOR_N; + need_factor = 1; + break; case GL_PRIMARY_COLOR: - args_RGB[i] = TEXBLENDARG_DIFFUSE; - break; + args_RGB[i] = TEXBLENDARG_DIFFUSE; + break; case GL_PREVIOUS: - args_RGB[i] = TEXBLENDARG_CURRENT; - break; - default: - return pass_through( state, blendUnit ); + args_RGB[i] = TEXBLENDARG_CURRENT; + break; + default: + return pass_through(state, blendUnit); } - switch(combine->OperandRGB[i]) { - case GL_SRC_COLOR: - args_RGB[i] |= 0; - break; - case GL_ONE_MINUS_SRC_COLOR: - args_RGB[i] |= TEXBLENDARG_INV_ARG; - break; - case GL_SRC_ALPHA: - args_RGB[i] |= TEXBLENDARG_REPLICATE_ALPHA; - break; - case GL_ONE_MINUS_SRC_ALPHA: - args_RGB[i] |= (TEXBLENDARG_REPLICATE_ALPHA | - TEXBLENDARG_INV_ARG); - break; - default: - return pass_through( state, blendUnit ); + switch (combine->OperandRGB[i]) { + case GL_SRC_COLOR: + args_RGB[i] |= 0; + break; + case GL_ONE_MINUS_SRC_COLOR: + args_RGB[i] |= TEXBLENDARG_INV_ARG; + break; + case GL_SRC_ALPHA: + args_RGB[i] |= TEXBLENDARG_REPLICATE_ALPHA; + break; + case GL_ONE_MINUS_SRC_ALPHA: + args_RGB[i] |= (TEXBLENDARG_REPLICATE_ALPHA | TEXBLENDARG_INV_ARG); + break; + default: + return pass_through(state, blendUnit); } } @@ -275,76 +274,76 @@ i830SetTexEnvCombine(i830ContextPtr i830, * Note - the global factor is set up with alpha == .5, so * the alpha part of the DOT4 calculation should be zero. */ - if ( combine->ModeRGB == GL_DOT3_RGBA_EXT || - combine->ModeRGB == GL_DOT3_RGBA ) { + if (combine->ModeRGB == GL_DOT3_RGBA_EXT || + combine->ModeRGB == GL_DOT3_RGBA) { ablendop = TEXBLENDOP_DOT4; - args_A[0] = TEXBLENDARG_FACTOR; /* the global factor */ + args_A[0] = TEXBLENDARG_FACTOR; /* the global factor */ args_A[1] = TEXBLENDARG_FACTOR; args_A[2] = TEXBLENDARG_FACTOR; } else { - switch(combine->ModeA) { - case GL_REPLACE: - ablendop = TEXBLENDOP_ARG1; - break; - case GL_MODULATE: - ablendop = TEXBLENDOP_MODULATE; - break; - case GL_ADD: - ablendop = TEXBLENDOP_ADD; - break; + switch (combine->ModeA) { + case GL_REPLACE: + ablendop = TEXBLENDOP_ARG1; + break; + case GL_MODULATE: + ablendop = TEXBLENDOP_MODULATE; + break; + case GL_ADD: + ablendop = TEXBLENDOP_ADD; + break; case GL_ADD_SIGNED: - ablendop = TEXBLENDOP_ADDSIGNED; - break; + ablendop = TEXBLENDOP_ADDSIGNED; + break; case GL_INTERPOLATE: - ablendop = TEXBLENDOP_BLEND; - break; - case GL_SUBTRACT: - ablendop = TEXBLENDOP_SUBTRACT; - break; + ablendop = TEXBLENDOP_BLEND; + break; + case GL_SUBTRACT: + ablendop = TEXBLENDOP_SUBTRACT; + break; default: - return pass_through( state, blendUnit ); + return pass_through(state, blendUnit); } ablendop |= (alpha_shift << TEXOP_SCALE_SHIFT); /* Handle A args */ - for(i = 0; i < 3; i++) { - switch(combine->SourceA[i]) { - case GL_TEXTURE: - args_A[i] = texel_op; - break; - case GL_TEXTURE0: - case GL_TEXTURE1: - case GL_TEXTURE2: - case GL_TEXTURE3: - args_A[i] = GetTexelOp( combine->SourceA[i] - GL_TEXTURE0 ); - break; - case GL_CONSTANT: - args_A[i] = TEXBLENDARG_FACTOR_N; - need_factor = 1; - break; - case GL_PRIMARY_COLOR: - args_A[i] = TEXBLENDARG_DIFFUSE; - break; - case GL_PREVIOUS: - args_A[i] = TEXBLENDARG_CURRENT; - break; - default: - return pass_through( state, blendUnit ); - } - - switch(combine->OperandA[i]) { - case GL_SRC_ALPHA: - args_A[i] |= 0; - break; - case GL_ONE_MINUS_SRC_ALPHA: - args_A[i] |= TEXBLENDARG_INV_ARG; - break; - default: - return pass_through( state, blendUnit ); - } + for (i = 0; i < 3; i++) { + switch (combine->SourceA[i]) { + case GL_TEXTURE: + args_A[i] = texel_op; + break; + case GL_TEXTURE0: + case GL_TEXTURE1: + case GL_TEXTURE2: + case GL_TEXTURE3: + args_A[i] = GetTexelOp(combine->SourceA[i] - GL_TEXTURE0); + break; + case GL_CONSTANT: + args_A[i] = TEXBLENDARG_FACTOR_N; + need_factor = 1; + break; + case GL_PRIMARY_COLOR: + args_A[i] = TEXBLENDARG_DIFFUSE; + break; + case GL_PREVIOUS: + args_A[i] = TEXBLENDARG_CURRENT; + break; + default: + return pass_through(state, blendUnit); + } + + switch (combine->OperandA[i]) { + case GL_SRC_ALPHA: + args_A[i] |= 0; + break; + case GL_ONE_MINUS_SRC_ALPHA: + args_A[i] |= TEXBLENDARG_INV_ARG; + break; + default: + return pass_through(state, blendUnit); + } } } @@ -363,86 +362,86 @@ i830SetTexEnvCombine(i830ContextPtr i830, used = 0; state[used++] = (_3DSTATE_MAP_BLEND_OP_CMD(blendUnit) | - TEXPIPE_COLOR | - ENABLE_TEXOUTPUT_WRT_SEL | - TEXOP_OUTPUT_CURRENT | - DISABLE_TEX_CNTRL_STAGE | - TEXOP_MODIFY_PARMS | - blendop); + TEXPIPE_COLOR | + ENABLE_TEXOUTPUT_WRT_SEL | + TEXOP_OUTPUT_CURRENT | + DISABLE_TEX_CNTRL_STAGE | TEXOP_MODIFY_PARMS | blendop); state[used++] = (_3DSTATE_MAP_BLEND_OP_CMD(blendUnit) | - TEXPIPE_ALPHA | - ENABLE_TEXOUTPUT_WRT_SEL | - TEXOP_OUTPUT_CURRENT | - TEXOP_MODIFY_PARMS | - ablendop); + TEXPIPE_ALPHA | + ENABLE_TEXOUTPUT_WRT_SEL | + TEXOP_OUTPUT_CURRENT | TEXOP_MODIFY_PARMS | ablendop); - for ( i = 0 ; i < numColorArgs ; i++ ) { + for (i = 0; i < numColorArgs; i++) { state[used++] = (_3DSTATE_MAP_BLEND_ARG_CMD(blendUnit) | - tex_blend_rgb[i] | args_RGB[i]); + tex_blend_rgb[i] | args_RGB[i]); } - for ( i = 0 ; i < numAlphaArgs ; i++ ) { + for (i = 0; i < numAlphaArgs; i++) { state[used++] = (_3DSTATE_MAP_BLEND_ARG_CMD(blendUnit) | - tex_blend_a[i] | args_A[i]); + tex_blend_a[i] | args_A[i]); } - if (need_factor) - return emit_factor( blendUnit, state, used, factor ); - else + if (need_factor) + return emit_factor(blendUnit, state, used, factor); + else return used; } -static void emit_texblend( i830ContextPtr i830, GLuint unit, GLuint blendUnit, - GLboolean last_stage ) +static void +emit_texblend(struct i830_context *i830, GLuint unit, GLuint blendUnit, + GLboolean last_stage) { struct gl_texture_unit *texUnit = &i830->intel.ctx.Texture.Unit[unit]; GLuint tmp[I830_TEXBLEND_SIZE], tmp_sz; - if (0) fprintf(stderr, "%s unit %d\n", __FUNCTION__, unit); + if (0) + fprintf(stderr, "%s unit %d\n", __FUNCTION__, unit); /* Update i830->state.TexBlend - */ - tmp_sz = i830SetTexEnvCombine(i830, texUnit->_CurrentCombine, blendUnit, - GetTexelOp(unit), tmp, - texUnit->EnvColor ); + */ + tmp_sz = i830SetTexEnvCombine(i830, texUnit->_CurrentCombine, blendUnit, + GetTexelOp(unit), tmp, texUnit->EnvColor); - if (last_stage) + if (last_stage) tmp[0] |= TEXOP_LAST_STAGE; if (tmp_sz != i830->state.TexBlendWordsUsed[blendUnit] || - memcmp( tmp, i830->state.TexBlend[blendUnit], tmp_sz * sizeof(GLuint))) { - - I830_STATECHANGE( i830, I830_UPLOAD_TEXBLEND(blendUnit) ); - memcpy( i830->state.TexBlend[blendUnit], tmp, tmp_sz * sizeof(GLuint)); + memcmp(tmp, i830->state.TexBlend[blendUnit], + tmp_sz * sizeof(GLuint))) { + + I830_STATECHANGE(i830, I830_UPLOAD_TEXBLEND(blendUnit)); + memcpy(i830->state.TexBlend[blendUnit], tmp, tmp_sz * sizeof(GLuint)); i830->state.TexBlendWordsUsed[blendUnit] = tmp_sz; } I830_ACTIVESTATE(i830, I830_UPLOAD_TEXBLEND(blendUnit), GL_TRUE); } -static void emit_passthrough( i830ContextPtr i830 ) +static void +emit_passthrough(struct i830_context *i830) { GLuint tmp[I830_TEXBLEND_SIZE], tmp_sz; GLuint unit = 0; - tmp_sz = pass_through( tmp, unit ); + tmp_sz = pass_through(tmp, unit); tmp[0] |= TEXOP_LAST_STAGE; if (tmp_sz != i830->state.TexBlendWordsUsed[unit] || - memcmp( tmp, i830->state.TexBlend[unit], tmp_sz * sizeof(GLuint))) { - - I830_STATECHANGE( i830, I830_UPLOAD_TEXBLEND(unit) ); - memcpy( i830->state.TexBlend[unit], tmp, tmp_sz * sizeof(GLuint)); + memcmp(tmp, i830->state.TexBlend[unit], tmp_sz * sizeof(GLuint))) { + + I830_STATECHANGE(i830, I830_UPLOAD_TEXBLEND(unit)); + memcpy(i830->state.TexBlend[unit], tmp, tmp_sz * sizeof(GLuint)); i830->state.TexBlendWordsUsed[unit] = tmp_sz; } I830_ACTIVESTATE(i830, I830_UPLOAD_TEXBLEND(unit), GL_TRUE); } -void i830EmitTextureBlend( i830ContextPtr i830 ) +void +i830EmitTextureBlend(struct i830_context *i830) { GLcontext *ctx = &i830->intel.ctx; GLuint unit, last_stage = 0, blendunit = 0; @@ -450,16 +449,15 @@ void i830EmitTextureBlend( i830ContextPtr i830 ) I830_ACTIVESTATE(i830, I830_UPLOAD_TEXBLEND_ALL, GL_FALSE); if (ctx->Texture._EnabledUnits) { - for (unit = 0 ; unit < ctx->Const.MaxTextureUnits ; unit++) - if (ctx->Texture.Unit[unit]._ReallyEnabled) - last_stage = unit; + for (unit = 0; unit < ctx->Const.MaxTextureUnits; unit++) + if (ctx->Texture.Unit[unit]._ReallyEnabled) + last_stage = unit; - for (unit = 0 ; unit < ctx->Const.MaxTextureUnits ; unit++) - if (ctx->Texture.Unit[unit]._ReallyEnabled) - emit_texblend( i830, unit, blendunit++, last_stage == unit ); + for (unit = 0; unit < ctx->Const.MaxTextureUnits; unit++) + if (ctx->Texture.Unit[unit]._ReallyEnabled) + emit_texblend(i830, unit, blendunit++, last_stage == unit); } else { - emit_passthrough( i830 ); + emit_passthrough(i830); } } - diff --git a/src/mesa/drivers/dri/i915/i830_texstate.c b/src/mesa/drivers/dri/i915/i830_texstate.c index ba972dac8f..7613b9d2a6 100644 --- a/src/mesa/drivers/dri/i915/i830_texstate.c +++ b/src/mesa/drivers/dri/i915/i830_texstate.c @@ -25,459 +25,319 @@ * **************************************************************************/ -#include "glheader.h" -#include "macros.h" #include "mtypes.h" -#include "simple_list.h" #include "enums.h" #include "texformat.h" -#include "texstore.h" +#include "dri_bufmgr.h" -#include "mm.h" - -#include "intel_screen.h" -#include "intel_ioctl.h" +#include "intel_mipmap_tree.h" #include "intel_tex.h" #include "i830_context.h" #include "i830_reg.h" -static const GLint initial_offsets[6][2] = { {0,0}, - {0,2}, - {1,0}, - {1,2}, - {1,1}, - {1,3} }; - -static const GLint step_offsets[6][2] = { {0,2}, - {0,2}, - {-1,2}, - {-1,2}, - {-1,1}, - {-1,1} }; -#define I830_TEX_UNIT_ENABLED(unit) (1<<unit) -static GLboolean i830SetTexImages( i830ContextPtr i830, - struct gl_texture_object *tObj ) +static GLuint +translate_texture_format(GLuint mesa_format) { - GLuint total_height, pitch, i, textureFormat; - i830TextureObjectPtr t = (i830TextureObjectPtr) tObj->DriverData; - const struct gl_texture_image *baseImage = tObj->Image[0][tObj->BaseLevel]; - GLint firstLevel, lastLevel, numLevels; - - switch( baseImage->TexFormat->MesaFormat ) { + switch (mesa_format) { case MESA_FORMAT_L8: - t->intel.texelBytes = 1; - textureFormat = MAPSURF_8BIT | MT_8BIT_L8; - break; - + return MAPSURF_8BIT | MT_8BIT_L8; case MESA_FORMAT_I8: - t->intel.texelBytes = 1; - textureFormat = MAPSURF_8BIT | MT_8BIT_I8; - break; - + return MAPSURF_8BIT | MT_8BIT_I8; case MESA_FORMAT_A8: - t->intel.texelBytes = 1; - textureFormat = MAPSURF_8BIT | MT_8BIT_I8; /* Kludge -- check with conform, glean */ - break; - + return MAPSURF_8BIT | MT_8BIT_I8; /* Kludge! */ case MESA_FORMAT_AL88: - t->intel.texelBytes = 2; - textureFormat = MAPSURF_16BIT | MT_16BIT_AY88; - break; - + return MAPSURF_16BIT | MT_16BIT_AY88; case MESA_FORMAT_RGB565: - t->intel.texelBytes = 2; - textureFormat = MAPSURF_16BIT | MT_16BIT_RGB565; - break; - + return MAPSURF_16BIT | MT_16BIT_RGB565; case MESA_FORMAT_ARGB1555: - t->intel.texelBytes = 2; - textureFormat = MAPSURF_16BIT | MT_16BIT_ARGB1555; - break; - + return MAPSURF_16BIT | MT_16BIT_ARGB1555; case MESA_FORMAT_ARGB4444: - t->intel.texelBytes = 2; - textureFormat = MAPSURF_16BIT | MT_16BIT_ARGB4444; - break; - + return MAPSURF_16BIT | MT_16BIT_ARGB4444; case MESA_FORMAT_ARGB8888: - t->intel.texelBytes = 4; - textureFormat = MAPSURF_32BIT | MT_32BIT_ARGB8888; - break; - + return MAPSURF_32BIT | MT_32BIT_ARGB8888; case MESA_FORMAT_YCBCR_REV: - t->intel.texelBytes = 2; - textureFormat = (MAPSURF_422 | MT_422_YCRCB_NORMAL | - TM0S1_COLORSPACE_CONVERSION); - break; - + return (MAPSURF_422 | MT_422_YCRCB_NORMAL); case MESA_FORMAT_YCBCR: - t->intel.texelBytes = 2; - textureFormat = (MAPSURF_422 | MT_422_YCRCB_SWAPY | /* ??? */ - TM0S1_COLORSPACE_CONVERSION); - break; - + return (MAPSURF_422 | MT_422_YCRCB_SWAPY); case MESA_FORMAT_RGB_FXT1: case MESA_FORMAT_RGBA_FXT1: - t->intel.texelBytes = 2; - textureFormat = MAPSURF_COMPRESSED | MT_COMPRESS_FXT1; - break; - + return (MAPSURF_COMPRESSED | MT_COMPRESS_FXT1); case MESA_FORMAT_RGBA_DXT1: case MESA_FORMAT_RGB_DXT1: - /* - * DXTn pitches are Width/4 * blocksize in bytes - * for DXT1: blocksize=8 so Width/4*8 = Width * 2 - * for DXT3/5: blocksize=16 so Width/4*16 = Width * 4 - */ - t->intel.texelBytes = 2; - textureFormat = (MAPSURF_COMPRESSED | MT_COMPRESS_DXT1); - break; + return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT1); case MESA_FORMAT_RGBA_DXT3: - t->intel.texelBytes = 4; - textureFormat = (MAPSURF_COMPRESSED | MT_COMPRESS_DXT2_3); - break; + return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT2_3); case MESA_FORMAT_RGBA_DXT5: - t->intel.texelBytes = 4; - textureFormat = (MAPSURF_COMPRESSED | MT_COMPRESS_DXT4_5); - break; - + return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT4_5); default: - fprintf(stderr, "%s: bad image format\n", __FUNCTION__); + fprintf(stderr, "%s: bad image format %x\n", __FUNCTION__, mesa_format); abort(); + return 0; } - - /* Compute which mipmap levels we really want to send to the hardware. - * This depends on the base image size, GL_TEXTURE_MIN_LOD, - * GL_TEXTURE_MAX_LOD, GL_TEXTURE_BASE_LEVEL, and GL_TEXTURE_MAX_LEVEL. - * Yes, this looks overly complicated, but it's all needed. - */ - driCalculateTextureFirstLastLevel( (driTextureObject *) t ); +} - /* Figure out the amount of memory required to hold all the mipmap - * levels. Choose the smallest pitch to accomodate the largest - * mipmap: - */ - firstLevel = t->intel.base.firstLevel; - lastLevel = t->intel.base.lastLevel; - numLevels = lastLevel - firstLevel + 1; - /* All images must be loaded at this pitch. Count the number of - * lines required: - */ - switch (tObj->Target) { - case GL_TEXTURE_CUBE_MAP: { - const GLuint dim = tObj->Image[0][firstLevel]->Width; - GLuint face; - - pitch = dim * t->intel.texelBytes; - pitch *= 2; /* double pitch for cube layouts */ - pitch = (pitch + 3) & ~3; - - total_height = dim * 4; - - for ( face = 0 ; face < 6 ; face++) { - GLuint x = initial_offsets[face][0] * dim; - GLuint y = initial_offsets[face][1] * dim; - GLuint d = dim; - - t->intel.base.dirty_images[face] = ~0; - - assert(tObj->Image[face][firstLevel]->Width == dim); - assert(tObj->Image[face][firstLevel]->Height == dim); - - for (i = 0; i < numLevels; i++) { - t->intel.image[face][i].image = tObj->Image[face][firstLevel + i]; - if (!t->intel.image[face][i].image) { - fprintf(stderr, "no image %d %d\n", face, i); - break; /* can't happen */ - } - - t->intel.image[face][i].offset = - y * pitch + x * t->intel.texelBytes; - t->intel.image[face][i].internalFormat = baseImage->_BaseFormat; - - d >>= 1; - x += step_offsets[face][0] * d; - y += step_offsets[face][1] * d; - } - } - break; - } +/* The i915 (and related graphics cores) do not support GL_CLAMP. The + * Intel drivers for "other operating systems" implement GL_CLAMP as + * GL_CLAMP_TO_EDGE, so the same is done here. + */ +static GLuint +translate_wrap_mode(GLenum wrap) +{ + switch (wrap) { + case GL_REPEAT: + return TEXCOORDMODE_WRAP; + case GL_CLAMP: + case GL_CLAMP_TO_EDGE: + return TEXCOORDMODE_CLAMP; /* not really correct */ + case GL_CLAMP_TO_BORDER: + return TEXCOORDMODE_CLAMP_BORDER; + case GL_MIRRORED_REPEAT: + return TEXCOORDMODE_MIRROR; default: - pitch = tObj->Image[0][firstLevel]->Width * t->intel.texelBytes; - pitch = (pitch + 3) & ~3; - t->intel.base.dirty_images[0] = ~0; - - for ( total_height = i = 0 ; i < numLevels ; i++ ) { - t->intel.image[0][i].image = tObj->Image[0][firstLevel + i]; - if (!t->intel.image[0][i].image) - break; - - t->intel.image[0][i].offset = total_height * pitch; - t->intel.image[0][i].internalFormat = baseImage->_BaseFormat; - if (t->intel.image[0][i].image->IsCompressed) - { - if (t->intel.image[0][i].image->Height > 4) - total_height += t->intel.image[0][i].image->Height/4; - else - total_height += 1; - } - else - total_height += MAX2(2, t->intel.image[0][i].image->Height); - } - break; + return TEXCOORDMODE_WRAP; } - - t->intel.Pitch = pitch; - t->intel.base.totalSize = total_height*pitch; - t->intel.max_level = i-1; - t->Setup[I830_TEXREG_TM0S1] = - (((tObj->Image[0][firstLevel]->Height - 1) << TM0S1_HEIGHT_SHIFT) | - ((tObj->Image[0][firstLevel]->Width - 1) << TM0S1_WIDTH_SHIFT) | - textureFormat); - t->Setup[I830_TEXREG_TM0S2] = - (((pitch / 4) - 1) << TM0S2_PITCH_SHIFT) | - TM0S2_CUBE_FACE_ENA_MASK; - t->Setup[I830_TEXREG_TM0S3] &= ~TM0S3_MAX_MIP_MASK; - t->Setup[I830_TEXREG_TM0S3] &= ~TM0S3_MIN_MIP_MASK; - t->Setup[I830_TEXREG_TM0S3] |= ((numLevels - 1)*4) << TM0S3_MIN_MIP_SHIFT; - t->intel.dirty = I830_UPLOAD_TEX_ALL; - - return intelUploadTexImages( &i830->intel, &t->intel, 0 ); } -static void i830_import_tex_unit( i830ContextPtr i830, - i830TextureObjectPtr t, - GLuint unit ) +/* Recalculate all state from scratch. Perhaps not the most + * efficient, but this has gotten complex enough that we need + * something which is understandable and reliable. + */ +static GLboolean +i830_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3) { - if(INTEL_DEBUG&DEBUG_TEXTURE) - fprintf(stderr, "%s unit(%d)\n", __FUNCTION__, unit); - - if (i830->intel.CurrentTexObj[unit]) - i830->intel.CurrentTexObj[unit]->base.bound &= ~(1U << unit); - - i830->intel.CurrentTexObj[unit] = (intelTextureObjectPtr)t; - t->intel.base.bound |= (1 << unit); - - I830_STATECHANGE( i830, I830_UPLOAD_TEX(unit) ); - - i830->state.Tex[unit][I830_TEXREG_TM0LI] = (_3DSTATE_LOAD_STATE_IMMEDIATE_2 | - (LOAD_TEXTURE_MAP0 << unit) | 4); - i830->state.Tex[unit][I830_TEXREG_TM0S0] = (TM0S0_USE_FENCE | - t->intel.TextureOffset); - - i830->state.Tex[unit][I830_TEXREG_TM0S1] = t->Setup[I830_TEXREG_TM0S1]; - i830->state.Tex[unit][I830_TEXREG_TM0S2] = t->Setup[I830_TEXREG_TM0S2]; - - i830->state.Tex[unit][I830_TEXREG_TM0S3] &= TM0S3_LOD_BIAS_MASK; - i830->state.Tex[unit][I830_TEXREG_TM0S3] |= (t->Setup[I830_TEXREG_TM0S3] & - ~TM0S3_LOD_BIAS_MASK); - - i830->state.Tex[unit][I830_TEXREG_TM0S4] = t->Setup[I830_TEXREG_TM0S4]; - i830->state.Tex[unit][I830_TEXREG_MCS] = (t->Setup[I830_TEXREG_MCS] & - ~MAP_UNIT_MASK); - i830->state.Tex[unit][I830_TEXREG_CUBE] = t->Setup[I830_TEXREG_CUBE]; - i830->state.Tex[unit][I830_TEXREG_MCS] |= MAP_UNIT(unit); - - t->intel.dirty &= ~I830_UPLOAD_TEX(unit); -} - + GLcontext *ctx = &intel->ctx; + struct i830_context *i830 = i830_context(ctx); + struct gl_texture_object *tObj = ctx->Texture.Unit[unit]._Current; + struct intel_texture_object *intelObj = intel_texture_object(tObj); + struct gl_texture_image *firstImage; + GLuint *state = i830->state.Tex[unit], format, pitch; + memset(state, 0, sizeof(state)); -static GLboolean enable_tex_common( GLcontext *ctx, GLuint unit ) -{ - i830ContextPtr i830 = I830_CONTEXT(ctx); - struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; - struct gl_texture_object *tObj = texUnit->_Current; - i830TextureObjectPtr t = (i830TextureObjectPtr)tObj->DriverData; + /*We need to refcount these. */ - if (0) fprintf(stderr, "%s\n", __FUNCTION__); + if (i830->state.tex_buffer[unit] != NULL) { + dri_bo_unreference(i830->state.tex_buffer[unit]); + i830->state.tex_buffer[unit] = NULL; + } - /* Fallback if there's a texture border */ - if ( tObj->Image[0][tObj->BaseLevel]->Border > 0 ) { - fprintf(stderr, "Texture border\n"); + if (!intelObj->imageOverride && !intel_finalize_mipmap_tree(intel, unit)) return GL_FALSE; - } - /* Upload teximages (not pipelined) + /* Get first image here, since intelObj->firstLevel will get set in + * the intel_finalize_mipmap_tree() call above. */ - if (t->intel.base.dirty_images[0]) { - if (!i830SetTexImages( i830, tObj )) { - return GL_FALSE; + firstImage = tObj->Image[0][intelObj->firstLevel]; + + if (intelObj->imageOverride) { + i830->state.tex_buffer[unit] = NULL; + i830->state.tex_offset[unit] = intelObj->textureOffset; + + switch (intelObj->depthOverride) { + case 32: + format = MAPSURF_32BIT | MT_32BIT_ARGB8888; + break; + case 24: + default: + format = MAPSURF_32BIT | MT_32BIT_XRGB8888; + break; + case 16: + format = MAPSURF_16BIT | MT_16BIT_RGB565; + break; } - } - /* Update state if this is a different texture object to last - * time. - */ - if (i830->intel.CurrentTexObj[unit] != &t->intel || - (t->intel.dirty & I830_UPLOAD_TEX(unit))) { - i830_import_tex_unit( i830, t, unit); - } - - I830_ACTIVESTATE(i830, I830_UPLOAD_TEX(unit), GL_TRUE); + pitch = intelObj->pitchOverride; + } else { + dri_bo_reference(intelObj->mt->region->buffer); + i830->state.tex_buffer[unit] = intelObj->mt->region->buffer; + i830->state.tex_offset[unit] = intel_miptree_image_offset(intelObj->mt, + 0, intelObj-> + firstLevel); - return GL_TRUE; -} - -static GLboolean enable_tex_rect( GLcontext *ctx, GLuint unit ) -{ - i830ContextPtr i830 = I830_CONTEXT(ctx); - GLuint mcs = i830->state.Tex[unit][I830_TEXREG_MCS]; - - mcs &= ~TEXCOORDS_ARE_NORMAL; - mcs |= TEXCOORDS_ARE_IN_TEXELUNITS; - - if ((mcs != i830->state.Tex[unit][I830_TEXREG_MCS]) - || (0 != i830->state.Tex[unit][I830_TEXREG_CUBE])) { - I830_STATECHANGE(i830, I830_UPLOAD_TEX(unit)); - i830->state.Tex[unit][I830_TEXREG_MCS] = mcs; - i830->state.Tex[unit][I830_TEXREG_CUBE] = 0; + format = translate_texture_format(firstImage->TexFormat->MesaFormat); + pitch = intelObj->mt->pitch * intelObj->mt->cpp; } - return GL_TRUE; -} + state[I830_TEXREG_TM0LI] = (_3DSTATE_LOAD_STATE_IMMEDIATE_2 | + (LOAD_TEXTURE_MAP0 << unit) | 4); +/* state[I830_TEXREG_TM0S0] = (TM0S0_USE_FENCE | */ +/* t->intel.TextureOffset); */ -static GLboolean enable_tex_2d( GLcontext *ctx, GLuint unit ) -{ - i830ContextPtr i830 = I830_CONTEXT(ctx); - GLuint mcs = i830->state.Tex[unit][I830_TEXREG_MCS]; - mcs &= ~TEXCOORDS_ARE_IN_TEXELUNITS; - mcs |= TEXCOORDS_ARE_NORMAL; + state[I830_TEXREG_TM0S1] = + (((firstImage->Height - 1) << TM0S1_HEIGHT_SHIFT) | + ((firstImage->Width - 1) << TM0S1_WIDTH_SHIFT) | format); + + state[I830_TEXREG_TM0S2] = + ((((pitch / 4) - 1) << TM0S2_PITCH_SHIFT) | TM0S2_CUBE_FACE_ENA_MASK); - if ((mcs != i830->state.Tex[unit][I830_TEXREG_MCS]) - || (0 != i830->state.Tex[unit][I830_TEXREG_CUBE])) { - I830_STATECHANGE(i830, I830_UPLOAD_TEX(unit)); - i830->state.Tex[unit][I830_TEXREG_MCS] = mcs; - i830->state.Tex[unit][I830_TEXREG_CUBE] = 0; + { + if (tObj->Target == GL_TEXTURE_CUBE_MAP) + state[I830_TEXREG_CUBE] = (_3DSTATE_MAP_CUBE | MAP_UNIT(unit) | + CUBE_NEGX_ENABLE | + CUBE_POSX_ENABLE | + CUBE_NEGY_ENABLE | + CUBE_POSY_ENABLE | + CUBE_NEGZ_ENABLE | CUBE_POSZ_ENABLE); + else + state[I830_TEXREG_CUBE] = (_3DSTATE_MAP_CUBE | MAP_UNIT(unit)); } - return GL_TRUE; -} - -static GLboolean enable_tex_cube( GLcontext *ctx, GLuint unit ) -{ - i830ContextPtr i830 = I830_CONTEXT(ctx); - struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; - struct gl_texture_object *tObj = texUnit->_Current; - i830TextureObjectPtr t = (i830TextureObjectPtr)tObj->DriverData; - GLuint mcs = i830->state.Tex[unit][I830_TEXREG_MCS]; - const GLuint cube = CUBE_NEGX_ENABLE | CUBE_POSX_ENABLE - | CUBE_NEGY_ENABLE | CUBE_POSY_ENABLE - | CUBE_NEGZ_ENABLE | CUBE_POSZ_ENABLE; - GLuint face; - - mcs &= ~TEXCOORDS_ARE_IN_TEXELUNITS; - mcs |= TEXCOORDS_ARE_NORMAL; - - if ((mcs != i830->state.Tex[unit][I830_TEXREG_MCS]) - || (cube != i830->state.Tex[unit][I830_TEXREG_CUBE])) { - I830_STATECHANGE(i830, I830_UPLOAD_TEX(unit)); - i830->state.Tex[unit][I830_TEXREG_MCS] = mcs; - i830->state.Tex[unit][I830_TEXREG_CUBE] = cube; - } - /* Upload teximages (not pipelined) - */ - if ( t->intel.base.dirty_images[0] || t->intel.base.dirty_images[1] || - t->intel.base.dirty_images[2] || t->intel.base.dirty_images[3] || - t->intel.base.dirty_images[4] || t->intel.base.dirty_images[5] ) { - i830SetTexImages( i830, tObj ); - } - /* upload (per face) */ - for (face = 0; face < 6; face++) { - if (t->intel.base.dirty_images[face]) { - if (!intelUploadTexImages( &i830->intel, &t->intel, face )) { - return GL_FALSE; - } + { + GLuint minFilt, mipFilt, magFilt; + + switch (tObj->MinFilter) { + case GL_NEAREST: + minFilt = FILTER_NEAREST; + mipFilt = MIPFILTER_NONE; + break; + case GL_LINEAR: + minFilt = FILTER_LINEAR; + mipFilt = MIPFILTER_NONE; + break; + case GL_NEAREST_MIPMAP_NEAREST: + minFilt = FILTER_NEAREST; + mipFilt = MIPFILTER_NEAREST; + break; + case GL_LINEAR_MIPMAP_NEAREST: + minFilt = FILTER_LINEAR; + mipFilt = MIPFILTER_NEAREST; + break; + case GL_NEAREST_MIPMAP_LINEAR: + minFilt = FILTER_NEAREST; + mipFilt = MIPFILTER_LINEAR; + break; + case GL_LINEAR_MIPMAP_LINEAR: + minFilt = FILTER_LINEAR; + mipFilt = MIPFILTER_LINEAR; + break; + default: + return GL_FALSE; } - } + if (tObj->MaxAnisotropy > 1.0) { + minFilt = FILTER_ANISOTROPIC; + magFilt = FILTER_ANISOTROPIC; + } + else { + switch (tObj->MagFilter) { + case GL_NEAREST: + magFilt = FILTER_NEAREST; + break; + case GL_LINEAR: + magFilt = FILTER_LINEAR; + break; + default: + return GL_FALSE; + } + } - return GL_TRUE; -} + state[I830_TEXREG_TM0S3] = i830->lodbias_tm0s3[unit]; +#if 0 + /* YUV conversion: + */ + if (firstImage->TexFormat->MesaFormat == MESA_FORMAT_YCBCR || + firstImage->TexFormat->MesaFormat == MESA_FORMAT_YCBCR_REV) + state[I830_TEXREG_TM0S3] |= SS2_COLORSPACE_CONVERSION; +#endif + + state[I830_TEXREG_TM0S3] |= ((intelObj->lastLevel - + intelObj->firstLevel) * + 4) << TM0S3_MIN_MIP_SHIFT; + + state[I830_TEXREG_TM0S3] |= ((minFilt << TM0S3_MIN_FILTER_SHIFT) | + (mipFilt << TM0S3_MIP_FILTER_SHIFT) | + (magFilt << TM0S3_MAG_FILTER_SHIFT)); + } -static GLboolean disable_tex( GLcontext *ctx, GLuint unit ) -{ - i830ContextPtr i830 = I830_CONTEXT(ctx); + { + GLenum ws = tObj->WrapS; + GLenum wt = tObj->WrapT; - /* This is happening too often. I need to conditionally send diffuse - * state to the card. Perhaps a diffuse dirty flag of some kind. - * Will need to change this logic if more than 2 texture units are - * used. We need to only do this up to the last unit enabled, or unit - * one if nothing is enabled. - */ - if ( i830->intel.CurrentTexObj[unit] != NULL ) { - /* The old texture is no longer bound to this texture unit. - * Mark it as such. + /* 3D textures not available on i830 */ - - i830->intel.CurrentTexObj[unit]->base.bound &= ~(1U << 0); - i830->intel.CurrentTexObj[unit] = NULL; + if (tObj->Target == GL_TEXTURE_3D) + return GL_FALSE; + + state[I830_TEXREG_MCS] = (_3DSTATE_MAP_COORD_SET_CMD | + MAP_UNIT(unit) | + ENABLE_TEXCOORD_PARAMS | + ss3 | + ENABLE_ADDR_V_CNTL | + TEXCOORD_ADDR_V_MODE(translate_wrap_mode(wt)) + | ENABLE_ADDR_U_CNTL | + TEXCOORD_ADDR_U_MODE(translate_wrap_mode + (ws))); } - return GL_TRUE; -} -static GLboolean i830UpdateTexUnit( GLcontext *ctx, GLuint unit ) -{ - struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; + state[I830_TEXREG_TM0S4] = INTEL_PACKCOLOR8888(tObj->_BorderChan[0], + tObj->_BorderChan[1], + tObj->_BorderChan[2], + tObj->_BorderChan[3]); - if (texUnit->_ReallyEnabled && - INTEL_CONTEXT(ctx)->intelScreen->tex.size < 2048 * 1024) - return GL_FALSE; - switch(texUnit->_ReallyEnabled) { - case TEXTURE_1D_BIT: - case TEXTURE_2D_BIT: - return (enable_tex_common( ctx, unit ) && - enable_tex_2d( ctx, unit )); - case TEXTURE_RECT_BIT: - return (enable_tex_common( ctx, unit ) && - enable_tex_rect( ctx, unit )); - case TEXTURE_CUBE_BIT: - return (enable_tex_common( ctx, unit ) && - enable_tex_cube( ctx, unit )); - case 0: - return disable_tex( ctx, unit ); - default: - return GL_FALSE; - } + I830_ACTIVESTATE(i830, I830_UPLOAD_TEX(unit), GL_TRUE); + /* memcmp was already disabled, but definitely won't work as the + * region might now change and that wouldn't be detected: + */ + I830_STATECHANGE(i830, I830_UPLOAD_TEX(unit)); + return GL_TRUE; } -void i830UpdateTextureState( intelContextPtr intel ) -{ - i830ContextPtr i830 = I830_CONTEXT(intel); - GLcontext *ctx = &intel->ctx; - GLboolean ok; - - if (0) fprintf(stderr, "%s\n", __FUNCTION__); - I830_ACTIVESTATE(i830, I830_UPLOAD_TEX_ALL, GL_FALSE); - ok = (i830UpdateTexUnit( ctx, 0 ) && - i830UpdateTexUnit( ctx, 1 ) && - i830UpdateTexUnit( ctx, 2 ) && - i830UpdateTexUnit( ctx, 3 )); +void +i830UpdateTextureState(struct intel_context *intel) +{ + struct i830_context *i830 = i830_context(&intel->ctx); + GLboolean ok = GL_TRUE; + GLuint i; + + for (i = 0; i < I830_TEX_UNITS && ok; i++) { + switch (intel->ctx.Texture.Unit[i]._ReallyEnabled) { + case TEXTURE_1D_BIT: + case TEXTURE_2D_BIT: + case TEXTURE_CUBE_BIT: + ok = i830_update_tex_unit(intel, i, TEXCOORDS_ARE_NORMAL); + break; + case TEXTURE_RECT_BIT: + ok = i830_update_tex_unit(intel, i, TEXCOORDS_ARE_IN_TEXELUNITS); + break; + case 0:{ + struct i830_context *i830 = i830_context(&intel->ctx); + if (i830->state.active & I830_UPLOAD_TEX(i)) + I830_ACTIVESTATE(i830, I830_UPLOAD_TEX(i), GL_FALSE); + + if (i830->state.tex_buffer[i] != NULL) { + dri_bo_unreference(i830->state.tex_buffer[i]); + i830->state.tex_buffer[i] = NULL; + } + break; + } + case TEXTURE_3D_BIT: + default: + ok = GL_FALSE; + break; + } + } - FALLBACK( intel, I830_FALLBACK_TEXTURE, !ok ); + FALLBACK(intel, I830_FALLBACK_TEXTURE, !ok); if (ok) - i830EmitTextureBlend( i830 ); + i830EmitTextureBlend(i830); } - - - diff --git a/src/mesa/drivers/dri/i915/i830_vtbl.c b/src/mesa/drivers/dri/i915/i830_vtbl.c index d40cf705a3..eecff2729f 100644 --- a/src/mesa/drivers/dri/i915/i830_vtbl.c +++ b/src/mesa/drivers/dri/i915/i830_vtbl.c @@ -28,14 +28,15 @@ #include "i830_context.h" #include "i830_reg.h" - #include "intel_batchbuffer.h" - +#include "intel_regions.h" #include "tnl/t_context.h" #include "tnl/t_vertex.h" -static GLboolean i830_check_vertex_size( intelContextPtr intel, - GLuint expected ); +#define FILE_DEBUG_FLAG DEBUG_STATE + +static GLboolean i830_check_vertex_size(struct intel_context *intel, + GLuint expected); #define SZ_TO_HW(sz) ((sz-2)&0x3) #define EMIT_SZ(sz) (EMIT_1F + (sz) - 1) @@ -59,10 +60,16 @@ do { \ #define VRTX_TEX_SET_FMT(n, x) ((x)<<((n)*2)) #define TEXBIND_SET(n, x) ((x)<<((n)*4)) -static void i830_render_start( intelContextPtr intel ) +static void +i830_render_prevalidate(struct intel_context *intel) +{ +} + +static void +i830_render_start(struct intel_context *intel) { GLcontext *ctx = &intel->ctx; - i830ContextPtr i830 = I830_CONTEXT(intel); + struct i830_context *i830 = i830_context(ctx); TNLcontext *tnl = TNL_CONTEXT(ctx); struct vertex_buffer *VB = &tnl->vb; DECLARE_RENDERINPUTS(index_bitset); @@ -70,7 +77,7 @@ static void i830_render_start( intelContextPtr intel ) GLuint v2 = _3DSTATE_VFT1_CMD; GLuint mcsb1 = 0; - RENDERINPUTS_COPY( index_bitset, tnl->render_inputs_bitset ); + RENDERINPUTS_COPY(index_bitset, tnl->render_inputs_bitset); /* Important: */ @@ -80,196 +87,215 @@ static void i830_render_start( intelContextPtr intel ) /* EMIT_ATTR's must be in order as they tell t_vertex.c how to * build up a hardware vertex. */ - if (RENDERINPUTS_TEST_RANGE( index_bitset, _TNL_FIRST_TEX, _TNL_LAST_TEX )) { - EMIT_ATTR( _TNL_ATTRIB_POS, EMIT_4F_VIEWPORT, VFT0_XYZW ); + if (RENDERINPUTS_TEST_RANGE(index_bitset, _TNL_FIRST_TEX, _TNL_LAST_TEX)) { + EMIT_ATTR(_TNL_ATTRIB_POS, EMIT_4F_VIEWPORT, VFT0_XYZW); intel->coloroffset = 4; } else { - EMIT_ATTR( _TNL_ATTRIB_POS, EMIT_3F_VIEWPORT, VFT0_XYZ ); + EMIT_ATTR(_TNL_ATTRIB_POS, EMIT_3F_VIEWPORT, VFT0_XYZ); intel->coloroffset = 3; } - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_POINTSIZE )) { - EMIT_ATTR( _TNL_ATTRIB_POINTSIZE, EMIT_1F, VFT0_POINT_WIDTH ); + if (RENDERINPUTS_TEST(index_bitset, _TNL_ATTRIB_POINTSIZE)) { + EMIT_ATTR(_TNL_ATTRIB_POINTSIZE, EMIT_1F, VFT0_POINT_WIDTH); } - EMIT_ATTR( _TNL_ATTRIB_COLOR0, EMIT_4UB_4F_BGRA, VFT0_DIFFUSE ); - + EMIT_ATTR(_TNL_ATTRIB_COLOR0, EMIT_4UB_4F_BGRA, VFT0_DIFFUSE); + intel->specoffset = 0; - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_COLOR1 ) || - RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_FOG )) { - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_COLOR1 )) { + if (RENDERINPUTS_TEST(index_bitset, _TNL_ATTRIB_COLOR1) || + RENDERINPUTS_TEST(index_bitset, _TNL_ATTRIB_FOG)) { + if (RENDERINPUTS_TEST(index_bitset, _TNL_ATTRIB_COLOR1)) { intel->specoffset = intel->coloroffset + 1; - EMIT_ATTR( _TNL_ATTRIB_COLOR1, EMIT_3UB_3F_BGR, VFT0_SPEC ); + EMIT_ATTR(_TNL_ATTRIB_COLOR1, EMIT_3UB_3F_BGR, VFT0_SPEC); } else - EMIT_PAD( 3 ); + EMIT_PAD(3); - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_FOG )) - EMIT_ATTR( _TNL_ATTRIB_FOG, EMIT_1UB_1F, VFT0_SPEC ); + if (RENDERINPUTS_TEST(index_bitset, _TNL_ATTRIB_FOG)) + EMIT_ATTR(_TNL_ATTRIB_FOG, EMIT_1UB_1F, VFT0_SPEC); else - EMIT_PAD( 1 ); + EMIT_PAD(1); } - if (RENDERINPUTS_TEST_RANGE( index_bitset, _TNL_FIRST_TEX, _TNL_LAST_TEX )) { + if (RENDERINPUTS_TEST_RANGE(index_bitset, _TNL_FIRST_TEX, _TNL_LAST_TEX)) { int i, count = 0; for (i = 0; i < I830_TEX_UNITS; i++) { - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX(i) )) { + if (RENDERINPUTS_TEST(index_bitset, _TNL_ATTRIB_TEX(i))) { GLuint sz = VB->TexCoordPtr[i]->size; GLuint emit; - GLuint mcs = (i830->state.Tex[i][I830_TEXREG_MCS] & + GLuint mcs = (i830->state.Tex[i][I830_TEXREG_MCS] & ~TEXCOORDTYPE_MASK); - switch (sz) { - case 1: - case 2: - emit = EMIT_2F; - sz = 2; - mcs |= TEXCOORDTYPE_CARTESIAN; - break; - case 3: - emit = EMIT_3F; - sz = 3; - mcs |= TEXCOORDTYPE_VECTOR; - break; - case 4: - emit = EMIT_3F_XYW; - sz = 3; - mcs |= TEXCOORDTYPE_HOMOGENEOUS; - break; - default: - continue; - }; - - - EMIT_ATTR( _TNL_ATTRIB_TEX0+i, emit, 0 ); - v2 |= VRTX_TEX_SET_FMT(count, SZ_TO_HW(sz)); - mcsb1 |= (count+8)<<(i*4); - - if (mcs != i830->state.Tex[i][I830_TEXREG_MCS]) { - I830_STATECHANGE(i830, I830_UPLOAD_TEX(i)); - i830->state.Tex[i][I830_TEXREG_MCS] = mcs; - } - - count++; - } + switch (sz) { + case 1: + case 2: + emit = EMIT_2F; + sz = 2; + mcs |= TEXCOORDTYPE_CARTESIAN; + break; + case 3: + emit = EMIT_3F; + sz = 3; + mcs |= TEXCOORDTYPE_VECTOR; + break; + case 4: + emit = EMIT_3F_XYW; + sz = 3; + mcs |= TEXCOORDTYPE_HOMOGENEOUS; + break; + default: + continue; + }; + + + EMIT_ATTR(_TNL_ATTRIB_TEX0 + i, emit, 0); + v2 |= VRTX_TEX_SET_FMT(count, SZ_TO_HW(sz)); + mcsb1 |= (count + 8) << (i * 4); + + if (mcs != i830->state.Tex[i][I830_TEXREG_MCS]) { + I830_STATECHANGE(i830, I830_UPLOAD_TEX(i)); + i830->state.Tex[i][I830_TEXREG_MCS] = mcs; + } + + count++; + } } v0 |= VFT0_TEX_COUNT(count); } - + /* Only need to change the vertex emit code if there has been a * statechange to a new hardware vertex format: */ if (v0 != i830->state.Ctx[I830_CTXREG_VF] || v2 != i830->state.Ctx[I830_CTXREG_VF2] || mcsb1 != i830->state.Ctx[I830_CTXREG_MCSB1] || - !RENDERINPUTS_EQUAL( index_bitset, i830->last_index_bitset )) { - - I830_STATECHANGE( i830, I830_UPLOAD_CTX ); + !RENDERINPUTS_EQUAL(index_bitset, i830->last_index_bitset)) { + int k; + + I830_STATECHANGE(i830, I830_UPLOAD_CTX); /* Must do this *after* statechange, so as not to affect * buffered vertices reliant on the old state: */ - intel->vertex_size = - _tnl_install_attrs( ctx, - intel->vertex_attrs, - intel->vertex_attr_count, - intel->ViewportMatrix.m, 0 ); + intel->vertex_size = + _tnl_install_attrs(ctx, + intel->vertex_attrs, + intel->vertex_attr_count, + intel->ViewportMatrix.m, 0); intel->vertex_size >>= 2; i830->state.Ctx[I830_CTXREG_VF] = v0; i830->state.Ctx[I830_CTXREG_VF2] = v2; i830->state.Ctx[I830_CTXREG_MCSB1] = mcsb1; - RENDERINPUTS_COPY( i830->last_index_bitset, index_bitset ); + RENDERINPUTS_COPY(i830->last_index_bitset, index_bitset); - assert(i830_check_vertex_size( intel, intel->vertex_size )); + k = i830_check_vertex_size(intel, intel->vertex_size); + assert(k); } } -static void i830_reduced_primitive_state( intelContextPtr intel, - GLenum rprim ) +static void +i830_reduced_primitive_state(struct intel_context *intel, GLenum rprim) { - i830ContextPtr i830 = I830_CONTEXT(intel); - GLuint st1 = i830->state.Stipple[I830_STPREG_ST1]; - - st1 &= ~ST1_ENABLE; - - switch (rprim) { - case GL_TRIANGLES: - if (intel->ctx.Polygon.StippleFlag && - intel->hw_stipple) - st1 |= ST1_ENABLE; - break; - case GL_LINES: - case GL_POINTS: - default: - break; - } - - i830->intel.reduced_primitive = rprim; - - if (st1 != i830->state.Stipple[I830_STPREG_ST1]) { - I830_STATECHANGE(i830, I830_UPLOAD_STIPPLE); - i830->state.Stipple[I830_STPREG_ST1] = st1; - } + struct i830_context *i830 = i830_context(&intel->ctx); + GLuint st1 = i830->state.Stipple[I830_STPREG_ST1]; + + st1 &= ~ST1_ENABLE; + + switch (rprim) { + case GL_TRIANGLES: + if (intel->ctx.Polygon.StippleFlag && intel->hw_stipple) + st1 |= ST1_ENABLE; + break; + case GL_LINES: + case GL_POINTS: + default: + break; + } + + i830->intel.reduced_primitive = rprim; + + if (st1 != i830->state.Stipple[I830_STPREG_ST1]) { + INTEL_FIREVERTICES(intel); + + I830_STATECHANGE(i830, I830_UPLOAD_STIPPLE); + i830->state.Stipple[I830_STPREG_ST1] = st1; + } } /* Pull apart the vertex format registers and figure out how large a * vertex is supposed to be. */ -static GLboolean i830_check_vertex_size( intelContextPtr intel, - GLuint expected ) +static GLboolean +i830_check_vertex_size(struct intel_context *intel, GLuint expected) { - i830ContextPtr i830 = I830_CONTEXT(intel); + struct i830_context *i830 = i830_context(&intel->ctx); int vft0 = i830->current->Ctx[I830_CTXREG_VF]; int vft1 = i830->current->Ctx[I830_CTXREG_VF2]; int nrtex = (vft0 & VFT0_TEX_COUNT_MASK) >> VFT0_TEX_COUNT_SHIFT; int i, sz = 0; switch (vft0 & VFT0_XYZW_MASK) { - case VFT0_XY: sz = 2; break; - case VFT0_XYZ: sz = 3; break; - case VFT0_XYW: sz = 3; break; - case VFT0_XYZW: sz = 4; break; - default: + case VFT0_XY: + sz = 2; + break; + case VFT0_XYZ: + sz = 3; + break; + case VFT0_XYW: + sz = 3; + break; + case VFT0_XYZW: + sz = 4; + break; + default: fprintf(stderr, "no xyzw specified\n"); return 0; } - if (vft0 & VFT0_SPEC) sz++; - if (vft0 & VFT0_DIFFUSE) sz++; - if (vft0 & VFT0_DEPTH_OFFSET) sz++; - if (vft0 & VFT0_POINT_WIDTH) sz++; - - for (i = 0 ; i < nrtex ; i++) { + if (vft0 & VFT0_SPEC) + sz++; + if (vft0 & VFT0_DIFFUSE) + sz++; + if (vft0 & VFT0_DEPTH_OFFSET) + sz++; + if (vft0 & VFT0_POINT_WIDTH) + sz++; + + for (i = 0; i < nrtex; i++) { switch (vft1 & VFT1_TEX0_MASK) { - case TEXCOORDFMT_2D: sz += 2; break; - case TEXCOORDFMT_3D: sz += 3; break; - case TEXCOORDFMT_4D: sz += 4; break; - case TEXCOORDFMT_1D: sz += 1; break; + case TEXCOORDFMT_2D: + sz += 2; + break; + case TEXCOORDFMT_3D: + sz += 3; + break; + case TEXCOORDFMT_4D: + sz += 4; + break; + case TEXCOORDFMT_1D: + sz += 1; + break; } vft1 >>= VFT1_TEX1_SHIFT; } - - if (sz != expected) + + if (sz != expected) fprintf(stderr, "vertex size mismatch %d/%d\n", sz, expected); - + return sz == expected; } -static void i830_emit_invarient_state( intelContextPtr intel ) +static void +i830_emit_invarient_state(struct intel_context *intel) { BATCH_LOCALS; - BEGIN_BATCH( 40 ); - - OUT_BATCH(_3DSTATE_MAP_CUBE | MAP_UNIT(0)); - OUT_BATCH(_3DSTATE_MAP_CUBE | MAP_UNIT(1)); - OUT_BATCH(_3DSTATE_MAP_CUBE | MAP_UNIT(2)); - OUT_BATCH(_3DSTATE_MAP_CUBE | MAP_UNIT(3)); + BEGIN_BATCH(40, 0); OUT_BATCH(_3DSTATE_DFLT_DIFFUSE_CMD); OUT_BATCH(0); @@ -282,37 +308,35 @@ static void i830_emit_invarient_state( intelContextPtr intel ) OUT_BATCH(_3DSTATE_FOG_MODE_CMD); OUT_BATCH(FOGFUNC_ENABLE | - FOG_LINEAR_CONST | - FOGSRC_INDEX_Z | - ENABLE_FOG_DENSITY); + FOG_LINEAR_CONST | FOGSRC_INDEX_Z | ENABLE_FOG_DENSITY); OUT_BATCH(0); OUT_BATCH(0); OUT_BATCH(_3DSTATE_MAP_TEX_STREAM_CMD | - MAP_UNIT(0) | - DISABLE_TEX_STREAM_BUMP | - ENABLE_TEX_STREAM_COORD_SET | - TEX_STREAM_COORD_SET(0) | - ENABLE_TEX_STREAM_MAP_IDX | TEX_STREAM_MAP_IDX(0)); + MAP_UNIT(0) | + DISABLE_TEX_STREAM_BUMP | + ENABLE_TEX_STREAM_COORD_SET | + TEX_STREAM_COORD_SET(0) | + ENABLE_TEX_STREAM_MAP_IDX | TEX_STREAM_MAP_IDX(0)); OUT_BATCH(_3DSTATE_MAP_TEX_STREAM_CMD | - MAP_UNIT(1) | - DISABLE_TEX_STREAM_BUMP | - ENABLE_TEX_STREAM_COORD_SET | - TEX_STREAM_COORD_SET(1) | - ENABLE_TEX_STREAM_MAP_IDX | TEX_STREAM_MAP_IDX(1)); + MAP_UNIT(1) | + DISABLE_TEX_STREAM_BUMP | + ENABLE_TEX_STREAM_COORD_SET | + TEX_STREAM_COORD_SET(1) | + ENABLE_TEX_STREAM_MAP_IDX | TEX_STREAM_MAP_IDX(1)); OUT_BATCH(_3DSTATE_MAP_TEX_STREAM_CMD | - MAP_UNIT(2) | - DISABLE_TEX_STREAM_BUMP | - ENABLE_TEX_STREAM_COORD_SET | - TEX_STREAM_COORD_SET(2) | - ENABLE_TEX_STREAM_MAP_IDX | TEX_STREAM_MAP_IDX(2)); + MAP_UNIT(2) | + DISABLE_TEX_STREAM_BUMP | + ENABLE_TEX_STREAM_COORD_SET | + TEX_STREAM_COORD_SET(2) | + ENABLE_TEX_STREAM_MAP_IDX | TEX_STREAM_MAP_IDX(2)); OUT_BATCH(_3DSTATE_MAP_TEX_STREAM_CMD | - MAP_UNIT(3) | - DISABLE_TEX_STREAM_BUMP | - ENABLE_TEX_STREAM_COORD_SET | - TEX_STREAM_COORD_SET(3) | - ENABLE_TEX_STREAM_MAP_IDX | TEX_STREAM_MAP_IDX(3)); + MAP_UNIT(3) | + DISABLE_TEX_STREAM_BUMP | + ENABLE_TEX_STREAM_COORD_SET | + TEX_STREAM_COORD_SET(3) | + ENABLE_TEX_STREAM_MAP_IDX | TEX_STREAM_MAP_IDX(3)); OUT_BATCH(_3DSTATE_MAP_COORD_TRANSFORM); OUT_BATCH(DISABLE_TEX_TRANSFORM | TEXTURE_SET(0)); @@ -324,21 +348,13 @@ static void i830_emit_invarient_state( intelContextPtr intel ) OUT_BATCH(DISABLE_TEX_TRANSFORM | TEXTURE_SET(3)); OUT_BATCH(_3DSTATE_RASTER_RULES_CMD | - ENABLE_POINT_RASTER_RULE | - OGL_POINT_RASTER_RULE | - ENABLE_LINE_STRIP_PROVOKE_VRTX | - ENABLE_TRI_FAN_PROVOKE_VRTX | - ENABLE_TRI_STRIP_PROVOKE_VRTX | - LINE_STRIP_PROVOKE_VRTX(1) | - TRI_FAN_PROVOKE_VRTX(2) | - TRI_STRIP_PROVOKE_VRTX(2)); - - OUT_BATCH(_3DSTATE_SCISSOR_ENABLE_CMD | - DISABLE_SCISSOR_RECT); - - OUT_BATCH(_3DSTATE_SCISSOR_RECT_0_CMD); - OUT_BATCH(0); - OUT_BATCH(0); + ENABLE_POINT_RASTER_RULE | + OGL_POINT_RASTER_RULE | + ENABLE_LINE_STRIP_PROVOKE_VRTX | + ENABLE_TRI_FAN_PROVOKE_VRTX | + ENABLE_TRI_STRIP_PROVOKE_VRTX | + LINE_STRIP_PROVOKE_VRTX(1) | + TRI_FAN_PROVOKE_VRTX(2) | TRI_STRIP_PROVOKE_VRTX(2)); OUT_BATCH(_3DSTATE_VERTEX_TRANSFORM); OUT_BATCH(DISABLE_VIEWPORT_TRANSFORM | DISABLE_PERSPECTIVE_DIVIDE); @@ -349,7 +365,7 @@ static void i830_emit_invarient_state( intelContextPtr intel ) OUT_BATCH(_3DSTATE_COLOR_FACTOR_CMD); - OUT_BATCH(0x80808080); /* .5 required in alpha for GL_DOT3_RGBA_EXT */ + OUT_BATCH(0x80808080); /* .5 required in alpha for GL_DOT3_RGBA_EXT */ ADVANCE_BATCH(); } @@ -358,13 +374,16 @@ static void i830_emit_invarient_state( intelContextPtr intel ) #define emit( intel, state, size ) \ do { \ int k; \ - BEGIN_BATCH( size / sizeof(GLuint)); \ - for (k = 0 ; k < size / sizeof(GLuint) ; k++) \ + BEGIN_BATCH(size / sizeof(GLuint), 0); \ + for (k = 0 ; k < size / sizeof(GLuint) ; k++) { \ + if (0) _mesa_printf(" 0x%08x\n", state[k]); \ OUT_BATCH(state[k]); \ + } \ ADVANCE_BATCH(); \ -} while (0); +} while (0) -static GLuint get_state_size( struct i830_hw_state *state ) +static GLuint +get_state_size(struct i830_hw_state *state) { GLuint dirty = state->active & ~state->emitted; GLuint sz = 0; @@ -373,21 +392,21 @@ static GLuint get_state_size( struct i830_hw_state *state ) if (dirty & I830_UPLOAD_INVARIENT) sz += 40 * sizeof(int); - if (dirty & I830_UPLOAD_CTX) + if (dirty & I830_UPLOAD_CTX) sz += sizeof(state->Ctx); - if (dirty & I830_UPLOAD_BUFFERS) + if (dirty & I830_UPLOAD_BUFFERS) sz += sizeof(state->Buffer); - if (dirty & I830_UPLOAD_STIPPLE) + if (dirty & I830_UPLOAD_STIPPLE) sz += sizeof(state->Stipple); for (i = 0; i < I830_TEX_UNITS; i++) { - if ((dirty & I830_UPLOAD_TEX(i))) - sz += sizeof(state->Tex[i]); + if ((dirty & I830_UPLOAD_TEX(i))) + sz += sizeof(state->Tex[i]); - if (dirty & I830_UPLOAD_TEXBLEND(i)) - sz += state->TexBlendWordsUsed[i] * 4; + if (dirty & I830_UPLOAD_TEXBLEND(i)) + sz += state->TexBlendWordsUsed[i] * 4; } return sz; @@ -396,139 +415,260 @@ static GLuint get_state_size( struct i830_hw_state *state ) /* Push the state into the sarea and/or texture memory. */ -static void i830_emit_state( intelContextPtr intel ) +static void +i830_emit_state(struct intel_context *intel) { - i830ContextPtr i830 = I830_CONTEXT(intel); + struct i830_context *i830 = i830_context(&intel->ctx); struct i830_hw_state *state = i830->current; int i; - GLuint dirty = state->active & ~state->emitted; - GLuint counter = intel->batch.counter; + GLuint dirty; BATCH_LOCALS; - if (intel->batch.space < get_state_size(state)) { - intelFlushBatch(intel, GL_TRUE); - dirty = state->active & ~state->emitted; - counter = intel->batch.counter; - } + /* We don't hold the lock at this point, so want to make sure that + * there won't be a buffer wrap. + * + * It might be better to talk about explicit places where + * scheduling is allowed, rather than assume that it is whenever a + * batchbuffer fills up. + */ + intel_batchbuffer_require_space(intel->batch, get_state_size(state), 0); + + /* Do this here as we may have flushed the batchbuffer above, + * causing more state to be dirty! + */ + dirty = state->active & ~state->emitted; if (dirty & I830_UPLOAD_INVARIENT) { - if (VERBOSE) fprintf(stderr, "I830_UPLOAD_INVARIENT:\n"); - i830_emit_invarient_state( intel ); + DBG("I830_UPLOAD_INVARIENT:\n"); + i830_emit_invarient_state(intel); } if (dirty & I830_UPLOAD_CTX) { - if (VERBOSE) fprintf(stderr, "I830_UPLOAD_CTX:\n"); - emit( i830, state->Ctx, sizeof(state->Ctx) ); + DBG("I830_UPLOAD_CTX:\n"); + emit(i830, state->Ctx, sizeof(state->Ctx)); + } if (dirty & I830_UPLOAD_BUFFERS) { - if (VERBOSE) fprintf(stderr, "I830_UPLOAD_BUFFERS:\n"); - emit( i830, state->Buffer, sizeof(state->Buffer) ); - } + DBG("I830_UPLOAD_BUFFERS:\n"); + BEGIN_BATCH(I830_DEST_SETUP_SIZE + 2, 0); + OUT_BATCH(state->Buffer[I830_DESTREG_CBUFADDR0]); + OUT_BATCH(state->Buffer[I830_DESTREG_CBUFADDR1]); + OUT_RELOC(state->draw_region->buffer, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, + state->draw_region->draw_offset); + + if (state->depth_region) { + OUT_BATCH(state->Buffer[I830_DESTREG_DBUFADDR0]); + OUT_BATCH(state->Buffer[I830_DESTREG_DBUFADDR1]); + OUT_RELOC(state->depth_region->buffer, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, + state->depth_region->draw_offset); + } + OUT_BATCH(state->Buffer[I830_DESTREG_DV0]); + OUT_BATCH(state->Buffer[I830_DESTREG_DV1]); + OUT_BATCH(state->Buffer[I830_DESTREG_SENABLE]); + OUT_BATCH(state->Buffer[I830_DESTREG_SR0]); + OUT_BATCH(state->Buffer[I830_DESTREG_SR1]); + OUT_BATCH(state->Buffer[I830_DESTREG_SR2]); + ADVANCE_BATCH(); + } + if (dirty & I830_UPLOAD_STIPPLE) { - if (VERBOSE) fprintf(stderr, "I830_UPLOAD_STIPPLE:\n"); - emit( i830, state->Stipple, sizeof(state->Stipple) ); + DBG("I830_UPLOAD_STIPPLE:\n"); + emit(i830, state->Stipple, sizeof(state->Stipple)); } for (i = 0; i < I830_TEX_UNITS; i++) { - if ((dirty & I830_UPLOAD_TEX(i))) { - if (VERBOSE) fprintf(stderr, "I830_UPLOAD_TEX(%d):\n", i); - emit( i830, state->Tex[i], sizeof(state->Tex[i])); - } + if ((dirty & I830_UPLOAD_TEX(i))) { + DBG("I830_UPLOAD_TEX(%d):\n", i); + + BEGIN_BATCH(I830_TEX_SETUP_SIZE + 1, 0); + OUT_BATCH(state->Tex[i][I830_TEXREG_TM0LI]); + + if (state->tex_buffer[i]) { + OUT_RELOC(state->tex_buffer[i], + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, + state->tex_offset[i] | TM0S0_USE_FENCE); + } + else if (state == &i830->meta) { + assert(i == 0); + OUT_BATCH(0); + } + else { + OUT_BATCH(state->tex_offset[i]); + } + + OUT_BATCH(state->Tex[i][I830_TEXREG_TM0S1]); + OUT_BATCH(state->Tex[i][I830_TEXREG_TM0S2]); + OUT_BATCH(state->Tex[i][I830_TEXREG_TM0S3]); + OUT_BATCH(state->Tex[i][I830_TEXREG_TM0S4]); + OUT_BATCH(state->Tex[i][I830_TEXREG_MCS]); + OUT_BATCH(state->Tex[i][I830_TEXREG_CUBE]); + } if (dirty & I830_UPLOAD_TEXBLEND(i)) { - if (VERBOSE) fprintf(stderr, "I830_UPLOAD_TEXBLEND(%d):\n", i); - emit( i830, state->TexBlend[i], - state->TexBlendWordsUsed[i] * 4 ); + DBG("I830_UPLOAD_TEXBLEND(%d): %d words\n", i, + state->TexBlendWordsUsed[i]); + emit(i830, state->TexBlend[i], state->TexBlendWordsUsed[i] * 4); } } state->emitted |= dirty; - intel->batch.last_emit_state = counter; - assert(counter == intel->batch.counter); } -static void i830_destroy_context( intelContextPtr intel ) +static void +i830_destroy_context(struct intel_context *intel) { + GLuint i; + struct i830_context *i830 = i830_context(&intel->ctx); + + for (i = 0; i < I830_TEX_UNITS; i++) { + if (i830->state.tex_buffer[i] != NULL) { + dri_bo_unreference(i830->state.tex_buffer[i]); + i830->state.tex_buffer[i] = NULL; + } + } + _tnl_free_vertices(&intel->ctx); } -static void -i830_set_color_region(intelContextPtr intel, const intelRegion *region) + +void +i830_state_draw_region(struct intel_context *intel, + struct i830_hw_state *state, + struct intel_region *color_region, + struct intel_region *depth_region) { - i830ContextPtr i830 = I830_CONTEXT(intel); - I830_STATECHANGE( i830, I830_UPLOAD_BUFFERS ); - i830->state.Buffer[I830_DESTREG_CBUFADDR1] = - (BUF_3D_ID_COLOR_BACK | BUF_3D_PITCH(region->pitch) | BUF_3D_USE_FENCE); - i830->state.Buffer[I830_DESTREG_CBUFADDR2] = region->offset; + struct i830_context *i830 = i830_context(&intel->ctx); + GLuint value; + + ASSERT(state == &i830->state || state == &i830->meta); + + if (state->draw_region != color_region) { + intel_region_release(&state->draw_region); + intel_region_reference(&state->draw_region, color_region); + } + if (state->depth_region != depth_region) { + intel_region_release(&state->depth_region); + intel_region_reference(&state->depth_region, depth_region); + } + + /* + * Set stride/cpp values + */ + if (color_region) { + state->Buffer[I830_DESTREG_CBUFADDR0] = _3DSTATE_BUF_INFO_CMD; + state->Buffer[I830_DESTREG_CBUFADDR1] = + (BUF_3D_ID_COLOR_BACK | + BUF_3D_PITCH(color_region->pitch * color_region->cpp) | + BUF_3D_USE_FENCE); + } + + if (depth_region) { + state->Buffer[I830_DESTREG_DBUFADDR0] = _3DSTATE_BUF_INFO_CMD; + state->Buffer[I830_DESTREG_DBUFADDR1] = + (BUF_3D_ID_DEPTH | + BUF_3D_PITCH(depth_region->pitch * depth_region->cpp) | + BUF_3D_USE_FENCE); + } + + /* + * Compute/set I830_DESTREG_DV1 value + */ + value = (DSTORG_HORT_BIAS(0x8) | /* .5 */ + DSTORG_VERT_BIAS(0x8) | DEPTH_IS_Z); /* .5 */ + + if (color_region && color_region->cpp == 4) { + value |= DV_PF_8888; + } + else { + value |= DV_PF_565; + } + if (depth_region && depth_region->cpp == 4) { + value |= DEPTH_FRMT_24_FIXED_8_OTHER; + } + else { + value |= DEPTH_FRMT_16_FIXED; + } + state->Buffer[I830_DESTREG_DV1] = value; + + I830_STATECHANGE(i830, I830_UPLOAD_BUFFERS); + + } static void -i830_set_z_region(intelContextPtr intel, const intelRegion *region) +i830_set_draw_region(struct intel_context *intel, + struct intel_region *color_region, + struct intel_region *depth_region) { - i830ContextPtr i830 = I830_CONTEXT(intel); - I830_STATECHANGE( i830, I830_UPLOAD_BUFFERS ); - i830->state.Buffer[I830_DESTREG_DBUFADDR1] = - (BUF_3D_ID_DEPTH | BUF_3D_PITCH(region->pitch) | BUF_3D_USE_FENCE); - i830->state.Buffer[I830_DESTREG_DBUFADDR2] = region->offset; + struct i830_context *i830 = i830_context(&intel->ctx); + i830_state_draw_region(intel, &i830->state, color_region, depth_region); } - +#if 0 static void i830_update_color_z_regions(intelContextPtr intel, - const intelRegion *colorRegion, - const intelRegion *depthRegion) + const intelRegion * colorRegion, + const intelRegion * depthRegion) { i830ContextPtr i830 = I830_CONTEXT(intel); i830->state.Buffer[I830_DESTREG_CBUFADDR1] = - (BUF_3D_ID_COLOR_BACK | BUF_3D_PITCH(colorRegion->pitch) | BUF_3D_USE_FENCE); + (BUF_3D_ID_COLOR_BACK | BUF_3D_PITCH(colorRegion->pitch) | + BUF_3D_USE_FENCE); i830->state.Buffer[I830_DESTREG_CBUFADDR2] = colorRegion->offset; i830->state.Buffer[I830_DESTREG_DBUFADDR1] = (BUF_3D_ID_DEPTH | BUF_3D_PITCH(depthRegion->pitch) | BUF_3D_USE_FENCE); i830->state.Buffer[I830_DESTREG_DBUFADDR2] = depthRegion->offset; } +#endif /* This isn't really handled at the moment. */ -static void i830_lost_hardware( intelContextPtr intel ) +static void +i830_lost_hardware(struct intel_context *intel) { - I830_CONTEXT(intel)->state.emitted = 0; + struct i830_context *i830 = i830_context(&intel->ctx); + i830->state.emitted = 0; } -static void i830_emit_flush( intelContextPtr intel ) +static GLuint +i830_flush_cmd(void) { - BATCH_LOCALS; - - BEGIN_BATCH(2); - OUT_BATCH( MI_FLUSH | FLUSH_MAP_CACHE ); - OUT_BATCH( 0 ); - ADVANCE_BATCH(); + return MI_FLUSH | FLUSH_MAP_CACHE; } +static void +i830_assert_not_dirty( struct intel_context *intel ) +{ + struct i830_context *i830 = i830_context(&intel->ctx); + struct i830_hw_state *state = i830->current; + GLuint dirty = state->active & ~state->emitted; + assert(!dirty); +} -void i830InitVtbl( i830ContextPtr i830 ) +void +i830InitVtbl(struct i830_context *i830) { - i830->intel.vtbl.alloc_tex_obj = i830AllocTexObj; i830->intel.vtbl.check_vertex_size = i830_check_vertex_size; - i830->intel.vtbl.clear_with_tris = i830ClearWithTris; - i830->intel.vtbl.rotate_window = i830RotateWindow; i830->intel.vtbl.destroy = i830_destroy_context; i830->intel.vtbl.emit_state = i830_emit_state; i830->intel.vtbl.lost_hardware = i830_lost_hardware; i830->intel.vtbl.reduced_primitive_state = i830_reduced_primitive_state; - i830->intel.vtbl.set_color_region = i830_set_color_region; - i830->intel.vtbl.set_z_region = i830_set_z_region; - i830->intel.vtbl.update_color_z_regions = i830_update_color_z_regions; + i830->intel.vtbl.set_draw_region = i830_set_draw_region; i830->intel.vtbl.update_texture_state = i830UpdateTextureState; - i830->intel.vtbl.emit_flush = i830_emit_flush; + i830->intel.vtbl.flush_cmd = i830_flush_cmd; i830->intel.vtbl.render_start = i830_render_start; + i830->intel.vtbl.render_prevalidate = i830_render_prevalidate; + i830->intel.vtbl.assert_not_dirty = i830_assert_not_dirty; } diff --git a/src/mesa/drivers/dri/i915/i915_context.c b/src/mesa/drivers/dri/i915/i915_context.c index 2bc1cae9c3..49e30141e4 100644 --- a/src/mesa/drivers/dri/i915/i915_context.c +++ b/src/mesa/drivers/dri/i915/i915_context.c @@ -36,84 +36,93 @@ #include "swrast/swrast.h" #include "swrast_setup/swrast_setup.h" #include "tnl/tnl.h" -#include "vbo/vbo.h" - #include "utils.h" #include "i915_reg.h" +#include "intel_regions.h" +#include "intel_batchbuffer.h" + /*************************************** * Mesa's Driver Functions ***************************************/ -static const struct dri_extension i915_extensions[] = -{ - { "GL_ARB_depth_texture", NULL }, - { "GL_ARB_fragment_program", NULL }, - { "GL_ARB_shadow", NULL }, - { "GL_ARB_texture_env_crossbar", NULL }, - { "GL_EXT_shadow_funcs", NULL }, - /* ARB extn won't work if not enabled */ - { "GL_SGIX_depth_texture", NULL }, - { NULL, NULL } +static const struct dri_extension i915_extensions[] = { + {"GL_ARB_depth_texture", NULL}, + {"GL_ARB_fragment_program", NULL}, + {"GL_ARB_shadow", NULL}, + {"GL_ARB_texture_env_crossbar", NULL}, + {"GL_ARB_texture_non_power_of_two", NULL}, + {"GL_EXT_shadow_funcs", NULL}, + /* ARB extn won't work if not enabled */ + {"GL_SGIX_depth_texture", NULL}, + {NULL, NULL} }; /* Override intel default. */ -static void i915InvalidateState( GLcontext *ctx, GLuint new_state ) +static void +i915InvalidateState(GLcontext * ctx, GLuint new_state) { - _swrast_InvalidateState( ctx, new_state ); - _swsetup_InvalidateState( ctx, new_state ); - _vbo_InvalidateState( ctx, new_state ); - _tnl_InvalidateState( ctx, new_state ); - _tnl_invalidate_vertex_state( ctx, new_state ); - INTEL_CONTEXT(ctx)->NewGLState |= new_state; + _swrast_InvalidateState(ctx, new_state); + _swsetup_InvalidateState(ctx, new_state); + _vbo_InvalidateState(ctx, new_state); + _tnl_InvalidateState(ctx, new_state); + _tnl_invalidate_vertex_state(ctx, new_state); + intel_context(ctx)->NewGLState |= new_state; /* Todo: gather state values under which tracked parameters become * invalidated, add callbacks for things like * ProgramLocalParameters, etc. */ { - struct i915_fragment_program *p = - (struct i915_fragment_program *)ctx->FragmentProgram._Current; + struct i915_fragment_program *p = + (struct i915_fragment_program *) ctx->FragmentProgram._Current; if (p && p->nr_params) - p->params_uptodate = 0; + p->params_uptodate = 0; } - if (new_state & (_NEW_FOG|_NEW_HINT|_NEW_PROGRAM)) + if (new_state & (_NEW_FOG | _NEW_HINT | _NEW_PROGRAM)) i915_update_fog(ctx); } -static void i915InitDriverFunctions( struct dd_function_table *functions ) +static void +i915InitDriverFunctions(struct dd_function_table *functions) { - intelInitDriverFunctions( functions ); - i915InitStateFunctions( functions ); - i915InitTextureFuncs( functions ); - i915InitFragProgFuncs( functions ); + intelInitDriverFunctions(functions); + i915InitStateFunctions(functions); + i915InitTextureFuncs(functions); + i915InitFragProgFuncs(functions); functions->UpdateState = i915InvalidateState; } -GLboolean i915CreateContext( const __GLcontextModes *mesaVis, - __DRIcontextPrivate *driContextPriv, - void *sharedContextPrivate) +GLboolean +i915CreateContext(const __GLcontextModes * mesaVis, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate) { struct dd_function_table functions; - i915ContextPtr i915 = (i915ContextPtr) CALLOC_STRUCT(i915_context); - intelContextPtr intel = &i915->intel; + struct i915_context *i915 = + (struct i915_context *) CALLOC_STRUCT(i915_context); + struct intel_context *intel = &i915->intel; GLcontext *ctx = &intel->ctx; - GLuint i; - if (!i915) return GL_FALSE; + if (!i915) + return GL_FALSE; + + if (0) + _mesa_printf("\ntexmem-0-3 branch\n\n"); - i915InitVtbl( i915 ); + i915InitVtbl(i915); + i915InitMetaFuncs(i915); - i915InitDriverFunctions( &functions ); + i915InitDriverFunctions(&functions); - if (!intelInitContext( intel, mesaVis, driContextPriv, - sharedContextPrivate, &functions )) { + if (!intelInitContext(intel, mesaVis, driContextPriv, + sharedContextPrivate, &functions)) { FREE(i915); return GL_FALSE; } @@ -122,65 +131,44 @@ GLboolean i915CreateContext( const __GLcontextModes *mesaVis, ctx->Const.MaxTextureImageUnits = I915_TEX_UNITS; ctx->Const.MaxTextureCoordUnits = I915_TEX_UNITS; - intel->nr_heaps = 1; - intel->texture_heaps[0] = - driCreateTextureHeap( 0, intel, - intel->intelScreen->tex.size, - 12, - I830_NR_TEX_REGIONS, - intel->sarea->texList, - (unsigned *) & intel->sarea->texAge, - & intel->swapped, - sizeof( struct i915_texture_object ), - (destroy_texture_object_t *)intelDestroyTexObj ); - - /* FIXME: driCalculateMaxTextureLevels assumes that mipmaps are - * tightly packed, but they're not in Intel graphics - * hardware. + + /* Advertise the full hardware capabilities. The new memory + * manager should cope much better with overload situations: */ + ctx->Const.MaxTextureLevels = 12; + ctx->Const.Max3DTextureLevels = 9; + ctx->Const.MaxCubeTextureLevels = 12; + ctx->Const.MaxTextureRectSize = (1 << 11); ctx->Const.MaxTextureUnits = I915_TEX_UNITS; - i = driQueryOptioni( &intel->optionCache, "allow_large_textures"); - driCalculateMaxTextureLevels( intel->texture_heaps, - intel->nr_heaps, - &intel->ctx.Const, - 4, - 11, /* max 2D texture size is 2048x2048 */ - 8, /* 3D texture */ - 11, /* cube texture. */ - 11, /* rect texture */ - 12, - GL_FALSE, - i ); /* GL_ARB_fragment_program limits - don't think Mesa actually * validates programs against these, and in any case one ARB * instruction can translate to more than one HW instruction, so * we'll still have to check and fallback each time. */ - ctx->Const.FragmentProgram.MaxNativeTemps = I915_MAX_TEMPORARY; - ctx->Const.FragmentProgram.MaxNativeAttribs = 11; /* 8 tex, 2 color, fog */ + ctx->Const.FragmentProgram.MaxNativeAttribs = 11; /* 8 tex, 2 color, fog */ ctx->Const.FragmentProgram.MaxNativeParameters = I915_MAX_CONSTANT; ctx->Const.FragmentProgram.MaxNativeAluInstructions = I915_MAX_ALU_INSN; ctx->Const.FragmentProgram.MaxNativeTexInstructions = I915_MAX_TEX_INSN; - ctx->Const.FragmentProgram.MaxNativeInstructions = (I915_MAX_ALU_INSN + - I915_MAX_TEX_INSN); - ctx->Const.FragmentProgram.MaxNativeTexIndirections = I915_MAX_TEX_INDIRECT; + ctx->Const.FragmentProgram.MaxNativeInstructions = (I915_MAX_ALU_INSN + + I915_MAX_TEX_INSN); + ctx->Const.FragmentProgram.MaxNativeTexIndirections = + I915_MAX_TEX_INDIRECT; ctx->Const.FragmentProgram.MaxNativeAddressRegs = 0; /* I don't think we have one */ + ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE; ctx->FragmentProgram._UseTexEnvProgram = GL_TRUE; + driInitExtensions(ctx, i915_extensions, GL_FALSE); - driInitExtensions( ctx, i915_extensions, GL_FALSE ); - - _tnl_init_vertices( ctx, ctx->Const.MaxArrayLockSize + 12, - 36 * sizeof(GLfloat) ); + _tnl_init_vertices(ctx, ctx->Const.MaxArrayLockSize + 12, + 36 * sizeof(GLfloat)); intel->verts = TNL_CONTEXT(ctx)->clipspace.vertex_buf; - i915InitState( i915 ); + i915InitState(i915); return GL_TRUE; } - diff --git a/src/mesa/drivers/dri/i915/i915_context.h b/src/mesa/drivers/dri/i915/i915_context.h index ec1550126a..1070de1a54 100644 --- a/src/mesa/drivers/dri/i915/i915_context.h +++ b/src/mesa/drivers/dri/i915/i915_context.h @@ -46,6 +46,7 @@ #define I915_UPLOAD_CONSTANTS 0x10 #define I915_UPLOAD_FOG 0x20 #define I915_UPLOAD_INVARIENT 0x40 +#define I915_UPLOAD_DEFAULTS 0x80 #define I915_UPLOAD_TEX(i) (0x00010000<<(i)) #define I915_UPLOAD_TEX_ALL (0x00ff0000) #define I915_UPLOAD_TEX_0_SHIFT 16 @@ -55,10 +56,8 @@ */ #define I915_DESTREG_CBUFADDR0 0 #define I915_DESTREG_CBUFADDR1 1 -#define I915_DESTREG_CBUFADDR2 2 #define I915_DESTREG_DBUFADDR0 3 #define I915_DESTREG_DBUFADDR1 4 -#define I915_DESTREG_DBUFADDR2 5 #define I915_DESTREG_DV0 6 #define I915_DESTREG_DV1 7 #define I915_DESTREG_SENABLE 8 @@ -89,7 +88,6 @@ #define I915_STPREG_ST1 1 #define I915_STP_SETUP_SIZE 2 -#define I915_TEXREG_MS2 0 #define I915_TEXREG_MS3 1 #define I915_TEXREG_MS4 2 #define I915_TEXREG_SS2 3 @@ -97,6 +95,15 @@ #define I915_TEXREG_SS4 5 #define I915_TEX_SETUP_SIZE 6 +#define I915_DEFREG_C0 0 +#define I915_DEFREG_C1 1 +#define I915_DEFREG_S0 2 +#define I915_DEFREG_S1 3 +#define I915_DEFREG_Z0 4 +#define I915_DEFREG_Z1 5 +#define I915_DEF_SETUP_SIZE 6 + + #define I915_MAX_CONSTANT 32 #define I915_CONSTANT_SIZE (2+(4*I915_MAX_CONSTANT)) @@ -107,13 +114,14 @@ /* Hardware version of a parsed fragment program. "Derived" from the * mesa fragment_program struct. */ -struct i915_fragment_program { +struct i915_fragment_program +{ struct gl_fragment_program FragProg; GLboolean translated; GLboolean params_uptodate; GLboolean on_hardware; - GLboolean error; /* If program is malformed for any reason. */ + GLboolean error; /* If program is malformed for any reason. */ GLuint nr_tex_indirect; GLuint nr_tex_insn; @@ -135,22 +143,22 @@ struct i915_fragment_program { GLuint constant_flags[I915_MAX_CONSTANT]; GLuint nr_constants; - GLuint *csr; /* Cursor, points into program. - */ + GLuint *csr; /* Cursor, points into program. + */ - GLuint *decl; /* Cursor, points into declarations. - */ - - GLuint decl_s; /* flags for which s regs need to be decl'd */ - GLuint decl_t; /* flags for which t regs need to be decl'd */ + GLuint *decl; /* Cursor, points into declarations. + */ - GLuint temp_flag; /* Tracks temporary regs which are in - * use. - */ + GLuint decl_s; /* flags for which s regs need to be decl'd */ + GLuint decl_t; /* flags for which t regs need to be decl'd */ - GLuint utemp_flag; /* Tracks TYPE_U temporary regs which are in - * use. - */ + GLuint temp_flag; /* Tracks temporary regs which are in + * use. + */ + + GLuint utemp_flag; /* Tracks TYPE_U temporary regs which are in + * use. + */ @@ -159,28 +167,12 @@ struct i915_fragment_program { GLuint wpos_tex; GLboolean depth_written; - struct { - GLuint reg; /* Hardware constant idx */ - const GLfloat *values; /* Pointer to tracked values */ + struct + { + GLuint reg; /* Hardware constant idx */ + const GLfloat *values; /* Pointer to tracked values */ } param[I915_MAX_CONSTANT]; GLuint nr_params; - - - - - /* Helpers for i915_texprog.c: - */ - GLuint src_texture; /* Reg containing sampled texture color, - * else UREG_BAD. - */ - - GLuint src_previous; /* Reg containing color from previous - * stage. May need to be decl'd. - */ - - GLuint last_tex_stage; /* Number of last enabled texture unit */ - - struct vertex_buffer *VB; }; @@ -188,67 +180,68 @@ struct i915_fragment_program { -struct i915_texture_object -{ - struct intel_texture_object intel; - GLenum lastTarget; - GLboolean refs_border_color; - GLuint Setup[I915_TEX_SETUP_SIZE]; -}; #define I915_TEX_UNITS 8 -struct i915_hw_state { +struct i915_hw_state +{ GLuint Ctx[I915_CTX_SETUP_SIZE]; GLuint Buffer[I915_DEST_SETUP_SIZE]; GLuint Stipple[I915_STP_SETUP_SIZE]; GLuint Fog[I915_FOG_SETUP_SIZE]; + GLuint Defaults[I915_DEF_SETUP_SIZE]; GLuint Tex[I915_TEX_UNITS][I915_TEX_SETUP_SIZE]; GLuint Constant[I915_CONSTANT_SIZE]; GLuint ConstantSize; GLuint Program[I915_PROGRAM_SIZE]; GLuint ProgramSize; - GLuint active; /* I915_UPLOAD_* */ - GLuint emitted; /* I915_UPLOAD_* */ + + /* Region pointers for relocation: + */ + struct intel_region *draw_region; + struct intel_region *depth_region; +/* struct intel_region *tex_region[I915_TEX_UNITS]; */ + + /* Regions aren't actually that appropriate here as the memory may + * be from a PBO or FBO. Will have to do this for draw and depth for + * FBO's... + */ + dri_bo *tex_buffer[I915_TEX_UNITS]; + GLuint tex_offset[I915_TEX_UNITS]; + + + GLuint active; /* I915_UPLOAD_* */ + GLuint emitted; /* I915_UPLOAD_* */ }; #define I915_FOG_PIXEL 2 #define I915_FOG_VERTEX 1 #define I915_FOG_NONE 0 -struct i915_context +struct i915_context { struct intel_context intel; GLuint last_ReallyEnabled; GLuint vertex_fog; + GLuint lodbias_ss2[MAX_TEXTURE_UNITS]; + - struct i915_fragment_program tex_program; struct i915_fragment_program *current_program; struct i915_hw_state meta, initial, state, *current; }; -typedef struct i915_context *i915ContextPtr; -typedef struct i915_texture_object *i915TextureObjectPtr; - -#define I915_CONTEXT(ctx) ((i915ContextPtr)(ctx)) - - - #define I915_STATECHANGE(i915, flag) \ do { \ - if (0) fprintf(stderr, "I915_STATECHANGE %x in %s\n", flag, __FUNCTION__); \ INTEL_FIREVERTICES( &(i915)->intel ); \ (i915)->state.emitted &= ~(flag); \ } while (0) #define I915_ACTIVESTATE(i915, flag, mode) \ do { \ - if (0) fprintf(stderr, "I915_ACTIVESTATE %x %d in %s\n", \ - flag, mode, __FUNCTION__); \ INTEL_FIREVERTICES( &(i915)->intel ); \ if (mode) \ (i915)->state.active |= (flag); \ @@ -260,7 +253,13 @@ do { \ /*====================================================================== * i915_vtbl.c */ -extern void i915InitVtbl( i915ContextPtr i915 ); +extern void i915InitVtbl(struct i915_context *i915); + +extern void +i915_state_draw_region(struct intel_context *intel, + struct i915_hw_state *state, + struct intel_region *color_region, + struct intel_region *depth_region); @@ -289,70 +288,58 @@ do { \ /*====================================================================== * i915_context.c */ -extern GLboolean i915CreateContext( const __GLcontextModes *mesaVis, - __DRIcontextPrivate *driContextPriv, - void *sharedContextPrivate); - - -/*====================================================================== - * i915_texprog.c - */ -extern void i915ValidateTextureProgram( i915ContextPtr i915 ); +extern GLboolean i915CreateContext(const __GLcontextModes * mesaVis, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate); /*====================================================================== * i915_debug.c */ -extern void i915_disassemble_program( const GLuint *program, GLuint sz ); -extern void i915_print_ureg( const char *msg, GLuint ureg ); +extern void i915_disassemble_program(const GLuint * program, GLuint sz); +extern void i915_print_ureg(const char *msg, GLuint ureg); /*====================================================================== * i915_state.c */ -extern void i915InitStateFunctions( struct dd_function_table *functions ); -extern void i915InitState( i915ContextPtr i915 ); -extern void i915_update_fog(GLcontext *ctxx); +extern void i915InitStateFunctions(struct dd_function_table *functions); +extern void i915InitState(struct i915_context *i915); +extern void i915_update_fog(GLcontext * ctx); /*====================================================================== * i915_tex.c */ -extern void i915UpdateTextureState( intelContextPtr intel ); -extern void i915InitTextureFuncs( struct dd_function_table *functions ); -extern intelTextureObjectPtr i915AllocTexObj( struct gl_texture_object *texObj ); +extern void i915UpdateTextureState(struct intel_context *intel); +extern void i915InitTextureFuncs(struct dd_function_table *functions); /*====================================================================== * i915_metaops.c */ -extern GLboolean -i915TryTextureReadPixels( GLcontext *ctx, - GLint x, GLint y, GLsizei width, GLsizei height, - GLenum format, GLenum type, - const struct gl_pixelstore_attrib *pack, - GLvoid *pixels ); - -extern GLboolean -i915TryTextureDrawPixels( GLcontext *ctx, - GLint x, GLint y, GLsizei width, GLsizei height, - GLenum format, GLenum type, - const struct gl_pixelstore_attrib *unpack, - const GLvoid *pixels ); +void i915InitMetaFuncs(struct i915_context *i915); -extern void -i915ClearWithTris( intelContextPtr intel, GLbitfield mask, - GLboolean all, GLint cx, GLint cy, GLint cw, GLint ch); - - -extern void -i915RotateWindow(intelContextPtr intel, __DRIdrawablePrivate *dPriv, - GLuint srcBuf); /*====================================================================== * i915_fragprog.c */ -extern void i915ValidateFragmentProgram( i915ContextPtr i915 ); -extern void i915InitFragProgFuncs( struct dd_function_table *functions ); - -#endif +extern void i915ValidateFragmentProgram(struct i915_context *i915); +extern void i915InitFragProgFuncs(struct dd_function_table *functions); + +/*====================================================================== + * Inline conversion functions. These are better-typed than the + * macros used previously: + */ +static INLINE struct i915_context * +i915_context(GLcontext * ctx) +{ + return (struct i915_context *) ctx; +} + + +#define I915_CONTEXT(ctx) i915_context(ctx) + + + +#endif diff --git a/src/mesa/drivers/dri/i915/i915_debug.c b/src/mesa/drivers/dri/i915/i915_debug.c index 054b561605..8eb1c5b49e 100644 --- a/src/mesa/drivers/dri/i915/i915_debug.c +++ b/src/mesa/drivers/dri/i915/i915_debug.c @@ -25,275 +25,824 @@ * **************************************************************************/ +#include "imports.h" + #include "i915_reg.h" #include "i915_context.h" -#include <stdio.h> - - -static const char *opcodes[0x20] = { - "NOP", - "ADD", - "MOV", - "MUL", - "MAD", - "DP2ADD", - "DP3", - "DP4", - "FRC", - "RCP", - "RSQ", - "EXP", - "LOG", - "CMP", - "MIN", - "MAX", - "FLR", - "MOD", - "TRC", - "SGE", - "SLT", - "TEXLD", - "TEXLDP", - "TEXLDB", - "TEXKILL", - "DCL", - "0x1a", - "0x1b", - "0x1c", - "0x1d", - "0x1e", - "0x1f", -}; - - -static const int args[0x20] = { - 0, /* 0 nop */ - 2, /* 1 add */ - 1, /* 2 mov */ - 2, /* 3 m ul */ - 3, /* 4 mad */ - 3, /* 5 dp2add */ - 2, /* 6 dp3 */ - 2, /* 7 dp4 */ - 1, /* 8 frc */ - 1, /* 9 rcp */ - 1, /* a rsq */ - 1, /* b exp */ - 1, /* c log */ - 3, /* d cmp */ - 2, /* e min */ - 2, /* f max */ - 1, /* 10 flr */ - 1, /* 11 mod */ - 1, /* 12 trc */ - 2, /* 13 sge */ - 2, /* 14 slt */ - 1, - 1, - 1, - 1, - 0, - 0, - 0, - 0, - 0, - 0, - 0, -}; - - -static const char *regname[0x8] = { - "R", - "T", - "CONST", - "S", - "OC", - "OD", - "U", - "UNKNOWN", -}; - -static void print_reg_type_nr( GLuint type, GLuint nr ) +#include "i915_debug.h" + +#define PRINTF( ... ) _mesa_printf( __VA_ARGS__ ) + +static GLboolean debug( struct debug_stream *stream, const char *name, GLuint len ) { - switch (type) { - case REG_TYPE_T: - switch (nr) { - case T_DIFFUSE: fprintf(stderr, "T_DIFFUSE"); return; - case T_SPECULAR: fprintf(stderr, "T_SPECULAR"); return; - case T_FOG_W: fprintf(stderr, "T_FOG_W"); return; - default: fprintf(stderr, "T_TEX%d", nr); return; - } - case REG_TYPE_OC: - if (nr == 0) { - fprintf(stderr, "oC"); - return; - } - break; - case REG_TYPE_OD: - if (nr == 0) { - fprintf(stderr, "oD"); - return; - } - break; - default: - break; + GLuint i; + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + + if (len == 0) { + PRINTF("Error - zero length packet (0x%08x)\n", stream->ptr[0]); + assert(0); + return GL_FALSE; } - fprintf(stderr, "%s[%d]", regname[type], nr); -} + if (stream->print_addresses) + PRINTF("%08x: ", stream->offset); + -#define REG_SWIZZLE_MASK 0x7777 -#define REG_NEGATE_MASK 0x8888 + PRINTF("%s (%d dwords):\n", name, len); + for (i = 0; i < len; i++) + PRINTF("\t0x%08x\n", ptr[i]); + PRINTF("\n"); -#define REG_SWIZZLE_XYZW ((SRC_X << A2_SRC2_CHANNEL_X_SHIFT) | \ - (SRC_Y << A2_SRC2_CHANNEL_Y_SHIFT) | \ - (SRC_Z << A2_SRC2_CHANNEL_Z_SHIFT) | \ - (SRC_W << A2_SRC2_CHANNEL_W_SHIFT)) + stream->offset += len * sizeof(GLuint); + + return GL_TRUE; +} -static void print_reg_neg_swizzle( GLuint reg ) +static const char *get_prim_name( GLuint val ) { - int i; - - if ((reg & REG_SWIZZLE_MASK) == REG_SWIZZLE_XYZW && - (reg & REG_NEGATE_MASK) == 0) - return; - - fprintf(stderr, "."); - - for (i = 3 ; i >= 0; i--) { - if (reg & (1<<((i*4)+3))) - fprintf(stderr, "-"); - - switch ((reg>>(i*4)) & 0x7) { - case 0: fprintf(stderr, "x"); break; - case 1: fprintf(stderr, "y"); break; - case 2: fprintf(stderr, "z"); break; - case 3: fprintf(stderr, "w"); break; - case 4: fprintf(stderr, "0"); break; - case 5: fprintf(stderr, "1"); break; - default: fprintf(stderr, "?"); break; - } + switch (val & PRIM3D_MASK) { + case PRIM3D_TRILIST: return "TRILIST"; break; + case PRIM3D_TRISTRIP: return "TRISTRIP"; break; + case PRIM3D_TRISTRIP_RVRSE: return "TRISTRIP_RVRSE"; break; + case PRIM3D_TRIFAN: return "TRIFAN"; break; + case PRIM3D_POLY: return "POLY"; break; + case PRIM3D_LINELIST: return "LINELIST"; break; + case PRIM3D_LINESTRIP: return "LINESTRIP"; break; + case PRIM3D_RECTLIST: return "RECTLIST"; break; + case PRIM3D_POINTLIST: return "POINTLIST"; break; + case PRIM3D_DIB: return "DIB"; break; + case PRIM3D_CLEAR_RECT: return "CLEAR_RECT"; break; + case PRIM3D_ZONE_INIT: return "ZONE_INIT"; break; + default: return "????"; break; } } - -static void print_src_reg( GLuint dword ) +static GLboolean debug_prim( struct debug_stream *stream, const char *name, + GLboolean dump_floats, + GLuint len ) { - GLuint nr = (dword >> A2_SRC2_NR_SHIFT) & REG_NR_MASK; - GLuint type = (dword >> A2_SRC2_TYPE_SHIFT) & REG_TYPE_MASK; - print_reg_type_nr( type, nr ); - print_reg_neg_swizzle( dword ); + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + const char *prim = get_prim_name( ptr[0] ); + GLuint i; + + + + PRINTF("%s %s (%d dwords):\n", name, prim, len); + PRINTF("\t0x%08x\n", ptr[0]); + for (i = 1; i < len; i++) { + if (dump_floats) + PRINTF("\t0x%08x // %f\n", ptr[i], *(GLfloat *)&ptr[i]); + else + PRINTF("\t0x%08x\n", ptr[i]); + } + + + PRINTF("\n"); + + stream->offset += len * sizeof(GLuint); + + return GL_TRUE; } + + -void i915_print_ureg( const char *msg, GLuint ureg ) + +static GLboolean debug_program( struct debug_stream *stream, const char *name, GLuint len ) { - fprintf(stderr, "%s: ", msg); - print_src_reg( ureg >> 8 ); - fprintf(stderr, "\n"); + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + + if (len == 0) { + PRINTF("Error - zero length packet (0x%08x)\n", stream->ptr[0]); + assert(0); + return GL_FALSE; + } + + if (stream->print_addresses) + PRINTF("%08x: ", stream->offset); + + PRINTF("%s (%d dwords):\n", name, len); + i915_disassemble_program( ptr, len ); + + stream->offset += len * sizeof(GLuint); + return GL_TRUE; } -static void print_dest_reg( GLuint dword ) + +static GLboolean debug_chain( struct debug_stream *stream, const char *name, GLuint len ) { - GLuint nr = (dword >> A0_DEST_NR_SHIFT) & REG_NR_MASK; - GLuint type = (dword >> A0_DEST_TYPE_SHIFT) & REG_TYPE_MASK; - print_reg_type_nr( type, nr ); - if ((dword & A0_DEST_CHANNEL_ALL) == A0_DEST_CHANNEL_ALL) - return; - fprintf(stderr, "."); - if (dword & A0_DEST_CHANNEL_X) fprintf(stderr, "x"); - if (dword & A0_DEST_CHANNEL_Y) fprintf(stderr, "y"); - if (dword & A0_DEST_CHANNEL_Z) fprintf(stderr, "z"); - if (dword & A0_DEST_CHANNEL_W) fprintf(stderr, "w"); + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + GLuint old_offset = stream->offset + len * sizeof(GLuint); + GLuint i; + + PRINTF("%s (%d dwords):\n", name, len); + for (i = 0; i < len; i++) + PRINTF("\t0x%08x\n", ptr[i]); + + stream->offset = ptr[1] & ~0x3; + + if (stream->offset < old_offset) + PRINTF("\n... skipping backwards from 0x%x --> 0x%x ...\n\n", + old_offset, stream->offset ); + else + PRINTF("\n... skipping from 0x%x --> 0x%x ...\n\n", + old_offset, stream->offset ); + + + return GL_TRUE; } -#define GET_SRC0_REG(r0, r1) ((r0<<14)|(r1>>A1_SRC0_CHANNEL_W_SHIFT)) -#define GET_SRC1_REG(r0, r1) ((r0<<8)|(r1>>A2_SRC1_CHANNEL_W_SHIFT)) -#define GET_SRC2_REG(r) (r) +static GLboolean debug_variable_length_prim( struct debug_stream *stream ) +{ + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + const char *prim = get_prim_name( ptr[0] ); + GLuint i, len; + + GLushort *idx = (GLushort *)(ptr+1); + for (i = 0; idx[i] != 0xffff; i++) + ; + + len = 1+(i+2)/2; + + PRINTF("3DPRIM, %s variable length %d indicies (%d dwords):\n", prim, i, len); + for (i = 0; i < len; i++) + PRINTF("\t0x%08x\n", ptr[i]); + PRINTF("\n"); + + stream->offset += len * sizeof(GLuint); + return GL_TRUE; +} -static void print_arith_op( GLuint opcode, const GLuint *program ) +#define BITS( dw, hi, lo, ... ) \ +do { \ + unsigned himask = ~0UL >> (31 - (hi)); \ + PRINTF("\t\t "); \ + PRINTF(__VA_ARGS__); \ + PRINTF(": 0x%x\n", ((dw) & himask) >> (lo)); \ +} while (0) + +#define MBZ( dw, hi, lo) do { \ + unsigned x = (dw) >> (lo); \ + unsigned lomask = (1 << (lo)) - 1; \ + unsigned himask; \ + himask = (1UL << (hi)) - 1; \ + assert ((x & himask & ~lomask) == 0); \ +} while (0) + +#define FLAG( dw, bit, ... ) \ +do { \ + if (((dw) >> (bit)) & 1) { \ + PRINTF("\t\t "); \ + PRINTF(__VA_ARGS__); \ + PRINTF("\n"); \ + } \ +} while (0) + +static GLboolean debug_load_immediate( struct debug_stream *stream, + const char *name, + GLuint len ) { - if (opcode != A0_NOP) { - print_dest_reg(program[0]); - if (program[0] & A0_DEST_SATURATE) - fprintf(stderr, " = SATURATE "); - else - fprintf(stderr, " = "); + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + GLuint bits = (ptr[0] >> 4) & 0xff; + GLuint j = 0; + + PRINTF("%s (%d dwords, flags: %x):\n", name, len, bits); + PRINTF("\t0x%08x\n", ptr[j++]); + + if (bits & (1<<0)) { + PRINTF("\t LIS0: 0x%08x\n", ptr[j]); + PRINTF("\t vb address: 0x%08x\n", (ptr[j] & ~0x3)); + BITS(ptr[j], 0, 0, "vb invalidate disable"); + j++; + } + if (bits & (1<<1)) { + PRINTF("\t LIS1: 0x%08x\n", ptr[j]); + BITS(ptr[j], 29, 24, "vb dword width"); + BITS(ptr[j], 21, 16, "vb dword pitch"); + BITS(ptr[j], 15, 0, "vb max index"); + j++; } + if (bits & (1<<2)) { + int i; + PRINTF("\t LIS2: 0x%08x\n", ptr[j]); + for (i = 0; i < 8; i++) { + unsigned tc = (ptr[j] >> (i * 4)) & 0xf; + if (tc != 0xf) + BITS(tc, 3, 0, "tex coord %d", i); + } + j++; + } + if (bits & (1<<3)) { + PRINTF("\t LIS3: 0x%08x\n", ptr[j]); + j++; + } + if (bits & (1<<4)) { + PRINTF("\t LIS4: 0x%08x\n", ptr[j]); + BITS(ptr[j], 31, 23, "point width"); + BITS(ptr[j], 22, 19, "line width"); + FLAG(ptr[j], 18, "alpha flatshade"); + FLAG(ptr[j], 17, "fog flatshade"); + FLAG(ptr[j], 16, "spec flatshade"); + FLAG(ptr[j], 15, "rgb flatshade"); + BITS(ptr[j], 14, 13, "cull mode"); + FLAG(ptr[j], 12, "vfmt: point width"); + FLAG(ptr[j], 11, "vfmt: specular/fog"); + FLAG(ptr[j], 10, "vfmt: rgba"); + FLAG(ptr[j], 9, "vfmt: depth offset"); + BITS(ptr[j], 8, 6, "vfmt: position (2==xyzw)"); + FLAG(ptr[j], 5, "force dflt diffuse"); + FLAG(ptr[j], 4, "force dflt specular"); + FLAG(ptr[j], 3, "local depth offset enable"); + FLAG(ptr[j], 2, "vfmt: fp32 fog coord"); + FLAG(ptr[j], 1, "sprite point"); + FLAG(ptr[j], 0, "antialiasing"); + j++; + } + if (bits & (1<<5)) { + PRINTF("\t LIS5: 0x%08x\n", ptr[j]); + BITS(ptr[j], 31, 28, "rgba write disables"); + FLAG(ptr[j], 27, "force dflt point width"); + FLAG(ptr[j], 26, "last pixel enable"); + FLAG(ptr[j], 25, "global z offset enable"); + FLAG(ptr[j], 24, "fog enable"); + BITS(ptr[j], 23, 16, "stencil ref"); + BITS(ptr[j], 15, 13, "stencil test"); + BITS(ptr[j], 12, 10, "stencil fail op"); + BITS(ptr[j], 9, 7, "stencil pass z fail op"); + BITS(ptr[j], 6, 4, "stencil pass z pass op"); + FLAG(ptr[j], 3, "stencil write enable"); + FLAG(ptr[j], 2, "stencil test enable"); + FLAG(ptr[j], 1, "color dither enable"); + FLAG(ptr[j], 0, "logiop enable"); + j++; + } + if (bits & (1<<6)) { + PRINTF("\t LIS6: 0x%08x\n", ptr[j]); + FLAG(ptr[j], 31, "alpha test enable"); + BITS(ptr[j], 30, 28, "alpha func"); + BITS(ptr[j], 27, 20, "alpha ref"); + FLAG(ptr[j], 19, "depth test enable"); + BITS(ptr[j], 18, 16, "depth func"); + FLAG(ptr[j], 15, "blend enable"); + BITS(ptr[j], 14, 12, "blend func"); + BITS(ptr[j], 11, 8, "blend src factor"); + BITS(ptr[j], 7, 4, "blend dst factor"); + FLAG(ptr[j], 3, "depth write enable"); + FLAG(ptr[j], 2, "color write enable"); + BITS(ptr[j], 1, 0, "provoking vertex"); + j++; + } + - fprintf(stderr, "%s ", opcodes[opcode]); + PRINTF("\n"); - print_src_reg(GET_SRC0_REG(program[0], program[1])); - if (args[opcode] == 1) { - fprintf(stderr, "\n"); - return; + assert(j == len); + + stream->offset += len * sizeof(GLuint); + + return GL_TRUE; +} + + + +static GLboolean debug_load_indirect( struct debug_stream *stream, + const char *name, + GLuint len ) +{ + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + GLuint bits = (ptr[0] >> 8) & 0x3f; + GLuint i, j = 0; + + PRINTF("%s (%d dwords):\n", name, len); + PRINTF("\t0x%08x\n", ptr[j++]); + + for (i = 0; i < 6; i++) { + if (bits & (1<<i)) { + switch (1<<(8+i)) { + case LI0_STATE_STATIC_INDIRECT: + PRINTF(" STATIC: 0x%08x | %x\n", ptr[j]&~3, ptr[j]&3); j++; + PRINTF(" 0x%08x\n", ptr[j++]); + break; + case LI0_STATE_DYNAMIC_INDIRECT: + PRINTF(" DYNAMIC: 0x%08x | %x\n", ptr[j]&~3, ptr[j]&3); j++; + break; + case LI0_STATE_SAMPLER: + PRINTF(" SAMPLER: 0x%08x | %x\n", ptr[j]&~3, ptr[j]&3); j++; + PRINTF(" 0x%08x\n", ptr[j++]); + break; + case LI0_STATE_MAP: + PRINTF(" MAP: 0x%08x | %x\n", ptr[j]&~3, ptr[j]&3); j++; + PRINTF(" 0x%08x\n", ptr[j++]); + break; + case LI0_STATE_PROGRAM: + PRINTF(" PROGRAM: 0x%08x | %x\n", ptr[j]&~3, ptr[j]&3); j++; + PRINTF(" 0x%08x\n", ptr[j++]); + break; + case LI0_STATE_CONSTANTS: + PRINTF(" CONSTANTS: 0x%08x | %x\n", ptr[j]&~3, ptr[j]&3); j++; + PRINTF(" 0x%08x\n", ptr[j++]); + break; + default: + assert(0); + break; + } + } } - fprintf(stderr, ", "); - print_src_reg(GET_SRC1_REG(program[1], program[2])); - if (args[opcode] == 2) { - fprintf(stderr, "\n"); - return; + if (bits == 0) { + PRINTF("\t DUMMY: 0x%08x\n", ptr[j++]); } - fprintf(stderr, ", "); - print_src_reg(GET_SRC2_REG(program[2])); - fprintf(stderr, "\n"); - return; + PRINTF("\n"); + + + assert(j == len); + + stream->offset += len * sizeof(GLuint); + + return GL_TRUE; +} + +static void BR13( struct debug_stream *stream, + GLuint val ) +{ + PRINTF("\t0x%08x\n", val); + FLAG(val, 30, "clipping enable"); + BITS(val, 25, 24, "color depth (3==32bpp)"); + BITS(val, 23, 16, "raster op"); + BITS(val, 15, 0, "dest pitch"); } -static void print_tex_op( GLuint opcode, const GLuint *program ) +static void BR2223( struct debug_stream *stream, + GLuint val22, GLuint val23 ) { - print_dest_reg(program[0] | A0_DEST_CHANNEL_ALL); - fprintf(stderr, " = "); + union { GLuint val; short field[2]; } BR22, BR23; - fprintf(stderr, "%s ", opcodes[opcode]); + BR22.val = val22; + BR23.val = val23; - fprintf(stderr, "S[%d],", - program[0] & T0_SAMPLER_NR_MASK); + PRINTF("\t0x%08x\n", val22); + BITS(val22, 31, 16, "dest y1"); + BITS(val22, 15, 0, "dest x1"); - print_reg_type_nr( (program[1]>>T1_ADDRESS_REG_TYPE_SHIFT) & REG_TYPE_MASK, - (program[1]>>T1_ADDRESS_REG_NR_SHIFT) & REG_NR_MASK ); - fprintf(stderr, "\n"); + PRINTF("\t0x%08x\n", val23); + BITS(val23, 31, 16, "dest y2"); + BITS(val23, 15, 0, "dest x2"); + + /* The blit engine may produce unexpected results when these aren't met */ + assert(BR22.field[0] < BR23.field[0]); + assert(BR22.field[1] < BR23.field[1]); } -static void print_dcl_op( GLuint opcode, const GLuint *program ) +static void BR09( struct debug_stream *stream, + GLuint val ) { - fprintf(stderr, "%s ", opcodes[opcode]); - print_dest_reg(program[0] | A0_DEST_CHANNEL_ALL); - fprintf(stderr, "\n"); + PRINTF("\t0x%08x -- dest address\n", val); } +static void BR26( struct debug_stream *stream, + GLuint val ) +{ + PRINTF("\t0x%08x\n", val); + BITS(val, 31, 16, "src y1"); + BITS(val, 15, 0, "src x1"); +} + +static void BR11( struct debug_stream *stream, + GLuint val ) +{ + PRINTF("\t0x%08x\n", val); + BITS(val, 15, 0, "src pitch"); +} -void i915_disassemble_program( const GLuint *program, GLuint sz ) +static void BR12( struct debug_stream *stream, + GLuint val ) { - GLuint size = program[0] & 0x1ff; - GLint i; + PRINTF("\t0x%08x -- src address\n", val); +} + +static void BR16( struct debug_stream *stream, + GLuint val ) +{ + PRINTF("\t0x%08x -- color\n", val); +} + +static GLboolean debug_copy_blit( struct debug_stream *stream, + const char *name, + GLuint len ) +{ + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + int j = 0; + + PRINTF("%s (%d dwords):\n", name, len); + PRINTF("\t0x%08x\n", ptr[j++]); + + BR13(stream, ptr[j++]); + BR2223(stream, ptr[j], ptr[j+1]); + j += 2; + BR09(stream, ptr[j++]); + BR26(stream, ptr[j++]); + BR11(stream, ptr[j++]); + BR12(stream, ptr[j++]); + + stream->offset += len * sizeof(GLuint); + assert(j == len); + return GL_TRUE; +} + +static GLboolean debug_color_blit( struct debug_stream *stream, + const char *name, + GLuint len ) +{ + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + int j = 0; + + PRINTF("%s (%d dwords):\n", name, len); + PRINTF("\t0x%08x\n", ptr[j++]); + + BR13(stream, ptr[j++]); + BR2223(stream, ptr[j], ptr[j+1]); + j += 2; + BR09(stream, ptr[j++]); + BR16(stream, ptr[j++]); + + stream->offset += len * sizeof(GLuint); + assert(j == len); + return GL_TRUE; +} + +static GLboolean debug_modes4( struct debug_stream *stream, + const char *name, + GLuint len ) +{ + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + int j = 0; + + PRINTF("%s (%d dwords):\n", name, len); + PRINTF("\t0x%08x\n", ptr[j]); + BITS(ptr[j], 21, 18, "logicop func"); + FLAG(ptr[j], 17, "stencil test mask modify-enable"); + FLAG(ptr[j], 16, "stencil write mask modify-enable"); + BITS(ptr[j], 15, 8, "stencil test mask"); + BITS(ptr[j], 7, 0, "stencil write mask"); + j++; + + stream->offset += len * sizeof(GLuint); + assert(j == len); + return GL_TRUE; +} + +static GLboolean debug_map_state( struct debug_stream *stream, + const char *name, + GLuint len ) +{ + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + int j = 0; + + PRINTF("%s (%d dwords):\n", name, len); + PRINTF("\t0x%08x\n", ptr[j++]); + + { + PRINTF("\t0x%08x\n", ptr[j]); + BITS(ptr[j], 15, 0, "map mask"); + j++; + } + + while (j < len) { + { + PRINTF("\t TMn.0: 0x%08x\n", ptr[j]); + PRINTF("\t map address: 0x%08x\n", (ptr[j] & ~0x3)); + FLAG(ptr[j], 1, "vertical line stride"); + FLAG(ptr[j], 0, "vertical line stride offset"); + j++; + } + + { + PRINTF("\t TMn.1: 0x%08x\n", ptr[j]); + BITS(ptr[j], 31, 21, "height"); + BITS(ptr[j], 20, 10, "width"); + BITS(ptr[j], 9, 7, "surface format"); + BITS(ptr[j], 6, 3, "texel format"); + FLAG(ptr[j], 2, "use fence regs"); + FLAG(ptr[j], 1, "tiled surface"); + FLAG(ptr[j], 0, "tile walk ymajor"); + j++; + } + { + PRINTF("\t TMn.2: 0x%08x\n", ptr[j]); + BITS(ptr[j], 31, 21, "dword pitch"); + BITS(ptr[j], 20, 15, "cube face enables"); + BITS(ptr[j], 14, 9, "max lod"); + FLAG(ptr[j], 8, "mip layout right"); + BITS(ptr[j], 7, 0, "depth"); + j++; + } + } + + stream->offset += len * sizeof(GLuint); + assert(j == len); + return GL_TRUE; +} + +static GLboolean debug_sampler_state( struct debug_stream *stream, + const char *name, + GLuint len ) +{ + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + int j = 0; + + PRINTF("%s (%d dwords):\n", name, len); + PRINTF("\t0x%08x\n", ptr[j++]); + + { + PRINTF("\t0x%08x\n", ptr[j]); + BITS(ptr[j], 15, 0, "sampler mask"); + j++; + } + + while (j < len) { + { + PRINTF("\t TSn.0: 0x%08x\n", ptr[j]); + FLAG(ptr[j], 31, "reverse gamma"); + FLAG(ptr[j], 30, "planar to packed"); + FLAG(ptr[j], 29, "yuv->rgb"); + BITS(ptr[j], 28, 27, "chromakey index"); + BITS(ptr[j], 26, 22, "base mip level"); + BITS(ptr[j], 21, 20, "mip mode filter"); + BITS(ptr[j], 19, 17, "mag mode filter"); + BITS(ptr[j], 16, 14, "min mode filter"); + BITS(ptr[j], 13, 5, "lod bias (s4.4)"); + FLAG(ptr[j], 4, "shadow enable"); + FLAG(ptr[j], 3, "max-aniso-4"); + BITS(ptr[j], 2, 0, "shadow func"); + j++; + } + + { + PRINTF("\t TSn.1: 0x%08x\n", ptr[j]); + BITS(ptr[j], 31, 24, "min lod"); + MBZ( ptr[j], 23, 18 ); + FLAG(ptr[j], 17, "kill pixel enable"); + FLAG(ptr[j], 16, "keyed tex filter mode"); + FLAG(ptr[j], 15, "chromakey enable"); + BITS(ptr[j], 14, 12, "tcx wrap mode"); + BITS(ptr[j], 11, 9, "tcy wrap mode"); + BITS(ptr[j], 8, 6, "tcz wrap mode"); + FLAG(ptr[j], 5, "normalized coords"); + BITS(ptr[j], 4, 1, "map (surface) index"); + FLAG(ptr[j], 0, "EAST deinterlacer enable"); + j++; + } + { + PRINTF("\t TSn.2: 0x%08x (default color)\n", ptr[j]); + j++; + } + } + + stream->offset += len * sizeof(GLuint); + assert(j == len); + return GL_TRUE; +} + +static GLboolean debug_dest_vars( struct debug_stream *stream, + const char *name, + GLuint len ) +{ + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + int j = 0; + + PRINTF("%s (%d dwords):\n", name, len); + PRINTF("\t0x%08x\n", ptr[j++]); + + { + PRINTF("\t0x%08x\n", ptr[j]); + FLAG(ptr[j], 31, "early classic ztest"); + FLAG(ptr[j], 30, "opengl tex default color"); + FLAG(ptr[j], 29, "bypass iz"); + FLAG(ptr[j], 28, "lod preclamp"); + BITS(ptr[j], 27, 26, "dither pattern"); + FLAG(ptr[j], 25, "linear gamma blend"); + FLAG(ptr[j], 24, "debug dither"); + BITS(ptr[j], 23, 20, "dstorg x"); + BITS(ptr[j], 19, 16, "dstorg y"); + MBZ (ptr[j], 15, 15 ); + BITS(ptr[j], 14, 12, "422 write select"); + BITS(ptr[j], 11, 8, "cbuf format"); + BITS(ptr[j], 3, 2, "zbuf format"); + FLAG(ptr[j], 1, "vert line stride"); + FLAG(ptr[j], 1, "vert line stride offset"); + j++; + } - fprintf(stderr, "BEGIN\n"); + stream->offset += len * sizeof(GLuint); + assert(j == len); + return GL_TRUE; +} - if (size+2 != sz) { - fprintf(stderr, "%s: program size mismatch %d/%d\n", __FUNCTION__, - size+2, sz); - exit(1); +static GLboolean debug_buf_info( struct debug_stream *stream, + const char *name, + GLuint len ) +{ + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + int j = 0; + + PRINTF("%s (%d dwords):\n", name, len); + PRINTF("\t0x%08x\n", ptr[j++]); + + { + PRINTF("\t0x%08x\n", ptr[j]); + BITS(ptr[j], 28, 28, "aux buffer id"); + BITS(ptr[j], 27, 24, "buffer id (7=depth, 3=back)"); + FLAG(ptr[j], 23, "use fence regs"); + FLAG(ptr[j], 22, "tiled surface"); + FLAG(ptr[j], 21, "tile walk ymajor"); + MBZ (ptr[j], 20, 14); + BITS(ptr[j], 13, 2, "dword pitch"); + MBZ (ptr[j], 2, 0); + j++; } + + PRINTF("\t0x%08x -- buffer base address\n", ptr[j++]); + + stream->offset += len * sizeof(GLuint); + assert(j == len); + return GL_TRUE; +} - program ++; - for (i = 1 ; i < sz ; i+=3, program+=3) { - GLuint opcode = program[0] & (0x1f<<24); - - if ((GLint) opcode >= A0_NOP && opcode <= A0_SLT) - print_arith_op(opcode >> 24, program); - else if (opcode >= T0_TEXLD && opcode <= T0_TEXKILL) - print_tex_op(opcode >> 24, program); - else if (opcode == D0_DCL) - print_dcl_op(opcode >> 24, program); - else - fprintf(stderr, "Unknown opcode 0x%x\n", opcode); +static GLboolean i915_debug_packet( struct debug_stream *stream ) +{ + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + GLuint cmd = *ptr; + + switch (((cmd >> 29) & 0x7)) { + case 0x0: + switch ((cmd >> 23) & 0x3f) { + case 0x0: + return debug(stream, "MI_NOOP", 1); + case 0x3: + return debug(stream, "MI_WAIT_FOR_EVENT", 1); + case 0x4: + return debug(stream, "MI_FLUSH", 1); + case 0xA: + debug(stream, "MI_BATCH_BUFFER_END", 1); + return GL_FALSE; + case 0x22: + return debug(stream, "MI_LOAD_REGISTER_IMM", 3); + case 0x31: + return debug_chain(stream, "MI_BATCH_BUFFER_START", 2); + default: + break; + } + break; + case 0x1: + break; + case 0x2: + switch ((cmd >> 22) & 0xff) { + case 0x50: + return debug_color_blit(stream, "XY_COLOR_BLT", (cmd & 0xff) + 2); + case 0x53: + return debug_copy_blit(stream, "XY_SRC_COPY_BLT", (cmd & 0xff) + 2); + default: + return debug(stream, "blit command", (cmd & 0xff) + 2); + } + break; + case 0x3: + switch ((cmd >> 24) & 0x1f) { + case 0x6: + return debug(stream, "3DSTATE_ANTI_ALIASING", 1); + case 0x7: + return debug(stream, "3DSTATE_RASTERIZATION_RULES", 1); + case 0x8: + return debug(stream, "3DSTATE_BACKFACE_STENCIL_OPS", 2); + case 0x9: + return debug(stream, "3DSTATE_BACKFACE_STENCIL_MASKS", 1); + case 0xb: + return debug(stream, "3DSTATE_INDEPENDENT_ALPHA_BLEND", 1); + case 0xc: + return debug(stream, "3DSTATE_MODES5", 1); + case 0xd: + return debug_modes4(stream, "3DSTATE_MODES4", 1); + case 0x15: + return debug(stream, "3DSTATE_FOG_COLOR", 1); + case 0x16: + return debug(stream, "3DSTATE_COORD_SET_BINDINGS", 1); + case 0x1c: + /* 3DState16NP */ + switch((cmd >> 19) & 0x1f) { + case 0x10: + return debug(stream, "3DSTATE_SCISSOR_ENABLE", 1); + case 0x11: + return debug(stream, "3DSTATE_DEPTH_SUBRECTANGLE_DISABLE", 1); + default: + break; + } + break; + case 0x1d: + /* 3DStateMW */ + switch ((cmd >> 16) & 0xff) { + case 0x0: + return debug_map_state(stream, "3DSTATE_MAP_STATE", (cmd & 0x1f) + 2); + case 0x1: + return debug_sampler_state(stream, "3DSTATE_SAMPLER_STATE", (cmd & 0x1f) + 2); + case 0x4: + return debug_load_immediate(stream, "3DSTATE_LOAD_STATE_IMMEDIATE", (cmd & 0xf) + 2); + case 0x5: + return debug_program(stream, "3DSTATE_PIXEL_SHADER_PROGRAM", (cmd & 0x1ff) + 2); + case 0x6: + return debug(stream, "3DSTATE_PIXEL_SHADER_CONSTANTS", (cmd & 0xff) + 2); + case 0x7: + return debug_load_indirect(stream, "3DSTATE_LOAD_INDIRECT", (cmd & 0xff) + 2); + case 0x80: + return debug(stream, "3DSTATE_DRAWING_RECTANGLE", (cmd & 0xffff) + 2); + case 0x81: + return debug(stream, "3DSTATE_SCISSOR_RECTANGLE", (cmd & 0xffff) + 2); + case 0x83: + return debug(stream, "3DSTATE_SPAN_STIPPLE", (cmd & 0xffff) + 2); + case 0x85: + return debug_dest_vars(stream, "3DSTATE_DEST_BUFFER_VARS", (cmd & 0xffff) + 2); + case 0x88: + return debug(stream, "3DSTATE_CONSTANT_BLEND_COLOR", (cmd & 0xffff) + 2); + case 0x89: + return debug(stream, "3DSTATE_FOG_MODE", (cmd & 0xffff) + 2); + case 0x8e: + return debug_buf_info(stream, "3DSTATE_BUFFER_INFO", (cmd & 0xffff) + 2); + case 0x97: + return debug(stream, "3DSTATE_DEPTH_OFFSET_SCALE", (cmd & 0xffff) + 2); + case 0x98: + return debug(stream, "3DSTATE_DEFAULT_Z", (cmd & 0xffff) + 2); + case 0x99: + return debug(stream, "3DSTATE_DEFAULT_DIFFUSE", (cmd & 0xffff) + 2); + case 0x9a: + return debug(stream, "3DSTATE_DEFAULT_SPECULAR", (cmd & 0xffff) + 2); + case 0x9c: + return debug(stream, "3DSTATE_CLEAR_PARAMETERS", (cmd & 0xffff) + 2); + default: + assert(0); + return 0; + } + break; + case 0x1e: + if (cmd & (1 << 23)) + return debug(stream, "???", (cmd & 0xffff) + 1); + else + return debug(stream, "", 1); + break; + case 0x1f: + if ((cmd & (1 << 23)) == 0) + return debug_prim(stream, "3DPRIM (inline)", 1, (cmd & 0x1ffff) + 2); + else if (cmd & (1 << 17)) + { + if ((cmd & 0xffff) == 0) + return debug_variable_length_prim(stream); + else + return debug_prim(stream, "3DPRIM (indexed)", 0, (((cmd & 0xffff) + 1) / 2) + 1); + } + else + return debug_prim(stream, "3DPRIM (indirect sequential)", 0, 2); + break; + default: + return debug(stream, "", 0); + } + default: + assert(0); + return 0; + } + + assert(0); + return 0; +} + + + +void +i915_dump_batchbuffer( GLuint *start, + GLuint *end ) +{ + struct debug_stream stream; + GLuint bytes = (end - start) * 4; + GLboolean done = GL_FALSE; + + PRINTF("\n\nBATCH: (%d)\n", bytes / 4); + + stream.offset = 0; + stream.ptr = (char *)start; + stream.print_addresses = 0; + + while (!done && + stream.offset < bytes && + stream.offset >= 0) + { + if (!i915_debug_packet( &stream )) + break; + + assert(stream.offset <= bytes && + stream.offset >= 0); } - fprintf(stderr, "END\n\n"); + PRINTF("END-BATCH\n\n\n"); } + + diff --git a/src/mesa/drivers/dri/i915/i915_debug.h b/src/mesa/drivers/dri/i915/i915_debug.h new file mode 100644 index 0000000000..0643a8c631 --- /dev/null +++ b/src/mesa/drivers/dri/i915/i915_debug.h @@ -0,0 +1,55 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell <keith@tungstengraphics.com> + */ + +#ifndef I915_DEBUG_H +#define I915_DEBUG_H + +struct i915_context; + +struct debug_stream +{ + unsigned offset; /* current gtt offset */ + char *ptr; /* pointer to gtt offset zero */ + char *end; /* pointer to gtt offset zero */ + unsigned print_addresses; +}; + + + +extern void i915_disassemble_program(const unsigned *program, unsigned sz); +extern void i915_print_ureg(const char *msg, unsigned ureg); + + +void +i915_dump_batchbuffer( unsigned *start, + unsigned *end ); + + +#endif diff --git a/src/mesa/drivers/dri/i915/i915_debug_fp.c b/src/mesa/drivers/dri/i915/i915_debug_fp.c new file mode 100644 index 0000000000..84347a01ef --- /dev/null +++ b/src/mesa/drivers/dri/i915/i915_debug_fp.c @@ -0,0 +1,333 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include <stdio.h> + +#include "i915_reg.h" +#include "i915_debug.h" +#include "main/imports.h" +#include "shader/program.h" +#include "shader/prog_instruction.h" +#include "shader/prog_print.h" + +#define PRINTF( ... ) _mesa_printf( __VA_ARGS__ ) + +static const char *opcodes[0x20] = { + "NOP", + "ADD", + "MOV", + "MUL", + "MAD", + "DP2ADD", + "DP3", + "DP4", + "FRC", + "RCP", + "RSQ", + "EXP", + "LOG", + "CMP", + "MIN", + "MAX", + "FLR", + "MOD", + "TRC", + "SGE", + "SLT", + "TEXLD", + "TEXLDP", + "TEXLDB", + "TEXKILL", + "DCL", + "0x1a", + "0x1b", + "0x1c", + "0x1d", + "0x1e", + "0x1f", +}; + + +static const int args[0x20] = { + 0, /* 0 nop */ + 2, /* 1 add */ + 1, /* 2 mov */ + 2, /* 3 m ul */ + 3, /* 4 mad */ + 3, /* 5 dp2add */ + 2, /* 6 dp3 */ + 2, /* 7 dp4 */ + 1, /* 8 frc */ + 1, /* 9 rcp */ + 1, /* a rsq */ + 1, /* b exp */ + 1, /* c log */ + 3, /* d cmp */ + 2, /* e min */ + 2, /* f max */ + 1, /* 10 flr */ + 1, /* 11 mod */ + 1, /* 12 trc */ + 2, /* 13 sge */ + 2, /* 14 slt */ + 1, + 1, + 1, + 1, + 0, + 0, + 0, + 0, + 0, + 0, + 0, +}; + + +static const char *regname[0x8] = { + "R", + "T", + "CONST", + "S", + "OC", + "OD", + "U", + "UNKNOWN", +}; + +static void +print_reg_type_nr(GLuint type, GLuint nr) +{ + switch (type) { + case REG_TYPE_T: + switch (nr) { + case T_DIFFUSE: + PRINTF("T_DIFFUSE"); + return; + case T_SPECULAR: + PRINTF("T_SPECULAR"); + return; + case T_FOG_W: + PRINTF("T_FOG_W"); + return; + default: + PRINTF("T_TEX%d", nr); + return; + } + case REG_TYPE_OC: + if (nr == 0) { + PRINTF("oC"); + return; + } + break; + case REG_TYPE_OD: + if (nr == 0) { + PRINTF("oD"); + return; + } + break; + default: + break; + } + + PRINTF("%s[%d]", regname[type], nr); +} + +#define REG_SWIZZLE_MASK 0x7777 +#define REG_NEGATE_MASK 0x8888 + +#define REG_SWIZZLE_XYZW ((SRC_X << A2_SRC2_CHANNEL_X_SHIFT) | \ + (SRC_Y << A2_SRC2_CHANNEL_Y_SHIFT) | \ + (SRC_Z << A2_SRC2_CHANNEL_Z_SHIFT) | \ + (SRC_W << A2_SRC2_CHANNEL_W_SHIFT)) + + +static void +print_reg_neg_swizzle(GLuint reg) +{ + int i; + + if ((reg & REG_SWIZZLE_MASK) == REG_SWIZZLE_XYZW && + (reg & REG_NEGATE_MASK) == 0) + return; + + PRINTF("."); + + for (i = 3; i >= 0; i--) { + if (reg & (1 << ((i * 4) + 3))) + PRINTF("-"); + + switch ((reg >> (i * 4)) & 0x7) { + case 0: + PRINTF("x"); + break; + case 1: + PRINTF("y"); + break; + case 2: + PRINTF("z"); + break; + case 3: + PRINTF("w"); + break; + case 4: + PRINTF("0"); + break; + case 5: + PRINTF("1"); + break; + default: + PRINTF("?"); + break; + } + } +} + + +static void +print_src_reg(GLuint dword) +{ + GLuint nr = (dword >> A2_SRC2_NR_SHIFT) & REG_NR_MASK; + GLuint type = (dword >> A2_SRC2_TYPE_SHIFT) & REG_TYPE_MASK; + print_reg_type_nr(type, nr); + print_reg_neg_swizzle(dword); +} + + +static void +print_dest_reg(GLuint dword) +{ + GLuint nr = (dword >> A0_DEST_NR_SHIFT) & REG_NR_MASK; + GLuint type = (dword >> A0_DEST_TYPE_SHIFT) & REG_TYPE_MASK; + print_reg_type_nr(type, nr); + if ((dword & A0_DEST_CHANNEL_ALL) == A0_DEST_CHANNEL_ALL) + return; + PRINTF("."); + if (dword & A0_DEST_CHANNEL_X) + PRINTF("x"); + if (dword & A0_DEST_CHANNEL_Y) + PRINTF("y"); + if (dword & A0_DEST_CHANNEL_Z) + PRINTF("z"); + if (dword & A0_DEST_CHANNEL_W) + PRINTF("w"); +} + + +#define GET_SRC0_REG(r0, r1) ((r0<<14)|(r1>>A1_SRC0_CHANNEL_W_SHIFT)) +#define GET_SRC1_REG(r0, r1) ((r0<<8)|(r1>>A2_SRC1_CHANNEL_W_SHIFT)) +#define GET_SRC2_REG(r) (r) + + +static void +print_arith_op(GLuint opcode, const GLuint * program) +{ + if (opcode != A0_NOP) { + print_dest_reg(program[0]); + if (program[0] & A0_DEST_SATURATE) + PRINTF(" = SATURATE "); + else + PRINTF(" = "); + } + + PRINTF("%s ", opcodes[opcode]); + + print_src_reg(GET_SRC0_REG(program[0], program[1])); + if (args[opcode] == 1) { + PRINTF("\n"); + return; + } + + PRINTF(", "); + print_src_reg(GET_SRC1_REG(program[1], program[2])); + if (args[opcode] == 2) { + PRINTF("\n"); + return; + } + + PRINTF(", "); + print_src_reg(GET_SRC2_REG(program[2])); + PRINTF("\n"); + return; +} + + +static void +print_tex_op(GLuint opcode, const GLuint * program) +{ + print_dest_reg(program[0] | A0_DEST_CHANNEL_ALL); + PRINTF(" = "); + + PRINTF("%s ", opcodes[opcode]); + + PRINTF("S[%d],", program[0] & T0_SAMPLER_NR_MASK); + + print_reg_type_nr((program[1] >> T1_ADDRESS_REG_TYPE_SHIFT) & + REG_TYPE_MASK, + (program[1] >> T1_ADDRESS_REG_NR_SHIFT) & REG_NR_MASK); + PRINTF("\n"); +} + +static void +print_dcl_op(GLuint opcode, const GLuint * program) +{ + PRINTF("%s ", opcodes[opcode]); + print_dest_reg(program[0] | A0_DEST_CHANNEL_ALL); + PRINTF("\n"); +} + + +void +i915_disassemble_program(const GLuint * program, GLuint sz) +{ + GLuint size = program[0] & 0x1ff; + GLint i; + + PRINTF("\t\tBEGIN\n"); + + assert(size + 2 == sz); + + program++; + for (i = 1; i < sz; i += 3, program += 3) { + GLuint opcode = program[0] & (0x1f << 24); + + PRINTF("\t\t"); + + if ((GLint) opcode >= A0_NOP && opcode <= A0_SLT) + print_arith_op(opcode >> 24, program); + else if (opcode >= T0_TEXLD && opcode <= T0_TEXKILL) + print_tex_op(opcode >> 24, program); + else if (opcode == D0_DCL) + print_dcl_op(opcode >> 24, program); + else + PRINTF("Unknown opcode 0x%x\n", opcode); + } + + PRINTF("\t\tEND\n\n"); +} + + diff --git a/src/mesa/drivers/dri/i915/i915_fragprog.c b/src/mesa/drivers/dri/i915/i915_fragprog.c index 702b878828..4c3f2236e5 100644 --- a/src/mesa/drivers/dri/i915/i915_fragprog.c +++ b/src/mesa/drivers/dri/i915/i915_fragprog.c @@ -29,40 +29,44 @@ #include "macros.h" #include "enums.h" +#include "shader/prog_instruction.h" +#include "shader/prog_parameter.h" +#include "shader/program.h" +#include "shader/programopt.h" + #include "tnl/tnl.h" #include "tnl/t_context.h" + #include "intel_batchbuffer.h" #include "i915_reg.h" #include "i915_context.h" #include "i915_program.h" -#include "prog_instruction.h" -#include "prog_parameter.h" -#include "program.h" -#include "programopt.h" - /* 1, -1/3!, 1/5!, -1/7! */ -static const GLfloat sin_constants[4] = { 1.0, - -1.0/(3*2*1), - 1.0/(5*4*3*2*1), - -1.0/(7*6*5*4*3*2*1) }; +static const GLfloat sin_constants[4] = { 1.0, + -1.0 / (3 * 2 * 1), + 1.0 / (5 * 4 * 3 * 2 * 1), + -1.0 / (7 * 6 * 5 * 4 * 3 * 2 * 1) +}; /* 1, -1/2!, 1/4!, -1/6! */ -static const GLfloat cos_constants[4] = { 1.0, - -1.0/(2*1), - 1.0/(4*3*2*1), - -1.0/(6*5*4*3*2*1) }; +static const GLfloat cos_constants[4] = { 1.0, + -1.0 / (2 * 1), + 1.0 / (4 * 3 * 2 * 1), + -1.0 / (6 * 5 * 4 * 3 * 2 * 1) +}; /** * Retrieve a ureg for the given source register. Will emit * constants, apply swizzling and negation as needed. */ -static GLuint src_vector( struct i915_fragment_program *p, - const struct prog_src_register *source, - const struct gl_fragment_program *program ) +static GLuint +src_vector(struct i915_fragment_program *p, + const struct prog_src_register *source, + const struct gl_fragment_program *program) { GLuint src; @@ -70,136 +74,152 @@ static GLuint src_vector( struct i915_fragment_program *p, /* Registers: */ - case PROGRAM_TEMPORARY: - if (source->Index >= I915_MAX_TEMPORARY) { - i915_program_error( p, "Exceeded max temporary reg" ); - return 0; - } - src = UREG( REG_TYPE_R, source->Index ); + case PROGRAM_TEMPORARY: + if (source->Index >= I915_MAX_TEMPORARY) { + i915_program_error(p, "Exceeded max temporary reg"); + return 0; + } + src = UREG(REG_TYPE_R, source->Index); + break; + case PROGRAM_INPUT: + switch (source->Index) { + case FRAG_ATTRIB_WPOS: + src = i915_emit_decl(p, REG_TYPE_T, p->wpos_tex, D0_CHANNEL_ALL); break; - case PROGRAM_INPUT: - switch (source->Index) { - case FRAG_ATTRIB_WPOS: - src = i915_emit_decl( p, REG_TYPE_T, p->wpos_tex, D0_CHANNEL_ALL ); - break; - case FRAG_ATTRIB_COL0: - src = i915_emit_decl( p, REG_TYPE_T, T_DIFFUSE, D0_CHANNEL_ALL ); - break; - case FRAG_ATTRIB_COL1: - src = i915_emit_decl( p, REG_TYPE_T, T_SPECULAR, D0_CHANNEL_XYZ ); - src = swizzle( src, X, Y, Z, ONE ); - break; - case FRAG_ATTRIB_FOGC: - src = i915_emit_decl( p, REG_TYPE_T, T_FOG_W, D0_CHANNEL_W ); - src = swizzle( src, W, W, W, W ); - break; - case FRAG_ATTRIB_TEX0: - case FRAG_ATTRIB_TEX1: - case FRAG_ATTRIB_TEX2: - case FRAG_ATTRIB_TEX3: - case FRAG_ATTRIB_TEX4: - case FRAG_ATTRIB_TEX5: - case FRAG_ATTRIB_TEX6: - case FRAG_ATTRIB_TEX7: - src = i915_emit_decl( p, REG_TYPE_T, - T_TEX0 + (source->Index - FRAG_ATTRIB_TEX0), - D0_CHANNEL_ALL ); - break; - - default: - i915_program_error( p, "Bad source->Index" ); - return 0; - } + case FRAG_ATTRIB_COL0: + src = i915_emit_decl(p, REG_TYPE_T, T_DIFFUSE, D0_CHANNEL_ALL); + break; + case FRAG_ATTRIB_COL1: + src = i915_emit_decl(p, REG_TYPE_T, T_SPECULAR, D0_CHANNEL_XYZ); + src = swizzle(src, X, Y, Z, ONE); + break; + case FRAG_ATTRIB_FOGC: + src = i915_emit_decl(p, REG_TYPE_T, T_FOG_W, D0_CHANNEL_W); + src = swizzle(src, W, W, W, W); + break; + case FRAG_ATTRIB_TEX0: + case FRAG_ATTRIB_TEX1: + case FRAG_ATTRIB_TEX2: + case FRAG_ATTRIB_TEX3: + case FRAG_ATTRIB_TEX4: + case FRAG_ATTRIB_TEX5: + case FRAG_ATTRIB_TEX6: + case FRAG_ATTRIB_TEX7: + src = i915_emit_decl(p, REG_TYPE_T, + T_TEX0 + (source->Index - FRAG_ATTRIB_TEX0), + D0_CHANNEL_ALL); break; - - /* Various paramters and env values. All emitted to - * hardware as program constants. - */ - case PROGRAM_LOCAL_PARAM: - src = i915_emit_param4fv( - p, program->Base.LocalParams[source->Index]); - break; - - case PROGRAM_ENV_PARAM: - src = i915_emit_param4fv( - p, p->ctx->FragmentProgram.Parameters[source->Index]); - break; - - case PROGRAM_CONSTANT: - case PROGRAM_STATE_VAR: - case PROGRAM_NAMED_PARAM: - src = i915_emit_param4fv( - p, program->Base.Parameters->ParameterValues[source->Index] ); - break; default: - i915_program_error( p, "Bad source->File" ); - return 0; + i915_program_error(p, "Bad source->Index"); + return 0; + } + break; + + /* Various paramters and env values. All emitted to + * hardware as program constants. + */ + case PROGRAM_LOCAL_PARAM: + src = i915_emit_param4fv(p, program->Base.LocalParams[source->Index]); + break; + + case PROGRAM_ENV_PARAM: + src = + i915_emit_param4fv(p, + p->ctx->FragmentProgram.Parameters[source-> + Index]); + break; + + case PROGRAM_CONSTANT: + case PROGRAM_STATE_VAR: + case PROGRAM_NAMED_PARAM: + src = + i915_emit_param4fv(p, + program->Base.Parameters->ParameterValues[source-> + Index]); + break; + + default: + i915_program_error(p, "Bad source->File"); + return 0; } - src = swizzle(src, - GET_SWZ(source->Swizzle, 0), - GET_SWZ(source->Swizzle, 1), - GET_SWZ(source->Swizzle, 2), - GET_SWZ(source->Swizzle, 3)); + src = swizzle(src, + GET_SWZ(source->Swizzle, 0), + GET_SWZ(source->Swizzle, 1), + GET_SWZ(source->Swizzle, 2), GET_SWZ(source->Swizzle, 3)); if (source->NegateBase) - src = negate( src, - GET_BIT(source->NegateBase, 0), - GET_BIT(source->NegateBase, 1), - GET_BIT(source->NegateBase, 2), - GET_BIT(source->NegateBase, 3)); + src = negate(src, + GET_BIT(source->NegateBase, 0), + GET_BIT(source->NegateBase, 1), + GET_BIT(source->NegateBase, 2), + GET_BIT(source->NegateBase, 3)); return src; } -static GLuint get_result_vector( struct i915_fragment_program *p, - const struct prog_instruction *inst ) +static GLuint +get_result_vector(struct i915_fragment_program *p, + const struct prog_instruction *inst) { switch (inst->DstReg.File) { case PROGRAM_OUTPUT: switch (inst->DstReg.Index) { - case FRAG_RESULT_COLR: - return UREG(REG_TYPE_OC, 0); - case FRAG_RESULT_DEPR: - p->depth_written = 1; - return UREG(REG_TYPE_OD, 0); - default: - i915_program_error( p, "Bad inst->DstReg.Index" ); - return 0; + case FRAG_RESULT_COLR: + return UREG(REG_TYPE_OC, 0); + case FRAG_RESULT_DEPR: + p->depth_written = 1; + return UREG(REG_TYPE_OD, 0); + default: + i915_program_error(p, "Bad inst->DstReg.Index"); + return 0; } case PROGRAM_TEMPORARY: return UREG(REG_TYPE_R, inst->DstReg.Index); default: - i915_program_error( p, "Bad inst->DstReg.File" ); + i915_program_error(p, "Bad inst->DstReg.File"); return 0; } } - -static GLuint get_result_flags( const struct prog_instruction *inst ) + +static GLuint +get_result_flags(const struct prog_instruction *inst) { GLuint flags = 0; - if (inst->SaturateMode == SATURATE_ZERO_ONE) flags |= A0_DEST_SATURATE; - if (inst->DstReg.WriteMask & WRITEMASK_X) flags |= A0_DEST_CHANNEL_X; - if (inst->DstReg.WriteMask & WRITEMASK_Y) flags |= A0_DEST_CHANNEL_Y; - if (inst->DstReg.WriteMask & WRITEMASK_Z) flags |= A0_DEST_CHANNEL_Z; - if (inst->DstReg.WriteMask & WRITEMASK_W) flags |= A0_DEST_CHANNEL_W; + if (inst->SaturateMode == SATURATE_ZERO_ONE) + flags |= A0_DEST_SATURATE; + if (inst->DstReg.WriteMask & WRITEMASK_X) + flags |= A0_DEST_CHANNEL_X; + if (inst->DstReg.WriteMask & WRITEMASK_Y) + flags |= A0_DEST_CHANNEL_Y; + if (inst->DstReg.WriteMask & WRITEMASK_Z) + flags |= A0_DEST_CHANNEL_Z; + if (inst->DstReg.WriteMask & WRITEMASK_W) + flags |= A0_DEST_CHANNEL_W; return flags; } -static GLuint translate_tex_src_target( struct i915_fragment_program *p, - GLubyte bit ) +static GLuint +translate_tex_src_target(struct i915_fragment_program *p, GLubyte bit) { switch (bit) { - case TEXTURE_1D_INDEX: return D0_SAMPLE_TYPE_2D; - case TEXTURE_2D_INDEX: return D0_SAMPLE_TYPE_2D; - case TEXTURE_RECT_INDEX: return D0_SAMPLE_TYPE_2D; - case TEXTURE_3D_INDEX: return D0_SAMPLE_TYPE_VOLUME; - case TEXTURE_CUBE_INDEX: return D0_SAMPLE_TYPE_CUBE; - default: i915_program_error(p, "TexSrcBit"); return 0; + case TEXTURE_1D_INDEX: + return D0_SAMPLE_TYPE_2D; + case TEXTURE_2D_INDEX: + return D0_SAMPLE_TYPE_2D; + case TEXTURE_RECT_INDEX: + return D0_SAMPLE_TYPE_2D; + case TEXTURE_3D_INDEX: + return D0_SAMPLE_TYPE_VOLUME; + case TEXTURE_CUBE_INDEX: + return D0_SAMPLE_TYPE_CUBE; + default: + i915_program_error(p, "TexSrcBit"); + return 0; } } @@ -246,9 +266,11 @@ do { \ * can lead to confusion -- hopefully we cope with it ok now. * */ -static void upload_program( struct i915_fragment_program *p ) +static void +upload_program(struct i915_fragment_program *p) { - const struct gl_fragment_program *program = p->ctx->FragmentProgram._Current; + const struct gl_fragment_program *program = + p->ctx->FragmentProgram._Current; const struct prog_instruction *inst = program->Base.Instructions; /* _mesa_debug_fp_inst(program->Base.NumInstructions, inst); */ @@ -258,12 +280,12 @@ static void upload_program( struct i915_fragment_program *p ) * this being uploaded to hardware. */ if (inst[0].Opcode == OPCODE_END) { - GLuint tmp = i915_get_utemp( p ); - i915_emit_arith( p, - A0_MOV, - UREG(REG_TYPE_OC, 0), - A0_DEST_CHANNEL_ALL, 0, - swizzle(tmp,ONE,ZERO,ONE,ONE), 0, 0); + GLuint tmp = i915_get_utemp(p); + i915_emit_arith(p, + A0_MOV, + UREG(REG_TYPE_OC, 0), + A0_DEST_CHANNEL_ALL, 0, + swizzle(tmp, ONE, ZERO, ONE, ONE), 0, 0); return; } @@ -272,496 +294,465 @@ static void upload_program( struct i915_fragment_program *p ) GLuint tmp = 0; switch (inst->Opcode) { - case OPCODE_ABS: - src0 = src_vector( p, &inst->SrcReg[0], program); - i915_emit_arith( p, - A0_MAX, - get_result_vector( p, inst ), - get_result_flags( inst ), 0, - src0, negate(src0, 1,1,1,1), 0); - break; - - case OPCODE_ADD: - EMIT_2ARG_ARITH( A0_ADD ); - break; - - case OPCODE_CMP: - src0 = src_vector( p, &inst->SrcReg[0], program); - src1 = src_vector( p, &inst->SrcReg[1], program); - src2 = src_vector( p, &inst->SrcReg[2], program); - i915_emit_arith( p, - A0_CMP, - get_result_vector( p, inst ), - get_result_flags( inst ), 0, - src0, src2, src1); /* NOTE: order of src2, src1 */ - break; + case OPCODE_ABS: + src0 = src_vector(p, &inst->SrcReg[0], program); + i915_emit_arith(p, + A0_MAX, + get_result_vector(p, inst), + get_result_flags(inst), 0, + src0, negate(src0, 1, 1, 1, 1), 0); + break; + + case OPCODE_ADD: + EMIT_2ARG_ARITH(A0_ADD); + break; + + case OPCODE_CMP: + src0 = src_vector(p, &inst->SrcReg[0], program); + src1 = src_vector(p, &inst->SrcReg[1], program); + src2 = src_vector(p, &inst->SrcReg[2], program); + i915_emit_arith(p, A0_CMP, get_result_vector(p, inst), get_result_flags(inst), 0, src0, src2, src1); /* NOTE: order of src2, src1 */ + break; case OPCODE_COS: - src0 = src_vector( p, &inst->SrcReg[0], program); - tmp = i915_get_utemp( p ); - - i915_emit_arith( p, - A0_MUL, - tmp, A0_DEST_CHANNEL_X, 0, - src0, - i915_emit_const1f(p, 1.0/(M_PI * 2)), - 0); - - i915_emit_arith( p, - A0_MOD, - tmp, A0_DEST_CHANNEL_X, 0, - tmp, - 0, 0 ); - - /* By choosing different taylor constants, could get rid of this mul: - */ - i915_emit_arith( p, - A0_MUL, - tmp, A0_DEST_CHANNEL_X, 0, - tmp, - i915_emit_const1f(p, (M_PI * 2)), - 0); - - /* - * t0.xy = MUL x.xx11, x.x1111 ; x^2, x, 1, 1 - * t0 = MUL t0.xyxy t0.xx11 ; x^4, x^3, x^2, 1 - * t0 = MUL t0.xxz1 t0.z111 ; x^6 x^4 x^2 1 - * result = DP4 t0, cos_constants - */ - i915_emit_arith( p, - A0_MUL, - tmp, A0_DEST_CHANNEL_XY, 0, - swizzle(tmp, X,X,ONE,ONE), - swizzle(tmp, X,ONE,ONE,ONE), 0); - - i915_emit_arith( p, - A0_MUL, - tmp, A0_DEST_CHANNEL_XYZ, 0, - swizzle(tmp, X,Y,X,ONE), - swizzle(tmp, X,X,ONE,ONE), 0); - - i915_emit_arith( p, - A0_MUL, - tmp, A0_DEST_CHANNEL_XYZ, 0, - swizzle(tmp, X,X,Z,ONE), - swizzle(tmp, Z,ONE,ONE,ONE), 0); - - i915_emit_arith( p, - A0_DP4, - get_result_vector( p, inst ), - get_result_flags( inst ), 0, - swizzle(tmp, ONE,Z,Y,X), - i915_emit_const4fv( p, cos_constants ), 0); - - break; - - case OPCODE_DP3: - EMIT_2ARG_ARITH( A0_DP3 ); - break; - - case OPCODE_DP4: - EMIT_2ARG_ARITH( A0_DP4 ); - break; - - case OPCODE_DPH: - src0 = src_vector( p, &inst->SrcReg[0], program); - src1 = src_vector( p, &inst->SrcReg[1], program); - - i915_emit_arith( p, - A0_DP4, - get_result_vector( p, inst ), - get_result_flags( inst ), 0, - swizzle(src0, X,Y,Z,ONE), src1, 0); - break; - - case OPCODE_DST: - src0 = src_vector( p, &inst->SrcReg[0], program); - src1 = src_vector( p, &inst->SrcReg[1], program); - - /* result[0] = 1 * 1; - * result[1] = a[1] * b[1]; - * result[2] = a[2] * 1; - * result[3] = 1 * b[3]; - */ - i915_emit_arith( p, - A0_MUL, - get_result_vector( p, inst ), - get_result_flags( inst ), 0, - swizzle(src0, ONE, Y, Z, ONE), - swizzle(src1, ONE, Y, ONE, W ), - 0); - break; - - case OPCODE_EX2: - src0 = src_vector( p, &inst->SrcReg[0], program); - - i915_emit_arith( p, - A0_EXP, - get_result_vector( p, inst ), - get_result_flags( inst ), 0, - swizzle(src0,X,X,X,X), 0, 0); - break; - - case OPCODE_FLR: - EMIT_1ARG_ARITH( A0_FLR ); - break; - - case OPCODE_FRC: - EMIT_1ARG_ARITH( A0_FRC ); - break; + src0 = src_vector(p, &inst->SrcReg[0], program); + tmp = i915_get_utemp(p); + + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_X, 0, + src0, i915_emit_const1f(p, 1.0 / (M_PI)), 0); + + i915_emit_arith(p, A0_MOD, tmp, A0_DEST_CHANNEL_X, 0, tmp, 0, 0); + + /* By choosing different taylor constants, could get rid of this mul: + */ + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_X, 0, + tmp, i915_emit_const1f(p, (M_PI)), 0); + + /* + * t0.xy = MUL x.xx11, x.x1111 ; x^2, x, 1, 1 + * t0 = MUL t0.xyxy t0.xx11 ; x^4, x^3, x^2, 1 + * t0 = MUL t0.xxz1 t0.z111 ; x^6 x^4 x^2 1 + * result = DP4 t0, cos_constants + */ + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_XY, 0, + swizzle(tmp, X, X, ONE, ONE), + swizzle(tmp, X, ONE, ONE, ONE), 0); + + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_XYZ, 0, + swizzle(tmp, X, Y, X, ONE), + swizzle(tmp, X, X, ONE, ONE), 0); + + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_XYZ, 0, + swizzle(tmp, X, X, Z, ONE), + swizzle(tmp, Z, ONE, ONE, ONE), 0); + + i915_emit_arith(p, + A0_DP4, + get_result_vector(p, inst), + get_result_flags(inst), 0, + swizzle(tmp, ONE, Z, Y, X), + i915_emit_const4fv(p, cos_constants), 0); + + break; + + case OPCODE_DP3: + EMIT_2ARG_ARITH(A0_DP3); + break; + + case OPCODE_DP4: + EMIT_2ARG_ARITH(A0_DP4); + break; + + case OPCODE_DPH: + src0 = src_vector(p, &inst->SrcReg[0], program); + src1 = src_vector(p, &inst->SrcReg[1], program); + + i915_emit_arith(p, + A0_DP4, + get_result_vector(p, inst), + get_result_flags(inst), 0, + swizzle(src0, X, Y, Z, ONE), src1, 0); + break; + + case OPCODE_DST: + src0 = src_vector(p, &inst->SrcReg[0], program); + src1 = src_vector(p, &inst->SrcReg[1], program); + + /* result[0] = 1 * 1; + * result[1] = a[1] * b[1]; + * result[2] = a[2] * 1; + * result[3] = 1 * b[3]; + */ + i915_emit_arith(p, + A0_MUL, + get_result_vector(p, inst), + get_result_flags(inst), 0, + swizzle(src0, ONE, Y, Z, ONE), + swizzle(src1, ONE, Y, ONE, W), 0); + break; + + case OPCODE_EX2: + src0 = src_vector(p, &inst->SrcReg[0], program); + + i915_emit_arith(p, + A0_EXP, + get_result_vector(p, inst), + get_result_flags(inst), 0, + swizzle(src0, X, X, X, X), 0, 0); + break; + + case OPCODE_FLR: + EMIT_1ARG_ARITH(A0_FLR); + break; + + case OPCODE_FRC: + EMIT_1ARG_ARITH(A0_FRC); + break; case OPCODE_KIL: - src0 = src_vector( p, &inst->SrcReg[0], program); - tmp = i915_get_utemp( p ); - - i915_emit_texld( p, - tmp, A0_DEST_CHANNEL_ALL, /* use a dummy dest reg */ - 0, - src0, - T0_TEXKILL ); - break; - - case OPCODE_LG2: - src0 = src_vector( p, &inst->SrcReg[0], program); - - i915_emit_arith( p, - A0_LOG, - get_result_vector( p, inst ), - get_result_flags( inst ), 0, - swizzle(src0,X,X,X,X), 0, 0); - break; - - case OPCODE_LIT: - src0 = src_vector( p, &inst->SrcReg[0], program); - tmp = i915_get_utemp( p ); - - /* tmp = max( a.xyzw, a.00zw ) - * XXX: Clamp tmp.w to -128..128 - * tmp.y = log(tmp.y) - * tmp.y = tmp.w * tmp.y - * tmp.y = exp(tmp.y) - * result = cmp (a.11-x1, a.1x01, a.1xy1 ) - */ - i915_emit_arith( p, A0_MAX, tmp, A0_DEST_CHANNEL_ALL, 0, - src0, swizzle(src0, ZERO, ZERO, Z, W), 0 ); - - i915_emit_arith( p, A0_LOG, tmp, A0_DEST_CHANNEL_Y, 0, - swizzle(tmp, Y, Y, Y, Y), 0, 0 ); - - i915_emit_arith( p, A0_MUL, tmp, A0_DEST_CHANNEL_Y, 0, - swizzle(tmp, ZERO, Y, ZERO, ZERO), - swizzle(tmp, ZERO, W, ZERO, ZERO), 0 ); - - i915_emit_arith( p, A0_EXP, tmp, A0_DEST_CHANNEL_Y, 0, - swizzle(tmp, Y, Y, Y, Y), 0, 0 ); - - i915_emit_arith( p, A0_CMP, - get_result_vector( p, inst ), - get_result_flags( inst ), 0, - negate(swizzle(tmp, ONE, ONE, X, ONE),0,0,1,0), - swizzle(tmp, ONE, X, ZERO, ONE), - swizzle(tmp, ONE, X, Y, ONE)); - - break; - - case OPCODE_LRP: - src0 = src_vector( p, &inst->SrcReg[0], program); - src1 = src_vector( p, &inst->SrcReg[1], program); - src2 = src_vector( p, &inst->SrcReg[2], program); - flags = get_result_flags( inst ); - tmp = i915_get_utemp( p ); - - /* b*a + c*(1-a) - * - * b*a + c - ca - * - * tmp = b*a + c, - * result = (-c)*a + tmp - */ - i915_emit_arith( p, A0_MAD, tmp, - flags & A0_DEST_CHANNEL_ALL, 0, - src1, src0, src2 ); - - i915_emit_arith( p, A0_MAD, - get_result_vector( p, inst ), - flags, 0, - negate(src2, 1,1,1,1), src0, tmp ); - break; + src0 = src_vector(p, &inst->SrcReg[0], program); + tmp = i915_get_utemp(p); + + i915_emit_texld(p, tmp, A0_DEST_CHANNEL_ALL, /* use a dummy dest reg */ + 0, src0, T0_TEXKILL); + break; + + case OPCODE_LG2: + src0 = src_vector(p, &inst->SrcReg[0], program); + + i915_emit_arith(p, + A0_LOG, + get_result_vector(p, inst), + get_result_flags(inst), 0, + swizzle(src0, X, X, X, X), 0, 0); + break; + + case OPCODE_LIT: + src0 = src_vector(p, &inst->SrcReg[0], program); + tmp = i915_get_utemp(p); + + /* tmp = max( a.xyzw, a.00zw ) + * XXX: Clamp tmp.w to -128..128 + * tmp.y = log(tmp.y) + * tmp.y = tmp.w * tmp.y + * tmp.y = exp(tmp.y) + * result = cmp (a.11-x1, a.1x01, a.1xy1 ) + */ + i915_emit_arith(p, A0_MAX, tmp, A0_DEST_CHANNEL_ALL, 0, + src0, swizzle(src0, ZERO, ZERO, Z, W), 0); + + i915_emit_arith(p, A0_LOG, tmp, A0_DEST_CHANNEL_Y, 0, + swizzle(tmp, Y, Y, Y, Y), 0, 0); + + i915_emit_arith(p, A0_MUL, tmp, A0_DEST_CHANNEL_Y, 0, + swizzle(tmp, ZERO, Y, ZERO, ZERO), + swizzle(tmp, ZERO, W, ZERO, ZERO), 0); + + i915_emit_arith(p, A0_EXP, tmp, A0_DEST_CHANNEL_Y, 0, + swizzle(tmp, Y, Y, Y, Y), 0, 0); + + i915_emit_arith(p, A0_CMP, + get_result_vector(p, inst), + get_result_flags(inst), 0, + negate(swizzle(tmp, ONE, ONE, X, ONE), 0, 0, 1, 0), + swizzle(tmp, ONE, X, ZERO, ONE), + swizzle(tmp, ONE, X, Y, ONE)); + + break; + + case OPCODE_LRP: + src0 = src_vector(p, &inst->SrcReg[0], program); + src1 = src_vector(p, &inst->SrcReg[1], program); + src2 = src_vector(p, &inst->SrcReg[2], program); + flags = get_result_flags(inst); + tmp = i915_get_utemp(p); + + /* b*a + c*(1-a) + * + * b*a + c - ca + * + * tmp = b*a + c, + * result = (-c)*a + tmp + */ + i915_emit_arith(p, A0_MAD, tmp, + flags & A0_DEST_CHANNEL_ALL, 0, src1, src0, src2); + + i915_emit_arith(p, A0_MAD, + get_result_vector(p, inst), + flags, 0, negate(src2, 1, 1, 1, 1), src0, tmp); + break; case OPCODE_MAD: - EMIT_3ARG_ARITH( A0_MAD ); - break; + EMIT_3ARG_ARITH(A0_MAD); + break; case OPCODE_MAX: - EMIT_2ARG_ARITH( A0_MAX ); - break; - - case OPCODE_MIN: - src0 = src_vector( p, &inst->SrcReg[0], program); - src1 = src_vector( p, &inst->SrcReg[1], program); - tmp = i915_get_utemp( p ); - flags = get_result_flags( inst ); - - i915_emit_arith( p, - A0_MAX, - tmp, flags & A0_DEST_CHANNEL_ALL, 0, - negate(src0,1,1,1,1), - negate(src1,1,1,1,1), 0); - - i915_emit_arith( p, - A0_MOV, - get_result_vector( p, inst ), - flags, 0, - negate(tmp, 1,1,1,1), 0, 0); - break; - - case OPCODE_MOV: - EMIT_1ARG_ARITH( A0_MOV ); - break; - - case OPCODE_MUL: - EMIT_2ARG_ARITH( A0_MUL ); - break; - - case OPCODE_POW: - src0 = src_vector( p, &inst->SrcReg[0], program); - src1 = src_vector( p, &inst->SrcReg[1], program); - tmp = i915_get_utemp( p ); - flags = get_result_flags( inst ); - - /* XXX: masking on intermediate values, here and elsewhere. - */ - i915_emit_arith( p, - A0_LOG, - tmp, A0_DEST_CHANNEL_X, 0, - swizzle(src0,X,X,X,X), 0, 0); - - i915_emit_arith( p, - A0_MUL, - tmp, A0_DEST_CHANNEL_X, 0, - tmp, src1, 0); - - - i915_emit_arith( p, - A0_EXP, - get_result_vector( p, inst ), - flags, 0, - swizzle(tmp,X,X,X,X), 0, 0); - - break; - - case OPCODE_RCP: - src0 = src_vector( p, &inst->SrcReg[0], program); - - i915_emit_arith( p, - A0_RCP, - get_result_vector( p, inst ), - get_result_flags( inst ), 0, - swizzle(src0,X,X,X,X), 0, 0); - break; - - case OPCODE_RSQ: - - src0 = src_vector( p, &inst->SrcReg[0], program); - - i915_emit_arith( p, - A0_RSQ, - get_result_vector( p, inst ), - get_result_flags( inst ), 0, - swizzle(src0,X,X,X,X), 0, 0); - break; - + EMIT_2ARG_ARITH(A0_MAX); + break; + + case OPCODE_MIN: + src0 = src_vector(p, &inst->SrcReg[0], program); + src1 = src_vector(p, &inst->SrcReg[1], program); + tmp = i915_get_utemp(p); + flags = get_result_flags(inst); + + i915_emit_arith(p, + A0_MAX, + tmp, flags & A0_DEST_CHANNEL_ALL, 0, + negate(src0, 1, 1, 1, 1), + negate(src1, 1, 1, 1, 1), 0); + + i915_emit_arith(p, + A0_MOV, + get_result_vector(p, inst), + flags, 0, negate(tmp, 1, 1, 1, 1), 0, 0); + break; + + case OPCODE_MOV: + EMIT_1ARG_ARITH(A0_MOV); + break; + + case OPCODE_MUL: + EMIT_2ARG_ARITH(A0_MUL); + break; + + case OPCODE_POW: + src0 = src_vector(p, &inst->SrcReg[0], program); + src1 = src_vector(p, &inst->SrcReg[1], program); + tmp = i915_get_utemp(p); + flags = get_result_flags(inst); + + /* XXX: masking on intermediate values, here and elsewhere. + */ + i915_emit_arith(p, + A0_LOG, + tmp, A0_DEST_CHANNEL_X, 0, + swizzle(src0, X, X, X, X), 0, 0); + + i915_emit_arith(p, A0_MUL, tmp, A0_DEST_CHANNEL_X, 0, tmp, src1, 0); + + + i915_emit_arith(p, + A0_EXP, + get_result_vector(p, inst), + flags, 0, swizzle(tmp, X, X, X, X), 0, 0); + + break; + + case OPCODE_RCP: + src0 = src_vector(p, &inst->SrcReg[0], program); + + i915_emit_arith(p, + A0_RCP, + get_result_vector(p, inst), + get_result_flags(inst), 0, + swizzle(src0, X, X, X, X), 0, 0); + break; + + case OPCODE_RSQ: + + src0 = src_vector(p, &inst->SrcReg[0], program); + + i915_emit_arith(p, + A0_RSQ, + get_result_vector(p, inst), + get_result_flags(inst), 0, + swizzle(src0, X, X, X, X), 0, 0); + break; + case OPCODE_SCS: - src0 = src_vector( p, &inst->SrcReg[0], program); - tmp = i915_get_utemp( p ); - - /* - * t0.xy = MUL x.xx11, x.x1111 ; x^2, x, 1, 1 - * t0 = MUL t0.xyxy t0.xx11 ; x^4, x^3, x^2, x - * t1 = MUL t0.xyyw t0.yz11 ; x^7 x^5 x^3 x - * scs.x = DP4 t1, sin_constants - * t1 = MUL t0.xxz1 t0.z111 ; x^6 x^4 x^2 1 - * scs.y = DP4 t1, cos_constants - */ - i915_emit_arith( p, - A0_MUL, - tmp, A0_DEST_CHANNEL_XY, 0, - swizzle(src0, X,X,ONE,ONE), - swizzle(src0, X,ONE,ONE,ONE), 0); - - i915_emit_arith( p, - A0_MUL, - tmp, A0_DEST_CHANNEL_ALL, 0, - swizzle(tmp, X,Y,X,Y), - swizzle(tmp, X,X,ONE,ONE), 0); - - if (inst->DstReg.WriteMask & WRITEMASK_Y) { - GLuint tmp1; - - if (inst->DstReg.WriteMask & WRITEMASK_X) - tmp1 = i915_get_utemp( p ); - else - tmp1 = tmp; - - i915_emit_arith( p, - A0_MUL, - tmp1, A0_DEST_CHANNEL_ALL, 0, - swizzle(tmp, X,Y,Y,W), - swizzle(tmp, X,Z,ONE,ONE), 0); - - i915_emit_arith( p, - A0_DP4, - get_result_vector( p, inst ), - A0_DEST_CHANNEL_Y, 0, - swizzle(tmp1, W,Z,Y,X), - i915_emit_const4fv( p, sin_constants ), 0); - } - - if (inst->DstReg.WriteMask & WRITEMASK_X) { - i915_emit_arith( p, - A0_MUL, - tmp, A0_DEST_CHANNEL_XYZ, 0, - swizzle(tmp, X,X,Z,ONE), - swizzle(tmp, Z,ONE,ONE,ONE), 0); - - i915_emit_arith( p, - A0_DP4, - get_result_vector( p, inst ), - A0_DEST_CHANNEL_X, 0, - swizzle(tmp, ONE,Z,Y,X), - i915_emit_const4fv( p, cos_constants ), 0); - } - break; - - case OPCODE_SGE: - EMIT_2ARG_ARITH( A0_SGE ); - break; + src0 = src_vector(p, &inst->SrcReg[0], program); + tmp = i915_get_utemp(p); + + /* + * t0.xy = MUL x.xx11, x.x1111 ; x^2, x, 1, 1 + * t0 = MUL t0.xyxy t0.xx11 ; x^4, x^3, x^2, x + * t1 = MUL t0.xyyw t0.yz11 ; x^7 x^5 x^3 x + * scs.x = DP4 t1, sin_constants + * t1 = MUL t0.xxz1 t0.z111 ; x^6 x^4 x^2 1 + * scs.y = DP4 t1, cos_constants + */ + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_XY, 0, + swizzle(src0, X, X, ONE, ONE), + swizzle(src0, X, ONE, ONE, ONE), 0); + + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_ALL, 0, + swizzle(tmp, X, Y, X, Y), + swizzle(tmp, X, X, ONE, ONE), 0); + + if (inst->DstReg.WriteMask & WRITEMASK_Y) { + GLuint tmp1; + + if (inst->DstReg.WriteMask & WRITEMASK_X) + tmp1 = i915_get_utemp(p); + else + tmp1 = tmp; + + i915_emit_arith(p, + A0_MUL, + tmp1, A0_DEST_CHANNEL_ALL, 0, + swizzle(tmp, X, Y, Y, W), + swizzle(tmp, X, Z, ONE, ONE), 0); + + i915_emit_arith(p, + A0_DP4, + get_result_vector(p, inst), + A0_DEST_CHANNEL_Y, 0, + swizzle(tmp1, W, Z, Y, X), + i915_emit_const4fv(p, sin_constants), 0); + } + + if (inst->DstReg.WriteMask & WRITEMASK_X) { + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_XYZ, 0, + swizzle(tmp, X, X, Z, ONE), + swizzle(tmp, Z, ONE, ONE, ONE), 0); + + i915_emit_arith(p, + A0_DP4, + get_result_vector(p, inst), + A0_DEST_CHANNEL_X, 0, + swizzle(tmp, ONE, Z, Y, X), + i915_emit_const4fv(p, cos_constants), 0); + } + break; + + case OPCODE_SGE: + EMIT_2ARG_ARITH(A0_SGE); + break; case OPCODE_SIN: - src0 = src_vector( p, &inst->SrcReg[0], program); - tmp = i915_get_utemp( p ); - - i915_emit_arith( p, - A0_MUL, - tmp, A0_DEST_CHANNEL_X, 0, - src0, - i915_emit_const1f(p, 1.0/(M_PI * 2)), - 0); - - i915_emit_arith( p, - A0_MOD, - tmp, A0_DEST_CHANNEL_X, 0, - tmp, - 0, 0 ); - - /* By choosing different taylor constants, could get rid of this mul: - */ - i915_emit_arith( p, - A0_MUL, - tmp, A0_DEST_CHANNEL_X, 0, - tmp, - i915_emit_const1f(p, (M_PI * 2)), - 0); - - /* - * t0.xy = MUL x.xx11, x.x1111 ; x^2, x, 1, 1 - * t0 = MUL t0.xyxy t0.xx11 ; x^4, x^3, x^2, x - * t1 = MUL t0.xyyw t0.yz11 ; x^7 x^5 x^3 x - * result = DP4 t1.wzyx, sin_constants - */ - i915_emit_arith( p, - A0_MUL, - tmp, A0_DEST_CHANNEL_XY, 0, - swizzle(tmp, X,X,ONE,ONE), - swizzle(tmp, X,ONE,ONE,ONE), 0); - - i915_emit_arith( p, - A0_MUL, - tmp, A0_DEST_CHANNEL_ALL, 0, - swizzle(tmp, X,Y,X,Y), - swizzle(tmp, X,X,ONE,ONE), 0); - - i915_emit_arith( p, - A0_MUL, - tmp, A0_DEST_CHANNEL_ALL, 0, - swizzle(tmp, X,Y,Y,W), - swizzle(tmp, X,Z,ONE,ONE), 0); - - i915_emit_arith( p, - A0_DP4, - get_result_vector( p, inst ), - get_result_flags( inst ), 0, - swizzle(tmp, W, Z, Y, X ), - i915_emit_const4fv( p, sin_constants ), 0); - break; - - case OPCODE_SLT: - EMIT_2ARG_ARITH( A0_SLT ); - break; - - case OPCODE_SUB: - src0 = src_vector( p, &inst->SrcReg[0], program); - src1 = src_vector( p, &inst->SrcReg[1], program); - - i915_emit_arith( p, - A0_ADD, - get_result_vector( p, inst ), - get_result_flags( inst ), 0, - src0, negate(src1, 1,1,1,1), 0); - break; - - case OPCODE_SWZ: - EMIT_1ARG_ARITH( A0_MOV ); /* extended swizzle handled natively */ - break; - - case OPCODE_TEX: - EMIT_TEX( T0_TEXLD ); - break; + src0 = src_vector(p, &inst->SrcReg[0], program); + tmp = i915_get_utemp(p); + + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_X, 0, + src0, i915_emit_const1f(p, 1.0 / (M_PI)), 0); + + i915_emit_arith(p, A0_MOD, tmp, A0_DEST_CHANNEL_X, 0, tmp, 0, 0); + + /* By choosing different taylor constants, could get rid of this mul: + */ + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_X, 0, + tmp, i915_emit_const1f(p, (M_PI)), 0); + + /* + * t0.xy = MUL x.xx11, x.x1111 ; x^2, x, 1, 1 + * t0 = MUL t0.xyxy t0.xx11 ; x^4, x^3, x^2, x + * t1 = MUL t0.xyyw t0.yz11 ; x^7 x^5 x^3 x + * result = DP4 t1.wzyx, sin_constants + */ + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_XY, 0, + swizzle(tmp, X, X, ONE, ONE), + swizzle(tmp, X, ONE, ONE, ONE), 0); + + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_ALL, 0, + swizzle(tmp, X, Y, X, Y), + swizzle(tmp, X, X, ONE, ONE), 0); + + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_ALL, 0, + swizzle(tmp, X, Y, Y, W), + swizzle(tmp, X, Z, ONE, ONE), 0); + + i915_emit_arith(p, + A0_DP4, + get_result_vector(p, inst), + get_result_flags(inst), 0, + swizzle(tmp, W, Z, Y, X), + i915_emit_const4fv(p, sin_constants), 0); + break; + + case OPCODE_SLT: + EMIT_2ARG_ARITH(A0_SLT); + break; + + case OPCODE_SUB: + src0 = src_vector(p, &inst->SrcReg[0], program); + src1 = src_vector(p, &inst->SrcReg[1], program); + + i915_emit_arith(p, + A0_ADD, + get_result_vector(p, inst), + get_result_flags(inst), 0, + src0, negate(src1, 1, 1, 1, 1), 0); + break; + + case OPCODE_SWZ: + EMIT_1ARG_ARITH(A0_MOV); /* extended swizzle handled natively */ + break; + + case OPCODE_TEX: + EMIT_TEX(T0_TEXLD); + break; case OPCODE_TXB: - EMIT_TEX( T0_TEXLDB ); - break; + EMIT_TEX(T0_TEXLDB); + break; case OPCODE_TXP: - EMIT_TEX( T0_TEXLDP ); - break; + EMIT_TEX(T0_TEXLDP); + break; case OPCODE_XPD: - /* Cross product: - * result.x = src0.y * src1.z - src0.z * src1.y; - * result.y = src0.z * src1.x - src0.x * src1.z; - * result.z = src0.x * src1.y - src0.y * src1.x; - * result.w = undef; - */ - src0 = src_vector( p, &inst->SrcReg[0], program); - src1 = src_vector( p, &inst->SrcReg[1], program); - tmp = i915_get_utemp( p ); - - i915_emit_arith( p, - A0_MUL, - tmp, A0_DEST_CHANNEL_ALL, 0, - swizzle(src0,Z,X,Y,ONE), - swizzle(src1,Y,Z,X,ONE), 0); - - i915_emit_arith( p, - A0_MAD, - get_result_vector( p, inst ), - get_result_flags( inst ), 0, - swizzle(src0,Y,Z,X,ONE), - swizzle(src1,Z,X,Y,ONE), - negate(tmp,1,1,1,0)); - break; + /* Cross product: + * result.x = src0.y * src1.z - src0.z * src1.y; + * result.y = src0.z * src1.x - src0.x * src1.z; + * result.z = src0.x * src1.y - src0.y * src1.x; + * result.w = undef; + */ + src0 = src_vector(p, &inst->SrcReg[0], program); + src1 = src_vector(p, &inst->SrcReg[1], program); + tmp = i915_get_utemp(p); + + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_ALL, 0, + swizzle(src0, Z, X, Y, ONE), + swizzle(src1, Y, Z, X, ONE), 0); + + i915_emit_arith(p, + A0_MAD, + get_result_vector(p, inst), + get_result_flags(inst), 0, + swizzle(src0, Y, Z, X, ONE), + swizzle(src1, Z, X, Y, ONE), + negate(tmp, 1, 1, 1, 0)); + break; case OPCODE_END: - return; - + return; + default: - i915_program_error( p, "bad opcode" ); - return; + i915_program_error(p, "bad opcode"); + return; } inst++; - i915_release_utemps( p ); + i915_release_utemps(p); } } @@ -769,21 +760,22 @@ static void upload_program( struct i915_fragment_program *p ) * emit, just move the value into its correct position at the end of * the program: */ -static void fixup_depth_write( struct i915_fragment_program *p ) +static void +fixup_depth_write(struct i915_fragment_program *p) { if (p->depth_written) { GLuint depth = UREG(REG_TYPE_OD, 0); - i915_emit_arith( p, - A0_MOV, - depth, A0_DEST_CHANNEL_W, 0, - swizzle(depth,X,Y,Z,Z), - 0, 0); + i915_emit_arith(p, + A0_MOV, + depth, A0_DEST_CHANNEL_W, 0, + swizzle(depth, X, Y, Z, Z), 0, 0); } } -static void check_wpos( struct i915_fragment_program *p ) +static void +check_wpos(struct i915_fragment_program *p) { GLuint inputs = p->FragProg.Base.InputsRead; GLint i; @@ -791,12 +783,12 @@ static void check_wpos( struct i915_fragment_program *p ) p->wpos_tex = -1; for (i = 0; i < p->ctx->Const.MaxTextureCoordUnits; i++) { - if (inputs & FRAG_BIT_TEX(i)) - continue; + if (inputs & FRAG_BIT_TEX(i)) + continue; else if (inputs & FRAG_BIT_WPOS) { - p->wpos_tex = i; - inputs &= ~FRAG_BIT_WPOS; - } + p->wpos_tex = i; + inputs &= ~FRAG_BIT_WPOS; + } } if (inputs & FRAG_BIT_WPOS) { @@ -805,53 +797,54 @@ static void check_wpos( struct i915_fragment_program *p ) } -static void translate_program( struct i915_fragment_program *p ) +static void +translate_program(struct i915_fragment_program *p) { - i915ContextPtr i915 = I915_CONTEXT(p->ctx); - - i915_init_program( i915, p ); - check_wpos( p ); - upload_program( p ); - fixup_depth_write( p ); - i915_fini_program( p ); - + struct i915_context *i915 = I915_CONTEXT(p->ctx); + + i915_init_program(i915, p); + check_wpos(p); + upload_program(p); + fixup_depth_write(p); + i915_fini_program(p); + p->translated = 1; } -static void track_params( struct i915_fragment_program *p ) +static void +track_params(struct i915_fragment_program *p) { GLint i; if (p->nr_params) - _mesa_load_state_parameters(p->ctx, p->FragProg.Base.Parameters); + _mesa_load_state_parameters(p->ctx, p->FragProg.Base.Parameters); for (i = 0; i < p->nr_params; i++) { GLint reg = p->param[i].reg; - COPY_4V( p->constant[reg], p->param[i].values ); + COPY_4V(p->constant[reg], p->param[i].values); } - + p->params_uptodate = 1; - p->on_hardware = 0; /* overkill */ + p->on_hardware = 0; /* overkill */ } -static void i915BindProgram( GLcontext *ctx, - GLenum target, - struct gl_program *prog ) +static void +i915BindProgram(GLcontext * ctx, GLenum target, struct gl_program *prog) { if (target == GL_FRAGMENT_PROGRAM_ARB) { - i915ContextPtr i915 = I915_CONTEXT(ctx); - struct i915_fragment_program *p = (struct i915_fragment_program *)prog; + struct i915_context *i915 = I915_CONTEXT(ctx); + struct i915_fragment_program *p = (struct i915_fragment_program *) prog; + + if (i915->current_program == p) + return; - if (i915->current_program == p) - return; - if (i915->current_program) { - i915->current_program->on_hardware = 0; - i915->current_program->params_uptodate = 0; + i915->current_program->on_hardware = 0; + i915->current_program->params_uptodate = 0; } - + i915->current_program = p; assert(p->on_hardware == 0); @@ -860,71 +853,70 @@ static void i915BindProgram( GLcontext *ctx, } } -static struct gl_program *i915NewProgram( GLcontext *ctx, - GLenum target, - GLuint id ) +static struct gl_program * +i915NewProgram(GLcontext * ctx, GLenum target, GLuint id) { switch (target) { case GL_VERTEX_PROGRAM_ARB: - return _mesa_init_vertex_program( ctx, CALLOC_STRUCT(gl_vertex_program), - target, id ); - - case GL_FRAGMENT_PROGRAM_ARB: { - struct i915_fragment_program *prog = CALLOC_STRUCT(i915_fragment_program); - if (prog) { - i915_init_program( I915_CONTEXT(ctx), prog ); - - return _mesa_init_fragment_program( ctx, &prog->FragProg, - target, id ); + return _mesa_init_vertex_program(ctx, CALLOC_STRUCT(gl_vertex_program), + target, id); + + case GL_FRAGMENT_PROGRAM_ARB:{ + struct i915_fragment_program *prog = + CALLOC_STRUCT(i915_fragment_program); + if (prog) { + i915_init_program(I915_CONTEXT(ctx), prog); + + return _mesa_init_fragment_program(ctx, &prog->FragProg, + target, id); + } + else + return NULL; } - else - return NULL; - } default: /* Just fallback: */ - return _mesa_new_program( ctx, target, id ); + return _mesa_new_program(ctx, target, id); } } -static void i915DeleteProgram( GLcontext *ctx, - struct gl_program *prog ) +static void +i915DeleteProgram(GLcontext * ctx, struct gl_program *prog) { if (prog->Target == GL_FRAGMENT_PROGRAM_ARB) { - i915ContextPtr i915 = I915_CONTEXT(ctx); - struct i915_fragment_program *p = (struct i915_fragment_program *)prog; - - if (i915->current_program == p) - i915->current_program = 0; + struct i915_context *i915 = I915_CONTEXT(ctx); + struct i915_fragment_program *p = (struct i915_fragment_program *) prog; + + if (i915->current_program == p) + i915->current_program = 0; } - _mesa_delete_program( ctx, prog ); + _mesa_delete_program(ctx, prog); } -static GLboolean i915IsProgramNative( GLcontext *ctx, - GLenum target, - struct gl_program *prog ) +static GLboolean +i915IsProgramNative(GLcontext * ctx, GLenum target, struct gl_program *prog) { if (target == GL_FRAGMENT_PROGRAM_ARB) { - struct i915_fragment_program *p = (struct i915_fragment_program *)prog; + struct i915_fragment_program *p = (struct i915_fragment_program *) prog; if (!p->translated) - translate_program( p ); - + translate_program(p); + return !p->error; } else return GL_TRUE; } -static void i915ProgramStringNotify( GLcontext *ctx, - GLenum target, - struct gl_program *prog ) +static void +i915ProgramStringNotify(GLcontext * ctx, + GLenum target, struct gl_program *prog) { if (target == GL_FRAGMENT_PROGRAM_ARB) { - struct i915_fragment_program *p = (struct i915_fragment_program *)prog; + struct i915_fragment_program *p = (struct i915_fragment_program *) prog; p->translated = 0; /* Hack: make sure fog is correctly enabled according to this @@ -941,28 +933,28 @@ static void i915ProgramStringNotify( GLcontext *ctx, } -void i915ValidateFragmentProgram( i915ContextPtr i915 ) +void +i915ValidateFragmentProgram(struct i915_context *i915) { GLcontext *ctx = &i915->intel.ctx; - intelContextPtr intel = INTEL_CONTEXT(ctx); + struct intel_context *intel = intel_context(ctx); TNLcontext *tnl = TNL_CONTEXT(ctx); struct vertex_buffer *VB = &tnl->vb; - struct i915_fragment_program *p = - (struct i915_fragment_program *)ctx->FragmentProgram._Current; + struct i915_fragment_program *p = + (struct i915_fragment_program *) ctx->FragmentProgram._Current; const GLuint inputsRead = p->FragProg.Base.InputsRead; GLuint s4 = i915->state.Ctx[I915_CTXREG_LIS4] & ~S4_VFMT_MASK; GLuint s2 = S2_TEXCOORD_NONE; int i, offset = 0; - if (i915->current_program != p) - { + if (i915->current_program != p) { if (i915->current_program) { - i915->current_program->on_hardware = 0; - i915->current_program->params_uptodate = 0; + i915->current_program->on_hardware = 0; + i915->current_program->params_uptodate = 0; } - + i915->current_program = p; } @@ -971,8 +963,8 @@ void i915ValidateFragmentProgram( i915ContextPtr i915 ) */ VB->AttribPtr[VERT_ATTRIB_POS] = VB->NdcPtr; - if (!p->translated) - translate_program( p ); + if (!p->translated) + translate_program(p); intel->vertex_attr_count = 0; intel->wpos_offset = 0; @@ -981,31 +973,31 @@ void i915ValidateFragmentProgram( i915ContextPtr i915 ) intel->specoffset = 0; if (inputsRead & FRAG_BITS_TEX_ANY) { - EMIT_ATTR( _TNL_ATTRIB_POS, EMIT_4F_VIEWPORT, S4_VFMT_XYZW, 16 ); + EMIT_ATTR(_TNL_ATTRIB_POS, EMIT_4F_VIEWPORT, S4_VFMT_XYZW, 16); } else { - EMIT_ATTR( _TNL_ATTRIB_POS, EMIT_3F_VIEWPORT, S4_VFMT_XYZ, 12 ); + EMIT_ATTR(_TNL_ATTRIB_POS, EMIT_3F_VIEWPORT, S4_VFMT_XYZ, 12); } if (inputsRead & FRAG_BIT_COL0) { intel->coloroffset = offset / 4; - EMIT_ATTR( _TNL_ATTRIB_COLOR0, EMIT_4UB_4F_BGRA, S4_VFMT_COLOR, 4 ); + EMIT_ATTR(_TNL_ATTRIB_COLOR0, EMIT_4UB_4F_BGRA, S4_VFMT_COLOR, 4); } - - if ((inputsRead & (FRAG_BIT_COL1|FRAG_BIT_FOGC)) || + + if ((inputsRead & (FRAG_BIT_COL1 | FRAG_BIT_FOGC)) || i915->vertex_fog != I915_FOG_NONE) { if (inputsRead & FRAG_BIT_COL1) { - intel->specoffset = offset / 4; - EMIT_ATTR( _TNL_ATTRIB_COLOR1, EMIT_3UB_3F_BGR, S4_VFMT_SPEC_FOG, 3 ); + intel->specoffset = offset / 4; + EMIT_ATTR(_TNL_ATTRIB_COLOR1, EMIT_3UB_3F_BGR, S4_VFMT_SPEC_FOG, 3); } else - EMIT_PAD(3); + EMIT_PAD(3); - if ((inputsRead & FRAG_BIT_FOGC) || i915->vertex_fog != I915_FOG_NONE) - EMIT_ATTR( _TNL_ATTRIB_FOG, EMIT_1UB_1F, S4_VFMT_SPEC_FOG, 1 ); + if ((inputsRead & FRAG_BIT_FOGC) || i915->vertex_fog != I915_FOG_NONE) + EMIT_ATTR(_TNL_ATTRIB_FOG, EMIT_1UB_1F, S4_VFMT_SPEC_FOG, 1); else - EMIT_PAD( 1 ); + EMIT_PAD(1); } /* XXX this was disabled, but enabling this code helped fix the Glean @@ -1013,63 +1005,66 @@ void i915ValidateFragmentProgram( i915ContextPtr i915 ) */ #if 1 if ((inputsRead & FRAG_BIT_FOGC) || i915->vertex_fog != I915_FOG_NONE) { - EMIT_ATTR( _TNL_ATTRIB_FOG, EMIT_1F, S4_VFMT_FOG_PARAM, 4 ); + EMIT_ATTR(_TNL_ATTRIB_FOG, EMIT_1F, S4_VFMT_FOG_PARAM, 4); } #endif for (i = 0; i < p->ctx->Const.MaxTextureCoordUnits; i++) { if (inputsRead & FRAG_BIT_TEX(i)) { - int sz = VB->TexCoordPtr[i]->size; - - s2 &= ~S2_TEXCOORD_FMT(i, S2_TEXCOORD_FMT0_MASK); - s2 |= S2_TEXCOORD_FMT(i, SZ_TO_HW(sz)); + int sz = VB->TexCoordPtr[i]->size; + + s2 &= ~S2_TEXCOORD_FMT(i, S2_TEXCOORD_FMT0_MASK); + s2 |= S2_TEXCOORD_FMT(i, SZ_TO_HW(sz)); - EMIT_ATTR( _TNL_ATTRIB_TEX0+i, EMIT_SZ(sz), 0, sz * 4 ); + EMIT_ATTR(_TNL_ATTRIB_TEX0 + i, EMIT_SZ(sz), 0, sz * 4); } else if (i == p->wpos_tex) { - - /* If WPOS is required, duplicate the XYZ position data in an - * unused texture coordinate: - */ - s2 &= ~S2_TEXCOORD_FMT(i, S2_TEXCOORD_FMT0_MASK); - s2 |= S2_TEXCOORD_FMT(i, SZ_TO_HW(3)); - - intel->wpos_offset = offset; - intel->wpos_size = 3 * sizeof(GLuint); - - EMIT_PAD( intel->wpos_size ); - } + + /* If WPOS is required, duplicate the XYZ position data in an + * unused texture coordinate: + */ + s2 &= ~S2_TEXCOORD_FMT(i, S2_TEXCOORD_FMT0_MASK); + s2 |= S2_TEXCOORD_FMT(i, SZ_TO_HW(3)); + + intel->wpos_offset = offset; + intel->wpos_size = 3 * sizeof(GLuint); + + EMIT_PAD(intel->wpos_size); + } } if (s2 != i915->state.Ctx[I915_CTXREG_LIS2] || s4 != i915->state.Ctx[I915_CTXREG_LIS4]) { - - I915_STATECHANGE( i915, I915_UPLOAD_CTX ); + int k; + + I915_STATECHANGE(i915, I915_UPLOAD_CTX); /* Must do this *after* statechange, so as not to affect * buffered vertices reliant on the old state: */ - intel->vertex_size = _tnl_install_attrs( &intel->ctx, - intel->vertex_attrs, - intel->vertex_attr_count, - intel->ViewportMatrix.m, 0 ); + intel->vertex_size = _tnl_install_attrs(&intel->ctx, + intel->vertex_attrs, + intel->vertex_attr_count, + intel->ViewportMatrix.m, 0); intel->vertex_size >>= 2; i915->state.Ctx[I915_CTXREG_LIS2] = s2; i915->state.Ctx[I915_CTXREG_LIS4] = s4; - assert(intel->vtbl.check_vertex_size( intel, intel->vertex_size )); + k = intel->vtbl.check_vertex_size(intel, intel->vertex_size); + assert(k); } - if (!p->params_uptodate) - track_params( p ); + if (!p->params_uptodate) + track_params(p); - if (!p->on_hardware) - i915_upload_program( i915, p ); + if (!p->on_hardware) + i915_upload_program(i915, p); } -void i915InitFragProgFuncs( struct dd_function_table *functions ) +void +i915InitFragProgFuncs(struct dd_function_table *functions) { functions->BindProgram = i915BindProgram; functions->NewProgram = i915NewProgram; diff --git a/src/mesa/drivers/dri/i915/i915_metaops.c b/src/mesa/drivers/dri/i915/i915_metaops.c index 1be7ac4c48..a739bd6581 100644 --- a/src/mesa/drivers/dri/i915/i915_metaops.c +++ b/src/mesa/drivers/dri/i915/i915_metaops.c @@ -34,128 +34,170 @@ #include "intel_screen.h" #include "intel_batchbuffer.h" #include "intel_ioctl.h" +#include "intel_regions.h" #include "intel_rotate.h" #include "i915_context.h" #include "i915_reg.h" -/* A large amount of state doesn't need to be uploaded. +/* We touch almost everything: */ -#define ACTIVE (I915_UPLOAD_INVARIENT | \ - I915_UPLOAD_PROGRAM | \ - I915_UPLOAD_STIPPLE | \ +#define ACTIVE (I915_UPLOAD_INVARIENT | \ I915_UPLOAD_CTX | \ I915_UPLOAD_BUFFERS | \ - I915_UPLOAD_TEX(0)) + I915_UPLOAD_STIPPLE | \ + I915_UPLOAD_PROGRAM | \ + I915_UPLOAD_FOG | \ + I915_UPLOAD_TEX(0)) -#define SET_STATE( i915, STATE ) \ +#define SET_STATE( i915, STATE ) \ do { \ i915->current->emitted &= ~ACTIVE; \ - i915->current = &i915->STATE; \ + i915->current = &i915->STATE; \ i915->current->emitted &= ~ACTIVE; \ } while (0) -/* Operations where the 3D engine is decoupled temporarily from the - * current GL state and used for other purposes than simply rendering - * incoming triangles. - */ -static void set_initial_state( i915ContextPtr i915 ) -{ - memcpy(&i915->meta, &i915->initial, sizeof(i915->meta) ); - i915->meta.active = ACTIVE; - i915->meta.emitted = 0; -} - -static void set_no_depth_stencil_write( i915ContextPtr i915 ) +static void +meta_no_stencil_write(struct intel_context *intel) { + struct i915_context *i915 = i915_context(&intel->ctx); + /* ctx->Driver.Enable( ctx, GL_STENCIL_TEST, GL_FALSE ) */ - i915->meta.Ctx[I915_CTXREG_LIS5] &= ~(S5_STENCIL_TEST_ENABLE | - S5_STENCIL_WRITE_ENABLE); + i915->meta.Ctx[I915_CTXREG_LIS5] &= ~(S5_STENCIL_TEST_ENABLE | + S5_STENCIL_WRITE_ENABLE); + + i915->meta.emitted &= ~I915_UPLOAD_CTX; +} + +static void +meta_no_depth_write(struct intel_context *intel) +{ + struct i915_context *i915 = i915_context(&intel->ctx); /* ctx->Driver.Enable( ctx, GL_DEPTH_TEST, GL_FALSE ) */ i915->meta.Ctx[I915_CTXREG_LIS6] &= ~(S6_DEPTH_TEST_ENABLE | - S6_DEPTH_WRITE_ENABLE); + S6_DEPTH_WRITE_ENABLE); i915->meta.emitted &= ~I915_UPLOAD_CTX; } +static void +meta_depth_replace(struct intel_context *intel) +{ + struct i915_context *i915 = i915_context(&intel->ctx); + + /* ctx->Driver.Enable( ctx, GL_DEPTH_TEST, GL_TRUE ) + * ctx->Driver.DepthMask( ctx, GL_TRUE ) + */ + i915->meta.Ctx[I915_CTXREG_LIS6] |= (S6_DEPTH_TEST_ENABLE | + S6_DEPTH_WRITE_ENABLE); + + /* ctx->Driver.DepthFunc( ctx, GL_ALWAYS ) + */ + i915->meta.Ctx[I915_CTXREG_LIS6] &= ~S6_DEPTH_TEST_FUNC_MASK; + i915->meta.Ctx[I915_CTXREG_LIS6] |= + COMPAREFUNC_ALWAYS << S6_DEPTH_TEST_FUNC_SHIFT; + + i915->meta.emitted &= ~I915_UPLOAD_CTX; +} + + /* Set stencil unit to replace always with the reference value. */ -static void set_stencil_replace( i915ContextPtr i915, - GLuint s_mask, - GLuint s_clear) +static void +meta_stencil_replace(struct intel_context *intel, + GLuint s_mask, GLuint s_clear) { + struct i915_context *i915 = i915_context(&intel->ctx); GLuint op = STENCILOP_REPLACE; GLuint func = COMPAREFUNC_ALWAYS; /* ctx->Driver.Enable( ctx, GL_STENCIL_TEST, GL_TRUE ) */ - i915->meta.Ctx[I915_CTXREG_LIS5] |= (S5_STENCIL_TEST_ENABLE | - S5_STENCIL_WRITE_ENABLE); - - - /* ctx->Driver.Enable( ctx, GL_DEPTH_TEST, GL_FALSE ) - */ - i915->meta.Ctx[I915_CTXREG_LIS6] &= ~(S6_DEPTH_TEST_ENABLE | - S6_DEPTH_WRITE_ENABLE); - + i915->meta.Ctx[I915_CTXREG_LIS5] |= (S5_STENCIL_TEST_ENABLE | + S5_STENCIL_WRITE_ENABLE); /* ctx->Driver.StencilMask( ctx, s_mask ) */ i915->meta.Ctx[I915_CTXREG_STATE4] &= ~MODE4_ENABLE_STENCIL_WRITE_MASK; i915->meta.Ctx[I915_CTXREG_STATE4] |= (ENABLE_STENCIL_WRITE_MASK | - STENCIL_WRITE_MASK(s_mask)); - + STENCIL_WRITE_MASK(s_mask)); /* ctx->Driver.StencilOp( ctx, GL_REPLACE, GL_REPLACE, GL_REPLACE ) */ i915->meta.Ctx[I915_CTXREG_LIS5] &= ~(S5_STENCIL_FAIL_MASK | - S5_STENCIL_PASS_Z_FAIL_MASK | - S5_STENCIL_PASS_Z_PASS_MASK); + S5_STENCIL_PASS_Z_FAIL_MASK | + S5_STENCIL_PASS_Z_PASS_MASK); i915->meta.Ctx[I915_CTXREG_LIS5] |= ((op << S5_STENCIL_FAIL_SHIFT) | - (op << S5_STENCIL_PASS_Z_FAIL_SHIFT) | - (op << S5_STENCIL_PASS_Z_PASS_SHIFT)); + (op << S5_STENCIL_PASS_Z_FAIL_SHIFT) | + (op << S5_STENCIL_PASS_Z_PASS_SHIFT)); /* ctx->Driver.StencilFunc( ctx, GL_ALWAYS, s_ref, ~0 ) */ i915->meta.Ctx[I915_CTXREG_STATE4] &= ~MODE4_ENABLE_STENCIL_TEST_MASK; i915->meta.Ctx[I915_CTXREG_STATE4] |= (ENABLE_STENCIL_TEST_MASK | - STENCIL_TEST_MASK(0xff)); + STENCIL_TEST_MASK(0xff)); i915->meta.Ctx[I915_CTXREG_LIS5] &= ~(S5_STENCIL_REF_MASK | - S5_STENCIL_TEST_FUNC_MASK); - - i915->meta.Ctx[I915_CTXREG_LIS5] |= ((s_clear << S5_STENCIL_REF_SHIFT) | - (func << S5_STENCIL_TEST_FUNC_SHIFT)); + S5_STENCIL_TEST_FUNC_MASK); + + i915->meta.Ctx[I915_CTXREG_LIS5] |= ((s_clear << S5_STENCIL_REF_SHIFT) | + (func << S5_STENCIL_TEST_FUNC_SHIFT)); i915->meta.emitted &= ~I915_UPLOAD_CTX; } -static void set_color_mask( i915ContextPtr i915, GLboolean state ) +static void +meta_color_mask(struct intel_context *intel, GLboolean state) { + struct i915_context *i915 = i915_context(&intel->ctx); const GLuint mask = (S5_WRITEDISABLE_RED | - S5_WRITEDISABLE_GREEN | - S5_WRITEDISABLE_BLUE | - S5_WRITEDISABLE_ALPHA); + S5_WRITEDISABLE_GREEN | + S5_WRITEDISABLE_BLUE | S5_WRITEDISABLE_ALPHA); /* Copy colormask state from "regular" hw context. */ if (state) { i915->meta.Ctx[I915_CTXREG_LIS5] &= ~mask; - i915->meta.Ctx[I915_CTXREG_LIS5] |= - (i915->state.Ctx[I915_CTXREG_LIS5] & mask); + i915->meta.Ctx[I915_CTXREG_LIS5] |= + (i915->state.Ctx[I915_CTXREG_LIS5] & mask); } - else + else i915->meta.Ctx[I915_CTXREG_LIS5] |= mask; - + + i915->meta.emitted &= ~I915_UPLOAD_CTX; +} + + + +static void +meta_import_pixel_state(struct intel_context *intel) +{ + struct i915_context *i915 = i915_context(&intel->ctx); + memcpy(i915->meta.Fog, i915->state.Fog, I915_FOG_SETUP_SIZE * 4); + + i915->meta.Ctx[I915_CTXREG_LIS5] = i915->state.Ctx[I915_CTXREG_LIS5]; + i915->meta.Ctx[I915_CTXREG_LIS6] = i915->state.Ctx[I915_CTXREG_LIS6]; + i915->meta.Ctx[I915_CTXREG_STATE4] = i915->state.Ctx[I915_CTXREG_STATE4]; + i915->meta.Ctx[I915_CTXREG_BLENDCOLOR1] = + i915->state.Ctx[I915_CTXREG_BLENDCOLOR1]; + i915->meta.Ctx[I915_CTXREG_IAB] = i915->state.Ctx[I915_CTXREG_IAB]; + + i915->meta.Buffer[I915_DESTREG_SENABLE] = + i915->state.Buffer[I915_DESTREG_SENABLE]; + i915->meta.Buffer[I915_DESTREG_SR1] = i915->state.Buffer[I915_DESTREG_SR1]; + i915->meta.Buffer[I915_DESTREG_SR2] = i915->state.Buffer[I915_DESTREG_SR2]; + + i915->meta.emitted &= ~I915_UPLOAD_FOG; + i915->meta.emitted &= ~I915_UPLOAD_BUFFERS; i915->meta.emitted &= ~I915_UPLOAD_CTX; } @@ -212,69 +254,64 @@ static void set_color_mask( i915ContextPtr i915, GLboolean state ) -static void set_no_texture( i915ContextPtr i915 ) +static void +meta_no_texture(struct intel_context *intel) { + struct i915_context *i915 = i915_context(&intel->ctx); + static const GLuint prog[] = { _3DSTATE_PIXEL_SHADER_PROGRAM, /* Declare incoming diffuse color: */ - (D0_DCL | - D0_DECL_REG( REG_T_DIFFUSE ) | - D0_CHANNEL_ALL), + (D0_DCL | D0_DECL_REG(REG_T_DIFFUSE) | D0_CHANNEL_ALL), D1_MBZ, D2_MBZ, /* output-color = mov(t_diffuse) */ (A0_MOV | - A0_DEST_REG( REG_OC ) | - A0_DEST_CHANNEL_ALL | - A0_SRC0_REG( REG_T_DIFFUSE )), + A0_DEST_REG(REG_OC) | + A0_DEST_CHANNEL_ALL | A0_SRC0_REG(REG_T_DIFFUSE)), (A1_SRC0_XYZW), 0, }; - - memcpy( i915->meta.Program, prog, sizeof(prog) ); + + memcpy(i915->meta.Program, prog, sizeof(prog)); i915->meta.ProgramSize = sizeof(prog) / sizeof(*prog); i915->meta.Program[0] |= i915->meta.ProgramSize - 2; i915->meta.emitted &= ~I915_UPLOAD_PROGRAM; } - -static void enable_texture_blend_replace( i915ContextPtr i915 ) +static void +meta_texture_blend_replace(struct intel_context *intel) { + struct i915_context *i915 = i915_context(&intel->ctx); + static const GLuint prog[] = { _3DSTATE_PIXEL_SHADER_PROGRAM, /* Declare the sampler: */ - (D0_DCL | - D0_DECL_REG( REG_S(0) ) | - D0_SAMPLE_TYPE_2D | - D0_CHANNEL_NONE), + (D0_DCL | D0_DECL_REG(REG_S(0)) | D0_SAMPLE_TYPE_2D | D0_CHANNEL_NONE), D1_MBZ, D2_MBZ, /* Declare the interpolated texture coordinate: */ - (D0_DCL | - D0_DECL_REG( REG_T_TEX(0) ) | - D0_CHANNEL_ALL), + (D0_DCL | D0_DECL_REG(REG_T_TEX(0)) | D0_CHANNEL_ALL), D1_MBZ, D2_MBZ, /* output-color = texld(sample0, texcoord0) */ - (T0_TEXLD | - T0_DEST_REG( REG_OC ) | - T0_SAMPLER( 0 )), + (T0_TEXLD | T0_DEST_REG(REG_OC) | T0_SAMPLER(0)), T1_ADDRESS_REG(REG_TYPE_T, 0), T2_MBZ }; - memcpy( i915->meta.Program, prog, sizeof(prog) ); + memcpy(i915->meta.Program, prog, sizeof(prog)); i915->meta.ProgramSize = sizeof(prog) / sizeof(*prog); i915->meta.Program[0] |= i915->meta.ProgramSize - 2; i915->meta.emitted &= ~I915_UPLOAD_PROGRAM; @@ -287,425 +324,186 @@ static void enable_texture_blend_replace( i915ContextPtr i915 ) /* Set up an arbitary piece of memory as a rectangular texture * (including the front or back buffer). */ -static void set_tex_rect_source( i915ContextPtr i915, - GLuint offset, - GLuint width, - GLuint height, - GLuint pitch, /* in bytes! */ - GLuint textureFormat ) +static GLboolean +meta_tex_rect_source(struct intel_context *intel, + dri_bo *buffer, + GLuint offset, + GLuint pitch, GLuint height, GLenum format, GLenum type) { + struct i915_context *i915 = i915_context(&intel->ctx); GLuint unit = 0; GLint numLevels = 1; GLuint *state = i915->meta.Tex[0]; + GLuint textureFormat; + GLuint cpp; -#if 0 - printf("TexRect source offset 0x%x pitch %d\n", offset, pitch); -#endif + /* A full implementation of this would do the upload through + * glTexImage2d, and get all the conversion operations at that + * point. We are restricted, but still at least have access to the + * fragment program swizzle. + */ + switch (format) { + case GL_BGRA: + switch (type) { + case GL_UNSIGNED_INT_8_8_8_8_REV: + case GL_UNSIGNED_BYTE: + textureFormat = (MAPSURF_32BIT | MT_32BIT_ARGB8888); + cpp = 4; + break; + default: + return GL_FALSE; + } + break; + case GL_RGBA: + switch (type) { + case GL_UNSIGNED_INT_8_8_8_8_REV: + case GL_UNSIGNED_BYTE: + textureFormat = (MAPSURF_32BIT | MT_32BIT_ABGR8888); + cpp = 4; + break; + default: + return GL_FALSE; + } + break; + case GL_BGR: + switch (type) { + case GL_UNSIGNED_SHORT_5_6_5_REV: + textureFormat = (MAPSURF_16BIT | MT_16BIT_RGB565); + cpp = 2; + break; + default: + return GL_FALSE; + } + break; + case GL_RGB: + switch (type) { + case GL_UNSIGNED_SHORT_5_6_5: + textureFormat = (MAPSURF_16BIT | MT_16BIT_RGB565); + cpp = 2; + break; + default: + return GL_FALSE; + } + break; -/* fprintf(stderr, "%s: offset: %x w: %d h: %d pitch %d format %x\n", */ -/* __FUNCTION__, offset, width, height, pitch, textureFormat ); */ + default: + return GL_FALSE; + } + + + if ((pitch * cpp) & 3) { + _mesa_printf("%s: texture is not dword pitch\n", __FUNCTION__); + return GL_FALSE; + } + +/* intel_region_release(&i915->meta.tex_region[0]); */ +/* intel_region_reference(&i915->meta.tex_region[0], region); */ + i915->meta.tex_buffer[0] = buffer; + i915->meta.tex_offset[0] = offset; - state[I915_TEXREG_MS2] = offset; state[I915_TEXREG_MS3] = (((height - 1) << MS3_HEIGHT_SHIFT) | - ((width - 1) << MS3_WIDTH_SHIFT) | - textureFormat | - MS3_USE_FENCE_REGS); + ((pitch - 1) << MS3_WIDTH_SHIFT) | + textureFormat | MS3_USE_FENCE_REGS); - state[I915_TEXREG_MS4] = ((((pitch / 4) - 1) << MS4_PITCH_SHIFT) | - ((((numLevels-1) * 4)) << MS4_MAX_LOD_SHIFT)); + state[I915_TEXREG_MS4] = (((((pitch * cpp) / 4) - 1) << MS4_PITCH_SHIFT) | + MS4_CUBE_FACE_ENA_MASK | + ((((numLevels - 1) * 4)) << MS4_MAX_LOD_SHIFT)); state[I915_TEXREG_SS2] = ((FILTER_NEAREST << SS2_MIN_FILTER_SHIFT) | - (MIPFILTER_NONE << SS2_MIP_FILTER_SHIFT) | - (FILTER_NEAREST << SS2_MAG_FILTER_SHIFT)); + (MIPFILTER_NONE << SS2_MIP_FILTER_SHIFT) | + (FILTER_NEAREST << SS2_MAG_FILTER_SHIFT)); + state[I915_TEXREG_SS3] = ((TEXCOORDMODE_WRAP << SS3_TCX_ADDR_MODE_SHIFT) | - (TEXCOORDMODE_WRAP << SS3_TCY_ADDR_MODE_SHIFT) | - (TEXCOORDMODE_WRAP << SS3_TCZ_ADDR_MODE_SHIFT) | - (unit<<SS3_TEXTUREMAP_INDEX_SHIFT)); + (TEXCOORDMODE_WRAP << SS3_TCY_ADDR_MODE_SHIFT) | + (TEXCOORDMODE_WRAP << SS3_TCZ_ADDR_MODE_SHIFT) | + (unit << SS3_TEXTUREMAP_INDEX_SHIFT)); state[I915_TEXREG_SS4] = 0; i915->meta.emitted &= ~I915_UPLOAD_TEX(0); + return GL_TRUE; } -/* Select between front and back draw buffers. +/** + * Set the color and depth drawing region for meta ops. */ -static void set_draw_region( i915ContextPtr i915, const intelRegion *region ) +static void +meta_draw_region(struct intel_context *intel, + struct intel_region *color_region, + struct intel_region *depth_region) { -#if 0 - printf("Rotate into region: offset 0x%x pitch %d\n", - region->offset, region->pitch); -#endif - i915->meta.Buffer[I915_DESTREG_CBUFADDR1] = - (BUF_3D_ID_COLOR_BACK | BUF_3D_PITCH(region->pitch) | BUF_3D_USE_FENCE); - i915->meta.Buffer[I915_DESTREG_CBUFADDR2] = region->offset; - i915->meta.emitted &= ~I915_UPLOAD_BUFFERS; + struct i915_context *i915 = i915_context(&intel->ctx); + i915_state_draw_region(intel, &i915->meta, color_region, depth_region); } -#if 0 -/* Setup an arbitary draw format, useful for targeting texture or agp - * memory. - */ -static void set_draw_format( i915ContextPtr i915, - GLuint format, - GLuint depth_format) +static void +set_vertex_format(struct intel_context *intel) { - i915->meta.Buffer[I915_DESTREG_DV1] = (DSTORG_HORT_BIAS(0x8) | /* .5 */ - DSTORG_VERT_BIAS(0x8) | /* .5 */ - format | - LOD_PRECLAMP_OGL | - TEX_DEFAULT_COLOR_OGL | - depth_format); - - i915->meta.emitted &= ~I915_UPLOAD_BUFFERS; -/* fprintf(stderr, "%s: DV1: %x\n", */ -/* __FUNCTION__, i915->meta.Buffer[I915_DESTREG_DV1]); */ -} -#endif + struct i915_context *i915 = i915_context(&intel->ctx); -static void set_vertex_format( i915ContextPtr i915 ) -{ - i915->meta.Ctx[I915_CTXREG_LIS2] = + i915->meta.Ctx[I915_CTXREG_LIS2] = (S2_TEXCOORD_FMT(0, TEXCOORDFMT_2D) | - S2_TEXCOORD_FMT(1, TEXCOORDFMT_NOT_PRESENT) | + S2_TEXCOORD_FMT(1, TEXCOORDFMT_NOT_PRESENT) | S2_TEXCOORD_FMT(2, TEXCOORDFMT_NOT_PRESENT) | S2_TEXCOORD_FMT(3, TEXCOORDFMT_NOT_PRESENT) | S2_TEXCOORD_FMT(4, TEXCOORDFMT_NOT_PRESENT) | - S2_TEXCOORD_FMT(5, TEXCOORDFMT_NOT_PRESENT) | + S2_TEXCOORD_FMT(5, TEXCOORDFMT_NOT_PRESENT) | S2_TEXCOORD_FMT(6, TEXCOORDFMT_NOT_PRESENT) | S2_TEXCOORD_FMT(7, TEXCOORDFMT_NOT_PRESENT)); i915->meta.Ctx[I915_CTXREG_LIS4] &= ~S4_VFMT_MASK; - i915->meta.Ctx[I915_CTXREG_LIS4] |= - (S4_VFMT_COLOR | - S4_VFMT_SPEC_FOG | - S4_VFMT_XYZW); + i915->meta.Ctx[I915_CTXREG_LIS4] |= (S4_VFMT_COLOR | S4_VFMT_XYZ); i915->meta.emitted &= ~I915_UPLOAD_CTX; - } -static void draw_quad(i915ContextPtr i915, - GLfloat x0, GLfloat x1, - GLfloat y0, GLfloat y1, - GLubyte red, GLubyte green, - GLubyte blue, GLubyte alpha, - GLfloat s0, GLfloat s1, - GLfloat t0, GLfloat t1 ) -{ - GLuint vertex_size = 8; - GLuint *vb = intelEmitInlinePrimitiveLocked( &i915->intel, - PRIM3D_TRIFAN, - 4 * vertex_size, - vertex_size ); - intelVertex tmp; - int i; - - if (0) - fprintf(stderr, "%s: %f,%f-%f,%f 0x%x%x%x%x %f,%f-%f,%f\n", - __FUNCTION__, - x0,y0,x1,y1,red,green,blue,alpha,s0,t0,s1,t1); - - - /* initial vertex, left bottom */ - tmp.v.x = x0; - tmp.v.y = y0; - tmp.v.z = 1.0; - tmp.v.w = 1.0; - tmp.v.color.red = red; - tmp.v.color.green = green; - tmp.v.color.blue = blue; - tmp.v.color.alpha = alpha; - tmp.v.specular.red = 0; - tmp.v.specular.green = 0; - tmp.v.specular.blue = 0; - tmp.v.specular.alpha = 0; - tmp.v.u0 = s0; - tmp.v.v0 = t0; - - for (i = 0 ; i < vertex_size ; i++) - vb[i] = tmp.ui[i]; - - /* right bottom */ - vb += vertex_size; - tmp.v.x = x1; - tmp.v.u0 = s1; - for (i = 0 ; i < vertex_size ; i++) - vb[i] = tmp.ui[i]; - - /* right top */ - vb += vertex_size; - tmp.v.y = y1; - tmp.v.v0 = t1; - for (i = 0 ; i < vertex_size ; i++) - vb[i] = tmp.ui[i]; - - /* left top */ - vb += vertex_size; - tmp.v.x = x0; - tmp.v.u0 = s0; - for (i = 0 ; i < vertex_size ; i++) - vb[i] = tmp.ui[i]; -} - -static void draw_poly(i915ContextPtr i915, - GLubyte red, GLubyte green, GLubyte blue, GLubyte alpha, - GLuint numVerts, - /*const*/ GLfloat verts[][2], - /*const*/ GLfloat texcoords[][2]) +/* Operations where the 3D engine is decoupled temporarily from the + * current GL state and used for other purposes than simply rendering + * incoming triangles. + */ +static void +install_meta_state(struct intel_context *intel) { - GLuint vertex_size = 8; - GLuint *vb = intelEmitInlinePrimitiveLocked( &i915->intel, - PRIM3D_TRIFAN, - numVerts * vertex_size, - vertex_size ); - intelVertex tmp; - int i, k; - - /* initial constant vertex fields */ - tmp.v.z = 1.0; - tmp.v.w = 1.0; - tmp.v.color.red = red; - tmp.v.color.green = green; - tmp.v.color.blue = blue; - tmp.v.color.alpha = alpha; - tmp.v.specular.red = 0; - tmp.v.specular.green = 0; - tmp.v.specular.blue = 0; - tmp.v.specular.alpha = 0; - - for (k = 0; k < numVerts; k++) { - tmp.v.x = verts[k][0]; - tmp.v.y = verts[k][1]; - tmp.v.u0 = texcoords[k][0]; - tmp.v.v0 = texcoords[k][1]; - - for (i = 0 ; i < vertex_size ; i++) - vb[i] = tmp.ui[i]; - - vb += vertex_size; - } -} + struct i915_context *i915 = i915_context(&intel->ctx); + memcpy(&i915->meta, &i915->initial, sizeof(i915->meta)); + i915->meta.active = ACTIVE; + i915->meta.emitted = 0; + SET_STATE(i915, meta); + set_vertex_format(intel); + meta_no_texture(intel); +} -void -i915ClearWithTris(intelContextPtr intel, GLbitfield buffers, - GLboolean allFoo, - GLint cxFoo, GLint cyFoo, GLint cwFoo, GLint chFoo) +static void +leave_meta_state(struct intel_context *intel) { - i915ContextPtr i915 = I915_CONTEXT( intel ); - __DRIdrawablePrivate *dPriv = intel->driDrawable; - intelScreenPrivate *screen = intel->intelScreen; - int x0, y0, x1, y1; - GLint cx, cy, cw, ch; - GLboolean all; - - SET_STATE( i915, meta ); - set_initial_state( i915 ); - set_no_texture( i915 ); - set_vertex_format( i915 ); - - LOCK_HARDWARE(intel); - - /* get clear bounds after locking */ - cx = intel->ctx.DrawBuffer->_Xmin; - cy = intel->ctx.DrawBuffer->_Ymin; - cw = intel->ctx.DrawBuffer->_Xmax - cx; - ch = intel->ctx.DrawBuffer->_Ymax - cy; - all = (cw == intel->ctx.DrawBuffer->Width && - ch == intel->ctx.DrawBuffer->Height); - - if (!all) { - x0 = cx; - y0 = cy; - x1 = x0 + cw; - y1 = y0 + ch; - } else { - x0 = 0; - y0 = 0; - x1 = x0 + dPriv->w; - y1 = y0 + dPriv->h; - } - - /* Don't do any clipping to screen - these are window coordinates. - * The active cliprects will be applied as for any other geometry. - */ - - if (buffers & BUFFER_BIT_FRONT_LEFT) { - set_no_depth_stencil_write( i915 ); - set_color_mask( i915, GL_TRUE ); - set_draw_region( i915, &screen->front ); - - draw_quad(i915, x0, x1, y0, y1, - intel->clear_red, intel->clear_green, - intel->clear_blue, intel->clear_alpha, - 0, 0, 0, 0); - } - - if (buffers & BUFFER_BIT_BACK_LEFT) { - set_no_depth_stencil_write( i915 ); - set_color_mask( i915, GL_TRUE ); - set_draw_region( i915, &screen->back ); - - draw_quad(i915, x0, x1, y0, y1, - intel->clear_red, intel->clear_green, - intel->clear_blue, intel->clear_alpha, - 0, 0, 0, 0); - } - - if (buffers & BUFFER_BIT_STENCIL) { - set_stencil_replace( i915, - intel->ctx.Stencil.WriteMask[0], - intel->ctx.Stencil.Clear); - - set_color_mask( i915, GL_FALSE ); - set_draw_region( i915, &screen->front ); /* could be either? */ - - draw_quad( i915, x0, x1, y0, y1, 0, 0, 0, 0, 0, 0, 0, 0 ); - } - - UNLOCK_HARDWARE(intel); - - SET_STATE( i915, state ); + struct i915_context *i915 = i915_context(&intel->ctx); + intel_region_release(&i915->meta.draw_region); + intel_region_release(&i915->meta.depth_region); +/* intel_region_release(&i915->meta.tex_region[0]); */ + SET_STATE(i915, state); } -/** - * Copy the window contents named by dPriv to the rotated (or reflected) - * color buffer. - * srcBuf is BUFFER_BIT_FRONT_LEFT or BUFFER_BIT_BACK_LEFT to indicate the source. - */ + void -i915RotateWindow(intelContextPtr intel, __DRIdrawablePrivate *dPriv, - GLuint srcBuf) +i915InitMetaFuncs(struct i915_context *i915) { - i915ContextPtr i915 = I915_CONTEXT( intel ); - intelScreenPrivate *screen = intel->intelScreen; - const GLuint cpp = screen->cpp; - drm_clip_rect_t fullRect; - GLuint textureFormat, srcOffset, srcPitch; - const drm_clip_rect_t *clipRects; - int numClipRects; - int i; - - int xOrig, yOrig; - int origNumClipRects; - drm_clip_rect_t *origRects; - - /* - * set up hardware state - */ - intelFlush( &intel->ctx ); - - SET_STATE( i915, meta ); - set_initial_state( i915 ); - set_no_texture( i915 ); - set_vertex_format( i915 ); - set_no_depth_stencil_write( i915 ); - set_color_mask( i915, GL_TRUE ); - - LOCK_HARDWARE(intel); - - /* save current drawing origin and cliprects (restored at end) */ - xOrig = intel->drawX; - yOrig = intel->drawY; - origNumClipRects = intel->numClipRects; - origRects = intel->pClipRects; - - if (!intel->numClipRects) - goto done; - - /* - * set drawing origin, cliprects for full-screen access to rotated screen - */ - fullRect.x1 = 0; - fullRect.y1 = 0; - fullRect.x2 = screen->rotatedWidth; - fullRect.y2 = screen->rotatedHeight; - intel->drawX = 0; - intel->drawY = 0; - intel->numClipRects = 1; - intel->pClipRects = &fullRect; - - set_draw_region( i915, &screen->rotated ); - - if (cpp == 4) - textureFormat = MAPSURF_32BIT | MT_32BIT_ARGB8888; - else - textureFormat = MAPSURF_16BIT | MT_16BIT_RGB565; - - if (srcBuf == BUFFER_BIT_FRONT_LEFT) { - srcPitch = screen->front.pitch; /* in bytes */ - srcOffset = screen->front.offset; /* bytes */ - clipRects = dPriv->pClipRects; - numClipRects = dPriv->numClipRects; - } - else { - srcPitch = screen->back.pitch; /* in bytes */ - srcOffset = screen->back.offset; /* bytes */ - clipRects = dPriv->pBackClipRects; - numClipRects = dPriv->numBackClipRects; - } - - /* set the whole screen up as a texture to avoid alignment issues */ - set_tex_rect_source(i915, - srcOffset, - screen->width, - screen->height, - srcPitch, - textureFormat); - - enable_texture_blend_replace(i915); - - /* - * loop over the source window's cliprects - */ - for (i = 0; i < numClipRects; i++) { - int srcX0 = clipRects[i].x1; - int srcY0 = clipRects[i].y1; - int srcX1 = clipRects[i].x2; - int srcY1 = clipRects[i].y2; - GLfloat verts[4][2], tex[4][2]; - int j; - - /* build vertices for four corners of clip rect */ - verts[0][0] = srcX0; verts[0][1] = srcY0; - verts[1][0] = srcX1; verts[1][1] = srcY0; - verts[2][0] = srcX1; verts[2][1] = srcY1; - verts[3][0] = srcX0; verts[3][1] = srcY1; - - /* .. and texcoords */ - tex[0][0] = srcX0; tex[0][1] = srcY0; - tex[1][0] = srcX1; tex[1][1] = srcY0; - tex[2][0] = srcX1; tex[2][1] = srcY1; - tex[3][0] = srcX0; tex[3][1] = srcY1; - - /* transform coords to rotated screen coords */ - for (j = 0; j < 4; j++) { - matrix23TransformCoordf(&screen->rotMatrix, - &verts[j][0], &verts[j][1]); - } - - /* draw polygon to map source image to dest region */ - draw_poly(i915, 255, 255, 255, 255, 4, verts, tex); - - } /* cliprect loop */ - - intelFlushBatchLocked( intel, GL_FALSE, GL_FALSE, GL_FALSE ); - - done: - /* restore original drawing origin and cliprects */ - intel->drawX = xOrig; - intel->drawY = yOrig; - intel->numClipRects = origNumClipRects; - intel->pClipRects = origRects; - - UNLOCK_HARDWARE(intel); - - SET_STATE( i915, state ); + i915->intel.vtbl.install_meta_state = install_meta_state; + i915->intel.vtbl.leave_meta_state = leave_meta_state; + i915->intel.vtbl.meta_no_depth_write = meta_no_depth_write; + i915->intel.vtbl.meta_no_stencil_write = meta_no_stencil_write; + i915->intel.vtbl.meta_stencil_replace = meta_stencil_replace; + i915->intel.vtbl.meta_depth_replace = meta_depth_replace; + i915->intel.vtbl.meta_color_mask = meta_color_mask; + i915->intel.vtbl.meta_no_texture = meta_no_texture; + i915->intel.vtbl.meta_texture_blend_replace = meta_texture_blend_replace; + i915->intel.vtbl.meta_tex_rect_source = meta_tex_rect_source; + i915->intel.vtbl.meta_draw_region = meta_draw_region; + i915->intel.vtbl.meta_import_pixel_state = meta_import_pixel_state; } - diff --git a/src/mesa/drivers/dri/i915/i915_program.c b/src/mesa/drivers/dri/i915/i915_program.c index 6849112444..c6c6434023 100644 --- a/src/mesa/drivers/dri/i915/i915_program.c +++ b/src/mesa/drivers/dri/i915/i915_program.c @@ -72,58 +72,62 @@ #define I915_CONSTFLAG_PARAM 0x1f -GLuint i915_get_temp( struct i915_fragment_program *p ) +GLuint +i915_get_temp(struct i915_fragment_program *p) { - int bit = ffs( ~p->temp_flag ); + int bit = ffs(~p->temp_flag); if (!bit) { fprintf(stderr, "%s: out of temporaries\n", __FILE__); exit(1); } - p->temp_flag |= 1<<(bit-1); - return UREG(REG_TYPE_R, (bit-1)); + p->temp_flag |= 1 << (bit - 1); + return UREG(REG_TYPE_R, (bit - 1)); } -GLuint i915_get_utemp( struct i915_fragment_program *p ) +GLuint +i915_get_utemp(struct i915_fragment_program * p) { - int bit = ffs( ~p->utemp_flag ); + int bit = ffs(~p->utemp_flag); if (!bit) { fprintf(stderr, "%s: out of temporaries\n", __FILE__); exit(1); } - p->utemp_flag |= 1<<(bit-1); - return UREG(REG_TYPE_U, (bit-1)); + p->utemp_flag |= 1 << (bit - 1); + return UREG(REG_TYPE_U, (bit - 1)); } -void i915_release_utemps( struct i915_fragment_program *p ) +void +i915_release_utemps(struct i915_fragment_program *p) { p->utemp_flag = ~0x7; } -GLuint i915_emit_decl( struct i915_fragment_program *p, - GLuint type, GLuint nr, GLuint d0_flags ) +GLuint +i915_emit_decl(struct i915_fragment_program *p, + GLuint type, GLuint nr, GLuint d0_flags) { GLuint reg = UREG(type, nr); if (type == REG_TYPE_T) { - if (p->decl_t & (1<<nr)) - return reg; + if (p->decl_t & (1 << nr)) + return reg; - p->decl_t |= (1<<nr); + p->decl_t |= (1 << nr); } else if (type == REG_TYPE_S) { - if (p->decl_s & (1<<nr)) - return reg; + if (p->decl_s & (1 << nr)) + return reg; - p->decl_s |= (1<<nr); + p->decl_s |= (1 << nr); } - else + else return reg; - *(p->decl++) = (D0_DCL | D0_DEST( reg ) | d0_flags); + *(p->decl++) = (D0_DCL | D0_DEST(reg) | d0_flags); *(p->decl++) = D1_MBZ; *(p->decl++) = D2_MBZ; @@ -131,24 +135,26 @@ GLuint i915_emit_decl( struct i915_fragment_program *p, return reg; } -GLuint i915_emit_arith( struct i915_fragment_program *p, - GLuint op, - GLuint dest, - GLuint mask, - GLuint saturate, - GLuint src0, - GLuint src1, - GLuint src2 ) +GLuint +i915_emit_arith(struct i915_fragment_program * p, + GLuint op, + GLuint dest, + GLuint mask, + GLuint saturate, GLuint src0, GLuint src1, GLuint src2) { GLuint c[3]; GLuint nr_const = 0; assert(GET_UREG_TYPE(dest) != REG_TYPE_CONST); - assert(dest = UREG(GET_UREG_TYPE(dest), GET_UREG_NR(dest))); + dest = UREG(GET_UREG_TYPE(dest), GET_UREG_NR(dest)); + assert(dest); - if (GET_UREG_TYPE(src0) == REG_TYPE_CONST) c[nr_const++] = 0; - if (GET_UREG_TYPE(src1) == REG_TYPE_CONST) c[nr_const++] = 1; - if (GET_UREG_TYPE(src2) == REG_TYPE_CONST) c[nr_const++] = 2; + if (GET_UREG_TYPE(src0) == REG_TYPE_CONST) + c[nr_const++] = 0; + if (GET_UREG_TYPE(src1) == REG_TYPE_CONST) + c[nr_const++] = 1; + if (GET_UREG_TYPE(src2) == REG_TYPE_CONST) + c[nr_const++] = 2; /* Recursively call this function to MOV additional const values * into temporary registers. Use utemp registers for this - @@ -164,31 +170,25 @@ GLuint i915_emit_arith( struct i915_fragment_program *p, old_utemp_flag = p->utemp_flag; first = GET_UREG_NR(s[c[0]]); - for (i = 1 ; i < nr_const ; i++) { - if (GET_UREG_NR(s[c[i]]) != first) { - GLuint tmp = i915_get_utemp(p); - - i915_emit_arith( p, A0_MOV, tmp, A0_DEST_CHANNEL_ALL, 0, - s[c[i]], 0, 0 ); - s[c[i]] = tmp; - } + for (i = 1; i < nr_const; i++) { + if (GET_UREG_NR(s[c[i]]) != first) { + GLuint tmp = i915_get_utemp(p); + + i915_emit_arith(p, A0_MOV, tmp, A0_DEST_CHANNEL_ALL, 0, + s[c[i]], 0, 0); + s[c[i]] = tmp; + } } src0 = s[0]; src1 = s[1]; src2 = s[2]; - p->utemp_flag = old_utemp_flag; /* restore */ + p->utemp_flag = old_utemp_flag; /* restore */ } - *(p->csr++) = (op | - A0_DEST( dest ) | - mask | - saturate | - A0_SRC0( src0 )); - *(p->csr++) = (A1_SRC0( src0 ) | - A1_SRC1( src1 )); - *(p->csr++) = (A2_SRC1( src1 ) | - A2_SRC2( src2 )); + *(p->csr++) = (op | A0_DEST(dest) | mask | saturate | A0_SRC0(src0)); + *(p->csr++) = (A1_SRC0(src0) | A1_SRC1(src1)); + *(p->csr++) = (A2_SRC1(src1) | A2_SRC2(src2)); p->nr_alu_insn++; return dest; @@ -239,24 +239,28 @@ GLuint i915_emit_texld( struct i915_fragment_program *p, } -GLuint i915_emit_const1f( struct i915_fragment_program *p, GLfloat c0 ) +GLuint +i915_emit_const1f(struct i915_fragment_program * p, GLfloat c0) { GLint reg, idx; - if (c0 == 0.0) return swizzle(UREG(REG_TYPE_R, 0), ZERO, ZERO, ZERO, ZERO); - if (c0 == 1.0) return swizzle(UREG(REG_TYPE_R, 0), ONE, ONE, ONE, ONE ); + if (c0 == 0.0) + return swizzle(UREG(REG_TYPE_R, 0), ZERO, ZERO, ZERO, ZERO); + if (c0 == 1.0) + return swizzle(UREG(REG_TYPE_R, 0), ONE, ONE, ONE, ONE); for (reg = 0; reg < I915_MAX_CONSTANT; reg++) { if (p->constant_flags[reg] == I915_CONSTFLAG_PARAM) - continue; + continue; for (idx = 0; idx < 4; idx++) { - if (!(p->constant_flags[reg] & (1<<idx)) || - p->constant[reg][idx] == c0) { - p->constant[reg][idx] = c0; - p->constant_flags[reg] |= 1<<idx; - if (reg+1 > p->nr_constants) p->nr_constants = reg+1; - return swizzle(UREG(REG_TYPE_CONST, reg),idx,ZERO,ZERO,ONE); - } + if (!(p->constant_flags[reg] & (1 << idx)) || + p->constant[reg][idx] == c0) { + p->constant[reg][idx] = c0; + p->constant_flags[reg] |= 1 << idx; + if (reg + 1 > p->nr_constants) + p->nr_constants = reg + 1; + return swizzle(UREG(REG_TYPE_CONST, reg), idx, ZERO, ZERO, ONE); + } } } @@ -265,29 +269,35 @@ GLuint i915_emit_const1f( struct i915_fragment_program *p, GLfloat c0 ) return 0; } -GLuint i915_emit_const2f( struct i915_fragment_program *p, - GLfloat c0, GLfloat c1 ) +GLuint +i915_emit_const2f(struct i915_fragment_program * p, GLfloat c0, GLfloat c1) { GLint reg, idx; - if (c0 == 0.0) return swizzle(i915_emit_const1f(p, c1), ZERO, X, Z, W); - if (c0 == 1.0) return swizzle(i915_emit_const1f(p, c1), ONE, X, Z, W); + if (c0 == 0.0) + return swizzle(i915_emit_const1f(p, c1), ZERO, X, Z, W); + if (c0 == 1.0) + return swizzle(i915_emit_const1f(p, c1), ONE, X, Z, W); - if (c1 == 0.0) return swizzle(i915_emit_const1f(p, c0), X, ZERO, Z, W); - if (c1 == 1.0) return swizzle(i915_emit_const1f(p, c0), X, ONE, Z, W); + if (c1 == 0.0) + return swizzle(i915_emit_const1f(p, c0), X, ZERO, Z, W); + if (c1 == 1.0) + return swizzle(i915_emit_const1f(p, c0), X, ONE, Z, W); for (reg = 0; reg < I915_MAX_CONSTANT; reg++) { if (p->constant_flags[reg] == 0xf || - p->constant_flags[reg] == I915_CONSTFLAG_PARAM) - continue; + p->constant_flags[reg] == I915_CONSTFLAG_PARAM) + continue; for (idx = 0; idx < 3; idx++) { - if (!(p->constant_flags[reg] & (3<<idx))) { - p->constant[reg][idx] = c0; - p->constant[reg][idx+1] = c1; - p->constant_flags[reg] |= 3<<idx; - if (reg+1 > p->nr_constants) p->nr_constants = reg+1; - return swizzle(UREG(REG_TYPE_CONST, reg),idx,idx+1,ZERO,ONE); - } + if (!(p->constant_flags[reg] & (3 << idx))) { + p->constant[reg][idx] = c0; + p->constant[reg][idx + 1] = c1; + p->constant_flags[reg] |= 3 << idx; + if (reg + 1 > p->nr_constants) + p->nr_constants = reg + 1; + return swizzle(UREG(REG_TYPE_CONST, reg), idx, idx + 1, ZERO, + ONE); + } } } @@ -298,27 +308,28 @@ GLuint i915_emit_const2f( struct i915_fragment_program *p, -GLuint i915_emit_const4f( struct i915_fragment_program *p, - GLfloat c0, GLfloat c1, GLfloat c2, GLfloat c3 ) +GLuint +i915_emit_const4f(struct i915_fragment_program * p, + GLfloat c0, GLfloat c1, GLfloat c2, GLfloat c3) { GLint reg; for (reg = 0; reg < I915_MAX_CONSTANT; reg++) { if (p->constant_flags[reg] == 0xf && - p->constant[reg][0] == c0 && - p->constant[reg][1] == c1 && - p->constant[reg][2] == c2 && - p->constant[reg][3] == c3) { - return UREG(REG_TYPE_CONST, reg); + p->constant[reg][0] == c0 && + p->constant[reg][1] == c1 && + p->constant[reg][2] == c2 && p->constant[reg][3] == c3) { + return UREG(REG_TYPE_CONST, reg); } else if (p->constant_flags[reg] == 0) { - p->constant[reg][0] = c0; - p->constant[reg][1] = c1; - p->constant[reg][2] = c2; - p->constant[reg][3] = c3; - p->constant_flags[reg] = 0xf; - if (reg+1 > p->nr_constants) p->nr_constants = reg+1; - return UREG(REG_TYPE_CONST, reg); + p->constant[reg][0] = c0; + p->constant[reg][1] = c1; + p->constant[reg][2] = c2; + p->constant[reg][3] = c3; + p->constant_flags[reg] = 0xf; + if (reg + 1 > p->nr_constants) + p->nr_constants = reg + 1; + return UREG(REG_TYPE_CONST, reg); } } @@ -328,34 +339,36 @@ GLuint i915_emit_const4f( struct i915_fragment_program *p, } -GLuint i915_emit_const4fv( struct i915_fragment_program *p, const GLfloat *c ) +GLuint +i915_emit_const4fv(struct i915_fragment_program * p, const GLfloat * c) { - return i915_emit_const4f( p, c[0], c[1], c[2], c[3] ); + return i915_emit_const4f(p, c[0], c[1], c[2], c[3]); } -GLuint i915_emit_param4fv( struct i915_fragment_program *p, - const GLfloat *values ) +GLuint +i915_emit_param4fv(struct i915_fragment_program * p, const GLfloat * values) { GLint reg, i; for (i = 0; i < p->nr_params; i++) { if (p->param[i].values == values) - return UREG(REG_TYPE_CONST, p->param[i].reg); + return UREG(REG_TYPE_CONST, p->param[i].reg); } for (reg = 0; reg < I915_MAX_CONSTANT; reg++) { if (p->constant_flags[reg] == 0) { - p->constant_flags[reg] = I915_CONSTFLAG_PARAM; - i = p->nr_params++; + p->constant_flags[reg] = I915_CONSTFLAG_PARAM; + i = p->nr_params++; - p->param[i].values = values; - p->param[i].reg = reg; - p->params_uptodate = 0; + p->param[i].values = values; + p->param[i].reg = reg; + p->params_uptodate = 0; - if (reg+1 > p->nr_constants) p->nr_constants = reg+1; - return UREG(REG_TYPE_CONST, reg); + if (reg + 1 > p->nr_constants) + p->nr_constants = reg + 1; + return UREG(REG_TYPE_CONST, reg); } } @@ -366,30 +379,31 @@ GLuint i915_emit_param4fv( struct i915_fragment_program *p, - -void i915_program_error( struct i915_fragment_program *p, const char *msg ) +void +i915_program_error(struct i915_fragment_program *p, const char *msg) { _mesa_problem(NULL, "i915_program_error: %s", msg); p->error = 1; } -void i915_init_program( i915ContextPtr i915, struct i915_fragment_program *p ) + +void +i915_init_program(struct i915_context *i915, struct i915_fragment_program *p) { GLcontext *ctx = &i915->intel.ctx; - TNLcontext *tnl = TNL_CONTEXT( ctx ); - + p->translated = 0; p->params_uptodate = 0; p->on_hardware = 0; p->error = 0; - p->nr_tex_indirect = 1; /* correct? */ + p->nr_tex_indirect = 1; /* correct? */ p->nr_tex_insn = 0; p->nr_alu_insn = 0; p->nr_decl_insn = 0; - p->ctx = ctx; - memset( p->constant_flags, 0, sizeof(p->constant_flags) ); + p->ctx = ctx; + memset(p->constant_flags, 0, sizeof(p->constant_flags)); p->nr_constants = 0; p->csr = p->program; @@ -402,21 +416,17 @@ void i915_init_program( i915ContextPtr i915, struct i915_fragment_program *p ) p->depth_written = 0; p->nr_params = 0; - p->src_texture = UREG_BAD; - p->src_previous = UREG(REG_TYPE_T, T_DIFFUSE); - p->last_tex_stage = 0; - p->VB = &tnl->vb; - *(p->decl++) = _3DSTATE_PIXEL_SHADER_PROGRAM; } -void i915_fini_program( struct i915_fragment_program *p ) +void +i915_fini_program(struct i915_fragment_program *p) { GLuint program_size = p->csr - p->program; GLuint decl_size = p->decl - p->declarations; - - if (p->nr_tex_indirect > I915_MAX_TEX_INDIRECT) + + if (p->nr_tex_indirect > I915_MAX_TEX_INDIRECT) i915_program_error(p, "Exceeded max nr indirect texture lookups"); if (p->nr_tex_insn > I915_MAX_TEX_INSN) @@ -446,22 +456,24 @@ void i915_fini_program( struct i915_fragment_program *p ) p->declarations[0] |= program_size + decl_size - 2; } -void i915_upload_program( i915ContextPtr i915, struct i915_fragment_program *p ) +void +i915_upload_program(struct i915_context *i915, + struct i915_fragment_program *p) { GLuint program_size = p->csr - p->program; GLuint decl_size = p->decl - p->declarations; - FALLBACK( &i915->intel, I915_FALLBACK_PROGRAM, p->error ); + FALLBACK(&i915->intel, I915_FALLBACK_PROGRAM, p->error); /* Could just go straight to the batchbuffer from here: */ if (i915->state.ProgramSize != (program_size + decl_size) || - memcmp(i915->state.Program + decl_size, p->program, - program_size*sizeof(int)) != 0) { - I915_STATECHANGE( i915, I915_UPLOAD_PROGRAM ); - memcpy(i915->state.Program, p->declarations, decl_size*sizeof(int)); + memcmp(i915->state.Program + decl_size, p->program, + program_size * sizeof(int)) != 0) { + I915_STATECHANGE(i915, I915_UPLOAD_PROGRAM); + memcpy(i915->state.Program, p->declarations, decl_size * sizeof(int)); memcpy(i915->state.Program + decl_size, p->program, - program_size*sizeof(int)); + program_size * sizeof(int)); i915->state.ProgramSize = decl_size + program_size; } @@ -470,30 +482,28 @@ void i915_upload_program( i915ContextPtr i915, struct i915_fragment_program *p ) */ if (p->nr_constants) { GLuint nr = p->nr_constants; - - I915_ACTIVESTATE( i915, I915_UPLOAD_CONSTANTS, 1 ); - I915_STATECHANGE( i915, I915_UPLOAD_CONSTANTS ); + + I915_ACTIVESTATE(i915, I915_UPLOAD_CONSTANTS, 1); + I915_STATECHANGE(i915, I915_UPLOAD_CONSTANTS); i915->state.Constant[0] = _3DSTATE_PIXEL_SHADER_CONSTANTS | ((nr) * 4); - i915->state.Constant[1] = (1<<(nr-1)) | ((1<<(nr-1))-1); - - memcpy(&i915->state.Constant[2], p->constant, 4*sizeof(int)*(nr)); + i915->state.Constant[1] = (1 << (nr - 1)) | ((1 << (nr - 1)) - 1); + + memcpy(&i915->state.Constant[2], p->constant, 4 * sizeof(int) * (nr)); i915->state.ConstantSize = 2 + (nr) * 4; if (0) { - GLuint i; - for (i = 0; i < nr; i++) { - fprintf(stderr, "const[%d]: %f %f %f %f\n", i, - p->constant[i][0], - p->constant[i][1], - p->constant[i][2], - p->constant[i][3]); - } + GLuint i; + for (i = 0; i < nr; i++) { + fprintf(stderr, "const[%d]: %f %f %f %f\n", i, + p->constant[i][0], + p->constant[i][1], p->constant[i][2], p->constant[i][3]); + } } } else { - I915_ACTIVESTATE( i915, I915_UPLOAD_CONSTANTS, 0 ); - } + I915_ACTIVESTATE(i915, I915_UPLOAD_CONSTANTS, 0); + } p->on_hardware = 1; } diff --git a/src/mesa/drivers/dri/i915/i915_program.h b/src/mesa/drivers/dri/i915/i915_program.h index 8891a17785..3c12b34f16 100644 --- a/src/mesa/drivers/dri/i915/i915_program.h +++ b/src/mesa/drivers/dri/i915/i915_program.h @@ -48,11 +48,11 @@ #define UREG_CHANNEL_W_NEGATE_SHIFT 11 #define UREG_CHANNEL_W_SHIFT 8 #define UREG_CHANNEL_ZERO_NEGATE_MBZ 5 -#define UREG_CHANNEL_ZERO_SHIFT 4 +#define UREG_CHANNEL_ZERO_SHIFT 4 #define UREG_CHANNEL_ONE_NEGATE_MBZ 1 -#define UREG_CHANNEL_ONE_SHIFT 0 +#define UREG_CHANNEL_ONE_SHIFT 0 -#define UREG_BAD 0xffffffff /* not a valid ureg */ +#define UREG_BAD 0xffffffff /* not a valid ureg */ #define X SRC_X #define Y SRC_Y @@ -84,78 +84,75 @@ /* One neat thing about the UREG representation: */ -static __inline int swizzle( int reg, int x, int y, int z, int w ) +static INLINE int +swizzle(int reg, int x, int y, int z, int w) { return ((reg & ~UREG_XYZW_CHANNEL_MASK) | - CHANNEL_SRC( GET_CHANNEL_SRC( reg, x ), 0 ) | - CHANNEL_SRC( GET_CHANNEL_SRC( reg, y ), 1 ) | - CHANNEL_SRC( GET_CHANNEL_SRC( reg, z ), 2 ) | - CHANNEL_SRC( GET_CHANNEL_SRC( reg, w ), 3 )); + CHANNEL_SRC(GET_CHANNEL_SRC(reg, x), 0) | + CHANNEL_SRC(GET_CHANNEL_SRC(reg, y), 1) | + CHANNEL_SRC(GET_CHANNEL_SRC(reg, z), 2) | + CHANNEL_SRC(GET_CHANNEL_SRC(reg, w), 3)); } /* Another neat thing about the UREG representation: */ -static __inline int negate( int reg, int x, int y, int z, int w ) +static INLINE int +negate(int reg, int x, int y, int z, int w) { - return reg ^ (((x&1)<<UREG_CHANNEL_X_NEGATE_SHIFT)| - ((y&1)<<UREG_CHANNEL_Y_NEGATE_SHIFT)| - ((z&1)<<UREG_CHANNEL_Z_NEGATE_SHIFT)| - ((w&1)<<UREG_CHANNEL_W_NEGATE_SHIFT)); + return reg ^ (((x & 1) << UREG_CHANNEL_X_NEGATE_SHIFT) | + ((y & 1) << UREG_CHANNEL_Y_NEGATE_SHIFT) | + ((z & 1) << UREG_CHANNEL_Z_NEGATE_SHIFT) | + ((w & 1) << UREG_CHANNEL_W_NEGATE_SHIFT)); } -extern GLuint i915_get_temp( struct i915_fragment_program *p ); -extern GLuint i915_get_utemp( struct i915_fragment_program *p ); -extern void i915_release_utemps( struct i915_fragment_program *p ); +extern GLuint i915_get_temp(struct i915_fragment_program *p); +extern GLuint i915_get_utemp(struct i915_fragment_program *p); +extern void i915_release_utemps(struct i915_fragment_program *p); -extern GLuint i915_emit_texld( struct i915_fragment_program *p, - GLuint dest, - GLuint destmask, - GLuint sampler, - GLuint coord, - GLuint op ); +extern GLuint i915_emit_texld(struct i915_fragment_program *p, + GLuint dest, + GLuint destmask, + GLuint sampler, GLuint coord, GLuint op); -extern GLuint i915_emit_arith( struct i915_fragment_program *p, - GLuint op, - GLuint dest, - GLuint mask, - GLuint saturate, - GLuint src0, - GLuint src1, - GLuint src2 ); +extern GLuint i915_emit_arith(struct i915_fragment_program *p, + GLuint op, + GLuint dest, + GLuint mask, + GLuint saturate, + GLuint src0, GLuint src1, GLuint src2); -extern GLuint i915_emit_decl( struct i915_fragment_program *p, - GLuint type, GLuint nr, GLuint d0_flags ); +extern GLuint i915_emit_decl(struct i915_fragment_program *p, + GLuint type, GLuint nr, GLuint d0_flags); -extern GLuint i915_emit_const1f( struct i915_fragment_program *p, - GLfloat c0 ); +extern GLuint i915_emit_const1f(struct i915_fragment_program *p, GLfloat c0); -extern GLuint i915_emit_const2f( struct i915_fragment_program *p, - GLfloat c0, GLfloat c1 ); +extern GLuint i915_emit_const2f(struct i915_fragment_program *p, + GLfloat c0, GLfloat c1); -extern GLuint i915_emit_const4fv( struct i915_fragment_program *p, - const GLfloat *c ); +extern GLuint i915_emit_const4fv(struct i915_fragment_program *p, + const GLfloat * c); -extern GLuint i915_emit_const4f( struct i915_fragment_program *p, - GLfloat c0, GLfloat c1, - GLfloat c2, GLfloat c3 ); +extern GLuint i915_emit_const4f(struct i915_fragment_program *p, + GLfloat c0, GLfloat c1, + GLfloat c2, GLfloat c3); -extern GLuint i915_emit_param4fv( struct i915_fragment_program *p, - const GLfloat *values ); +extern GLuint i915_emit_param4fv(struct i915_fragment_program *p, + const GLfloat * values); -extern void i915_program_error( struct i915_fragment_program *p, - const char *msg ); +extern void i915_program_error(struct i915_fragment_program *p, + const char *msg); -extern void i915_init_program( i915ContextPtr i915, - struct i915_fragment_program *p ); +extern void i915_init_program(struct i915_context *i915, + struct i915_fragment_program *p); -extern void i915_upload_program( i915ContextPtr i915, - struct i915_fragment_program *p ); +extern void i915_upload_program(struct i915_context *i915, + struct i915_fragment_program *p); -extern void i915_fini_program( struct i915_fragment_program *p ); +extern void i915_fini_program(struct i915_fragment_program *p); diff --git a/src/mesa/drivers/dri/i915/i915_reg.h b/src/mesa/drivers/dri/i915/i915_reg.h index 694cd4c8c3..b5585e70e7 100644 --- a/src/mesa/drivers/dri/i915/i915_reg.h +++ b/src/mesa/drivers/dri/i915/i915_reg.h @@ -34,8 +34,6 @@ #define I915_SET_FIELD( var, mask, value ) (var &= ~(mask), var |= value) -#define CMD_3D (0x3<<29) - #define PRIM3D_INLINE (CMD_3D | (0x1f<<24)) #define PRIM3D_TRILIST (0x0<<18) #define PRIM3D_TRISTRIP (0x1<<18) @@ -112,6 +110,20 @@ /* 3DSTATE_CHROMA_KEY */ /* 3DSTATE_CLEAR_PARAMETERS, p150 */ +/* + * Sets the color, depth and stencil clear values used by the + * CLEAR_RECT and ZONE_INIT primitive types, respectively. These + * primitives set override most 3d state and only take a minimal x/y + * vertex. The color/z/stencil information is supplied here and + * therefore cannot vary per vertex. + */ +#define _3DSTATE_CLEAR_PARAMETERS (CMD_3D | (0x1d<<24) | (0x9c<<16) | 5) +/* Dword 1 */ +#define CLEARPARAM_CLEAR_RECT (1 << 16) +#define CLEARPARAM_ZONE_INIT (0 << 16) +#define CLEARPARAM_WRITE_COLOR (1 << 2) +#define CLEARPARAM_WRITE_DEPTH (1 << 1) +#define CLEARPARAM_WRITE_STENCIL (1 << 0) /* 3DSTATE_CONSTANT_BLEND_COLOR, p153 */ #define _3DSTATE_CONST_BLEND_COLOR_CMD (CMD_3D | (0x1d<<24) | (0x88<<16)) @@ -424,9 +436,22 @@ #define S7_DEPTH_OFFSET_CONST_MASK ~0 + +/* Helper macros for blend factors + */ +#define DST_BLND_FACT(f) ((f)<<S6_CBUF_DST_BLEND_FACT_SHIFT) +#define SRC_BLND_FACT(f) ((f)<<S6_CBUF_SRC_BLEND_FACT_SHIFT) +#define DST_ABLND_FACT(f) ((f)<<IAB_DST_FACTOR_SHIFT) +#define SRC_ABLND_FACT(f) ((f)<<IAB_SRC_FACTOR_SHIFT) + + + + /* 3DSTATE_MAP_DEINTERLACER_PARAMETERS */ -/* 3DSTATE_MAP_PALETTE_LOAD_32, p206 */ +/* 3DSTATE_MAP_PALETTE_LOAD_32, p206 */ +#define _3DSTATE_MAP_PALETTE_LOAD_32 (CMD_3D|(0x1d<<24)|(0x8f<<16)) +/* subsequent dwords up to length (max 16) are ARGB8888 color values */ /* _3DSTATE_MODES_4, p218 */ #define _3DSTATE_MODES_4_CMD (CMD_3D|(0x0d<<24)) @@ -435,7 +460,7 @@ #define LOGICOP_MASK (0xf<<18) #define MODE4_ENABLE_STENCIL_TEST_MASK ((1<<17)|(0xff00)) #define ENABLE_STENCIL_TEST_MASK (1<<17) -#define STENCIL_TEST_MASK(x) ((x)<<8) +#define STENCIL_TEST_MASK(x) (((x)&0xff)<<8) #define MODE4_ENABLE_STENCIL_WRITE_MASK ((1<<16)|(0x00ff)) #define ENABLE_STENCIL_WRITE_MASK (1<<16) #define STENCIL_WRITE_MASK(x) ((x)&0xff) @@ -458,7 +483,7 @@ #define I915_MAX_TEX_INDIRECT 4 -#define I915_MAX_TEX_INSN 32 +#define I915_MAX_TEX_INSN 32 #define I915_MAX_ALU_INSN 64 #define I915_MAX_DECL_INSN 27 #define I915_MAX_TEMPORARY 16 @@ -470,33 +495,33 @@ */ #define _3DSTATE_PIXEL_SHADER_PROGRAM (CMD_3D|(0x1d<<24)|(0x5<<16)) -#define REG_TYPE_R 0 /* temporary regs, no need to - * dcl, must be written before - * read -- Preserved between - * phases. - */ -#define REG_TYPE_T 1 /* Interpolated values, must be - * dcl'ed before use. - * - * 0..7: texture coord, - * 8: diffuse spec, - * 9: specular color, - * 10: fog parameter in w. - */ -#define REG_TYPE_CONST 2 /* Restriction: only one const - * can be referenced per - * instruction, though it may be - * selected for multiple inputs. - * Constants not initialized - * default to zero. - */ -#define REG_TYPE_S 3 /* sampler */ -#define REG_TYPE_OC 4 /* output color (rgba) */ -#define REG_TYPE_OD 5 /* output depth (w), xyz are - * temporaries. If not written, - * interpolated depth is used? - */ -#define REG_TYPE_U 6 /* unpreserved temporaries */ +#define REG_TYPE_R 0 /* temporary regs, no need to + * dcl, must be written before + * read -- Preserved between + * phases. + */ +#define REG_TYPE_T 1 /* Interpolated values, must be + * dcl'ed before use. + * + * 0..7: texture coord, + * 8: diffuse spec, + * 9: specular color, + * 10: fog parameter in w. + */ +#define REG_TYPE_CONST 2 /* Restriction: only one const + * can be referenced per + * instruction, though it may be + * selected for multiple inputs. + * Constants not initialized + * default to zero. + */ +#define REG_TYPE_S 3 /* sampler */ +#define REG_TYPE_OC 4 /* output color (rgba) */ +#define REG_TYPE_OD 5 /* output depth (w), xyz are + * temporaries. If not written, + * interpolated depth is used? + */ +#define REG_TYPE_U 6 /* unpreserved temporaries */ #define REG_TYPE_MASK 0x7 #define REG_NR_MASK 0xf @@ -513,34 +538,34 @@ #define T_TEX7 7 #define T_DIFFUSE 8 #define T_SPECULAR 9 -#define T_FOG_W 10 /* interpolated fog is in W coord */ +#define T_FOG_W 10 /* interpolated fog is in W coord */ /* Arithmetic instructions */ /* .replicate_swizzle == selection and replication of a particular * scalar channel, ie., .xxxx, .yyyy, .zzzz or .wwww */ -#define A0_NOP (0x0<<24) /* no operation */ -#define A0_ADD (0x1<<24) /* dst = src0 + src1 */ -#define A0_MOV (0x2<<24) /* dst = src0 */ -#define A0_MUL (0x3<<24) /* dst = src0 * src1 */ -#define A0_MAD (0x4<<24) /* dst = src0 * src1 + src2 */ -#define A0_DP2ADD (0x5<<24) /* dst.xyzw = src0.xy dot src1.xy + src2.replicate_swizzle */ -#define A0_DP3 (0x6<<24) /* dst.xyzw = src0.xyz dot src1.xyz */ -#define A0_DP4 (0x7<<24) /* dst.xyzw = src0.xyzw dot src1.xyzw */ -#define A0_FRC (0x8<<24) /* dst = src0 - floor(src0) */ -#define A0_RCP (0x9<<24) /* dst.xyzw = 1/(src0.replicate_swizzle) */ -#define A0_RSQ (0xa<<24) /* dst.xyzw = 1/(sqrt(abs(src0.replicate_swizzle))) */ -#define A0_EXP (0xb<<24) /* dst.xyzw = exp2(src0.replicate_swizzle) */ -#define A0_LOG (0xc<<24) /* dst.xyzw = log2(abs(src0.replicate_swizzle)) */ -#define A0_CMP (0xd<<24) /* dst = (src0 >= 0.0) ? src1 : src2 */ -#define A0_MIN (0xe<<24) /* dst = (src0 < src1) ? src0 : src1 */ -#define A0_MAX (0xf<<24) /* dst = (src0 >= src1) ? src0 : src1 */ -#define A0_FLR (0x10<<24) /* dst = floor(src0) */ -#define A0_MOD (0x11<<24) /* dst = src0 fmod 1.0 */ -#define A0_TRC (0x12<<24) /* dst = int(src0) */ -#define A0_SGE (0x13<<24) /* dst = src0 >= src1 ? 1.0 : 0.0 */ -#define A0_SLT (0x14<<24) /* dst = src0 < src1 ? 1.0 : 0.0 */ +#define A0_NOP (0x0<<24) /* no operation */ +#define A0_ADD (0x1<<24) /* dst = src0 + src1 */ +#define A0_MOV (0x2<<24) /* dst = src0 */ +#define A0_MUL (0x3<<24) /* dst = src0 * src1 */ +#define A0_MAD (0x4<<24) /* dst = src0 * src1 + src2 */ +#define A0_DP2ADD (0x5<<24) /* dst.xyzw = src0.xy dot src1.xy + src2.replicate_swizzle */ +#define A0_DP3 (0x6<<24) /* dst.xyzw = src0.xyz dot src1.xyz */ +#define A0_DP4 (0x7<<24) /* dst.xyzw = src0.xyzw dot src1.xyzw */ +#define A0_FRC (0x8<<24) /* dst = src0 - floor(src0) */ +#define A0_RCP (0x9<<24) /* dst.xyzw = 1/(src0.replicate_swizzle) */ +#define A0_RSQ (0xa<<24) /* dst.xyzw = 1/(sqrt(abs(src0.replicate_swizzle))) */ +#define A0_EXP (0xb<<24) /* dst.xyzw = exp2(src0.replicate_swizzle) */ +#define A0_LOG (0xc<<24) /* dst.xyzw = log2(abs(src0.replicate_swizzle)) */ +#define A0_CMP (0xd<<24) /* dst = (src0 >= 0.0) ? src1 : src2 */ +#define A0_MIN (0xe<<24) /* dst = (src0 < src1) ? src0 : src1 */ +#define A0_MAX (0xf<<24) /* dst = (src0 >= src1) ? src0 : src1 */ +#define A0_FLR (0x10<<24) /* dst = floor(src0) */ +#define A0_MOD (0x11<<24) /* dst = src0 fmod 1.0 */ +#define A0_TRC (0x12<<24) /* dst = int(src0) */ +#define A0_SGE (0x13<<24) /* dst = src0 >= src1 ? 1.0 : 0.0 */ +#define A0_SLT (0x14<<24) /* dst = src0 < src1 ? 1.0 : 0.0 */ #define A0_DEST_SATURATE (1<<22) #define A0_DEST_TYPE_SHIFT 19 /* Allow: R, OC, OD, U */ @@ -599,23 +624,23 @@ /* Texture instructions */ -#define T0_TEXLD (0x15<<24) /* Sample texture using predeclared - * sampler and address, and output - * filtered texel data to destination - * register */ -#define T0_TEXLDP (0x16<<24) /* Same as texld but performs a - * perspective divide of the texture - * coordinate .xyz values by .w before - * sampling. */ -#define T0_TEXLDB (0x17<<24) /* Same as texld but biases the - * computed LOD by w. Only S4.6 two's - * comp is used. This implies that a - * float to fixed conversion is - * done. */ -#define T0_TEXKILL (0x18<<24) /* Does not perform a sampling - * operation. Simply kills the pixel - * if any channel of the address - * register is < 0.0. */ +#define T0_TEXLD (0x15<<24) /* Sample texture using predeclared + * sampler and address, and output + * filtered texel data to destination + * register */ +#define T0_TEXLDP (0x16<<24) /* Same as texld but performs a + * perspective divide of the texture + * coordinate .xyz values by .w before + * sampling. */ +#define T0_TEXLDB (0x17<<24) /* Same as texld but biases the + * computed LOD by w. Only S4.6 two's + * comp is used. This implies that a + * float to fixed conversion is + * done. */ +#define T0_TEXKILL (0x18<<24) /* Does not perform a sampling + * operation. Simply kills the pixel + * if any channel of the address + * register is < 0.0. */ #define T0_DEST_TYPE_SHIFT 19 /* Allow: R, OC, OD, U */ /* Note: U (unpreserved) regs do not retain their values between @@ -627,18 +652,18 @@ */ #define T0_DEST_NR_SHIFT 14 /* Allow R: 0..15, OC,OD: 0..0, U: 0..2 */ -#define T0_SAMPLER_NR_SHIFT 0 /* This field ignored for TEXKILL */ +#define T0_SAMPLER_NR_SHIFT 0 /* This field ignored for TEXKILL */ #define T0_SAMPLER_NR_MASK (0xf<<0) -#define T1_ADDRESS_REG_TYPE_SHIFT 24 /* Reg to use as texture coord */ +#define T1_ADDRESS_REG_TYPE_SHIFT 24 /* Reg to use as texture coord */ /* Allow R, T, OC, OD -- R, OC, OD are 'dependent' reads, new program phase */ #define T1_ADDRESS_REG_NR_SHIFT 17 #define T2_MBZ 0 /* Declaration instructions */ -#define D0_DCL (0x19<<24) /* Declare a t (interpolated attrib) - * register or an s (sampler) - * register. */ +#define D0_DCL (0x19<<24) /* Declare a t (interpolated attrib) + * register or an s (sampler) + * register. */ #define D0_SAMPLE_TYPE_SHIFT 22 #define D0_SAMPLE_TYPE_2D (0x0<<22) #define D0_SAMPLE_TYPE_CUBE (0x1<<22) @@ -695,12 +720,12 @@ #define MAPSURF_4BIT_INDEXED (7<<7) #define MS3_MT_FORMAT_MASK (0x7 << 3) #define MS3_MT_FORMAT_SHIFT 3 -#define MT_4BIT_IDX_ARGB8888 (7<<3) /* SURFACE_4BIT_INDEXED */ -#define MT_8BIT_I8 (0<<3) /* SURFACE_8BIT */ +#define MT_4BIT_IDX_ARGB8888 (7<<3) /* SURFACE_4BIT_INDEXED */ +#define MT_8BIT_I8 (0<<3) /* SURFACE_8BIT */ #define MT_8BIT_L8 (1<<3) #define MT_8BIT_A8 (4<<3) #define MT_8BIT_MONO8 (5<<3) -#define MT_16BIT_RGB565 (0<<3) /* SURFACE_16BIT */ +#define MT_16BIT_RGB565 (0<<3) /* SURFACE_16BIT */ #define MT_16BIT_ARGB1555 (1<<3) #define MT_16BIT_ARGB4444 (2<<3) #define MT_16BIT_AY88 (3<<3) @@ -709,7 +734,7 @@ #define MT_16BIT_I16 (7<<3) #define MT_16BIT_L16 (8<<3) #define MT_16BIT_A16 (9<<3) -#define MT_32BIT_ARGB8888 (0<<3) /* SURFACE_32BIT */ +#define MT_32BIT_ARGB8888 (0<<3) /* SURFACE_32BIT */ #define MT_32BIT_ABGR8888 (1<<3) #define MT_32BIT_XRGB8888 (2<<3) #define MT_32BIT_XBGR8888 (3<<3) @@ -725,11 +750,11 @@ #define MT_32BIT_xI824 (0xD<<3) #define MT_32BIT_xA824 (0xE<<3) #define MT_32BIT_xL824 (0xF<<3) -#define MT_422_YCRCB_SWAPY (0<<3) /* SURFACE_422 */ +#define MT_422_YCRCB_SWAPY (0<<3) /* SURFACE_422 */ #define MT_422_YCRCB_NORMAL (1<<3) #define MT_422_YCRCB_SWAPUV (2<<3) #define MT_422_YCRCB_SWAPUVY (3<<3) -#define MT_COMPRESS_DXT1 (0<<3) /* SURFACE_COMPRESSED */ +#define MT_COMPRESS_DXT1 (0<<3) /* SURFACE_COMPRESSED */ #define MT_COMPRESS_DXT2_3 (1<<3) #define MT_COMPRESS_DXT4_5 (2<<3) #define MT_COMPRESS_FXT1 (3<<3) @@ -751,7 +776,7 @@ #define MS4_MIP_LAYOUT_LEGACY (0<<8) #define MS4_MIP_LAYOUT_BELOW_LPT (0<<8) #define MS4_MIP_LAYOUT_RIGHT_LPT (1<<8) -#define MS4_VOLUME_DEPTH_SHIFT 0 +#define MS4_VOLUME_DEPTH_SHIFT 0 #define MS4_VOLUME_DEPTH_MASK (0xff<<0) /* p244 */ @@ -779,7 +804,7 @@ #define FILTER_4X4_1 3 #define FILTER_4X4_2 4 #define FILTER_4X4_FLAT 5 -#define FILTER_6X5_MONO 6 /* XXX - check */ +#define FILTER_6X5_MONO 6 /* XXX - check */ #define SS2_MIN_FILTER_SHIFT 14 #define SS2_MIN_FILTER_MASK (0x7<<14) #define SS2_LOD_BIAS_SHIFT 5 @@ -826,10 +851,14 @@ #define ST1_ENABLE (1<<16) #define ST1_MASK (0xffff) +#define _3DSTATE_DEFAULT_Z ((0x3<<29)|(0x1d<<24)|(0x98<<16)) +#define _3DSTATE_DEFAULT_DIFFUSE ((0x3<<29)|(0x1d<<24)|(0x99<<16)) +#define _3DSTATE_DEFAULT_SPECULAR ((0x3<<29)|(0x1d<<24)|(0x9a<<16)) + -#define MI_FLUSH ((0<<29)|(4<<23)) -#define FLUSH_MAP_CACHE (1<<0) -#define FLUSH_RENDER_CACHE (1<<1) +#define MI_FLUSH ((0<<29)|(4<<23)) +#define FLUSH_MAP_CACHE (1<<0) +#define INHIBIT_FLUSH_RENDER_CACHE (1<<2) #endif diff --git a/src/mesa/drivers/dri/i915/i915_state.c b/src/mesa/drivers/dri/i915/i915_state.c index 1c4ec74755..e5d8d27993 100644 --- a/src/mesa/drivers/dri/i915/i915_state.c +++ b/src/mesa/drivers/dri/i915/i915_state.c @@ -38,97 +38,99 @@ #include "drivers/common/driverfuncs.h" +#include "intel_fbo.h" #include "intel_screen.h" #include "intel_batchbuffer.h" #include "i915_context.h" #include "i915_reg.h" - +#define FILE_DEBUG_FLAG DEBUG_STATE static void -i915StencilFuncSeparate(GLcontext *ctx, GLenum face, GLenum func, GLint ref, +i915StencilFuncSeparate(GLcontext * ctx, GLenum face, GLenum func, GLint ref, GLuint mask) { - i915ContextPtr i915 = I915_CONTEXT(ctx); - int test = intel_translate_compare_func( func ); + struct i915_context *i915 = I915_CONTEXT(ctx); + int test = intel_translate_compare_func(func); mask = mask & 0xff; - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s : func: %s, ref : 0x%x, mask: 0x%x\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(func), ref, mask); + DBG("%s : func: %s, ref : 0x%x, mask: 0x%x\n", __FUNCTION__, + _mesa_lookup_enum_by_nr(func), ref, mask); I915_STATECHANGE(i915, I915_UPLOAD_CTX); i915->state.Ctx[I915_CTXREG_STATE4] &= ~MODE4_ENABLE_STENCIL_TEST_MASK; i915->state.Ctx[I915_CTXREG_STATE4] |= (ENABLE_STENCIL_TEST_MASK | - STENCIL_TEST_MASK(mask)); + STENCIL_TEST_MASK(mask)); i915->state.Ctx[I915_CTXREG_LIS5] &= ~(S5_STENCIL_REF_MASK | - S5_STENCIL_TEST_FUNC_MASK); - - i915->state.Ctx[I915_CTXREG_LIS5] |= ((ref << S5_STENCIL_REF_SHIFT) | - (test << S5_STENCIL_TEST_FUNC_SHIFT)); + S5_STENCIL_TEST_FUNC_MASK); + + i915->state.Ctx[I915_CTXREG_LIS5] |= ((ref << S5_STENCIL_REF_SHIFT) | + (test << + S5_STENCIL_TEST_FUNC_SHIFT)); } static void -i915StencilMaskSeparate(GLcontext *ctx, GLenum face, GLuint mask) +i915StencilMaskSeparate(GLcontext * ctx, GLenum face, GLuint mask) { - i915ContextPtr i915 = I915_CONTEXT(ctx); - - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s : mask 0x%x\n", __FUNCTION__, mask); + struct i915_context *i915 = I915_CONTEXT(ctx); + DBG("%s : mask 0x%x\n", __FUNCTION__, mask); + mask = mask & 0xff; I915_STATECHANGE(i915, I915_UPLOAD_CTX); i915->state.Ctx[I915_CTXREG_STATE4] &= ~MODE4_ENABLE_STENCIL_WRITE_MASK; i915->state.Ctx[I915_CTXREG_STATE4] |= (ENABLE_STENCIL_WRITE_MASK | - STENCIL_WRITE_MASK(mask)); + STENCIL_WRITE_MASK(mask)); } static void -i915StencilOpSeparate(GLcontext *ctx, GLenum face, GLenum fail, GLenum zfail, +i915StencilOpSeparate(GLcontext * ctx, GLenum face, GLenum fail, GLenum zfail, GLenum zpass) { - i915ContextPtr i915 = I915_CONTEXT(ctx); - int fop = intel_translate_stencil_op(fail); - int dfop = intel_translate_stencil_op(zfail); + struct i915_context *i915 = I915_CONTEXT(ctx); + int fop = intel_translate_stencil_op(fail); + int dfop = intel_translate_stencil_op(zfail); int dpop = intel_translate_stencil_op(zpass); - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s: fail : %s, zfail: %s, zpass : %s\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(fail), - _mesa_lookup_enum_by_nr(zfail), - _mesa_lookup_enum_by_nr(zpass)); + DBG("%s: fail : %s, zfail: %s, zpass : %s\n", __FUNCTION__, + _mesa_lookup_enum_by_nr(fail), + _mesa_lookup_enum_by_nr(zfail), _mesa_lookup_enum_by_nr(zpass)); I915_STATECHANGE(i915, I915_UPLOAD_CTX); i915->state.Ctx[I915_CTXREG_LIS5] &= ~(S5_STENCIL_FAIL_MASK | - S5_STENCIL_PASS_Z_FAIL_MASK | - S5_STENCIL_PASS_Z_PASS_MASK); + S5_STENCIL_PASS_Z_FAIL_MASK | + S5_STENCIL_PASS_Z_PASS_MASK); i915->state.Ctx[I915_CTXREG_LIS5] |= ((fop << S5_STENCIL_FAIL_SHIFT) | - (dfop << S5_STENCIL_PASS_Z_FAIL_SHIFT) | - (dpop << S5_STENCIL_PASS_Z_PASS_SHIFT)); + (dfop << + S5_STENCIL_PASS_Z_FAIL_SHIFT) | + (dpop << + S5_STENCIL_PASS_Z_PASS_SHIFT)); } -static void i915AlphaFunc(GLcontext *ctx, GLenum func, GLfloat ref) +static void +i915AlphaFunc(GLcontext * ctx, GLenum func, GLfloat ref) { - i915ContextPtr i915 = I915_CONTEXT(ctx); - int test = intel_translate_compare_func( func ); + struct i915_context *i915 = I915_CONTEXT(ctx); + int test = intel_translate_compare_func(func); GLubyte refByte; UNCLAMPED_FLOAT_TO_UBYTE(refByte, ref); I915_STATECHANGE(i915, I915_UPLOAD_CTX); i915->state.Ctx[I915_CTXREG_LIS6] &= ~(S6_ALPHA_TEST_FUNC_MASK | - S6_ALPHA_REF_MASK); + S6_ALPHA_REF_MASK); i915->state.Ctx[I915_CTXREG_LIS6] |= ((test << S6_ALPHA_TEST_FUNC_SHIFT) | - (((GLuint)refByte) << S6_ALPHA_REF_SHIFT)); + (((GLuint) refByte) << + S6_ALPHA_REF_SHIFT)); } /* This function makes sure that the proper enables are @@ -137,41 +139,45 @@ static void i915AlphaFunc(GLcontext *ctx, GLenum func, GLfloat ref) * could change the LogicOp or Independant Alpha Blend without subsequent * calls to glEnable. */ -static void i915EvalLogicOpBlendState(GLcontext *ctx) +static void +i915EvalLogicOpBlendState(GLcontext * ctx) { - i915ContextPtr i915 = I915_CONTEXT(ctx); + struct i915_context *i915 = I915_CONTEXT(ctx); I915_STATECHANGE(i915, I915_UPLOAD_CTX); if (RGBA_LOGICOP_ENABLED(ctx)) { i915->state.Ctx[I915_CTXREG_LIS5] |= S5_LOGICOP_ENABLE; i915->state.Ctx[I915_CTXREG_LIS6] &= ~S6_CBUF_BLEND_ENABLE; - } else { + } + else { i915->state.Ctx[I915_CTXREG_LIS5] &= ~S5_LOGICOP_ENABLE; if (ctx->Color.BlendEnabled) { - i915->state.Ctx[I915_CTXREG_LIS6] |= S6_CBUF_BLEND_ENABLE; - } else { - i915->state.Ctx[I915_CTXREG_LIS6] &= ~S6_CBUF_BLEND_ENABLE; + i915->state.Ctx[I915_CTXREG_LIS6] |= S6_CBUF_BLEND_ENABLE; + } + else { + i915->state.Ctx[I915_CTXREG_LIS6] &= ~S6_CBUF_BLEND_ENABLE; } } } -static void i915BlendColor(GLcontext *ctx, const GLfloat color[4]) +static void +i915BlendColor(GLcontext * ctx, const GLfloat color[4]) { - i915ContextPtr i915 = I915_CONTEXT(ctx); + struct i915_context *i915 = I915_CONTEXT(ctx); GLubyte r, g, b, a; - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s\n", __FUNCTION__); - + DBG("%s\n", __FUNCTION__); + UNCLAMPED_FLOAT_TO_UBYTE(r, color[RCOMP]); UNCLAMPED_FLOAT_TO_UBYTE(g, color[GCOMP]); UNCLAMPED_FLOAT_TO_UBYTE(b, color[BCOMP]); UNCLAMPED_FLOAT_TO_UBYTE(a, color[ACOMP]); I915_STATECHANGE(i915, I915_UPLOAD_CTX); - i915->state.Ctx[I915_CTXREG_BLENDCOLOR1] = (a<<24) | (r<<16) | (g<<8) | b; + i915->state.Ctx[I915_CTXREG_BLENDCOLOR1] = + (a << 24) | (r << 16) | (g << 8) | b; } @@ -182,31 +188,37 @@ static void i915BlendColor(GLcontext *ctx, const GLfloat color[4]) -static GLuint translate_blend_equation( GLenum mode ) +static GLuint +translate_blend_equation(GLenum mode) { switch (mode) { - case GL_FUNC_ADD: return BLENDFUNC_ADD; - case GL_MIN: return BLENDFUNC_MIN; - case GL_MAX: return BLENDFUNC_MAX; - case GL_FUNC_SUBTRACT: return BLENDFUNC_SUBTRACT; - case GL_FUNC_REVERSE_SUBTRACT: return BLENDFUNC_REVERSE_SUBTRACT; - default: return 0; + case GL_FUNC_ADD: + return BLENDFUNC_ADD; + case GL_MIN: + return BLENDFUNC_MIN; + case GL_MAX: + return BLENDFUNC_MAX; + case GL_FUNC_SUBTRACT: + return BLENDFUNC_SUBTRACT; + case GL_FUNC_REVERSE_SUBTRACT: + return BLENDFUNC_REVERSE_SUBTRACT; + default: + return 0; } } -static void i915UpdateBlendState( GLcontext *ctx ) +static void +i915UpdateBlendState(GLcontext * ctx) { - i915ContextPtr i915 = I915_CONTEXT(ctx); - GLuint iab = (i915->state.Ctx[I915_CTXREG_IAB] & - ~(IAB_SRC_FACTOR_MASK | - IAB_DST_FACTOR_MASK | - (BLENDFUNC_MASK << IAB_FUNC_SHIFT) | - IAB_ENABLE)); - - GLuint lis6 = (i915->state.Ctx[I915_CTXREG_LIS6] & - ~(S6_CBUF_SRC_BLEND_FACT_MASK | - S6_CBUF_DST_BLEND_FACT_MASK | - S6_CBUF_BLEND_FUNC_MASK)); + struct i915_context *i915 = I915_CONTEXT(ctx); + GLuint iab = (i915->state.Ctx[I915_CTXREG_IAB] & + ~(IAB_SRC_FACTOR_MASK | + IAB_DST_FACTOR_MASK | + (BLENDFUNC_MASK << IAB_FUNC_SHIFT) | IAB_ENABLE)); + + GLuint lis6 = (i915->state.Ctx[I915_CTXREG_LIS6] & + ~(S6_CBUF_SRC_BLEND_FACT_MASK | + S6_CBUF_DST_BLEND_FACT_MASK | S6_CBUF_BLEND_FUNC_MASK)); GLuint eqRGB = ctx->Color.BlendEquationRGB; GLuint eqA = ctx->Color.BlendEquationA; @@ -223,15 +235,15 @@ static void i915UpdateBlendState( GLcontext *ctx ) srcA = dstA = GL_ONE; } - lis6 |= SRC_BLND_FACT(intel_translate_blend_factor(srcRGB)); - lis6 |= DST_BLND_FACT(intel_translate_blend_factor(dstRGB)); - lis6 |= translate_blend_equation( eqRGB ) << S6_CBUF_BLEND_FUNC_SHIFT; + lis6 |= SRC_BLND_FACT(intel_translate_blend_factor(srcRGB)); + lis6 |= DST_BLND_FACT(intel_translate_blend_factor(dstRGB)); + lis6 |= translate_blend_equation(eqRGB) << S6_CBUF_BLEND_FUNC_SHIFT; - iab |= SRC_ABLND_FACT(intel_translate_blend_factor(srcA)); - iab |= DST_ABLND_FACT(intel_translate_blend_factor(dstA)); - iab |= translate_blend_equation( eqA ) << IAB_FUNC_SHIFT; + iab |= SRC_ABLND_FACT(intel_translate_blend_factor(srcA)); + iab |= DST_ABLND_FACT(intel_translate_blend_factor(dstA)); + iab |= translate_blend_equation(eqA) << IAB_FUNC_SHIFT; - if (srcA != srcRGB || dstA != dstRGB || eqA != eqRGB) + if (srcA != srcRGB || dstA != dstRGB || eqA != eqRGB) iab |= IAB_ENABLE; if (iab != i915->state.Ctx[I915_CTXREG_IAB] || @@ -246,41 +258,41 @@ static void i915UpdateBlendState( GLcontext *ctx ) } -static void i915BlendFuncSeparate(GLcontext *ctx, GLenum srcRGB, - GLenum dstRGB, GLenum srcA, - GLenum dstA ) -{ - i915UpdateBlendState( ctx ); +static void +i915BlendFuncSeparate(GLcontext * ctx, GLenum srcRGB, + GLenum dstRGB, GLenum srcA, GLenum dstA) +{ + i915UpdateBlendState(ctx); } -static void i915BlendEquationSeparate(GLcontext *ctx, GLenum eqRGB, - GLenum eqA) +static void +i915BlendEquationSeparate(GLcontext * ctx, GLenum eqRGB, GLenum eqA) { - i915UpdateBlendState( ctx ); + i915UpdateBlendState(ctx); } -static void i915DepthFunc(GLcontext *ctx, GLenum func) +static void +i915DepthFunc(GLcontext * ctx, GLenum func) { - i915ContextPtr i915 = I915_CONTEXT(ctx); - int test = intel_translate_compare_func( func ); - - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s\n", __FUNCTION__); + struct i915_context *i915 = I915_CONTEXT(ctx); + int test = intel_translate_compare_func(func); + DBG("%s\n", __FUNCTION__); + I915_STATECHANGE(i915, I915_UPLOAD_CTX); i915->state.Ctx[I915_CTXREG_LIS6] &= ~S6_DEPTH_TEST_FUNC_MASK; i915->state.Ctx[I915_CTXREG_LIS6] |= test << S6_DEPTH_TEST_FUNC_SHIFT; } -static void i915DepthMask(GLcontext *ctx, GLboolean flag) +static void +i915DepthMask(GLcontext * ctx, GLboolean flag) { - i915ContextPtr i915 = I915_CONTEXT(ctx); - - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s flag (%d)\n", __FUNCTION__, flag); + struct i915_context *i915 = I915_CONTEXT(ctx); + DBG("%s flag (%d)\n", __FUNCTION__, flag); + I915_STATECHANGE(i915, I915_UPLOAD_CTX); if (flag && ctx->Depth.Test) @@ -295,14 +307,15 @@ static void i915DepthMask(GLcontext *ctx, GLboolean flag) * The i915 supports a 4x4 stipple natively, GL wants 32x32. * Fortunately stipple is usually a repeating pattern. */ -static void i915PolygonStipple( GLcontext *ctx, const GLubyte *mask ) +static void +i915PolygonStipple(GLcontext * ctx, const GLubyte * mask) { - i915ContextPtr i915 = I915_CONTEXT(ctx); + struct i915_context *i915 = I915_CONTEXT(ctx); const GLubyte *m = mask; GLubyte p[4]; - int i,j,k; + int i, j, k; int active = (ctx->Polygon.StippleFlag && - i915->intel.reduced_primitive == GL_TRIANGLES); + i915->intel.reduced_primitive == GL_TRIANGLES); GLuint newMask; if (active) { @@ -310,23 +323,26 @@ static void i915PolygonStipple( GLcontext *ctx, const GLubyte *mask ) i915->state.Stipple[I915_STPREG_ST1] &= ~ST1_ENABLE; } - p[0] = mask[12] & 0xf; p[0] |= p[0] << 4; - p[1] = mask[8] & 0xf; p[1] |= p[1] << 4; - p[2] = mask[4] & 0xf; p[2] |= p[2] << 4; - p[3] = mask[0] & 0xf; p[3] |= p[3] << 4; - - for (k = 0 ; k < 8 ; k++) - for (j = 3 ; j >= 0; j--) - for (i = 0 ; i < 4 ; i++, m++) - if (*m != p[j]) { - i915->intel.hw_stipple = 0; - return; - } + p[0] = mask[12] & 0xf; + p[0] |= p[0] << 4; + p[1] = mask[8] & 0xf; + p[1] |= p[1] << 4; + p[2] = mask[4] & 0xf; + p[2] |= p[2] << 4; + p[3] = mask[0] & 0xf; + p[3] |= p[3] << 4; + + for (k = 0; k < 8; k++) + for (j = 3; j >= 0; j--) + for (i = 0; i < 4; i++, m++) + if (*m != p[j]) { + i915->intel.hw_stipple = 0; + return; + } newMask = (((p[0] & 0xf) << 0) | - ((p[1] & 0xf) << 4) | - ((p[2] & 0xf) << 8) | - ((p[3] & 0xf) << 12)); + ((p[1] & 0xf) << 4) | + ((p[2] & 0xf) << 8) | ((p[3] & 0xf) << 12)); if (newMask == 0xffff || newMask == 0x0) { @@ -347,49 +363,54 @@ static void i915PolygonStipple( GLcontext *ctx, const GLubyte *mask ) /* ============================================================= * Hardware clipping */ -static void i915Scissor(GLcontext *ctx, GLint x, GLint y, - GLsizei w, GLsizei h) +static void +i915Scissor(GLcontext * ctx, GLint x, GLint y, GLsizei w, GLsizei h) { - i915ContextPtr i915 = I915_CONTEXT(ctx); - intelScreenPrivate *screen = i915->intel.intelScreen; + struct i915_context *i915 = I915_CONTEXT(ctx); int x1, y1, x2, y2; - if (!i915->intel.driDrawable) + if (!ctx->DrawBuffer) return; - x1 = x; - y1 = i915->intel.driDrawable->h - (y + h); - x2 = x + w - 1; - y2 = y1 + h - 1; - - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "[%s] x(%d) y(%d) w(%d) h(%d)\n", __FUNCTION__, - x, y, w, h); - - if (x1 < 0) x1 = 0; - if (y1 < 0) y1 = 0; - if (x2 < 0) x2 = 0; - if (y2 < 0) y2 = 0; - - if (x2 >= screen->width) x2 = screen->width-1; - if (y2 >= screen->height) y2 = screen->height-1; - if (x1 >= screen->width) x1 = screen->width-1; - if (y1 >= screen->height) y1 = screen->height-1; + DBG("%s %d,%d %dx%d\n", __FUNCTION__, x, y, w, h); + if (ctx->DrawBuffer->Name == 0) { + x1 = x; + y1 = ctx->DrawBuffer->Height - (y + h); + x2 = x + w - 1; + y2 = y1 + h - 1; + DBG("%s %d..%d,%d..%d (inverted)\n", __FUNCTION__, x1, x2, y1, y2); + } + else { + /* FBO - not inverted + */ + x1 = x; + y1 = y; + x2 = x + w - 1; + y2 = y + h - 1; + DBG("%s %d..%d,%d..%d (not inverted)\n", __FUNCTION__, x1, x2, y1, y2); + } + + x1 = CLAMP(x1, 0, ctx->DrawBuffer->Width - 1); + y1 = CLAMP(y1, 0, ctx->DrawBuffer->Height - 1); + x2 = CLAMP(x2, 0, ctx->DrawBuffer->Width - 1); + y2 = CLAMP(y2, 0, ctx->DrawBuffer->Height - 1); + + DBG("%s %d..%d,%d..%d (clamped)\n", __FUNCTION__, x1, x2, y1, y2); I915_STATECHANGE(i915, I915_UPLOAD_BUFFERS); i915->state.Buffer[I915_DESTREG_SR1] = (y1 << 16) | (x1 & 0xffff); i915->state.Buffer[I915_DESTREG_SR2] = (y2 << 16) | (x2 & 0xffff); } -static void i915LogicOp(GLcontext *ctx, GLenum opcode) +static void +i915LogicOp(GLcontext * ctx, GLenum opcode) { - i915ContextPtr i915 = I915_CONTEXT(ctx); + struct i915_context *i915 = I915_CONTEXT(ctx); int tmp = intel_translate_logic_op(opcode); - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s\n", __FUNCTION__); - + DBG("%s\n", __FUNCTION__); + I915_STATECHANGE(i915, I915_UPLOAD_CTX); i915->state.Ctx[I915_CTXREG_STATE4] &= ~LOGICOP_MASK; i915->state.Ctx[I915_CTXREG_STATE4] |= LOGIC_OP_FUNC(tmp); @@ -397,13 +418,14 @@ static void i915LogicOp(GLcontext *ctx, GLenum opcode) -static void i915CullFaceFrontFace(GLcontext *ctx, GLenum unused) +static void +i915CullFaceFrontFace(GLcontext * ctx, GLenum unused) { - i915ContextPtr i915 = I915_CONTEXT(ctx); + struct i915_context *i915 = I915_CONTEXT(ctx); GLuint mode; - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s\n", __FUNCTION__); + DBG("%s %d\n", __FUNCTION__, + ctx->DrawBuffer ? ctx->DrawBuffer->Name : 0); if (!ctx->Polygon.CullFlag) { mode = S4_CULLMODE_NONE; @@ -411,10 +433,12 @@ static void i915CullFaceFrontFace(GLcontext *ctx, GLenum unused) else if (ctx->Polygon.CullFaceMode != GL_FRONT_AND_BACK) { mode = S4_CULLMODE_CW; + if (ctx->DrawBuffer && ctx->DrawBuffer->Name != 0) + mode ^= (S4_CULLMODE_CW ^ S4_CULLMODE_CCW); if (ctx->Polygon.CullFaceMode == GL_FRONT) - mode ^= (S4_CULLMODE_CW ^ S4_CULLMODE_CCW); + mode ^= (S4_CULLMODE_CW ^ S4_CULLMODE_CCW); if (ctx->Polygon.FrontFace != GL_CCW) - mode ^= (S4_CULLMODE_CW ^ S4_CULLMODE_CCW); + mode ^= (S4_CULLMODE_CW ^ S4_CULLMODE_CCW); } else { mode = S4_CULLMODE_BOTH; @@ -425,16 +449,16 @@ static void i915CullFaceFrontFace(GLcontext *ctx, GLenum unused) i915->state.Ctx[I915_CTXREG_LIS4] |= mode; } -static void i915LineWidth( GLcontext *ctx, GLfloat widthf ) +static void +i915LineWidth(GLcontext * ctx, GLfloat widthf) { - i915ContextPtr i915 = I915_CONTEXT( ctx ); + struct i915_context *i915 = I915_CONTEXT(ctx); int lis4 = i915->state.Ctx[I915_CTXREG_LIS4] & ~S4_LINE_WIDTH_MASK; int width; - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s\n", __FUNCTION__); - - width = (int)(widthf * 2); + DBG("%s\n", __FUNCTION__); + + width = (int) (widthf * 2); CLAMP_SELF(width, 1, 0xf); lis4 |= width << S4_LINE_WIDTH_SHIFT; @@ -444,15 +468,15 @@ static void i915LineWidth( GLcontext *ctx, GLfloat widthf ) } } -static void i915PointSize(GLcontext *ctx, GLfloat size) +static void +i915PointSize(GLcontext * ctx, GLfloat size) { - i915ContextPtr i915 = I915_CONTEXT(ctx); + struct i915_context *i915 = I915_CONTEXT(ctx); int lis4 = i915->state.Ctx[I915_CTXREG_LIS4] & ~S4_POINT_WIDTH_MASK; - GLint point_size = (int)size; - - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s\n", __FUNCTION__); + GLint point_size = (int) size; + DBG("%s\n", __FUNCTION__); + CLAMP_SELF(point_size, 1, 255); lis4 |= point_size << S4_POINT_WIDTH_SHIFT; @@ -467,20 +491,24 @@ static void i915PointSize(GLcontext *ctx, GLfloat size) * Color masks */ -static void i915ColorMask(GLcontext *ctx, - GLboolean r, GLboolean g, - GLboolean b, GLboolean a) +static void +i915ColorMask(GLcontext * ctx, + GLboolean r, GLboolean g, GLboolean b, GLboolean a) { - i915ContextPtr i915 = I915_CONTEXT( ctx ); + struct i915_context *i915 = I915_CONTEXT(ctx); GLuint tmp = i915->state.Ctx[I915_CTXREG_LIS5] & ~S5_WRITEDISABLE_MASK; - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s r(%d) g(%d) b(%d) a(%d)\n", __FUNCTION__, r, g, b, a); + DBG("%s r(%d) g(%d) b(%d) a(%d)\n", __FUNCTION__, r, g, b, + a); - if (!r) tmp |= S5_WRITEDISABLE_RED; - if (!g) tmp |= S5_WRITEDISABLE_GREEN; - if (!b) tmp |= S5_WRITEDISABLE_BLUE; - if (!a) tmp |= S5_WRITEDISABLE_ALPHA; + if (!r) + tmp |= S5_WRITEDISABLE_RED; + if (!g) + tmp |= S5_WRITEDISABLE_GREEN; + if (!b) + tmp |= S5_WRITEDISABLE_BLUE; + if (!a) + tmp |= S5_WRITEDISABLE_ALPHA; if (tmp != i915->state.Ctx[I915_CTXREG_LIS5]) { I915_STATECHANGE(i915, I915_UPLOAD_CTX); @@ -488,54 +516,55 @@ static void i915ColorMask(GLcontext *ctx, } } -static void update_specular( GLcontext *ctx ) +static void +update_specular(GLcontext * ctx) { /* A hack to trigger the rebuild of the fragment program. */ - INTEL_CONTEXT(ctx)->NewGLState |= _NEW_TEXTURE; - I915_CONTEXT(ctx)->tex_program.translated = 0; + intel_context(ctx)->NewGLState |= _NEW_TEXTURE; } -static void i915LightModelfv(GLcontext *ctx, GLenum pname, - const GLfloat *param) +static void +i915LightModelfv(GLcontext * ctx, GLenum pname, const GLfloat * param) { - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s\n", __FUNCTION__); - + DBG("%s\n", __FUNCTION__); + if (pname == GL_LIGHT_MODEL_COLOR_CONTROL) { - update_specular( ctx ); + update_specular(ctx); } } -static void i915ShadeModel(GLcontext *ctx, GLenum mode) +static void +i915ShadeModel(GLcontext * ctx, GLenum mode) { - i915ContextPtr i915 = I915_CONTEXT(ctx); + struct i915_context *i915 = I915_CONTEXT(ctx); I915_STATECHANGE(i915, I915_UPLOAD_CTX); if (mode == GL_SMOOTH) { - i915->state.Ctx[I915_CTXREG_LIS4] &= ~(S4_FLATSHADE_ALPHA | - S4_FLATSHADE_COLOR | - S4_FLATSHADE_SPECULAR); - } else { - i915->state.Ctx[I915_CTXREG_LIS4] |= (S4_FLATSHADE_ALPHA | - S4_FLATSHADE_COLOR | - S4_FLATSHADE_SPECULAR); + i915->state.Ctx[I915_CTXREG_LIS4] &= ~(S4_FLATSHADE_ALPHA | + S4_FLATSHADE_COLOR | + S4_FLATSHADE_SPECULAR); + } + else { + i915->state.Ctx[I915_CTXREG_LIS4] |= (S4_FLATSHADE_ALPHA | + S4_FLATSHADE_COLOR | + S4_FLATSHADE_SPECULAR); } } /* ============================================================= * Fog */ -void i915_update_fog( GLcontext *ctx ) +void +i915_update_fog(GLcontext * ctx) { - i915ContextPtr i915 = I915_CONTEXT(ctx); + struct i915_context *i915 = I915_CONTEXT(ctx); GLenum mode; GLboolean enabled; GLboolean try_pixel_fog; - + if (ctx->FragmentProgram._Active) { /* Pull in static fog state from program */ - mode = ctx->FragmentProgram._Current->FogOption; enabled = (mode != GL_NONE); try_pixel_fog = 0; @@ -546,7 +575,7 @@ void i915_update_fog( GLcontext *ctx ) #if 0 /* XXX - DISABLED -- Need ortho fallback */ try_pixel_fog = (ctx->Fog.FogCoordinateSource == GL_FRAGMENT_DEPTH_EXT - &&ctx->Hint.Fog == GL_NICEST); + && ctx->Hint.Fog == GL_NICEST); #else try_pixel_fog = 0; #endif @@ -559,48 +588,49 @@ void i915_update_fog( GLcontext *ctx ) I915_STATECHANGE(i915, I915_UPLOAD_FOG); i915->state.Fog[I915_FOGREG_MODE1] &= ~FMC1_FOGFUNC_MASK; i915->vertex_fog = I915_FOG_PIXEL; - + switch (mode) { case GL_LINEAR: - if (ctx->Fog.End <= ctx->Fog.Start) { - /* XXX - this won't work with fragment programs. Need to - * either fallback or append fog instructions to end of - * program in the case of linear fog. - */ - i915->state.Fog[I915_FOGREG_MODE1] |= FMC1_FOGFUNC_VERTEX; - i915->vertex_fog = I915_FOG_VERTEX; - } - else { + if (ctx->Fog.End <= ctx->Fog.Start) { + /* XXX - this won't work with fragment programs. Need to + * either fallback or append fog instructions to end of + * program in the case of linear fog. + */ + printf("vertex fog!\n"); + i915->state.Fog[I915_FOGREG_MODE1] |= FMC1_FOGFUNC_VERTEX; + i915->vertex_fog = I915_FOG_VERTEX; + } + else { GLfloat c2 = 1.0 / (ctx->Fog.End - ctx->Fog.Start); GLfloat c1 = ctx->Fog.End * c2; - i915->state.Fog[I915_FOGREG_MODE1] &= ~FMC1_C1_MASK; - i915->state.Fog[I915_FOGREG_MODE1] |= FMC1_FOGFUNC_PIXEL_LINEAR; - i915->state.Fog[I915_FOGREG_MODE1] |= - ((GLuint)(c1 * FMC1_C1_ONE)) & FMC1_C1_MASK; - - if (i915->state.Fog[I915_FOGREG_MODE1] & FMC1_FOGINDEX_Z) { - i915->state.Fog[I915_FOGREG_MODE2] - = (GLuint)(c2 * FMC2_C2_ONE); - } - else { - fi_type fi; - fi.f = c2; - i915->state.Fog[I915_FOGREG_MODE2] = fi.i; - } - } - break; + i915->state.Fog[I915_FOGREG_MODE1] &= ~FMC1_C1_MASK; + i915->state.Fog[I915_FOGREG_MODE1] |= FMC1_FOGFUNC_PIXEL_LINEAR; + i915->state.Fog[I915_FOGREG_MODE1] |= + ((GLuint) (c1 * FMC1_C1_ONE)) & FMC1_C1_MASK; + + if (i915->state.Fog[I915_FOGREG_MODE1] & FMC1_FOGINDEX_Z) { + i915->state.Fog[I915_FOGREG_MODE2] + = (GLuint) (c2 * FMC2_C2_ONE); + } + else { + fi_type fi; + fi.f = c2; + i915->state.Fog[I915_FOGREG_MODE2] = fi.i; + } + } + break; case GL_EXP: - i915->state.Fog[I915_FOGREG_MODE1] |= FMC1_FOGFUNC_PIXEL_EXP; - break; + i915->state.Fog[I915_FOGREG_MODE1] |= FMC1_FOGFUNC_PIXEL_EXP; + break; case GL_EXP2: - i915->state.Fog[I915_FOGREG_MODE1] |= FMC1_FOGFUNC_PIXEL_EXP2; - break; + i915->state.Fog[I915_FOGREG_MODE1] |= FMC1_FOGFUNC_PIXEL_EXP2; + break; default: - break; + break; } } - else /* if (i915->vertex_fog != I915_FOG_VERTEX) */ { + else { /* if (i915->vertex_fog != I915_FOG_VERTEX) */ I915_STATECHANGE(i915, I915_UPLOAD_FOG); i915->state.Fog[I915_FOGREG_MODE1] &= ~FMC1_FOGFUNC_MASK; i915->state.Fog[I915_FOGREG_MODE1] |= FMC1_FOGFUNC_VERTEX; @@ -622,38 +652,38 @@ void i915_update_fog( GLcontext *ctx ) } static void -i915Fogfv(GLcontext *ctx, GLenum pname, const GLfloat *param) +i915Fogfv(GLcontext * ctx, GLenum pname, const GLfloat * param) { - i915ContextPtr i915 = I915_CONTEXT(ctx); + struct i915_context *i915 = I915_CONTEXT(ctx); switch (pname) { - case GL_FOG_COORDINATE_SOURCE_EXT: + case GL_FOG_COORDINATE_SOURCE_EXT: case GL_FOG_MODE: case GL_FOG_START: - case GL_FOG_END: + case GL_FOG_END: break; case GL_FOG_DENSITY: I915_STATECHANGE(i915, I915_UPLOAD_FOG); if (i915->state.Fog[I915_FOGREG_MODE1] & FMC1_FOGINDEX_Z) { - i915->state.Fog[I915_FOGREG_MODE3] - = (GLuint)(ctx->Fog.Density * FMC3_D_ONE); + i915->state.Fog[I915_FOGREG_MODE3] = + (GLuint) (ctx->Fog.Density * FMC3_D_ONE); } else { - union { float f; int i; } fi; - fi.f = ctx->Fog.Density; - i915->state.Fog[I915_FOGREG_MODE3] = fi.i; + fi_type fi; + fi.f = ctx->Fog.Density; + i915->state.Fog[I915_FOGREG_MODE3] = fi.i; } break; - case GL_FOG_COLOR: + case GL_FOG_COLOR: I915_STATECHANGE(i915, I915_UPLOAD_FOG); - i915->state.Fog[I915_FOGREG_COLOR] = - (_3DSTATE_FOG_COLOR_CMD | - ((GLubyte)(ctx->Fog.Color[0]*255.0F) << 16) | - ((GLubyte)(ctx->Fog.Color[1]*255.0F) << 8) | - ((GLubyte)(ctx->Fog.Color[2]*255.0F) << 0)); + i915->state.Fog[I915_FOGREG_COLOR] = + (_3DSTATE_FOG_COLOR_CMD | + ((GLubyte) (ctx->Fog.Color[0] * 255.0F) << 16) | + ((GLubyte) (ctx->Fog.Color[1] * 255.0F) << 8) | + ((GLubyte) (ctx->Fog.Color[2] * 255.0F) << 0)); break; default: @@ -661,7 +691,8 @@ i915Fogfv(GLcontext *ctx, GLenum pname, const GLfloat *param) } } -static void i915Hint(GLcontext *ctx, GLenum target, GLenum state) +static void +i915Hint(GLcontext * ctx, GLenum target, GLenum state) { switch (target) { case GL_FOG_HINT: @@ -674,25 +705,26 @@ static void i915Hint(GLcontext *ctx, GLenum target, GLenum state) /* ============================================================= */ -static void i915Enable(GLcontext *ctx, GLenum cap, GLboolean state) +static void +i915Enable(GLcontext * ctx, GLenum cap, GLboolean state) { - i915ContextPtr i915 = I915_CONTEXT(ctx); + struct i915_context *i915 = I915_CONTEXT(ctx); - switch(cap) { + switch (cap) { case GL_TEXTURE_2D: break; case GL_LIGHTING: case GL_COLOR_SUM: - update_specular( ctx ); + update_specular(ctx); break; case GL_ALPHA_TEST: I915_STATECHANGE(i915, I915_UPLOAD_CTX); if (state) - i915->state.Ctx[I915_CTXREG_LIS6] |= S6_ALPHA_TEST_ENABLE; + i915->state.Ctx[I915_CTXREG_LIS6] |= S6_ALPHA_TEST_ENABLE; else - i915->state.Ctx[I915_CTXREG_LIS6] &= ~S6_ALPHA_TEST_ENABLE; + i915->state.Ctx[I915_CTXREG_LIS6] &= ~S6_ALPHA_TEST_ENABLE; break; case GL_BLEND: @@ -704,8 +736,8 @@ static void i915Enable(GLcontext *ctx, GLenum cap, GLboolean state) /* Logicop doesn't seem to work at 16bpp: */ - if (i915->intel.intelScreen->cpp == 2) - FALLBACK( &i915->intel, I915_FALLBACK_LOGICOP, state ); + if (i915->intel.intelScreen->cpp == 2) /* XXX FBO fix */ + FALLBACK(&i915->intel, I915_FALLBACK_LOGICOP, state); break; case GL_FRAGMENT_PROGRAM_ARB: @@ -714,37 +746,37 @@ static void i915Enable(GLcontext *ctx, GLenum cap, GLboolean state) case GL_DITHER: I915_STATECHANGE(i915, I915_UPLOAD_CTX); if (state) - i915->state.Ctx[I915_CTXREG_LIS5] |= S5_COLOR_DITHER_ENABLE; + i915->state.Ctx[I915_CTXREG_LIS5] |= S5_COLOR_DITHER_ENABLE; else - i915->state.Ctx[I915_CTXREG_LIS5] &= ~S5_COLOR_DITHER_ENABLE; + i915->state.Ctx[I915_CTXREG_LIS5] &= ~S5_COLOR_DITHER_ENABLE; break; case GL_DEPTH_TEST: I915_STATECHANGE(i915, I915_UPLOAD_CTX); if (state) - i915->state.Ctx[I915_CTXREG_LIS6] |= S6_DEPTH_TEST_ENABLE; + i915->state.Ctx[I915_CTXREG_LIS6] |= S6_DEPTH_TEST_ENABLE; else - i915->state.Ctx[I915_CTXREG_LIS6] &= ~S6_DEPTH_TEST_ENABLE; + i915->state.Ctx[I915_CTXREG_LIS6] &= ~S6_DEPTH_TEST_ENABLE; - i915DepthMask( ctx, ctx->Depth.Mask ); + i915DepthMask(ctx, ctx->Depth.Mask); break; case GL_SCISSOR_TEST: I915_STATECHANGE(i915, I915_UPLOAD_BUFFERS); if (state) - i915->state.Buffer[I915_DESTREG_SENABLE] = (_3DSTATE_SCISSOR_ENABLE_CMD | - ENABLE_SCISSOR_RECT); + i915->state.Buffer[I915_DESTREG_SENABLE] = + (_3DSTATE_SCISSOR_ENABLE_CMD | ENABLE_SCISSOR_RECT); else - i915->state.Buffer[I915_DESTREG_SENABLE] = (_3DSTATE_SCISSOR_ENABLE_CMD | - DISABLE_SCISSOR_RECT); + i915->state.Buffer[I915_DESTREG_SENABLE] = + (_3DSTATE_SCISSOR_ENABLE_CMD | DISABLE_SCISSOR_RECT); break; case GL_LINE_SMOOTH: I915_STATECHANGE(i915, I915_UPLOAD_CTX); if (state) - i915->state.Ctx[I915_CTXREG_LIS4] |= S4_LINE_ANTIALIAS_ENABLE; + i915->state.Ctx[I915_CTXREG_LIS4] |= S4_LINE_ANTIALIAS_ENABLE; else - i915->state.Ctx[I915_CTXREG_LIS4] &= ~S4_LINE_ANTIALIAS_ENABLE; + i915->state.Ctx[I915_CTXREG_LIS4] &= ~S4_LINE_ANTIALIAS_ENABLE; break; case GL_FOG: @@ -755,16 +787,25 @@ static void i915Enable(GLcontext *ctx, GLenum cap, GLboolean state) break; case GL_STENCIL_TEST: - if (i915->intel.hw_stencil) { - I915_STATECHANGE(i915, I915_UPLOAD_CTX); - if (state) - i915->state.Ctx[I915_CTXREG_LIS5] |= (S5_STENCIL_TEST_ENABLE | - S5_STENCIL_WRITE_ENABLE); - else - i915->state.Ctx[I915_CTXREG_LIS5] &= ~(S5_STENCIL_TEST_ENABLE | - S5_STENCIL_WRITE_ENABLE); - } else { - FALLBACK( &i915->intel, I915_FALLBACK_STENCIL, state ); + { + GLboolean hw_stencil = GL_FALSE; + if (ctx->DrawBuffer) { + struct intel_renderbuffer *irbStencil + = intel_get_renderbuffer(ctx->DrawBuffer, BUFFER_STENCIL); + hw_stencil = (irbStencil && irbStencil->region); + } + if (hw_stencil) { + I915_STATECHANGE(i915, I915_UPLOAD_CTX); + if (state) + i915->state.Ctx[I915_CTXREG_LIS5] |= (S5_STENCIL_TEST_ENABLE | + S5_STENCIL_WRITE_ENABLE); + else + i915->state.Ctx[I915_CTXREG_LIS5] &= ~(S5_STENCIL_TEST_ENABLE | + S5_STENCIL_WRITE_ENABLE); + } + else { + FALLBACK(&i915->intel, I915_FALLBACK_STENCIL, state); + } } break; @@ -773,23 +814,20 @@ static void i915Enable(GLcontext *ctx, GLenum cap, GLboolean state) * I'll do more testing later to find out exactly which hardware * supports it. Disabled for now. */ - if (i915->intel.hw_stipple && - i915->intel.reduced_primitive == GL_TRIANGLES) - { - I915_STATECHANGE(i915, I915_UPLOAD_STIPPLE); - if (state) - i915->state.Stipple[I915_STPREG_ST1] |= ST1_ENABLE; - else - i915->state.Stipple[I915_STPREG_ST1] &= ~ST1_ENABLE; + if (i915->intel.hw_stipple && + i915->intel.reduced_primitive == GL_TRIANGLES) { + I915_STATECHANGE(i915, I915_UPLOAD_STIPPLE); + if (state) + i915->state.Stipple[I915_STPREG_ST1] |= ST1_ENABLE; + else + i915->state.Stipple[I915_STPREG_ST1] &= ~ST1_ENABLE; } break; case GL_POLYGON_SMOOTH: - FALLBACK( &i915->intel, I915_FALLBACK_POLYGON_SMOOTH, state ); break; case GL_POINT_SMOOTH: - FALLBACK( &i915->intel, I915_FALLBACK_POINT_SMOOTH, state ); break; default: @@ -798,7 +836,8 @@ static void i915Enable(GLcontext *ctx, GLenum cap, GLboolean state) } -static void i915_init_packets( i915ContextPtr i915 ) +static void +i915_init_packets(struct i915_context *i915) { intelScreenPrivate *screen = i915->intel.intelScreen; @@ -811,39 +850,35 @@ static void i915_init_packets( i915ContextPtr i915 ) /* Probably don't want to upload all this stuff every time one * piece changes. */ - i915->state.Ctx[I915_CTXREG_LI] = (_3DSTATE_LOAD_STATE_IMMEDIATE_1 | - I1_LOAD_S(2) | - I1_LOAD_S(4) | - I1_LOAD_S(5) | - I1_LOAD_S(6) | - (3)); + i915->state.Ctx[I915_CTXREG_LI] = (_3DSTATE_LOAD_STATE_IMMEDIATE_1 | + I1_LOAD_S(2) | + I1_LOAD_S(4) | + I1_LOAD_S(5) | I1_LOAD_S(6) | (3)); i915->state.Ctx[I915_CTXREG_LIS2] = 0; i915->state.Ctx[I915_CTXREG_LIS4] = 0; i915->state.Ctx[I915_CTXREG_LIS5] = 0; - if (screen->cpp == 2) - i915->state.Ctx[I915_CTXREG_LIS5] |= S5_COLOR_DITHER_ENABLE; + if (screen->cpp == 2) /* XXX FBO fix */ + i915->state.Ctx[I915_CTXREG_LIS5] |= S5_COLOR_DITHER_ENABLE; i915->state.Ctx[I915_CTXREG_LIS6] = (S6_COLOR_WRITE_ENABLE | - (2 << S6_TRISTRIP_PV_SHIFT)); + (2 << S6_TRISTRIP_PV_SHIFT)); i915->state.Ctx[I915_CTXREG_STATE4] = (_3DSTATE_MODES_4_CMD | - ENABLE_LOGIC_OP_FUNC | - LOGIC_OP_FUNC(LOGICOP_COPY) | - ENABLE_STENCIL_TEST_MASK | - STENCIL_TEST_MASK(0xff) | - ENABLE_STENCIL_WRITE_MASK | - STENCIL_WRITE_MASK(0xff)); - - - i915->state.Ctx[I915_CTXREG_IAB] = (_3DSTATE_INDEPENDENT_ALPHA_BLEND_CMD | - IAB_MODIFY_ENABLE | - IAB_MODIFY_FUNC | - IAB_MODIFY_SRC_FACTOR | - IAB_MODIFY_DST_FACTOR); - - i915->state.Ctx[I915_CTXREG_BLENDCOLOR0] = _3DSTATE_CONST_BLEND_COLOR_CMD; + ENABLE_LOGIC_OP_FUNC | + LOGIC_OP_FUNC(LOGICOP_COPY) | + ENABLE_STENCIL_TEST_MASK | + STENCIL_TEST_MASK(0xff) | + ENABLE_STENCIL_WRITE_MASK | + STENCIL_WRITE_MASK(0xff)); + + i915->state.Ctx[I915_CTXREG_IAB] = + (_3DSTATE_INDEPENDENT_ALPHA_BLEND_CMD | IAB_MODIFY_ENABLE | + IAB_MODIFY_FUNC | IAB_MODIFY_SRC_FACTOR | IAB_MODIFY_DST_FACTOR); + + i915->state.Ctx[I915_CTXREG_BLENDCOLOR0] = + _3DSTATE_CONST_BLEND_COLOR_CMD; i915->state.Ctx[I915_CTXREG_BLENDCOLOR1] = 0; } @@ -858,11 +893,11 @@ static void i915_init_packets( i915ContextPtr i915 ) I915_STATECHANGE(i915, I915_UPLOAD_FOG); i915->state.Fog[I915_FOGREG_MODE0] = _3DSTATE_FOG_MODE_CMD; i915->state.Fog[I915_FOGREG_MODE1] = (FMC1_FOGFUNC_MODIFY_ENABLE | - FMC1_FOGFUNC_VERTEX | - FMC1_FOGINDEX_MODIFY_ENABLE | - FMC1_FOGINDEX_W | - FMC1_C1_C2_MODIFY_ENABLE | - FMC1_DENSITY_MODIFY_ENABLE); + FMC1_FOGFUNC_VERTEX | + FMC1_FOGINDEX_MODIFY_ENABLE | + FMC1_FOGINDEX_W | + FMC1_C1_C2_MODIFY_ENABLE | + FMC1_DENSITY_MODIFY_ENABLE); i915->state.Fog[I915_FOGREG_COLOR] = _3DSTATE_FOG_COLOR_CMD; } @@ -871,66 +906,74 @@ static void i915_init_packets( i915ContextPtr i915 ) I915_STATECHANGE(i915, I915_UPLOAD_BUFFERS); /* color buffer offset/stride */ i915->state.Buffer[I915_DESTREG_CBUFADDR0] = _3DSTATE_BUF_INFO_CMD; - i915->state.Buffer[I915_DESTREG_CBUFADDR1] = - (BUF_3D_ID_COLOR_BACK | - BUF_3D_PITCH(screen->front.pitch) | /* pitch in bytes */ - BUF_3D_USE_FENCE); - /*i915->state.Buffer[I915_DESTREG_CBUFADDR2] is the offset */ - + /* XXX FBO: remove this? Also get set in i915_set_draw_region() */ + i915->state.Buffer[I915_DESTREG_CBUFADDR1] = (BUF_3D_ID_COLOR_BACK | BUF_3D_PITCH(screen->front.pitch) | /* pitch in bytes */ + BUF_3D_USE_FENCE); - /* depth/Z buffer offset/stride */ i915->state.Buffer[I915_DESTREG_DBUFADDR0] = _3DSTATE_BUF_INFO_CMD; - i915->state.Buffer[I915_DESTREG_DBUFADDR1] = - (BUF_3D_ID_DEPTH | - BUF_3D_PITCH(screen->depth.pitch) | /* pitch in bytes */ - BUF_3D_USE_FENCE); - i915->state.Buffer[I915_DESTREG_DBUFADDR2] = screen->depth.offset; - + /* XXX FBO: remove this? Also get set in i915_set_draw_region() */ + i915->state.Buffer[I915_DESTREG_DBUFADDR1] = (BUF_3D_ID_DEPTH | BUF_3D_PITCH(screen->depth.pitch) | /* pitch in bytes */ + BUF_3D_USE_FENCE); i915->state.Buffer[I915_DESTREG_DV0] = _3DSTATE_DST_BUF_VARS_CMD; - /* color/depth pixel format */ + /* XXX FBO: remove this? Also get set in i915_set_draw_region() */ +#if 0 /* seems we don't need this */ switch (screen->fbFormat) { - case DV_PF_555: case DV_PF_565: - i915->state.Buffer[I915_DESTREG_DV1] = (DSTORG_HORT_BIAS(0x8) | /* .5 */ - DSTORG_VERT_BIAS(0x8) | /* .5 */ - LOD_PRECLAMP_OGL | - TEX_DEFAULT_COLOR_OGL | - DITHER_FULL_ALWAYS | - screen->fbFormat | - DEPTH_FRMT_16_FIXED); - break; + i915->state.Buffer[I915_DESTREG_DV1] = (DSTORG_HORT_BIAS(0x8) | /* .5 */ + DSTORG_VERT_BIAS(0x8) | /* .5 */ + LOD_PRECLAMP_OGL | + TEX_DEFAULT_COLOR_OGL | + DITHER_FULL_ALWAYS | + screen->fbFormat | + DEPTH_FRMT_16_FIXED); + break; case DV_PF_8888: - i915->state.Buffer[I915_DESTREG_DV1] = (DSTORG_HORT_BIAS(0x8) | /* .5 */ - DSTORG_VERT_BIAS(0x8) | /* .5 */ - LOD_PRECLAMP_OGL | - TEX_DEFAULT_COLOR_OGL | - screen->fbFormat | - DEPTH_FRMT_24_FIXED_8_OTHER); - break; + i915->state.Buffer[I915_DESTREG_DV1] = (DSTORG_HORT_BIAS(0x8) | /* .5 */ + DSTORG_VERT_BIAS(0x8) | /* .5 */ + LOD_PRECLAMP_OGL | + TEX_DEFAULT_COLOR_OGL | + screen->fbFormat | + DEPTH_FRMT_24_FIXED_8_OTHER); + break; } +#endif + /* scissor */ - i915->state.Buffer[I915_DESTREG_SENABLE] = (_3DSTATE_SCISSOR_ENABLE_CMD | - DISABLE_SCISSOR_RECT); + i915->state.Buffer[I915_DESTREG_SENABLE] = + (_3DSTATE_SCISSOR_ENABLE_CMD | DISABLE_SCISSOR_RECT); i915->state.Buffer[I915_DESTREG_SR0] = _3DSTATE_SCISSOR_RECT_0_CMD; i915->state.Buffer[I915_DESTREG_SR1] = 0; i915->state.Buffer[I915_DESTREG_SR2] = 0; } +#if 0 + { + I915_STATECHANGE(i915, I915_UPLOAD_DEFAULTS); + i915->state.Default[I915_DEFREG_C0] = _3DSTATE_DEFAULT_DIFFUSE; + i915->state.Default[I915_DEFREG_C1] = 0; + i915->state.Default[I915_DEFREG_S0] = _3DSTATE_DEFAULT_SPECULAR; + i915->state.Default[I915_DEFREG_S1] = 0; + i915->state.Default[I915_DEFREG_Z0] = _3DSTATE_DEFAULT_Z; + i915->state.Default[I915_DEFREG_Z1] = 0; + } +#endif + + /* These will be emitted every at the head of every buffer, unless * we get hardware contexts working. */ - i915->state.active = (I915_UPLOAD_PROGRAM | - I915_UPLOAD_STIPPLE | - I915_UPLOAD_CTX | - I915_UPLOAD_BUFFERS | - I915_UPLOAD_INVARIENT); + i915->state.active = (I915_UPLOAD_PROGRAM | + I915_UPLOAD_STIPPLE | + I915_UPLOAD_CTX | + I915_UPLOAD_BUFFERS | I915_UPLOAD_INVARIENT); } -void i915InitStateFunctions( struct dd_function_table *functions ) +void +i915InitStateFunctions(struct dd_function_table *functions) { functions->AlphaFunc = i915AlphaFunc; functions->BlendColor = i915BlendColor; @@ -957,14 +1000,15 @@ void i915InitStateFunctions( struct dd_function_table *functions ) } -void i915InitState( i915ContextPtr i915 ) +void +i915InitState(struct i915_context *i915) { GLcontext *ctx = &i915->intel.ctx; - i915_init_packets( i915 ); + i915_init_packets(i915); _mesa_init_driver_state(ctx); - memcpy( &i915->initial, &i915->state, sizeof(i915->state) ); + memcpy(&i915->initial, &i915->state, sizeof(i915->state)); i915->current = &i915->state; } diff --git a/src/mesa/drivers/dri/i915/i915_tex.c b/src/mesa/drivers/dri/i915/i915_tex.c index d9609d3193..59e148ca04 100644 --- a/src/mesa/drivers/dri/i915/i915_tex.c +++ b/src/mesa/drivers/dri/i915/i915_tex.c @@ -45,109 +45,25 @@ - - - -/** - * Allocate space for and load the mesa images into the texture memory block. - * This will happen before drawing with a new texture, or drawing with a - * texture after it was swapped out or teximaged again. - */ - -intelTextureObjectPtr i915AllocTexObj( struct gl_texture_object *texObj ) +static void +i915TexEnv(GLcontext * ctx, GLenum target, + GLenum pname, const GLfloat * param) { - i915TextureObjectPtr t = CALLOC_STRUCT( i915_texture_object ); - if ( !t ) - return NULL; + struct i915_context *i915 = I915_CONTEXT(ctx); - texObj->DriverData = t; - t->intel.base.tObj = texObj; - t->intel.dirty = I915_UPLOAD_TEX_ALL; - make_empty_list( &t->intel.base ); - return &t->intel; -} - - -static void i915TexParameter( GLcontext *ctx, GLenum target, - struct gl_texture_object *tObj, - GLenum pname, const GLfloat *params ) -{ - i915TextureObjectPtr t = (i915TextureObjectPtr) tObj->DriverData; - switch (pname) { - case GL_TEXTURE_MIN_FILTER: - case GL_TEXTURE_MAG_FILTER: - case GL_TEXTURE_MAX_ANISOTROPY_EXT: - case GL_TEXTURE_WRAP_S: - case GL_TEXTURE_WRAP_T: - case GL_TEXTURE_WRAP_R: - case GL_TEXTURE_BORDER_COLOR: - t->intel.dirty = I915_UPLOAD_TEX_ALL; - break; - - case GL_TEXTURE_COMPARE_MODE: - t->intel.dirty = I915_UPLOAD_TEX_ALL; - break; - case GL_TEXTURE_COMPARE_FUNC: - t->intel.dirty = I915_UPLOAD_TEX_ALL; - break; - - case GL_TEXTURE_BASE_LEVEL: - case GL_TEXTURE_MAX_LEVEL: - case GL_TEXTURE_MIN_LOD: - case GL_TEXTURE_MAX_LOD: - /* The i915 and its successors can do a lot of this without - * reloading the textures. A project for someone? - */ - intelFlush( ctx ); - driSwapOutTextureObject( (driTextureObject *) t ); - t->intel.dirty = I915_UPLOAD_TEX_ALL; - break; - - default: - return; - } -} - - -static void i915TexEnv( GLcontext *ctx, GLenum target, - GLenum pname, const GLfloat *param ) -{ - i915ContextPtr i915 = I915_CONTEXT( ctx ); - GLuint unit = ctx->Texture.CurrentUnit; - - switch (pname) { - case GL_TEXTURE_ENV_COLOR: /* Should be a tracked param */ - case GL_TEXTURE_ENV_MODE: - case GL_COMBINE_RGB: - case GL_COMBINE_ALPHA: - case GL_SOURCE0_RGB: - case GL_SOURCE1_RGB: - case GL_SOURCE2_RGB: - case GL_SOURCE0_ALPHA: - case GL_SOURCE1_ALPHA: - case GL_SOURCE2_ALPHA: - case GL_OPERAND0_RGB: - case GL_OPERAND1_RGB: - case GL_OPERAND2_RGB: - case GL_OPERAND0_ALPHA: - case GL_OPERAND1_ALPHA: - case GL_OPERAND2_ALPHA: - case GL_RGB_SCALE: - case GL_ALPHA_SCALE: - i915->tex_program.translated = 0; - break; - - case GL_TEXTURE_LOD_BIAS: { - int b = (int) ((*param) * 16.0); - if (b > 255) b = 255; - if (b < -256) b = -256; - I915_STATECHANGE(i915, I915_UPLOAD_TEX(unit)); - i915->state.Tex[unit][I915_TEXREG_SS2] &= ~SS2_LOD_BIAS_MASK; - i915->state.Tex[unit][I915_TEXREG_SS2] |= - ((b << SS2_LOD_BIAS_SHIFT) & SS2_LOD_BIAS_MASK); - break; - } + case GL_TEXTURE_LOD_BIAS:{ + GLuint unit = ctx->Texture.CurrentUnit; + GLint b = (int) ((*param) * 16.0); + if (b > 255) + b = 255; + if (b < -256) + b = -256; + I915_STATECHANGE(i915, I915_UPLOAD_TEX(unit)); + i915->lodbias_ss2[unit] = + ((b << SS2_LOD_BIAS_SHIFT) & SS2_LOD_BIAS_MASK); + break; + } default: break; @@ -155,33 +71,8 @@ static void i915TexEnv( GLcontext *ctx, GLenum target, } -static void i915BindTexture( GLcontext *ctx, GLenum target, - struct gl_texture_object *texObj ) -{ - i915TextureObjectPtr tex; - - if (!texObj->DriverData) - i915AllocTexObj( texObj ); - - tex = (i915TextureObjectPtr)texObj->DriverData; - - if (tex->lastTarget != texObj->Target) { - tex->intel.dirty = I915_UPLOAD_TEX_ALL; - tex->lastTarget = texObj->Target; - } - - /* Need this if image format changes between bound textures. - * Could try and shortcircuit by checking for differences in - * state between incoming and outgoing textures: - */ - I915_CONTEXT(ctx)->tex_program.translated = 0; -} - - - -void i915InitTextureFuncs( struct dd_function_table *functions ) +void +i915InitTextureFuncs(struct dd_function_table *functions) { - functions->BindTexture = i915BindTexture; functions->TexEnv = i915TexEnv; - functions->TexParameter = i915TexParameter; } diff --git a/src/mesa/drivers/dri/i915/i915_tex_layout.c b/src/mesa/drivers/dri/i915/i915_tex_layout.c new file mode 100644 index 0000000000..7b761a7b22 --- /dev/null +++ b/src/mesa/drivers/dri/i915/i915_tex_layout.c @@ -0,0 +1,343 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Code to layout images in a mipmap tree for i915 and i945 + * respectively. + */ + +#include "intel_mipmap_tree.h" +#include "intel_tex_layout.h" +#include "macros.h" +#include "intel_context.h" + +#define FILE_DEBUG_FLAG DEBUG_TEXTURE + +static GLint initial_offsets[6][2] = { {0, 0}, +{0, 2}, +{1, 0}, +{1, 2}, +{1, 1}, +{1, 3} +}; + + +static GLint step_offsets[6][2] = { {0, 2}, +{0, 2}, +{-1, 2}, +{-1, 2}, +{-1, 1}, +{-1, 1} +}; + +GLboolean +i915_miptree_layout(struct intel_mipmap_tree * mt) +{ + GLint level; + + switch (mt->target) { + case GL_TEXTURE_CUBE_MAP:{ + const GLuint dim = mt->width0; + GLuint face; + GLuint lvlWidth = mt->width0, lvlHeight = mt->height0; + + assert(lvlWidth == lvlHeight); /* cubemap images are square */ + + /* double pitch for cube layouts */ + mt->pitch = ((dim * mt->cpp * 2 + 3) & ~3) / mt->cpp; + mt->total_height = dim * 4; + + for (level = mt->first_level; level <= mt->last_level; level++) { + intel_miptree_set_level_info(mt, level, 6, + 0, 0, + /*OLD: mt->pitch, mt->total_height,*/ + lvlWidth, lvlHeight, + 1); + lvlWidth /= 2; + lvlHeight /= 2; + } + + for (face = 0; face < 6; face++) { + GLuint x = initial_offsets[face][0] * dim; + GLuint y = initial_offsets[face][1] * dim; + GLuint d = dim; + + for (level = mt->first_level; level <= mt->last_level; level++) { + intel_miptree_set_image_offset(mt, level, face, x, y); + + if (d == 0) + _mesa_printf("cube mipmap %d/%d (%d..%d) is 0x0\n", + face, level, mt->first_level, mt->last_level); + + d >>= 1; + x += step_offsets[face][0] * d; + y += step_offsets[face][1] * d; + } + } + break; + } + case GL_TEXTURE_3D:{ + GLuint width = mt->width0; + GLuint height = mt->height0; + GLuint depth = mt->depth0; + GLuint stack_height = 0; + + /* Calculate the size of a single slice. + */ + mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; + + /* XXX: hardware expects/requires 9 levels at minimum. + */ + for (level = mt->first_level; level <= MAX2(8, mt->last_level); + level++) { + intel_miptree_set_level_info(mt, level, depth, 0, mt->total_height, + width, height, depth); + + + stack_height += MAX2(2, height); + + width = minify(width); + height = minify(height); + depth = minify(depth); + } + + /* Fixup depth image_offsets: + */ + depth = mt->depth0; + for (level = mt->first_level; level <= mt->last_level; level++) { + GLuint i; + for (i = 0; i < depth; i++) + intel_miptree_set_image_offset(mt, level, i, + 0, i * stack_height); + + depth = minify(depth); + } + + + /* Multiply slice size by texture depth for total size. It's + * remarkable how wasteful of memory the i915 texture layouts + * are. They are largely fixed in the i945. + */ + mt->total_height = stack_height * mt->depth0; + break; + } + + default:{ + GLuint width = mt->width0; + GLuint height = mt->height0; + GLuint img_height; + + mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; + mt->total_height = 0; + + for (level = mt->first_level; level <= mt->last_level; level++) { + intel_miptree_set_level_info(mt, level, 1, + 0, mt->total_height, + width, height, 1); + + if (mt->compressed) + img_height = MAX2(1, height / 4); + else + img_height = (MAX2(2, height) + 1) & ~1; + + mt->total_height += img_height; + + width = minify(width); + height = minify(height); + } + break; + } + } + DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, + mt->pitch, + mt->total_height, mt->cpp, mt->pitch * mt->total_height * mt->cpp); + + return GL_TRUE; +} + + +GLboolean +i945_miptree_layout(struct intel_mipmap_tree * mt) +{ + GLint level; + + switch (mt->target) { + case GL_TEXTURE_CUBE_MAP:{ + const GLuint dim = mt->width0; + GLuint face; + GLuint lvlWidth = mt->width0, lvlHeight = mt->height0; + + assert(lvlWidth == lvlHeight); /* cubemap images are square */ + + /* Depending on the size of the largest images, pitch can be + * determined either by the old-style packing of cubemap faces, + * or the final row of 4x4, 2x2 and 1x1 faces below this. + */ + if (dim > 32) + mt->pitch = ((dim * mt->cpp * 2 + 3) & ~3) / mt->cpp; + else + mt->pitch = 14 * 8; + + mt->total_height = dim * 4 + 4; + + /* Set all the levels to effectively occupy the whole rectangular region. + */ + for (level = mt->first_level; level <= mt->last_level; level++) { + intel_miptree_set_level_info(mt, level, 6, + 0, 0, + lvlWidth, lvlHeight, 1); + lvlWidth /= 2; + lvlHeight /= 2; + } + + + for (face = 0; face < 6; face++) { + GLuint x = initial_offsets[face][0] * dim; + GLuint y = initial_offsets[face][1] * dim; + GLuint d = dim; + + if (dim == 4 && face >= 4) { + y = mt->total_height - 4; + x = (face - 4) * 8; + } + else if (dim < 4 && (face > 0 || mt->first_level > 0)) { + y = mt->total_height - 4; + x = face * 8; + } + + for (level = mt->first_level; level <= mt->last_level; level++) { + intel_miptree_set_image_offset(mt, level, face, x, y); + + d >>= 1; + + switch (d) { + case 4: + switch (face) { + case FACE_POS_X: + case FACE_NEG_X: + x += step_offsets[face][0] * d; + y += step_offsets[face][1] * d; + break; + case FACE_POS_Y: + case FACE_NEG_Y: + y += 12; + x -= 8; + break; + case FACE_POS_Z: + case FACE_NEG_Z: + y = mt->total_height - 4; + x = (face - 4) * 8; + break; + } + + case 2: + y = mt->total_height - 4; + x = 16 + face * 8; + break; + + case 1: + x += 48; + break; + + default: + x += step_offsets[face][0] * d; + y += step_offsets[face][1] * d; + break; + } + } + } + break; + } + case GL_TEXTURE_3D:{ + GLuint width = mt->width0; + GLuint height = mt->height0; + GLuint depth = mt->depth0; + GLuint pack_x_pitch, pack_x_nr; + GLuint pack_y_pitch; + GLuint level; + + mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; + mt->total_height = 0; + + pack_y_pitch = MAX2(mt->height0, 2); + pack_x_pitch = mt->pitch; + pack_x_nr = 1; + + for (level = mt->first_level; level <= mt->last_level; level++) { + GLuint nr_images = mt->target == GL_TEXTURE_3D ? depth : 6; + GLint x = 0; + GLint y = 0; + GLint q, j; + + intel_miptree_set_level_info(mt, level, nr_images, + 0, mt->total_height, + width, height, depth); + + for (q = 0; q < nr_images;) { + for (j = 0; j < pack_x_nr && q < nr_images; j++, q++) { + intel_miptree_set_image_offset(mt, level, q, x, y); + x += pack_x_pitch; + } + + x = 0; + y += pack_y_pitch; + } + + + mt->total_height += y; + + if (pack_x_pitch > 4) { + pack_x_pitch >>= 1; + pack_x_nr <<= 1; + assert(pack_x_pitch * pack_x_nr <= mt->pitch); + } + + if (pack_y_pitch > 2) { + pack_y_pitch >>= 1; + } + + width = minify(width); + height = minify(height); + depth = minify(depth); + } + break; + } + + case GL_TEXTURE_1D: + case GL_TEXTURE_2D: + case GL_TEXTURE_RECTANGLE_ARB: + i945_miptree_layout_2d(mt); + break; + default: + _mesa_problem(NULL, "Unexpected tex target in i945_miptree_layout()"); + } + + DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, + mt->pitch, + mt->total_height, mt->cpp, mt->pitch * mt->total_height * mt->cpp); + + return GL_TRUE; +} diff --git a/src/mesa/drivers/dri/i915/i915_texprog.c b/src/mesa/drivers/dri/i915/i915_texprog.c deleted file mode 100644 index f6a8b0205a..0000000000 --- a/src/mesa/drivers/dri/i915/i915_texprog.c +++ /dev/null @@ -1,676 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include <strings.h> - -#include "glheader.h" -#include "macros.h" -#include "enums.h" - -#include "tnl/t_context.h" -#include "intel_batchbuffer.h" - -#include "i915_reg.h" -#include "i915_context.h" -#include "i915_program.h" - -static GLuint translate_tex_src_bit( struct i915_fragment_program *p, - GLubyte bit ) -{ - switch (bit) { - case TEXTURE_1D_BIT: return D0_SAMPLE_TYPE_2D; - case TEXTURE_2D_BIT: return D0_SAMPLE_TYPE_2D; - case TEXTURE_RECT_BIT: return D0_SAMPLE_TYPE_2D; - case TEXTURE_3D_BIT: return D0_SAMPLE_TYPE_VOLUME; - case TEXTURE_CUBE_BIT: return D0_SAMPLE_TYPE_CUBE; - default: i915_program_error(p, "TexSrcBit"); return 0; - } -} - -static GLuint get_source( struct i915_fragment_program *p, - GLenum src, GLuint unit ) -{ - switch (src) { - case GL_TEXTURE: - if (p->src_texture == UREG_BAD) { - - /* TODO: Use D0_CHANNEL_XY where possible. - */ - GLuint dim = translate_tex_src_bit( p, p->ctx->Texture.Unit[unit]._ReallyEnabled); - GLuint sampler = i915_emit_decl(p, REG_TYPE_S, unit, dim); - GLuint texcoord = i915_emit_decl(p, REG_TYPE_T, unit, D0_CHANNEL_ALL); - GLuint tmp = i915_get_temp( p ); - GLuint op = T0_TEXLD; - - if (p->VB->TexCoordPtr[unit]->size == 4) - op = T0_TEXLDP; - - p->src_texture = i915_emit_texld( p, tmp, A0_DEST_CHANNEL_ALL, - sampler, texcoord, op ); - } - - return p->src_texture; - - /* Crossbar: */ - case GL_TEXTURE0: - case GL_TEXTURE1: - case GL_TEXTURE2: - case GL_TEXTURE3: - case GL_TEXTURE4: - case GL_TEXTURE5: - case GL_TEXTURE6: - case GL_TEXTURE7: { - return UREG_BAD; - } - - case GL_CONSTANT: - return i915_emit_const4fv( p, p->ctx->Texture.Unit[unit].EnvColor ); - case GL_PRIMARY_COLOR: - return i915_emit_decl(p, REG_TYPE_T, T_DIFFUSE, D0_CHANNEL_ALL); - case GL_PREVIOUS: - default: - i915_emit_decl(p, - GET_UREG_TYPE(p->src_previous), - GET_UREG_NR(p->src_previous), D0_CHANNEL_ALL); - return p->src_previous; - } -} - - -static GLuint emit_combine_source( struct i915_fragment_program *p, - GLuint mask, - GLuint unit, - GLenum source, - GLenum operand ) -{ - GLuint arg, src; - - src = get_source(p, source, unit); - - switch (operand) { - case GL_ONE_MINUS_SRC_COLOR: - /* Get unused tmp, - * Emit tmp = 1.0 + arg.-x-y-z-w - */ - arg = i915_get_temp( p ); - return i915_emit_arith( p, A0_ADD, arg, mask, 0, - swizzle(src, ONE, ONE, ONE, ONE ), - negate(src, 1,1,1,1), 0); - - case GL_SRC_ALPHA: - if (mask == A0_DEST_CHANNEL_W) - return src; - else - return swizzle( src, W, W, W, W ); - case GL_ONE_MINUS_SRC_ALPHA: - /* Get unused tmp, - * Emit tmp = 1.0 + arg.-w-w-w-w - */ - arg = i915_get_temp( p ); - return i915_emit_arith( p, A0_ADD, arg, mask, 0, - swizzle(src, ONE, ONE, ONE, ONE ), - negate( swizzle(src,W,W,W,W), 1,1,1,1), 0); - case GL_SRC_COLOR: - default: - return src; - } -} - - - -static int nr_args( GLenum mode ) -{ - switch (mode) { - case GL_REPLACE: return 1; - case GL_MODULATE: return 2; - case GL_ADD: return 2; - case GL_ADD_SIGNED: return 2; - case GL_INTERPOLATE: return 3; - case GL_SUBTRACT: return 2; - case GL_DOT3_RGB_EXT: return 2; - case GL_DOT3_RGBA_EXT: return 2; - case GL_DOT3_RGB: return 2; - case GL_DOT3_RGBA: return 2; - default: return 0; - } -} - - -static GLboolean args_match( struct gl_texture_unit *texUnit ) -{ - int i, nr = nr_args(texUnit->Combine.ModeRGB); - - for (i = 0 ; i < nr ; i++) { - if (texUnit->Combine.SourceA[i] != texUnit->Combine.SourceRGB[i]) - return GL_FALSE; - - switch(texUnit->Combine.OperandA[i]) { - case GL_SRC_ALPHA: - switch(texUnit->Combine.OperandRGB[i]) { - case GL_SRC_COLOR: - case GL_SRC_ALPHA: - break; - default: - return GL_FALSE; - } - break; - case GL_ONE_MINUS_SRC_ALPHA: - switch(texUnit->Combine.OperandRGB[i]) { - case GL_ONE_MINUS_SRC_COLOR: - case GL_ONE_MINUS_SRC_ALPHA: - break; - default: - return GL_FALSE; - } - break; - default: - return GL_FALSE; /* impossible */ - } - } - - return GL_TRUE; -} - - -static GLuint emit_combine( struct i915_fragment_program *p, - GLuint dest, - GLuint mask, - GLuint saturate, - GLuint unit, - GLenum mode, - const GLenum *source, - const GLenum *operand) -{ - int tmp, src[3], nr = nr_args(mode); - int i; - - for (i = 0; i < nr; i++) - src[i] = emit_combine_source( p, mask, unit, source[i], operand[i] ); - - switch (mode) { - case GL_REPLACE: - if (mask == A0_DEST_CHANNEL_ALL && !saturate) - return src[0]; - else - return i915_emit_arith( p, A0_MOV, dest, mask, saturate, src[0], 0, 0 ); - case GL_MODULATE: - return i915_emit_arith( p, A0_MUL, dest, mask, saturate, - src[0], src[1], 0 ); - case GL_ADD: - return i915_emit_arith( p, A0_ADD, dest, mask, saturate, - src[0], src[1], 0 ); - case GL_ADD_SIGNED: - /* tmp = arg0 + arg1 - * result = tmp + -.5 - */ - tmp = i915_emit_const1f(p, .5); - tmp = negate(swizzle(tmp,X,X,X,X),1,1,1,1); - i915_emit_arith( p, A0_ADD, dest, mask, 0, src[0], src[1], 0 ); - i915_emit_arith( p, A0_ADD, dest, mask, saturate, dest, tmp, 0 ); - return dest; - case GL_INTERPOLATE: /* TWO INSTRUCTIONS */ - /* Arg0 * (Arg2) + Arg1 * (1-Arg2) - * - * Arg0*Arg2 + Arg1 - Arg1Arg2 - * - * tmp = Arg0*Arg2 + Arg1, - * result = (-Arg1)Arg2 + tmp - */ - tmp = i915_get_temp( p ); - i915_emit_arith( p, A0_MAD, tmp, mask, 0, src[0], src[2], src[1] ); - i915_emit_arith( p, A0_MAD, dest, mask, saturate, - negate(src[1], 1,1,1,1), src[2], tmp ); - return dest; - case GL_SUBTRACT: - /* negate src[1] */ - return i915_emit_arith( p, A0_ADD, dest, mask, saturate, src[0], - negate(src[1],1,1,1,1), 0 ); - - case GL_DOT3_RGBA: - case GL_DOT3_RGBA_EXT: - case GL_DOT3_RGB_EXT: - case GL_DOT3_RGB: { - GLuint tmp0 = i915_get_temp( p ); - GLuint tmp1 = i915_get_temp( p ); - GLuint neg1 = negate(swizzle(i915_emit_const1f(p, 1),X,X,X,X), 1,1,1,1); - GLuint two = swizzle(i915_emit_const1f(p, 2),X,X,X,X); - i915_emit_arith( p, A0_MAD, tmp0, A0_DEST_CHANNEL_ALL, 0, - two, src[0], neg1); - if (src[0] == src[1]) - tmp1 = tmp0; - else - i915_emit_arith( p, A0_MAD, tmp1, A0_DEST_CHANNEL_ALL, 0, - two, src[1], neg1); - i915_emit_arith( p, A0_DP3, dest, mask, saturate, tmp0, tmp1, 0); - return dest; - } - - default: - return src[0]; - } -} - -static GLuint get_dest( struct i915_fragment_program *p, int unit ) -{ - if (p->ctx->_TriangleCaps & DD_SEPARATE_SPECULAR) - return i915_get_temp( p ); - else if (unit != p->last_tex_stage) - return i915_get_temp( p ); - else - return UREG(REG_TYPE_OC, 0); -} - - - -static GLuint emit_texenv( struct i915_fragment_program *p, int unit ) -{ - struct gl_texture_unit *texUnit = &p->ctx->Texture.Unit[unit]; - GLenum envMode = texUnit->EnvMode; - struct gl_texture_object *tObj = texUnit->_Current; - GLenum format = tObj->Image[0][tObj->BaseLevel]->_BaseFormat; - GLuint saturate = unit < p->last_tex_stage ? A0_DEST_SATURATE : 0; - - switch(envMode) { - case GL_BLEND: { - const int cf = get_source(p, GL_PREVIOUS, unit); - const int cc = get_source(p, GL_CONSTANT, unit); - const int cs = get_source(p, GL_TEXTURE, unit); - const int out = get_dest(p, unit); - - if (format == GL_INTENSITY) { - /* cv = cf(1 - cs) + cc.cs - * cv = cf - cf.cs + cc.cs - */ - /* u[2] = MAD( -cf * cs + cf ) - * cv = MAD( cc * cs + u[2] ) - */ - - i915_emit_arith( p, A0_MAD, out, A0_DEST_CHANNEL_ALL, 0, - negate(cf,1,1,1,1), cs, cf ); - - i915_emit_arith( p, A0_MAD, out, A0_DEST_CHANNEL_ALL, saturate, - cc, cs, out ); - - return out; - } else { - /* cv = cf(1 - cs) + cc.cs - * cv = cf - cf.cs + cc.cs - * av = af.as - */ - /* u[2] = MAD( cf.-x-y-zw * cs.xyzw + cf.xyz0 ) - * oC = MAD( cc.xyz0 * cs.xyz0 + u[2].xyzw ) - */ - i915_emit_arith( p, A0_MAD, out, A0_DEST_CHANNEL_ALL, 0, - negate(cf,1,1,1,0), - cs, - swizzle(cf,X,Y,Z,ZERO) ); - - - i915_emit_arith( p, A0_MAD, out, A0_DEST_CHANNEL_ALL, saturate, - swizzle(cc,X,Y,Z,ZERO), - swizzle(cs,X,Y,Z,ZERO), - out ); - - return out; - } - } - - case GL_DECAL: { - if (format == GL_RGB || - format == GL_RGBA) { - int cf = get_source( p, GL_PREVIOUS, unit ); - int cs = get_source( p, GL_TEXTURE, unit ); - int out = get_dest(p, unit); - - /* cv = cf(1-as) + cs.as - * cv = cf.(-as) + cf + cs.as - * av = af - */ - - /* u[2] = mad( cf.xyzw * cs.-w-w-w1 + cf.xyz0 ) - * oc = mad( cs.xyz0 * cs.www0 + u[2].xyzw ) - */ - i915_emit_arith( p, A0_MAD, out, A0_DEST_CHANNEL_ALL, 0, - cf, - negate(swizzle(cs,W,W,W,ONE),1,1,1,0), - swizzle(cf,X,Y,Z,ZERO) ); - - i915_emit_arith( p, A0_MAD, out, A0_DEST_CHANNEL_ALL, saturate, - swizzle(cs,X,Y,Z,ZERO), - swizzle(cs,W,W,W,ZERO), - out ); - return out; - } - else { - return get_source( p, GL_PREVIOUS, unit ); - } - } - - case GL_REPLACE: { - const int cs = get_source( p, GL_TEXTURE, unit ); /* saturated */ - switch (format) { - case GL_ALPHA: { - const int cf = get_source( p, GL_PREVIOUS, unit ); /* saturated */ - i915_emit_arith( p, A0_MOV, cs, A0_DEST_CHANNEL_XYZ, 0, cf, 0, 0 ); - return cs; - } - case GL_RGB: - case GL_LUMINANCE: { - const int cf = get_source( p, GL_PREVIOUS, unit ); /* saturated */ - i915_emit_arith( p, A0_MOV, cs, A0_DEST_CHANNEL_W, 0, cf, 0, 0 ); - return cs; - } - default: - return cs; - } - } - - case GL_MODULATE: { - const int cf = get_source( p, GL_PREVIOUS, unit ); - const int cs = get_source( p, GL_TEXTURE, unit ); - const int out = get_dest(p, unit); - switch (format) { - case GL_ALPHA: - i915_emit_arith( p, A0_MUL, out, A0_DEST_CHANNEL_ALL, saturate, - swizzle(cs, ONE, ONE, ONE, W), cf, 0 ); - break; - default: - i915_emit_arith( p, A0_MUL, out, A0_DEST_CHANNEL_ALL, saturate, - cs, cf, 0 ); - break; - } - return out; - } - case GL_ADD: { - int cf = get_source( p, GL_PREVIOUS, unit ); - int cs = get_source( p, GL_TEXTURE, unit ); - const int out = get_dest( p, unit ); - - if (format == GL_INTENSITY) { - /* output-color.rgba = add( incoming, u[1] ) - */ - i915_emit_arith( p, A0_ADD, out, A0_DEST_CHANNEL_ALL, saturate, - cs, cf, 0 ); - return out; - } - else { - /* cv.xyz = cf.xyz + cs.xyz - * cv.w = cf.w * cs.w - * - * cv.xyzw = MAD( cf.111w * cs.xyzw + cf.xyz0 ) - */ - i915_emit_arith( p, A0_MAD, out, A0_DEST_CHANNEL_ALL, saturate, - swizzle(cf,ONE,ONE,ONE,W), - cs, - swizzle(cf,X,Y,Z,ZERO) ); - return out; - } - break; - } - case GL_COMBINE: { - GLuint rgb_shift, alpha_shift, out, shift; - GLuint dest = get_dest(p, unit); - - /* The EXT version of the DOT3 extension does not support the - * scale factor, but the ARB version (and the version in OpenGL - * 1.3) does. - */ - switch (texUnit->Combine.ModeRGB) { - case GL_DOT3_RGB_EXT: - alpha_shift = texUnit->Combine.ScaleShiftA; - rgb_shift = 0; - break; - - case GL_DOT3_RGBA_EXT: - alpha_shift = 0; - rgb_shift = 0; - break; - - default: - rgb_shift = texUnit->Combine.ScaleShiftRGB; - alpha_shift = texUnit->Combine.ScaleShiftA; - break; - } - - - /* Emit the RGB and A combine ops - */ - if (texUnit->Combine.ModeRGB == texUnit->Combine.ModeA && - args_match( texUnit )) { - out = emit_combine( p, dest, A0_DEST_CHANNEL_ALL, saturate, - unit, - texUnit->Combine.ModeRGB, - texUnit->Combine.SourceRGB, - texUnit->Combine.OperandRGB ); - } - else if (texUnit->Combine.ModeRGB == GL_DOT3_RGBA_EXT || - texUnit->Combine.ModeRGB == GL_DOT3_RGBA) { - - out = emit_combine( p, dest, A0_DEST_CHANNEL_ALL, saturate, - unit, - texUnit->Combine.ModeRGB, - texUnit->Combine.SourceRGB, - texUnit->Combine.OperandRGB ); - } - else { - /* Need to do something to stop from re-emitting identical - * argument calculations here: - */ - out = emit_combine( p, dest, A0_DEST_CHANNEL_XYZ, saturate, - unit, - texUnit->Combine.ModeRGB, - texUnit->Combine.SourceRGB, - texUnit->Combine.OperandRGB ); - out = emit_combine( p, dest, A0_DEST_CHANNEL_W, saturate, - unit, - texUnit->Combine.ModeA, - texUnit->Combine.SourceA, - texUnit->Combine.OperandA ); - } - - /* Deal with the final shift: - */ - if (alpha_shift || rgb_shift) { - if (rgb_shift == alpha_shift) { - shift = i915_emit_const1f(p, 1<<rgb_shift); - shift = swizzle(shift,X,X,X,X); - } - else { - shift = i915_emit_const2f(p, 1<<rgb_shift, 1<<alpha_shift); - shift = swizzle(shift,X,X,X,Y); - } - return i915_emit_arith( p, A0_MUL, dest, A0_DEST_CHANNEL_ALL, - saturate, out, shift, 0 ); - } - - return out; - } - - default: - return get_source(p, GL_PREVIOUS, 0); - } -} - -static void emit_program_fini( struct i915_fragment_program *p ) -{ - int cf = get_source( p, GL_PREVIOUS, 0 ); - int out = UREG( REG_TYPE_OC, 0 ); - - if (p->ctx->_TriangleCaps & DD_SEPARATE_SPECULAR) { - /* Emit specular add. - */ - GLuint s = i915_emit_decl(p, REG_TYPE_T, T_SPECULAR, D0_CHANNEL_ALL); - i915_emit_arith( p, A0_ADD, out, A0_DEST_CHANNEL_ALL, 0, cf, - swizzle(s, X,Y,Z,ZERO), 0 ); - } - else if (cf != out) { - /* Will wind up in here if no texture enabled or a couple of - * other scenarios (GL_REPLACE for instance). - */ - i915_emit_arith( p, A0_MOV, out, A0_DEST_CHANNEL_ALL, 0, cf, 0, 0 ); - } -} - - -static void i915EmitTextureProgram( i915ContextPtr i915 ) -{ - GLcontext *ctx = &i915->intel.ctx; - struct i915_fragment_program *p = &i915->tex_program; - GLuint unit; - - if (0) fprintf(stderr, "%s\n", __FUNCTION__); - - i915_init_program( i915, p ); - - if (ctx->Texture._EnabledUnits) { - for (unit = 0 ; unit < ctx->Const.MaxTextureUnits ; unit++) - if (ctx->Texture.Unit[unit]._ReallyEnabled) { - p->last_tex_stage = unit; - } - - for (unit = 0 ; unit < ctx->Const.MaxTextureUnits; unit++) - if (ctx->Texture.Unit[unit]._ReallyEnabled) { - p->src_previous = emit_texenv( p, unit ); - p->src_texture = UREG_BAD; - p->temp_flag = 0xffff000; - p->temp_flag |= 1 << GET_UREG_NR(p->src_previous); - } - } - - emit_program_fini( p ); - - i915_fini_program( p ); - i915_upload_program( i915, p ); - - p->translated = 1; -} - - -void i915ValidateTextureProgram( i915ContextPtr i915 ) -{ - intelContextPtr intel = &i915->intel; - GLcontext *ctx = &intel->ctx; - TNLcontext *tnl = TNL_CONTEXT(ctx); - struct vertex_buffer *VB = &tnl->vb; - DECLARE_RENDERINPUTS(index_bitset); - int i, offset; - GLuint s4 = i915->state.Ctx[I915_CTXREG_LIS4] & ~S4_VFMT_MASK; - GLuint s2 = S2_TEXCOORD_NONE; - - RENDERINPUTS_COPY( index_bitset, tnl->render_inputs_bitset ); - - /* Important: - */ - VB->AttribPtr[VERT_ATTRIB_POS] = VB->NdcPtr; - intel->vertex_attr_count = 0; - intel->coloroffset = 0; - intel->specoffset = 0; - offset = 0; - - if (i915->current_program) { - i915->current_program->on_hardware = 0; - i915->current_program->params_uptodate = 0; - } - - if (i915->vertex_fog == I915_FOG_PIXEL) { - EMIT_ATTR( _TNL_ATTRIB_POS, EMIT_4F_VIEWPORT, S4_VFMT_XYZW, 16 ); - RENDERINPUTS_CLEAR( index_bitset, _TNL_ATTRIB_FOG ); - } - else if (RENDERINPUTS_TEST_RANGE( index_bitset, _TNL_FIRST_TEX, _TNL_LAST_TEX )) { - EMIT_ATTR( _TNL_ATTRIB_POS, EMIT_4F_VIEWPORT, S4_VFMT_XYZW, 16 ); - } - else { - EMIT_ATTR( _TNL_ATTRIB_POS, EMIT_3F_VIEWPORT, S4_VFMT_XYZ, 12 ); - } - - /* How undefined is undefined? */ - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_POINTSIZE )) { - EMIT_ATTR( _TNL_ATTRIB_POINTSIZE, EMIT_1F, S4_VFMT_POINT_WIDTH, 4 ); - } - - intel->coloroffset = offset / 4; - EMIT_ATTR( _TNL_ATTRIB_COLOR0, EMIT_4UB_4F_BGRA, S4_VFMT_COLOR, 4 ); - - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_COLOR1 ) || - RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_FOG )) { - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_COLOR1 )) { - intel->specoffset = offset / 4; - EMIT_ATTR( _TNL_ATTRIB_COLOR1, EMIT_3UB_3F_BGR, S4_VFMT_SPEC_FOG, 3 ); - } else - EMIT_PAD( 3 ); - - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_FOG )) - EMIT_ATTR( _TNL_ATTRIB_FOG, EMIT_1UB_1F, S4_VFMT_SPEC_FOG, 1 ); - else - EMIT_PAD( 1 ); - } - - if (RENDERINPUTS_TEST_RANGE( index_bitset, _TNL_FIRST_TEX, _TNL_LAST_TEX )) { - for (i = 0; i < 8; i++) { - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX(i) )) { - int sz = VB->TexCoordPtr[i]->size; - - s2 &= ~S2_TEXCOORD_FMT(i, S2_TEXCOORD_FMT0_MASK); - s2 |= S2_TEXCOORD_FMT(i, SZ_TO_HW(sz)); - - EMIT_ATTR( _TNL_ATTRIB_TEX0+i, EMIT_SZ(sz), 0, sz * 4 ); - } - } - } - - /* Only need to change the vertex emit code if there has been a - * statechange to a new hardware vertex format: - */ - if (s2 != i915->state.Ctx[I915_CTXREG_LIS2] || - s4 != i915->state.Ctx[I915_CTXREG_LIS4]) { - - I915_STATECHANGE( i915, I915_UPLOAD_CTX ); - - i915->tex_program.translated = 0; - - /* Must do this *after* statechange, so as not to affect - * buffered vertices reliant on the old state: - */ - intel->vertex_size = _tnl_install_attrs( ctx, - intel->vertex_attrs, - intel->vertex_attr_count, - intel->ViewportMatrix.m, 0 ); - - intel->vertex_size >>= 2; - - i915->state.Ctx[I915_CTXREG_LIS2] = s2; - i915->state.Ctx[I915_CTXREG_LIS4] = s4; - - assert(intel->vtbl.check_vertex_size( intel, intel->vertex_size )); - } - - if (!i915->tex_program.translated || - i915->last_ReallyEnabled != ctx->Texture._EnabledUnits) { - i915EmitTextureProgram( i915 ); - i915->last_ReallyEnabled = ctx->Texture._EnabledUnits; - } -} diff --git a/src/mesa/drivers/dri/i915/i915_texstate.c b/src/mesa/drivers/dri/i915/i915_texstate.c index a19d4b6584..09684e87b4 100644 --- a/src/mesa/drivers/dri/i915/i915_texstate.c +++ b/src/mesa/drivers/dri/i915/i915_texstate.c @@ -25,902 +25,335 @@ * **************************************************************************/ -#include "glheader.h" -#include "macros.h" #include "mtypes.h" -#include "simple_list.h" #include "enums.h" #include "texformat.h" -#include "texstore.h" +#include "dri_bufmgr.h" -#include "mm.h" - -#include "intel_screen.h" -#include "intel_ioctl.h" +#include "intel_mipmap_tree.h" #include "intel_tex.h" #include "i915_context.h" #include "i915_reg.h" -static GLint initial_offsets[6][2] = { {0,0}, - {0,2}, - {1,0}, - {1,2}, - {1,1}, - {1,3} }; - - -static GLint step_offsets[6][2] = { {0,2}, - {0,2}, - {-1,2}, - {-1,2}, - {-1,1}, - {-1,1} }; - - -#define I915_TEX_UNIT_ENABLED(unit) (1<<unit) - -static void i915LayoutTextureImages( i915ContextPtr i915, - struct gl_texture_object *tObj ) -{ - const struct gl_texture_image *baseImage = tObj->Image[0][tObj->BaseLevel]; - i915TextureObjectPtr t = (i915TextureObjectPtr) tObj->DriverData; - GLint firstLevel, lastLevel, numLevels; - GLint i, total_height, pitch; - - /* Compute which mipmap levels we really want to send to the hardware. - */ - driCalculateTextureFirstLastLevel( (driTextureObject *) t ); - - /* Figure out the amount of memory required to hold all the mipmap - * levels. Choose the smallest pitch to accomodate the largest - * mipmap: - */ - firstLevel = t->intel.base.firstLevel; - lastLevel = t->intel.base.lastLevel; - numLevels = lastLevel - firstLevel + 1; - - - - /* All images must be loaded at this pitch. Count the number of - * lines required: - */ - switch (tObj->Target) { - case GL_TEXTURE_CUBE_MAP: { - const GLuint dim = tObj->Image[0][firstLevel]->Width; - GLuint face; - - pitch = dim * t->intel.texelBytes; - pitch *= 2; /* double pitch for cube layouts */ - pitch = (pitch + 3) & ~3; - - total_height = dim * 4; - - for ( face = 0 ; face < 6 ; face++) { - GLuint x = initial_offsets[face][0] * dim; - GLuint y = initial_offsets[face][1] * dim; - GLuint d = dim; - - t->intel.base.dirty_images[face] = ~0; - - assert(tObj->Image[face][firstLevel]->Width == dim); - assert(tObj->Image[face][firstLevel]->Height == dim); - - for (i = 0; i < numLevels; i++) { - t->intel.image[face][i].image = tObj->Image[face][firstLevel + i]; - if (!t->intel.image[face][i].image) { - fprintf(stderr, "no image %d %d\n", face, i); - break; /* can't happen */ - } - - t->intel.image[face][i].offset = - y * pitch + x * t->intel.texelBytes; - t->intel.image[face][i].internalFormat = baseImage->_BaseFormat; - - d >>= 1; - x += step_offsets[face][0] * d; - y += step_offsets[face][1] * d; - } - } - break; - } - case GL_TEXTURE_3D: { - GLuint virtual_height; - GLuint tmp_numLevels = numLevels; - pitch = tObj->Image[0][firstLevel]->Width * t->intel.texelBytes; - pitch = (pitch + 3) & ~3; - t->intel.base.dirty_images[0] = ~0; - - /* Calculate the size of a single slice. Hardware demands a - * minimum of 8 mipmaps, some of which might ultimately not be - * used: - */ - if (tmp_numLevels < 9) - tmp_numLevels = 9; - - virtual_height = tObj->Image[0][firstLevel]->Height; - - for ( total_height = i = 0 ; i < tmp_numLevels ; i++ ) { - t->intel.image[0][i].image = tObj->Image[0][firstLevel + i]; - if (t->intel.image[0][i].image) { - t->intel.image[0][i].offset = total_height * pitch; - t->intel.image[0][i].internalFormat = baseImage->_BaseFormat; - } - - total_height += MAX2(2, virtual_height); - virtual_height >>= 1; - } - - t->intel.depth_pitch = total_height * pitch; - - /* Multiply slice size by texture depth for total size. It's - * remarkable how wasteful of memory all the i8x0 texture - * layouts are. - */ - total_height *= t->intel.image[0][0].image->Depth; - break; - } - default: - pitch = tObj->Image[0][firstLevel]->Width * t->intel.texelBytes; - pitch = (pitch + 3) & ~3; - t->intel.base.dirty_images[0] = ~0; - - for ( total_height = i = 0 ; i < numLevels ; i++ ) { - t->intel.image[0][i].image = tObj->Image[0][firstLevel + i]; - if (!t->intel.image[0][i].image) - break; - - t->intel.image[0][i].offset = total_height * pitch; - t->intel.image[0][i].internalFormat = baseImage->_BaseFormat; - if (t->intel.image[0][i].image->IsCompressed) { - total_height += (t->intel.image[0][i].image->Height + 3) / 4; - } - else - total_height += MAX2(2, t->intel.image[0][i].image->Height); - } - break; - } - - t->intel.Pitch = pitch; - t->intel.base.totalSize = total_height*pitch; - t->intel.max_level = numLevels-1; -} - - -static void i945LayoutTextureImages( i915ContextPtr i915, - struct gl_texture_object *tObj ) -{ - const struct gl_texture_image *baseImage = tObj->Image[0][tObj->BaseLevel]; - i915TextureObjectPtr t = (i915TextureObjectPtr) tObj->DriverData; - GLint firstLevel, lastLevel, numLevels; - GLint i, total_height, pitch, sz, max_offset = 0, offset; - - - /* Compute which mipmap levels we really want to send to the hardware. - */ - driCalculateTextureFirstLastLevel( (driTextureObject *) t ); - - /* Figure out the amount of memory required to hold all the mipmap - * levels. Choose the smallest pitch to accomodate the largest - * mipmap: - */ - firstLevel = t->intel.base.firstLevel; - lastLevel = t->intel.base.lastLevel; - numLevels = lastLevel - firstLevel + 1; - - - - /* All images must be loaded at this pitch. Count the number of - * lines required: - */ - switch (tObj->Target) { - case GL_TEXTURE_CUBE_MAP: { - const GLuint dim = tObj->Image[0][firstLevel]->Width; - GLuint face; - - /* Depending on the size of the largest images, pitch can be - * determined either by the old-style packing of cubemap faces, - * or the final row of 4x4, 2x2 and 1x1 faces below this. - */ - if (dim > 32) { - pitch = dim * t->intel.texelBytes; - pitch *= 2; /* double pitch for cube layouts */ - pitch = (pitch + 3) & ~3; - } - else { - pitch = 14 * 8 * t->intel.texelBytes; /* determined by row of - * little maps at - * bottom */ - } - - total_height = dim * 4 + 4; - - for ( face = 0 ; face < 6 ; face++) { - GLuint x = initial_offsets[face][0] * dim; - GLuint y = initial_offsets[face][1] * dim; - GLuint d = dim; - - if (dim == 4 && face >= 4) { - y = total_height - 4; - x = (face - 4) * 8; - } - else if (dim < 4) { - y = total_height - 4; - x = face * 8; - } - - t->intel.base.dirty_images[face] = ~0; - - assert(tObj->Image[face][firstLevel]->Width == dim); - assert(tObj->Image[face][firstLevel]->Height == dim); - - for (i = 0; i < numLevels; i++) { - - - t->intel.image[face][i].image = tObj->Image[face][firstLevel + i]; - assert(t->intel.image[face][i].image); - - t->intel.image[face][i].offset = - y * pitch + x * t->intel.texelBytes; - t->intel.image[face][i].internalFormat = baseImage->_BaseFormat; - - d >>= 1; - - switch (d) { - case 4: - switch (face) { - case FACE_POS_X: - case FACE_NEG_X: - x += step_offsets[face][0] * d; - y += step_offsets[face][1] * d; - break; - case FACE_POS_Y: - case FACE_NEG_Y: - y += 12; - x -= 8; - break; - case FACE_POS_Z: - case FACE_NEG_Z: - y = total_height - 4; - x = (face - 4) * 8; - break; - } - - case 2: - y = total_height - 4; - x = 16 + face * 8; - break; - - case 1: - x += 48; - break; - - default: - x += step_offsets[face][0] * d; - y += step_offsets[face][1] * d; - break; - } - } - } - max_offset = total_height * pitch; - break; - } - case GL_TEXTURE_3D: { - GLuint depth_packing = 0, depth_pack_pitch; - GLuint tmp_numLevels = numLevels; - pitch = tObj->Image[0][firstLevel]->Width * t->intel.texelBytes; - pitch = (pitch + 3) & ~3; - depth_pack_pitch = pitch; - - t->intel.base.dirty_images[0] = ~0; - - - for ( total_height = i = 0 ; i < tmp_numLevels ; i++ ) { - t->intel.image[0][i].image = tObj->Image[0][firstLevel + i]; - if (!t->intel.image[0][i].image) - break; - - - t->intel.image[0][i].offset = total_height * pitch; - t->intel.image[0][i].internalFormat = baseImage->_BaseFormat; - - - - total_height += MAX2(2, t->intel.image[0][i].image->Height) * - MAX2((t->intel.image[0][i].image->Depth >> depth_packing), 1); - - /* When alignment dominates, can't increase depth packing? - * Or does pitch grow??? What are the alignment constraints, - * anyway? - */ - if (depth_pack_pitch > 4) { - depth_packing++; - depth_pack_pitch <<= 2; - } - } - - max_offset = total_height * pitch; - break; - } - default: - pitch = tObj->Image[0][firstLevel]->Width * t->intel.texelBytes; - pitch = (pitch + 3) & ~3; - t->intel.base.dirty_images[0] = ~0; - max_offset = 0; - - for ( offset = i = 0 ; i < numLevels ; i++ ) { - t->intel.image[0][i].image = tObj->Image[0][firstLevel + i]; - if (!t->intel.image[0][i].image) - break; - - t->intel.image[0][i].offset = offset; - t->intel.image[0][i].internalFormat = baseImage->_BaseFormat; - - if (t->intel.image[0][i].image->IsCompressed) - sz = MAX2(1, t->intel.image[0][i].image->Height/4) * pitch; - else - sz = MAX2(2, t->intel.image[0][i].image->Height) * pitch; - - /* Because the images are packed better, the final offset - * might not be the maximal one: - */ - max_offset = MAX2(max_offset, offset + sz); - - /* LPT change: step right after second mipmap. - */ - if (i == 1) - offset += pitch / 2; - else - offset += sz; - } - break; - } - - t->intel.Pitch = pitch; - t->intel.base.totalSize = max_offset; - t->intel.max_level = numLevels-1; -} - - - - -static void i915SetTexImages( i915ContextPtr i915, - struct gl_texture_object *tObj ) +static GLuint +translate_texture_format(GLuint mesa_format) { - GLuint textureFormat; - i915TextureObjectPtr t = (i915TextureObjectPtr) tObj->DriverData; - const struct gl_texture_image *baseImage = tObj->Image[0][tObj->BaseLevel]; - GLint ss2 = 0; - - switch( baseImage->TexFormat->MesaFormat ) { + switch (mesa_format) { case MESA_FORMAT_L8: - t->intel.texelBytes = 1; - textureFormat = MAPSURF_8BIT | MT_8BIT_L8; - break; - + return MAPSURF_8BIT | MT_8BIT_L8; case MESA_FORMAT_I8: - t->intel.texelBytes = 1; - textureFormat = MAPSURF_8BIT | MT_8BIT_I8; - break; - + return MAPSURF_8BIT | MT_8BIT_I8; case MESA_FORMAT_A8: - t->intel.texelBytes = 1; - textureFormat = MAPSURF_8BIT | MT_8BIT_A8; - break; - + return MAPSURF_8BIT | MT_8BIT_A8; case MESA_FORMAT_AL88: - t->intel.texelBytes = 2; - textureFormat = MAPSURF_16BIT | MT_16BIT_AY88; - break; - + return MAPSURF_16BIT | MT_16BIT_AY88; case MESA_FORMAT_RGB565: - t->intel.texelBytes = 2; - textureFormat = MAPSURF_16BIT | MT_16BIT_RGB565; - break; - + return MAPSURF_16BIT | MT_16BIT_RGB565; case MESA_FORMAT_ARGB1555: - t->intel.texelBytes = 2; - textureFormat = MAPSURF_16BIT | MT_16BIT_ARGB1555; - break; - + return MAPSURF_16BIT | MT_16BIT_ARGB1555; case MESA_FORMAT_ARGB4444: - t->intel.texelBytes = 2; - textureFormat = MAPSURF_16BIT | MT_16BIT_ARGB4444; - break; - + return MAPSURF_16BIT | MT_16BIT_ARGB4444; case MESA_FORMAT_ARGB8888: - t->intel.texelBytes = 4; - textureFormat = MAPSURF_32BIT | MT_32BIT_ARGB8888; - break; - + return MAPSURF_32BIT | MT_32BIT_ARGB8888; case MESA_FORMAT_YCBCR_REV: - t->intel.texelBytes = 2; - textureFormat = (MAPSURF_422 | MT_422_YCRCB_NORMAL); - ss2 |= SS2_COLORSPACE_CONVERSION; - break; - + return (MAPSURF_422 | MT_422_YCRCB_NORMAL); case MESA_FORMAT_YCBCR: - t->intel.texelBytes = 2; - textureFormat = (MAPSURF_422 | MT_422_YCRCB_SWAPY); - ss2 |= SS2_COLORSPACE_CONVERSION; - break; - + return (MAPSURF_422 | MT_422_YCRCB_SWAPY); case MESA_FORMAT_RGB_FXT1: case MESA_FORMAT_RGBA_FXT1: - t->intel.texelBytes = 2; - textureFormat = (MAPSURF_COMPRESSED | MT_COMPRESS_FXT1); - break; - + return (MAPSURF_COMPRESSED | MT_COMPRESS_FXT1); case MESA_FORMAT_Z16: - t->intel.texelBytes = 2; - textureFormat = (MAPSURF_16BIT | MT_16BIT_L16); - break; - + return (MAPSURF_16BIT | MT_16BIT_L16); case MESA_FORMAT_RGBA_DXT1: case MESA_FORMAT_RGB_DXT1: - /* - * DXTn pitches are Width/4 * blocksize in bytes - * for DXT1: blocksize=8 so Width/4*8 = Width * 2 - * for DXT3/5: blocksize=16 so Width/4*16 = Width * 4 - */ - t->intel.texelBytes = 2; - textureFormat = (MAPSURF_COMPRESSED | MT_COMPRESS_DXT1); - break; - + return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT1); case MESA_FORMAT_RGBA_DXT3: - t->intel.texelBytes = 4; - textureFormat = (MAPSURF_COMPRESSED | MT_COMPRESS_DXT2_3); - break; - + return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT2_3); case MESA_FORMAT_RGBA_DXT5: - t->intel.texelBytes = 4; - textureFormat = (MAPSURF_COMPRESSED | MT_COMPRESS_DXT4_5); - break; - -#if 0 + return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT4_5); case MESA_FORMAT_Z24_S8: - t->intel.texelBytes = 4; - textureFormat = (MAPSURF_32BIT | MT_32BIT_xL824); - break; -#endif - + return (MAPSURF_32BIT | MT_32BIT_xL824); default: - fprintf(stderr, "%s: bad image format %x\n", __FUNCTION__, - baseImage->TexFormat->MesaFormat); + fprintf(stderr, "%s: bad image format %x\n", __FUNCTION__, mesa_format); abort(); + return 0; } +} - switch (i915->intel.intelScreen->deviceID) { - case PCI_CHIP_I945_G: - case PCI_CHIP_I945_GM: - case PCI_CHIP_I945_GME: - case PCI_CHIP_G33_G: - case PCI_CHIP_Q33_G: - case PCI_CHIP_Q35_G: - i945LayoutTextureImages( i915, tObj ); - break; - default: - i915LayoutTextureImages( i915, tObj ); - break; - } - - t->Setup[I915_TEXREG_MS3] = - (((tObj->Image[0][t->intel.base.firstLevel]->Height - 1) << MS3_HEIGHT_SHIFT) | - ((tObj->Image[0][t->intel.base.firstLevel]->Width - 1) << MS3_WIDTH_SHIFT) | - textureFormat | - MS3_USE_FENCE_REGS); - - t->Setup[I915_TEXREG_MS4] = - ((((t->intel.Pitch / 4) - 1) << MS4_PITCH_SHIFT) | - MS4_CUBE_FACE_ENA_MASK | - (((t->intel.max_level * 4)) << MS4_MAX_LOD_SHIFT) | - ((tObj->Image[0][t->intel.base.firstLevel]->Depth - 1) << MS4_VOLUME_DEPTH_SHIFT)); - - t->Setup[I915_TEXREG_SS2] &= ~(SS2_COLORSPACE_CONVERSION); - t->Setup[I915_TEXREG_SS2] |= ss2; - - t->intel.dirty = I915_UPLOAD_TEX_ALL; -} /* The i915 (and related graphics cores) do not support GL_CLAMP. The * Intel drivers for "other operating systems" implement GL_CLAMP as * GL_CLAMP_TO_EDGE, so the same is done here. */ -static GLuint translate_wrap_mode( GLenum wrap ) +static GLuint +translate_wrap_mode(GLenum wrap) { - switch( wrap ) { - case GL_REPEAT: return TEXCOORDMODE_WRAP; - case GL_CLAMP: return TEXCOORDMODE_CLAMP_EDGE; /* not quite correct */ - case GL_CLAMP_TO_EDGE: return TEXCOORDMODE_CLAMP_EDGE; - case GL_CLAMP_TO_BORDER: return TEXCOORDMODE_CLAMP_BORDER; - case GL_MIRRORED_REPEAT: return TEXCOORDMODE_MIRROR; - default: return TEXCOORDMODE_WRAP; - } -} - - -/** - */ -static void i915ImportTexObjState( struct gl_texture_object *texObj ) -{ - i915TextureObjectPtr t = (i915TextureObjectPtr)texObj->DriverData; - int minFilt = 0, mipFilt = 0, magFilt = 0, shadow = 0; - - if(INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s\n", __FUNCTION__); - - switch (texObj->MinFilter) { - case GL_NEAREST: - minFilt = FILTER_NEAREST; - mipFilt = MIPFILTER_NONE; - break; - case GL_LINEAR: - minFilt = FILTER_LINEAR; - mipFilt = MIPFILTER_NONE; - break; - case GL_NEAREST_MIPMAP_NEAREST: - minFilt = FILTER_NEAREST; - mipFilt = MIPFILTER_NEAREST; - break; - case GL_LINEAR_MIPMAP_NEAREST: - minFilt = FILTER_LINEAR; - mipFilt = MIPFILTER_NEAREST; - break; - case GL_NEAREST_MIPMAP_LINEAR: - minFilt = FILTER_NEAREST; - mipFilt = MIPFILTER_LINEAR; - break; - case GL_LINEAR_MIPMAP_LINEAR: - minFilt = FILTER_LINEAR; - mipFilt = MIPFILTER_LINEAR; - break; + switch (wrap) { + case GL_REPEAT: + return TEXCOORDMODE_WRAP; + case GL_CLAMP: + return TEXCOORDMODE_CLAMP_EDGE; /* not quite correct */ + case GL_CLAMP_TO_EDGE: + return TEXCOORDMODE_CLAMP_EDGE; + case GL_CLAMP_TO_BORDER: + return TEXCOORDMODE_CLAMP_BORDER; + case GL_MIRRORED_REPEAT: + return TEXCOORDMODE_MIRROR; default: - break; - } - - if ( texObj->MaxAnisotropy > 1.0 ) { - minFilt = FILTER_ANISOTROPIC; - magFilt = FILTER_ANISOTROPIC; - } - else { - switch (texObj->MagFilter) { - case GL_NEAREST: - magFilt = FILTER_NEAREST; - break; - case GL_LINEAR: - magFilt = FILTER_LINEAR; - break; - default: - break; - } - } - - if (texObj->CompareMode == GL_COMPARE_R_TO_TEXTURE_ARB && - texObj->Target != GL_TEXTURE_3D) { - - shadow = SS2_SHADOW_ENABLE; - shadow |= intel_translate_compare_func( texObj->CompareFunc ); - - minFilt = FILTER_4X4_FLAT; - magFilt = FILTER_4X4_FLAT; - } - - - t->Setup[I915_TEXREG_SS2] &= ~(SS2_MIN_FILTER_MASK | - SS2_MIP_FILTER_MASK | - SS2_MAG_FILTER_MASK | - SS2_SHADOW_ENABLE | - SS2_SHADOW_FUNC_MASK); - t->Setup[I915_TEXREG_SS2] |= ((minFilt << SS2_MIN_FILTER_SHIFT) | - (mipFilt << SS2_MIP_FILTER_SHIFT) | - (magFilt << SS2_MAG_FILTER_SHIFT) | - shadow); - - { - GLuint ss3 = t->Setup[I915_TEXREG_SS3] & ~(SS3_TCX_ADDR_MODE_MASK | - SS3_TCY_ADDR_MODE_MASK | - SS3_TCZ_ADDR_MODE_MASK); - GLenum ws = texObj->WrapS; - GLenum wt = texObj->WrapT; - GLenum wr = texObj->WrapR; - - t->refs_border_color = 0; - - if (texObj->Target == GL_TEXTURE_3D && - (texObj->MinFilter != GL_NEAREST || - texObj->MagFilter != GL_NEAREST)) { - - /* Try to mimic GL_CLAMP functionality a little better - - * switch to CLAMP_TO_BORDER whenever a non-NEAREST filter is - * in use. Only do this for 3D textures at the moment -- - * doing it universally would fix the conform texbc.c - * failure, though. - */ - if (ws == GL_CLAMP) ws = GL_CLAMP_TO_BORDER; - if (wt == GL_CLAMP) wt = GL_CLAMP_TO_BORDER; - if (wr == GL_CLAMP) wr = GL_CLAMP_TO_BORDER; - - /* 3D textures don't seem to respect the border color. - * Fallback if there's ever a danger that they might refer to - * it. - */ - if (ws == GL_CLAMP_TO_BORDER) t->refs_border_color = 1; - if (wt == GL_CLAMP_TO_BORDER) t->refs_border_color = 1; - if (wr == GL_CLAMP_TO_BORDER) t->refs_border_color = 1; - } - - ss3 |= translate_wrap_mode(ws) << SS3_TCX_ADDR_MODE_SHIFT; - ss3 |= translate_wrap_mode(wt) << SS3_TCY_ADDR_MODE_SHIFT; - ss3 |= translate_wrap_mode(wr) << SS3_TCZ_ADDR_MODE_SHIFT; - - if (ss3 != t->Setup[I915_TEXREG_SS3]) { - t->intel.dirty = I915_UPLOAD_TEX_ALL; - t->Setup[I915_TEXREG_SS3] = ss3; - } - } - - { - const GLubyte *color = texObj->_BorderChan; - - t->Setup[I915_TEXREG_SS4] = INTEL_PACKCOLOR8888(color[0],color[1], - color[2],color[3]); + return TEXCOORDMODE_WRAP; } } -static void i915_import_tex_unit( i915ContextPtr i915, - i915TextureObjectPtr t, - GLuint unit ) +/* Recalculate all state from scratch. Perhaps not the most + * efficient, but this has gotten complex enough that we need + * something which is understandable and reliable. + */ +static GLboolean +i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3) { - GLuint state[I915_TEX_SETUP_SIZE]; + GLcontext *ctx = &intel->ctx; + struct i915_context *i915 = i915_context(ctx); + struct gl_texture_object *tObj = ctx->Texture.Unit[unit]._Current; + struct intel_texture_object *intelObj = intel_texture_object(tObj); + struct gl_texture_image *firstImage; + GLuint *state = i915->state.Tex[unit], format, pitch; - if(INTEL_DEBUG&DEBUG_TEXTURE) - fprintf(stderr, "%s unit(%d)\n", __FUNCTION__, unit); - - if (i915->intel.CurrentTexObj[unit]) - i915->intel.CurrentTexObj[unit]->base.bound &= ~(1U << unit); + memset(state, 0, sizeof(state)); - i915->intel.CurrentTexObj[unit] = (intelTextureObjectPtr)t; - t->intel.base.bound |= (1 << unit); + /*We need to refcount these. */ - if (t->intel.dirty & I915_UPLOAD_TEX(unit)) { - i915ImportTexObjState( t->intel.base.tObj ); - t->intel.dirty &= ~I915_UPLOAD_TEX(unit); + if (i915->state.tex_buffer[unit] != NULL) { + dri_bo_unreference(i915->state.tex_buffer[unit]); + i915->state.tex_buffer[unit] = NULL; } - state[I915_TEXREG_MS2] = t->intel.TextureOffset; - state[I915_TEXREG_MS3] = t->Setup[I915_TEXREG_MS3]; - state[I915_TEXREG_MS4] = t->Setup[I915_TEXREG_MS4]; - - state[I915_TEXREG_SS2] = (i915->state.Tex[unit][I915_TEXREG_SS2] & - SS2_LOD_BIAS_MASK); - state[I915_TEXREG_SS2] |= (t->Setup[I915_TEXREG_SS2] & ~SS2_LOD_BIAS_MASK); + if (!intelObj->imageOverride && !intel_finalize_mipmap_tree(intel, unit)) + return GL_FALSE; - state[I915_TEXREG_SS3] = (i915->state.Tex[unit][I915_TEXREG_SS3] & - SS3_NORMALIZED_COORDS); - state[I915_TEXREG_SS3] |= (t->Setup[I915_TEXREG_SS3] & - ~(SS3_NORMALIZED_COORDS| - SS3_TEXTUREMAP_INDEX_MASK)); + /* Get first image here, since intelObj->firstLevel will get set in + * the intel_finalize_mipmap_tree() call above. + */ + firstImage = tObj->Image[0][intelObj->firstLevel]; - state[I915_TEXREG_SS3] |= (unit<<SS3_TEXTUREMAP_INDEX_SHIFT); + if (intelObj->imageOverride) { + i915->state.tex_buffer[unit] = NULL; + i915->state.tex_offset[unit] = intelObj->textureOffset; - state[I915_TEXREG_SS4] = t->Setup[I915_TEXREG_SS4]; + switch (intelObj->depthOverride) { + case 32: + format = MAPSURF_32BIT | MT_32BIT_ARGB8888; + break; + case 24: + default: + format = MAPSURF_32BIT | MT_32BIT_XRGB8888; + break; + case 16: + format = MAPSURF_16BIT | MT_16BIT_RGB565; + break; + } + pitch = intelObj->pitchOverride; + } else { + dri_bo_reference(intelObj->mt->region->buffer); + i915->state.tex_buffer[unit] = intelObj->mt->region->buffer; + i915->state.tex_offset[unit] = intel_miptree_image_offset(intelObj->mt, + 0, intelObj-> + firstLevel); - if (memcmp(state, i915->state.Tex[unit], sizeof(state)) != 0) { - I915_STATECHANGE( i915, I915_UPLOAD_TEX(unit) ); - memcpy(i915->state.Tex[unit], state, sizeof(state)); + format = translate_texture_format(firstImage->TexFormat->MesaFormat); + pitch = intelObj->mt->pitch * intelObj->mt->cpp; } -} - + state[I915_TEXREG_MS3] = + (((firstImage->Height - 1) << MS3_HEIGHT_SHIFT) | + ((firstImage->Width - 1) << MS3_WIDTH_SHIFT) | format | + MS3_USE_FENCE_REGS); -static GLboolean enable_tex_common( GLcontext *ctx, GLuint unit ) -{ - i915ContextPtr i915 = I915_CONTEXT(ctx); - struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; - struct gl_texture_object *tObj = texUnit->_Current; - i915TextureObjectPtr t = (i915TextureObjectPtr)tObj->DriverData; - - if (0) fprintf(stderr, "%s %d\n", __FUNCTION__, unit); + state[I915_TEXREG_MS4] = + ((((pitch / 4) - 1) << MS4_PITCH_SHIFT) | MS4_CUBE_FACE_ENA_MASK | + ((((intelObj->lastLevel - intelObj->firstLevel) * 4)) << + MS4_MAX_LOD_SHIFT) | ((firstImage->Depth - 1) << + MS4_VOLUME_DEPTH_SHIFT)); - if (!(i915->state.active & I915_UPLOAD_TEX(unit))) { - I915_ACTIVESTATE(i915, I915_UPLOAD_TEX(unit), GL_TRUE); - } - /* Fallback if there's a texture border */ - if ( tObj->Image[0][tObj->BaseLevel]->Border > 0 ) { - return GL_FALSE; - } + { + GLuint minFilt, mipFilt, magFilt; + switch (tObj->MinFilter) { + case GL_NEAREST: + minFilt = FILTER_NEAREST; + mipFilt = MIPFILTER_NONE; + break; + case GL_LINEAR: + minFilt = FILTER_LINEAR; + mipFilt = MIPFILTER_NONE; + break; + case GL_NEAREST_MIPMAP_NEAREST: + minFilt = FILTER_NEAREST; + mipFilt = MIPFILTER_NEAREST; + break; + case GL_LINEAR_MIPMAP_NEAREST: + minFilt = FILTER_LINEAR; + mipFilt = MIPFILTER_NEAREST; + break; + case GL_NEAREST_MIPMAP_LINEAR: + minFilt = FILTER_NEAREST; + mipFilt = MIPFILTER_LINEAR; + break; + case GL_LINEAR_MIPMAP_LINEAR: + minFilt = FILTER_LINEAR; + mipFilt = MIPFILTER_LINEAR; + break; + default: + return GL_FALSE; + } - /* Update state if this is a different texture object to last - * time. - */ - if (i915->intel.CurrentTexObj[unit] != &t->intel || - (t->intel.dirty & I915_UPLOAD_TEX(unit))) { - i915_import_tex_unit( i915, t, unit); - i915->tex_program.translated = 0; - } + if (tObj->MaxAnisotropy > 1.0) { + minFilt = FILTER_ANISOTROPIC; + magFilt = FILTER_ANISOTROPIC; + } + else { + switch (tObj->MagFilter) { + case GL_NEAREST: + magFilt = FILTER_NEAREST; + break; + case GL_LINEAR: + magFilt = FILTER_LINEAR; + break; + default: + return GL_FALSE; + } + } - return GL_TRUE; -} + state[I915_TEXREG_SS2] = i915->lodbias_ss2[unit]; -static GLboolean enable_tex_rect( GLcontext *ctx, GLuint unit ) -{ - i915ContextPtr i915 = I915_CONTEXT(ctx); - struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; - struct gl_texture_object *tObj = texUnit->_Current; - i915TextureObjectPtr t = (i915TextureObjectPtr)tObj->DriverData; - GLuint ss3 = i915->state.Tex[unit][I915_TEXREG_SS3]; + /* YUV conversion: + */ + if (firstImage->TexFormat->MesaFormat == MESA_FORMAT_YCBCR || + firstImage->TexFormat->MesaFormat == MESA_FORMAT_YCBCR_REV) + state[I915_TEXREG_SS2] |= SS2_COLORSPACE_CONVERSION; - ss3 &= ~SS3_NORMALIZED_COORDS; + /* Shadow: + */ + if (tObj->CompareMode == GL_COMPARE_R_TO_TEXTURE_ARB && + tObj->Target != GL_TEXTURE_3D) { - if (ss3 != i915->state.Tex[unit][I915_TEXREG_SS3]) { - I915_STATECHANGE(i915, I915_UPLOAD_TEX(unit)); - i915->state.Tex[unit][I915_TEXREG_SS3] = ss3; - } + state[I915_TEXREG_SS2] |= + (SS2_SHADOW_ENABLE | + intel_translate_compare_func(tObj->CompareFunc)); - /* Upload teximages (not pipelined) - */ - if (t->intel.base.dirty_images[0]) { - i915SetTexImages( i915, tObj ); - if (!intelUploadTexImages( &i915->intel, &t->intel, 0 )) { - return GL_FALSE; + minFilt = FILTER_4X4_FLAT; + magFilt = FILTER_4X4_FLAT; } - } - - return GL_TRUE; -} + state[I915_TEXREG_SS2] |= ((minFilt << SS2_MIN_FILTER_SHIFT) | + (mipFilt << SS2_MIP_FILTER_SHIFT) | + (magFilt << SS2_MAG_FILTER_SHIFT)); + } -static GLboolean enable_tex_2d( GLcontext *ctx, GLuint unit ) -{ - i915ContextPtr i915 = I915_CONTEXT(ctx); - struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; - struct gl_texture_object *tObj = texUnit->_Current; - i915TextureObjectPtr t = (i915TextureObjectPtr)tObj->DriverData; - GLuint ss3 = i915->state.Tex[unit][I915_TEXREG_SS3]; + { + GLenum ws = tObj->WrapS; + GLenum wt = tObj->WrapT; + GLenum wr = tObj->WrapR; - ss3 |= SS3_NORMALIZED_COORDS; - if (ss3 != i915->state.Tex[unit][I915_TEXREG_SS3]) { - I915_STATECHANGE(i915, I915_UPLOAD_TEX(unit)); - i915->state.Tex[unit][I915_TEXREG_SS3] = ss3; - } - - /* Upload teximages (not pipelined) - */ - if (t->intel.base.dirty_images[0]) { - i915SetTexImages( i915, tObj ); - if (!intelUploadTexImages( &i915->intel, &t->intel, 0 )) { - return GL_FALSE; - } - } + /* 3D textures don't seem to respect the border color. + * Fallback if there's ever a danger that they might refer to + * it. + * + * Effectively this means fallback on 3D clamp or + * clamp_to_border. + */ + if (tObj->Target == GL_TEXTURE_3D && + (tObj->MinFilter != GL_NEAREST || + tObj->MagFilter != GL_NEAREST) && + (ws == GL_CLAMP || + wt == GL_CLAMP || + wr == GL_CLAMP || + ws == GL_CLAMP_TO_BORDER || + wt == GL_CLAMP_TO_BORDER || wr == GL_CLAMP_TO_BORDER)) + return GL_FALSE; - return GL_TRUE; -} -static GLboolean enable_tex_cube( GLcontext *ctx, GLuint unit ) -{ - i915ContextPtr i915 = I915_CONTEXT(ctx); - struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; - struct gl_texture_object *tObj = texUnit->_Current; - i915TextureObjectPtr t = (i915TextureObjectPtr)tObj->DriverData; - GLuint ss3 = i915->state.Tex[unit][I915_TEXREG_SS3]; - GLuint face; - - ss3 |= SS3_NORMALIZED_COORDS; - - if (ss3 != i915->state.Tex[unit][I915_TEXREG_SS3]) { - I915_STATECHANGE(i915, I915_UPLOAD_TEX(unit)); - i915->state.Tex[unit][I915_TEXREG_SS3] = ss3; - } + state[I915_TEXREG_SS3] = ss3; /* SS3_NORMALIZED_COORDS */ - /* Upload teximages (not pipelined) - */ - if ( t->intel.base.dirty_images[0] || t->intel.base.dirty_images[1] || - t->intel.base.dirty_images[2] || t->intel.base.dirty_images[3] || - t->intel.base.dirty_images[4] || t->intel.base.dirty_images[5] ) { - i915SetTexImages( i915, tObj ); - } + state[I915_TEXREG_SS3] |= + ((translate_wrap_mode(ws) << SS3_TCX_ADDR_MODE_SHIFT) | + (translate_wrap_mode(wt) << SS3_TCY_ADDR_MODE_SHIFT) | + (translate_wrap_mode(wr) << SS3_TCZ_ADDR_MODE_SHIFT)); - /* upload (per face) */ - for (face = 0; face < 6; face++) { - if (t->intel.base.dirty_images[face]) { - if (!intelUploadTexImages( &i915->intel, &t->intel, face )) { - return GL_FALSE; - } - } + state[I915_TEXREG_SS3] |= (unit << SS3_TEXTUREMAP_INDEX_SHIFT); } - return GL_TRUE; -} + state[I915_TEXREG_SS4] = INTEL_PACKCOLOR8888(tObj->_BorderChan[0], + tObj->_BorderChan[1], + tObj->_BorderChan[2], + tObj->_BorderChan[3]); -static GLboolean enable_tex_3d( GLcontext *ctx, GLuint unit ) -{ - struct gl_texture_object *tObj = ctx->Texture.Unit[unit]._Current; - i915TextureObjectPtr t = (i915TextureObjectPtr)tObj->DriverData; - /* 3D textures on I915 seem to get bogus border colors, hence this - * fallback: + I915_ACTIVESTATE(i915, I915_UPLOAD_TEX(unit), GL_TRUE); + /* memcmp was already disabled, but definitely won't work as the + * region might now change and that wouldn't be detected: */ - if (t->refs_border_color) - return GL_FALSE; + I915_STATECHANGE(i915, I915_UPLOAD_TEX(unit)); - return GL_TRUE; -} - - - -static GLboolean disable_tex( GLcontext *ctx, GLuint unit ) -{ - i915ContextPtr i915 = I915_CONTEXT(ctx); - - if (i915->state.active & I915_UPLOAD_TEX(unit)) { - I915_ACTIVESTATE(i915, I915_UPLOAD_TEX(unit), GL_FALSE); - } - - /* The old texture is no longer bound to this texture unit. - * Mark it as such. - */ - if ( i915->intel.CurrentTexObj[unit] != NULL ) { - i915->intel.CurrentTexObj[unit]->base.bound &= ~(1U << 0); - i915->intel.CurrentTexObj[unit] = NULL; - } +#if 0 + DBG(TEXTURE, "state[I915_TEXREG_SS2] = 0x%x\n", state[I915_TEXREG_SS2]); + DBG(TEXTURE, "state[I915_TEXREG_SS3] = 0x%x\n", state[I915_TEXREG_SS3]); + DBG(TEXTURE, "state[I915_TEXREG_SS4] = 0x%x\n", state[I915_TEXREG_SS4]); + DBG(TEXTURE, "state[I915_TEXREG_MS2] = 0x%x\n", state[I915_TEXREG_MS2]); + DBG(TEXTURE, "state[I915_TEXREG_MS3] = 0x%x\n", state[I915_TEXREG_MS3]); + DBG(TEXTURE, "state[I915_TEXREG_MS4] = 0x%x\n", state[I915_TEXREG_MS4]); +#endif return GL_TRUE; } -static GLboolean i915UpdateTexUnit( GLcontext *ctx, GLuint unit ) -{ - struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; - if (texUnit->_ReallyEnabled && - INTEL_CONTEXT(ctx)->intelScreen->tex.size < 2048 * 1024) - return GL_FALSE; - switch (texUnit->_ReallyEnabled) { - case TEXTURE_1D_BIT: - case TEXTURE_2D_BIT: - return (enable_tex_2d( ctx, unit ) && - enable_tex_common( ctx, unit )); - case TEXTURE_RECT_BIT: - return (enable_tex_rect( ctx, unit ) && - enable_tex_common( ctx, unit )); - case TEXTURE_CUBE_BIT: - return (enable_tex_cube( ctx, unit ) && - enable_tex_common( ctx, unit )); - case TEXTURE_3D_BIT: - return (enable_tex_2d( ctx, unit ) && - enable_tex_common( ctx, unit ) && - enable_tex_3d( ctx, unit)); - case 0: - return disable_tex( ctx, unit ); - default: - return GL_FALSE; - } -} - -void i915UpdateTextureState( intelContextPtr intel ) +void +i915UpdateTextureState(struct intel_context *intel) { - GLcontext *ctx = &intel->ctx; GLboolean ok = GL_TRUE; GLuint i; - for (i = 0 ; i < I915_TEX_UNITS && ok ; i++) { - ok = i915UpdateTexUnit( ctx, i ); + for (i = 0; i < I915_TEX_UNITS && ok; i++) { + switch (intel->ctx.Texture.Unit[i]._ReallyEnabled) { + case TEXTURE_1D_BIT: + case TEXTURE_2D_BIT: + case TEXTURE_CUBE_BIT: + case TEXTURE_3D_BIT: + ok = i915_update_tex_unit(intel, i, SS3_NORMALIZED_COORDS); + break; + case TEXTURE_RECT_BIT: + ok = i915_update_tex_unit(intel, i, 0); + break; + case 0:{ + struct i915_context *i915 = i915_context(&intel->ctx); + if (i915->state.active & I915_UPLOAD_TEX(i)) + I915_ACTIVESTATE(i915, I915_UPLOAD_TEX(i), GL_FALSE); + + if (i915->state.tex_buffer[i] != NULL) { + dri_bo_unreference(i915->state.tex_buffer[i]); + i915->state.tex_buffer[i] = NULL; + } + + break; + } + default: + ok = GL_FALSE; + break; + } } - FALLBACK( intel, I915_FALLBACK_TEXTURE, !ok ); + FALLBACK(intel, I915_FALLBACK_TEXTURE, !ok); } - - - diff --git a/src/mesa/drivers/dri/i915/i915_vtbl.c b/src/mesa/drivers/dri/i915/i915_vtbl.c index cc8a605e50..35757e17eb 100644 --- a/src/mesa/drivers/dri/i915/i915_vtbl.c +++ b/src/mesa/drivers/dri/i915/i915_vtbl.c @@ -37,115 +37,144 @@ #include "tnl/t_vertex.h" #include "intel_batchbuffer.h" +#include "intel_tex.h" +#include "intel_regions.h" #include "i915_reg.h" #include "i915_context.h" -static void i915_render_start( intelContextPtr intel ) +static void +i915_render_prevalidate(struct intel_context *intel) { - GLcontext *ctx = &intel->ctx; - i915ContextPtr i915 = I915_CONTEXT(intel); + struct i915_context *i915 = i915_context(&intel->ctx); - if (ctx->FragmentProgram._Active) - i915ValidateFragmentProgram( i915 ); - else { - assert(!ctx->FragmentProgram._MaintainTexEnvProgram); - i915ValidateTextureProgram( i915 ); - } + i915ValidateFragmentProgram(i915); +} + +static void +i915_render_start(struct intel_context *intel) +{ } -static void i915_reduced_primitive_state( intelContextPtr intel, - GLenum rprim ) +static void +i915_reduced_primitive_state(struct intel_context *intel, GLenum rprim) { - i915ContextPtr i915 = I915_CONTEXT(intel); - GLuint st1 = i915->state.Stipple[I915_STPREG_ST1]; - - st1 &= ~ST1_ENABLE; - - switch (rprim) { - case GL_QUADS: /* from RASTERIZE(GL_QUADS) in t_dd_tritemp.h */ - case GL_TRIANGLES: - if (intel->ctx.Polygon.StippleFlag && - intel->hw_stipple) - st1 |= ST1_ENABLE; - break; - case GL_LINES: - case GL_POINTS: - default: - break; - } - - i915->intel.reduced_primitive = rprim; - - if (st1 != i915->state.Stipple[I915_STPREG_ST1]) { - I915_STATECHANGE(i915, I915_UPLOAD_STIPPLE); - i915->state.Stipple[I915_STPREG_ST1] = st1; - } + struct i915_context *i915 = i915_context(&intel->ctx); + GLuint st1 = i915->state.Stipple[I915_STPREG_ST1]; + + st1 &= ~ST1_ENABLE; + + switch (rprim) { + case GL_QUADS: /* from RASTERIZE(GL_QUADS) in t_dd_tritemp.h */ + case GL_TRIANGLES: + if (intel->ctx.Polygon.StippleFlag && intel->hw_stipple) + st1 |= ST1_ENABLE; + break; + case GL_LINES: + case GL_POINTS: + default: + break; + } + + i915->intel.reduced_primitive = rprim; + + if (st1 != i915->state.Stipple[I915_STPREG_ST1]) { + INTEL_FIREVERTICES(intel); + + I915_STATECHANGE(i915, I915_UPLOAD_STIPPLE); + i915->state.Stipple[I915_STPREG_ST1] = st1; + } } /* Pull apart the vertex format registers and figure out how large a * vertex is supposed to be. */ -static GLboolean i915_check_vertex_size( intelContextPtr intel, - GLuint expected ) +static GLboolean +i915_check_vertex_size(struct intel_context *intel, GLuint expected) { - i915ContextPtr i915 = I915_CONTEXT(intel); + struct i915_context *i915 = i915_context(&intel->ctx); int lis2 = i915->current->Ctx[I915_CTXREG_LIS2]; int lis4 = i915->current->Ctx[I915_CTXREG_LIS4]; int i, sz = 0; switch (lis4 & S4_VFMT_XYZW_MASK) { - case S4_VFMT_XY: sz = 2; break; - case S4_VFMT_XYZ: sz = 3; break; - case S4_VFMT_XYW: sz = 3; break; - case S4_VFMT_XYZW: sz = 4; break; - default: + case S4_VFMT_XY: + sz = 2; + break; + case S4_VFMT_XYZ: + sz = 3; + break; + case S4_VFMT_XYW: + sz = 3; + break; + case S4_VFMT_XYZW: + sz = 4; + break; + default: fprintf(stderr, "no xyzw specified\n"); return 0; } - if (lis4 & S4_VFMT_SPEC_FOG) sz++; - if (lis4 & S4_VFMT_COLOR) sz++; - if (lis4 & S4_VFMT_DEPTH_OFFSET) sz++; - if (lis4 & S4_VFMT_POINT_WIDTH) sz++; - if (lis4 & S4_VFMT_FOG_PARAM) sz++; - - for (i = 0 ; i < 8 ; i++) { + if (lis4 & S4_VFMT_SPEC_FOG) + sz++; + if (lis4 & S4_VFMT_COLOR) + sz++; + if (lis4 & S4_VFMT_DEPTH_OFFSET) + sz++; + if (lis4 & S4_VFMT_POINT_WIDTH) + sz++; + if (lis4 & S4_VFMT_FOG_PARAM) + sz++; + + for (i = 0; i < 8; i++) { switch (lis2 & S2_TEXCOORD_FMT0_MASK) { - case TEXCOORDFMT_2D: sz += 2; break; - case TEXCOORDFMT_3D: sz += 3; break; - case TEXCOORDFMT_4D: sz += 4; break; - case TEXCOORDFMT_1D: sz += 1; break; - case TEXCOORDFMT_2D_16: sz += 1; break; - case TEXCOORDFMT_4D_16: sz += 2; break; - case TEXCOORDFMT_NOT_PRESENT: break; + case TEXCOORDFMT_2D: + sz += 2; + break; + case TEXCOORDFMT_3D: + sz += 3; + break; + case TEXCOORDFMT_4D: + sz += 4; + break; + case TEXCOORDFMT_1D: + sz += 1; + break; + case TEXCOORDFMT_2D_16: + sz += 1; + break; + case TEXCOORDFMT_4D_16: + sz += 2; + break; + case TEXCOORDFMT_NOT_PRESENT: + break; default: - fprintf(stderr, "bad texcoord fmt %d\n", i); - return GL_FALSE; + fprintf(stderr, "bad texcoord fmt %d\n", i); + return GL_FALSE; } lis2 >>= S2_TEXCOORD_FMT1_SHIFT; } - - if (sz != expected) + + if (sz != expected) fprintf(stderr, "vertex size mismatch %d/%d\n", sz, expected); - + return sz == expected; } -static void i915_emit_invarient_state( intelContextPtr intel ) +static void +i915_emit_invarient_state(struct intel_context *intel) { BATCH_LOCALS; - BEGIN_BATCH( 20 ); + BEGIN_BATCH(200, 0); OUT_BATCH(_3DSTATE_AA_CMD | - AA_LINE_ECAAR_WIDTH_ENABLE | - AA_LINE_ECAAR_WIDTH_1_0 | - AA_LINE_REGION_WIDTH_ENABLE | - AA_LINE_REGION_WIDTH_1_0); + AA_LINE_ECAAR_WIDTH_ENABLE | + AA_LINE_ECAAR_WIDTH_1_0 | + AA_LINE_REGION_WIDTH_ENABLE | AA_LINE_REGION_WIDTH_1_0); OUT_BATCH(_3DSTATE_DFLT_DIFFUSE_CMD); OUT_BATCH(0); @@ -158,35 +187,27 @@ static void i915_emit_invarient_state( intelContextPtr intel ) /* Don't support texture crossbar yet */ OUT_BATCH(_3DSTATE_COORD_SET_BINDINGS | - CSB_TCB(0, 0) | - CSB_TCB(1, 1) | - CSB_TCB(2, 2) | - CSB_TCB(3, 3) | - CSB_TCB(4, 4) | - CSB_TCB(5, 5) | - CSB_TCB(6, 6) | - CSB_TCB(7, 7)); + CSB_TCB(0, 0) | + CSB_TCB(1, 1) | + CSB_TCB(2, 2) | + CSB_TCB(3, 3) | + CSB_TCB(4, 4) | CSB_TCB(5, 5) | CSB_TCB(6, 6) | CSB_TCB(7, 7)); OUT_BATCH(_3DSTATE_RASTER_RULES_CMD | - ENABLE_POINT_RASTER_RULE | - OGL_POINT_RASTER_RULE | - ENABLE_LINE_STRIP_PROVOKE_VRTX | - ENABLE_TRI_FAN_PROVOKE_VRTX | - LINE_STRIP_PROVOKE_VRTX(1) | - TRI_FAN_PROVOKE_VRTX(2) | - ENABLE_TEXKILL_3D_4D | - TEXKILL_4D); + ENABLE_POINT_RASTER_RULE | + OGL_POINT_RASTER_RULE | + ENABLE_LINE_STRIP_PROVOKE_VRTX | + ENABLE_TRI_FAN_PROVOKE_VRTX | + LINE_STRIP_PROVOKE_VRTX(1) | + TRI_FAN_PROVOKE_VRTX(2) | ENABLE_TEXKILL_3D_4D | TEXKILL_4D); /* Need to initialize this to zero. */ - OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | - I1_LOAD_S(3) | - (0)); + OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | I1_LOAD_S(3) | (0)); OUT_BATCH(0); - + /* XXX: Use this */ - OUT_BATCH(_3DSTATE_SCISSOR_ENABLE_CMD | - DISABLE_SCISSOR_RECT); + OUT_BATCH(_3DSTATE_SCISSOR_ENABLE_CMD | DISABLE_SCISSOR_RECT); OUT_BATCH(_3DSTATE_SCISSOR_RECT_0_CMD); OUT_BATCH(0); @@ -194,29 +215,23 @@ static void i915_emit_invarient_state( intelContextPtr intel ) OUT_BATCH(_3DSTATE_DEPTH_SUBRECT_DISABLE); - OUT_BATCH(_3DSTATE_LOAD_INDIRECT | 0); /* disable indirect state */ + OUT_BATCH(_3DSTATE_LOAD_INDIRECT | 0); /* disable indirect state */ OUT_BATCH(0); /* Don't support twosided stencil yet */ - OUT_BATCH(_3DSTATE_BACKFACE_STENCIL_OPS | - BFO_ENABLE_STENCIL_TWO_SIDE | - 0 ); - + OUT_BATCH(_3DSTATE_BACKFACE_STENCIL_OPS | BFO_ENABLE_STENCIL_TWO_SIDE | 0); + OUT_BATCH(0); + ADVANCE_BATCH(); } -#define emit( intel, state, size ) \ -do { \ - int k; \ - BEGIN_BATCH( (size) / sizeof(GLuint)); \ - for (k = 0 ; k < (size) / sizeof(GLuint) ; k++) \ - OUT_BATCH((state)[k]); \ - ADVANCE_BATCH(); \ -} while (0); +#define emit(intel, state, size ) \ + intel_batchbuffer_data(intel->batch, state, size, 0 ) -static GLuint get_dirty( struct i915_hw_state *state ) +static GLuint +get_dirty(struct i915_hw_state *state) { GLuint dirty; @@ -227,45 +242,45 @@ static GLuint get_dirty( struct i915_hw_state *state ) if (dirty & I915_UPLOAD_TEX_ALL) state->emitted &= ~I915_UPLOAD_TEX_ALL; dirty = state->active & ~state->emitted; - return dirty; } -static GLuint get_state_size( struct i915_hw_state *state ) +static GLuint +get_state_size(struct i915_hw_state *state) { GLuint dirty = get_dirty(state); GLuint i; GLuint sz = 0; if (dirty & I915_UPLOAD_INVARIENT) - sz += 20 * sizeof(int); + sz += 30 * 4; if (dirty & I915_UPLOAD_CTX) sz += sizeof(state->Ctx); - if (dirty & I915_UPLOAD_BUFFERS) + if (dirty & I915_UPLOAD_BUFFERS) sz += sizeof(state->Buffer); if (dirty & I915_UPLOAD_STIPPLE) sz += sizeof(state->Stipple); - if (dirty & I915_UPLOAD_FOG) + if (dirty & I915_UPLOAD_FOG) sz += sizeof(state->Fog); if (dirty & I915_UPLOAD_TEX_ALL) { int nr = 0; - for (i = 0; i < I915_TEX_UNITS; i++) - if (dirty & I915_UPLOAD_TEX(i)) - nr++; + for (i = 0; i < I915_TEX_UNITS; i++) + if (dirty & I915_UPLOAD_TEX(i)) + nr++; - sz += (2+nr*3) * sizeof(GLuint) * 2; + sz += (2 + nr * 3) * sizeof(GLuint) * 2; } - if (dirty & I915_UPLOAD_CONSTANTS) + if (dirty & I915_UPLOAD_CONSTANTS) sz += state->ConstantSize * sizeof(GLuint); - if (dirty & I915_UPLOAD_PROGRAM) + if (dirty & I915_UPLOAD_PROGRAM) sz += state->ProgramSize * sizeof(GLuint); return sz; @@ -274,47 +289,83 @@ static GLuint get_state_size( struct i915_hw_state *state ) /* Push the state into the sarea and/or texture memory. */ -static void i915_emit_state( intelContextPtr intel ) +static void +i915_emit_state(struct intel_context *intel) { - i915ContextPtr i915 = I915_CONTEXT(intel); + struct i915_context *i915 = i915_context(&intel->ctx); struct i915_hw_state *state = i915->current; int i; - GLuint dirty = get_dirty(state); - GLuint counter = intel->batch.counter; + GLuint dirty; BATCH_LOCALS; - if (intel->batch.space < get_state_size(state)) { - intelFlushBatch(intel, GL_TRUE); - dirty = get_dirty(state); - counter = intel->batch.counter; - } + /* We don't hold the lock at this point, so want to make sure that + * there won't be a buffer wrap. + * + * It might be better to talk about explicit places where + * scheduling is allowed, rather than assume that it is whenever a + * batchbuffer fills up. + */ + intel_batchbuffer_require_space(intel->batch, get_state_size(state), 0); + + /* Do this here as we may have flushed the batchbuffer above, + * causing more state to be dirty! + */ + dirty = get_dirty(state); + state->emitted |= dirty; - if (VERBOSE) + if (INTEL_DEBUG & DEBUG_STATE) fprintf(stderr, "%s dirty: %x\n", __FUNCTION__, dirty); if (dirty & I915_UPLOAD_INVARIENT) { - if (VERBOSE) fprintf(stderr, "I915_UPLOAD_INVARIENT:\n"); - i915_emit_invarient_state( intel ); + if (INTEL_DEBUG & DEBUG_STATE) + fprintf(stderr, "I915_UPLOAD_INVARIENT:\n"); + i915_emit_invarient_state(intel); } if (dirty & I915_UPLOAD_CTX) { - if (VERBOSE) fprintf(stderr, "I915_UPLOAD_CTX:\n"); - emit( i915, state->Ctx, sizeof(state->Ctx) ); + if (INTEL_DEBUG & DEBUG_STATE) + fprintf(stderr, "I915_UPLOAD_CTX:\n"); + + emit(intel, state->Ctx, sizeof(state->Ctx)); } if (dirty & I915_UPLOAD_BUFFERS) { - if (VERBOSE) fprintf(stderr, "I915_UPLOAD_BUFFERS:\n"); - emit( i915, state->Buffer, sizeof(state->Buffer) ); + if (INTEL_DEBUG & DEBUG_STATE) + fprintf(stderr, "I915_UPLOAD_BUFFERS:\n"); + BEGIN_BATCH(I915_DEST_SETUP_SIZE + 2, 0); + OUT_BATCH(state->Buffer[I915_DESTREG_CBUFADDR0]); + OUT_BATCH(state->Buffer[I915_DESTREG_CBUFADDR1]); + OUT_RELOC(state->draw_region->buffer, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, + state->draw_region->draw_offset); + + if (state->depth_region) { + OUT_BATCH(state->Buffer[I915_DESTREG_DBUFADDR0]); + OUT_BATCH(state->Buffer[I915_DESTREG_DBUFADDR1]); + OUT_RELOC(state->depth_region->buffer, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, + state->depth_region->draw_offset); + } + + OUT_BATCH(state->Buffer[I915_DESTREG_DV0]); + OUT_BATCH(state->Buffer[I915_DESTREG_DV1]); + OUT_BATCH(state->Buffer[I915_DESTREG_SENABLE]); + OUT_BATCH(state->Buffer[I915_DESTREG_SR0]); + OUT_BATCH(state->Buffer[I915_DESTREG_SR1]); + OUT_BATCH(state->Buffer[I915_DESTREG_SR2]); + ADVANCE_BATCH(); } if (dirty & I915_UPLOAD_STIPPLE) { - if (VERBOSE) fprintf(stderr, "I915_UPLOAD_STIPPLE:\n"); - emit( i915, state->Stipple, sizeof(state->Stipple) ); + if (INTEL_DEBUG & DEBUG_STATE) + fprintf(stderr, "I915_UPLOAD_STIPPLE:\n"); + emit(intel, state->Stipple, sizeof(state->Stipple)); } if (dirty & I915_UPLOAD_FOG) { - if (VERBOSE) fprintf(stderr, "I915_UPLOAD_FOG:\n"); - emit( i915, state->Fog, sizeof(state->Fog) ); + if (INTEL_DEBUG & DEBUG_STATE) + fprintf(stderr, "I915_UPLOAD_FOG:\n"); + emit(intel, state->Fog, sizeof(state->Fog)); } /* Combine all the dirty texture state into a single command to @@ -323,141 +374,198 @@ static void i915_emit_state( intelContextPtr intel ) if (dirty & I915_UPLOAD_TEX_ALL) { int nr = 0; - for (i = 0; i < I915_TEX_UNITS; i++) - if (dirty & I915_UPLOAD_TEX(i)) - nr++; + for (i = 0; i < I915_TEX_UNITS; i++) + if (dirty & I915_UPLOAD_TEX(i)) + nr++; - BEGIN_BATCH(2+nr*3); - OUT_BATCH(_3DSTATE_MAP_STATE | (3*nr)); + BEGIN_BATCH(2 + nr * 3, 0); + OUT_BATCH(_3DSTATE_MAP_STATE | (3 * nr)); OUT_BATCH((dirty & I915_UPLOAD_TEX_ALL) >> I915_UPLOAD_TEX_0_SHIFT); - for (i = 0 ; i < I915_TEX_UNITS ; i++) - if (dirty & I915_UPLOAD_TEX(i)) { - OUT_BATCH(state->Tex[i][I915_TEXREG_MS2]); - OUT_BATCH(state->Tex[i][I915_TEXREG_MS3]); - OUT_BATCH(state->Tex[i][I915_TEXREG_MS4]); - } + for (i = 0; i < I915_TEX_UNITS; i++) + if (dirty & I915_UPLOAD_TEX(i)) { + + if (state->tex_buffer[i]) { + OUT_RELOC(state->tex_buffer[i], + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, + state->tex_offset[i]); + } + else if (state == &i915->meta) { + assert(i == 0); + OUT_BATCH(0); + } + else { + OUT_BATCH(state->tex_offset[i]); + } + + OUT_BATCH(state->Tex[i][I915_TEXREG_MS3]); + OUT_BATCH(state->Tex[i][I915_TEXREG_MS4]); + } ADVANCE_BATCH(); - BEGIN_BATCH(2+nr*3); - OUT_BATCH(_3DSTATE_SAMPLER_STATE | (3*nr)); + BEGIN_BATCH(2 + nr * 3, 0); + OUT_BATCH(_3DSTATE_SAMPLER_STATE | (3 * nr)); OUT_BATCH((dirty & I915_UPLOAD_TEX_ALL) >> I915_UPLOAD_TEX_0_SHIFT); - for (i = 0 ; i < I915_TEX_UNITS ; i++) - if (dirty & I915_UPLOAD_TEX(i)) { - OUT_BATCH(state->Tex[i][I915_TEXREG_SS2]); - OUT_BATCH(state->Tex[i][I915_TEXREG_SS3]); - OUT_BATCH(state->Tex[i][I915_TEXREG_SS4]); - } + for (i = 0; i < I915_TEX_UNITS; i++) + if (dirty & I915_UPLOAD_TEX(i)) { + OUT_BATCH(state->Tex[i][I915_TEXREG_SS2]); + OUT_BATCH(state->Tex[i][I915_TEXREG_SS3]); + OUT_BATCH(state->Tex[i][I915_TEXREG_SS4]); + } ADVANCE_BATCH(); } if (dirty & I915_UPLOAD_CONSTANTS) { - if (VERBOSE) fprintf(stderr, "I915_UPLOAD_CONSTANTS:\n"); - emit( i915, state->Constant, state->ConstantSize * sizeof(GLuint) ); + if (INTEL_DEBUG & DEBUG_STATE) + fprintf(stderr, "I915_UPLOAD_CONSTANTS:\n"); + emit(intel, state->Constant, state->ConstantSize * sizeof(GLuint)); } if (dirty & I915_UPLOAD_PROGRAM) { - if (VERBOSE) fprintf(stderr, "I915_UPLOAD_PROGRAM:\n"); + if (INTEL_DEBUG & DEBUG_STATE) + fprintf(stderr, "I915_UPLOAD_PROGRAM:\n"); + + assert((state->Program[0] & 0x1ff) + 2 == state->ProgramSize); - assert((state->Program[0] & 0x1ff)+2 == state->ProgramSize); - - emit( i915, state->Program, state->ProgramSize * sizeof(GLuint) ); - if (VERBOSE) - i915_disassemble_program( state->Program, state->ProgramSize ); + emit(intel, state->Program, state->ProgramSize * sizeof(GLuint)); + if (INTEL_DEBUG & DEBUG_STATE) + i915_disassemble_program(state->Program, state->ProgramSize); } - state->emitted |= dirty; - intel->batch.last_emit_state = counter; - assert(counter == intel->batch.counter); + assert(get_dirty(state) == 0); } -static void i915_destroy_context( intelContextPtr intel ) +static void +i915_destroy_context(struct intel_context *intel) { + GLuint i; + struct i915_context *i915 = i915_context(&intel->ctx); + + for (i = 0; i < I915_TEX_UNITS; i++) { + if (i915->state.tex_buffer[i] != NULL) { + dri_bo_unreference(i915->state.tex_buffer[i]); + i915->state.tex_buffer[i] = NULL; + } + } + _tnl_free_vertices(&intel->ctx); } /** - * Set the color buffer drawing region. + * Set the drawing regions for the color and depth/stencil buffers. + * This involves setting the pitch, cpp and buffer ID/location. + * Also set pixel format for color and Z rendering + * Used for setting both regular and meta state. */ -static void -i915_set_color_region( intelContextPtr intel, const intelRegion *region) +void +i915_state_draw_region(struct intel_context *intel, + struct i915_hw_state *state, + struct intel_region *color_region, + struct intel_region *depth_region) { - i915ContextPtr i915 = I915_CONTEXT(intel); - I915_STATECHANGE( i915, I915_UPLOAD_BUFFERS ); - i915->state.Buffer[I915_DESTREG_CBUFADDR1] = - (BUF_3D_ID_COLOR_BACK | BUF_3D_PITCH(region->pitch) | BUF_3D_USE_FENCE); - i915->state.Buffer[I915_DESTREG_CBUFADDR2] = region->offset; + struct i915_context *i915 = i915_context(&intel->ctx); + GLuint value; + + ASSERT(state == &i915->state || state == &i915->meta); + + if (state->draw_region != color_region) { + intel_region_release(&state->draw_region); + intel_region_reference(&state->draw_region, color_region); + } + if (state->depth_region != depth_region) { + intel_region_release(&state->depth_region); + intel_region_reference(&state->depth_region, depth_region); + } + + /* + * Set stride/cpp values + */ + if (color_region) { + state->Buffer[I915_DESTREG_CBUFADDR0] = _3DSTATE_BUF_INFO_CMD; + state->Buffer[I915_DESTREG_CBUFADDR1] = + (BUF_3D_ID_COLOR_BACK | + BUF_3D_PITCH(color_region->pitch * color_region->cpp) | + BUF_3D_USE_FENCE); + } + + if (depth_region) { + state->Buffer[I915_DESTREG_DBUFADDR0] = _3DSTATE_BUF_INFO_CMD; + state->Buffer[I915_DESTREG_DBUFADDR1] = + (BUF_3D_ID_DEPTH | + BUF_3D_PITCH(depth_region->pitch * depth_region->cpp) | + BUF_3D_USE_FENCE); + } + + /* + * Compute/set I915_DESTREG_DV1 value + */ + value = (DSTORG_HORT_BIAS(0x8) | /* .5 */ + DSTORG_VERT_BIAS(0x8) | /* .5 */ + LOD_PRECLAMP_OGL | TEX_DEFAULT_COLOR_OGL); + if (color_region && color_region->cpp == 4) { + value |= DV_PF_8888; + } + else { + value |= (DITHER_FULL_ALWAYS | DV_PF_565); + } + if (depth_region && depth_region->cpp == 4) { + value |= DEPTH_FRMT_24_FIXED_8_OTHER; + } + else { + value |= DEPTH_FRMT_16_FIXED; + } + state->Buffer[I915_DESTREG_DV1] = value; + + I915_STATECHANGE(i915, I915_UPLOAD_BUFFERS); } -/** - * specify the z-buffer/stencil region - */ static void -i915_set_z_region( intelContextPtr intel, const intelRegion *region) +i915_set_draw_region(struct intel_context *intel, + struct intel_region *color_region, + struct intel_region *depth_region) { - i915ContextPtr i915 = I915_CONTEXT(intel); - I915_STATECHANGE( i915, I915_UPLOAD_BUFFERS ); - i915->state.Buffer[I915_DESTREG_DBUFADDR1] = - (BUF_3D_ID_DEPTH | BUF_3D_PITCH(region->pitch) | BUF_3D_USE_FENCE); - i915->state.Buffer[I915_DESTREG_DBUFADDR2] = region->offset; + struct i915_context *i915 = i915_context(&intel->ctx); + i915_state_draw_region(intel, &i915->state, color_region, depth_region); } -/** - * Set both the color and Z/stencil drawing regions. - * Similar to two previous functions, but don't use I915_STATECHANGE() - */ + static void -i915_update_color_z_regions(intelContextPtr intel, - const intelRegion *colorRegion, - const intelRegion *depthRegion) +i915_lost_hardware(struct intel_context *intel) { - i915ContextPtr i915 = I915_CONTEXT(intel); - - i915->state.Buffer[I915_DESTREG_CBUFADDR1] = - (BUF_3D_ID_COLOR_BACK | BUF_3D_PITCH(colorRegion->pitch) | BUF_3D_USE_FENCE); - i915->state.Buffer[I915_DESTREG_CBUFADDR2] = colorRegion->offset; - - i915->state.Buffer[I915_DESTREG_DBUFADDR1] = - (BUF_3D_ID_DEPTH | - BUF_3D_PITCH(depthRegion->pitch) | /* pitch in bytes */ - BUF_3D_USE_FENCE); - i915->state.Buffer[I915_DESTREG_DBUFADDR2] = depthRegion->offset; + struct i915_context *i915 = i915_context(&intel->ctx); + i915->state.emitted = 0; } - -static void i915_lost_hardware( intelContextPtr intel ) +static GLuint +i915_flush_cmd(void) { - I915_CONTEXT(intel)->state.emitted = 0; + return MI_FLUSH | FLUSH_MAP_CACHE; } -static void i915_emit_flush( intelContextPtr intel ) +static void +i915_assert_not_dirty( struct intel_context *intel ) { - BATCH_LOCALS; - - BEGIN_BATCH(2); - OUT_BATCH( MI_FLUSH | FLUSH_MAP_CACHE | FLUSH_RENDER_CACHE ); - OUT_BATCH( 0 ); - ADVANCE_BATCH(); + struct i915_context *i915 = i915_context(&intel->ctx); + struct i915_hw_state *state = i915->current; + GLuint dirty = get_dirty(state); + assert(!dirty); } -void i915InitVtbl( i915ContextPtr i915 ) +void +i915InitVtbl(struct i915_context *i915) { - i915->intel.vtbl.alloc_tex_obj = i915AllocTexObj; i915->intel.vtbl.check_vertex_size = i915_check_vertex_size; - i915->intel.vtbl.clear_with_tris = i915ClearWithTris; - i915->intel.vtbl.rotate_window = i915RotateWindow; i915->intel.vtbl.destroy = i915_destroy_context; i915->intel.vtbl.emit_state = i915_emit_state; i915->intel.vtbl.lost_hardware = i915_lost_hardware; i915->intel.vtbl.reduced_primitive_state = i915_reduced_primitive_state; i915->intel.vtbl.render_start = i915_render_start; - i915->intel.vtbl.set_color_region = i915_set_color_region; - i915->intel.vtbl.set_z_region = i915_set_z_region; - i915->intel.vtbl.update_color_z_regions = i915_update_color_z_regions; + i915->intel.vtbl.render_prevalidate = i915_render_prevalidate; + i915->intel.vtbl.set_draw_region = i915_set_draw_region; i915->intel.vtbl.update_texture_state = i915UpdateTextureState; - i915->intel.vtbl.emit_flush = i915_emit_flush; + i915->intel.vtbl.flush_cmd = i915_flush_cmd; + i915->intel.vtbl.assert_not_dirty = i915_assert_not_dirty; } - diff --git a/src/mesa/drivers/dri/i915/intel_batchbuffer.c b/src/mesa/drivers/dri/i915/intel_batchbuffer.c index 803b41b256..74c75a3769 100644 --- a/src/mesa/drivers/dri/i915/intel_batchbuffer.c +++ b/src/mesa/drivers/dri/i915/intel_batchbuffer.c @@ -1,6 +1,6 @@ /************************************************************************** * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -25,805 +25,226 @@ * **************************************************************************/ - -#include <stdio.h> -#include <errno.h> - -#include "mtypes.h" -#include "context.h" -#include "enums.h" -#include "vblank.h" - -#include "intel_reg.h" #include "intel_batchbuffer.h" -#include "intel_context.h" - - - +#include "intel_ioctl.h" +#include "intel_decode.h" +#include "intel_reg.h" -/* ================================================================ - * Performance monitoring functions +/* Relocations in kernel space: + * - pass dma buffer seperately + * - memory manager knows how to patch + * - pass list of dependent buffers + * - pass relocation list + * + * Either: + * - get back an offset for buffer to fire + * - memory manager knows how to fire buffer + * + * Really want the buffer to be AGP and pinned. + * */ -static void intel_fill_box( intelContextPtr intel, - GLshort x, GLshort y, - GLshort w, GLshort h, - GLubyte r, GLubyte g, GLubyte b ) -{ - x += intel->drawX; - y += intel->drawY; - - if (x >= 0 && y >= 0 && - x+w < intel->intelScreen->width && - y+h < intel->intelScreen->height) - intelEmitFillBlitLocked( intel, - intel->intelScreen->cpp, - intel->intelScreen->back.pitch, - intel->intelScreen->back.offset, - x, y, w, h, - INTEL_PACKCOLOR(intel->intelScreen->fbFormat, - r,g,b,0xff)); -} +/* Cliprect fence: The highest fence protecting a dma buffer + * containing explicit cliprect information. Like the old drawable + * lock but irq-driven. X server must wait for this fence to expire + * before changing cliprects [and then doing sw rendering?]. For + * other dma buffers, the scheduler will grab current cliprect info + * and mix into buffer. X server must hold the lock while changing + * cliprects??? Make per-drawable. Need cliprects in shared memory + * -- beats storing them with every cmd buffer in the queue. + * + * ==> X server must wait for this fence to expire before touching the + * framebuffer with new cliprects. + * + * ==> Cliprect-dependent buffers associated with a + * cliprect-timestamp. All of the buffers associated with a timestamp + * must go to hardware before any buffer with a newer timestamp. + * + * ==> Dma should be queued per-drawable for correct X/GL + * synchronization. Or can fences be used for this? + * + * Applies to: Blit operations, metaops, X server operations -- X + * server automatically waits on its own dma to complete before + * modifying cliprects ??? + */ -static void intel_draw_performance_boxes( intelContextPtr intel ) +void +intel_batchbuffer_reset(struct intel_batchbuffer *batch) { - /* Purple box for page flipping - */ - if ( intel->perf_boxes & I830_BOX_FLIP ) - intel_fill_box( intel, 4, 4, 8, 8, 255, 0, 255 ); - - /* Red box if we have to wait for idle at any point - */ - if ( intel->perf_boxes & I830_BOX_WAIT ) - intel_fill_box( intel, 16, 4, 8, 8, 255, 0, 0 ); - - /* Blue box: lost context? - */ - if ( intel->perf_boxes & I830_BOX_LOST_CONTEXT ) - intel_fill_box( intel, 28, 4, 8, 8, 0, 0, 255 ); - - /* Yellow box for texture swaps - */ - if ( intel->perf_boxes & I830_BOX_TEXTURE_LOAD ) - intel_fill_box( intel, 40, 4, 8, 8, 255, 255, 0 ); - - /* Green box if hardware never idles (as far as we can tell) - */ - if ( !(intel->perf_boxes & I830_BOX_RING_EMPTY) ) - intel_fill_box( intel, 64, 4, 8, 8, 0, 255, 0 ); + struct intel_context *intel = batch->intel; - - /* Draw bars indicating number of buffers allocated - * (not a great measure, easily confused) - */ -#if 0 - if (intel->dma_used) { - int bar = intel->dma_used / 10240; - if (bar > 100) bar = 100; - if (bar < 1) bar = 1; - intel_fill_box( intel, 4, 16, bar, 4, 196, 128, 128 ); - intel->dma_used = 0; + if (batch->buf != NULL) { + dri_bo_unreference(batch->buf); + batch->buf = NULL; } -#endif - intel->perf_boxes = 0; + batch->buf = dri_bo_alloc(intel->intelScreen->bufmgr, "batchbuffer", + intel->intelScreen->maxBatchSize, 4096, + DRM_BO_FLAG_MEM_TT); + dri_bo_map(batch->buf, GL_TRUE); + batch->map = batch->buf->virtual; + batch->size = intel->intelScreen->maxBatchSize; + batch->ptr = batch->map; } - - - - - -static int bad_prim_vertex_nr( int primitive, int nr ) -{ - switch (primitive & PRIM3D_MASK) { - case PRIM3D_POINTLIST: - return nr < 1; - case PRIM3D_LINELIST: - return (nr & 1) || nr == 0; - case PRIM3D_LINESTRIP: - return nr < 2; - case PRIM3D_TRILIST: - case PRIM3D_RECTLIST: - return nr % 3 || nr == 0; - case PRIM3D_POLY: - case PRIM3D_TRIFAN: - case PRIM3D_TRISTRIP: - case PRIM3D_TRISTRIP_RVRSE: - return nr < 3; - default: - return 1; - } -} - -static void intel_flush_inline_primitive( GLcontext *ctx ) +struct intel_batchbuffer * +intel_batchbuffer_alloc(struct intel_context *intel) { - intelContextPtr intel = INTEL_CONTEXT( ctx ); - GLuint used = intel->batch.ptr - intel->prim.start_ptr; - GLuint vertcount; - - assert(intel->prim.primitive != ~0); - - if (1) { - /* Check vertex size against the vertex we're specifying to - * hardware. If it's wrong, ditch the primitive. - */ - if (!intel->vtbl.check_vertex_size( intel, intel->vertex_size )) - goto do_discard; + struct intel_batchbuffer *batch = calloc(sizeof(*batch), 1); - vertcount = (used - 4)/ (intel->vertex_size * 4); - - if (!vertcount) - goto do_discard; - - if (vertcount * intel->vertex_size * 4 != used - 4) { - fprintf(stderr, "vertex size confusion %d %d\n", used, - intel->vertex_size * vertcount * 4); - goto do_discard; - } + batch->intel = intel; + batch->last_fence = NULL; + intel_batchbuffer_reset(batch); - if (bad_prim_vertex_nr( intel->prim.primitive, vertcount )) { - fprintf(stderr, "bad_prim_vertex_nr %x %d\n", intel->prim.primitive, - vertcount); - goto do_discard; - } - } - - if (used < 8) - goto do_discard; - - *(int *)intel->prim.start_ptr = (_3DPRIMITIVE | - intel->prim.primitive | - (used/4-2)); - - goto finished; - - do_discard: - intel->batch.ptr -= used; - intel->batch.space += used; - assert(intel->batch.space >= 0); - - finished: - intel->prim.primitive = ~0; - intel->prim.start_ptr = 0; - intel->prim.flush = 0; + return batch; } - -/* Emit a primitive referencing vertices in a vertex buffer. - */ -void intelStartInlinePrimitive( intelContextPtr intel, GLuint prim ) +void +intel_batchbuffer_free(struct intel_batchbuffer *batch) { - BATCH_LOCALS; - - if (0) - fprintf(stderr, "%s %x\n", __FUNCTION__, prim); - - - /* Finish any in-progress primitive: - */ - INTEL_FIREVERTICES( intel ); - - /* Emit outstanding state: - */ - intel->vtbl.emit_state( intel ); - - /* Make sure there is some space in this buffer: - */ - if (intel->vertex_size * 10 * sizeof(GLuint) >= intel->batch.space) { - intelFlushBatch(intel, GL_TRUE); - intel->vtbl.emit_state( intel ); - } - -#if 1 - if (((unsigned long)intel->batch.ptr) & 0x4) { - BEGIN_BATCH(1); - OUT_BATCH(0); - ADVANCE_BATCH(); - } -#endif - - /* Emit a slot which will be filled with the inline primitive - * command later. - */ - BEGIN_BATCH(2); - OUT_BATCH( 0 ); - - intel->prim.start_ptr = batch_ptr; - intel->prim.primitive = prim; - intel->prim.flush = intel_flush_inline_primitive; - intel->batch.contains_geometry = 1; - - OUT_BATCH( 0 ); - ADVANCE_BATCH(); + if (batch->last_fence) { + dri_fence_wait(batch->last_fence); + dri_fence_unreference(batch->last_fence); + batch->last_fence = NULL; + } + if (batch->map) { + dri_bo_unmap(batch->buf); + batch->map = NULL; + } + dri_bo_unreference(batch->buf); + batch->buf = NULL; + free(batch); } -void intelRestartInlinePrimitive( intelContextPtr intel ) -{ - GLuint prim = intel->prim.primitive; - intel_flush_inline_primitive( &intel->ctx ); - if (1) intelFlushBatch(intel, GL_TRUE); /* GL_TRUE - is critical */ - intelStartInlinePrimitive( intel, prim ); -} - - - -void intelWrapInlinePrimitive( intelContextPtr intel ) -{ - GLuint prim = intel->prim.primitive; - - if (0) - fprintf(stderr, "%s\n", __FUNCTION__); - intel_flush_inline_primitive( &intel->ctx ); - intelFlushBatch(intel, GL_TRUE); - intelStartInlinePrimitive( intel, prim ); -} - - -/* Emit a primitive with space for inline vertices. +/* TODO: Push this whole function into bufmgr. */ -GLuint *intelEmitInlinePrimitiveLocked(intelContextPtr intel, - int primitive, - int dwords, - int vertex_size ) +static void +do_flush_locked(struct intel_batchbuffer *batch, + GLuint used, + GLboolean ignore_cliprects, GLboolean allow_unlock) { - GLuint *tmp = 0; - BATCH_LOCALS; - - if (0) - fprintf(stderr, "%s 0x%x %d\n", __FUNCTION__, primitive, dwords); - - /* Emit outstanding state: - */ - intel->vtbl.emit_state( intel ); + struct intel_context *intel = batch->intel; + void *start; + GLuint count; - if ((1+dwords)*4 >= intel->batch.space) { - intelFlushBatch(intel, GL_TRUE); - intel->vtbl.emit_state( intel ); - } - - - if (1) { - int used = dwords * 4; - int vertcount; - - /* Check vertex size against the vertex we're specifying to - * hardware. If it's wrong, ditch the primitive. - */ - if (!intel->vtbl.check_vertex_size( intel, vertex_size )) - goto do_discard; - - vertcount = dwords / vertex_size; - - if (dwords % vertex_size) { - fprintf(stderr, "did not request a whole number of vertices\n"); - goto do_discard; - } - - if (bad_prim_vertex_nr( primitive, vertcount )) { - fprintf(stderr, "bad_prim_vertex_nr %x %d\n", primitive, vertcount); - goto do_discard; - } + start = dri_process_relocs(batch->buf, &count); - if (used < 8) - goto do_discard; - } + batch->map = NULL; + batch->ptr = NULL; + batch->flags = 0; - /* Emit 3D_PRIMITIVE commands: + /* Throw away non-effective packets. Won't work once we have + * hardware contexts which would preserve statechanges beyond a + * single buffer. */ - BEGIN_BATCH(1 + dwords); - OUT_BATCH( _3DPRIMITIVE | - primitive | - (dwords-1) ); - - tmp = (GLuint *)batch_ptr; - batch_ptr += dwords * 4; - - ADVANCE_BATCH(); - - intel->batch.contains_geometry = 1; - - do_discard: - return tmp; -} - - -static void intelWaitForFrameCompletion( intelContextPtr intel ) -{ - drm_i915_sarea_t *sarea = (drm_i915_sarea_t *)intel->sarea; - - if (intel->do_irqs) { - if (intelGetLastFrame(intel) < sarea->last_dispatch) { - if (!intel->irqsEmitted) { - while (intelGetLastFrame (intel) < sarea->last_dispatch) - ; - } - else { - intelWaitIrq( intel, intel->alloc.irq_emitted ); - } - intel->irqsEmitted = 10; - } - - if (intel->irqsEmitted) { - LOCK_HARDWARE( intel ); - intelEmitIrqLocked( intel ); - intel->irqsEmitted--; - UNLOCK_HARDWARE( intel ); - } - } - else { - while (intelGetLastFrame (intel) < sarea->last_dispatch) { - if (intel->do_usleeps) - DO_USLEEP( 1 ); - } - } -} - -/* - * Copy the back buffer to the front buffer. - */ -void intelCopyBuffer( const __DRIdrawablePrivate *dPriv, - const drm_clip_rect_t *rect) -{ - intelContextPtr intel; - const intelScreenPrivate *intelScreen; - GLboolean missed_target; - int64_t ust; - - if (0) - fprintf(stderr, "%s\n", __FUNCTION__); - - assert(dPriv); - assert(dPriv->driContextPriv); - assert(dPriv->driContextPriv->driverPrivate); - - intel = (intelContextPtr) dPriv->driContextPriv->driverPrivate; - - intelFlush( &intel->ctx ); - - intelScreen = intel->intelScreen; - - if (!rect && !intel->swap_scheduled && intelScreen->drmMinor >= 6 && - !(intel->vblank_flags & VBLANK_FLAG_NO_IRQ) && - intelScreen->current_rotation == 0) { - unsigned int interval = driGetVBlankInterval(dPriv, intel->vblank_flags); - unsigned int target; - drm_i915_vblank_swap_t swap; - - swap.drawable = dPriv->hHWDrawable; - swap.seqtype = DRM_VBLANK_ABSOLUTE; - target = swap.sequence = intel->vbl_seq + interval; - if (intel->vblank_flags & VBLANK_FLAG_SYNC) { - swap.seqtype |= DRM_VBLANK_NEXTONMISS; - } else if (interval == 0) { - goto noschedule; + if (!(intel->numClipRects == 0 && !ignore_cliprects)) { + if (intel->intelScreen->ttm == GL_TRUE) { + intel_exec_ioctl(batch->intel, + used, ignore_cliprects, allow_unlock, + start, count, &batch->last_fence); + } else { + intel_batch_ioctl(batch->intel, + batch->buf->offset, + used, ignore_cliprects, allow_unlock); } - - if ( intel->vblank_flags & VBLANK_FLAG_SECONDARY ) { - swap.seqtype |= DRM_VBLANK_SECONDARY; - } - - if (!drmCommandWriteRead(intel->driFd, DRM_I915_VBLANK_SWAP, &swap, - sizeof(swap))) { - intel->swap_scheduled = 1; - intel->vbl_seq = swap.sequence; - swap.sequence -= target; - missed_target = swap.sequence > 0 && swap.sequence <= (1 << 23); - } - } else { - intel->swap_scheduled = 0; } -noschedule: - - if (!intel->swap_scheduled) { - intelWaitForFrameCompletion( intel ); - LOCK_HARDWARE( intel ); - - if (!rect) - { - UNLOCK_HARDWARE( intel ); - driWaitForVBlank( dPriv, &intel->vbl_seq, intel->vblank_flags, & missed_target ); - LOCK_HARDWARE( intel ); - } - { - const intelScreenPrivate *intelScreen = intel->intelScreen; - const __DRIdrawablePrivate *dPriv = intel->driDrawable; - const int nbox = dPriv->numClipRects; - const drm_clip_rect_t *pbox = dPriv->pClipRects; - drm_clip_rect_t box; - const int cpp = intelScreen->cpp; - const int pitch = intelScreen->front.pitch; /* in bytes */ - int i; - GLuint CMD, BR13; - BATCH_LOCALS; - - switch(cpp) { - case 2: - BR13 = (pitch) | (0xCC << 16) | (1<<24); - CMD = XY_SRC_COPY_BLT_CMD; - break; - case 4: - BR13 = (pitch) | (0xCC << 16) | (1<<24) | (1<<25); - CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | - XY_SRC_COPY_BLT_WRITE_RGB); - break; - default: - BR13 = (pitch) | (0xCC << 16) | (1<<24); - CMD = XY_SRC_COPY_BLT_CMD; - break; - } - - if (0) - intel_draw_performance_boxes( intel ); - - for (i = 0 ; i < nbox; i++, pbox++) - { - if (pbox->x1 > pbox->x2 || - pbox->y1 > pbox->y2 || - pbox->x2 > intelScreen->width || - pbox->y2 > intelScreen->height) { - _mesa_warning(&intel->ctx, "Bad cliprect in intelCopyBuffer()"); - continue; - } - - box = *pbox; - - if (rect) - { - if (rect->x1 > box.x1) - box.x1 = rect->x1; - if (rect->y1 > box.y1) - box.y1 = rect->y1; - if (rect->x2 < box.x2) - box.x2 = rect->x2; - if (rect->y2 < box.y2) - box.y2 = rect->y2; - - if (box.x1 > box.x2 || box.y1 > box.y2) - continue; - } - - BEGIN_BATCH( 8); - OUT_BATCH( CMD ); - OUT_BATCH( BR13 ); - OUT_BATCH( (box.y1 << 16) | box.x1 ); - OUT_BATCH( (box.y2 << 16) | box.x2 ); - - if (intel->sarea->pf_current_page == 0) - OUT_BATCH( intelScreen->front.offset ); - else - OUT_BATCH( intelScreen->back.offset ); - - OUT_BATCH( (box.y1 << 16) | box.x1 ); - OUT_BATCH( BR13 & 0xffff ); - - if (intel->sarea->pf_current_page == 0) - OUT_BATCH( intelScreen->back.offset ); - else - OUT_BATCH( intelScreen->front.offset ); + + dri_post_submit(batch->buf, &batch->last_fence); - ADVANCE_BATCH(); - } + if (intel->numClipRects == 0 && !ignore_cliprects) { + if (allow_unlock) { + /* If we are not doing any actual user-visible rendering, + * do a sched_yield to keep the app from pegging the cpu while + * achieving nothing. + */ + UNLOCK_HARDWARE(intel); + sched_yield(); + LOCK_HARDWARE(intel); } - intelFlushBatchLocked( intel, GL_TRUE, GL_TRUE, GL_TRUE ); - UNLOCK_HARDWARE( intel ); + intel->vtbl.lost_hardware(intel); } - if (!rect) - { - intel->swap_count++; - (*dri_interface->getUST)(&ust); - if (missed_target) { - intel->swap_missed_count++; - intel->swap_missed_ust = ust - intel->swap_ust; - } - - intel->swap_ust = ust; + if (INTEL_DEBUG & DEBUG_BATCH) { + // dri_bo_map(batch->buf, GL_FALSE); + // intel_decode(ptr, used / 4, batch->buf->offset, + // intel->intelScreen->deviceID); + // dri_bo_unmap(batch->buf); } } - - - -void intelEmitFillBlitLocked( intelContextPtr intel, - GLuint cpp, - GLshort dst_pitch, /* in bytes */ - GLuint dst_offset, - GLshort x, GLshort y, - GLshort w, GLshort h, - GLuint color ) +void +intel_batchbuffer_flush(struct intel_batchbuffer *batch) { - GLuint BR13, CMD; - BATCH_LOCALS; + struct intel_context *intel = batch->intel; + GLuint used = batch->ptr - batch->map; + GLboolean was_locked = intel->locked; - switch(cpp) { - case 1: - case 2: - case 3: - BR13 = dst_pitch | (0xF0 << 16) | (1<<24); - CMD = XY_COLOR_BLT_CMD; - break; - case 4: - BR13 = dst_pitch | (0xF0 << 16) | (1<<24) | (1<<25); - CMD = (XY_COLOR_BLT_CMD | XY_COLOR_BLT_WRITE_ALPHA | - XY_COLOR_BLT_WRITE_RGB); - break; - default: + if (used == 0) return; - } - - BEGIN_BATCH( 6); - OUT_BATCH( CMD ); - OUT_BATCH( BR13 ); - OUT_BATCH( (y << 16) | x ); - OUT_BATCH( ((y+h) << 16) | (x+w) ); - OUT_BATCH( dst_offset ); - OUT_BATCH( color ); - ADVANCE_BATCH(); -} - -/* Copy BitBlt - */ -void intelEmitCopyBlitLocked( intelContextPtr intel, - GLuint cpp, - GLshort src_pitch, - GLuint src_offset, - GLshort dst_pitch, - GLuint dst_offset, - GLshort src_x, GLshort src_y, - GLshort dst_x, GLshort dst_y, - GLshort w, GLshort h ) -{ - GLuint CMD, BR13; - int dst_y2 = dst_y + h; - int dst_x2 = dst_x + w; - BATCH_LOCALS; - - src_pitch *= cpp; - dst_pitch *= cpp; - - switch(cpp) { - case 1: - case 2: - case 3: - BR13 = dst_pitch | (0xCC << 16) | (1<<24); - CMD = XY_SRC_COPY_BLT_CMD; - break; - case 4: - BR13 = dst_pitch | (0xCC << 16) | (1<<24) | (1<<25); - CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | - XY_SRC_COPY_BLT_WRITE_RGB); - break; - default: - return; + /* Add the MI_BATCH_BUFFER_END. Always add an MI_FLUSH - this is a + * performance drain that we would like to avoid. + */ + if (used & 4) { + ((int *) batch->ptr)[0] = intel->vtbl.flush_cmd(); + ((int *) batch->ptr)[1] = 0; + ((int *) batch->ptr)[2] = MI_BATCH_BUFFER_END; + used += 12; } - - if (dst_y2 < dst_y || - dst_x2 < dst_x) { - return; + else { + ((int *) batch->ptr)[0] = intel->vtbl.flush_cmd(); + ((int *) batch->ptr)[1] = MI_BATCH_BUFFER_END; + used += 8; } - BEGIN_BATCH( 12); - OUT_BATCH( CMD ); - OUT_BATCH( BR13 ); - OUT_BATCH( (dst_y << 16) | dst_x ); - OUT_BATCH( (dst_y2 << 16) | dst_x2 ); - OUT_BATCH( dst_offset ); - OUT_BATCH( (src_y << 16) | src_x ); - OUT_BATCH( src_pitch ); - OUT_BATCH( src_offset ); - ADVANCE_BATCH(); -} + /* TODO: Just pass the relocation list and dma buffer up to the + * kernel. + */ + if (!was_locked) + LOCK_HARDWARE(intel); + do_flush_locked(batch, used, !(batch->flags & INTEL_BATCH_CLIPRECTS), + GL_FALSE); + + if (!was_locked) + UNLOCK_HARDWARE(intel); + /* Reset the buffer: + */ + intel_batchbuffer_reset(batch); +} -void intelClearWithBlit(GLcontext *ctx, GLbitfield buffers, GLboolean allFoo, - GLint cx1Foo, GLint cy1Foo, GLint cwFoo, GLint chFoo) +void +intel_batchbuffer_finish(struct intel_batchbuffer *batch) { - intelContextPtr intel = INTEL_CONTEXT( ctx ); - intelScreenPrivate *intelScreen = intel->intelScreen; - GLuint clear_depth, clear_color; - GLint cx, cy, cw, ch; - GLboolean all; - GLint pitch; - GLint cpp = intelScreen->cpp; - GLint i; - GLuint BR13, CMD, D_CMD; - BATCH_LOCALS; - - intelFlush( &intel->ctx ); - LOCK_HARDWARE( intel ); - - /* get clear bounds after locking */ - cx = intel->ctx.DrawBuffer->_Xmin; - cy = intel->ctx.DrawBuffer->_Ymin; - cw = intel->ctx.DrawBuffer->_Xmax - cx; - ch = intel->ctx.DrawBuffer->_Ymax - cy; - all = (cw == intel->ctx.DrawBuffer->Width && - ch == intel->ctx.DrawBuffer->Height); - - pitch = intelScreen->front.pitch; - - clear_color = intel->ClearColor; - clear_depth = 0; - - if (buffers & BUFFER_BIT_DEPTH) { - clear_depth = (GLuint)(ctx->Depth.Clear * intel->ClearDepth); - } - - if (buffers & BUFFER_BIT_STENCIL) { - clear_depth |= (ctx->Stencil.Clear & 0xff) << 24; - } - - switch(cpp) { - case 2: - BR13 = (0xF0 << 16) | (pitch) | (1<<24); - D_CMD = CMD = XY_COLOR_BLT_CMD; - break; - case 4: - BR13 = (0xF0 << 16) | (pitch) | (1<<24) | (1<<25); - CMD = (XY_COLOR_BLT_CMD | - XY_COLOR_BLT_WRITE_ALPHA | - XY_COLOR_BLT_WRITE_RGB); - D_CMD = XY_COLOR_BLT_CMD; - if (buffers & BUFFER_BIT_DEPTH) D_CMD |= XY_COLOR_BLT_WRITE_RGB; - if (buffers & BUFFER_BIT_STENCIL) D_CMD |= XY_COLOR_BLT_WRITE_ALPHA; - break; - default: - BR13 = (0xF0 << 16) | (pitch) | (1<<24); - D_CMD = CMD = XY_COLOR_BLT_CMD; - break; - } - - { - /* flip top to bottom */ - cy = intel->driDrawable->h - cy - ch; - cx = cx + intel->drawX; - cy += intel->drawY; - - /* adjust for page flipping */ - if ( intel->sarea->pf_current_page == 1 ) { - GLuint tmp = buffers; - - buffers &= ~(BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_BACK_LEFT); - if ( tmp & BUFFER_BIT_FRONT_LEFT ) buffers |= BUFFER_BIT_BACK_LEFT; - if ( tmp & BUFFER_BIT_BACK_LEFT ) buffers |= BUFFER_BIT_FRONT_LEFT; - } - - for (i = 0 ; i < intel->numClipRects ; i++) - { - drm_clip_rect_t *box = &intel->pClipRects[i]; - drm_clip_rect_t b; - - if (!all) { - GLint x = box->x1; - GLint y = box->y1; - GLint w = box->x2 - x; - GLint h = box->y2 - y; - - if (x < cx) w -= cx - x, x = cx; - if (y < cy) h -= cy - y, y = cy; - if (x + w > cx + cw) w = cx + cw - x; - if (y + h > cy + ch) h = cy + ch - y; - if (w <= 0) continue; - if (h <= 0) continue; - - b.x1 = x; - b.y1 = y; - b.x2 = x + w; - b.y2 = y + h; - } else { - b = *box; - } - - - if (b.x1 > b.x2 || - b.y1 > b.y2 || - b.x2 > intelScreen->width || - b.y2 > intelScreen->height) - continue; - - if ( buffers & BUFFER_BIT_FRONT_LEFT ) { - BEGIN_BATCH( 6); - OUT_BATCH( CMD ); - OUT_BATCH( BR13 ); - OUT_BATCH( (b.y1 << 16) | b.x1 ); - OUT_BATCH( (b.y2 << 16) | b.x2 ); - OUT_BATCH( intelScreen->front.offset ); - OUT_BATCH( clear_color ); - ADVANCE_BATCH(); - } - - if ( buffers & BUFFER_BIT_BACK_LEFT ) { - BEGIN_BATCH( 6); - OUT_BATCH( CMD ); - OUT_BATCH( BR13 ); - OUT_BATCH( (b.y1 << 16) | b.x1 ); - OUT_BATCH( (b.y2 << 16) | b.x2 ); - OUT_BATCH( intelScreen->back.offset ); - OUT_BATCH( clear_color ); - ADVANCE_BATCH(); - } - - if ( buffers & (BUFFER_BIT_STENCIL | BUFFER_BIT_DEPTH) ) { - BEGIN_BATCH( 6); - OUT_BATCH( D_CMD ); - OUT_BATCH( BR13 ); - OUT_BATCH( (b.y1 << 16) | b.x1 ); - OUT_BATCH( (b.y2 << 16) | b.x2 ); - OUT_BATCH( intelScreen->depth.offset ); - OUT_BATCH( clear_depth ); - ADVANCE_BATCH(); - } - } - } - intelFlushBatchLocked( intel, GL_TRUE, GL_FALSE, GL_TRUE ); - UNLOCK_HARDWARE( intel ); + intel_batchbuffer_flush(batch); + if (batch->last_fence != NULL) + dri_fence_wait(batch->last_fence); } - - -void intelDestroyBatchBuffer( GLcontext *ctx ) +/* This is the only way buffers get added to the validate list. + */ +GLboolean +intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch, + dri_bo *buffer, + GLuint flags, GLuint delta) { - intelContextPtr intel = INTEL_CONTEXT(ctx); - - if (intel->alloc.offset) { - intelFreeAGP( intel, intel->alloc.ptr ); - intel->alloc.ptr = NULL; - intel->alloc.offset = 0; - } - else if (intel->alloc.ptr) { - free(intel->alloc.ptr); - intel->alloc.ptr = NULL; - } + dri_emit_reloc(batch->buf, flags, delta, batch->ptr - batch->map, buffer); + batch->ptr += 4; - memset(&intel->batch, 0, sizeof(intel->batch)); + return GL_TRUE; } - -void intelInitBatchBuffer( GLcontext *ctx ) +void +intel_batchbuffer_data(struct intel_batchbuffer *batch, + const void *data, GLuint bytes, GLuint flags) { - intelContextPtr intel = INTEL_CONTEXT(ctx); - - /* This path isn't really safe with rotate: - */ - if (getenv("INTEL_BATCH") && intel->intelScreen->allow_batchbuffer) { - switch (intel->intelScreen->deviceID) { - case PCI_CHIP_I865_G: - /* HW bug? Seems to crash if batchbuffer crosses 4k boundary. - */ - intel->alloc.size = 8 * 1024; - break; - default: - /* This is the smallest amount of memory the kernel deals with. - * We'd ideally like to make this smaller. - */ - intel->alloc.size = 1 << intel->intelScreen->logTextureGranularity; - break; - } - - intel->alloc.ptr = intelAllocateAGP( intel, intel->alloc.size ); - if (intel->alloc.ptr) - intel->alloc.offset = - intelAgpOffsetFromVirtual( intel, intel->alloc.ptr ); - else - intel->alloc.offset = 0; /* OK? */ - } - - /* The default is now to use a local buffer and pass that to the - * kernel. This is also a fallback if allocation fails on the - * above path: - */ - if (!intel->alloc.ptr) { - intel->alloc.size = 8 * 1024; - intel->alloc.ptr = malloc( intel->alloc.size ); - intel->alloc.offset = 0; - } - - assert(intel->alloc.ptr); + assert((bytes & 3) == 0); + intel_batchbuffer_require_space(batch, bytes, flags); + __memcpy(batch->ptr, data, bytes); + batch->ptr += bytes; } diff --git a/src/mesa/drivers/dri/i915/intel_batchbuffer.h b/src/mesa/drivers/dri/i915/intel_batchbuffer.h index 577d07137f..b5c7a783a7 100644 --- a/src/mesa/drivers/dri/i915/intel_batchbuffer.h +++ b/src/mesa/drivers/dri/i915/intel_batchbuffer.h @@ -1,126 +1,109 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - #ifndef INTEL_BATCHBUFFER_H #define INTEL_BATCHBUFFER_H -#include "intel_context.h" -#include "intel_ioctl.h" +#include "mtypes.h" +#include "dri_bufmgr.h" -#define BATCH_LOCALS GLubyte *batch_ptr; +struct intel_context; -/* #define VERBOSE 0 */ -#ifndef VERBOSE -extern int VERBOSE; -#endif +#define BATCH_SZ 16384 +#define BATCH_RESERVED 16 +#define INTEL_BATCH_NO_CLIPRECTS 0x1 +#define INTEL_BATCH_CLIPRECTS 0x2 -#define BEGIN_BATCH(n) \ -do { \ - if (VERBOSE) fprintf(stderr, \ - "BEGIN_BATCH(%ld) in %s, %d dwords free\n", \ - ((unsigned long)n), __FUNCTION__, \ - intel->batch.space/4); \ - if (intel->batch.space < (n)*4) \ - intelFlushBatch(intel, GL_TRUE); \ - if (intel->batch.space == intel->batch.size) intel->batch.func = __FUNCTION__; \ - batch_ptr = intel->batch.ptr; \ -} while (0) +struct intel_batchbuffer +{ + struct intel_context *intel; + + dri_bo *buf; + dri_fence *last_fence; + GLuint flags; + + GLubyte *map; + GLubyte *ptr; + + GLuint size; +}; + +struct intel_batchbuffer *intel_batchbuffer_alloc(struct intel_context + *intel); + +void intel_batchbuffer_free(struct intel_batchbuffer *batch); + + +void intel_batchbuffer_finish(struct intel_batchbuffer *batch); + +void intel_batchbuffer_flush(struct intel_batchbuffer *batch); + +void intel_batchbuffer_reset(struct intel_batchbuffer *batch); -#define OUT_BATCH(n) \ -do { \ - *(GLuint *)batch_ptr = (n); \ - if (VERBOSE) fprintf(stderr, " -- %08x at %s/%d\n", (n), __FILE__, __LINE__); \ - batch_ptr += 4; \ -} while (0) -#define ADVANCE_BATCH() \ -do { \ - if (VERBOSE) fprintf(stderr, "ADVANCE_BATCH()\n"); \ - intel->batch.space -= (batch_ptr - intel->batch.ptr); \ - intel->batch.ptr = batch_ptr; \ - assert(intel->batch.space >= 0); \ -} while(0) - -extern void intelInitBatchBuffer( GLcontext *ctx ); -extern void intelDestroyBatchBuffer( GLcontext *ctx ); - -extern void intelStartInlinePrimitive( intelContextPtr intel, GLuint prim ); -extern void intelWrapInlinePrimitive( intelContextPtr intel ); -extern void intelRestartInlinePrimitive( intelContextPtr intel ); -extern GLuint *intelEmitInlinePrimitiveLocked(intelContextPtr intel, - int primitive, int dwords, - int vertex_size); -extern void intelCopyBuffer( const __DRIdrawablePrivate *dpriv, - const drm_clip_rect_t *rect); -extern void intelClearWithBlit(GLcontext *ctx, GLbitfield mask, GLboolean all, - GLint cx1, GLint cy1, GLint cw, GLint ch); - -extern void intelEmitCopyBlitLocked( intelContextPtr intel, - GLuint cpp, - GLshort src_pitch, - GLuint src_offset, - GLshort dst_pitch, - GLuint dst_offset, - GLshort srcx, GLshort srcy, - GLshort dstx, GLshort dsty, - GLshort w, GLshort h ); - -extern void intelEmitFillBlitLocked( intelContextPtr intel, - GLuint cpp, - GLshort dst_pitch, - GLuint dst_offset, - GLshort x, GLshort y, - GLshort w, GLshort h, - GLuint color ); - - - - -static __inline GLuint *intelExtendInlinePrimitive( intelContextPtr intel, - GLuint dwords ) +/* Unlike bmBufferData, this currently requires the buffer be mapped. + * Consider it a convenience function wrapping multple + * intel_buffer_dword() calls. + */ +void intel_batchbuffer_data(struct intel_batchbuffer *batch, + const void *data, GLuint bytes, GLuint flags); + +void intel_batchbuffer_release_space(struct intel_batchbuffer *batch, + GLuint bytes); + +GLboolean intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch, + dri_bo *buffer, + GLuint flags, GLuint offset); + +/* Inline functions - might actually be better off with these + * non-inlined. Certainly better off switching all command packets to + * be passed as structs rather than dwords, but that's a little bit of + * work... + */ +static INLINE GLuint +intel_batchbuffer_space(struct intel_batchbuffer *batch) { - GLuint sz = dwords * sizeof(GLuint); - GLuint *ptr; + return (batch->size - BATCH_RESERVED) - (batch->ptr - batch->map); +} - if (intel->batch.space < sz) { - intelWrapInlinePrimitive( intel ); -/* assert(intel->batch.space >= sz); */ - } -/* assert(intel->prim.primitive != ~0); */ - ptr = (GLuint *)intel->batch.ptr; - intel->batch.ptr += sz; - intel->batch.space -= sz; +static INLINE void +intel_batchbuffer_emit_dword(struct intel_batchbuffer *batch, GLuint dword) +{ + assert(batch->map); + assert(intel_batchbuffer_space(batch) >= 4); + *(GLuint *) (batch->ptr) = dword; + batch->ptr += 4; +} + +static INLINE void +intel_batchbuffer_require_space(struct intel_batchbuffer *batch, + GLuint sz, GLuint flags) +{ + assert(sz < batch->size - 8); + if (intel_batchbuffer_space(batch) < sz || + (batch->flags != 0 && flags != 0 && batch->flags != flags)) + intel_batchbuffer_flush(batch); - return ptr; + batch->flags |= flags; } +/* Here are the crusty old macros, to be removed: + */ +#define BATCH_LOCALS + +#define BEGIN_BATCH(n, flags) do { \ + assert(!intel->prim.flush); \ + intel_batchbuffer_require_space(intel->batch, (n)*4, flags); \ +} while (0) + +#define OUT_BATCH(d) intel_batchbuffer_emit_dword(intel->batch, d) + +#define OUT_RELOC(buf, flags, delta) do { \ + assert((delta) >= 0); \ + intel_batchbuffer_emit_reloc(intel->batch, buf, flags, delta); \ +} while (0) + +#define ADVANCE_BATCH() do { } while(0) #endif diff --git a/src/mesa/drivers/dri/i915/intel_blit.c b/src/mesa/drivers/dri/i915/intel_blit.c new file mode 100644 index 0000000000..2d4d00edef --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_blit.c @@ -0,0 +1,492 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include <stdio.h> +#include <errno.h> + +#include "mtypes.h" +#include "context.h" +#include "enums.h" + +#include "intel_batchbuffer.h" +#include "intel_blit.h" +#include "intel_buffers.h" +#include "intel_context.h" +#include "intel_fbo.h" +#include "intel_reg.h" +#include "intel_regions.h" +#include "vblank.h" + +#define FILE_DEBUG_FLAG DEBUG_BLIT + +/** + * Copy the back color buffer to the front color buffer. + * Used for SwapBuffers(). + */ +void +intelCopyBuffer(const __DRIdrawablePrivate * dPriv, + const drm_clip_rect_t * rect) +{ + + struct intel_context *intel; + const intelScreenPrivate *intelScreen; + + DBG("%s\n", __FUNCTION__); + + assert(dPriv); + + intel = intelScreenContext(dPriv->driScreenPriv->private); + if (!intel) + return; + + intelScreen = intel->intelScreen; + + if (intel->last_swap_fence) { + dri_fence_wait(intel->last_swap_fence); + dri_fence_unreference(intel->last_swap_fence); + intel->last_swap_fence = NULL; + } + intel->last_swap_fence = intel->first_swap_fence; + intel->first_swap_fence = NULL; + + /* The LOCK_HARDWARE is required for the cliprects. Buffer offsets + * should work regardless. + */ + LOCK_HARDWARE(intel); + + if (dPriv && dPriv->numClipRects) { + struct intel_framebuffer *intel_fb = dPriv->driverPrivate; + const struct intel_region *frontRegion + = intel_get_rb_region(&intel_fb->Base, BUFFER_FRONT_LEFT); + const struct intel_region *backRegion + = intel_get_rb_region(&intel_fb->Base, BUFFER_BACK_LEFT); + const int nbox = dPriv->numClipRects; + const drm_clip_rect_t *pbox = dPriv->pClipRects; + const int pitch = frontRegion->pitch; + const int cpp = frontRegion->cpp; + int BR13, CMD; + int i; + + ASSERT(intel_fb); + ASSERT(intel_fb->Base.Name == 0); /* Not a user-created FBO */ + ASSERT(frontRegion); + ASSERT(backRegion); + ASSERT(frontRegion->pitch == backRegion->pitch); + ASSERT(frontRegion->cpp == backRegion->cpp); + + if (cpp == 2) { + BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24); + CMD = XY_SRC_COPY_BLT_CMD; + } + else { + BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24) | (1 << 25); + CMD = (XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB); + } + + for (i = 0; i < nbox; i++, pbox++) { + drm_clip_rect_t box; + + if (pbox->x1 > pbox->x2 || + pbox->y1 > pbox->y2 || + pbox->x2 > intelScreen->width || pbox->y2 > intelScreen->height) + continue; + + box = *pbox; + + if (rect) { + if (rect->x1 > box.x1) + box.x1 = rect->x1; + if (rect->y1 > box.y1) + box.y1 = rect->y1; + if (rect->x2 < box.x2) + box.x2 = rect->x2; + if (rect->y2 < box.y2) + box.y2 = rect->y2; + + if (box.x1 > box.x2 || box.y1 > box.y2) + continue; + } + + BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH(CMD); + OUT_BATCH(BR13); + OUT_BATCH((pbox->y1 << 16) | pbox->x1); + OUT_BATCH((pbox->y2 << 16) | pbox->x2); + + OUT_RELOC(frontRegion->buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, + 0); + OUT_BATCH((pbox->y1 << 16) | pbox->x1); + OUT_BATCH(BR13 & 0xffff); + OUT_RELOC(backRegion->buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, + 0); + + ADVANCE_BATCH(); + } + + if (intel->first_swap_fence) + dri_fence_unreference(intel->first_swap_fence); + intel_batchbuffer_flush(intel->batch); + intel->first_swap_fence = intel->batch->last_fence; + dri_fence_reference(intel->first_swap_fence); + } + + UNLOCK_HARDWARE(intel); +} + + + + +void +intelEmitFillBlit(struct intel_context *intel, + GLuint cpp, + GLshort dst_pitch, + dri_bo *dst_buffer, + GLuint dst_offset, + GLshort x, GLshort y, GLshort w, GLshort h, GLuint color) +{ + GLuint BR13, CMD; + BATCH_LOCALS; + + dst_pitch *= cpp; + + switch (cpp) { + case 1: + case 2: + case 3: + BR13 = dst_pitch | (0xF0 << 16) | (1 << 24); + CMD = XY_COLOR_BLT_CMD; + break; + case 4: + BR13 = dst_pitch | (0xF0 << 16) | (1 << 24) | (1 << 25); + CMD = (XY_COLOR_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB); + break; + default: + return; + } + + DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", + __FUNCTION__, dst_buffer, dst_pitch, dst_offset, x, y, w, h); + + + BEGIN_BATCH(6, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH(CMD); + OUT_BATCH(BR13); + OUT_BATCH((y << 16) | x); + OUT_BATCH(((y + h) << 16) | (x + w)); + OUT_RELOC(dst_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, dst_offset); + OUT_BATCH(color); + ADVANCE_BATCH(); +} + + +static GLuint translate_raster_op(GLenum logicop) +{ + switch(logicop) { + case GL_CLEAR: return 0x00; + case GL_AND: return 0x88; + case GL_AND_REVERSE: return 0x44; + case GL_COPY: return 0xCC; + case GL_AND_INVERTED: return 0x22; + case GL_NOOP: return 0xAA; + case GL_XOR: return 0x66; + case GL_OR: return 0xEE; + case GL_NOR: return 0x11; + case GL_EQUIV: return 0x99; + case GL_INVERT: return 0x55; + case GL_OR_REVERSE: return 0xDD; + case GL_COPY_INVERTED: return 0x33; + case GL_OR_INVERTED: return 0xBB; + case GL_NAND: return 0x77; + case GL_SET: return 0xFF; + default: return 0; + } +} + + +/* Copy BitBlt + */ +void +intelEmitCopyBlit(struct intel_context *intel, + GLuint cpp, + GLshort src_pitch, + dri_bo *src_buffer, + GLuint src_offset, + GLshort dst_pitch, + dri_bo *dst_buffer, + GLuint dst_offset, + GLshort src_x, GLshort src_y, + GLshort dst_x, GLshort dst_y, + GLshort w, GLshort h, + GLenum logic_op) +{ + GLuint CMD, BR13; + int dst_y2 = dst_y + h; + int dst_x2 = dst_x + w; + BATCH_LOCALS; + + + DBG("%s src:buf(%p)/%d+%d %d,%d dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", + __FUNCTION__, + src_buffer, src_pitch, src_offset, src_x, src_y, + dst_buffer, dst_pitch, dst_offset, dst_x, dst_y, w, h); + + src_pitch *= cpp; + dst_pitch *= cpp; + + switch (cpp) { + case 1: + case 2: + case 3: + BR13 = (((GLint) dst_pitch) & 0xffff) | + (translate_raster_op(logic_op) << 16) | (1 << 24); + CMD = XY_SRC_COPY_BLT_CMD; + break; + case 4: + BR13 = + (((GLint) dst_pitch) & 0xffff) | + (translate_raster_op(logic_op) << 16) | (1 << 24) | (1 << 25); + CMD = + (XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB); + break; + default: + return; + } + + if (dst_y2 < dst_y || dst_x2 < dst_x) { + return; + } + + /* Initial y values don't seem to work with negative pitches. If + * we adjust the offsets manually (below), it seems to work fine. + * + * On the other hand, if we always adjust, the hardware doesn't + * know which blit directions to use, so overlapping copypixels get + * the wrong result. + */ + if (dst_pitch > 0 && src_pitch > 0) { + BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH(CMD); + OUT_BATCH(BR13); + OUT_BATCH((dst_y << 16) | dst_x); + OUT_BATCH((dst_y2 << 16) | dst_x2); + OUT_RELOC(dst_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, dst_offset); + OUT_BATCH((src_y << 16) | src_x); + OUT_BATCH(((GLint) src_pitch & 0xffff)); + OUT_RELOC(src_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, src_offset); + ADVANCE_BATCH(); + } + else { + BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH(CMD); + OUT_BATCH(BR13); + OUT_BATCH((0 << 16) | dst_x); + OUT_BATCH((h << 16) | dst_x2); + OUT_RELOC(dst_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, + dst_offset + dst_y * dst_pitch); + OUT_BATCH((0 << 16) | src_x); + OUT_BATCH(((GLint) src_pitch & 0xffff)); + OUT_RELOC(src_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, + src_offset + src_y * src_pitch); + ADVANCE_BATCH(); + } +} + + +/** + * Use blitting to clear the renderbuffers named by 'flags'. + * Note: we can't use the ctx->DrawBuffer->_ColorDrawBufferMask field + * since that might include software renderbuffers or renderbuffers + * which we're clearing with triangles. + * \param mask bitmask of BUFFER_BIT_* values indicating buffers to clear + */ +void +intelClearWithBlit(GLcontext * ctx, GLbitfield mask) +{ + struct intel_context *intel = intel_context(ctx); + struct gl_framebuffer *fb = ctx->DrawBuffer; + GLuint clear_depth; + GLbitfield skipBuffers = 0; + BATCH_LOCALS; + + DBG("%s %x\n", __FUNCTION__, mask); + + /* + * Compute values for clearing the buffers. + */ + clear_depth = 0; + if (mask & BUFFER_BIT_DEPTH) { + clear_depth = (GLuint) (fb->_DepthMax * ctx->Depth.Clear); + } + if (mask & BUFFER_BIT_STENCIL) { + clear_depth |= (ctx->Stencil.Clear & 0xff) << 24; + } + + /* If clearing both depth and stencil, skip BUFFER_BIT_STENCIL in + * the loop below. + */ + if ((mask & BUFFER_BIT_DEPTH) && (mask & BUFFER_BIT_STENCIL)) { + skipBuffers = BUFFER_BIT_STENCIL; + } + + /* XXX Move this flush/lock into the following conditional? */ + intelFlush(&intel->ctx); + LOCK_HARDWARE(intel); + + if (intel->numClipRects) { + GLint cx, cy, cw, ch; + drm_clip_rect_t clear; + int i; + + /* Get clear bounds after locking */ + cx = fb->_Xmin; + cy = fb->_Ymin; + cw = fb->_Xmax - cx; + ch = fb->_Ymax - cy; + + if (fb->Name == 0) { + /* clearing a window */ + + /* flip top to bottom */ + clear.x1 = cx + intel->drawX; + clear.y1 = intel->driDrawable->y + intel->driDrawable->h - cy - ch; + clear.x2 = clear.x1 + cw; + clear.y2 = clear.y1 + ch; + } + else { + /* clearing FBO */ + assert(intel->numClipRects == 1); + assert(intel->pClipRects == &intel->fboRect); + clear.x1 = cx; + clear.y1 = cy; + clear.x2 = clear.x1 + cw; + clear.y2 = clear.y1 + ch; + /* no change to mask */ + } + + for (i = 0; i < intel->numClipRects; i++) { + const drm_clip_rect_t *box = &intel->pClipRects[i]; + drm_clip_rect_t b; + GLuint buf; + GLuint clearMask = mask; /* use copy, since we modify it below */ + GLboolean all = (cw == fb->Width && ch == fb->Height); + + if (!all) { + intel_intersect_cliprects(&b, &clear, box); + } + else { + b = *box; + } + + if (b.x1 >= b.x2 || b.y1 >= b.y2) + continue; + + if (0) + _mesa_printf("clear %d,%d..%d,%d, mask %x\n", + b.x1, b.y1, b.x2, b.y2, mask); + + /* Loop over all renderbuffers */ + for (buf = 0; buf < BUFFER_COUNT && clearMask; buf++) { + const GLbitfield bufBit = 1 << buf; + if ((clearMask & bufBit) && !(bufBit & skipBuffers)) { + /* OK, clear this renderbuffer */ + struct intel_region *irb_region = + intel_get_rb_region(fb, buf); + dri_bo *write_buffer = + intel_region_buffer(intel->intelScreen, irb_region, + all ? INTEL_WRITE_FULL : + INTEL_WRITE_PART); + + GLuint clearVal; + GLint pitch, cpp; + GLuint BR13, CMD; + + ASSERT(irb_region); + + pitch = irb_region->pitch; + cpp = irb_region->cpp; + + DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", + __FUNCTION__, + irb_region->buffer, (pitch * cpp), + irb_region->draw_offset, + b.x1, b.y1, b.x2 - b.x1, b.y2 - b.y1); + + + /* Setup the blit command */ + if (cpp == 4) { + BR13 = (0xF0 << 16) | (pitch * cpp) | (1 << 24) | (1 << 25); + if (buf == BUFFER_DEPTH || buf == BUFFER_STENCIL) { + CMD = XY_COLOR_BLT_CMD; + if (clearMask & BUFFER_BIT_DEPTH) + CMD |= XY_BLT_WRITE_RGB; + if (clearMask & BUFFER_BIT_STENCIL) + CMD |= XY_BLT_WRITE_ALPHA; + } + else { + /* clearing RGBA */ + CMD = XY_COLOR_BLT_CMD | + XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB; + } + } + else { + ASSERT(cpp == 2 || cpp == 0); + BR13 = (0xF0 << 16) | (pitch * cpp) | (1 << 24); + CMD = XY_COLOR_BLT_CMD; + } + + if (buf == BUFFER_DEPTH || buf == BUFFER_STENCIL) { + clearVal = clear_depth; + } + else { + clearVal = (cpp == 4) + ? intel->ClearColor8888 : intel->ClearColor565; + } + /* + _mesa_debug(ctx, "hardware blit clear buf %d rb id %d\n", + buf, irb->Base.Name); + */ + intel_wait_flips(intel, INTEL_BATCH_NO_CLIPRECTS); + + BEGIN_BATCH(6, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH(CMD); + OUT_BATCH(BR13); + OUT_BATCH((b.y1 << 16) | b.x1); + OUT_BATCH((b.y2 << 16) | b.x2); + OUT_RELOC(write_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, + irb_region->draw_offset); + OUT_BATCH(clearVal); + ADVANCE_BATCH(); + clearMask &= ~bufBit; /* turn off bit, for faster loop exit */ + } + } + } + intel_batchbuffer_flush(intel->batch); + } + + UNLOCK_HARDWARE(intel); +} diff --git a/src/mesa/drivers/dri/i915/intel_blit.h b/src/mesa/drivers/dri/i915/intel_blit.h new file mode 100644 index 0000000000..a66af86359 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_blit.h @@ -0,0 +1,62 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_BLIT_H +#define INTEL_BLIT_H + +#include "intel_context.h" +#include "intel_ioctl.h" +#include "dri_bufmgr.h" + +extern void intelCopyBuffer(const __DRIdrawablePrivate * dpriv, + const drm_clip_rect_t * rect); + +extern void intelClearWithBlit(GLcontext * ctx, GLbitfield mask); + +extern void intelEmitCopyBlit(struct intel_context *intel, + GLuint cpp, + GLshort src_pitch, + dri_bo *src_buffer, + GLuint src_offset, + GLshort dst_pitch, + dri_bo *dst_buffer, + GLuint dst_offset, + GLshort srcx, GLshort srcy, + GLshort dstx, GLshort dsty, + GLshort w, GLshort h, + GLenum logicop ); + +extern void intelEmitFillBlit(struct intel_context *intel, + GLuint cpp, + GLshort dst_pitch, + dri_bo *dst_buffer, + GLuint dst_offset, + GLshort x, GLshort y, + GLshort w, GLshort h, GLuint color); + + +#endif diff --git a/src/mesa/drivers/dri/i915/intel_buffer_objects.c b/src/mesa/drivers/dri/i915/intel_buffer_objects.c new file mode 100644 index 0000000000..3c73c402d9 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_buffer_objects.c @@ -0,0 +1,268 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "imports.h" +#include "mtypes.h" +#include "bufferobj.h" + +#include "intel_context.h" +#include "intel_buffer_objects.h" +#include "intel_regions.h" +#include "dri_bufmgr.h" + +/** Allocates a new dri_bo to store the data for the buffer object. */ +static void +intel_bufferobj_alloc_buffer(struct intel_context *intel, + struct intel_buffer_object *intel_obj) +{ + intel_obj->buffer = dri_bo_alloc(intel->intelScreen->bufmgr, "bufferobj", + intel_obj->Base.Size, 64, + DRM_BO_FLAG_MEM_TT); +} + +/** + * There is some duplication between mesa's bufferobjects and our + * bufmgr buffers. Both have an integer handle and a hashtable to + * lookup an opaque structure. It would be nice if the handles and + * internal structure where somehow shared. + */ +static struct gl_buffer_object * +intel_bufferobj_alloc(GLcontext * ctx, GLuint name, GLenum target) +{ + struct intel_buffer_object *obj = CALLOC_STRUCT(intel_buffer_object); + + _mesa_initialize_buffer_object(&obj->Base, name, target); + + obj->buffer = NULL; + + return &obj->Base; +} + +/* Break the COW tie to the region. The region gets to keep the data. + */ +void +intel_bufferobj_release_region(struct intel_context *intel, + struct intel_buffer_object *intel_obj) +{ + assert(intel_obj->region->buffer == intel_obj->buffer); + intel_obj->region->pbo = NULL; + intel_obj->region = NULL; + + dri_bo_unreference(intel_obj->buffer); + intel_obj->buffer = NULL; +} + +/* Break the COW tie to the region. Both the pbo and the region end + * up with a copy of the data. + */ +void +intel_bufferobj_cow(struct intel_context *intel, + struct intel_buffer_object *intel_obj) +{ + assert(intel_obj->region); + intel_region_cow(intel->intelScreen, intel_obj->region); +} + + +/** + * Deallocate/free a vertex/pixel buffer object. + * Called via glDeleteBuffersARB(). + */ +static void +intel_bufferobj_free(GLcontext * ctx, struct gl_buffer_object *obj) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_buffer_object *intel_obj = intel_buffer_object(obj); + + assert(intel_obj); + + if (intel_obj->region) { + intel_bufferobj_release_region(intel, intel_obj); + } + else if (intel_obj->buffer) { + dri_bo_unreference(intel_obj->buffer); + } + + _mesa_free(intel_obj); +} + + + +/** + * Allocate space for and store data in a buffer object. Any data that was + * previously stored in the buffer object is lost. If data is NULL, + * memory will be allocated, but no copy will occur. + * Called via glBufferDataARB(). + */ +static void +intel_bufferobj_data(GLcontext * ctx, + GLenum target, + GLsizeiptrARB size, + const GLvoid * data, + GLenum usage, struct gl_buffer_object *obj) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_buffer_object *intel_obj = intel_buffer_object(obj); + + intel_obj->Base.Size = size; + intel_obj->Base.Usage = usage; + + if (intel_obj->region) + intel_bufferobj_release_region(intel, intel_obj); + + if (intel_obj->buffer != NULL && intel_obj->buffer->size != size) { + dri_bo_unreference(intel_obj->buffer); + intel_obj->buffer = NULL; + } + + intel_bufferobj_alloc_buffer(intel, intel_obj); + + if (data != NULL) + dri_bo_subdata(intel_obj->buffer, 0, size, data); +} + + +/** + * Replace data in a subrange of buffer object. If the data range + * specified by size + offset extends beyond the end of the buffer or + * if data is NULL, no copy is performed. + * Called via glBufferSubDataARB(). + */ +static void +intel_bufferobj_subdata(GLcontext * ctx, + GLenum target, + GLintptrARB offset, + GLsizeiptrARB size, + const GLvoid * data, struct gl_buffer_object *obj) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_buffer_object *intel_obj = intel_buffer_object(obj); + + assert(intel_obj); + + if (intel_obj->region) + intel_bufferobj_cow(intel, intel_obj); + + dri_bo_subdata(intel_obj->buffer, offset, size, data); +} + + +/** + * Called via glGetBufferSubDataARB(). + */ +static void +intel_bufferobj_get_subdata(GLcontext * ctx, + GLenum target, + GLintptrARB offset, + GLsizeiptrARB size, + GLvoid * data, struct gl_buffer_object *obj) +{ + struct intel_buffer_object *intel_obj = intel_buffer_object(obj); + + assert(intel_obj); + dri_bo_get_subdata(intel_obj->buffer, offset, size, data); +} + + + +/** + * Called via glMapBufferARB(). + */ +static void * +intel_bufferobj_map(GLcontext * ctx, + GLenum target, + GLenum access, struct gl_buffer_object *obj) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_buffer_object *intel_obj = intel_buffer_object(obj); + + /* XXX: Translate access to flags arg below: + */ + assert(intel_obj); + + if (intel_obj->region) + intel_bufferobj_cow(intel, intel_obj); + + if (intel_obj->buffer == NULL) { + obj->Pointer = NULL; + return NULL; + } + + dri_bo_map(intel_obj->buffer, GL_TRUE); + obj->Pointer = intel_obj->buffer->virtual; + return obj->Pointer; +} + + +/** + * Called via glMapBufferARB(). + */ +static GLboolean +intel_bufferobj_unmap(GLcontext * ctx, + GLenum target, struct gl_buffer_object *obj) +{ + struct intel_buffer_object *intel_obj = intel_buffer_object(obj); + + assert(intel_obj); + if (intel_obj->buffer != NULL) { + assert(obj->Pointer); + dri_bo_unmap(intel_obj->buffer); + obj->Pointer = NULL; + } + return GL_TRUE; +} + +dri_bo * +intel_bufferobj_buffer(struct intel_context *intel, + struct intel_buffer_object *intel_obj, GLuint flag) +{ + if (intel_obj->region) { + if (flag == INTEL_WRITE_PART) + intel_bufferobj_cow(intel, intel_obj); + else if (flag == INTEL_WRITE_FULL) { + intel_bufferobj_release_region(intel, intel_obj); + intel_bufferobj_alloc_buffer(intel, intel_obj); + } + } + + return intel_obj->buffer; +} + +void +intel_bufferobj_init(struct intel_context *intel) +{ + GLcontext *ctx = &intel->ctx; + + ctx->Driver.NewBufferObject = intel_bufferobj_alloc; + ctx->Driver.DeleteBuffer = intel_bufferobj_free; + ctx->Driver.BufferData = intel_bufferobj_data; + ctx->Driver.BufferSubData = intel_bufferobj_subdata; + ctx->Driver.GetBufferSubData = intel_bufferobj_get_subdata; + ctx->Driver.MapBuffer = intel_bufferobj_map; + ctx->Driver.UnmapBuffer = intel_bufferobj_unmap; +} diff --git a/src/mesa/drivers/dri/i915/intel_buffer_objects.h b/src/mesa/drivers/dri/i915/intel_buffer_objects.h new file mode 100644 index 0000000000..db579a8ae4 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_buffer_objects.h @@ -0,0 +1,86 @@ + /************************************************************************** + * + * Copyright 2005 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_BUFFEROBJ_H +#define INTEL_BUFFEROBJ_H + +#include "mtypes.h" + +struct intel_context; +struct intel_region; +struct gl_buffer_object; + + +/** + * Intel vertex/pixel buffer object, derived from Mesa's gl_buffer_object. + */ +struct intel_buffer_object +{ + struct gl_buffer_object Base; + dri_bo *buffer; /* the low-level buffer manager's buffer handle */ + + struct intel_region *region; /* Is there a zero-copy texture + associated with this (pixel) + buffer object? */ +}; + + +/* Get the bm buffer associated with a GL bufferobject: + */ +dri_bo *intel_bufferobj_buffer(struct intel_context *intel, + struct intel_buffer_object + *obj, GLuint flag); + +/* Hook the bufferobject implementation into mesa: + */ +void intel_bufferobj_init(struct intel_context *intel); + + + +/* Are the obj->Name tests necessary? Unfortunately yes, mesa + * allocates a couple of gl_buffer_object structs statically, and + * the Name == 0 test is the only way to identify them and avoid + * casting them erroneously to our structs. + */ +static INLINE struct intel_buffer_object * +intel_buffer_object(struct gl_buffer_object *obj) +{ + if (obj->Name) + return (struct intel_buffer_object *) obj; + else + return NULL; +} + +/* Helpers for zerocopy image uploads. See also intel_regions.h: + */ +void intel_bufferobj_cow(struct intel_context *intel, + struct intel_buffer_object *intel_obj); +void intel_bufferobj_release_region(struct intel_context *intel, + struct intel_buffer_object *intel_obj); + + +#endif diff --git a/src/mesa/drivers/dri/i915/intel_buffers.c b/src/mesa/drivers/dri/i915/intel_buffers.c new file mode 100644 index 0000000000..46a67b141e --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_buffers.c @@ -0,0 +1,1193 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "intel_screen.h" +#include "intel_context.h" +#include "intel_blit.h" +#include "intel_buffers.h" +#include "intel_depthstencil.h" +#include "intel_fbo.h" +#include "intel_tris.h" +#include "intel_regions.h" +#include "intel_batchbuffer.h" +#include "intel_reg.h" +#include "context.h" +#include "utils.h" +#include "drirenderbuffer.h" +#include "framebuffer.h" +#include "swrast/swrast.h" +#include "vblank.h" + + +/* This block can be removed when libdrm >= 2.3.1 is required */ + +#ifndef DRM_IOCTL_I915_FLIP + +#define DRM_VBLANK_FLIP 0x8000000 + +typedef struct drm_i915_flip { + int pipes; +} drm_i915_flip_t; + +#undef DRM_IOCTL_I915_FLIP +#define DRM_IOCTL_I915_FLIP DRM_IOW(DRM_COMMAND_BASE + DRM_I915_FLIP, \ + drm_i915_flip_t) + +#endif + + +/** + * XXX move this into a new dri/common/cliprects.c file. + */ +GLboolean +intel_intersect_cliprects(drm_clip_rect_t * dst, + const drm_clip_rect_t * a, + const drm_clip_rect_t * b) +{ + GLint bx = b->x1; + GLint by = b->y1; + GLint bw = b->x2 - bx; + GLint bh = b->y2 - by; + + if (bx < a->x1) + bw -= a->x1 - bx, bx = a->x1; + if (by < a->y1) + bh -= a->y1 - by, by = a->y1; + if (bx + bw > a->x2) + bw = a->x2 - bx; + if (by + bh > a->y2) + bh = a->y2 - by; + if (bw <= 0) + return GL_FALSE; + if (bh <= 0) + return GL_FALSE; + + dst->x1 = bx; + dst->y1 = by; + dst->x2 = bx + bw; + dst->y2 = by + bh; + + return GL_TRUE; +} + +/** + * Return pointer to current color drawing region, or NULL. + */ +struct intel_region * +intel_drawbuf_region(struct intel_context *intel) +{ + struct intel_renderbuffer *irbColor = + intel_renderbuffer(intel->ctx.DrawBuffer->_ColorDrawBuffers[0][0]); + if (irbColor) + return irbColor->region; + else + return NULL; +} + +/** + * Return pointer to current color reading region, or NULL. + */ +struct intel_region * +intel_readbuf_region(struct intel_context *intel) +{ + struct intel_renderbuffer *irb + = intel_renderbuffer(intel->ctx.ReadBuffer->_ColorReadBuffer); + if (irb) + return irb->region; + else + return NULL; +} + + + +/** + * Update the following fields for rendering to a user-created FBO: + * intel->numClipRects + * intel->pClipRects + * intel->drawX + * intel->drawY + */ +static void +intelSetRenderbufferClipRects(struct intel_context *intel) +{ + assert(intel->ctx.DrawBuffer->Width > 0); + assert(intel->ctx.DrawBuffer->Height > 0); + intel->fboRect.x1 = 0; + intel->fboRect.y1 = 0; + intel->fboRect.x2 = intel->ctx.DrawBuffer->Width; + intel->fboRect.y2 = intel->ctx.DrawBuffer->Height; + intel->numClipRects = 1; + intel->pClipRects = &intel->fboRect; + intel->drawX = 0; + intel->drawY = 0; +} + + +/** + * As above, but for rendering to front buffer of a window. + * \sa intelSetRenderbufferClipRects + */ +static void +intelSetFrontClipRects(struct intel_context *intel) +{ + __DRIdrawablePrivate *dPriv = intel->driDrawable; + + if (!dPriv) + return; + + intel->numClipRects = dPriv->numClipRects; + intel->pClipRects = dPriv->pClipRects; + intel->drawX = dPriv->x; + intel->drawY = dPriv->y; +} + + +/** + * As above, but for rendering to back buffer of a window. + */ +static void +intelSetBackClipRects(struct intel_context *intel) +{ + __DRIdrawablePrivate *dPriv = intel->driDrawable; + struct intel_framebuffer *intel_fb; + + if (!dPriv) + return; + + intel_fb = dPriv->driverPrivate; + + if (intel_fb->pf_active || dPriv->numBackClipRects == 0) { + /* use the front clip rects */ + intel->numClipRects = dPriv->numClipRects; + intel->pClipRects = dPriv->pClipRects; + intel->drawX = dPriv->x; + intel->drawY = dPriv->y; + } + else { + /* use the back clip rects */ + intel->numClipRects = dPriv->numBackClipRects; + intel->pClipRects = dPriv->pBackClipRects; + intel->drawX = dPriv->backX; + intel->drawY = dPriv->backY; + } +} + + +/** + * This will be called whenever the currently bound window is moved/resized. + * XXX: actually, it seems to NOT be called when the window is only moved (BP). + */ +void +intelWindowMoved(struct intel_context *intel) +{ + GLcontext *ctx = &intel->ctx; + __DRIdrawablePrivate *dPriv = intel->driDrawable; + struct intel_framebuffer *intel_fb = dPriv->driverPrivate; + + if (!intel->ctx.DrawBuffer) { + /* when would this happen? -BP */ + intelSetFrontClipRects(intel); + } + else if (intel->ctx.DrawBuffer->Name != 0) { + /* drawing to user-created FBO - do nothing */ + /* Cliprects would be set from intelDrawBuffer() */ + } + else { + /* drawing to a window */ + switch (intel_fb->Base._ColorDrawBufferMask[0]) { + case BUFFER_BIT_FRONT_LEFT: + intelSetFrontClipRects(intel); + break; + case BUFFER_BIT_BACK_LEFT: + intelSetBackClipRects(intel); + break; + default: + /* glDrawBuffer(GL_NONE or GL_FRONT_AND_BACK): software fallback */ + intelSetFrontClipRects(intel); + } + } + + if (intel->intelScreen->driScrnPriv->ddx_version.minor >= 7) { + drmI830Sarea *sarea = intel->sarea; + drm_clip_rect_t drw_rect = { .x1 = dPriv->x, .x2 = dPriv->x + dPriv->w, + .y1 = dPriv->y, .y2 = dPriv->y + dPriv->h }; + drm_clip_rect_t planeA_rect = { .x1 = sarea->planeA_x, .y1 = sarea->planeA_y, + .x2 = sarea->planeA_x + sarea->planeA_w, + .y2 = sarea->planeA_y + sarea->planeA_h }; + drm_clip_rect_t planeB_rect = { .x1 = sarea->planeB_x, .y1 = sarea->planeB_y, + .x2 = sarea->planeB_x + sarea->planeB_w, + .y2 = sarea->planeB_y + sarea->planeB_h }; + GLint areaA = driIntersectArea( drw_rect, planeA_rect ); + GLint areaB = driIntersectArea( drw_rect, planeB_rect ); + GLuint flags = intel_fb->vblank_flags; + GLboolean pf_active; + GLint pf_planes; + + /* Update page flipping info + */ + pf_planes = 0; + + if (areaA > 0) + pf_planes |= 1; + + if (areaB > 0) + pf_planes |= 2; + + intel_fb->pf_current_page = (intel->sarea->pf_current_page >> + (intel_fb->pf_planes & 0x2)) & 0x3; + + intel_fb->pf_num_pages = intel->intelScreen->third.handle ? 3 : 2; + + pf_active = pf_planes && (pf_planes & intel->sarea->pf_active) == pf_planes; + + if (INTEL_DEBUG & DEBUG_LOCK) + if (pf_active != intel_fb->pf_active) + _mesa_printf("%s - Page flipping %sactive\n", __progname, + pf_active ? "" : "in"); + + if (pf_active) { + /* Sync pages between planes if flipping on both at the same time */ + if (pf_planes == 0x3 && pf_planes != intel_fb->pf_planes && + (intel->sarea->pf_current_page & 0x3) != + (((intel->sarea->pf_current_page) >> 2) & 0x3)) { + drm_i915_flip_t flip; + + if (intel_fb->pf_current_page == + (intel->sarea->pf_current_page & 0x3)) { + /* XXX: This is ugly, but emitting two flips 'in a row' can cause + * lockups for unknown reasons. + */ + intel->sarea->pf_current_page = + intel->sarea->pf_current_page & 0x3; + intel->sarea->pf_current_page |= + ((intel_fb->pf_current_page + intel_fb->pf_num_pages - 1) % + intel_fb->pf_num_pages) << 2; + + flip.pipes = 0x2; + } else { + intel->sarea->pf_current_page = + intel->sarea->pf_current_page & (0x3 << 2); + intel->sarea->pf_current_page |= + (intel_fb->pf_current_page + intel_fb->pf_num_pages - 1) % + intel_fb->pf_num_pages; + + flip.pipes = 0x1; + } + + drmCommandWrite(intel->driFd, DRM_I915_FLIP, &flip, sizeof(flip)); + } + + intel_fb->pf_planes = pf_planes; + } + + intel_fb->pf_active = pf_active; + intel_flip_renderbuffers(intel_fb); + intel_draw_buffer(&intel->ctx, intel->ctx.DrawBuffer); + + /* Update vblank info + */ + if (areaB > areaA || (areaA == areaB && areaB > 0)) { + flags = intel_fb->vblank_flags | VBLANK_FLAG_SECONDARY; + } else { + flags = intel_fb->vblank_flags & ~VBLANK_FLAG_SECONDARY; + } + + if (flags != intel_fb->vblank_flags && intel_fb->vblank_flags && + !(intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ)) { + drmVBlank vbl; + int i; + + vbl.request.type = DRM_VBLANK_ABSOLUTE; + + if ( intel_fb->vblank_flags & VBLANK_FLAG_SECONDARY ) { + vbl.request.type |= DRM_VBLANK_SECONDARY; + } + + for (i = 0; i < intel_fb->pf_num_pages; i++) { + if (!intel_fb->color_rb[i] || + (intel_fb->vbl_waited - intel_fb->color_rb[i]->vbl_pending) <= + (1<<23)) + continue; + + vbl.request.sequence = intel_fb->color_rb[i]->vbl_pending; + drmWaitVBlank(intel->driFd, &vbl); + } + + intel_fb->vblank_flags = flags; + driGetCurrentVBlank(dPriv, intel_fb->vblank_flags, &intel_fb->vbl_seq); + intel_fb->vbl_waited = intel_fb->vbl_seq; + + for (i = 0; i < intel_fb->pf_num_pages; i++) { + if (intel_fb->color_rb[i]) + intel_fb->color_rb[i]->vbl_pending = intel_fb->vbl_waited; + } + } + } else { + intel_fb->vblank_flags &= ~VBLANK_FLAG_SECONDARY; + } + + /* Update Mesa's notion of window size */ + driUpdateFramebufferSize(ctx, dPriv); + intel_fb->Base.Initialized = GL_TRUE; /* XXX remove someday */ + + /* Update hardware scissor */ + ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y, + ctx->Scissor.Width, ctx->Scissor.Height); + + /* Re-calculate viewport related state */ + ctx->Driver.DepthRange( ctx, ctx->Viewport.Near, ctx->Viewport.Far ); +} + + + +/* A true meta version of this would be very simple and additionally + * machine independent. Maybe we'll get there one day. + */ +static void +intelClearWithTris(struct intel_context *intel, GLbitfield mask) +{ + GLcontext *ctx = &intel->ctx; + struct gl_framebuffer *fb = ctx->DrawBuffer; + drm_clip_rect_t clear; + + if (INTEL_DEBUG & DEBUG_BLIT) + _mesa_printf("%s 0x%x\n", __FUNCTION__, mask); + + LOCK_HARDWARE(intel); + + /* XXX FBO: was: intel->driDrawable->numClipRects */ + if (intel->numClipRects) { + GLint cx, cy, cw, ch; + GLuint buf; + + intel->vtbl.install_meta_state(intel); + + /* Get clear bounds after locking */ + cx = fb->_Xmin; + cy = fb->_Ymin; + ch = fb->_Ymax - cx; + cw = fb->_Xmax - cy; + + /* note: regardless of 'all', cx, cy, cw, ch are now correct */ + clear.x1 = cx; + clear.y1 = cy; + clear.x2 = cx + cw; + clear.y2 = cy + ch; + + /* Back and stencil cliprects are the same. Try and do both + * buffers at once: + */ + if (mask & + (BUFFER_BIT_BACK_LEFT | BUFFER_BIT_STENCIL | BUFFER_BIT_DEPTH)) { + struct intel_region *backRegion = + intel_get_rb_region(fb, BUFFER_BACK_LEFT); + struct intel_region *depthRegion = + intel_get_rb_region(fb, BUFFER_DEPTH); + const GLuint clearColor = (backRegion && backRegion->cpp == 4) + ? intel->ClearColor8888 : intel->ClearColor565; + + intel->vtbl.meta_draw_region(intel, backRegion, depthRegion); + + if (mask & BUFFER_BIT_BACK_LEFT) + intel->vtbl.meta_color_mask(intel, GL_TRUE); + else + intel->vtbl.meta_color_mask(intel, GL_FALSE); + + if (mask & BUFFER_BIT_STENCIL) + intel->vtbl.meta_stencil_replace(intel, + intel->ctx.Stencil.WriteMask[0], + intel->ctx.Stencil.Clear); + else + intel->vtbl.meta_no_stencil_write(intel); + + if (mask & BUFFER_BIT_DEPTH) + intel->vtbl.meta_depth_replace(intel); + else + intel->vtbl.meta_no_depth_write(intel); + + /* XXX: Using INTEL_BATCH_NO_CLIPRECTS here is dangerous as the + * drawing origin may not be correctly emitted. + */ + intel_meta_draw_quad(intel, clear.x1, clear.x2, clear.y1, clear.y2, intel->ctx.Depth.Clear, clearColor, 0, 0, 0, 0); /* texcoords */ + + mask &= + ~(BUFFER_BIT_BACK_LEFT | BUFFER_BIT_STENCIL | BUFFER_BIT_DEPTH); + } + + /* clear the remaining (color) renderbuffers */ + for (buf = 0; buf < BUFFER_COUNT && mask; buf++) { + const GLuint bufBit = 1 << buf; + if (mask & bufBit) { + struct intel_renderbuffer *irbColor = + intel_renderbuffer(fb->Attachment[buf].Renderbuffer); + GLuint color = (irbColor->region->cpp == 4) + ? intel->ClearColor8888 : intel->ClearColor565; + + ASSERT(irbColor); + + intel->vtbl.meta_no_depth_write(intel); + intel->vtbl.meta_no_stencil_write(intel); + intel->vtbl.meta_color_mask(intel, GL_TRUE); + intel->vtbl.meta_draw_region(intel, irbColor->region, NULL); + + /* XXX: Using INTEL_BATCH_NO_CLIPRECTS here is dangerous as the + * drawing origin may not be correctly emitted. + */ + intel_meta_draw_quad(intel, clear.x1, clear.x2, clear.y1, clear.y2, 0, /* depth clear val */ + color, 0, 0, 0, 0); /* texcoords */ + + mask &= ~bufBit; + } + } + + intel->vtbl.leave_meta_state(intel); + intel_batchbuffer_flush(intel->batch); + } + UNLOCK_HARDWARE(intel); +} + + + + +/** + * Copy the window contents named by dPriv to the rotated (or reflected) + * color buffer. + * srcBuf is BUFFER_BIT_FRONT_LEFT or BUFFER_BIT_BACK_LEFT to indicate the source. + */ +void +intelRotateWindow(struct intel_context *intel, + __DRIdrawablePrivate * dPriv, GLuint srcBuf) +{ + intelScreenPrivate *screen = intel->intelScreen; + drm_clip_rect_t fullRect; + struct intel_framebuffer *intel_fb; + struct intel_region *src; + const drm_clip_rect_t *clipRects; + int numClipRects; + int i; + GLenum format, type; + + int xOrig, yOrig; + int origNumClipRects; + drm_clip_rect_t *origRects; + + /* + * set up hardware state + */ + intelFlush(&intel->ctx); + + LOCK_HARDWARE(intel); + + if (!intel->numClipRects) { + UNLOCK_HARDWARE(intel); + return; + } + + intel->vtbl.install_meta_state(intel); + + intel->vtbl.meta_no_depth_write(intel); + intel->vtbl.meta_no_stencil_write(intel); + intel->vtbl.meta_color_mask(intel, GL_FALSE); + + + /* save current drawing origin and cliprects (restored at end) */ + xOrig = intel->drawX; + yOrig = intel->drawY; + origNumClipRects = intel->numClipRects; + origRects = intel->pClipRects; + + /* + * set drawing origin, cliprects for full-screen access to rotated screen + */ + fullRect.x1 = 0; + fullRect.y1 = 0; + fullRect.x2 = screen->rotatedWidth; + fullRect.y2 = screen->rotatedHeight; + intel->drawX = 0; + intel->drawY = 0; + intel->numClipRects = 1; + intel->pClipRects = &fullRect; + + intel->vtbl.meta_draw_region(intel, screen->rotated_region, NULL); /* ? */ + + intel_fb = dPriv->driverPrivate; + + if ((srcBuf == BUFFER_BIT_BACK_LEFT && !intel_fb->pf_active)) { + src = intel_get_rb_region(&intel_fb->Base, BUFFER_BACK_LEFT); + clipRects = dPriv->pBackClipRects; + numClipRects = dPriv->numBackClipRects; + } + else { + src = intel_get_rb_region(&intel_fb->Base, BUFFER_FRONT_LEFT); + clipRects = dPriv->pClipRects; + numClipRects = dPriv->numClipRects; + } + + if (src->cpp == 4) { + format = GL_BGRA; + type = GL_UNSIGNED_BYTE; + } + else { + format = GL_BGR; + type = GL_UNSIGNED_SHORT_5_6_5_REV; + } + + /* set the whole screen up as a texture to avoid alignment issues */ + intel->vtbl.meta_tex_rect_source(intel, + src->buffer, + screen->width, + screen->height, src->pitch, format, type); + + intel->vtbl.meta_texture_blend_replace(intel); + + /* + * loop over the source window's cliprects + */ + for (i = 0; i < numClipRects; i++) { + int srcX0 = clipRects[i].x1; + int srcY0 = clipRects[i].y1; + int srcX1 = clipRects[i].x2; + int srcY1 = clipRects[i].y2; + GLfloat verts[4][2], tex[4][2]; + int j; + + /* build vertices for four corners of clip rect */ + verts[0][0] = srcX0; + verts[0][1] = srcY0; + verts[1][0] = srcX1; + verts[1][1] = srcY0; + verts[2][0] = srcX1; + verts[2][1] = srcY1; + verts[3][0] = srcX0; + verts[3][1] = srcY1; + + /* .. and texcoords */ + tex[0][0] = srcX0; + tex[0][1] = srcY0; + tex[1][0] = srcX1; + tex[1][1] = srcY0; + tex[2][0] = srcX1; + tex[2][1] = srcY1; + tex[3][0] = srcX0; + tex[3][1] = srcY1; + + /* transform coords to rotated screen coords */ + + for (j = 0; j < 4; j++) { + matrix23TransformCoordf(&screen->rotMatrix, + &verts[j][0], &verts[j][1]); + } + + /* draw polygon to map source image to dest region */ + intel_meta_draw_poly(intel, 4, verts, 0, 0, tex); + + } /* cliprect loop */ + + intel->vtbl.leave_meta_state(intel); + intel_batchbuffer_flush(intel->batch); + + /* restore original drawing origin and cliprects */ + intel->drawX = xOrig; + intel->drawY = yOrig; + intel->numClipRects = origNumClipRects; + intel->pClipRects = origRects; + + UNLOCK_HARDWARE(intel); +} + + +/** + * Called by ctx->Driver.Clear. + */ +static void +intelClear(GLcontext *ctx, GLbitfield mask) +{ + struct intel_context *intel = intel_context(ctx); + const GLuint colorMask = *((GLuint *) & ctx->Color.ColorMask); + GLbitfield tri_mask = 0; + GLbitfield blit_mask = 0; + GLbitfield swrast_mask = 0; + struct gl_framebuffer *fb = ctx->DrawBuffer; + GLuint i; + + if (0) + fprintf(stderr, "%s\n", __FUNCTION__); + + /* HW color buffers (front, back, aux, generic FBO, etc) */ + if (colorMask == ~0) { + /* clear all R,G,B,A */ + /* XXX FBO: need to check if colorbuffers are software RBOs! */ + blit_mask |= (mask & BUFFER_BITS_COLOR); + } + else { + /* glColorMask in effect */ + tri_mask |= (mask & BUFFER_BITS_COLOR); + } + + /* HW stencil */ + if (mask & BUFFER_BIT_STENCIL) { + const struct intel_region *stencilRegion + = intel_get_rb_region(fb, BUFFER_STENCIL); + if (stencilRegion) { + /* have hw stencil */ + if ((ctx->Stencil.WriteMask[0] & 0xff) != 0xff) { + /* not clearing all stencil bits, so use triangle clearing */ + tri_mask |= BUFFER_BIT_STENCIL; + } + else { + /* clearing all stencil bits, use blitting */ + blit_mask |= BUFFER_BIT_STENCIL; + } + } + } + + /* HW depth */ + if (mask & BUFFER_BIT_DEPTH) { + /* clear depth with whatever method is used for stencil (see above) */ + if (tri_mask & BUFFER_BIT_STENCIL) + tri_mask |= BUFFER_BIT_DEPTH; + else + blit_mask |= BUFFER_BIT_DEPTH; + } + + /* SW fallback clearing */ + swrast_mask = mask & ~tri_mask & ~blit_mask; + + for (i = 0; i < BUFFER_COUNT; i++) { + GLuint bufBit = 1 << i; + if ((blit_mask | tri_mask) & bufBit) { + if (!fb->Attachment[i].Renderbuffer->ClassID) { + blit_mask &= ~bufBit; + tri_mask &= ~bufBit; + swrast_mask |= bufBit; + } + } + } + + + intelFlush(ctx); /* XXX intelClearWithBlit also does this */ + + if (blit_mask) + intelClearWithBlit(ctx, blit_mask); + + if (tri_mask) + intelClearWithTris(intel, tri_mask); + + if (swrast_mask) + _swrast_Clear(ctx, swrast_mask); +} + + +/* Emit wait for pending flips */ +void +intel_wait_flips(struct intel_context *intel, GLuint batch_flags) +{ + struct intel_framebuffer *intel_fb = + (struct intel_framebuffer *) intel->ctx.DrawBuffer; + struct intel_renderbuffer *intel_rb = + intel_get_renderbuffer(&intel_fb->Base, + intel_fb->Base._ColorDrawBufferMask[0] == + BUFFER_BIT_FRONT_LEFT ? BUFFER_FRONT_LEFT : + BUFFER_BACK_LEFT); + + if (intel_fb->Base.Name == 0 && intel_rb->pf_pending == intel_fb->pf_seq) { + GLint pf_planes = intel_fb->pf_planes; + BATCH_LOCALS; + + /* Wait for pending flips to take effect */ + BEGIN_BATCH(2, batch_flags); + OUT_BATCH(pf_planes & 0x1 ? (MI_WAIT_FOR_EVENT | MI_WAIT_FOR_PLANE_A_FLIP) + : 0); + OUT_BATCH(pf_planes & 0x2 ? (MI_WAIT_FOR_EVENT | MI_WAIT_FOR_PLANE_B_FLIP) + : 0); + ADVANCE_BATCH(); + + intel_rb->pf_pending--; + } +} + + +/* Flip the front & back buffers + */ +static GLboolean +intelPageFlip(const __DRIdrawablePrivate * dPriv) +{ + struct intel_context *intel; + int ret; + struct intel_framebuffer *intel_fb = dPriv->driverPrivate; + + if (INTEL_DEBUG & DEBUG_IOCTL) + fprintf(stderr, "%s\n", __FUNCTION__); + + assert(dPriv); + assert(dPriv->driContextPriv); + assert(dPriv->driContextPriv->driverPrivate); + + intel = (struct intel_context *) dPriv->driContextPriv->driverPrivate; + + if (intel->intelScreen->drmMinor < 9) + return GL_FALSE; + + intelFlush(&intel->ctx); + + ret = 0; + + LOCK_HARDWARE(intel); + + if (dPriv->numClipRects && intel_fb->pf_active) { + drm_i915_flip_t flip; + + flip.pipes = intel_fb->pf_planes; + + ret = drmCommandWrite(intel->driFd, DRM_I915_FLIP, &flip, sizeof(flip)); + } + + UNLOCK_HARDWARE(intel); + + if (ret || !intel_fb->pf_active) + return GL_FALSE; + + if (!dPriv->numClipRects) { + usleep(10000); /* throttle invisible client 10ms */ + } + + intel_fb->pf_current_page = (intel->sarea->pf_current_page >> + (intel_fb->pf_planes & 0x2)) & 0x3; + + if (dPriv->numClipRects != 0) { + intel_get_renderbuffer(&intel_fb->Base, BUFFER_FRONT_LEFT)->pf_pending = + intel_get_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT)->pf_pending = + ++intel_fb->pf_seq; + } + + intel_flip_renderbuffers(intel_fb); + intel_draw_buffer(&intel->ctx, &intel_fb->Base); + + return GL_TRUE; +} + +#if 0 +void +intelSwapBuffers(__DRIdrawablePrivate * dPriv) +{ + if (dPriv->driverPrivate) { + const struct gl_framebuffer *fb + = (struct gl_framebuffer *) dPriv->driverPrivate; + if (fb->Visual.doubleBufferMode) { + GET_CURRENT_CONTEXT(ctx); + if (ctx && ctx->DrawBuffer == fb) { + _mesa_notifySwapBuffers(ctx); /* flush pending rendering */ + } + if (intel->doPageFlip) { + intelPageFlip(dPriv); + } + else { + intelCopyBuffer(dPriv); + } + } + } + else { + _mesa_problem(NULL, + "dPriv has no gl_framebuffer pointer in intelSwapBuffers"); + } +} +#else +/* Trunk version: + */ + +static GLboolean +intelScheduleSwap(const __DRIdrawablePrivate * dPriv, GLboolean *missed_target) +{ + struct intel_framebuffer *intel_fb = dPriv->driverPrivate; + unsigned int interval = driGetVBlankInterval(dPriv, intel_fb->vblank_flags); + struct intel_context *intel = + intelScreenContext(dPriv->driScreenPriv->private); + const intelScreenPrivate *intelScreen = intel->intelScreen; + unsigned int target; + drm_i915_vblank_swap_t swap; + GLboolean ret; + + if (!intel_fb->vblank_flags || + (intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ) || + intelScreen->current_rotation != 0 || + intelScreen->drmMinor < (intel_fb->pf_active ? 9 : 6)) + return GL_FALSE; + + swap.seqtype = DRM_VBLANK_ABSOLUTE; + + if (intel_fb->vblank_flags & VBLANK_FLAG_SYNC) { + swap.seqtype |= DRM_VBLANK_NEXTONMISS; + } else if (interval == 0) { + return GL_FALSE; + } + + swap.drawable = dPriv->hHWDrawable; + target = swap.sequence = intel_fb->vbl_seq + interval; + + if ( intel_fb->vblank_flags & VBLANK_FLAG_SECONDARY ) { + swap.seqtype |= DRM_VBLANK_SECONDARY; + } + + LOCK_HARDWARE(intel); + + intel_batchbuffer_flush(intel->batch); + + if ( intel_fb->pf_active ) { + swap.seqtype |= DRM_VBLANK_FLIP; + + intel_fb->pf_current_page = (((intel->sarea->pf_current_page >> + (intel_fb->pf_planes & 0x2)) & 0x3) + 1) % + intel_fb->pf_num_pages; + } + + if (!drmCommandWriteRead(intel->driFd, DRM_I915_VBLANK_SWAP, &swap, + sizeof(swap))) { + intel_fb->vbl_seq = swap.sequence; + swap.sequence -= target; + *missed_target = swap.sequence > 0 && swap.sequence <= (1 << 23); + + intel_get_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT)->vbl_pending = + intel_get_renderbuffer(&intel_fb->Base, + BUFFER_FRONT_LEFT)->vbl_pending = + intel_fb->vbl_seq; + + if (swap.seqtype & DRM_VBLANK_FLIP) { + intel_flip_renderbuffers(intel_fb); + intel_draw_buffer(&intel->ctx, intel->ctx.DrawBuffer); + } + + ret = GL_TRUE; + } else { + if (swap.seqtype & DRM_VBLANK_FLIP) { + intel_fb->pf_current_page = ((intel->sarea->pf_current_page >> + (intel_fb->pf_planes & 0x2)) & 0x3) % + intel_fb->pf_num_pages; + } + + ret = GL_FALSE; + } + + UNLOCK_HARDWARE(intel); + + return ret; +} + +void +intelSwapBuffers(__DRIdrawablePrivate * dPriv) +{ + if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { + GET_CURRENT_CONTEXT(ctx); + struct intel_context *intel; + + if (ctx == NULL) + return; + + intel = intel_context(ctx); + + if (ctx->Visual.doubleBufferMode) { + intelScreenPrivate *screen = intel->intelScreen; + GLboolean missed_target; + struct intel_framebuffer *intel_fb = dPriv->driverPrivate; + int64_t ust; + + _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ + + if (screen->current_rotation != 0 || + !intelScheduleSwap(dPriv, &missed_target)) { + driWaitForVBlank(dPriv, &intel_fb->vbl_seq, intel_fb->vblank_flags, + &missed_target); + + if (screen->current_rotation != 0 || !intelPageFlip(dPriv)) { + intelCopyBuffer(dPriv, NULL); + } + + if (screen->current_rotation != 0) { + intelRotateWindow(intel, dPriv, BUFFER_BIT_FRONT_LEFT); + } + } + + intel_fb->swap_count++; + (*dri_interface->getUST) (&ust); + if (missed_target) { + intel_fb->swap_missed_count++; + intel_fb->swap_missed_ust = ust - intel_fb->swap_ust; + } + + intel_fb->swap_ust = ust; + } + } + else { + /* XXX this shouldn't be an error but we can't handle it for now */ + fprintf(stderr, "%s: drawable has no context!\n", __FUNCTION__); + } +} +#endif + +void +intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) +{ + if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { + struct intel_context *intel = + (struct intel_context *) dPriv->driContextPriv->driverPrivate; + GLcontext *ctx = &intel->ctx; + + if (ctx->Visual.doubleBufferMode) { + drm_clip_rect_t rect; + rect.x1 = x + dPriv->x; + rect.y1 = (dPriv->h - y - h) + dPriv->y; + rect.x2 = rect.x1 + w; + rect.y2 = rect.y1 + h; + _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ + intelCopyBuffer(dPriv, &rect); + } + } + else { + /* XXX this shouldn't be an error but we can't handle it for now */ + fprintf(stderr, "%s: drawable has no context!\n", __FUNCTION__); + } +} + + +/** + * Update the hardware state for drawing into a window or framebuffer object. + * + * Called by glDrawBuffer, glBindFramebufferEXT, MakeCurrent, and other + * places within the driver. + * + * Basically, this needs to be called any time the current framebuffer + * changes, the renderbuffers change, or we need to draw into different + * color buffers. + */ +void +intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_region *colorRegion, *depthRegion = NULL; + struct intel_renderbuffer *irbDepth = NULL, *irbStencil = NULL; + int front = 0; /* drawing to front color buffer? */ + + if (!fb) { + /* this can happen during the initial context initialization */ + return; + } + + /* Do this here, note core Mesa, since this function is called from + * many places within the driver. + */ + if (ctx->NewState & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) { + /* this updates the DrawBuffer->_NumColorDrawBuffers fields, etc */ + _mesa_update_framebuffer(ctx); + /* this updates the DrawBuffer's Width/Height if it's a FBO */ + _mesa_update_draw_buffer_bounds(ctx); + } + + if (fb->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) { + /* this may occur when we're called by glBindFrameBuffer() during + * the process of someone setting up renderbuffers, etc. + */ + /*_mesa_debug(ctx, "DrawBuffer: incomplete user FBO\n");*/ + return; + } + + if (fb->Name) + intel_validate_paired_depth_stencil(ctx, fb); + + /* + * How many color buffers are we drawing into? + */ + if (fb->_NumColorDrawBuffers[0] != 1 +#if 0 + /* XXX FBO temporary - always use software rendering */ + || 1 +#endif + ) { + /* writing to 0 or 2 or 4 color buffers */ + /*_mesa_debug(ctx, "Software rendering\n");*/ + FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_TRUE); + front = 1; /* might not have back color buffer */ + } + else { + /* draw to exactly one color buffer */ + /*_mesa_debug(ctx, "Hardware rendering\n");*/ + FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_FALSE); + if (fb->_ColorDrawBufferMask[0] == BUFFER_BIT_FRONT_LEFT) { + front = 1; + } + } + + /* + * Get the intel_renderbuffer for the colorbuffer we're drawing into. + * And set up cliprects. + */ + if (fb->Name == 0) { + /* drawing to window system buffer */ + if (front) { + intelSetFrontClipRects(intel); + colorRegion = intel_get_rb_region(fb, BUFFER_FRONT_LEFT); + } + else { + intelSetBackClipRects(intel); + colorRegion = intel_get_rb_region(fb, BUFFER_BACK_LEFT); + } + } + else { + /* drawing to user-created FBO */ + struct intel_renderbuffer *irb; + intelSetRenderbufferClipRects(intel); + irb = intel_renderbuffer(fb->_ColorDrawBuffers[0][0]); + colorRegion = (irb && irb->region) ? irb->region : NULL; + } + + /* Update culling direction which changes depending on the + * orientation of the buffer: + */ + if (ctx->Driver.FrontFace) + ctx->Driver.FrontFace(ctx, ctx->Polygon.FrontFace); + else + ctx->NewState |= _NEW_POLYGON; + + if (!colorRegion) { + FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_TRUE); + } + else { + FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_FALSE); + } + + /*** + *** Get depth buffer region and check if we need a software fallback. + *** Note that the depth buffer is usually a DEPTH_STENCIL buffer. + ***/ + if (fb->_DepthBuffer && fb->_DepthBuffer->Wrapped) { + irbDepth = intel_renderbuffer(fb->_DepthBuffer->Wrapped); + if (irbDepth && irbDepth->region) { + FALLBACK(intel, INTEL_FALLBACK_DEPTH_BUFFER, GL_FALSE); + depthRegion = irbDepth->region; + } + else { + FALLBACK(intel, INTEL_FALLBACK_DEPTH_BUFFER, GL_TRUE); + depthRegion = NULL; + } + } + else { + /* not using depth buffer */ + FALLBACK(intel, INTEL_FALLBACK_DEPTH_BUFFER, GL_FALSE); + depthRegion = NULL; + } + + /*** + *** Stencil buffer + *** This can only be hardware accelerated if we're using a + *** combined DEPTH_STENCIL buffer (for now anyway). + ***/ + if (fb->_StencilBuffer && fb->_StencilBuffer->Wrapped) { + irbStencil = intel_renderbuffer(fb->_StencilBuffer->Wrapped); + if (irbStencil && irbStencil->region) { + ASSERT(irbStencil->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); + FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_FALSE); + /* need to re-compute stencil hw state */ + ctx->Driver.Enable(ctx, GL_STENCIL_TEST, ctx->Stencil.Enabled); + if (!depthRegion) + depthRegion = irbStencil->region; + } + else { + FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_TRUE); + } + } + else { + /* XXX FBO: instead of FALSE, pass ctx->Stencil.Enabled ??? */ + FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_FALSE); + /* need to re-compute stencil hw state */ + ctx->Driver.Enable(ctx, GL_STENCIL_TEST, ctx->Stencil.Enabled); + } + + /* + * Update depth test state + */ + if (ctx->Depth.Test && fb->Visual.depthBits > 0) { + ctx->Driver.Enable(ctx, GL_DEPTH_TEST, GL_TRUE); + } + else { + ctx->Driver.Enable(ctx, GL_DEPTH_TEST, GL_FALSE); + } + + /** + ** Release old regions, reference new regions + **/ +#if 0 /* XXX FBO: this seems to be redundant with i915_state_draw_region() */ + if (intel->draw_region != colorRegion) { + intel_region_release(&intel->draw_region); + intel_region_reference(&intel->draw_region, colorRegion); + } + if (intel->intelScreen->depth_region != depthRegion) { + intel_region_release(&intel->intelScreen->depth_region); + intel_region_reference(&intel->intelScreen->depth_region, depthRegion); + } +#endif + + intel->vtbl.set_draw_region(intel, colorRegion, depthRegion); + + /* update viewport since it depends on window size */ + ctx->Driver.Viewport(ctx, ctx->Viewport.X, ctx->Viewport.Y, + ctx->Viewport.Width, ctx->Viewport.Height); + + /* Update hardware scissor */ + ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y, + ctx->Scissor.Width, ctx->Scissor.Height); +} + + +static void +intelDrawBuffer(GLcontext * ctx, GLenum mode) +{ + intel_draw_buffer(ctx, ctx->DrawBuffer); +} + + +static void +intelReadBuffer(GLcontext * ctx, GLenum mode) +{ + if (ctx->ReadBuffer == ctx->DrawBuffer) { + /* This will update FBO completeness status. + * A framebuffer will be incomplete if the GL_READ_BUFFER setting + * refers to a missing renderbuffer. Calling glReadBuffer can set + * that straight and can make the drawing buffer complete. + */ + intel_draw_buffer(ctx, ctx->DrawBuffer); + } + /* Generally, functions which read pixels (glReadPixels, glCopyPixels, etc) + * reference ctx->ReadBuffer and do appropriate state checks. + */ +} + + +void +intelInitBufferFuncs(struct dd_function_table *functions) +{ + functions->Clear = intelClear; + functions->DrawBuffer = intelDrawBuffer; + functions->ReadBuffer = intelReadBuffer; +} diff --git a/src/mesa/drivers/dri/i915/intel_buffers.h b/src/mesa/drivers/dri/i915/intel_buffers.h new file mode 100644 index 0000000000..3b686cb5c1 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_buffers.h @@ -0,0 +1,59 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_BUFFERS_H +#define INTEL_BUFFERS_H + + +struct intel_context; +struct intel_framebuffer; + + +extern GLboolean +intel_intersect_cliprects(drm_clip_rect_t * dest, + const drm_clip_rect_t * a, + const drm_clip_rect_t * b); + +extern struct intel_region *intel_readbuf_region(struct intel_context *intel); + +extern struct intel_region *intel_drawbuf_region(struct intel_context *intel); + +extern void intel_wait_flips(struct intel_context *intel, GLuint batch_flags); + +extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv); + +extern void intelWindowMoved(struct intel_context *intel); + +extern void intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb); + +extern void intelInitBufferFuncs(struct dd_function_table *functions); + +extern void +intelRotateWindow(struct intel_context *intel, + __DRIdrawablePrivate * dPriv, GLuint srcBuf); + +#endif /* INTEL_BUFFERS_H */ diff --git a/src/mesa/drivers/dri/i915/intel_bufmgr_ttm.c b/src/mesa/drivers/dri/i915/intel_bufmgr_ttm.c new file mode 100644 index 0000000000..32d9886091 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_bufmgr_ttm.c @@ -0,0 +1,833 @@ +/************************************************************************** + * + * Copyright © 2007 Red Hat Inc. + * Copyright © 2007 Intel Corporation + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Thomas Hellström <thomas-at-tungstengraphics-dot-com> + * Keith Whitwell <keithw-at-tungstengraphics-dot-com> + * Eric Anholt <eric@anholt.net> + * Dave Airlie <airlied@linux.ie> + */ + +#include <xf86drm.h> +#include <stdlib.h> +#include <unistd.h> +#include "glthread.h" +#include "errno.h" +#include "mtypes.h" +#include "dri_bufmgr.h" +#include "string.h" +#include "imports.h" + +#include "i915_drm.h" + +#include "intel_bufmgr_ttm.h" + +#define BUFMGR_DEBUG 0 + +struct intel_reloc_info +{ + GLuint type; + GLuint reloc; + GLuint delta; /* not needed? */ + GLuint index; + drm_handle_t handle; +}; + +struct intel_bo_node +{ + drmMMListHead head; + drmBO *buf; + struct drm_i915_op_arg bo_arg; + unsigned long arg0; + unsigned long arg1; + void (*destroy)(void *); + void *priv; +}; + +struct intel_bo_reloc_list +{ + drmMMListHead head; + drmBO buf; + uint32_t *relocs; +}; + +struct intel_bo_reloc_node +{ + drmMMListHead head; + drm_handle_t handle; + uint32_t nr_reloc_types; + struct intel_bo_reloc_list type_list; +}; + +struct intel_bo_list { + unsigned numCurrent; + drmMMListHead list; + void (*destroy)(void *node); +}; + +typedef struct _dri_bufmgr_ttm { + dri_bufmgr bufmgr; + + int fd; + _glthread_Mutex mutex; + unsigned int fence_type; + unsigned int fence_type_flush; + + uint32_t max_relocs; + /** ttm relocation list */ + struct intel_bo_list list; + struct intel_bo_list reloc_list; + +} dri_bufmgr_ttm; + +typedef struct _dri_bo_ttm { + dri_bo bo; + + int refcount; /* Protected by bufmgr->mutex */ + drmBO drm_bo; + const char *name; +} dri_bo_ttm; + +typedef struct _dri_fence_ttm +{ + dri_fence fence; + + int refcount; /* Protected by bufmgr->mutex */ + const char *name; + drmFence drm_fence; +} dri_fence_ttm; + + +static void intel_bo_free_list(struct intel_bo_list *list) +{ + struct intel_bo_node *node; + drmMMListHead *l; + + l = list->list.next; + while(l != &list->list) { + DRMLISTDEL(l); + node = DRMLISTENTRY(struct intel_bo_node, l, head); + list->destroy(node); + l = list->list.next; + list->numCurrent--; + } +} + +static void generic_destroy(void *nodep) +{ + free(nodep); +} + +static int intel_create_bo_list(int numTarget, struct intel_bo_list *list, void (*destroy)(void *)) +{ + DRMINITLISTHEAD(&list->list); + list->numCurrent = 0; + if (destroy) + list->destroy = destroy; + else + list->destroy = generic_destroy; + return 0; +} + + +static struct drm_i915_op_arg * +intel_setup_validate_list(int fd, struct intel_bo_list *list, struct intel_bo_list *reloc_list, GLuint *count_p) +{ + struct intel_bo_node *node; + struct intel_bo_reloc_node *rl_node; + drmMMListHead *l, *rl; + struct drm_i915_op_arg *arg, *first; + struct drm_bo_op_req *req; + uint64_t *prevNext = NULL; + GLuint count = 0; + + first = NULL; + + for (l = list->list.next; l != &list->list; l = l->next) { + node = DRMLISTENTRY(struct intel_bo_node, l, head); + + arg = &node->bo_arg; + req = &arg->d.req; + + if (!first) + first = arg; + + if (prevNext) + *prevNext = (unsigned long) arg; + + memset(arg, 0, sizeof(*arg)); + prevNext = &arg->next; + req->bo_req.handle = node->buf->handle; + req->op = drm_bo_validate; + req->bo_req.flags = node->arg0; + req->bo_req.hint = 0; + req->bo_req.mask = node->arg1; + req->bo_req.fence_class = 0; /* Backwards compat. */ + arg->reloc_handle = 0; + + for (rl = reloc_list->list.next; rl != &reloc_list->list; rl = rl->next) { + rl_node = DRMLISTENTRY(struct intel_bo_reloc_node, rl, head); + + if (rl_node->handle == node->buf->handle) { + arg->reloc_handle = rl_node->type_list.buf.handle; + } + } + count++; + } + + if (!first) + return 0; + + *count_p = count; + return first; +} + +static void intel_free_validate_list(int fd, struct intel_bo_list *list) +{ + struct intel_bo_node *node; + drmMMListHead *l; + + for (l = list->list.next; l != &list->list; l = l->next) { + node = DRMLISTENTRY(struct intel_bo_node, l, head); + + if (node->destroy) + (*node->destroy)(node->priv); + + } +} + +static void intel_free_reloc_list(int fd, struct intel_bo_list *reloc_list) +{ + struct intel_bo_reloc_node *reloc_node; + drmMMListHead *rl, *tmp; + + for (rl = reloc_list->list.next, tmp = rl->next; rl != &reloc_list->list; rl = tmp, tmp = rl->next) { + reloc_node = DRMLISTENTRY(struct intel_bo_reloc_node, rl, head); + + DRMLISTDEL(rl); + + if (reloc_node->nr_reloc_types > 1) { + + /* TODO */ + } + + drmBOUnmap(fd, &reloc_node->type_list.buf); + drmBOUnreference(fd, &reloc_node->type_list.buf); + free(reloc_node); + } +} + +static int intel_add_validate_buffer(struct intel_bo_list *list, dri_bo *buf, unsigned flags, + unsigned mask, int *itemLoc, void (*destroy_cb)(void *)) +{ + struct intel_bo_node *node, *cur; + drmMMListHead *l; + int count = 0; + int ret = 0; + drmBO *buf_bo = &((dri_bo_ttm *)buf)->drm_bo; + cur = NULL; + + for (l = list->list.next; l != &list->list; l = l->next) { + node = DRMLISTENTRY(struct intel_bo_node, l, head); + if (node->buf->handle == buf_bo->handle) { + cur = node; + break; + } + count++; + } + + if (!cur) { + cur = drmMalloc(sizeof(*cur)); + if (!cur) { + return -ENOMEM; + } + cur->buf = buf_bo; + cur->priv = buf; + cur->arg0 = flags; + cur->arg1 = mask; + cur->destroy = destroy_cb; + ret = 1; + + DRMLISTADDTAIL(&cur->head, &list->list); + + } else { + unsigned memMask = (cur->arg1 | mask) & DRM_BO_MASK_MEM; + unsigned memFlags = cur->arg0 & flags & memMask; + + if (!memFlags) { + return -EINVAL; + } + if (mask & cur->arg1 & ~DRM_BO_MASK_MEM & (cur->arg0 ^ flags)) { + return -EINVAL; + } + cur->arg1 |= mask; + cur->arg0 = memFlags | ((cur->arg0 | flags) & + cur->arg1 & ~DRM_BO_MASK_MEM); + } + *itemLoc = count; + return ret; +} + + +#define RELOC_BUF_SIZE(x) ((I915_RELOC_HEADER + x * I915_RELOC0_STRIDE) * sizeof(uint32_t)) + +static int intel_create_new_reloc_type_list(int fd, struct intel_bo_reloc_list *cur_type, int max_relocs) +{ + int ret; + + /* should allocate a drmBO here */ + ret = drmBOCreate(fd, RELOC_BUF_SIZE(max_relocs), 0, + NULL, + DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE | DRM_BO_FLAG_MAPPABLE | DRM_BO_FLAG_CACHED, + 0, &cur_type->buf); + if (ret) + return ret; + + ret = drmBOMap(fd, &cur_type->buf, DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0, (void **)&cur_type->relocs); + if (ret) + return ret; + return 0; +} + + +static int intel_add_validate_reloc(int fd, struct intel_bo_list *reloc_list, struct intel_reloc_info *reloc_info, uint32_t max_relocs) +{ + struct intel_bo_reloc_node *rl_node, *cur; + drmMMListHead *rl, *l; + int ret = 0; + uint32_t *reloc_start; + int num_relocs; + struct intel_bo_reloc_list *cur_type; + + cur = NULL; + + for (rl = reloc_list->list.next; rl != &reloc_list->list; rl = rl->next) { + rl_node = DRMLISTENTRY(struct intel_bo_reloc_node, rl, head); + if (rl_node->handle == reloc_info->handle) { + cur = rl_node; + break; + } + } + + if (!cur) { + + cur = malloc(sizeof(*cur)); + if (!cur) + return -ENOMEM; + + cur->nr_reloc_types = 1; + cur->handle = reloc_info->handle; + cur_type = &cur->type_list; + + DRMINITLISTHEAD(&cur->type_list.head); + ret = intel_create_new_reloc_type_list(fd, cur_type, max_relocs); + if (ret) { + return -1; + } + DRMLISTADDTAIL(&cur->head, &reloc_list->list); + + cur_type->relocs[0] = 0 | (reloc_info->type << 16); + cur_type->relocs[1] = 0; // next reloc buffer handle is 0 + + } else { + int found = 0; + if ((cur->type_list.relocs[0] >> 16) == reloc_info->type) { + cur_type = &cur->type_list; + found = 1; + } else { + for (l = cur->type_list.head.next; l != &cur->type_list.head; l = l->next) { + cur_type = DRMLISTENTRY(struct intel_bo_reloc_list, l, head); + if (((cur_type->relocs[0] >> 16) & 0xffff) == reloc_info->type) + found = 1; + break; + } + } + + /* didn't find the relocation type */ + if (!found) { + cur_type = malloc(sizeof(*cur_type)); + if (!cur_type) { + return -ENOMEM; + } + + ret = intel_create_new_reloc_type_list(fd, cur_type, max_relocs); + DRMLISTADDTAIL(&cur_type->head, &cur->type_list.head); + + cur_type->relocs[0] = (reloc_info->type << 16); + cur_type->relocs[1] = 0; + + cur->nr_reloc_types++; + } + } + + reloc_start = cur_type->relocs; + + num_relocs = (reloc_start[0] & 0xffff); + + reloc_start[num_relocs*I915_RELOC0_STRIDE + I915_RELOC_HEADER] = reloc_info->reloc; + reloc_start[num_relocs*I915_RELOC0_STRIDE + I915_RELOC_HEADER+1] = reloc_info->delta; + reloc_start[num_relocs*I915_RELOC0_STRIDE + I915_RELOC_HEADER+2] = reloc_info->index; + reloc_start[0]++; + if (((reloc_start[0] & 0xffff)) > (max_relocs)) { + return -ENOMEM; + } + return 0; +} + + +#if 0 +int +driFenceSignaled(DriFenceObject * fence, unsigned type) +{ + int signaled; + int ret; + + if (fence == NULL) + return GL_TRUE; + + _glthread_LOCK_MUTEX(fence->mutex); + ret = drmFenceSignaled(bufmgr_ttm->fd, &fence->fence, type, &signaled); + _glthread_UNLOCK_MUTEX(fence->mutex); + BM_CKFATAL(ret); + return signaled; +} +#endif + +static dri_bo * +dri_ttm_alloc(dri_bufmgr *bufmgr, const char *name, + unsigned long size, unsigned int alignment, + unsigned int location_mask) +{ + dri_bufmgr_ttm *ttm_bufmgr; + dri_bo_ttm *ttm_buf; + unsigned int pageSize = getpagesize(); + int ret; + unsigned int flags, hint; + + ttm_bufmgr = (dri_bufmgr_ttm *)bufmgr; + + ttm_buf = malloc(sizeof(*ttm_buf)); + if (!ttm_buf) + return NULL; + + /* The mask argument doesn't do anything for us that we want other than + * determine which pool (TTM or local) the buffer is allocated into, so just + * pass all of the allocation class flags. + */ + flags = location_mask | DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE | + DRM_BO_FLAG_EXE; + /* No hints we want to use. */ + hint = 0; + + ret = drmBOCreate(ttm_bufmgr->fd, size, alignment / pageSize, + NULL, flags, hint, &ttm_buf->drm_bo); + if (ret != 0) { + free(ttm_buf); + return NULL; + } + ttm_buf->bo.size = ttm_buf->drm_bo.size; + ttm_buf->bo.offset = ttm_buf->drm_bo.offset; + ttm_buf->bo.virtual = NULL; + ttm_buf->bo.bufmgr = bufmgr; + ttm_buf->name = name; + ttm_buf->refcount = 1; + +#if BUFMGR_DEBUG + fprintf(stderr, "bo_create: %p (%s)\n", &ttm_buf->bo, ttm_buf->name); +#endif + + return &ttm_buf->bo; +} + +/* Our TTM backend doesn't allow creation of static buffers, as that requires + * privelege for the non-fake case, and the lock in the fake case where we were + * working around the X Server not creating buffers and passing handles to us. + */ +static dri_bo * +dri_ttm_alloc_static(dri_bufmgr *bufmgr, const char *name, + unsigned long offset, unsigned long size, void *virtual, + unsigned int location_mask) +{ + return NULL; +} + +/** Returns a dri_bo wrapping the given buffer object handle. + * + * This can be used when one application needs to pass a buffer object + * to another. + */ +dri_bo * +intel_ttm_bo_create_from_handle(dri_bufmgr *bufmgr, const char *name, + unsigned int handle) +{ + dri_bufmgr_ttm *ttm_bufmgr; + dri_bo_ttm *ttm_buf; + int ret; + + ttm_bufmgr = (dri_bufmgr_ttm *)bufmgr; + + ttm_buf = malloc(sizeof(*ttm_buf)); + if (!ttm_buf) + return NULL; + + ret = drmBOReference(ttm_bufmgr->fd, handle, &ttm_buf->drm_bo); + if (ret != 0) { + free(ttm_buf); + return NULL; + } + ttm_buf->bo.size = ttm_buf->drm_bo.size; + ttm_buf->bo.offset = ttm_buf->drm_bo.offset; + ttm_buf->bo.virtual = NULL; + ttm_buf->bo.bufmgr = bufmgr; + ttm_buf->name = name; + ttm_buf->refcount = 1; + +#if BUFMGR_DEBUG + fprintf(stderr, "bo_create_from_handle: %p %08x (%s)\n", &ttm_buf->bo, handle, + ttm_buf->name); +#endif + + return &ttm_buf->bo; +} + +static void +dri_ttm_bo_reference(dri_bo *buf) +{ + dri_bufmgr_ttm *bufmgr_ttm = (dri_bufmgr_ttm *)buf->bufmgr; + dri_bo_ttm *ttm_buf = (dri_bo_ttm *)buf; + + _glthread_LOCK_MUTEX(bufmgr_ttm->mutex); + ttm_buf->refcount++; + _glthread_UNLOCK_MUTEX(bufmgr_ttm->mutex); +} + +static void +dri_ttm_bo_unreference(dri_bo *buf) +{ + dri_bufmgr_ttm *bufmgr_ttm = (dri_bufmgr_ttm *)buf->bufmgr; + dri_bo_ttm *ttm_buf = (dri_bo_ttm *)buf; + + if (!buf) + return; + + _glthread_LOCK_MUTEX(bufmgr_ttm->mutex); + if (--ttm_buf->refcount == 0) { + int ret; + + ret = drmBOUnreference(bufmgr_ttm->fd, &ttm_buf->drm_bo); + if (ret != 0) { + fprintf(stderr, "drmBOUnreference failed (%s): %s\n", ttm_buf->name, + strerror(-ret)); + } +#if BUFMGR_DEBUG + fprintf(stderr, "bo_unreference final: %p (%s)\n", + &ttm_buf->bo, ttm_buf->name); +#endif + _glthread_UNLOCK_MUTEX(bufmgr_ttm->mutex); + free(buf); + return; + } + _glthread_UNLOCK_MUTEX(bufmgr_ttm->mutex); +} + +static int +dri_ttm_bo_map(dri_bo *buf, GLboolean write_enable) +{ + dri_bufmgr_ttm *bufmgr_ttm; + dri_bo_ttm *ttm_buf = (dri_bo_ttm *)buf; + unsigned int flags; + + bufmgr_ttm = (dri_bufmgr_ttm *)buf->bufmgr; + + flags = DRM_BO_FLAG_READ; + if (write_enable) + flags |= DRM_BO_FLAG_WRITE; + + assert(buf->virtual == NULL); + +#if BUFMGR_DEBUG + fprintf(stderr, "bo_map: %p (%s)\n", &ttm_buf->bo, ttm_buf->name); +#endif + + return drmBOMap(bufmgr_ttm->fd, &ttm_buf->drm_bo, flags, 0, &buf->virtual); +} + +static int +dri_ttm_bo_unmap(dri_bo *buf) +{ + dri_bufmgr_ttm *bufmgr_ttm; + dri_bo_ttm *ttm_buf = (dri_bo_ttm *)buf; + + if (buf == NULL) + return 0; + + bufmgr_ttm = (dri_bufmgr_ttm *)buf->bufmgr; + + assert(buf->virtual != NULL); + + buf->virtual = NULL; + +#if BUFMGR_DEBUG + fprintf(stderr, "bo_unmap: %p (%s)\n", &ttm_buf->bo, ttm_buf->name); +#endif + + return drmBOUnmap(bufmgr_ttm->fd, &ttm_buf->drm_bo); +} + +/* Returns a dri_bo wrapping the given buffer object handle. + * + * This can be used when one application needs to pass a buffer object + * to another. + */ +dri_fence * +intel_ttm_fence_create_from_arg(dri_bufmgr *bufmgr, const char *name, + drm_fence_arg_t *arg) +{ + dri_bufmgr_ttm *ttm_bufmgr; + dri_fence_ttm *ttm_fence; + + ttm_bufmgr = (dri_bufmgr_ttm *)bufmgr; + + ttm_fence = malloc(sizeof(*ttm_fence)); + if (!ttm_fence) + return NULL; + + ttm_fence->drm_fence.handle = arg->handle; + ttm_fence->drm_fence.fence_class = arg->fence_class; + ttm_fence->drm_fence.type = arg->type; + ttm_fence->drm_fence.flags = arg->flags; + ttm_fence->drm_fence.signaled = 0; + ttm_fence->drm_fence.sequence = arg->sequence; + + ttm_fence->fence.bufmgr = bufmgr; + ttm_fence->name = name; + ttm_fence->refcount = 1; + +#if BUFMGR_DEBUG + fprintf(stderr, "fence_create_from_handle: %p (%s)\n", &ttm_fence->fence, + ttm_fence->name); +#endif + + return &ttm_fence->fence; +} + + +static void +dri_ttm_fence_reference(dri_fence *fence) +{ + dri_fence_ttm *fence_ttm = (dri_fence_ttm *)fence; + dri_bufmgr_ttm *bufmgr_ttm = (dri_bufmgr_ttm *)fence->bufmgr; + + _glthread_LOCK_MUTEX(bufmgr_ttm->mutex); + ++fence_ttm->refcount; + _glthread_UNLOCK_MUTEX(bufmgr_ttm->mutex); +#if BUFMGR_DEBUG + fprintf(stderr, "fence_reference: %p (%s)\n", &fence_ttm->fence, + fence_ttm->name); +#endif +} + +static void +dri_ttm_fence_unreference(dri_fence *fence) +{ + dri_fence_ttm *fence_ttm = (dri_fence_ttm *)fence; + dri_bufmgr_ttm *bufmgr_ttm = (dri_bufmgr_ttm *)fence->bufmgr; + + if (!fence) + return; + +#if BUFMGR_DEBUG + fprintf(stderr, "fence_unreference: %p (%s)\n", &fence_ttm->fence, + fence_ttm->name); +#endif + _glthread_LOCK_MUTEX(bufmgr_ttm->mutex); + if (--fence_ttm->refcount == 0) { + int ret; + + ret = drmFenceUnreference(bufmgr_ttm->fd, &fence_ttm->drm_fence); + if (ret != 0) { + fprintf(stderr, "drmFenceUnreference failed (%s): %s\n", + fence_ttm->name, strerror(-ret)); + } + + _glthread_UNLOCK_MUTEX(bufmgr_ttm->mutex); + free(fence); + return; + } + _glthread_UNLOCK_MUTEX(bufmgr_ttm->mutex); +} + +static void +dri_ttm_fence_wait(dri_fence *fence) +{ + dri_fence_ttm *fence_ttm = (dri_fence_ttm *)fence; + dri_bufmgr_ttm *bufmgr_ttm = (dri_bufmgr_ttm *)fence->bufmgr; + int ret; + + _glthread_LOCK_MUTEX(bufmgr_ttm->mutex); + ret = drmFenceWait(bufmgr_ttm->fd, 0, &fence_ttm->drm_fence, 0); + _glthread_UNLOCK_MUTEX(bufmgr_ttm->mutex); + if (ret != 0) { + _mesa_printf("%s:%d: Error %d waiting for fence %s.\n", + __FILE__, __LINE__, ret, fence_ttm->name); + abort(); + } + +#if BUFMGR_DEBUG + fprintf(stderr, "fence_wait: %p (%s)\n", &fence_ttm->fence, + fence_ttm->name); +#endif +} + +static void +dri_bufmgr_ttm_destroy(dri_bufmgr *bufmgr) +{ + dri_bufmgr_ttm *bufmgr_ttm = (dri_bufmgr_ttm *)bufmgr; + + intel_bo_free_list(&bufmgr_ttm->list); + intel_bo_free_list(&bufmgr_ttm->reloc_list); + + _glthread_DESTROY_MUTEX(bufmgr_ttm->mutex); + free(bufmgr); +} + + +static void intel_dribo_destroy_callback(void *priv) +{ + dri_bo *dribo = priv; + + if (dribo) { + dri_bo_unreference(dribo); + } +} + +static void +dri_ttm_emit_reloc(dri_bo *batch_buf, GLuint flags, GLuint delta, GLuint offset, + dri_bo *relocatee) +{ + dri_bo_ttm *ttm_buf = (dri_bo_ttm *)batch_buf; + dri_bufmgr_ttm *bufmgr_ttm = (dri_bufmgr_ttm *)batch_buf->bufmgr; + int newItem; + struct intel_reloc_info reloc; + int mask; + int ret; + + mask = DRM_BO_MASK_MEM; + mask |= flags & (DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE | DRM_BO_FLAG_EXE); + + ret = intel_add_validate_buffer(&bufmgr_ttm->list, relocatee, flags, mask, &newItem, intel_dribo_destroy_callback); + if (ret < 0) + return; + + if (ret == 1) { + dri_bo_reference(relocatee); + } + + reloc.type = I915_RELOC_TYPE_0; + reloc.reloc = offset; + reloc.delta = delta; + reloc.index = newItem; + reloc.handle = ttm_buf->drm_bo.handle; + + intel_add_validate_reloc(bufmgr_ttm->fd, &bufmgr_ttm->reloc_list, &reloc, bufmgr_ttm->max_relocs); + return; +} + + +static void * +dri_ttm_process_reloc(dri_bo *batch_buf, GLuint *count) +{ + dri_bufmgr_ttm *bufmgr_ttm = (dri_bufmgr_ttm *)batch_buf->bufmgr; + void *ptr; + int itemLoc; + + dri_bo_unmap(batch_buf); + + intel_add_validate_buffer(&bufmgr_ttm->list, batch_buf, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_EXE, + DRM_BO_MASK_MEM | DRM_BO_FLAG_EXE, &itemLoc, NULL); + + ptr = intel_setup_validate_list(bufmgr_ttm->fd, &bufmgr_ttm->list, &bufmgr_ttm->reloc_list, count); + + return ptr; +} + +static void +dri_ttm_post_submit(dri_bo *batch_buf, dri_fence **last_fence) +{ + dri_bufmgr_ttm *bufmgr_ttm = (dri_bufmgr_ttm *)batch_buf->bufmgr; + + intel_free_validate_list(bufmgr_ttm->fd, &bufmgr_ttm->list); + intel_free_reloc_list(bufmgr_ttm->fd, &bufmgr_ttm->reloc_list); + + intel_bo_free_list(&bufmgr_ttm->list); +} + +/** + * Initializes the TTM buffer manager, which uses the kernel to allocate, map, + * and manage map buffer objections. + * + * \param fd File descriptor of the opened DRM device. + * \param fence_type Driver-specific fence type used for fences with no flush. + * \param fence_type_flush Driver-specific fence type used for fences with a + * flush. + */ +dri_bufmgr * +intel_bufmgr_ttm_init(int fd, unsigned int fence_type, + unsigned int fence_type_flush, int batch_size) +{ + dri_bufmgr_ttm *bufmgr_ttm; + + bufmgr_ttm = malloc(sizeof(*bufmgr_ttm)); + bufmgr_ttm->fd = fd; + bufmgr_ttm->fence_type = fence_type; + bufmgr_ttm->fence_type_flush = fence_type_flush; + _glthread_INIT_MUTEX(bufmgr_ttm->mutex); + + /* lets go with one relocation per every four dwords - purely heuristic */ + bufmgr_ttm->max_relocs = batch_size / sizeof(uint32_t) / 4; + + intel_create_bo_list(10, &bufmgr_ttm->list, NULL); + intel_create_bo_list(1, &bufmgr_ttm->reloc_list, NULL); + + bufmgr_ttm->bufmgr.bo_alloc = dri_ttm_alloc; + bufmgr_ttm->bufmgr.bo_alloc_static = dri_ttm_alloc_static; + bufmgr_ttm->bufmgr.bo_reference = dri_ttm_bo_reference; + bufmgr_ttm->bufmgr.bo_unreference = dri_ttm_bo_unreference; + bufmgr_ttm->bufmgr.bo_map = dri_ttm_bo_map; + bufmgr_ttm->bufmgr.bo_unmap = dri_ttm_bo_unmap; + bufmgr_ttm->bufmgr.fence_reference = dri_ttm_fence_reference; + bufmgr_ttm->bufmgr.fence_unreference = dri_ttm_fence_unreference; + bufmgr_ttm->bufmgr.fence_wait = dri_ttm_fence_wait; + bufmgr_ttm->bufmgr.destroy = dri_bufmgr_ttm_destroy; + bufmgr_ttm->bufmgr.emit_reloc = dri_ttm_emit_reloc; + bufmgr_ttm->bufmgr.process_relocs = dri_ttm_process_reloc; + bufmgr_ttm->bufmgr.post_submit = dri_ttm_post_submit; + return &bufmgr_ttm->bufmgr; +} + diff --git a/src/mesa/drivers/dri/i915/intel_bufmgr_ttm.h b/src/mesa/drivers/dri/i915/intel_bufmgr_ttm.h new file mode 100644 index 0000000000..0738839cef --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_bufmgr_ttm.h @@ -0,0 +1,17 @@ + +#ifndef INTEL_BUFMGR_TTM_H +#define INTEL_BUFMGR_TTM_H + +#include "dri_bufmgr.h" + +extern dri_bo *intel_ttm_bo_create_from_handle(dri_bufmgr *bufmgr, const char *name, + unsigned int handle); + +dri_fence *intel_ttm_fence_create_from_arg(dri_bufmgr *bufmgr, const char *name, + drm_fence_arg_t *arg); + + +dri_bufmgr *intel_bufmgr_ttm_init(int fd, unsigned int fence_type, + unsigned int fence_type_flush, int batch_size); + +#endif diff --git a/src/mesa/drivers/dri/i915/intel_context.c b/src/mesa/drivers/dri/i915/intel_context.c index 11c23f24a1..d7af432ad6 100644 --- a/src/mesa/drivers/dri/i915/intel_context.c +++ b/src/mesa/drivers/dri/i915/intel_context.c @@ -38,7 +38,6 @@ #include "swrast/swrast.h" #include "swrast_setup/swrast_setup.h" #include "tnl/tnl.h" -#include "vbo/vbo.h" #include "tnl/t_pipeline.h" #include "tnl/t_vertex.h" @@ -48,17 +47,24 @@ #include "intel_screen.h" #include "i830_dri.h" -#include "i830_common.h" +#include "intel_buffers.h" #include "intel_tex.h" #include "intel_span.h" #include "intel_tris.h" #include "intel_ioctl.h" #include "intel_batchbuffer.h" - +#include "intel_blit.h" +#include "intel_pixel.h" +#include "intel_regions.h" +#include "intel_buffer_objects.h" +#include "intel_fbo.h" +#include "intel_decode.h" + +#include "drirenderbuffer.h" #include "vblank.h" #include "utils.h" -#include "xmlpool.h" /* for symbolic values of enum-type options */ +#include "xmlpool.h" /* for symbolic values of enum-type options */ #ifndef INTEL_DEBUG int INTEL_DEBUG = (0); #endif @@ -75,67 +81,74 @@ int INTEL_DEBUG = (0); #define need_GL_EXT_blend_minmax #define need_GL_EXT_cull_vertex #define need_GL_EXT_fog_coord +#define need_GL_EXT_framebuffer_object #define need_GL_EXT_multi_draw_arrays #define need_GL_EXT_secondary_color #define need_GL_NV_vertex_program #include "extension_helper.h" -#ifndef VERBOSE -int VERBOSE = 0; -#endif -#if DEBUG_LOCKING -char *prevLockFile; -int prevLockLine; -#endif +#define DRIVER_DATE "20061102" -/*************************************** - * Mesa's Driver Functions - ***************************************/ +_glthread_Mutex lockMutex; +static GLboolean lockMutexInit = GL_FALSE; -#define DRIVER_DATE "20061017" -const GLubyte *intelGetString( GLcontext *ctx, GLenum name ) +static const GLubyte * +intelGetString(GLcontext * ctx, GLenum name) { - const char * chipset; + const char *chipset; static char buffer[128]; switch (name) { case GL_VENDOR: - return (GLubyte *)"Tungsten Graphics, Inc"; + return (GLubyte *) "Tungsten Graphics, Inc"; break; - + case GL_RENDERER: - switch (INTEL_CONTEXT(ctx)->intelScreen->deviceID) { + switch (intel_context(ctx)->intelScreen->deviceID) { case PCI_CHIP_845_G: - chipset = "Intel(R) 845G"; break; + chipset = "Intel(R) 845G"; + break; case PCI_CHIP_I830_M: - chipset = "Intel(R) 830M"; break; + chipset = "Intel(R) 830M"; + break; case PCI_CHIP_I855_GM: - chipset = "Intel(R) 852GM/855GM"; break; + chipset = "Intel(R) 852GM/855GM"; + break; case PCI_CHIP_I865_G: - chipset = "Intel(R) 865G"; break; + chipset = "Intel(R) 865G"; + break; case PCI_CHIP_I915_G: - chipset = "Intel(R) 915G"; break; + chipset = "Intel(R) 915G"; + break; case PCI_CHIP_I915_GM: - chipset = "Intel(R) 915GM"; break; + chipset = "Intel(R) 915GM"; + break; case PCI_CHIP_I945_G: - chipset = "Intel(R) 945G"; break; + chipset = "Intel(R) 945G"; + break; case PCI_CHIP_I945_GM: - chipset = "Intel(R) 945GM"; break; + chipset = "Intel(R) 945GM"; + break; case PCI_CHIP_I945_GME: - chipset = "Intel(R) 945GME"; break; + chipset = "Intel(R) 945GME"; + break; case PCI_CHIP_G33_G: - chipset = "Intel(R) G33"; break; + chipset = "Intel(R) G33"; + break; case PCI_CHIP_Q35_G: - chipset = "Intel(R) Q35"; break; + chipset = "Intel(R) Q35"; + break; case PCI_CHIP_Q33_G: - chipset = "Intel(R) Q33"; break; + chipset = "Intel(R) Q33"; + break; default: - chipset = "Unknown Intel Chipset"; break; + chipset = "Unknown Intel Chipset"; + break; } - (void) driGetRendererString( buffer, chipset, DRIVER_DATE, 0 ); + (void) driGetRendererString(buffer, chipset, DRIVER_DATE, 0); return (GLubyte *) buffer; default: @@ -146,51 +159,56 @@ const GLubyte *intelGetString( GLcontext *ctx, GLenum name ) /** * Extension strings exported by the intel driver. - * - * \note - * It appears that ARB_texture_env_crossbar has "disappeared" compared to the - * old i830-specific driver. */ -const struct dri_extension card_extensions[] = -{ - { "GL_ARB_multisample", GL_ARB_multisample_functions }, - { "GL_ARB_multitexture", NULL }, - { "GL_ARB_point_parameters", GL_ARB_point_parameters_functions }, - { "GL_ARB_texture_border_clamp", NULL }, - { "GL_ARB_texture_compression", GL_ARB_texture_compression_functions }, - { "GL_ARB_texture_cube_map", NULL }, - { "GL_ARB_texture_env_add", NULL }, - { "GL_ARB_texture_env_combine", NULL }, - { "GL_ARB_texture_env_dot3", NULL }, - { "GL_ARB_texture_mirrored_repeat", NULL }, - { "GL_ARB_texture_rectangle", NULL }, - { "GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions }, - { "GL_ARB_vertex_program", GL_ARB_vertex_program_functions }, - { "GL_ARB_window_pos", GL_ARB_window_pos_functions }, - { "GL_EXT_blend_color", GL_EXT_blend_color_functions }, - { "GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions }, - { "GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions }, - { "GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions }, - { "GL_EXT_blend_subtract", NULL }, - { "GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions }, - { "GL_EXT_fog_coord", GL_EXT_fog_coord_functions }, - { "GL_EXT_multi_draw_arrays", GL_EXT_multi_draw_arrays_functions }, - { "GL_EXT_secondary_color", GL_EXT_secondary_color_functions }, - { "GL_EXT_stencil_wrap", NULL }, - { "GL_EXT_texture_edge_clamp", NULL }, - { "GL_EXT_texture_env_combine", NULL }, - { "GL_EXT_texture_env_dot3", NULL }, - { "GL_EXT_texture_filter_anisotropic", NULL }, - { "GL_EXT_texture_lod_bias", NULL }, - { "GL_3DFX_texture_compression_FXT1", NULL }, - { "GL_APPLE_client_storage", NULL }, - { "GL_MESA_pack_invert", NULL }, - { "GL_MESA_ycbcr_texture", NULL }, - { "GL_NV_blend_square", NULL }, - { "GL_NV_vertex_program", GL_NV_vertex_program_functions }, - { "GL_NV_vertex_program1_1", NULL }, - { "GL_SGIS_generate_mipmap", NULL }, - { NULL, NULL } +const struct dri_extension card_extensions[] = { + {"GL_ARB_multisample", GL_ARB_multisample_functions}, + {"GL_ARB_multitexture", NULL}, + {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions}, + {"GL_ARB_texture_border_clamp", NULL}, + {"GL_ARB_texture_compression", GL_ARB_texture_compression_functions}, + {"GL_ARB_texture_cube_map", NULL}, + {"GL_ARB_texture_env_add", NULL}, + {"GL_ARB_texture_env_combine", NULL}, + {"GL_ARB_texture_env_dot3", NULL}, + {"GL_ARB_texture_mirrored_repeat", NULL}, + {"GL_ARB_texture_rectangle", NULL}, + {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions}, + {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions}, + {"GL_ARB_window_pos", GL_ARB_window_pos_functions}, + {"GL_EXT_blend_color", GL_EXT_blend_color_functions}, + {"GL_EXT_blend_equation_separate", + GL_EXT_blend_equation_separate_functions}, + {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions}, + {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions}, + {"GL_EXT_blend_subtract", NULL}, + {"GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions}, + {"GL_EXT_fog_coord", GL_EXT_fog_coord_functions}, + {"GL_EXT_multi_draw_arrays", GL_EXT_multi_draw_arrays_functions}, +#if 1 /* XXX FBO temporary? */ + {"GL_EXT_packed_depth_stencil", NULL}, +#endif + {"GL_EXT_secondary_color", GL_EXT_secondary_color_functions}, + {"GL_EXT_stencil_wrap", NULL}, + {"GL_EXT_texture_edge_clamp", NULL}, + {"GL_EXT_texture_env_combine", NULL}, + {"GL_EXT_texture_env_dot3", NULL}, + {"GL_EXT_texture_filter_anisotropic", NULL}, + {"GL_EXT_texture_lod_bias", NULL}, + {"GL_3DFX_texture_compression_FXT1", NULL}, + {"GL_APPLE_client_storage", NULL}, + {"GL_MESA_pack_invert", NULL}, + {"GL_MESA_ycbcr_texture", NULL}, + {"GL_NV_blend_square", NULL}, + {"GL_NV_vertex_program", GL_NV_vertex_program_functions}, + {"GL_NV_vertex_program1_1", NULL}, +/* { "GL_SGIS_generate_mipmap", NULL }, */ + {NULL, NULL} +}; + +const struct dri_extension ttm_extensions[] = { + {"GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions}, + {"GL_ARB_pixel_buffer_object", NULL}, + {NULL, NULL} }; extern const struct tnl_pipeline_stage _intel_render_stage; @@ -206,82 +224,144 @@ static const struct tnl_pipeline_stage *intel_pipeline[] = { &_tnl_point_attenuation_stage, &_tnl_vertex_program_stage, #if 1 - &_intel_render_stage, /* ADD: unclipped rastersetup-to-dma */ + &_intel_render_stage, /* ADD: unclipped rastersetup-to-dma */ #endif &_tnl_render_stage, 0, }; -static const struct dri_debug_control debug_control[] = -{ - { "fall", DEBUG_FALLBACKS }, - { "tex", DEBUG_TEXTURE }, - { "ioctl", DEBUG_IOCTL }, - { "prim", DEBUG_PRIMS }, - { "vert", DEBUG_VERTS }, - { "state", DEBUG_STATE }, - { "verb", DEBUG_VERBOSE }, - { "dri", DEBUG_DRI }, - { "dma", DEBUG_DMA }, - { "san", DEBUG_SANITY }, - { "sync", DEBUG_SYNC }, - { "sleep", DEBUG_SLEEP }, - { "pix", DEBUG_PIXEL }, - { NULL, 0 } +static const struct dri_debug_control debug_control[] = { + {"tex", DEBUG_TEXTURE}, + {"state", DEBUG_STATE}, + {"ioctl", DEBUG_IOCTL}, + {"blit", DEBUG_BLIT}, + {"mip", DEBUG_MIPTREE}, + {"fall", DEBUG_FALLBACKS}, + {"verb", DEBUG_VERBOSE}, + {"bat", DEBUG_BATCH}, + {"pix", DEBUG_PIXEL}, + {"buf", DEBUG_BUFMGR}, + {"reg", DEBUG_REGION}, + {"fbo", DEBUG_FBO}, + {"lock", DEBUG_LOCK}, + {NULL, 0} }; -static void intelInvalidateState( GLcontext *ctx, GLuint new_state ) +static void +intelInvalidateState(GLcontext * ctx, GLuint new_state) { - _swrast_InvalidateState( ctx, new_state ); - _swsetup_InvalidateState( ctx, new_state ); - _vbo_InvalidateState( ctx, new_state ); - _tnl_InvalidateState( ctx, new_state ); - _tnl_invalidate_vertex_state( ctx, new_state ); - INTEL_CONTEXT(ctx)->NewGLState |= new_state; + _swrast_InvalidateState(ctx, new_state); + _swsetup_InvalidateState(ctx, new_state); + _vbo_InvalidateState(ctx, new_state); + _tnl_InvalidateState(ctx, new_state); + _tnl_invalidate_vertex_state(ctx, new_state); + intel_context(ctx)->NewGLState |= new_state; } -void intelInitDriverFunctions( struct dd_function_table *functions ) +void +intelFlush(GLcontext * ctx) { - _mesa_init_driver_functions( functions ); + struct intel_context *intel = intel_context(ctx); + + if (intel->Fallback) + _swrast_flush(ctx); + + INTEL_FIREVERTICES(intel); + + if (intel->batch->map != intel->batch->ptr) + intel_batchbuffer_flush(intel->batch); + + /* XXX: Need to do an MI_FLUSH here. + */ +} - functions->Clear = intelClear; - functions->Flush = intelglFlush; - functions->Finish = intelFinish; - functions->GetString = intelGetString; - functions->UpdateState = intelInvalidateState; - intelInitTextureFuncs( functions ); - intelInitPixelFuncs( functions ); - intelInitStateFuncs( functions ); +/** + * Check if we need to rotate/warp the front color buffer to the + * rotated screen. We generally need to do this when we get a glFlush + * or glFinish after drawing to the front color buffer. + */ +static void +intelCheckFrontRotate(GLcontext * ctx) +{ + struct intel_context *intel = intel_context(ctx); + if (intel->ctx.DrawBuffer->_ColorDrawBufferMask[0] == + BUFFER_BIT_FRONT_LEFT) { + intelScreenPrivate *screen = intel->intelScreen; + if (screen->current_rotation != 0) { + __DRIdrawablePrivate *dPriv = intel->driDrawable; + intelRotateWindow(intel, dPriv, BUFFER_BIT_FRONT_LEFT); + } + } } -static void intel_emit_invarient_state( GLcontext *ctx ) + +/** + * Called via glFlush. + */ +static void +intelglFlush(GLcontext * ctx) { + intelFlush(ctx); + intelCheckFrontRotate(ctx); } +void +intelFinish(GLcontext * ctx) +{ + struct intel_context *intel = intel_context(ctx); + intelFlush(ctx); + if (intel->batch->last_fence) { + dri_fence_wait(intel->batch->last_fence); + dri_fence_unreference(intel->batch->last_fence); + intel->batch->last_fence = NULL; + } + intelCheckFrontRotate(ctx); +} -GLboolean intelInitContext( intelContextPtr intel, - const __GLcontextModes *mesaVis, - __DRIcontextPrivate *driContextPriv, - void *sharedContextPrivate, - struct dd_function_table *functions ) +void +intelInitDriverFunctions(struct dd_function_table *functions) +{ + _mesa_init_driver_functions(functions); + + functions->Flush = intelglFlush; + functions->Finish = intelFinish; + functions->GetString = intelGetString; + functions->UpdateState = intelInvalidateState; + functions->CopyColorTable = _swrast_CopyColorTable; + functions->CopyColorSubTable = _swrast_CopyColorSubTable; + functions->CopyConvolutionFilter1D = _swrast_CopyConvolutionFilter1D; + functions->CopyConvolutionFilter2D = _swrast_CopyConvolutionFilter2D; + + intelInitTextureFuncs(functions); + intelInitPixelFuncs(functions); + intelInitStateFuncs(functions); + intelInitBufferFuncs(functions); +} + + +GLboolean +intelInitContext(struct intel_context *intel, + const __GLcontextModes * mesaVis, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate, + struct dd_function_table *functions) { GLcontext *ctx = &intel->ctx; GLcontext *shareCtx = (GLcontext *) sharedContextPrivate; __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; - intelScreenPrivate *intelScreen = (intelScreenPrivate *)sPriv->private; + intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; drmI830Sarea *saPriv = (drmI830Sarea *) - (((GLubyte *)sPriv->pSAREA)+intelScreen->sarea_priv_offset); + (((GLubyte *) sPriv->pSAREA) + intelScreen->sarea_priv_offset); int fthrottle_mode; if (!_mesa_initialize_context(&intel->ctx, - mesaVis, shareCtx, - functions, - (void*) intel)) + mesaVis, shareCtx, + functions, (void *) intel)) return GL_FALSE; driContextPriv->driverPrivate = intel; @@ -289,15 +369,27 @@ GLboolean intelInitContext( intelContextPtr intel, intel->driScreen = sPriv; intel->sarea = saPriv; + intel->width = intelScreen->width; + intel->height = intelScreen->height; + intel->current_rotation = intelScreen->current_rotation; - (void) memset( intel->texture_heaps, 0, sizeof( intel->texture_heaps ) ); - make_empty_list( & intel->swapped ); + if (!lockMutexInit) { + lockMutexInit = GL_TRUE; + _glthread_INIT_MUTEX(lockMutex); + } - driParseConfigFiles (&intel->optionCache, &intelScreen->optionCache, - intel->driScreen->myNum, "i915"); + driParseConfigFiles(&intel->optionCache, &intelScreen->optionCache, + intel->driScreen->myNum, "i915"); ctx->Const.MaxTextureMaxAnisotropy = 2.0; + /* This doesn't yet catch all non-conformant rendering, but it's a + * start. + */ + if (getenv("INTEL_STRICT_CONFORMANCE")) { + intel->strict_conformance = 1; + } + ctx->Const.MinLineWidth = 1.0; ctx->Const.MinLineWidthAA = 1.0; ctx->Const.MaxLineWidth = 3.0; @@ -315,362 +407,256 @@ GLboolean intelInitContext( intelContextPtr intel, */ _mesa_init_point(ctx); + ctx->Const.MaxColorAttachments = 4; /* XXX FBO: review this */ + /* Initialize the software rasterizer and helper modules. */ - _swrast_CreateContext( ctx ); - _vbo_CreateContext( ctx ); - _tnl_CreateContext( ctx ); - _swsetup_CreateContext( ctx ); + _swrast_CreateContext(ctx); + _vbo_CreateContext(ctx); + _tnl_CreateContext(ctx); + _swsetup_CreateContext(ctx); /* Install the customized pipeline: */ - _tnl_destroy_pipeline( ctx ); - _tnl_install_pipeline( ctx, intel_pipeline ); + _tnl_destroy_pipeline(ctx); + _tnl_install_pipeline(ctx, intel_pipeline); /* Configure swrast to match hardware characteristics: */ - _swrast_allow_pixel_fog( ctx, GL_FALSE ); - _swrast_allow_vertex_fog( ctx, GL_TRUE ); + _swrast_allow_pixel_fog(ctx, GL_FALSE); + _swrast_allow_vertex_fog(ctx, GL_TRUE); /* Dri stuff */ intel->hHWContext = driContextPriv->hHWContext; intel->driFd = sPriv->fd; - intel->driHwLock = (drmLock *) &sPriv->pSAREA->lock; + intel->driHwLock = (drmLock *) & sPriv->pSAREA->lock; - intel->hw_stencil = mesaVis->stencilBits && mesaVis->depthBits == 24; intel->hw_stipple = 1; - switch(mesaVis->depthBits) { - case 0: /* what to do in this case? */ + /* XXX FBO: this doesn't seem to be used anywhere */ + switch (mesaVis->depthBits) { + case 0: /* what to do in this case? */ case 16: - intel->depth_scale = 1.0/0xffff; - intel->polygon_offset_scale = 1.0/0xffff; - intel->depth_clear_mask = ~0; - intel->ClearDepth = 0xffff; + intel->polygon_offset_scale = 1.0 / 0xffff; break; case 24: - intel->depth_scale = 1.0/0xffffff; - intel->polygon_offset_scale = 2.0/0xffffff; /* req'd to pass glean */ - intel->depth_clear_mask = 0x00ffffff; - intel->stencil_clear_mask = 0xff000000; - intel->ClearDepth = 0x00ffffff; + intel->polygon_offset_scale = 2.0 / 0xffffff; /* req'd to pass glean */ break; default: - assert(0); + assert(0); break; } /* Initialize swrast, tnl driver tables: */ - intelInitSpanFuncs( ctx ); - intelInitTriFuncs( ctx ); + intelInitSpanFuncs(ctx); + intelInitTriFuncs(ctx); intel->RenderIndex = ~0; fthrottle_mode = driQueryOptioni(&intel->optionCache, "fthrottle_mode"); - intel->iw.irq_seq = -1; intel->irqsEmitted = 0; intel->do_irqs = (intel->intelScreen->irq_active && - fthrottle_mode == DRI_CONF_FTHROTTLE_IRQS); + fthrottle_mode == DRI_CONF_FTHROTTLE_IRQS); intel->do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS); - intel->vblank_flags = (intel->intelScreen->irq_active != 0) - ? driGetDefaultVBlankFlags(&intel->optionCache) : VBLANK_FLAG_NO_IRQ; + _math_matrix_ctr(&intel->ViewportMatrix); + + /* Disable imaging extension until convolution is working in + * teximage paths: + */ + driInitExtensions(ctx, card_extensions, +/* GL_TRUE, */ + GL_FALSE); + + if (intelScreen->ttm) + driInitExtensions(ctx, ttm_extensions, GL_FALSE); - (*dri_interface->getUST)(&intel->swap_ust); - _math_matrix_ctr (&intel->ViewportMatrix); - driInitExtensions( ctx, card_extensions, GL_TRUE ); + intel->batch = intel_batchbuffer_alloc(intel); + intel->last_swap_fence = NULL; + intel->first_swap_fence = NULL; + + intel_bufferobj_init(intel); + intel_fbo_init(intel); if (intel->ctx.Mesa_DXTn) { - _mesa_enable_extension( ctx, "GL_EXT_texture_compression_s3tc" ); - _mesa_enable_extension( ctx, "GL_S3_s3tc" ); + _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc"); + _mesa_enable_extension(ctx, "GL_S3_s3tc"); } - else if (driQueryOptionb (&intel->optionCache, "force_s3tc_enable")) { - _mesa_enable_extension( ctx, "GL_EXT_texture_compression_s3tc" ); + else if (driQueryOptionb(&intel->optionCache, "force_s3tc_enable")) { + _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc"); } -/* driInitTextureObjects( ctx, & intel->swapped, */ -/* DRI_TEXMGR_DO_TEXTURE_1D | */ -/* DRI_TEXMGR_DO_TEXTURE_2D | */ -/* DRI_TEXMGR_DO_TEXTURE_RECT ); */ - - - intelInitBatchBuffer(&intel->ctx); - intel->prim.flush = intel_emit_invarient_state; intel->prim.primitive = ~0; #if DO_DEBUG - INTEL_DEBUG = driParseDebugString( getenv( "INTEL_DEBUG" ), - debug_control ); - INTEL_DEBUG |= driParseDebugString( getenv( "INTEL_DEBUG" ), - debug_control ); -#endif - -#ifndef VERBOSE - if (getenv("INTEL_VERBOSE")) - VERBOSE=1; + INTEL_DEBUG = driParseDebugString(getenv("INTEL_DEBUG"), debug_control); #endif - if (getenv("INTEL_NO_RAST") || - getenv("INTEL_NO_RAST")) { + if (getenv("INTEL_NO_RAST")) { fprintf(stderr, "disabling 3D rasterization\n"); - FALLBACK(intel, INTEL_FALLBACK_USER, 1); + FALLBACK(intel, INTEL_FALLBACK_USER, 1); } return GL_TRUE; } -void intelDestroyContext(__DRIcontextPrivate *driContextPriv) +void +intelDestroyContext(__DRIcontextPrivate * driContextPriv) { - intelContextPtr intel = (intelContextPtr) driContextPriv->driverPrivate; + struct intel_context *intel = + (struct intel_context *) driContextPriv->driverPrivate; - assert(intel); /* should never be null */ + assert(intel); /* should never be null */ if (intel) { - GLboolean release_texture_heaps; + GLboolean release_texture_heaps; - INTEL_FIREVERTICES( intel ); + INTEL_FIREVERTICES(intel); - intel->vtbl.destroy( intel ); + intel->vtbl.destroy(intel); release_texture_heaps = (intel->ctx.Shared->RefCount == 1); - _swsetup_DestroyContext (&intel->ctx); - _tnl_DestroyContext (&intel->ctx); - _vbo_DestroyContext (&intel->ctx); + _swsetup_DestroyContext(&intel->ctx); + _tnl_DestroyContext(&intel->ctx); + _vbo_DestroyContext(&intel->ctx); - _swrast_DestroyContext (&intel->ctx); - intel->Fallback = 0; /* don't call _swrast_Flush later */ + _swrast_DestroyContext(&intel->ctx); + intel->Fallback = 0; /* don't call _swrast_Flush later */ - intelDestroyBatchBuffer(&intel->ctx); - + intel_batchbuffer_free(intel->batch); - if ( release_texture_heaps ) { + if (intel->last_swap_fence) { + dri_fence_wait(intel->last_swap_fence); + dri_fence_unreference(intel->last_swap_fence); + intel->last_swap_fence = NULL; + } + if (intel->first_swap_fence) { + dri_fence_wait(intel->first_swap_fence); + dri_fence_unreference(intel->first_swap_fence); + intel->first_swap_fence = NULL; + } + + + if (release_texture_heaps) { /* This share group is about to go away, free our private * texture object data. */ - int i; - - for ( i = 0 ; i < intel->nr_heaps ; i++ ) { - driDestroyTextureHeap( intel->texture_heaps[ i ] ); - intel->texture_heaps[ i ] = NULL; - } - - assert( is_empty_list( & intel->swapped ) ); + if (INTEL_DEBUG & DEBUG_TEXTURE) + fprintf(stderr, "do something to free texture heaps\n"); } /* free the Mesa context */ - _mesa_destroy_context(&intel->ctx); + _mesa_free_context_data(&intel->ctx); } } -void intelSetFrontClipRects( intelContextPtr intel ) +GLboolean +intelUnbindContext(__DRIcontextPrivate * driContextPriv) { - __DRIdrawablePrivate *dPriv = intel->driDrawable; - - if (!dPriv) return; - - intel->numClipRects = dPriv->numClipRects; - intel->pClipRects = dPriv->pClipRects; - intel->drawX = dPriv->x; - intel->drawY = dPriv->y; + return GL_TRUE; } - -void intelSetBackClipRects( intelContextPtr intel ) +GLboolean +intelMakeCurrent(__DRIcontextPrivate * driContextPriv, + __DRIdrawablePrivate * driDrawPriv, + __DRIdrawablePrivate * driReadPriv) { - __DRIdrawablePrivate *dPriv = intel->driDrawable; - if (!dPriv) return; - - if (intel->sarea->pf_enabled == 0 && dPriv->numBackClipRects == 0) { - intel->numClipRects = dPriv->numClipRects; - intel->pClipRects = dPriv->pClipRects; - intel->drawX = dPriv->x; - intel->drawY = dPriv->y; - } else { - intel->numClipRects = dPriv->numBackClipRects; - intel->pClipRects = dPriv->pBackClipRects; - intel->drawX = dPriv->backX; - intel->drawY = dPriv->backY; - - if (dPriv->numBackClipRects == 1 && - dPriv->x == dPriv->backX && - dPriv->y == dPriv->backY) { - - /* Repeat the calculation of the back cliprect dimensions here - * as early versions of dri.a in the Xserver are incorrect. Try - * very hard not to restrict future versions of dri.a which - * might eg. allocate truly private back buffers. - */ - int x1, y1; - int x2, y2; - - x1 = dPriv->x; - y1 = dPriv->y; - x2 = dPriv->x + dPriv->w; - y2 = dPriv->y + dPriv->h; - - if (x1 < 0) x1 = 0; - if (y1 < 0) y1 = 0; - if (x2 > intel->intelScreen->width) x2 = intel->intelScreen->width; - if (y2 > intel->intelScreen->height) y2 = intel->intelScreen->height; - - if (x1 == dPriv->pBackClipRects[0].x1 && - y1 == dPriv->pBackClipRects[0].y1) { - - dPriv->pBackClipRects[0].x2 = x2; - dPriv->pBackClipRects[0].y2 = y2; - } + if (driContextPriv) { + struct intel_context *intel = + (struct intel_context *) driContextPriv->driverPrivate; + struct intel_framebuffer *intel_fb = + (struct intel_framebuffer *) driDrawPriv->driverPrivate; + GLframebuffer *readFb = (GLframebuffer *) driReadPriv->driverPrivate; + + + /* XXX FBO temporary fix-ups! */ + /* if the renderbuffers don't have regions, init them from the context */ + { + struct intel_renderbuffer *irbDepth + = intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH); + struct intel_renderbuffer *irbStencil + = intel_get_renderbuffer(&intel_fb->Base, BUFFER_STENCIL); + + if (intel_fb->color_rb[0] && !intel_fb->color_rb[0]->region) { + intel_region_reference(&intel_fb->color_rb[0]->region, + intel->intelScreen->front_region); + } + if (intel_fb->color_rb[1] && !intel_fb->color_rb[1]->region) { + intel_region_reference(&intel_fb->color_rb[1]->region, + intel->intelScreen->back_region); + } + if (intel_fb->color_rb[2] && !intel_fb->color_rb[2]->region) { + intel_region_reference(&intel_fb->color_rb[2]->region, + intel->intelScreen->third_region); + } + if (irbDepth && !irbDepth->region) { + intel_region_reference(&irbDepth->region, intel->intelScreen->depth_region); + } + if (irbStencil && !irbStencil->region) { + intel_region_reference(&irbStencil->region, intel->intelScreen->depth_region); + } } - } -} + /* set GLframebuffer size to match window, if needed */ + driUpdateFramebufferSize(&intel->ctx, driDrawPriv); -void intelWindowMoved( intelContextPtr intel ) -{ - __DRIdrawablePrivate *dPriv = intel->driDrawable; - GLframebuffer *drawFb = (GLframebuffer *) dPriv->driverPrivate; - - if (!intel->ctx.DrawBuffer) { - intelSetFrontClipRects( intel ); - } - else { - driUpdateFramebufferSize(&intel->ctx, dPriv); - switch (drawFb->_ColorDrawBufferMask[0]) { - case BUFFER_BIT_FRONT_LEFT: - intelSetFrontClipRects( intel ); - break; - case BUFFER_BIT_BACK_LEFT: - intelSetBackClipRects( intel ); - break; - default: - /* glDrawBuffer(GL_NONE or GL_FRONT_AND_BACK): software fallback */ - intelSetFrontClipRects( intel ); + if (driReadPriv != driDrawPriv) { + driUpdateFramebufferSize(&intel->ctx, driReadPriv); } - } - if (drawFb->Width != dPriv->w || drawFb->Height != dPriv->h) { - /* update Mesa's notion of framebuffer/window size */ - _mesa_resize_framebuffer(&intel->ctx, drawFb, dPriv->w, dPriv->h); - drawFb->Initialized = GL_TRUE; /* XXX remove someday */ - } + _mesa_make_current(&intel->ctx, &intel_fb->Base, readFb); - /* Set state we know depends on drawable parameters: - */ - { - GLcontext *ctx = &intel->ctx; - - if (intel->intelScreen->driScrnPriv->ddxMinor >= 7) { - drmI830Sarea *sarea = intel->sarea; - drm_clip_rect_t drw_rect = { .x1 = dPriv->x, .x2 = dPriv->x + dPriv->w, - .y1 = dPriv->y, .y2 = dPriv->y + dPriv->h }; - drm_clip_rect_t pipeA_rect = { .x1 = sarea->pipeA_x, - .x2 = sarea->pipeA_x + sarea->pipeA_w, - .y1 = sarea->pipeA_y, - .y2 = sarea->pipeA_y + sarea->pipeA_h }; - drm_clip_rect_t pipeB_rect = { .x1 = sarea->pipeB_x, - .x2 = sarea->pipeB_x + sarea->pipeB_w, - .y1 = sarea->pipeB_y, - .y2 = sarea->pipeB_y + sarea->pipeB_h }; - GLint areaA = driIntersectArea( drw_rect, pipeA_rect ); - GLint areaB = driIntersectArea( drw_rect, pipeB_rect ); - GLuint flags = intel->vblank_flags; - - if (areaB > areaA || (areaA == areaB && areaB > 0)) { - flags = intel->vblank_flags | VBLANK_FLAG_SECONDARY; - } else { - flags = intel->vblank_flags & ~VBLANK_FLAG_SECONDARY; + /* The drawbuffer won't always be updated by _mesa_make_current: + */ + if (intel->ctx.DrawBuffer == &intel_fb->Base) { + + if (intel->driDrawable != driDrawPriv) { + if (driDrawPriv->swap_interval == (unsigned)-1) { + int i; + + intel_fb->vblank_flags = (intel->intelScreen->irq_active != 0) + ? driGetDefaultVBlankFlags(&intel->optionCache) + : VBLANK_FLAG_NO_IRQ; + + (*dri_interface->getUST) (&intel_fb->swap_ust); + driDrawableInitVBlank(driDrawPriv, intel_fb->vblank_flags, + &intel_fb->vbl_seq); + intel_fb->vbl_waited = intel_fb->vbl_seq; + + for (i = 0; i < (intel->intelScreen->third.handle ? 3 : 2); i++) { + if (intel_fb->color_rb[i]) + intel_fb->color_rb[i]->vbl_pending = intel_fb->vbl_seq; + } + } + intel->driDrawable = driDrawPriv; + intelWindowMoved(intel); } - if (flags != intel->vblank_flags) { - intel->vblank_flags = flags; - driGetCurrentVBlank(dPriv, intel->vblank_flags, &intel->vbl_seq); - } - } else { - intel->vblank_flags &= ~VBLANK_FLAG_SECONDARY; + intel_draw_buffer(&intel->ctx, &intel_fb->Base); } - - ctx->Driver.Scissor( ctx, ctx->Scissor.X, ctx->Scissor.Y, - ctx->Scissor.Width, ctx->Scissor.Height ); - - ctx->Driver.DepthRange( ctx, - ctx->Viewport.Near, - ctx->Viewport.Far ); } -} - -GLboolean intelUnbindContext(__DRIcontextPrivate *driContextPriv) -{ - return GL_TRUE; -} - -GLboolean intelMakeCurrent(__DRIcontextPrivate *driContextPriv, - __DRIdrawablePrivate *driDrawPriv, - __DRIdrawablePrivate *driReadPriv) -{ - - if (driContextPriv) { - intelContextPtr intel = (intelContextPtr) driContextPriv->driverPrivate; - - if ( intel->driDrawable != driDrawPriv ) { - /* Shouldn't the readbuffer be stored also? */ - driDrawableInitVBlank( driDrawPriv, intel->vblank_flags, - &intel->vbl_seq ); - - intel->driDrawable = driDrawPriv; - intelWindowMoved( intel ); - } - - _mesa_make_current(&intel->ctx, - (GLframebuffer *) driDrawPriv->driverPrivate, - (GLframebuffer *) driReadPriv->driverPrivate); - - intel->ctx.Driver.DrawBuffer( &intel->ctx, intel->ctx.Color.DrawBuffer[0] ); - } else { + else { _mesa_make_current(NULL, NULL, NULL); } return GL_TRUE; } -/** - * Use the information in the sarea to update the screen parameters - * related to screen rotation. - */ static void -intelUpdateScreenRotation(intelContextPtr intel, - __DRIscreenPrivate *sPriv, - drmI830Sarea *sarea) -{ - intelScreenPrivate *intelScreen = (intelScreenPrivate *)sPriv->private; - intelRegion *colorBuf; - - intelUnmapScreenRegions(intelScreen); - - intelUpdateScreenFromSAREA(intelScreen, sarea); - - /* update the current hw offsets for the color and depth buffers */ - if (intel->ctx.DrawBuffer->_ColorDrawBufferMask[0] == BUFFER_BIT_BACK_LEFT) - colorBuf = &intelScreen->back; - else - colorBuf = &intelScreen->front; - intel->vtbl.update_color_z_regions(intel, colorBuf, &intelScreen->depth); - - if (!intelMapScreenRegions(sPriv)) { - fprintf(stderr, "ERROR Remapping screen regions!!!\n"); - } -} - -void intelGetLock( intelContextPtr intel, GLuint flags ) +intelContendedLock(struct intel_context *intel, GLuint flags) { __DRIdrawablePrivate *dPriv = intel->driDrawable; __DRIscreenPrivate *sPriv = intel->driScreen; - intelScreenPrivate *intelScreen = (intelScreenPrivate *)sPriv->private; - drmI830Sarea * sarea = intel->sarea; - unsigned i; + intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; + drmI830Sarea *sarea = intel->sarea; drmGetLock(intel->driFd, intel->hHWContext, flags); + if (INTEL_DEBUG & DEBUG_LOCK) + _mesa_printf("%s - got contended lock\n", __progname); + /* If the window moved, may need to set a new cliprect now. * * NOTE: This releases and regains the hw lock, so all state @@ -679,98 +665,129 @@ void intelGetLock( intelContextPtr intel, GLuint flags ) if (dPriv) DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); - if (dPriv && intel->lastStamp != dPriv->lastStamp) { - intelWindowMoved( intel ); - intel->lastStamp = dPriv->lastStamp; - } - - /* If we lost context, need to dump all registers to hardware. - * Note that we don't care about 2d contexts, even if they perform - * accelerated commands, so the DRI locking in the X server is even - * more broken than usual. + /* If the last consumer of the texture memory wasn't us, notify the fake + * bufmgr and record the new owner. We should have the memory shared + * between contexts of a single fake bufmgr, but this will at least make + * things correct for now. */ + if (!intel->intelScreen->ttm && sarea->texAge != intel->hHWContext) { + sarea->texAge = intel->hHWContext; + dri_bufmgr_fake_contended_lock_take(intel->intelScreen->bufmgr); + if (INTEL_DEBUG & DEBUG_BATCH) + intel_decode_context_reset(); + } if (sarea->width != intelScreen->width || sarea->height != intelScreen->height || sarea->rotation != intelScreen->current_rotation) { - intelUpdateScreenRotation(intel, sPriv, sarea); - /* This will drop the outstanding batchbuffer on the floor */ - intel->batch.ptr -= (intel->batch.size - intel->batch.space); - intel->batch.space = intel->batch.size; - /* lose all primitives */ - intel->prim.primitive = ~0; - intel->prim.start_ptr = 0; - intel->prim.flush = 0; - intel->vtbl.lost_hardware( intel ); + intelUpdateScreenRotation(sPriv, sarea); + } - intel->lastStamp = 0; /* force window update */ + if (sarea->width != intel->width || + sarea->height != intel->height || + sarea->rotation != intel->current_rotation) { + int numClipRects = intel->numClipRects; - /* Release batch buffer + /* + * FIXME: Really only need to do this when drawing to a + * common back- or front buffer. */ - intelDestroyBatchBuffer(&intel->ctx); - intelInitBatchBuffer(&intel->ctx); - intel->prim.flush = intel_emit_invarient_state; - /* Still need to reset the global LRU? + /* + * This will essentially drop the outstanding batchbuffer on the floor. */ - intel_driReinitTextureHeap( intel->texture_heaps[0], intel->intelScreen->tex.size ); + intel->numClipRects = 0; + + if (intel->Fallback) + _swrast_flush(&intel->ctx); + + INTEL_FIREVERTICES(intel); + + if (intel->batch->map != intel->batch->ptr) + intel_batchbuffer_flush(intel->batch); + + intel->numClipRects = numClipRects; + + /* force window update */ + intel->lastStamp = 0; + + intel->width = sarea->width; + intel->height = sarea->height; + intel->current_rotation = sarea->rotation; } - /* Shared texture managment - if another client has played with - * texture space, figure out which if any of our textures have been - * ejected, and update our global LRU. + /* Drawable changed? */ - for ( i = 0 ; i < intel->nr_heaps ; i++ ) { - DRI_AGE_TEXTURES( intel->texture_heaps[ i ] ); + if (dPriv && intel->lastStamp != dPriv->lastStamp) { + intelWindowMoved(intel); + intel->lastStamp = dPriv->lastStamp; } } -void intelSwapBuffers( __DRIdrawablePrivate *dPriv ) + +/* Lock the hardware and validate our state. + */ +void LOCK_HARDWARE( struct intel_context *intel ) { - if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { - intelContextPtr intel; - GLcontext *ctx; - intel = (intelContextPtr) dPriv->driContextPriv->driverPrivate; - ctx = &intel->ctx; - if (ctx->Visual.doubleBufferMode) { - intelScreenPrivate *screen = intel->intelScreen; - _mesa_notifySwapBuffers( ctx ); /* flush pending rendering comands */ - if ( 0 /*intel->doPageFlip*/ ) { /* doPageFlip is never set !!! */ - intelPageFlip( dPriv ); - } else { - intelCopyBuffer( dPriv, NULL ); - } - if (screen->current_rotation != 0) { - intelRotateWindow(intel, dPriv, BUFFER_BIT_FRONT_LEFT); - } - } - } else { - /* XXX this shouldn't be an error but we can't handle it for now */ - fprintf(stderr, "%s: drawable has no context!\n", __FUNCTION__); - } + char __ret=0; + struct intel_framebuffer *intel_fb = NULL; + struct intel_renderbuffer *intel_rb = NULL; + _glthread_LOCK_MUTEX(lockMutex); + assert(!intel->locked); + + if (intel->driDrawable) { + intel_fb = intel->driDrawable->driverPrivate; + + if (intel_fb) + intel_rb = + intel_get_renderbuffer(&intel_fb->Base, + intel_fb->Base._ColorDrawBufferMask[0] == + BUFFER_BIT_FRONT_LEFT ? BUFFER_FRONT_LEFT : + BUFFER_BACK_LEFT); + } + + if (intel_rb && intel_fb->vblank_flags && + !(intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ) && + (intel_fb->vbl_waited - intel_rb->vbl_pending) > (1<<23)) { + drmVBlank vbl; + + vbl.request.type = DRM_VBLANK_ABSOLUTE; + + if ( intel_fb->vblank_flags & VBLANK_FLAG_SECONDARY ) { + vbl.request.type |= DRM_VBLANK_SECONDARY; + } + + vbl.request.sequence = intel_rb->vbl_pending; + drmWaitVBlank(intel->driFd, &vbl); + intel_fb->vbl_waited = vbl.reply.sequence; + } + + DRM_CAS(intel->driHwLock, intel->hHWContext, + (DRM_LOCK_HELD|intel->hHWContext), __ret); + + if (__ret) + intelContendedLock( intel, 0 ); + + if (INTEL_DEBUG & DEBUG_LOCK) + _mesa_printf("%s - locked\n", __progname); + + intel->locked = 1; } -void intelCopySubBuffer( __DRIdrawablePrivate *dPriv, - int x, int y, int w, int h ) + + /* Unlock the hardware using the global current context + */ +void UNLOCK_HARDWARE( struct intel_context *intel ) { - if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { - intelContextPtr intel; - GLcontext *ctx; - intel = (intelContextPtr) dPriv->driContextPriv->driverPrivate; - ctx = &intel->ctx; - if (ctx->Visual.doubleBufferMode) { - drm_clip_rect_t rect; - rect.x1 = x + dPriv->x; - rect.y1 = (dPriv->h - y - h) + dPriv->y; - rect.x2 = rect.x1 + w; - rect.y2 = rect.y1 + h; - _mesa_notifySwapBuffers( ctx ); /* flush pending rendering comands */ - intelCopyBuffer( dPriv, &rect ); - } - } else { - /* XXX this shouldn't be an error but we can't handle it for now */ - fprintf(stderr, "%s: drawable has no context!\n", __FUNCTION__); - } -} + intel->locked = 0; + + DRM_UNLOCK(intel->driFd, intel->driHwLock, intel->hHWContext); + + _glthread_UNLOCK_MUTEX(lockMutex); + + if (INTEL_DEBUG & DEBUG_LOCK) + _mesa_printf("%s - unlocked\n", __progname); +} + diff --git a/src/mesa/drivers/dri/i915/intel_context.h b/src/mesa/drivers/dri/i915/intel_context.h index 3b50107d73..ce9a362944 100644 --- a/src/mesa/drivers/dri/i915/intel_context.h +++ b/src/mesa/drivers/dri/i915/intel_context.h @@ -34,7 +34,6 @@ #include "drm.h" #include "mm.h" #include "texmem.h" -#include "vblank.h" #include "intel_screen.h" #include "i915_drm.h" @@ -49,156 +48,177 @@ #define DV_PF_565 (2<<8) #define DV_PF_8888 (3<<8) -#define INTEL_CONTEXT(ctx) ((intelContextPtr)(ctx)) +struct intel_region; +struct intel_context; -typedef struct intel_context intelContext; -typedef struct intel_context *intelContextPtr; -typedef struct intel_texture_object *intelTextureObjectPtr; - -typedef void (*intel_tri_func)(intelContextPtr, intelVertex *, intelVertex *, - intelVertex *); -typedef void (*intel_line_func)(intelContextPtr, intelVertex *, intelVertex *); -typedef void (*intel_point_func)(intelContextPtr, intelVertex *); +typedef void (*intel_tri_func) (struct intel_context *, intelVertex *, + intelVertex *, intelVertex *); +typedef void (*intel_line_func) (struct intel_context *, intelVertex *, + intelVertex *); +typedef void (*intel_point_func) (struct intel_context *, intelVertex *); #define INTEL_FALLBACK_DRAW_BUFFER 0x1 #define INTEL_FALLBACK_READ_BUFFER 0x2 -#define INTEL_FALLBACK_USER 0x4 -#define INTEL_FALLBACK_NO_BATCHBUFFER 0x8 -#define INTEL_FALLBACK_NO_TEXMEM 0x10 +#define INTEL_FALLBACK_DEPTH_BUFFER 0x4 +#define INTEL_FALLBACK_STENCIL_BUFFER 0x8 +#define INTEL_FALLBACK_USER 0x10 #define INTEL_FALLBACK_RENDERMODE 0x20 -extern void intelFallback( intelContextPtr intel, GLuint bit, GLboolean mode ); +extern void intelFallback(struct intel_context *intel, GLuint bit, + GLboolean mode); #define FALLBACK( intel, bit, mode ) intelFallback( intel, bit, mode ) -#define INTEL_TEX_MAXLEVELS 10 - +#define INTEL_WRITE_PART 0x1 +#define INTEL_WRITE_FULL 0x2 +#define INTEL_READ 0x4 struct intel_texture_object { - driTextureObject base; /* the parent class */ - - GLuint texelBytes; - GLuint age; - GLuint Pitch; - GLuint Height; - GLuint TextureOffset; - GLubyte *BufAddr; - - GLuint min_level; - GLuint max_level; - GLuint depth_pitch; - - struct { - const struct gl_texture_image *image; - GLuint offset; /* into BufAddr */ - GLuint height; - GLuint internalFormat; - } image[6][INTEL_TEX_MAXLEVELS]; - - GLuint dirty; - GLuint firstLevel,lastLevel; + struct gl_texture_object base; /* The "parent" object */ + + /* The mipmap tree must include at least these levels once + * validated: + */ + GLuint firstLevel; + GLuint lastLevel; + + /* Offset for firstLevel image: + */ + GLuint textureOffset; + + /* On validation any active images held in main memory or in other + * regions will be copied to this region and the old storage freed. + */ + struct intel_mipmap_tree *mt; + + GLboolean imageOverride; + GLint depthOverride; + GLuint pitchOverride; +}; + + + +struct intel_texture_image +{ + struct gl_texture_image base; + + /* These aren't stored in gl_texture_image + */ + GLuint level; + GLuint face; + + /* If intelImage->mt != NULL, image data is stored here. + * Else if intelImage->base.Data != NULL, image is stored there. + * Else there is no image data. + */ + struct intel_mipmap_tree *mt; }; +#define INTEL_MAX_FIXUP 64 + struct intel_context { - GLcontext ctx; /* the parent class */ + GLcontext ctx; /* the parent class */ + + struct + { + void (*destroy) (struct intel_context * intel); + void (*emit_state) (struct intel_context * intel); + void (*lost_hardware) (struct intel_context * intel); + void (*update_texture_state) (struct intel_context * intel); + + void (*render_start) (struct intel_context * intel); + void (*render_prevalidate) (struct intel_context * intel); + void (*set_draw_region) (struct intel_context * intel, + struct intel_region * draw_region, + struct intel_region * depth_region); + + GLuint(*flush_cmd) (void); + + void (*reduced_primitive_state) (struct intel_context * intel, + GLenum rprim); + + GLboolean(*check_vertex_size) (struct intel_context * intel, + GLuint expected); + + + /* Metaops: + */ + void (*install_meta_state) (struct intel_context * intel); + void (*leave_meta_state) (struct intel_context * intel); + + void (*meta_draw_region) (struct intel_context * intel, + struct intel_region * draw_region, + struct intel_region * depth_region); - struct { - void (*destroy)( intelContextPtr intel ); - void (*emit_state)( intelContextPtr intel ); - void (*lost_hardware)( intelContextPtr intel ); - void (*update_texture_state)( intelContextPtr intel ); + void (*meta_color_mask) (struct intel_context * intel, GLboolean); - void (*render_start)( intelContextPtr intel ); - void (*set_color_region)( intelContextPtr intel, const intelRegion *reg ); - void (*set_z_region)( intelContextPtr intel, const intelRegion *reg ); - void (*update_color_z_regions)(intelContextPtr intel, - const intelRegion *colorRegion, - const intelRegion *depthRegion); - void (*emit_flush)( intelContextPtr intel ); - void (*reduced_primitive_state)( intelContextPtr intel, GLenum rprim ); + void (*meta_stencil_replace) (struct intel_context * intel, + GLuint mask, GLuint clear); - GLboolean (*check_vertex_size)( intelContextPtr intel, GLuint expected ); + void (*meta_depth_replace) (struct intel_context * intel); - void (*clear_with_tris)( intelContextPtr intel, GLbitfield mask, - GLboolean all, - GLint cx, GLint cy, GLint cw, GLint ch); + void (*meta_texture_blend_replace) (struct intel_context * intel); - void (*rotate_window)( intelContextPtr intel, - __DRIdrawablePrivate *dPriv, GLuint srcBuf); + void (*meta_no_stencil_write) (struct intel_context * intel); + void (*meta_no_depth_write) (struct intel_context * intel); + void (*meta_no_texture) (struct intel_context * intel); - intelTextureObjectPtr (*alloc_tex_obj)( struct gl_texture_object *tObj ); + void (*meta_import_pixel_state) (struct intel_context * intel); + + GLboolean(*meta_tex_rect_source) (struct intel_context * intel, + dri_bo * buffer, + GLuint offset, + GLuint pitch, + GLuint height, + GLenum format, GLenum type); + void (*rotate_window) (struct intel_context * intel, + __DRIdrawablePrivate * dPriv, GLuint srcBuf); + + void (*assert_not_dirty) (struct intel_context *intel); } vtbl; - GLint refcount; + GLint refcount; GLuint Fallback; GLuint NewGLState; - - struct { - GLuint start_offset; - GLint size; - GLint space; - GLubyte *ptr; - GLuint counter; - GLuint last_emit_state; - GLboolean contains_geometry; - const char *func; - GLuint last_swap; - } batch; - - struct { - void *ptr; - GLint size; - GLuint offset; - GLuint active_buf; - GLuint irq_emitted; - } alloc; - - struct { + + dri_fence *last_swap_fence; + dri_fence *first_swap_fence; + + struct intel_batchbuffer *batch; + + struct + { + GLuint id; GLuint primitive; - GLubyte *start_ptr; - void (*flush)( GLcontext * ); + GLubyte *start_ptr; + void (*flush) (struct intel_context *); } prim; GLboolean locked; + char *prevLockFile; + int prevLockLine; - GLubyte clear_red; - GLubyte clear_green; - GLubyte clear_blue; - GLubyte clear_alpha; - GLuint ClearColor; - GLuint ClearDepth; + GLuint ClearColor565; + GLuint ClearColor8888; + /* Offsets of fields within the current vertex: + */ GLuint coloroffset; GLuint specoffset; - - /* Support for duplicating XYZW as WPOS parameter (crutch for I915). - */ GLuint wpos_offset; GLuint wpos_size; struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX]; GLuint vertex_attr_count; - GLfloat depth_scale; - GLfloat polygon_offset_scale; /* dependent on depth_scale, bpp */ - GLuint depth_clear_mask; - GLuint stencil_clear_mask; + GLfloat polygon_offset_scale; /* dependent on depth_scale, bpp */ - GLboolean hw_stencil; GLboolean hw_stipple; - - /* Texture object bookkeeping - */ - GLuint nr_heaps; - driTexHeap * texture_heaps[1]; - driTextureObject swapped; - GLuint lastStamp; - - struct intel_texture_object *CurrentTexObj[MAX_TEXTURE_UNITS]; + GLboolean strict_conformance; /* State for intelvb.c and inteltris.c. */ @@ -207,8 +227,15 @@ struct intel_context GLenum render_primitive; GLenum reduced_primitive; GLuint vertex_size; - unsigned char *verts; /* points to tnl->clipspace.vertex_buf */ - + GLubyte *verts; /* points to tnl->clipspace.vertex_buf */ + +#if 0 + struct intel_region *front_region; /* XXX FBO: obsolete */ + struct intel_region *rotated_region; /* XXX FBO: obsolete */ + struct intel_region *back_region; /* XXX FBO: obsolete */ + struct intel_region *draw_region; /* XXX FBO: rename to color_region */ + struct intel_region *depth_region; /**< currently bound depth/Z region */ +#endif /* Fallback rasterization functions */ @@ -216,27 +243,18 @@ struct intel_context intel_line_func draw_line; intel_tri_func draw_tri; - /* Drawing buffer state + /* These refer to the current drawing buffer: */ - intelRegion *drawRegion; /* current drawing buffer */ - intelRegion *readRegion; /* current reading buffer */ - - int drawX; /* origin of drawable in draw buffer */ - int drawY; - GLuint numClipRects; /* cliprects for that buffer */ + int drawX, drawY; /**< origin of drawing area within region */ + GLuint numClipRects; /**< cliprects for drawing */ drm_clip_rect_t *pClipRects; + drm_clip_rect_t fboRect; /**< cliprect for FBO rendering */ - int dirtyAge; int perf_boxes; GLuint do_usleeps; int do_irqs; GLuint irqsEmitted; - drm_i915_irq_wait_t iw; - - GLboolean scissor; - drm_clip_rect_t draw_rect; - drm_clip_rect_t scissor_rect; drm_context_t hHWContext; drmLock *driHwLock; @@ -244,127 +262,42 @@ struct intel_context __DRIdrawablePrivate *driDrawable; __DRIscreenPrivate *driScreen; - intelScreenPrivate *intelScreen; - drmI830Sarea *sarea; + intelScreenPrivate *intelScreen; + drmI830Sarea *sarea; + + GLuint lastStamp; /** * Configuration cache */ driOptionCache optionCache; - /* VBI - */ - GLuint vbl_seq; - GLuint vblank_flags; - - int64_t swap_ust; - int64_t swap_missed_ust; - - GLuint swap_count; - GLuint swap_missed_count; + /* Rotation. Need to match that of the + * current screen. + */ - GLuint swap_scheduled; + int width; + int height; + int current_rotation; }; - -#define DEBUG_LOCKING 1 - -#if DEBUG_LOCKING -extern char *prevLockFile; -extern int prevLockLine; - -#define DEBUG_LOCK() \ - do { \ - prevLockFile = (__FILE__); \ - prevLockLine = (__LINE__); \ - } while (0) - -#define DEBUG_RESET() \ - do { \ - prevLockFile = 0; \ - prevLockLine = 0; \ - } while (0) - -/* Slightly less broken way of detecting recursive locking in a - * threaded environment. The right way to do this would be to make - * prevLockFile, prevLockLine thread-local. - * - * This technique instead checks to see if the same context is - * requesting the lock twice -- this will not catch application - * breakages where the same context is active in two different threads - * at once, but it will catch driver breakages (recursive locking) in - * threaded apps. +/* These are functions now: */ -#define DEBUG_CHECK_LOCK() \ - do { \ - if ( *((volatile int *)intel->driHwLock) == \ - (DRM_LOCK_HELD | intel->hHWContext) ) { \ - fprintf( stderr, \ - "LOCK SET!\n\tPrevious %s:%d\n\tCurrent: %s:%d\n", \ - prevLockFile, prevLockLine, __FILE__, __LINE__ ); \ - abort(); \ - } \ - } while (0) - -#else - -#define DEBUG_LOCK() -#define DEBUG_RESET() -#define DEBUG_CHECK_LOCK() - -#endif - - +void LOCK_HARDWARE( struct intel_context *intel ); +void UNLOCK_HARDWARE( struct intel_context *intel ); - -/* Lock the hardware and validate our state. - */ -#define LOCK_HARDWARE( intel ) \ -do { \ - char __ret=0; \ - DEBUG_CHECK_LOCK(); \ - assert(!(intel)->locked); \ - if ((intel)->swap_scheduled) { \ - drmVBlank vbl; \ - vbl.request.type = DRM_VBLANK_ABSOLUTE; \ - if ((intel)->vblank_flags & \ - VBLANK_FLAG_SECONDARY) { \ - vbl.request.type |= DRM_VBLANK_SECONDARY; \ - } \ - vbl.request.sequence = (intel)->vbl_seq; \ - drmWaitVBlank((intel)->driFd, &vbl); \ - (intel)->swap_scheduled = 0; \ - } \ - DRM_CAS((intel)->driHwLock, (intel)->hHWContext, \ - (DRM_LOCK_HELD|(intel)->hHWContext), __ret); \ - if (__ret) \ - intelGetLock( (intel), 0 ); \ - DEBUG_LOCK(); \ - (intel)->locked = 1; \ -}while (0) - - - /* Unlock the hardware using the global current context - */ -#define UNLOCK_HARDWARE(intel) \ -do { \ - intel->locked = 0; \ - if (0) { \ - intel->perf_boxes |= intel->sarea->perf_boxes; \ - intel->sarea->perf_boxes = 0; \ - } \ - DRM_UNLOCK((intel)->driFd, (intel)->driHwLock, (intel)->hHWContext); \ - DEBUG_RESET(); \ -} while (0) +extern char *__progname; #define SUBPIXEL_X 0.125 #define SUBPIXEL_Y 0.125 +#define ALIGN(value, alignment) ((value + alignment - 1) & ~(alignment - 1)) + #define INTEL_FIREVERTICES(intel) \ do { \ if ((intel)->prim.flush) \ - (intel)->prim.flush(&(intel)->ctx); \ + (intel)->prim.flush(intel); \ } while (0) /* ================================================================ @@ -385,34 +318,26 @@ do { \ ((a<<24) | (r<<16) | (g<<8) | b) -#define INTEL_PACKCOLOR(format, r, g, b, a) \ -(format == DV_PF_555 ? INTEL_PACKCOLOR1555(r,g,b,a) : \ - (format == DV_PF_565 ? INTEL_PACKCOLOR565(r,g,b) : \ - (format == DV_PF_8888 ? INTEL_PACKCOLOR8888(r,g,b,a) : \ - 0))) - - /* ================================================================ * From linux kernel i386 header files, copes with odd sizes better * than COPY_DWORDS would: + * XXX Put this in src/mesa/main/imports.h ??? */ #if defined(i386) || defined(__i386__) -static __inline__ void * __memcpy(void * to, const void * from, size_t n) +static INLINE void * +__memcpy(void *to, const void *from, size_t n) { int d0, d1, d2; - __asm__ __volatile__( - "rep ; movsl\n\t" - "testb $2,%b4\n\t" - "je 1f\n\t" - "movsw\n" - "1:\ttestb $1,%b4\n\t" - "je 2f\n\t" - "movsb\n" - "2:" - : "=&c" (d0), "=&D" (d1), "=&S" (d2) - :"0" (n/4), "q" (n),"1" ((long) to),"2" ((long) from) - : "memory"); + __asm__ __volatile__("rep ; movsl\n\t" + "testb $2,%b4\n\t" + "je 1f\n\t" + "movsw\n" + "1:\ttestb $1,%b4\n\t" + "je 2f\n\t" + "movsb\n" "2:":"=&c"(d0), "=&D"(d1), "=&S"(d2) + :"0"(n / 4), "q"(n), "1"((long) to), "2"((long) from) + :"memory"); return (to); } #else @@ -434,16 +359,18 @@ extern int INTEL_DEBUG; #define DEBUG_TEXTURE 0x1 #define DEBUG_STATE 0x2 #define DEBUG_IOCTL 0x4 -#define DEBUG_PRIMS 0x8 -#define DEBUG_VERTS 0x10 +#define DEBUG_BLIT 0x8 +#define DEBUG_MIPTREE 0x10 #define DEBUG_FALLBACKS 0x20 #define DEBUG_VERBOSE 0x40 -#define DEBUG_DRI 0x80 -#define DEBUG_DMA 0x100 -#define DEBUG_SANITY 0x200 -#define DEBUG_SYNC 0x400 -#define DEBUG_SLEEP 0x800 -#define DEBUG_PIXEL 0x1000 +#define DEBUG_BATCH 0x80 +#define DEBUG_PIXEL 0x100 +#define DEBUG_BUFMGR 0x200 +#define DEBUG_REGION 0x400 +#define DEBUG_FBO 0x800 +#define DEBUG_LOCK 0x1000 + +#define DBG(...) do { if (INTEL_DEBUG & FILE_DEBUG_FLAG) _mesa_printf(__VA_ARGS__); } while(0) #define PCI_CHIP_845_G 0x2562 @@ -464,26 +391,24 @@ extern int INTEL_DEBUG; * intel_context.c: */ -extern void intelInitDriverFunctions( struct dd_function_table *functions ); +extern GLboolean intelInitContext(struct intel_context *intel, + const __GLcontextModes * mesaVis, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate, + struct dd_function_table *functions); -extern GLboolean intelInitContext( intelContextPtr intel, - const __GLcontextModes *mesaVis, - __DRIcontextPrivate *driContextPriv, - void *sharedContextPrivate, - struct dd_function_table *functions ); +extern void intelGetLock(struct intel_context *intel, GLuint flags); -extern void intelGetLock(intelContextPtr intel, GLuint flags); -extern void intelSetBackClipRects(intelContextPtr intel); -extern void intelSetFrontClipRects(intelContextPtr intel); -extern void intelWindowMoved( intelContextPtr intel ); +extern void intelFinish(GLcontext * ctx); +extern void intelFlush(GLcontext * ctx); -extern const GLubyte *intelGetString( GLcontext *ctx, GLenum name ); +extern void intelInitDriverFunctions(struct dd_function_table *functions); /* ================================================================ * intel_state.c: */ -extern void intelInitStateFuncs( struct dd_function_table *functions ); +extern void intelInitStateFuncs(struct dd_function_table *functions); #define COMPAREFUNC_ALWAYS 0 #define COMPAREFUNC_NEVER 0x1 @@ -537,27 +462,36 @@ extern void intelInitStateFuncs( struct dd_function_table *functions ); #define BLENDFACT_INV_CONST_ALPHA 0x0f #define BLENDFACT_MASK 0x0f - -extern int intel_translate_compare_func( GLenum func ); -extern int intel_translate_stencil_op( GLenum op ); -extern int intel_translate_blend_factor( GLenum factor ); -extern int intel_translate_logic_op( GLenum opcode ); +extern int intel_translate_compare_func(GLenum func); +extern int intel_translate_stencil_op(GLenum op); +extern int intel_translate_blend_factor(GLenum factor); +extern int intel_translate_logic_op(GLenum opcode); -/* ================================================================ - * intel_ioctl.c: +/*====================================================================== + * Inline conversion functions. + * These are better-typed than the macros used previously: */ -extern void intel_dump_batchbuffer( long offset, - int *ptr, - int count ); +static INLINE struct intel_context * +intel_context(GLcontext * ctx) +{ + return (struct intel_context *) ctx; +} +static INLINE struct intel_texture_object * +intel_texture_object(struct gl_texture_object *obj) +{ + return (struct intel_texture_object *) obj; +} -/* ================================================================ - * intel_pixel.c: - */ -extern void intelInitPixelFuncs( struct dd_function_table *functions ); +static INLINE struct intel_texture_image * +intel_texture_image(struct gl_texture_image *img) +{ + return (struct intel_texture_image *) img; +} +extern struct intel_renderbuffer *intel_renderbuffer(struct gl_renderbuffer + *rb); #endif - diff --git a/src/mesa/drivers/dri/i915/intel_decode.c b/src/mesa/drivers/dri/i915/intel_decode.c new file mode 120000 index 0000000000..f671b6cbb1 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_decode.c @@ -0,0 +1 @@ +../intel/intel_decode.c
\ No newline at end of file diff --git a/src/mesa/drivers/dri/i915/intel_depthstencil.c b/src/mesa/drivers/dri/i915/intel_depthstencil.c new file mode 100644 index 0000000000..d269a85a3c --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_depthstencil.c @@ -0,0 +1,282 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "glheader.h" +#include "imports.h" +#include "context.h" +#include "depthstencil.h" +#include "fbobject.h" +#include "framebuffer.h" +#include "hash.h" +#include "mtypes.h" +#include "renderbuffer.h" + +#include "intel_context.h" +#include "intel_fbo.h" +#include "intel_depthstencil.h" +#include "intel_regions.h" + + +/** + * The GL_EXT_framebuffer_object allows the user to create their own + * framebuffer objects consisting of color renderbuffers (0 or more), + * depth renderbuffers (0 or 1) and stencil renderbuffers (0 or 1). + * + * The spec considers depth and stencil renderbuffers to be totally independent + * buffers. In reality, most graphics hardware today uses a combined + * depth+stencil buffer (one 32-bit pixel = 24 bits of Z + 8 bits of stencil). + * + * This causes difficulty because the user may create some number of depth + * renderbuffers and some number of stencil renderbuffers and bind them + * together in framebuffers in any combination. + * + * This code manages all that. + * + * 1. Depth renderbuffers are always allocated in hardware as 32bpp + * GL_DEPTH24_STENCIL8 buffers. + * + * 2. Stencil renderbuffers are initially allocated in software as 8bpp + * GL_STENCIL_INDEX8 buffers. + * + * 3. Depth and Stencil renderbuffers use the PairedStencil and PairedDepth + * fields (respectively) to indicate if the buffer's currently paired + * with another stencil or depth buffer (respectively). + * + * 4. When a depth and stencil buffer are initially both attached to the + * current framebuffer, we merge the stencil buffer values into the + * depth buffer (really a depth+stencil buffer). The then hardware uses + * the combined buffer. + * + * 5. Whenever a depth or stencil buffer is reallocated (with + * glRenderbufferStorage) we undo the pairing and copy the stencil values + * from the combined depth/stencil buffer back to the stencil-only buffer. + * + * 6. We also undo the pairing when we find a change in buffer bindings. + * + * 7. If a framebuffer is only using a depth renderbuffer (no stencil), we + * just use the combined depth/stencil buffer and ignore the stencil values. + * + * 8. If a framebuffer is only using a stencil renderbuffer (no depth) we have + * to promote the 8bpp software stencil buffer to a 32bpp hardware + * depth+stencil buffer. + * + */ + + + +static void +map_regions(GLcontext * ctx, + struct intel_renderbuffer *depthRb, + struct intel_renderbuffer *stencilRb) +{ + struct intel_context *intel = intel_context(ctx); + if (depthRb && depthRb->region) { + intel_region_map(intel->intelScreen, depthRb->region); + depthRb->pfMap = depthRb->region->map; + depthRb->pfPitch = depthRb->region->pitch; + } + if (stencilRb && stencilRb->region) { + intel_region_map(intel->intelScreen, stencilRb->region); + stencilRb->pfMap = stencilRb->region->map; + stencilRb->pfPitch = stencilRb->region->pitch; + } +} + +static void +unmap_regions(GLcontext * ctx, + struct intel_renderbuffer *depthRb, + struct intel_renderbuffer *stencilRb) +{ + struct intel_context *intel = intel_context(ctx); + if (depthRb && depthRb->region) { + intel_region_unmap(intel->intelScreen, depthRb->region); + depthRb->pfMap = NULL; + depthRb->pfPitch = 0; + } + if (stencilRb && stencilRb->region) { + intel_region_unmap(intel->intelScreen, stencilRb->region); + stencilRb->pfMap = NULL; + stencilRb->pfPitch = 0; + } +} + + + +/** + * Undo the pairing/interleaving between depth and stencil buffers. + * irb should be a depth/stencil or stencil renderbuffer. + */ +void +intel_unpair_depth_stencil(GLcontext * ctx, struct intel_renderbuffer *irb) +{ + if (irb->PairedStencil) { + /* irb is a depth/stencil buffer */ + struct gl_renderbuffer *stencilRb; + struct intel_renderbuffer *stencilIrb; + + ASSERT(irb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); + + stencilRb = _mesa_lookup_renderbuffer(ctx, irb->PairedStencil); + stencilIrb = intel_renderbuffer(stencilRb); + if (stencilIrb) { + /* need to extract stencil values from the depth buffer */ + ASSERT(stencilIrb->PairedDepth == irb->Base.Name); + map_regions(ctx, irb, stencilIrb); + _mesa_extract_stencil(ctx, &irb->Base, &stencilIrb->Base); + unmap_regions(ctx, irb, stencilIrb); + stencilIrb->PairedDepth = 0; + } + irb->PairedStencil = 0; + } + else if (irb->PairedDepth) { + /* irb is a stencil buffer */ + struct gl_renderbuffer *depthRb; + struct intel_renderbuffer *depthIrb; + + ASSERT(irb->Base._ActualFormat == GL_STENCIL_INDEX8_EXT || + irb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); + + depthRb = _mesa_lookup_renderbuffer(ctx, irb->PairedDepth); + depthIrb = intel_renderbuffer(depthRb); + if (depthIrb) { + /* need to extract stencil values from the depth buffer */ + ASSERT(depthIrb->PairedStencil == irb->Base.Name); + map_regions(ctx, depthIrb, irb); + _mesa_extract_stencil(ctx, &depthIrb->Base, &irb->Base); + unmap_regions(ctx, depthIrb, irb); + depthIrb->PairedStencil = 0; + } + irb->PairedDepth = 0; + } + else { + _mesa_problem(ctx, "Problem in undo_depth_stencil_pairing"); + } + + ASSERT(irb->PairedStencil == 0); + ASSERT(irb->PairedDepth == 0); +} + + +/** + * Examine the depth and stencil renderbuffers which are attached to the + * framebuffer. If both depth and stencil are attached, make sure that the + * renderbuffers are 'paired' (combined). If only depth or only stencil is + * attached, undo any previous pairing. + * + * Must be called if NewState & _NEW_BUFFER (when renderbuffer attachments + * change, for example). + */ +void +intel_validate_paired_depth_stencil(GLcontext * ctx, + struct gl_framebuffer *fb) +{ + struct intel_renderbuffer *depthRb, *stencilRb; + + depthRb = intel_get_renderbuffer(fb, BUFFER_DEPTH); + stencilRb = intel_get_renderbuffer(fb, BUFFER_STENCIL); + + if (depthRb && stencilRb) { + if (depthRb == stencilRb) { + /* Using a user-created combined depth/stencil buffer. + * Nothing to do. + */ + ASSERT(depthRb->Base._BaseFormat == GL_DEPTH_STENCIL_EXT); + ASSERT(depthRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); + } + else { + /* Separate depth/stencil buffers, need to interleave now */ + ASSERT(depthRb->Base._BaseFormat == GL_DEPTH_COMPONENT); + ASSERT(stencilRb->Base._BaseFormat == GL_STENCIL_INDEX); + /* may need to interleave depth/stencil now */ + if (depthRb->PairedStencil == stencilRb->Base.Name) { + /* OK, the depth and stencil buffers are already interleaved */ + ASSERT(stencilRb->PairedDepth == depthRb->Base.Name); + } + else { + /* need to setup new pairing/interleaving */ + if (depthRb->PairedStencil) { + intel_unpair_depth_stencil(ctx, depthRb); + } + if (stencilRb->PairedDepth) { + intel_unpair_depth_stencil(ctx, stencilRb); + } + + ASSERT(depthRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); + ASSERT(stencilRb->Base._ActualFormat == GL_STENCIL_INDEX8_EXT || + stencilRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); + + /* establish new pairing: interleave stencil into depth buffer */ + map_regions(ctx, depthRb, stencilRb); + _mesa_insert_stencil(ctx, &depthRb->Base, &stencilRb->Base); + unmap_regions(ctx, depthRb, stencilRb); + depthRb->PairedStencil = stencilRb->Base.Name; + stencilRb->PairedDepth = depthRb->Base.Name; + } + + } + } + else if (depthRb) { + /* Depth buffer but no stencil buffer. + * We'll use a GL_DEPTH24_STENCIL8 buffer and ignore the stencil bits. + */ + /* can't assert this until storage is allocated: + ASSERT(depthRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); + */ + /* intel_undo any previous pairing */ + if (depthRb->PairedStencil) { + intel_unpair_depth_stencil(ctx, depthRb); + } + } + else if (stencilRb) { + /* Stencil buffer but no depth buffer. + * Since h/w doesn't typically support just 8bpp stencil w/out Z, + * we'll use a GL_DEPTH24_STENCIL8 buffer and ignore the depth bits. + */ + /* undo any previous pairing */ + if (stencilRb->PairedDepth) { + intel_unpair_depth_stencil(ctx, stencilRb); + } + if (stencilRb->Base._ActualFormat == GL_STENCIL_INDEX8_EXT) { + /* promote buffer to GL_DEPTH24_STENCIL8 for hw rendering */ + _mesa_promote_stencil(ctx, &stencilRb->Base); + ASSERT(stencilRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); + } + } + + /* Finally, update the fb->_DepthBuffer and fb->_StencilBuffer fields */ + _mesa_update_depth_buffer(ctx, fb, BUFFER_DEPTH); + if (depthRb && depthRb->PairedStencil) + _mesa_update_stencil_buffer(ctx, fb, BUFFER_DEPTH); + else + _mesa_update_stencil_buffer(ctx, fb, BUFFER_STENCIL); + + + /* The hardware should use fb->Attachment[BUFFER_DEPTH].Renderbuffer + * first, if present, then fb->Attachment[BUFFER_STENCIL].Renderbuffer + * if present. + */ +} diff --git a/src/mesa/drivers/dri/i915/intel_depthstencil.h b/src/mesa/drivers/dri/i915/intel_depthstencil.h new file mode 100644 index 0000000000..2d3fc48b3a --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_depthstencil.h @@ -0,0 +1,14 @@ + +#ifndef INTEL_DEPTH_STENCIL_H +#define INTEL_DEPTH_STENCIL_H + + +extern void +intel_unpair_depth_stencil(GLcontext * ctx, struct intel_renderbuffer *irb); + +extern void +intel_validate_paired_depth_stencil(GLcontext * ctx, + struct gl_framebuffer *fb); + + +#endif /* INTEL_DEPTH_STENCIL_H */ diff --git a/src/mesa/drivers/dri/i915/intel_fbo.c b/src/mesa/drivers/dri/i915/intel_fbo.c new file mode 100644 index 0000000000..6f99f401c7 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_fbo.c @@ -0,0 +1,687 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "imports.h" +#include "mtypes.h" +#include "fbobject.h" +#include "framebuffer.h" +#include "renderbuffer.h" +#include "context.h" +#include "texformat.h" +#include "texrender.h" + +#include "intel_context.h" +#include "intel_buffers.h" +#include "intel_depthstencil.h" +#include "intel_fbo.h" +#include "intel_mipmap_tree.h" +#include "intel_regions.h" +#include "intel_span.h" + + +#define FILE_DEBUG_FLAG DEBUG_FBO + +#define INTEL_RB_CLASS 0x12345678 + + +/* XXX FBO: move this to intel_context.h (inlined) */ +/** + * Return a gl_renderbuffer ptr casted to intel_renderbuffer. + * NULL will be returned if the rb isn't really an intel_renderbuffer. + * This is determiend by checking the ClassID. + */ +struct intel_renderbuffer * +intel_renderbuffer(struct gl_renderbuffer *rb) +{ + struct intel_renderbuffer *irb = (struct intel_renderbuffer *) rb; + if (irb && irb->Base.ClassID == INTEL_RB_CLASS) { + /*_mesa_warning(NULL, "Returning non-intel Rb\n");*/ + return irb; + } + else + return NULL; +} + + +struct intel_renderbuffer * +intel_get_renderbuffer(struct gl_framebuffer *fb, GLuint attIndex) +{ + return intel_renderbuffer(fb->Attachment[attIndex].Renderbuffer); +} + + +void +intel_flip_renderbuffers(struct intel_framebuffer *intel_fb) +{ + int current_page = intel_fb->pf_current_page; + int next_page = (current_page + 1) % intel_fb->pf_num_pages; + struct gl_renderbuffer *tmp_rb; + + /* Exchange renderbuffers if necessary but make sure their reference counts + * are preserved. + */ + if (intel_fb->color_rb[current_page] && + intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer != + &intel_fb->color_rb[current_page]->Base) { + tmp_rb = NULL; + _mesa_reference_renderbuffer(&tmp_rb, + intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer); + tmp_rb = &intel_fb->color_rb[current_page]->Base; + _mesa_reference_renderbuffer( + &intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer, tmp_rb); + _mesa_reference_renderbuffer(&tmp_rb, NULL); + } + + if (intel_fb->color_rb[next_page] && + intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer != + &intel_fb->color_rb[next_page]->Base) { + tmp_rb = NULL; + _mesa_reference_renderbuffer(&tmp_rb, + intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer); + tmp_rb = &intel_fb->color_rb[next_page]->Base; + _mesa_reference_renderbuffer( + &intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer, tmp_rb); + _mesa_reference_renderbuffer(&tmp_rb, NULL); + } +} + + +struct intel_region * +intel_get_rb_region(struct gl_framebuffer *fb, GLuint attIndex) +{ + struct intel_renderbuffer *irb = intel_get_renderbuffer(fb, attIndex); + + if (irb) + return irb->region; + else + return NULL; +} + + + +/** + * Create a new framebuffer object. + */ +static struct gl_framebuffer * +intel_new_framebuffer(GLcontext * ctx, GLuint name) +{ + /* Only drawable state in intel_framebuffer at this time, just use Mesa's + * class + */ + return _mesa_new_framebuffer(ctx, name); +} + + +static void +intel_delete_renderbuffer(struct gl_renderbuffer *rb) +{ + GET_CURRENT_CONTEXT(ctx); + struct intel_context *intel = intel_context(ctx); + struct intel_renderbuffer *irb = intel_renderbuffer(rb); + + ASSERT(irb); + + if (irb->PairedStencil || irb->PairedDepth) { + intel_unpair_depth_stencil(ctx, irb); + } + + if (intel && irb->region) { + intel_region_release(&irb->region); + } + + _mesa_free(irb); +} + + + +/** + * Return a pointer to a specific pixel in a renderbuffer. + */ +static void * +intel_get_pointer(GLcontext * ctx, struct gl_renderbuffer *rb, + GLint x, GLint y) +{ + /* By returning NULL we force all software rendering to go through + * the span routines. + */ + return NULL; +} + + + +/** + * Called via glRenderbufferStorageEXT() to set the format and allocate + * storage for a user-created renderbuffer. + */ +static GLboolean +intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, + GLenum internalFormat, + GLuint width, GLuint height) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_renderbuffer *irb = intel_renderbuffer(rb); + GLboolean softwareBuffer = GL_FALSE; + int cpp; + + ASSERT(rb->Name != 0); + + switch (internalFormat) { + case GL_R3_G3_B2: + case GL_RGB4: + case GL_RGB5: + rb->_ActualFormat = GL_RGB5; + rb->DataType = GL_UNSIGNED_BYTE; + rb->RedBits = 5; + rb->GreenBits = 6; + rb->BlueBits = 5; + cpp = 2; + break; + case GL_RGB: + case GL_RGB8: + case GL_RGB10: + case GL_RGB12: + case GL_RGB16: + case GL_RGBA: + case GL_RGBA2: + case GL_RGBA4: + case GL_RGB5_A1: + case GL_RGBA8: + case GL_RGB10_A2: + case GL_RGBA12: + case GL_RGBA16: + rb->_ActualFormat = GL_RGBA8; + rb->DataType = GL_UNSIGNED_BYTE; + rb->RedBits = 8; + rb->GreenBits = 8; + rb->BlueBits = 8; + rb->AlphaBits = 8; + cpp = 4; + break; + case GL_STENCIL_INDEX: + case GL_STENCIL_INDEX1_EXT: + case GL_STENCIL_INDEX4_EXT: + case GL_STENCIL_INDEX8_EXT: + case GL_STENCIL_INDEX16_EXT: + /* alloc a depth+stencil buffer */ + rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT; + rb->DataType = GL_UNSIGNED_INT_24_8_EXT; + rb->StencilBits = 8; + cpp = 4; + break; + case GL_DEPTH_COMPONENT16: + rb->_ActualFormat = GL_DEPTH_COMPONENT16; + rb->DataType = GL_UNSIGNED_SHORT; + rb->DepthBits = 16; + cpp = 2; + break; + case GL_DEPTH_COMPONENT: + case GL_DEPTH_COMPONENT24: + case GL_DEPTH_COMPONENT32: + rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT; + rb->DataType = GL_UNSIGNED_INT_24_8_EXT; + rb->DepthBits = 24; + cpp = 4; + break; + case GL_DEPTH_STENCIL_EXT: + case GL_DEPTH24_STENCIL8_EXT: + rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT; + rb->DataType = GL_UNSIGNED_INT_24_8_EXT; + rb->DepthBits = 24; + rb->StencilBits = 8; + cpp = 4; + break; + default: + _mesa_problem(ctx, + "Unexpected format in intel_alloc_renderbuffer_storage"); + return GL_FALSE; + } + + intelFlush(ctx); + + /* free old region */ + if (irb->region) { + intel_region_release(&irb->region); + } + + /* allocate new memory region/renderbuffer */ + if (softwareBuffer) { + return _mesa_soft_renderbuffer_storage(ctx, rb, internalFormat, + width, height); + } + else { + /* Choose a pitch to match hardware requirements: + */ + GLuint pitch = ((cpp * width + 63) & ~63) / cpp; + + /* alloc hardware renderbuffer */ + DBG("Allocating %d x %d Intel RBO (pitch %d)\n", width, + height, pitch); + + irb->region = intel_region_alloc(intel->intelScreen, cpp, pitch, height); + if (!irb->region) + return GL_FALSE; /* out of memory? */ + + ASSERT(irb->region->buffer); + + rb->Width = width; + rb->Height = height; + + /* This sets the Get/PutRow/Value functions */ + intel_set_span_functions(&irb->Base); + + return GL_TRUE; + } +} + + + +/** + * Called for each hardware renderbuffer when a _window_ is resized. + * Just update fields. + * Not used for user-created renderbuffers! + */ +static GLboolean +intel_alloc_window_storage(GLcontext * ctx, struct gl_renderbuffer *rb, + GLenum internalFormat, GLuint width, GLuint height) +{ + ASSERT(rb->Name == 0); + rb->Width = width; + rb->Height = height; + rb->_ActualFormat = internalFormat; + + return GL_TRUE; +} + +static void +intel_resize_buffers(GLcontext *ctx, struct gl_framebuffer *fb, + GLuint width, GLuint height) +{ + struct intel_framebuffer *intel_fb = (struct intel_framebuffer*)fb; + int i; + + _mesa_resize_framebuffer(ctx, fb, width, height); + + fb->Initialized = GL_TRUE; /* XXX remove someday */ + + if (fb->Name != 0) { + return; + } + + /* Make sure all window system renderbuffers are up to date */ + for (i = 0; i < 3; i++) { + struct gl_renderbuffer *rb = &intel_fb->color_rb[i]->Base; + + /* only resize if size is changing */ + if (rb && (rb->Width != width || rb->Height != height)) { + rb->AllocStorage(ctx, rb, rb->InternalFormat, width, height); + } + } +} + +static GLboolean +intel_nop_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, + GLenum internalFormat, GLuint width, GLuint height) +{ + _mesa_problem(ctx, "intel_op_alloc_storage should never be called."); + return GL_FALSE; +} + + + +/** + * Create a new intel_renderbuffer which corresponds to an on-screen window, + * not a user-created renderbuffer. + * \param width the screen width + * \param height the screen height + */ +struct intel_renderbuffer * +intel_create_renderbuffer(GLenum intFormat, GLsizei width, GLsizei height, + int offset, int pitch, int cpp, void *map) +{ + GET_CURRENT_CONTEXT(ctx); + + struct intel_renderbuffer *irb; + const GLuint name = 0; + + irb = CALLOC_STRUCT(intel_renderbuffer); + if (!irb) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "creating renderbuffer"); + return NULL; + } + + _mesa_init_renderbuffer(&irb->Base, name); + irb->Base.ClassID = INTEL_RB_CLASS; + + switch (intFormat) { + case GL_RGB5: + irb->Base._ActualFormat = GL_RGB5; + irb->Base._BaseFormat = GL_RGBA; + irb->Base.RedBits = 5; + irb->Base.GreenBits = 6; + irb->Base.BlueBits = 5; + irb->Base.DataType = GL_UNSIGNED_BYTE; + cpp = 2; + break; + case GL_RGBA8: + irb->Base._ActualFormat = GL_RGBA8; + irb->Base._BaseFormat = GL_RGBA; + irb->Base.RedBits = 8; + irb->Base.GreenBits = 8; + irb->Base.BlueBits = 8; + irb->Base.AlphaBits = 8; + irb->Base.DataType = GL_UNSIGNED_BYTE; + cpp = 4; + break; + case GL_STENCIL_INDEX8_EXT: + irb->Base._ActualFormat = GL_STENCIL_INDEX8_EXT; + irb->Base._BaseFormat = GL_STENCIL_INDEX; + irb->Base.StencilBits = 8; + irb->Base.DataType = GL_UNSIGNED_BYTE; + cpp = 1; + break; + case GL_DEPTH_COMPONENT16: + irb->Base._ActualFormat = GL_DEPTH_COMPONENT16; + irb->Base._BaseFormat = GL_DEPTH_COMPONENT; + irb->Base.DepthBits = 16; + irb->Base.DataType = GL_UNSIGNED_SHORT; + cpp = 2; + break; + case GL_DEPTH_COMPONENT24: + irb->Base._ActualFormat = GL_DEPTH24_STENCIL8_EXT; + irb->Base._BaseFormat = GL_DEPTH_COMPONENT; + irb->Base.DepthBits = 24; + irb->Base.DataType = GL_UNSIGNED_INT; + cpp = 4; + break; + case GL_DEPTH24_STENCIL8_EXT: + irb->Base._ActualFormat = GL_DEPTH24_STENCIL8_EXT; + irb->Base._BaseFormat = GL_DEPTH_STENCIL_EXT; + irb->Base.DepthBits = 24; + irb->Base.StencilBits = 8; + irb->Base.DataType = GL_UNSIGNED_INT_24_8_EXT; + cpp = 4; + break; + default: + _mesa_problem(NULL, + "Unexpected intFormat in intel_create_renderbuffer"); + return NULL; + } + + irb->Base.InternalFormat = intFormat; + + /* intel-specific methods */ + irb->Base.Delete = intel_delete_renderbuffer; + irb->Base.AllocStorage = intel_alloc_window_storage; + irb->Base.GetPointer = intel_get_pointer; + /* This sets the Get/PutRow/Value functions */ + intel_set_span_functions(&irb->Base); + + irb->pfMap = map; + irb->pfPitch = pitch / cpp; /* in pixels */ + +#if 00 + irb->region = intel_region_create_static(intel, + DRM_MM_TT, + offset, map, cpp, width, height); +#endif + + return irb; +} + + +/** + * Create a new renderbuffer object. + * Typically called via glBindRenderbufferEXT(). + */ +static struct gl_renderbuffer * +intel_new_renderbuffer(GLcontext * ctx, GLuint name) +{ + /*struct intel_context *intel = intel_context(ctx); */ + struct intel_renderbuffer *irb; + + irb = CALLOC_STRUCT(intel_renderbuffer); + if (!irb) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "creating renderbuffer"); + return NULL; + } + + _mesa_init_renderbuffer(&irb->Base, name); + irb->Base.ClassID = INTEL_RB_CLASS; + + /* intel-specific methods */ + irb->Base.Delete = intel_delete_renderbuffer; + irb->Base.AllocStorage = intel_alloc_renderbuffer_storage; + irb->Base.GetPointer = intel_get_pointer; + /* span routines set in alloc_storage function */ + + return &irb->Base; +} + + +/** + * Called via glBindFramebufferEXT(). + */ +static void +intel_bind_framebuffer(GLcontext * ctx, GLenum target, + struct gl_framebuffer *fb, struct gl_framebuffer *fbread) +{ + if (target == GL_FRAMEBUFFER_EXT || target == GL_DRAW_FRAMEBUFFER_EXT) { + intel_draw_buffer(ctx, fb); + /* Integer depth range depends on depth buffer bits */ + ctx->Driver.DepthRange(ctx, ctx->Viewport.Near, ctx->Viewport.Far); + } + else { + /* don't need to do anything if target == GL_READ_FRAMEBUFFER_EXT */ + } +} + + +/** + * Called via glFramebufferRenderbufferEXT(). + */ +static void +intel_framebuffer_renderbuffer(GLcontext * ctx, + struct gl_framebuffer *fb, + GLenum attachment, struct gl_renderbuffer *rb) +{ + DBG("Intel FramebufferRenderbuffer %u %u\n", fb->Name, rb ? rb->Name : 0); + + intelFlush(ctx); + + _mesa_framebuffer_renderbuffer(ctx, fb, attachment, rb); + intel_draw_buffer(ctx, fb); +} + + +/** + * When glFramebufferTexture[123]D is called this function sets up the + * gl_renderbuffer wrapper around the texture image. + * This will have the region info needed for hardware rendering. + */ +static struct intel_renderbuffer * +intel_wrap_texture(GLcontext * ctx, struct gl_texture_image *texImage) +{ + const GLuint name = ~0; /* not significant, but distinct for debugging */ + struct intel_renderbuffer *irb; + + /* make an intel_renderbuffer to wrap the texture image */ + irb = CALLOC_STRUCT(intel_renderbuffer); + if (!irb) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "glFramebufferTexture"); + return NULL; + } + + _mesa_init_renderbuffer(&irb->Base, name); + irb->Base.ClassID = INTEL_RB_CLASS; + + if (texImage->TexFormat == &_mesa_texformat_argb8888) { + irb->Base._ActualFormat = GL_RGBA8; + irb->Base._BaseFormat = GL_RGBA; + DBG("Render to RGBA8 texture OK\n"); + } + else if (texImage->TexFormat == &_mesa_texformat_rgb565) { + irb->Base._ActualFormat = GL_RGB5; + irb->Base._BaseFormat = GL_RGB; + DBG("Render to RGB5 texture OK\n"); + } + else if (texImage->TexFormat == &_mesa_texformat_z16) { + irb->Base._ActualFormat = GL_DEPTH_COMPONENT16; + irb->Base._BaseFormat = GL_DEPTH_COMPONENT; + DBG("Render to DEPTH16 texture OK\n"); + } + else { + DBG("Render to texture BAD FORMAT %d\n", + texImage->TexFormat->MesaFormat); + _mesa_free(irb); + return NULL; + } + + irb->Base.InternalFormat = irb->Base._ActualFormat; + irb->Base.Width = texImage->Width; + irb->Base.Height = texImage->Height; + irb->Base.DataType = GL_UNSIGNED_BYTE; /* FBO XXX fix */ + irb->Base.RedBits = texImage->TexFormat->RedBits; + irb->Base.GreenBits = texImage->TexFormat->GreenBits; + irb->Base.BlueBits = texImage->TexFormat->BlueBits; + irb->Base.AlphaBits = texImage->TexFormat->AlphaBits; + irb->Base.DepthBits = texImage->TexFormat->DepthBits; + + irb->Base.Delete = intel_delete_renderbuffer; + irb->Base.AllocStorage = intel_nop_alloc_storage; + intel_set_span_functions(&irb->Base); + + irb->RenderToTexture = GL_TRUE; + + return irb; +} + + +/** + * Called by glFramebufferTexture[123]DEXT() (and other places) to + * prepare for rendering into texture memory. This might be called + * many times to choose different texture levels, cube faces, etc + * before intel_finish_render_texture() is ever called. + */ +static void +intel_render_texture(GLcontext * ctx, + struct gl_framebuffer *fb, + struct gl_renderbuffer_attachment *att) +{ + struct gl_texture_image *newImage + = att->Texture->Image[att->CubeMapFace][att->TextureLevel]; + struct intel_renderbuffer *irb = intel_renderbuffer(att->Renderbuffer); + struct intel_texture_image *intel_image; + GLuint imageOffset; + + (void) fb; + + ASSERT(newImage); + + if (!irb) { + irb = intel_wrap_texture(ctx, newImage); + if (irb) { + /* bind the wrapper to the attachment point */ + _mesa_reference_renderbuffer(&att->Renderbuffer, &irb->Base); + } + else { + /* fallback to software rendering */ + _mesa_render_texture(ctx, fb, att); + return; + } + } + + DBG("Begin render texture tid %x tex=%u w=%d h=%d refcount=%d\n", + _glthread_GetID(), + att->Texture->Name, newImage->Width, newImage->Height, + irb->Base.RefCount); + + /* point the renderbufer's region to the texture image region */ + intel_image = intel_texture_image(newImage); + if (irb->region != intel_image->mt->region) { + if (irb->region) + intel_region_release(&irb->region); + intel_region_reference(&irb->region, intel_image->mt->region); + } + + /* compute offset of the particular 2D image within the texture region */ + imageOffset = intel_miptree_image_offset(intel_image->mt, + att->CubeMapFace, + att->TextureLevel); + + if (att->Texture->Target == GL_TEXTURE_3D) { + const GLuint *offsets = intel_miptree_depth_offsets(intel_image->mt, + att->TextureLevel); + imageOffset += offsets[att->Zoffset]; + } + + /* store that offset in the region */ + intel_image->mt->region->draw_offset = imageOffset; + + /* update drawing region, etc */ + intel_draw_buffer(ctx, fb); +} + + +/** + * Called by Mesa when rendering to a texture is done. + */ +static void +intel_finish_render_texture(GLcontext * ctx, + struct gl_renderbuffer_attachment *att) +{ + struct intel_renderbuffer *irb = intel_renderbuffer(att->Renderbuffer); + + DBG("End render texture (tid %x) tex %u\n", _glthread_GetID(), att->Texture->Name); + + if (irb) { + /* just release the region */ + intel_region_release(&irb->region); + } + else if (att->Renderbuffer) { + /* software fallback */ + _mesa_finish_render_texture(ctx, att); + /* XXX FBO: Need to unmap the buffer (or in intelSpanRenderStart???) */ + } +} + + +/** + * Do one-time context initializations related to GL_EXT_framebuffer_object. + * Hook in device driver functions. + */ +void +intel_fbo_init(struct intel_context *intel) +{ + intel->ctx.Driver.NewFramebuffer = intel_new_framebuffer; + intel->ctx.Driver.NewRenderbuffer = intel_new_renderbuffer; + intel->ctx.Driver.BindFramebuffer = intel_bind_framebuffer; + intel->ctx.Driver.FramebufferRenderbuffer = intel_framebuffer_renderbuffer; + intel->ctx.Driver.RenderTexture = intel_render_texture; + intel->ctx.Driver.FinishRenderTexture = intel_finish_render_texture; + intel->ctx.Driver.ResizeBuffers = intel_resize_buffers; +} diff --git a/src/mesa/drivers/dri/i915/intel_fbo.h b/src/mesa/drivers/dri/i915/intel_fbo.h new file mode 100644 index 0000000000..411d634231 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_fbo.h @@ -0,0 +1,115 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_FBO_H +#define INTEL_FBO_H + + +struct intel_context; +struct intel_region; + +/** + * Intel framebuffer, derived from gl_framebuffer. + */ +struct intel_framebuffer +{ + struct gl_framebuffer Base; + + struct intel_renderbuffer *color_rb[3]; + + /* Drawable page flipping state */ + GLboolean pf_active; + GLuint pf_seq; + GLint pf_planes; + GLint pf_current_page; + GLint pf_num_pages; + + /* VBI + */ + GLuint vbl_seq; + GLuint vblank_flags; + GLuint vbl_waited; + + int64_t swap_ust; + int64_t swap_missed_ust; + + GLuint swap_count; + GLuint swap_missed_count; +}; + + +/** + * Intel renderbuffer, derived from gl_renderbuffer. + * Note: The PairedDepth and PairedStencil fields use renderbuffer IDs, + * not pointers because in some circumstances a deleted renderbuffer could + * result in a dangling pointer here. + */ +struct intel_renderbuffer +{ + struct gl_renderbuffer Base; + struct intel_region *region; + void *pfMap; /* possibly paged flipped map pointer */ + GLuint pfPitch; /* possibly paged flipped pitch */ + GLboolean RenderToTexture; /* RTT? */ + + GLuint PairedDepth; /**< only used if this is a depth renderbuffer */ + GLuint PairedStencil; /**< only used if this is a stencil renderbuffer */ + + GLuint pf_pending; /**< sequence number of pending flip */ + + GLuint vbl_pending; /**< vblank sequence number of pending flip */ +}; + + +extern struct intel_renderbuffer *intel_create_renderbuffer(GLenum intFormat, + GLsizei width, + GLsizei height, + int offset, + int pitch, + int cpp, + void *map); + + +extern void intel_fbo_init(struct intel_context *intel); + + +/* XXX make inline or macro */ +extern struct intel_renderbuffer *intel_get_renderbuffer(struct gl_framebuffer + *fb, + GLuint attIndex); + +extern void intel_flip_renderbuffers(struct intel_framebuffer *intel_fb); + + +/* XXX make inline or macro */ +extern struct intel_region *intel_get_rb_region(struct gl_framebuffer *fb, + GLuint attIndex); + + + + +#endif /* INTEL_FBO_H */ diff --git a/src/mesa/drivers/dri/i915/intel_ioctl.c b/src/mesa/drivers/dri/i915/intel_ioctl.c index ede3b6378f..94f7e73ecf 100644 --- a/src/mesa/drivers/dri/i915/intel_ioctl.c +++ b/src/mesa/drivers/dri/i915/intel_ioctl.c @@ -38,622 +38,150 @@ #include "intel_context.h" #include "intel_ioctl.h" #include "intel_batchbuffer.h" +#include "intel_blit.h" +#include "intel_regions.h" #include "drm.h" -u_int32_t intelGetLastFrame (intelContextPtr intel) -{ - int ret; - u_int32_t frame; - drm_i915_getparam_t gp; - - gp.param = I915_PARAM_LAST_DISPATCH; - gp.value = (int *)&frame; - ret = drmCommandWriteRead( intel->driFd, DRM_I915_GETPARAM, - &gp, sizeof(gp) ); - return frame; -} +#include "intel_bufmgr_ttm.h" -int intelEmitIrqLocked( intelContextPtr intel ) +#define FILE_DEBUG_FLAG DEBUG_IOCTL + +int +intelEmitIrqLocked(intelScreenPrivate *intelScreen) { drmI830IrqEmit ie; int ret, seq; - - assert(((*(int *)intel->driHwLock) & ~DRM_LOCK_CONT) == - (DRM_LOCK_HELD|intel->hHWContext)); ie.irq_seq = &seq; - - ret = drmCommandWriteRead( intel->driFd, DRM_I830_IRQ_EMIT, - &ie, sizeof(ie) ); - if ( ret ) { - fprintf( stderr, "%s: drmI830IrqEmit: %d\n", __FUNCTION__, ret ); + + ret = drmCommandWriteRead(intelScreen->driScrnPriv->fd, + DRM_I830_IRQ_EMIT, &ie, sizeof(ie)); + if (ret) { + fprintf(stderr, "%s: drmI830IrqEmit: %d\n", __FUNCTION__, ret); exit(1); } - - if (0) - fprintf(stderr, "%s --> %d\n", __FUNCTION__, seq ); + + DBG("%s --> %d\n", __FUNCTION__, seq); return seq; } -void intelWaitIrq( intelContextPtr intel, int seq ) +void +intelWaitIrq(intelScreenPrivate *intelScreen, int seq) { + drm_i915_irq_wait_t iw; int ret; - - if (0) - fprintf(stderr, "%s %d\n", __FUNCTION__, seq ); - intel->iw.irq_seq = seq; - + DBG("%s %d\n", __FUNCTION__, seq); + + iw.irq_seq = seq; + do { - ret = drmCommandWrite( intel->driFd, DRM_I830_IRQ_WAIT, &intel->iw, sizeof(intel->iw) ); + ret = drmCommandWrite(intelScreen->driScrnPriv->fd, + DRM_I830_IRQ_WAIT, &iw, sizeof(iw)); } while (ret == -EAGAIN || ret == -EINTR); - if ( ret ) { - fprintf( stderr, "%s: drmI830IrqWait: %d\n", __FUNCTION__, ret ); - if (0) - intel_dump_batchbuffer( intel->alloc.offset, - intel->alloc.ptr, - intel->alloc.size ); + if (ret) { + fprintf(stderr, "%s: drmI830IrqWait: %d\n", __FUNCTION__, ret); exit(1); } } - -static void age_intel( intelContextPtr intel, int age ) -{ - GLuint i; - - for (i = 0 ; i < MAX_TEXTURE_UNITS ; i++) - if (intel->CurrentTexObj[i]) - intel->CurrentTexObj[i]->age = age; -} - -void intel_dump_batchbuffer( long offset, - int *ptr, - int count ) -{ - int i; - fprintf(stderr, "\n\n\nSTART BATCH (%d dwords):\n", count); - for (i = 0; i < count/4; i += 4) - fprintf(stderr, "\t0x%x: 0x%08x 0x%08x 0x%08x 0x%08x\n", - (unsigned int)offset + i*4, ptr[i], ptr[i+1], ptr[i+2], ptr[i+3]); - fprintf(stderr, "END BATCH\n\n\n"); -} - -void intelRefillBatchLocked( intelContextPtr intel, GLboolean allow_unlock ) -{ - GLuint last_irq = intel->alloc.irq_emitted; - GLuint half = intel->alloc.size / 2; - GLuint buf = (intel->alloc.active_buf ^= 1); - - intel->alloc.irq_emitted = intelEmitIrqLocked( intel ); - - if (last_irq) { - if (allow_unlock) UNLOCK_HARDWARE( intel ); - intelWaitIrq( intel, last_irq ); - if (allow_unlock) LOCK_HARDWARE( intel ); - } - - if (0) - fprintf(stderr, "%s: now using half %d\n", __FUNCTION__, buf); - - intel->batch.start_offset = intel->alloc.offset + buf * half; - intel->batch.ptr = (unsigned char *)intel->alloc.ptr + buf * half; - intel->batch.size = half - 8; - intel->batch.space = half - 8; - assert(intel->batch.space >= 0); -} - -#define MI_BATCH_BUFFER_END (0xA<<23) - - -void intelFlushBatchLocked( intelContextPtr intel, - GLboolean ignore_cliprects, - GLboolean refill, - GLboolean allow_unlock) +void +intel_batch_ioctl(struct intel_context *intel, + GLuint start_offset, + GLuint used, + GLboolean ignore_cliprects, GLboolean allow_unlock) { drmI830BatchBuffer batch; assert(intel->locked); + assert(used); - if (0) - fprintf(stderr, "%s used %d of %d offset %x..%x refill %d (started in %s)\n", - __FUNCTION__, - (intel->batch.size - intel->batch.space), - intel->batch.size, - intel->batch.start_offset, - intel->batch.start_offset + - (intel->batch.size - intel->batch.space), - refill, - intel->batch.func); + DBG("%s used %d offset %x..%x ignore_cliprects %d\n", + __FUNCTION__, + used, start_offset, start_offset + used, ignore_cliprects); /* Throw away non-effective packets. Won't work once we have * hardware contexts which would preserve statechanges beyond a * single buffer. */ - if (intel->numClipRects == 0 && !ignore_cliprects) { - - /* Without this yeild, an application with no cliprects can hog - * the hardware. Without unlocking, the effect is much worse - - * effectively a lock-out of other contexts. - */ - if (allow_unlock) { - UNLOCK_HARDWARE( intel ); - sched_yield(); - LOCK_HARDWARE( intel ); - } - - /* Note that any state thought to have been emitted actually - * hasn't: - */ - intel->batch.ptr -= (intel->batch.size - intel->batch.space); - intel->batch.space = intel->batch.size; - intel->vtbl.lost_hardware( intel ); - } - - if (intel->batch.space != intel->batch.size) { - - if (intel->sarea->ctxOwner != intel->hHWContext) { - intel->perf_boxes |= I830_BOX_LOST_CONTEXT; - intel->sarea->ctxOwner = intel->hHWContext; - } - - batch.start = intel->batch.start_offset; - batch.used = intel->batch.size - intel->batch.space; - batch.cliprects = intel->pClipRects; - batch.num_cliprects = ignore_cliprects ? 0 : intel->numClipRects; - batch.DR1 = 0; - batch.DR4 = ((((GLuint)intel->drawX) & 0xffff) | - (((GLuint)intel->drawY) << 16)); - - if (intel->alloc.offset) { - if ((batch.used & 0x4) == 0) { - ((int *)intel->batch.ptr)[0] = 0; - ((int *)intel->batch.ptr)[1] = MI_BATCH_BUFFER_END; - batch.used += 0x8; - intel->batch.ptr += 0x8; - } - else { - ((int *)intel->batch.ptr)[0] = MI_BATCH_BUFFER_END; - batch.used += 0x4; - intel->batch.ptr += 0x4; - } - } - - if (0) - intel_dump_batchbuffer( batch.start, - (int *)(intel->batch.ptr - batch.used), - batch.used ); - - intel->batch.start_offset += batch.used; - intel->batch.size -= batch.used; - - if (intel->batch.size < 8) { - refill = GL_TRUE; - intel->batch.space = intel->batch.size = 0; - } - else { - intel->batch.size -= 8; - intel->batch.space = intel->batch.size; - } - - - assert(intel->batch.space >= 0); - assert(batch.start >= intel->alloc.offset); - assert(batch.start < intel->alloc.offset + intel->alloc.size); - assert(batch.start + batch.used > intel->alloc.offset); - assert(batch.start + batch.used <= - intel->alloc.offset + intel->alloc.size); - - - if (intel->alloc.offset) { - if (drmCommandWrite (intel->driFd, DRM_I830_BATCHBUFFER, &batch, - sizeof(batch))) { - fprintf(stderr, "DRM_I830_BATCHBUFFER: %d\n", -errno); - UNLOCK_HARDWARE(intel); - exit(1); - } - } else { - drmI830CmdBuffer cmd; - cmd.buf = (char *)intel->alloc.ptr + batch.start; - cmd.sz = batch.used; - cmd.DR1 = batch.DR1; - cmd.DR4 = batch.DR4; - cmd.num_cliprects = batch.num_cliprects; - cmd.cliprects = batch.cliprects; - - if (drmCommandWrite (intel->driFd, DRM_I830_CMDBUFFER, &cmd, - sizeof(cmd))) { - fprintf(stderr, "DRM_I830_CMDBUFFER: %d\n", -errno); - UNLOCK_HARDWARE(intel); - exit(1); - } - } - - - age_intel(intel, intel->sarea->last_enqueue); - - /* FIXME: use hardware contexts to avoid 'losing' hardware after - * each buffer flush. - */ - if (intel->batch.contains_geometry) - assert(intel->batch.last_emit_state == intel->batch.counter); - - intel->batch.counter++; - intel->batch.contains_geometry = 0; - intel->batch.func = 0; - intel->vtbl.lost_hardware( intel ); - } - - if (refill) - intelRefillBatchLocked( intel, allow_unlock ); -} - -void intelFlushBatch( intelContextPtr intel, GLboolean refill ) -{ - if (intel->locked) { - intelFlushBatchLocked( intel, GL_FALSE, refill, GL_FALSE ); - } - else { - LOCK_HARDWARE(intel); - intelFlushBatchLocked( intel, GL_FALSE, refill, GL_TRUE ); + batch.start = start_offset; + batch.used = used; + batch.cliprects = intel->pClipRects; + batch.num_cliprects = ignore_cliprects ? 0 : intel->numClipRects; + batch.DR1 = 0; + batch.DR4 = ((((GLuint) intel->drawX) & 0xffff) | + (((GLuint) intel->drawY) << 16)); + + DBG("%s: 0x%x..0x%x DR4: %x cliprects: %d\n", + __FUNCTION__, + batch.start, + batch.start + batch.used * 4, batch.DR4, batch.num_cliprects); + + if (drmCommandWrite(intel->driFd, DRM_I830_BATCHBUFFER, &batch, + sizeof(batch))) { + fprintf(stderr, "DRM_I830_BATCHBUFFER: %d\n", -errno); UNLOCK_HARDWARE(intel); + exit(1); } -} - - -void intelWaitForIdle( intelContextPtr intel ) -{ - if (0) - fprintf(stderr, "%s\n", __FUNCTION__); - - intel->vtbl.emit_flush( intel ); - intelFlushBatch( intel, GL_TRUE ); - /* Use an irq to wait for dma idle -- Need to track lost contexts - * to shortcircuit consecutive calls to this function: + /* FIXME: use hardware contexts to avoid 'losing' hardware after + * each buffer flush. */ - intelWaitIrq( intel, intel->alloc.irq_emitted ); - intel->alloc.irq_emitted = 0; -} - - -/** - * Check if we need to rotate/warp the front color buffer to the - * rotated screen. We generally need to do this when we get a glFlush - * or glFinish after drawing to the front color buffer. - */ -static void -intelCheckFrontRotate(GLcontext *ctx) -{ - intelContextPtr intel = INTEL_CONTEXT( ctx ); - if (intel->ctx.DrawBuffer->_ColorDrawBufferMask[0] == BUFFER_BIT_FRONT_LEFT) { - intelScreenPrivate *screen = intel->intelScreen; - if (screen->current_rotation != 0) { - __DRIdrawablePrivate *dPriv = intel->driDrawable; - intelRotateWindow(intel, dPriv, BUFFER_BIT_FRONT_LEFT); - } - } -} - - -/** - * NOT directly called via glFlush. - */ -void intelFlush( GLcontext *ctx ) -{ - intelContextPtr intel = INTEL_CONTEXT( ctx ); - - if (intel->Fallback) - _swrast_flush( ctx ); - - INTEL_FIREVERTICES( intel ); - - if (intel->batch.size != intel->batch.space) - intelFlushBatch( intel, GL_FALSE ); -} - - -/** - * Called via glFlush. - */ -void intelglFlush( GLcontext *ctx ) -{ - intelFlush(ctx); - intelCheckFrontRotate(ctx); + intel->vtbl.lost_hardware(intel); } - -void intelFinish( GLcontext *ctx ) -{ - intelContextPtr intel = INTEL_CONTEXT( ctx ); - intelFlush( ctx ); - intelWaitForIdle( intel ); - intelCheckFrontRotate(ctx); -} - - -void intelClear(GLcontext *ctx, GLbitfield mask) -{ - intelContextPtr intel = INTEL_CONTEXT( ctx ); - const GLuint colorMask = *((GLuint *) &ctx->Color.ColorMask); - GLbitfield tri_mask = 0; - GLbitfield blit_mask = 0; - GLbitfield swrast_mask = 0; - - if (0) - fprintf(stderr, "%s\n", __FUNCTION__); - - /* Take care of cliprects, which are handled differently for - * clears, etc. - */ - intelFlush( &intel->ctx ); - - if (mask & BUFFER_BIT_FRONT_LEFT) { - if (colorMask == ~0) { - blit_mask |= BUFFER_BIT_FRONT_LEFT; - } - else { - tri_mask |= BUFFER_BIT_FRONT_LEFT; - } - } - - if (mask & BUFFER_BIT_BACK_LEFT) { - if (colorMask == ~0) { - blit_mask |= BUFFER_BIT_BACK_LEFT; - } - else { - tri_mask |= BUFFER_BIT_BACK_LEFT; - } - } - - if (mask & BUFFER_BIT_DEPTH) { - blit_mask |= BUFFER_BIT_DEPTH; - } - - if (mask & BUFFER_BIT_STENCIL) { - if (!intel->hw_stencil) { - swrast_mask |= BUFFER_BIT_STENCIL; - } - else if ((ctx->Stencil.WriteMask[0] & 0xff) != 0xff) { - tri_mask |= BUFFER_BIT_STENCIL; - } - else { - blit_mask |= BUFFER_BIT_STENCIL; - } - } - - swrast_mask |= (mask & BUFFER_BIT_ACCUM); - - if (blit_mask) - intelClearWithBlit( ctx, blit_mask, 0, 0, 0, 0, 0); - - if (tri_mask) - intel->vtbl.clear_with_tris( intel, tri_mask, 0, 0, 0, 0, 0); - - if (swrast_mask) - _swrast_Clear( ctx, swrast_mask ); -} - - void -intelRotateWindow(intelContextPtr intel, __DRIdrawablePrivate *dPriv, - GLuint srcBuffer) -{ - if (intel->vtbl.rotate_window) { - intel->vtbl.rotate_window(intel, dPriv, srcBuffer); - } -} - - -void *intelAllocateAGP( intelContextPtr intel, GLsizei size ) -{ - int region_offset; - drmI830MemAlloc alloc; - int ret; - - if (0) - fprintf(stderr, "%s: %d bytes\n", __FUNCTION__, size); - - alloc.region = I830_MEM_REGION_AGP; - alloc.alignment = 0; - alloc.size = size; - alloc.region_offset = ®ion_offset; - - LOCK_HARDWARE(intel); - - /* Make sure the global heap is initialized - */ - if (intel->texture_heaps[0]) - driAgeTextures( intel->texture_heaps[0] ); - - - ret = drmCommandWriteRead( intel->driFd, - DRM_I830_ALLOC, - &alloc, sizeof(alloc)); - - if (ret) { - fprintf(stderr, "%s: DRM_I830_ALLOC ret %d\n", __FUNCTION__, ret); - UNLOCK_HARDWARE(intel); - return NULL; - } - - if (0) - fprintf(stderr, "%s: allocated %d bytes\n", __FUNCTION__, size); - - /* Need to propogate this information (agp memory in use) to our - * local texture lru. The kernel has already updated the global - * lru. An alternative would have been to allocate memory the - * usual way and then notify the kernel to pin the allocation. - */ - if (intel->texture_heaps[0]) - driAgeTextures( intel->texture_heaps[0] ); - - UNLOCK_HARDWARE(intel); - - return (void *)((char *)intel->intelScreen->tex.map + region_offset); -} - -void intelFreeAGP( intelContextPtr intel, void *pointer ) +intel_exec_ioctl(struct intel_context *intel, + GLuint used, + GLboolean ignore_cliprects, GLboolean allow_unlock, + void *start, GLuint count, dri_fence **fence) { - int region_offset; - drmI830MemFree memfree; - int ret; + struct drm_i915_execbuffer execbuf; + dri_fence *fo; - region_offset = (char *)pointer - (char *)intel->intelScreen->tex.map; + assert(intel->locked); + assert(used); - if (region_offset < 0 || - region_offset > intel->intelScreen->tex.size) { - fprintf(stderr, "offset %d outside range 0..%d\n", region_offset, - intel->intelScreen->tex.size); - return; + if (*fence) { + dri_fence_unreference(*fence); } - memfree.region = I830_MEM_REGION_AGP; - memfree.region_offset = region_offset; - - ret = drmCommandWrite( intel->driFd, - DRM_I830_FREE, - &memfree, sizeof(memfree)); - - if (ret) - fprintf(stderr, "%s: DRM_I830_FREE ret %d\n", __FUNCTION__, ret); -} - -/* This version of AllocateMemoryMESA allocates only agp memory, and - * only does so after the point at which the driver has been - * initialized. - * - * Theoretically a valid context isn't required. However, in this - * implementation, it is, as I'm using the hardware lock to protect - * the kernel data structures, and the current context to get the - * device fd. - */ -void *intelAllocateMemoryMESA(__DRInativeDisplay *dpy, int scrn, - GLsizei size, GLfloat readfreq, - GLfloat writefreq, GLfloat priority) -{ - GET_CURRENT_CONTEXT(ctx); - - if (INTEL_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "%s sz %d %f/%f/%f\n", __FUNCTION__, size, readfreq, - writefreq, priority); - - if (getenv("INTEL_NO_ALLOC")) - return NULL; - - if (!ctx || INTEL_CONTEXT(ctx) == 0) - return NULL; - - return intelAllocateAGP( INTEL_CONTEXT(ctx), size ); -} - + memset(&execbuf, 0, sizeof(execbuf)); -/* Called via glXFreeMemoryMESA() */ -void intelFreeMemoryMESA(__DRInativeDisplay *dpy, int scrn, GLvoid *pointer) -{ - GET_CURRENT_CONTEXT(ctx); - if (INTEL_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "%s %p\n", __FUNCTION__, pointer); + execbuf.num_buffers = count; + execbuf.batch.used = used; + execbuf.batch.cliprects = intel->pClipRects; + execbuf.batch.num_cliprects = ignore_cliprects ? 0 : intel->numClipRects; + execbuf.batch.DR1 = 0; + execbuf.batch.DR4 = ((((GLuint) intel->drawX) & 0xffff) | + (((GLuint) intel->drawY) << 16)); - if (!ctx || INTEL_CONTEXT(ctx) == 0) { - fprintf(stderr, "%s: no context\n", __FUNCTION__); - return; - } - - intelFreeAGP( INTEL_CONTEXT(ctx), pointer ); -} + execbuf.ops_list = (unsigned)start; // TODO + execbuf.fence_arg.flags = DRM_FENCE_FLAG_SHAREABLE | DRM_I915_FENCE_FLAG_FLUSHED; -/* Called via glXGetMemoryOffsetMESA() - * - * Returns offset of pointer from the start of agp aperture. - */ -GLuint intelGetMemoryOffsetMESA(__DRInativeDisplay *dpy, int scrn, - const GLvoid *pointer) -{ - GET_CURRENT_CONTEXT(ctx); - intelContextPtr intel; - - if (!ctx || !(intel = INTEL_CONTEXT(ctx)) ) { - fprintf(stderr, "%s: no context\n", __FUNCTION__); - return ~0; + if (drmCommandWriteRead(intel->driFd, DRM_I915_EXECBUFFER, &execbuf, + sizeof(execbuf))) { + fprintf(stderr, "DRM_I830_EXECBUFFER: %d\n", -errno); + UNLOCK_HARDWARE(intel); + exit(1); } - if (!intelIsAgpMemory( intel, pointer, 0 )) - return ~0; - - return intelAgpOffsetFromVirtual( intel, pointer ); -} - - -GLboolean intelIsAgpMemory( intelContextPtr intel, const GLvoid *pointer, - GLint size ) -{ - int offset = (char *)pointer - (char *)intel->intelScreen->tex.map; - int valid = (size >= 0 && - offset >= 0 && - offset + size < intel->intelScreen->tex.size); - - if (INTEL_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "intelIsAgpMemory( %p ) : %d\n", pointer, valid ); - - return valid; -} - - -GLuint intelAgpOffsetFromVirtual( intelContextPtr intel, const GLvoid *pointer ) -{ - int offset = (char *)pointer - (char *)intel->intelScreen->tex.map; - - if (offset < 0 || offset > intel->intelScreen->tex.size) - return ~0; - else - return intel->intelScreen->tex.offset + offset; -} - - - - - -/* Flip the front & back buffes - */ -void intelPageFlip( const __DRIdrawablePrivate *dPriv ) -{ -#if 0 - intelContextPtr intel; - int tmp, ret; - - if (INTEL_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "%s\n", __FUNCTION__); - assert(dPriv); - assert(dPriv->driContextPriv); - assert(dPriv->driContextPriv->driverPrivate); - - intel = (intelContextPtr) dPriv->driContextPriv->driverPrivate; - - intelFlush( &intel->ctx ); - LOCK_HARDWARE( intel ); - - if (dPriv->pClipRects) { - *(drm_clip_rect_t *)intel->sarea->boxes = dPriv->pClipRects[0]; - intel->sarea->nbox = 1; - } - - ret = drmCommandNone(intel->driFd, DRM_I830_FLIP); - if (ret) { - fprintf(stderr, "%s: %d\n", __FUNCTION__, ret); - UNLOCK_HARDWARE( intel ); + fo = intel_ttm_fence_create_from_arg(intel->intelScreen->bufmgr, "fence buffers", + &execbuf.fence_arg); + if (!fo) { + fprintf(stderr, "failed to fence handle: %08x\n", execbuf.fence_arg.handle); + UNLOCK_HARDWARE(intel); exit(1); } + *fence = fo; - tmp = intel->sarea->last_enqueue; - intelRefillBatchLocked( intel ); - UNLOCK_HARDWARE( intel ); - + /* FIXME: use hardware contexts to avoid 'losing' hardware after + * each buffer flush. + */ + intel->vtbl.lost_hardware(intel); - intelSetDrawBuffer( &intel->ctx, intel->ctx.Color.DriverDrawBuffer ); -#endif } diff --git a/src/mesa/drivers/dri/i915/intel_ioctl.h b/src/mesa/drivers/dri/i915/intel_ioctl.h index 6ea47e462e..953fee9240 100644 --- a/src/mesa/drivers/dri/i915/intel_ioctl.h +++ b/src/mesa/drivers/dri/i915/intel_ioctl.h @@ -30,43 +30,15 @@ #include "intel_context.h" -extern void intelWaitAgeLocked( intelContextPtr intel, int age, GLboolean unlock ); - -extern void intelClear(GLcontext *ctx, GLbitfield mask); - -extern void intelPageFlip( const __DRIdrawablePrivate *dpriv ); - -extern void intelRotateWindow(intelContextPtr intel, - __DRIdrawablePrivate *dPriv, GLuint srcBuffer); - -extern void intelWaitForIdle( intelContextPtr intel ); -extern void intelFlushBatch( intelContextPtr intel, GLboolean refill ); -extern void intelFlushBatchLocked( intelContextPtr intel, - GLboolean ignore_cliprects, - GLboolean refill, - GLboolean allow_unlock); -extern void intelRefillBatchLocked( intelContextPtr intel, GLboolean allow_unlock ); -extern void intelFinish( GLcontext *ctx ); -extern void intelFlush( GLcontext *ctx ); -extern void intelglFlush( GLcontext *ctx ); - -extern void *intelAllocateAGP( intelContextPtr intel, GLsizei size ); -extern void intelFreeAGP( intelContextPtr intel, void *pointer ); - -extern void *intelAllocateMemoryMESA( __DRInativeDisplay *dpy, int scrn, - GLsizei size, GLfloat readfreq, - GLfloat writefreq, GLfloat priority ); - -extern void intelFreeMemoryMESA( __DRInativeDisplay *dpy, int scrn, - GLvoid *pointer ); - -extern GLuint intelGetMemoryOffsetMESA( __DRInativeDisplay *dpy, int scrn, const GLvoid *pointer ); -extern GLboolean intelIsAgpMemory( intelContextPtr intel, const GLvoid *pointer, - GLint size ); - -extern GLuint intelAgpOffsetFromVirtual( intelContextPtr intel, const GLvoid *p ); - -extern void intelWaitIrq( intelContextPtr intel, int seq ); -extern u_int32_t intelGetLastFrame (intelContextPtr intel); -extern int intelEmitIrqLocked( intelContextPtr intel ); +void intelWaitIrq(intelScreenPrivate *intelScreen, int seq); +int intelEmitIrqLocked(intelScreenPrivate *intelScreen); + +void intel_batch_ioctl(struct intel_context *intel, + GLuint start_offset, + GLuint used, + GLboolean ignore_cliprects, GLboolean allow_unlock); +void intel_exec_ioctl(struct intel_context *intel, + GLuint used, + GLboolean ignore_cliprects, GLboolean allow_unlock, + void *start, GLuint count, dri_fence **fence); #endif diff --git a/src/mesa/drivers/dri/i915/intel_mipmap_tree.c b/src/mesa/drivers/dri/i915/intel_mipmap_tree.c new file mode 100644 index 0000000000..2c167a9ab7 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_mipmap_tree.c @@ -0,0 +1,388 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "intel_context.h" +#include "intel_mipmap_tree.h" +#include "intel_regions.h" +#include "enums.h" + +#define FILE_DEBUG_FLAG DEBUG_MIPTREE + +static GLenum +target_to_target(GLenum target) +{ + switch (target) { + case GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB: + case GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB: + case GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB: + return GL_TEXTURE_CUBE_MAP_ARB; + default: + return target; + } +} + +struct intel_mipmap_tree * +intel_miptree_create(struct intel_context *intel, + GLenum target, + GLenum internal_format, + GLuint first_level, + GLuint last_level, + GLuint width0, + GLuint height0, + GLuint depth0, GLuint cpp, GLuint compress_byte) +{ + GLboolean ok; + struct intel_mipmap_tree *mt = calloc(sizeof(*mt), 1); + + DBG("%s target %s format %s level %d..%d\n", __FUNCTION__, + _mesa_lookup_enum_by_nr(target), + _mesa_lookup_enum_by_nr(internal_format), first_level, last_level); + + mt->target = target_to_target(target); + mt->internal_format = internal_format; + mt->first_level = first_level; + mt->last_level = last_level; + mt->width0 = width0; + mt->height0 = height0; + mt->depth0 = depth0; + mt->cpp = compress_byte ? compress_byte : cpp; + mt->compressed = compress_byte ? 1 : 0; + mt->refcount = 1; + + switch (intel->intelScreen->deviceID) { + case PCI_CHIP_I945_G: + case PCI_CHIP_I945_GM: + case PCI_CHIP_I945_GME: + case PCI_CHIP_G33_G: + case PCI_CHIP_Q33_G: + case PCI_CHIP_Q35_G: + ok = i945_miptree_layout(mt); + break; + case PCI_CHIP_I915_G: + case PCI_CHIP_I915_GM: + case PCI_CHIP_I830_M: + case PCI_CHIP_I855_GM: + case PCI_CHIP_I865_G: + default: + /* All the i830 chips and the i915 use this layout: + */ + ok = i915_miptree_layout(mt); + break; + } + + if (ok) { + if (!mt->compressed) { + int align; + + if (intel->intelScreen->ttm) { + /* XXX: Align pitch to multiple of 64 bytes for now to allow + * render-to-texture to work in all cases. This should probably be + * replaced at some point by some scheme to only do this when really + * necessary. + */ + align = 63; + } else { + align = 3; + } + + mt->pitch = (mt->pitch * cpp + align) & ~align; + + /* XXX: At least the i915 seems very upset when the pitch is a multiple + * of 1024 and sometimes 512 bytes - performance can drop by several + * times. Go to the next multiple of the required alignment for now. + */ + if (!(mt->pitch & 511)) + mt->pitch += align + 1; + + mt->pitch /= cpp; + } + + mt->region = intel_region_alloc(intel->intelScreen, + mt->cpp, mt->pitch, mt->total_height); + } + + if (!mt->region) { + free(mt); + return NULL; + } + + return mt; +} + + +void +intel_miptree_reference(struct intel_mipmap_tree **dst, + struct intel_mipmap_tree *src) +{ + src->refcount++; + *dst = src; + DBG("%s %p refcount now %d\n", __FUNCTION__, src, src->refcount); +} + +void +intel_miptree_release(struct intel_context *intel, + struct intel_mipmap_tree **mt) +{ + if (!*mt) + return; + + DBG("%s %p refcount will be %d\n", __FUNCTION__, *mt, (*mt)->refcount - 1); + if (--(*mt)->refcount <= 0) { + GLuint i; + + DBG("%s deleting %p\n", __FUNCTION__, *mt); + + intel_region_release(&((*mt)->region)); + + for (i = 0; i < MAX_TEXTURE_LEVELS; i++) + if ((*mt)->level[i].image_offset) + free((*mt)->level[i].image_offset); + + free(*mt); + } + *mt = NULL; +} + + + + +/* Can the image be pulled into a unified mipmap tree. This mirrors + * the completeness test in a lot of ways. + * + * Not sure whether I want to pass gl_texture_image here. + */ +GLboolean +intel_miptree_match_image(struct intel_mipmap_tree *mt, + struct gl_texture_image *image, + GLuint face, GLuint level) +{ + /* Images with borders are never pulled into mipmap trees. + */ + if (image->Border) + return GL_FALSE; + + if (image->InternalFormat != mt->internal_format || + image->IsCompressed != mt->compressed) + return GL_FALSE; + + /* Test image dimensions against the base level image adjusted for + * minification. This will also catch images not present in the + * tree, changed targets, etc. + */ + if (image->Width != mt->level[level].width || + image->Height != mt->level[level].height || + image->Depth != mt->level[level].depth) + return GL_FALSE; + + return GL_TRUE; +} + + +void +intel_miptree_set_level_info(struct intel_mipmap_tree *mt, + GLuint level, + GLuint nr_images, + GLuint x, GLuint y, GLuint w, GLuint h, GLuint d) +{ + + mt->level[level].width = w; + mt->level[level].height = h; + mt->level[level].depth = d; + mt->level[level].level_offset = (x + y * mt->pitch) * mt->cpp; + mt->level[level].nr_images = nr_images; + + DBG("%s level %d size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, + level, w, h, d, x, y, mt->level[level].level_offset); + + /* Not sure when this would happen, but anyway: + */ + if (mt->level[level].image_offset) { + free(mt->level[level].image_offset); + mt->level[level].image_offset = NULL; + } + + assert(nr_images); + + mt->level[level].image_offset = malloc(nr_images * sizeof(GLuint)); + mt->level[level].image_offset[0] = 0; +} + + + +void +intel_miptree_set_image_offset(struct intel_mipmap_tree *mt, + GLuint level, GLuint img, GLuint x, GLuint y) +{ + if (img == 0 && level == 0) + assert(x == 0 && y == 0); + + assert(img < mt->level[level].nr_images); + + mt->level[level].image_offset[img] = (x + y * mt->pitch); + + DBG("%s level %d img %d pos %d,%d image_offset %x\n", + __FUNCTION__, level, img, x, y, mt->level[level].image_offset[img]); +} + + +/* Although we use the image_offset[] array to store relative offsets + * to cube faces, Mesa doesn't know anything about this and expects + * each cube face to be treated as a separate image. + * + * These functions present that view to mesa: + */ +const GLuint * +intel_miptree_depth_offsets(struct intel_mipmap_tree *mt, GLuint level) +{ + static const GLuint zero = 0; + + if (mt->target != GL_TEXTURE_3D || mt->level[level].nr_images == 1) + return &zero; + else + return mt->level[level].image_offset; +} + + +GLuint +intel_miptree_image_offset(struct intel_mipmap_tree * mt, + GLuint face, GLuint level) +{ + if (mt->target == GL_TEXTURE_CUBE_MAP_ARB) + return (mt->level[level].level_offset + + mt->level[level].image_offset[face] * mt->cpp); + else + return mt->level[level].level_offset; +} + + + +/** + * Map a teximage in a mipmap tree. + * \param row_stride returns row stride in bytes + * \param image_stride returns image stride in bytes (for 3D textures). + * \return address of mapping + */ +GLubyte * +intel_miptree_image_map(struct intel_context * intel, + struct intel_mipmap_tree * mt, + GLuint face, + GLuint level, + GLuint * row_stride, GLuint * image_offsets) +{ + DBG("%s \n", __FUNCTION__); + + if (row_stride) + *row_stride = mt->pitch * mt->cpp; + + if (image_offsets) + memcpy(image_offsets, mt->level[level].image_offset, + mt->level[level].depth * sizeof(GLuint)); + + return (intel_region_map(intel->intelScreen, mt->region) + + intel_miptree_image_offset(mt, face, level)); +} + +void +intel_miptree_image_unmap(struct intel_context *intel, + struct intel_mipmap_tree *mt) +{ + DBG("%s\n", __FUNCTION__); + intel_region_unmap(intel->intelScreen, mt->region); +} + + + +/* Upload data for a particular image. + */ +void +intel_miptree_image_data(struct intel_context *intel, + struct intel_mipmap_tree *dst, + GLuint face, + GLuint level, + void *src, + GLuint src_row_pitch, GLuint src_image_pitch) +{ + GLuint depth = dst->level[level].depth; + GLuint dst_offset = intel_miptree_image_offset(dst, face, level); + const GLuint *dst_depth_offset = intel_miptree_depth_offsets(dst, level); + GLuint i; + GLuint height = 0; + + DBG("%s\n", __FUNCTION__); + for (i = 0; i < depth; i++) { + height = dst->level[level].height; + if(dst->compressed) + height /= 4; + intel_region_data(intel->intelScreen, dst->region, + dst_offset + dst_depth_offset[i], /* dst_offset */ + 0, 0, /* dstx, dsty */ + src, + src_row_pitch, + 0, 0, /* source x, y */ + dst->level[level].width, height); /* width, height */ + + src += src_image_pitch * dst->cpp; + } +} + +extern GLuint intel_compressed_alignment(GLenum); +/* Copy mipmap image between trees + */ +void +intel_miptree_image_copy(struct intel_context *intel, + struct intel_mipmap_tree *dst, + GLuint face, GLuint level, + struct intel_mipmap_tree *src) +{ + GLuint width = src->level[level].width; + GLuint height = src->level[level].height; + GLuint depth = src->level[level].depth; + GLuint dst_offset = intel_miptree_image_offset(dst, face, level); + GLuint src_offset = intel_miptree_image_offset(src, face, level); + const GLuint *dst_depth_offset = intel_miptree_depth_offsets(dst, level); + const GLuint *src_depth_offset = intel_miptree_depth_offsets(src, level); + GLuint i; + + if (dst->compressed) { + GLuint alignment = intel_compressed_alignment(dst->internal_format); + height = (height + 3) / 4; + width = ((width + alignment - 1) & ~(alignment - 1)); + } + + for (i = 0; i < depth; i++) { + intel_region_copy(intel->intelScreen, + dst->region, dst_offset + dst_depth_offset[i], + 0, + 0, + src->region, src_offset + src_depth_offset[i], + 0, 0, width, height); + } + +} diff --git a/src/mesa/drivers/dri/i915/intel_mipmap_tree.h b/src/mesa/drivers/dri/i915/intel_mipmap_tree.h new file mode 100644 index 0000000000..ecdb7be244 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_mipmap_tree.h @@ -0,0 +1,198 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_MIPMAP_TREE_H +#define INTEL_MIPMAP_TREE_H + +#include "intel_regions.h" + +/* A layer on top of the intel_regions code which adds: + * + * - Code to size and layout a region to hold a set of mipmaps. + * - Query to determine if a new image fits in an existing tree. + * - More refcounting + * - maybe able to remove refcounting from intel_region? + * - ? + * + * The fixed mipmap layout of intel hardware where one offset + * specifies the position of all images in a mipmap hierachy + * complicates the implementation of GL texture image commands, + * compared to hardware where each image is specified with an + * independent offset. + * + * In an ideal world, each texture object would be associated with a + * single bufmgr buffer or 2d intel_region, and all the images within + * the texture object would slot into the tree as they arrive. The + * reality can be a little messier, as images can arrive from the user + * with sizes that don't fit in the existing tree, or in an order + * where the tree layout cannot be guessed immediately. + * + * This structure encodes an idealized mipmap tree. The GL image + * commands build these where possible, otherwise store the images in + * temporary system buffers. + */ + + +/** + * Describes the location of each texture image within a texture region. + */ +struct intel_mipmap_level +{ + GLuint level_offset; + GLuint width; + GLuint height; + GLuint depth; + GLuint nr_images; + + /* Explicitly store the offset of each image for each cube face or + * depth value. Pretty much have to accept that hardware formats + * are going to be so diverse that there is no unified way to + * compute the offsets of depth/cube images within a mipmap level, + * so have to store them as a lookup table: + */ + GLuint *image_offset; +}; + +struct intel_mipmap_tree +{ + /* Effectively the key: + */ + GLenum target; + GLenum internal_format; + + GLuint first_level; + GLuint last_level; + + GLuint width0, height0, depth0; /**< Level zero image dimensions */ + GLuint cpp; + GLboolean compressed; + + /* Derived from the above: + */ + GLuint pitch; + GLuint depth_pitch; /* per-image on i945? */ + GLuint total_height; + + /* Includes image offset tables: + */ + struct intel_mipmap_level level[MAX_TEXTURE_LEVELS]; + + /* The data is held here: + */ + struct intel_region *region; + + /* These are also refcounted: + */ + GLuint refcount; +}; + + + +struct intel_mipmap_tree *intel_miptree_create(struct intel_context *intel, + GLenum target, + GLenum internal_format, + GLuint first_level, + GLuint last_level, + GLuint width0, + GLuint height0, + GLuint depth0, + GLuint cpp, + GLuint compress_byte); + +void intel_miptree_reference(struct intel_mipmap_tree **dst, + struct intel_mipmap_tree *src); + +void intel_miptree_release(struct intel_context *intel, + struct intel_mipmap_tree **mt); + +/* Check if an image fits an existing mipmap tree layout + */ +GLboolean intel_miptree_match_image(struct intel_mipmap_tree *mt, + struct gl_texture_image *image, + GLuint face, GLuint level); + +/* Return a pointer to an image within a tree. Return image stride as + * well. + */ +GLubyte *intel_miptree_image_map(struct intel_context *intel, + struct intel_mipmap_tree *mt, + GLuint face, + GLuint level, + GLuint * row_stride, GLuint * image_stride); + +void intel_miptree_image_unmap(struct intel_context *intel, + struct intel_mipmap_tree *mt); + + +/* Return the linear offset of an image relative to the start of the + * tree: + */ +GLuint intel_miptree_image_offset(struct intel_mipmap_tree *mt, + GLuint face, GLuint level); + +/* Return pointers to each 2d slice within an image. Indexed by depth + * value. + */ +const GLuint *intel_miptree_depth_offsets(struct intel_mipmap_tree *mt, + GLuint level); + + +void intel_miptree_set_level_info(struct intel_mipmap_tree *mt, + GLuint level, + GLuint nr_images, + GLuint x, GLuint y, + GLuint w, GLuint h, GLuint d); + +void intel_miptree_set_image_offset(struct intel_mipmap_tree *mt, + GLuint level, + GLuint img, GLuint x, GLuint y); + + +/* Upload an image into a tree + */ +void intel_miptree_image_data(struct intel_context *intel, + struct intel_mipmap_tree *dst, + GLuint face, + GLuint level, + void *src, + GLuint src_row_pitch, GLuint src_image_pitch); + +/* Copy an image between two trees + */ +void intel_miptree_image_copy(struct intel_context *intel, + struct intel_mipmap_tree *dst, + GLuint face, GLuint level, + struct intel_mipmap_tree *src); + +/* i915_mipmap_tree.c: + */ +GLboolean i915_miptree_layout(struct intel_mipmap_tree *mt); +GLboolean i945_miptree_layout(struct intel_mipmap_tree *mt); + + + +#endif diff --git a/src/mesa/drivers/dri/i915/intel_pixel.c b/src/mesa/drivers/dri/i915/intel_pixel.c index c3030d42b0..9018e3daef 100644 --- a/src/mesa/drivers/dri/i915/intel_pixel.c +++ b/src/mesa/drivers/dri/i915/intel_pixel.c @@ -1,6 +1,6 @@ /************************************************************************** * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -12,7 +12,7 @@ * the following conditions: * * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions + * next paragraph) shall be included in all copies or substantial portionsalloc * of the Software. * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS @@ -25,500 +25,96 @@ * **************************************************************************/ -#include "glheader.h" #include "enums.h" -#include "mtypes.h" -#include "macros.h" +#include "state.h" #include "swrast/swrast.h" -#include "intel_screen.h" #include "intel_context.h" -#include "intel_ioctl.h" -#include "intel_batchbuffer.h" - - - -static GLboolean -check_color( const GLcontext *ctx, GLenum type, GLenum format, - const struct gl_pixelstore_attrib *packing, - const void *pixels, GLint sz, GLint pitch ) -{ - intelContextPtr intel = INTEL_CONTEXT(ctx); - GLuint cpp = intel->intelScreen->cpp; - - if (INTEL_DEBUG & DEBUG_PIXEL) - fprintf(stderr, "%s\n", __FUNCTION__); - - if ( (pitch & 63) || - ctx->_ImageTransferState || - packing->SwapBytes || - packing->LsbFirst) { - if (INTEL_DEBUG & DEBUG_PIXEL) - fprintf(stderr, "%s: failed 1\n", __FUNCTION__); - return GL_FALSE; - } - - if ( type == GL_UNSIGNED_INT_8_8_8_8_REV && - cpp == 4 && - format == GL_BGRA ) { - if (INTEL_DEBUG & DEBUG_PIXEL) - fprintf(stderr, "%s: passed 2\n", __FUNCTION__); - return GL_TRUE; - } - - if (INTEL_DEBUG & DEBUG_PIXEL) - fprintf(stderr, "%s: failed\n", __FUNCTION__); - - return GL_FALSE; -} - -static GLboolean -check_color_per_fragment_ops( const GLcontext *ctx ) -{ - int result; - result = (!( ctx->Color.AlphaEnabled || - ctx->Depth.Test || - ctx->Fog.Enabled || - ctx->Scissor.Enabled || - ctx->Stencil.Enabled || - !ctx->Color.ColorMask[0] || - !ctx->Color.ColorMask[1] || - !ctx->Color.ColorMask[2] || - !ctx->Color.ColorMask[3] || - ctx->Color.ColorLogicOpEnabled || - ctx->Texture._EnabledUnits - ) && - ctx->Current.RasterPosValid); - - return result; -} - - -/** - * Clip the given rectangle against the buffer's bounds (including scissor). - * \param size returns the - * \return GL_TRUE if any pixels remain, GL_FALSE if totally clipped. - * - * XXX Replace this with _mesa_clip_drawpixels() and _mesa_clip_readpixels() - * from Mesa 6.4. We shouldn't apply scissor for ReadPixels. - */ -static GLboolean -clip_pixelrect( const GLcontext *ctx, - const GLframebuffer *buffer, - GLint *x, GLint *y, - GLsizei *width, GLsizei *height) -{ - /* left clipping */ - if (*x < buffer->_Xmin) { - *width -= (buffer->_Xmin - *x); - *x = buffer->_Xmin; - } - - /* right clipping */ - if (*x + *width > buffer->_Xmax) - *width -= (*x + *width - buffer->_Xmax - 1); - - if (*width <= 0) - return GL_FALSE; - - /* bottom clipping */ - if (*y < buffer->_Ymin) { - *height -= (buffer->_Ymin - *y); - *y = buffer->_Ymin; - } - - /* top clipping */ - if (*y + *height > buffer->_Ymax) - *height -= (*y + *height - buffer->_Ymax - 1); - - if (*height <= 0) - return GL_FALSE; - - return GL_TRUE; -} +#include "intel_pixel.h" +#include "intel_regions.h" /** - * Compute intersection of a clipping rectangle and pixel rectangle, - * returning results in x/y/w/hOut vars. - * \return GL_TRUE if there's intersection, GL_FALSE if disjoint. + * Check if any fragment operations are in effect which might effect + * glDraw/CopyPixels. */ -static INLINE GLboolean -intersect_region(const drm_clip_rect_t *box, - GLint x, GLint y, GLsizei width, GLsizei height, - GLint *xOut, GLint *yOut, GLint *wOut, GLint *hOut) -{ - GLint bx = box->x1; - GLint by = box->y1; - GLint bw = box->x2 - bx; - GLint bh = box->y2 - by; - - if (bx < x) bw -= x - bx, bx = x; - if (by < y) bh -= y - by, by = y; - if (bx + bw > x + width) bw = x + width - bx; - if (by + bh > y + height) bh = y + height - by; - - *xOut = bx; - *yOut = by; - *wOut = bw; - *hOut = bh; - - if (bw <= 0) return GL_FALSE; - if (bh <= 0) return GL_FALSE; - - return GL_TRUE; -} - - - -static GLboolean -intelTryReadPixels( GLcontext *ctx, - GLint x, GLint y, GLsizei width, GLsizei height, - GLenum format, GLenum type, - const struct gl_pixelstore_attrib *pack, - GLvoid *pixels ) +GLboolean +intel_check_blit_fragment_ops(GLcontext * ctx) { - intelContextPtr intel = INTEL_CONTEXT(ctx); - GLint size = 0; /* not really used */ - GLint pitch = pack->RowLength ? pack->RowLength : width; - - if (INTEL_DEBUG & DEBUG_PIXEL) - fprintf(stderr, "%s\n", __FUNCTION__); - - /* Only accelerate reading to agp buffers. - */ - if ( !intelIsAgpMemory(intel, pixels, - pitch * height * intel->intelScreen->cpp ) ) { - if (INTEL_DEBUG & DEBUG_PIXEL) - fprintf(stderr, "%s: dest not agp\n", __FUNCTION__); - return GL_FALSE; - } - - /* Need GL_PACK_INVERT_MESA to cope with upsidedown results from - * blitter: - */ - if (!pack->Invert) { - if (INTEL_DEBUG & DEBUG_PIXEL) - fprintf(stderr, "%s: MESA_PACK_INVERT not set\n", __FUNCTION__); - return GL_FALSE; - } + if (ctx->NewState) + _mesa_update_state(ctx); - if (!check_color(ctx, type, format, pack, pixels, size, pitch)) - return GL_FALSE; - - switch ( intel->intelScreen->cpp ) { - case 4: - break; - default: - return GL_FALSE; - } - - - /* Although the blits go on the command buffer, need to do this and - * fire with lock held to guarentee cliprects and drawing offset are - * correct. - * - * This is an unusual situation however, as the code which flushes - * a full command buffer expects to be called unlocked. As a - * workaround, immediately flush the buffer on aquiring the lock. + /* XXX Note: Scissor could be done with the blitter: */ - intelFlush( &intel->ctx ); - LOCK_HARDWARE( intel ); - { - __DRIdrawablePrivate *dPriv = intel->driDrawable; - int nbox = dPriv->numClipRects; - int src_offset = intel->readRegion->offset; - int src_pitch = intel->intelScreen->front.pitch; - int dst_offset = intelAgpOffsetFromVirtual( intel, pixels); - drm_clip_rect_t *box = dPriv->pClipRects; - int i; - - assert(dst_offset != ~0); /* should have been caught above */ - - if (!clip_pixelrect(ctx, ctx->ReadBuffer, &x, &y, &width, &height)) { - UNLOCK_HARDWARE( intel ); - if (INTEL_DEBUG & DEBUG_PIXEL) - fprintf(stderr, "%s totally clipped -- nothing to do\n", - __FUNCTION__); - return GL_TRUE; - } - - /* convert to screen coords (y=0=top) */ - y = dPriv->h - y - height; - x += dPriv->x; - y += dPriv->y; - - if (INTEL_DEBUG & DEBUG_PIXEL) - fprintf(stderr, "readpixel blit src_pitch %d dst_pitch %d\n", - src_pitch, pitch); - - /* We don't really have to do window clipping for readpixels. - * The OpenGL spec says that pixels read from outside the - * visible window region (pixel ownership) have undefined value. - */ - for (i = 0 ; i < nbox ; i++) - { - GLint bx, by, bw, bh; - if (intersect_region(box+i, x, y, width, height, - &bx, &by, &bw, &bh)) { - intelEmitCopyBlitLocked( intel, - intel->intelScreen->cpp, - src_pitch, src_offset, - pitch, dst_offset, - bx, by, - bx - x, by - y, - bw, bh ); - } - } - } - UNLOCK_HARDWARE( intel ); - intelFinish( &intel->ctx ); - - return GL_TRUE; -} - -static void -intelReadPixels( GLcontext *ctx, - GLint x, GLint y, GLsizei width, GLsizei height, - GLenum format, GLenum type, - const struct gl_pixelstore_attrib *pack, - GLvoid *pixels ) -{ - if (INTEL_DEBUG & DEBUG_PIXEL) - fprintf(stderr, "%s\n", __FUNCTION__); - - if (!intelTryReadPixels( ctx, x, y, width, height, format, type, pack, - pixels)) - _swrast_ReadPixels( ctx, x, y, width, height, format, type, pack, - pixels); + return !(ctx->_ImageTransferState || + ctx->Color.AlphaEnabled || + ctx->Depth.Test || + ctx->Fog.Enabled || + ctx->Scissor.Enabled || + ctx->Stencil.Enabled || + !ctx->Color.ColorMask[0] || + !ctx->Color.ColorMask[1] || + !ctx->Color.ColorMask[2] || + !ctx->Color.ColorMask[3] || + ctx->Texture._EnabledUnits || + ctx->FragmentProgram._Enabled || + ctx->Color.BlendEnabled); } - - -static void do_draw_pix( GLcontext *ctx, - GLint x, GLint y, GLsizei width, GLsizei height, - GLint pitch, - const void *pixels, - GLuint dest ) +GLboolean +intel_check_meta_tex_fragment_ops(GLcontext * ctx) { - intelContextPtr intel = INTEL_CONTEXT(ctx); - __DRIdrawablePrivate *dPriv = intel->driDrawable; - drm_clip_rect_t *box = dPriv->pClipRects; - int nbox = dPriv->numClipRects; - int i; - int src_offset = intelAgpOffsetFromVirtual( intel, pixels); - int src_pitch = pitch; - - assert(src_offset != ~0); /* should be caught earlier */ + if (ctx->NewState) + _mesa_update_state(ctx); - if (INTEL_DEBUG & DEBUG_PIXEL) - fprintf(stderr, "%s\n", __FUNCTION__); - - intelFlush( &intel->ctx ); - LOCK_HARDWARE( intel ); - if (ctx->DrawBuffer) - { - y -= height; /* cope with pixel zoom */ - - if (!clip_pixelrect(ctx, ctx->DrawBuffer, - &x, &y, &width, &height)) { - UNLOCK_HARDWARE( intel ); - return; - } - - y = dPriv->h - y - height; /* convert from gl to hardware coords */ - x += dPriv->x; - y += dPriv->y; - - for (i = 0 ; i < nbox ; i++ ) - { - GLint bx, by, bw, bh; - if (intersect_region(box + i, x, y, width, height, - &bx, &by, &bw, &bh)) { - intelEmitCopyBlitLocked( intel, - intel->intelScreen->cpp, - src_pitch, src_offset, - intel->intelScreen->front.pitch, - intel->drawRegion->offset, - bx - x, by - y, - bx, by, - bw, bh ); - } - } - } - UNLOCK_HARDWARE( intel ); - intelFinish( &intel->ctx ); + /* Some of _ImageTransferState (scale, bias) could be done with + * fragment programs on i915. + */ + return !(ctx->_ImageTransferState || ctx->Fog.Enabled || /* not done yet */ + ctx->Texture._EnabledUnits || ctx->FragmentProgram._Enabled); } - - -static GLboolean -intelTryDrawPixels( GLcontext *ctx, - GLint x, GLint y, GLsizei width, GLsizei height, - GLenum format, GLenum type, - const struct gl_pixelstore_attrib *unpack, - const GLvoid *pixels ) +/* The intel_region struct doesn't really do enough to capture the + * format of the pixels in the region. For now this code assumes that + * the region is a display surface and hence is either ARGB8888 or + * RGB565. + * XXX FBO: If we'd pass in the intel_renderbuffer instead of region, we'd + * know the buffer's pixel format. + * + * \param format as given to glDraw/ReadPixels + * \param type as given to glDraw/ReadPixels + */ +GLboolean +intel_check_blit_format(struct intel_region * region, + GLenum format, GLenum type) { - intelContextPtr intel = INTEL_CONTEXT(ctx); - GLint pitch = unpack->RowLength ? unpack->RowLength : width; - GLuint dest; - GLuint cpp = intel->intelScreen->cpp; - GLint size = width * pitch * cpp; - - if (INTEL_DEBUG & DEBUG_PIXEL) - fprintf(stderr, "%s\n", __FUNCTION__); - - switch (format) { - case GL_RGB: - case GL_RGBA: - case GL_BGRA: - dest = intel->drawRegion->offset; - - /* Planemask doesn't have full support in blits. - */ - if (!ctx->Color.ColorMask[RCOMP] || - !ctx->Color.ColorMask[GCOMP] || - !ctx->Color.ColorMask[BCOMP] || - !ctx->Color.ColorMask[ACOMP]) { - if (INTEL_DEBUG & DEBUG_PIXEL) - fprintf(stderr, "%s: planemask\n", __FUNCTION__); - return GL_FALSE; - } - - /* Can't do conversions on agp reads/draws. - */ - if ( !intelIsAgpMemory( intel, pixels, size ) ) { - if (INTEL_DEBUG & DEBUG_PIXEL) - fprintf(stderr, "%s: not agp memory\n", __FUNCTION__); - return GL_FALSE; - } - - if (!check_color(ctx, type, format, unpack, pixels, size, pitch)) { - return GL_FALSE; - } - if (!check_color_per_fragment_ops(ctx)) { - return GL_FALSE; - } - - if (ctx->Pixel.ZoomX != 1.0F || - ctx->Pixel.ZoomY != -1.0F) - return GL_FALSE; - break; - - default: - return GL_FALSE; + if (region->cpp == 4 && + (type == GL_UNSIGNED_INT_8_8_8_8_REV || + type == GL_UNSIGNED_BYTE) && format == GL_BGRA) { + return GL_TRUE; } - if ( intelIsAgpMemory(intel, pixels, size) ) - { - do_draw_pix( ctx, x, y, width, height, pitch, pixels, dest ); + if (region->cpp == 2 && + type == GL_UNSIGNED_SHORT_5_6_5_REV && format == GL_BGR) { return GL_TRUE; } - else if (0) - { - /* Pixels is in regular memory -- get dma buffers and perform - * upload through them. No point doing this for regular uploads - * but once we remove some of the restrictions above (colormask, - * pixelformat conversion, zoom?, etc), this could be a win. - */ - } - else - return GL_FALSE; - - return GL_FALSE; -} -static void -intelDrawPixels( GLcontext *ctx, - GLint x, GLint y, GLsizei width, GLsizei height, - GLenum format, GLenum type, - const struct gl_pixelstore_attrib *unpack, - const GLvoid *pixels ) -{ if (INTEL_DEBUG & DEBUG_PIXEL) - fprintf(stderr, "%s\n", __FUNCTION__); - - if (intelTryDrawPixels( ctx, x, y, width, height, format, type, - unpack, pixels )) - return; - - if (ctx->FragmentProgram._Current == ctx->FragmentProgram._TexEnvProgram) { - /* - * We don't want the i915 texenv program to be applied to DrawPixels. - * This is really just a performance optimization (mesa will other- - * wise happily run the fragment program on each pixel in the image). - */ - struct gl_fragment_program *fpSave = ctx->FragmentProgram._Current; - /* can't just set current frag prog to 0 here as on buffer resize - we'll get new state checks which will segfault. Remains a hack. */ - ctx->FragmentProgram._Current = NULL; - ctx->FragmentProgram._UseTexEnvProgram = GL_FALSE; - ctx->FragmentProgram._Active = GL_FALSE; - _swrast_DrawPixels( ctx, x, y, width, height, format, type, - unpack, pixels ); - ctx->FragmentProgram._Current = fpSave; - ctx->FragmentProgram._UseTexEnvProgram = GL_TRUE; - ctx->FragmentProgram._Active = GL_TRUE; - } - else { - _swrast_DrawPixels( ctx, x, y, width, height, format, type, - unpack, pixels ); - } -} - - - - -/** - * Implement glCopyPixels for the front color buffer (or back buffer Pixmap) - * for the color buffer. Don't support zooming, pixel transfer, etc. - * We do support copying from one window to another, ala glXMakeCurrentRead. - */ -static void -intelCopyPixels( GLcontext *ctx, - GLint srcx, GLint srcy, GLsizei width, GLsizei height, - GLint destx, GLint desty, GLenum type ) -{ -#if 0 - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - const SWcontext *swrast = SWRAST_CONTEXT( ctx ); - XMesaDisplay *dpy = xmesa->xm_visual->display; - const XMesaDrawable drawBuffer = xmesa->xm_draw_buffer->buffer; - const XMesaDrawable readBuffer = xmesa->xm_read_buffer->buffer; - const XMesaGC gc = xmesa->xm_draw_buffer->gc; - - ASSERT(dpy); - ASSERT(gc); + fprintf(stderr, "%s: bad format for blit (cpp %d, type %s format %s)\n", + __FUNCTION__, region->cpp, + _mesa_lookup_enum_by_nr(type), _mesa_lookup_enum_by_nr(format)); - if (drawBuffer && /* buffer != 0 means it's a Window or Pixmap */ - readBuffer && - type == GL_COLOR && - (swrast->_RasterMask & ~CLIP_BIT) == 0 && /* no blend, z-test, etc */ - ctx->_ImageTransferState == 0 && /* no color tables, scale/bias, etc */ - ctx->Pixel.ZoomX == 1.0 && /* no zooming */ - ctx->Pixel.ZoomY == 1.0) { - /* Note: we don't do any special clipping work here. We could, - * but X will do it for us. - */ - srcy = FLIP(xmesa->xm_read_buffer, srcy) - height + 1; - desty = FLIP(xmesa->xm_draw_buffer, desty) - height + 1; - XCopyArea(dpy, readBuffer, drawBuffer, gc, - srcx, srcy, width, height, destx, desty); - } -#else - _swrast_CopyPixels(ctx, srcx, srcy, width, height, destx, desty, type ); -#endif + return GL_FALSE; } - - -void intelInitPixelFuncs( struct dd_function_table *functions ) +void +intelInitPixelFuncs(struct dd_function_table *functions) { + functions->Accum = _swrast_Accum; + functions->Bitmap = _swrast_Bitmap; functions->CopyPixels = intelCopyPixels; - if (!getenv("INTEL_NO_BLITS")) { - functions->ReadPixels = intelReadPixels; - functions->DrawPixels = intelDrawPixels; - } + functions->ReadPixels = intelReadPixels; + functions->DrawPixels = intelDrawPixels; } diff --git a/src/mesa/drivers/dri/i915/intel_pixel.h b/src/mesa/drivers/dri/i915/intel_pixel.h new file mode 100644 index 0000000000..a6fcf90ce0 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_pixel.h @@ -0,0 +1,63 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_PIXEL_H +#define INTEL_PIXEL_H + +#include "mtypes.h" + +void intelInitPixelFuncs(struct dd_function_table *functions); + +GLboolean intel_check_blit_fragment_ops(GLcontext * ctx); + +GLboolean intel_check_meta_tex_fragment_ops(GLcontext * ctx); + +GLboolean intel_check_blit_format(struct intel_region *region, + GLenum format, GLenum type); + + +void intelReadPixels(GLcontext * ctx, + GLint x, GLint y, + GLsizei width, GLsizei height, + GLenum format, GLenum type, + const struct gl_pixelstore_attrib *pack, + GLvoid * pixels); + +void intelDrawPixels(GLcontext * ctx, + GLint x, GLint y, + GLsizei width, GLsizei height, + GLenum format, + GLenum type, + const struct gl_pixelstore_attrib *unpack, + const GLvoid * pixels); + +void intelCopyPixels(GLcontext * ctx, + GLint srcx, GLint srcy, + GLsizei width, GLsizei height, + GLint destx, GLint desty, GLenum type); + +#endif diff --git a/src/mesa/drivers/dri/i915/intel_pixel_bitmap.c b/src/mesa/drivers/dri/i915/intel_pixel_bitmap.c new file mode 100644 index 0000000000..65bf338589 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_pixel_bitmap.c @@ -0,0 +1,350 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portionsalloc + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "glheader.h" +#include "enums.h" +#include "image.h" +#include "colormac.h" +#include "mtypes.h" +#include "macros.h" +#include "bufferobj.h" +#include "swrast/swrast.h" + +#include "intel_screen.h" +#include "intel_context.h" +#include "intel_ioctl.h" +#include "intel_batchbuffer.h" +#include "intel_blit.h" +#include "intel_regions.h" +#include "intel_buffer_objects.h" + + + +#define FILE_DEBUG_FLAG DEBUG_PIXEL + + +/* Unlike the other intel_pixel_* functions, the expectation here is + * that the incoming data is not in a PBO. With the XY_TEXT blit + * method, there's no benefit haveing it in a PBO, but we could + * implement a path based on XY_MONO_SRC_COPY_BLIT which might benefit + * PBO bitmaps. I think they are probably pretty rare though - I + * wonder if Xgl uses them? + */ +static const GLubyte *map_pbo( GLcontext *ctx, + GLsizei width, GLsizei height, + const struct gl_pixelstore_attrib *unpack, + const GLubyte *bitmap ) +{ + GLubyte *buf; + + if (!_mesa_validate_pbo_access(2, unpack, width, height, 1, + GL_COLOR_INDEX, GL_BITMAP, + (GLvoid *) bitmap)) { + _mesa_error(ctx, GL_INVALID_OPERATION,"glBitmap(invalid PBO access)"); + return NULL; + } + + buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT, + GL_READ_ONLY_ARB, + unpack->BufferObj); + if (!buf) { + _mesa_error(ctx, GL_INVALID_OPERATION, "glBitmap(PBO is mapped)"); + return NULL; + } + + return ADD_POINTERS(buf, bitmap); +} + +static GLboolean test_bit( const GLubyte *src, + GLuint bit ) +{ + return (src[bit/8] & (1<<(bit % 8))) ? 1 : 0; +} + +static void set_bit( GLubyte *dest, + GLuint bit ) +{ + dest[bit/8] |= 1 << (bit % 8); +} + +static int align(int x, int align) +{ + return (x + align - 1) & ~(align - 1); +} + +/* Extract a rectangle's worth of data from the bitmap. Called + * per-cliprect. + */ +static GLuint get_bitmap_rect(GLsizei width, GLsizei height, + const struct gl_pixelstore_attrib *unpack, + const GLubyte *bitmap, + GLuint x, GLuint y, + GLuint w, GLuint h, + GLubyte *dest, + GLuint row_align, + GLboolean invert) +{ + GLuint src_offset = (x + unpack->SkipPixels) & 0x7; + GLuint mask = unpack->LsbFirst ? 0 : 7; + GLuint bit = 0; + GLint row, col; + GLint first, last; + GLint incr; + GLuint count = 0; + + if (INTEL_DEBUG & DEBUG_PIXEL) + _mesa_printf("%s %d,%d %dx%d bitmap %dx%d skip %d src_offset %d mask %d\n", + __FUNCTION__, x,y,w,h,width,height,unpack->SkipPixels, src_offset, mask); + + if (invert) { + first = h-1; + last = 0; + incr = -1; + } + else { + first = 0; + last = h-1; + incr = 1; + } + + /* Require that dest be pre-zero'd. + */ + for (row = first; row != (last+incr); row += incr) { + const GLubyte *rowsrc = _mesa_image_address2d(unpack, bitmap, + width, height, + GL_COLOR_INDEX, GL_BITMAP, + y + row, x); + + for (col = 0; col < w; col++, bit++) { + if (test_bit(rowsrc, (col + src_offset) ^ mask)) { + set_bit(dest, bit ^ 7); + count++; + } + } + + if (row_align) + bit = (bit + row_align - 1) & ~(row_align - 1); + } + + return count; +} + + + + +/* + * Render a bitmap. + */ +static GLboolean +do_blit_bitmap( GLcontext *ctx, + GLint dstx, GLint dsty, + GLsizei width, GLsizei height, + const struct gl_pixelstore_attrib *unpack, + const GLubyte *bitmap ) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_region *dst = intel_drawbuf_region(intel); + + union { + GLuint ui; + GLubyte ub[4]; + } color; + + + if (unpack->BufferObj->Name) { + bitmap = map_pbo(ctx, width, height, unpack, bitmap); + if (bitmap == NULL) + return GL_TRUE; /* even though this is an error, we're done */ + } + + UNCLAMPED_FLOAT_TO_CHAN(color.ub[0], ctx->Current.RasterColor[2]); + UNCLAMPED_FLOAT_TO_CHAN(color.ub[1], ctx->Current.RasterColor[1]); + UNCLAMPED_FLOAT_TO_CHAN(color.ub[2], ctx->Current.RasterColor[0]); + UNCLAMPED_FLOAT_TO_CHAN(color.ub[3], ctx->Current.RasterColor[3]); + + /* Does zoom apply to bitmaps? + */ + if (!intel_check_blit_fragment_ops(ctx) || + ctx->Pixel.ZoomX != 1.0F || + ctx->Pixel.ZoomY != 1.0F) + return GL_FALSE; + + LOCK_HARDWARE(intel); + + if (intel->driDrawable->numClipRects) { + __DRIdrawablePrivate *dPriv = intel->driDrawable; + drm_clip_rect_t *box = dPriv->pClipRects; + drm_clip_rect_t dest_rect; + GLint nbox = dPriv->numClipRects; + GLint srcx = 0, srcy = 0; + GLint orig_screen_x1, orig_screen_y2; + GLuint i; + + + orig_screen_x1 = dPriv->x + dstx; + orig_screen_y2 = dPriv->y + (dPriv->h - dsty); + + /* Do scissoring in GL coordinates: + */ +x if (ctx->Scissor.Enabled) + { + GLint x = ctx->Scissor.X; + GLint y = ctx->Scissor.Y; + GLuint w = ctx->Scissor.Width; + GLuint h = ctx->Scissor.Height; + + if (!_mesa_clip_to_region(x, y, x+w-1, y+h-1, &dstx, &dsty, &width, &height)) + goto out; + } + + /* Convert from GL to hardware coordinates: + */ + dsty = dPriv->y + (dPriv->h - dsty - height); + dstx = dPriv->x + dstx; + + dest_rect.x1 = dstx; + dest_rect.y1 = dsty; + dest_rect.x2 = dstx + width; + dest_rect.y2 = dsty + height; + + for (i = 0; i < nbox; i++) { + drm_clip_rect_t rect; + int box_w, box_h; + GLint px, py; + GLuint stipple[32]; + + if (!intel_intersect_cliprects(&rect, &dest_rect, &box[i])) + continue; + + /* Now go back to GL coordinates to figure out what subset of + * the bitmap we are uploading for this cliprect: + */ + box_w = rect.x2 - rect.x1; + box_h = rect.y2 - rect.y1; + srcx = rect.x1 - orig_screen_x1; + srcy = orig_screen_y2 - rect.y2; + + +#define DY 32 +#define DX 32 + + /* Then, finally, chop it all into chunks that can be + * digested by hardware: + */ + for (py = 0; py < box_h; py += DY) { + for (px = 0; px < box_w; px += DX) { + int h = MIN2(DY, box_h - py); + int w = MIN2(DX, box_w - px); + GLuint sz = align(align(w,8) * h, 64)/8; + + assert(sz <= sizeof(stipple)); + memset(stipple, 0, sz); + + /* May need to adjust this when padding has been introduced in + * sz above: + */ + if (get_bitmap_rect(width, height, unpack, + bitmap, + srcx + px, srcy + py, w, h, + (GLubyte *)stipple, + 8, + GL_TRUE) == 0) + continue; + + /* + */ + intelEmitImmediateColorExpandBlit( intel, + dst->cpp, + (GLubyte *)stipple, + sz, + color.ui, + dst->pitch, + dst->buffer, + 0, + dst->tiled, + rect.x1 + px, + rect.y2 - (py + h), + w, h); + } + } + } + intel->need_flush = GL_TRUE; + out: + intel_batchbuffer_flush(intel->batch); + } + UNLOCK_HARDWARE(intel); + + + if (unpack->BufferObj->Name) { + /* done with PBO so unmap it now */ + ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT, + unpack->BufferObj); + } + + return GL_TRUE; +} + + + + + +/* There are a large number of possible ways to implement bitmap on + * this hardware, most of them have some sort of drawback. Here are a + * few that spring to mind: + * + * Blit: + * - XY_MONO_SRC_BLT_CMD + * - use XY_SETUP_CLIP_BLT for cliprect clipping. + * - XY_TEXT_BLT + * - XY_TEXT_IMMEDIATE_BLT + * - blit per cliprect, subject to maximum immediate data size. + * - XY_COLOR_BLT + * - per pixel or run of pixels + * - XY_PIXEL_BLT + * - good for sparse bitmaps + * + * 3D engine: + * - Point per pixel + * - Translate bitmap to an alpha texture and render as a quad + * - Chop bitmap up into 32x32 squares and render w/polygon stipple. + */ +void +intelBitmap(GLcontext * ctx, + GLint x, GLint y, + GLsizei width, GLsizei height, + const struct gl_pixelstore_attrib *unpack, + const GLubyte * pixels) +{ + if (do_blit_bitmap(ctx, x, y, width, height, + unpack, pixels)) + return; + + if (INTEL_DEBUG & DEBUG_PIXEL) + _mesa_printf("%s: fallback to swrast\n", __FUNCTION__); + + _swrast_Bitmap(ctx, x, y, width, height, unpack, pixels); +} diff --git a/src/mesa/drivers/dri/i915/intel_pixel_copy.c b/src/mesa/drivers/dri/i915/intel_pixel_copy.c new file mode 100644 index 0000000000..9d478283e4 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_pixel_copy.c @@ -0,0 +1,383 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "glheader.h" +#include "enums.h" +#include "image.h" +#include "state.h" +#include "mtypes.h" +#include "macros.h" +#include "swrast/swrast.h" + +#include "intel_screen.h" +#include "intel_context.h" +#include "intel_ioctl.h" +#include "intel_batchbuffer.h" +#include "intel_buffers.h" +#include "intel_blit.h" +#include "intel_regions.h" +#include "intel_tris.h" +#include "intel_pixel.h" + +#define FILE_DEBUG_FLAG DEBUG_PIXEL + +static struct intel_region * +copypix_src_region(struct intel_context *intel, GLenum type) +{ + switch (type) { + case GL_COLOR: + return intel_readbuf_region(intel); + case GL_DEPTH: + /* Don't think this is really possible execpt at 16bpp, when we have no stencil. + */ + if (intel->intelScreen->depth_region && + intel->intelScreen->depth_region->cpp == 2) + return intel->intelScreen->depth_region; + case GL_STENCIL: + /* Don't think this is really possible. + */ + break; + case GL_DEPTH_STENCIL_EXT: + /* Does it matter whether it is stencil/depth or depth/stencil? + */ + return intel->intelScreen->depth_region; + default: + break; + } + + return NULL; +} + + +/** + * Check if any fragment operations are in effect which might effect + * glCopyPixels. Differs from intel_check_blit_fragment_ops in that + * we allow Scissor. + */ +static GLboolean +intel_check_copypixel_blit_fragment_ops(GLcontext * ctx) +{ + if (ctx->NewState) + _mesa_update_state(ctx); + + /* Could do logicop with the blitter: + */ + return !(ctx->_ImageTransferState || + ctx->Color.AlphaEnabled || + ctx->Depth.Test || + ctx->Fog.Enabled || + ctx->Stencil.Enabled || + !ctx->Color.ColorMask[0] || + !ctx->Color.ColorMask[1] || + !ctx->Color.ColorMask[2] || + !ctx->Color.ColorMask[3] || + ctx->Texture._EnabledUnits || + ctx->FragmentProgram._Enabled || + ctx->Color.BlendEnabled); +} + +/* Doesn't work for overlapping regions. Could do a double copy or + * just fallback. + */ +static GLboolean +do_texture_copypixels(GLcontext * ctx, + GLint srcx, GLint srcy, + GLsizei width, GLsizei height, + GLint dstx, GLint dsty, GLenum type) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_region *dst = intel_drawbuf_region(intel); + struct intel_region *src = copypix_src_region(intel, type); + GLenum src_format; + GLenum src_type; + + DBG("%s %d,%d %dx%d --> %d,%d\n", __FUNCTION__, + srcx, srcy, width, height, dstx, dsty); + + if (!src || !dst || type != GL_COLOR) + return GL_FALSE; + + /* Can't handle overlapping regions. Don't have sufficient control + * over rasterization to pull it off in-place. Punt on these for + * now. + * + * XXX: do a copy to a temporary. + */ + if (src->buffer == dst->buffer) { + drm_clip_rect_t srcbox; + drm_clip_rect_t dstbox; + drm_clip_rect_t tmp; + + srcbox.x1 = srcx; + srcbox.y1 = srcy; + srcbox.x2 = srcx + width; + srcbox.y2 = srcy + height; + + dstbox.x1 = dstx; + dstbox.y1 = dsty; + dstbox.x2 = dstx + width * ctx->Pixel.ZoomX; + dstbox.y2 = dsty + height * ctx->Pixel.ZoomY; + + DBG("src %d,%d %d,%d\n", srcbox.x1, srcbox.y1, srcbox.x2, srcbox.y2); + DBG("dst %d,%d %d,%d (%dx%d) (%f,%f)\n", dstbox.x1, dstbox.y1, dstbox.x2, dstbox.y2, + width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY); + + if (intel_intersect_cliprects(&tmp, &srcbox, &dstbox)) { + DBG("%s: regions overlap\n", __FUNCTION__); + return GL_FALSE; + } + } + + intelFlush(&intel->ctx); + + intel->vtbl.install_meta_state(intel); + + /* Is this true? Also will need to turn depth testing on according + * to state: + */ + intel->vtbl.meta_no_stencil_write(intel); + intel->vtbl.meta_no_depth_write(intel); + + /* Set the 3d engine to draw into the destination region: + */ + intel->vtbl.meta_draw_region(intel, dst, intel->intelScreen->depth_region); + + intel->vtbl.meta_import_pixel_state(intel); + + if (src->cpp == 2) { + src_format = GL_RGB; + src_type = GL_UNSIGNED_SHORT_5_6_5; + } + else { + src_format = GL_BGRA; + src_type = GL_UNSIGNED_BYTE; + } + + /* Set the frontbuffer up as a large rectangular texture. + */ + if (!intel->vtbl.meta_tex_rect_source(intel, src->buffer, 0, + src->pitch, + src->height, src_format, src_type)) { + intel->vtbl.leave_meta_state(intel); + return GL_FALSE; + } + + + intel->vtbl.meta_texture_blend_replace(intel); + + LOCK_HARDWARE(intel); + + if (intel->driDrawable->numClipRects) { + __DRIdrawablePrivate *dPriv = intel->driDrawable; + + + srcy = dPriv->h - srcy - height; /* convert from gl to hardware coords */ + + srcx += dPriv->x; + srcy += dPriv->y; + + /* Clip against the source region. This is the only source + * clipping we do. XXX: Just set the texcord wrap mode to clamp + * or similar. + * + */ + if (0) { + GLint orig_x = srcx; + GLint orig_y = srcy; + + if (!_mesa_clip_to_region(0, 0, src->pitch, src->height, + &srcx, &srcy, &width, &height)) + goto out; + + dstx += srcx - orig_x; + dsty += (srcy - orig_y) * ctx->Pixel.ZoomY; + } + + /* Just use the regular cliprect mechanism... Does this need to + * even hold the lock??? + */ + intel_meta_draw_quad(intel, + dstx, + dstx + width * ctx->Pixel.ZoomX, + dPriv->h - (dsty + height * ctx->Pixel.ZoomY), + dPriv->h - (dsty), 0, /* XXX: what z value? */ + 0x00ff00ff, + srcx, srcx + width, srcy, srcy + height); + + out: + intel->vtbl.leave_meta_state(intel); + intel_batchbuffer_flush(intel->batch); + } + UNLOCK_HARDWARE(intel); + + DBG("%s: success\n", __FUNCTION__); + return GL_TRUE; +} + + + + + +/** + * CopyPixels with the blitter. Don't support zooming, pixel transfer, etc. + */ +static GLboolean +do_blit_copypixels(GLcontext * ctx, + GLint srcx, GLint srcy, + GLsizei width, GLsizei height, + GLint dstx, GLint dsty, GLenum type) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_region *dst = intel_drawbuf_region(intel); + struct intel_region *src = copypix_src_region(intel, type); + + /* Copypixels can be more than a straight copy. Ensure all the + * extra operations are disabled: + */ + if (!intel_check_copypixel_blit_fragment_ops(ctx) || + ctx->Pixel.ZoomX != 1.0F || ctx->Pixel.ZoomY != 1.0F) + return GL_FALSE; + + if (!src || !dst) + return GL_FALSE; + + + + intelFlush(&intel->ctx); + + LOCK_HARDWARE(intel); + + if (intel->driDrawable->numClipRects) { + __DRIdrawablePrivate *dPriv = intel->driDrawable; + drm_clip_rect_t *box = dPriv->pClipRects; + drm_clip_rect_t dest_rect; + GLint nbox = dPriv->numClipRects; + GLint delta_x = 0; + GLint delta_y = 0; + GLuint i; + + /* Do scissoring in GL coordinates: + */ + if (ctx->Scissor.Enabled) + { + GLint x = ctx->Scissor.X; + GLint y = ctx->Scissor.Y; + GLuint w = ctx->Scissor.Width; + GLuint h = ctx->Scissor.Height; + GLint dx = dstx - srcx; + GLint dy = dsty - srcy; + + if (!_mesa_clip_to_region(x, y, x+w-1, y+h-1, &dstx, &dsty, &width, &height)) + goto out; + + srcx = dstx - dx; + srcy = dsty - dy; + } + + /* Convert from GL to hardware coordinates: + */ + dsty = dPriv->h - dsty - height; + srcy = dPriv->h - srcy - height; + dstx += dPriv->x; + dsty += dPriv->y; + srcx += dPriv->x; + srcy += dPriv->y; + + /* Clip against the source region. This is the only source + * clipping we do. Dst is clipped with cliprects below. + */ + { + delta_x = srcx - dstx; + delta_y = srcy - dsty; + + if (!_mesa_clip_to_region(0, 0, src->pitch, src->height, + &srcx, &srcy, &width, &height)) + goto out; + + dstx = srcx - delta_x; + dsty = srcy - delta_y; + } + + dest_rect.x1 = dstx; + dest_rect.y1 = dsty; + dest_rect.x2 = dstx + width; + dest_rect.y2 = dsty + height; + + /* Could do slightly more clipping: Eg, take the intersection of + * the existing set of cliprects and those cliprects translated + * by delta_x, delta_y: + * + * This code will not overwrite other windows, but will + * introduce garbage when copying from obscured window regions. + */ + for (i = 0; i < nbox; i++) { + drm_clip_rect_t rect; + + if (!intel_intersect_cliprects(&rect, &dest_rect, &box[i])) + continue; + + + intelEmitCopyBlit(intel, dst->cpp, + src->pitch, src->buffer, 0, + dst->pitch, dst->buffer, 0, + rect.x1 + delta_x, + rect.y1 + delta_y, /* srcx, srcy */ + rect.x1, rect.y1, /* dstx, dsty */ + rect.x2 - rect.x1, rect.y2 - rect.y1, + ctx->Color.ColorLogicOpEnabled ? + ctx->Color.LogicOp : GL_COPY); + } + + out: + intel_batchbuffer_flush(intel->batch); + } + UNLOCK_HARDWARE(intel); + + DBG("%s: success\n", __FUNCTION__); + return GL_TRUE; +} + + +void +intelCopyPixels(GLcontext * ctx, + GLint srcx, GLint srcy, + GLsizei width, GLsizei height, + GLint destx, GLint desty, GLenum type) +{ + if (INTEL_DEBUG & DEBUG_PIXEL) + fprintf(stderr, "%s\n", __FUNCTION__); + + if (do_blit_copypixels(ctx, srcx, srcy, width, height, destx, desty, type)) + return; + + if (do_texture_copypixels(ctx, srcx, srcy, width, height, destx, desty, type)) + return; + + DBG("fallback to _swrast_CopyPixels\n"); + + _swrast_CopyPixels(ctx, srcx, srcy, width, height, destx, desty, type); +} diff --git a/src/mesa/drivers/dri/i915/intel_pixel_draw.c b/src/mesa/drivers/dri/i915/intel_pixel_draw.c new file mode 100644 index 0000000000..afb586b4a3 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_pixel_draw.c @@ -0,0 +1,388 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portionsalloc + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "glheader.h" +#include "enums.h" +#include "image.h" +#include "mtypes.h" +#include "macros.h" +#include "bufferobj.h" +#include "swrast/swrast.h" + +#include "intel_screen.h" +#include "intel_context.h" +#include "intel_ioctl.h" +#include "intel_batchbuffer.h" +#include "intel_blit.h" +#include "intel_buffers.h" +#include "intel_regions.h" +#include "intel_pixel.h" +#include "intel_buffer_objects.h" +#include "intel_tris.h" + + + +static GLboolean +do_texture_drawpixels(GLcontext * ctx, + GLint x, GLint y, + GLsizei width, GLsizei height, + GLenum format, GLenum type, + const struct gl_pixelstore_attrib *unpack, + const GLvoid * pixels) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_region *dst = intel_drawbuf_region(intel); + struct intel_buffer_object *src = intel_buffer_object(unpack->BufferObj); + GLuint rowLength = unpack->RowLength ? unpack->RowLength : width; + GLuint src_offset; + + if (INTEL_DEBUG & DEBUG_PIXEL) + fprintf(stderr, "%s\n", __FUNCTION__); + + intelFlush(&intel->ctx); + intel->vtbl.render_start(intel); + intel->vtbl.emit_state(intel); + + if (!dst) + return GL_FALSE; + + if (src) { + if (!_mesa_validate_pbo_access(2, unpack, width, height, 1, + format, type, pixels)) { + _mesa_error(ctx, GL_INVALID_OPERATION, "glDrawPixels"); + return GL_TRUE; + } + } + else { + /* PBO only for now: + */ +/* _mesa_printf("%s - not PBO\n", __FUNCTION__); */ + return GL_FALSE; + } + + /* There are a couple of things we can't do yet, one of which is + * set the correct state for pixel operations when GL texturing is + * enabled. That's a pretty rare state and probably not worth the + * effort. A completely device-independent version of this may do + * more. + * + * Similarly, we make no attempt to merge metaops processing with + * an enabled fragment program, though it would certainly be + * possible. + */ + if (!intel_check_meta_tex_fragment_ops(ctx)) { + if (INTEL_DEBUG & DEBUG_PIXEL) + _mesa_printf("%s - bad GL fragment state for metaops texture\n", + __FUNCTION__); + return GL_FALSE; + } + + intel->vtbl.install_meta_state(intel); + + + /* Is this true? Also will need to turn depth testing on according + * to state: + */ + intel->vtbl.meta_no_stencil_write(intel); + intel->vtbl.meta_no_depth_write(intel); + + /* Set the 3d engine to draw into the destination region: + */ + intel->vtbl.meta_draw_region(intel, dst, intel->intelScreen->depth_region); + + intel->vtbl.meta_import_pixel_state(intel); + + src_offset = (GLuint) _mesa_image_address(2, unpack, pixels, width, height, + format, type, 0, 0, 0); + + + /* Setup the pbo up as a rectangular texture, if possible. + * + * TODO: This is almost always possible if the i915 fragment + * program is adjusted to correctly swizzle the sampled colors. + * The major exception is any 24bit texture, like RGB888, for which + * there is no hardware support. + */ + if (!intel->vtbl.meta_tex_rect_source(intel, src->buffer, src_offset, + rowLength, height, format, type)) { + intel->vtbl.leave_meta_state(intel); + return GL_FALSE; + } + + intel->vtbl.meta_texture_blend_replace(intel); + + + LOCK_HARDWARE(intel); + + if (intel->driDrawable->numClipRects) { + __DRIdrawablePrivate *dPriv = intel->driDrawable; + GLint srcx, srcy; + GLint dstx, dsty; + + dstx = x; + dsty = dPriv->h - (y + height); + + srcx = 0; /* skiprows/pixels already done */ + srcy = 0; + + if (0) { + const GLint orig_x = dstx; + const GLint orig_y = dsty; + + if (!_mesa_clip_to_region(0, 0, dst->pitch, dst->height, + &dstx, &dsty, &width, &height)) + goto out; + + srcx += dstx - orig_x; + srcy += dsty - orig_y; + } + + + if (INTEL_DEBUG & DEBUG_PIXEL) + _mesa_printf("draw %d,%d %dx%d\n", dstx, dsty, width, height); + + /* Must use the regular cliprect mechanism in order to get the + * drawing origin set correctly. Otherwise scissor state is in + * incorrect coordinate space. Does this even need to hold the + * lock??? + */ + intel_meta_draw_quad(intel, + dstx, dstx + width * ctx->Pixel.ZoomX, + dPriv->h - (y + height * ctx->Pixel.ZoomY), + dPriv->h - (y), + -ctx->Current.RasterPos[2] * .5, + 0x00ff00ff, + srcx, srcx + width, srcy + height, srcy); + out: + intel->vtbl.leave_meta_state(intel); + intel_batchbuffer_flush(intel->batch); + } + UNLOCK_HARDWARE(intel); + return GL_TRUE; +} + + + + + +/* Pros: + * - no waiting for idle before updating framebuffer. + * + * Cons: + * - if upload is by memcpy, this may actually be slower than fallback path. + * - uploads the whole image even if destination is clipped + * + * Need to benchmark. + * + * Given the questions about performance, implement for pbo's only. + * This path is definitely a win if the pbo is already in agp. If it + * turns out otherwise, we can add the code necessary to upload client + * data to agp space before performing the blit. (Though it may turn + * out to be better/simpler just to use the texture engine). + */ +static GLboolean +do_blit_drawpixels(GLcontext * ctx, + GLint x, GLint y, + GLsizei width, GLsizei height, + GLenum format, GLenum type, + const struct gl_pixelstore_attrib *unpack, + const GLvoid * pixels) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_region *dest = intel_drawbuf_region(intel); + struct intel_buffer_object *src = intel_buffer_object(unpack->BufferObj); + GLuint src_offset; + GLuint rowLength; + dri_fence *fence = NULL; + + if (INTEL_DEBUG & DEBUG_PIXEL) + _mesa_printf("%s\n", __FUNCTION__); + + + if (!dest) { + if (INTEL_DEBUG & DEBUG_PIXEL) + _mesa_printf("%s - no dest\n", __FUNCTION__); + return GL_FALSE; + } + + if (src) { + /* This validation should be done by core mesa: + */ + if (!_mesa_validate_pbo_access(2, unpack, width, height, 1, + format, type, pixels)) { + _mesa_error(ctx, GL_INVALID_OPERATION, "glDrawPixels"); + return GL_TRUE; + } + } + else { + /* PBO only for now: + */ + if (INTEL_DEBUG & DEBUG_PIXEL) + _mesa_printf("%s - not PBO\n", __FUNCTION__); + return GL_FALSE; + } + + if (!intel_check_blit_format(dest, format, type)) { + if (INTEL_DEBUG & DEBUG_PIXEL) + _mesa_printf("%s - bad format for blit\n", __FUNCTION__); + return GL_FALSE; + } + + if (!intel_check_blit_fragment_ops(ctx)) { + if (INTEL_DEBUG & DEBUG_PIXEL) + _mesa_printf("%s - bad GL fragment state for blitter\n", + __FUNCTION__); + return GL_FALSE; + } + + if (ctx->Pixel.ZoomX != 1.0F) { + if (INTEL_DEBUG & DEBUG_PIXEL) + _mesa_printf("%s - bad PixelZoomX for blit\n", __FUNCTION__); + return GL_FALSE; + } + + + if (unpack->RowLength > 0) + rowLength = unpack->RowLength; + else + rowLength = width; + + if (ctx->Pixel.ZoomY == -1.0F) { + if (INTEL_DEBUG & DEBUG_PIXEL) + _mesa_printf("%s - bad PixelZoomY for blit\n", __FUNCTION__); + return GL_FALSE; /* later */ + y -= height; + } + else if (ctx->Pixel.ZoomY == 1.0F) { + rowLength = -rowLength; + } + else { + if (INTEL_DEBUG & DEBUG_PIXEL) + _mesa_printf("%s - bad PixelZoomY for blit\n", __FUNCTION__); + return GL_FALSE; + } + + src_offset = (GLuint) _mesa_image_address(2, unpack, pixels, width, height, + format, type, 0, 0, 0); + + intelFlush(&intel->ctx); + LOCK_HARDWARE(intel); + + if (intel->driDrawable->numClipRects) { + __DRIdrawablePrivate *dPriv = intel->driDrawable; + int nbox = dPriv->numClipRects; + drm_clip_rect_t *box = dPriv->pClipRects; + drm_clip_rect_t rect; + drm_clip_rect_t dest_rect; + dri_bo *src_buffer = intel_bufferobj_buffer(intel, src, INTEL_READ); + int i; + + dest_rect.x1 = dPriv->x + x; + dest_rect.y1 = dPriv->y + dPriv->h - (y + height); + dest_rect.x2 = dest_rect.x1 + width; + dest_rect.y2 = dest_rect.y1 + height; + + for (i = 0; i < nbox; i++) { + if (!intel_intersect_cliprects(&rect, &dest_rect, &box[i])) + continue; + + intelEmitCopyBlit(intel, + dest->cpp, + rowLength, + src_buffer, src_offset, + dest->pitch, + dest->buffer, 0, + rect.x1 - dest_rect.x1, + rect.y2 - dest_rect.y2, + rect.x1, + rect.y1, rect.x2 - rect.x1, rect.y2 - rect.y1, + ctx->Color.ColorLogicOpEnabled ? + ctx->Color.LogicOp : GL_COPY); + } + intel_batchbuffer_flush(intel->batch); + fence = intel->batch->last_fence; + dri_fence_reference(fence); + } + UNLOCK_HARDWARE(intel); + + if (fence) { + dri_fence_wait(fence); + dri_fence_unreference(fence); + } + + if (INTEL_DEBUG & DEBUG_PIXEL) + _mesa_printf("%s - DONE\n", __FUNCTION__); + + return GL_TRUE; +} + + + +void +intelDrawPixels(GLcontext * ctx, + GLint x, GLint y, + GLsizei width, GLsizei height, + GLenum format, + GLenum type, + const struct gl_pixelstore_attrib *unpack, + const GLvoid * pixels) +{ + if (do_blit_drawpixels(ctx, x, y, width, height, format, type, + unpack, pixels)) + return; + + if (do_texture_drawpixels(ctx, x, y, width, height, format, type, + unpack, pixels)) + return; + + + if (INTEL_DEBUG & DEBUG_PIXEL) + _mesa_printf("%s: fallback to swrast\n", __FUNCTION__); + + if (ctx->FragmentProgram._Current == ctx->FragmentProgram._TexEnvProgram) { + /* + * We don't want the i915 texenv program to be applied to DrawPixels. + * This is really just a performance optimization (mesa will other- + * wise happily run the fragment program on each pixel in the image). + */ + struct gl_fragment_program *fpSave = ctx->FragmentProgram._Current; + /* can't just set current frag prog to 0 here as on buffer resize + we'll get new state checks which will segfault. Remains a hack. */ + ctx->FragmentProgram._Current = NULL; + ctx->FragmentProgram._UseTexEnvProgram = GL_FALSE; + ctx->FragmentProgram._Active = GL_FALSE; + _swrast_DrawPixels( ctx, x, y, width, height, format, type, + unpack, pixels ); + ctx->FragmentProgram._Current = fpSave; + ctx->FragmentProgram._UseTexEnvProgram = GL_TRUE; + ctx->FragmentProgram._Active = GL_TRUE; + } + else { + _swrast_DrawPixels( ctx, x, y, width, height, format, type, + unpack, pixels ); + } +} diff --git a/src/mesa/drivers/dri/i915/intel_pixel_read.c b/src/mesa/drivers/dri/i915/intel_pixel_read.c new file mode 100644 index 0000000000..a22844926c --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_pixel_read.c @@ -0,0 +1,319 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "glheader.h" +#include "enums.h" +#include "mtypes.h" +#include "macros.h" +#include "image.h" +#include "bufferobj.h" +#include "swrast/swrast.h" + +#include "intel_screen.h" +#include "intel_context.h" +#include "intel_ioctl.h" +#include "intel_batchbuffer.h" +#include "intel_blit.h" +#include "intel_buffers.h" +#include "intel_regions.h" +#include "intel_pixel.h" +#include "intel_buffer_objects.h" + +/* For many applications, the new ability to pull the source buffers + * back out of the GTT and then do the packing/conversion operations + * in software will be as much of an improvement as trying to get the + * blitter and/or texture engine to do the work. + * + * This step is gated on private backbuffers. + * + * Obviously the frontbuffer can't be pulled back, so that is either + * an argument for blit/texture readpixels, or for blitting to a + * temporary and then pulling that back. + * + * When the destination is a pbo, however, it's not clear if it is + * ever going to be pulled to main memory (though the access param + * will be a good hint). So it sounds like we do want to be able to + * choose between blit/texture implementation on the gpu and pullback + * and cpu-based copying. + * + * Unless you can magically turn client memory into a PBO for the + * duration of this call, there will be a cpu-based copying step in + * any case. + */ + + +static GLboolean +do_texture_readpixels(GLcontext * ctx, + GLint x, GLint y, GLsizei width, GLsizei height, + GLenum format, GLenum type, + const struct gl_pixelstore_attrib *pack, + struct intel_region *dest_region) +{ +#if 0 + struct intel_context *intel = intel_context(ctx); + intelScreenPrivate *screen = intel->intelScreen; + GLint pitch = pack->RowLength ? pack->RowLength : width; + __DRIdrawablePrivate *dPriv = intel->driDrawable; + int textureFormat; + GLenum glTextureFormat; + int destFormat, depthFormat, destPitch; + drm_clip_rect_t tmp; + + if (INTEL_DEBUG & DEBUG_PIXEL) + fprintf(stderr, "%s\n", __FUNCTION__); + + + if (ctx->_ImageTransferState || + pack->SwapBytes || pack->LsbFirst || !pack->Invert) { + if (INTEL_DEBUG & DEBUG_PIXEL) + fprintf(stderr, "%s: check_color failed\n", __FUNCTION__); + return GL_FALSE; + } + + intel->vtbl.meta_texrect_source(intel, intel_readbuf_region(intel)); + + if (!intel->vtbl.meta_render_dest(intel, dest_region, type, format)) { + if (INTEL_DEBUG & DEBUG_PIXEL) + fprintf(stderr, "%s: couldn't set dest %s/%s\n", + __FUNCTION__, + _mesa_lookup_enum_by_nr(type), + _mesa_lookup_enum_by_nr(format)); + return GL_FALSE; + } + + LOCK_HARDWARE(intel); + + if (intel->driDrawable->numClipRects) { + intel->vtbl.install_meta_state(intel); + intel->vtbl.meta_no_depth_write(intel); + intel->vtbl.meta_no_stencil_write(intel); + + if (!driClipRectToFramebuffer(ctx->ReadBuffer, &x, &y, &width, &height)) { + UNLOCK_HARDWARE(intel); + SET_STATE(i830, state); + if (INTEL_DEBUG & DEBUG_PIXEL) + fprintf(stderr, "%s: cliprect failed\n", __FUNCTION__); + return GL_TRUE; + } + + y = dPriv->h - y - height; + x += dPriv->x; + y += dPriv->y; + + + /* Set the frontbuffer up as a large rectangular texture. + */ + intel->vtbl.meta_tex_rect_source(intel, src_region, textureFormat); + + + intel->vtbl.meta_texture_blend_replace(i830, glTextureFormat); + + + /* Set the 3d engine to draw into the destination region: + */ + + intel->vtbl.meta_draw_region(intel, dest_region); + intel->vtbl.meta_draw_format(intel, destFormat, depthFormat); /* ?? */ + + + /* Draw a single quad, no cliprects: + */ + intel->vtbl.meta_disable_cliprects(intel); + + intel->vtbl.draw_quad(intel, + 0, width, 0, height, + 0x00ff00ff, x, x + width, y, y + height); + + intel->vtbl.leave_meta_state(intel); + } + UNLOCK_HARDWARE(intel); + + intel_region_wait_fence(ctx, dest_region); /* required by GL */ + return GL_TRUE; +#endif + + return GL_FALSE; +} + + + + +static GLboolean +do_blit_readpixels(GLcontext * ctx, + GLint x, GLint y, GLsizei width, GLsizei height, + GLenum format, GLenum type, + const struct gl_pixelstore_attrib *pack, GLvoid * pixels) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_region *src = intel_readbuf_region(intel); + struct intel_buffer_object *dst = intel_buffer_object(pack->BufferObj); + GLuint dst_offset; + GLuint rowLength; + dri_fence *fence = NULL; + + if (INTEL_DEBUG & DEBUG_PIXEL) + _mesa_printf("%s\n", __FUNCTION__); + + if (!src) + return GL_FALSE; + + if (dst) { + /* XXX This validation should be done by core mesa: + */ + if (!_mesa_validate_pbo_access(2, pack, width, height, 1, + format, type, pixels)) { + _mesa_error(ctx, GL_INVALID_OPERATION, "glDrawPixels"); + return GL_TRUE; + } + } + else { + /* PBO only for now: + */ + if (INTEL_DEBUG & DEBUG_PIXEL) + _mesa_printf("%s - not PBO\n", __FUNCTION__); + return GL_FALSE; + } + + + if (ctx->_ImageTransferState || + !intel_check_blit_format(src, format, type)) { + if (INTEL_DEBUG & DEBUG_PIXEL) + _mesa_printf("%s - bad format for blit\n", __FUNCTION__); + return GL_FALSE; + } + + if (pack->Alignment != 1 || pack->SwapBytes || pack->LsbFirst) { + if (INTEL_DEBUG & DEBUG_PIXEL) + _mesa_printf("%s: bad packing params\n", __FUNCTION__); + return GL_FALSE; + } + + if (pack->RowLength > 0) + rowLength = pack->RowLength; + else + rowLength = width; + + if (pack->Invert) { + if (INTEL_DEBUG & DEBUG_PIXEL) + _mesa_printf("%s: MESA_PACK_INVERT not done yet\n", __FUNCTION__); + return GL_FALSE; + } + else { + rowLength = -rowLength; + } + + /* XXX 64-bit cast? */ + dst_offset = (GLuint) _mesa_image_address(2, pack, pixels, width, height, + format, type, 0, 0, 0); + + + /* Although the blits go on the command buffer, need to do this and + * fire with lock held to guarentee cliprects are correct. + */ + intelFlush(&intel->ctx); + LOCK_HARDWARE(intel); + + if (intel->driDrawable->numClipRects) { + GLboolean all = (width * height * src->cpp == dst->Base.Size && + x == 0 && dst_offset == 0); + + dri_bo *dst_buffer = intel_bufferobj_buffer(intel, dst, + all ? INTEL_WRITE_FULL : + INTEL_WRITE_PART); + __DRIdrawablePrivate *dPriv = intel->driDrawable; + int nbox = dPriv->numClipRects; + drm_clip_rect_t *box = dPriv->pClipRects; + drm_clip_rect_t rect; + drm_clip_rect_t src_rect; + int i; + + src_rect.x1 = dPriv->x + x; + src_rect.y1 = dPriv->y + dPriv->h - (y + height); + src_rect.x2 = src_rect.x1 + width; + src_rect.y2 = src_rect.y1 + height; + + + + for (i = 0; i < nbox; i++) { + if (!intel_intersect_cliprects(&rect, &src_rect, &box[i])) + continue; + + intelEmitCopyBlit(intel, + src->cpp, + src->pitch, src->buffer, 0, + rowLength, + dst_buffer, dst_offset, + rect.x1, + rect.y1, + rect.x1 - src_rect.x1, + rect.y2 - src_rect.y2, + rect.x2 - rect.x1, rect.y2 - rect.y1, + GL_COPY); + } + + intel_batchbuffer_flush(intel->batch); + fence = intel->batch->last_fence; + dri_fence_reference(fence); + + } + UNLOCK_HARDWARE(intel); + + if (fence) { + dri_fence_wait(fence); + dri_fence_unreference(fence); + } + + if (INTEL_DEBUG & DEBUG_PIXEL) + _mesa_printf("%s - DONE\n", __FUNCTION__); + + return GL_TRUE; +} + +void +intelReadPixels(GLcontext * ctx, + GLint x, GLint y, GLsizei width, GLsizei height, + GLenum format, GLenum type, + const struct gl_pixelstore_attrib *pack, GLvoid * pixels) +{ + if (INTEL_DEBUG & DEBUG_PIXEL) + fprintf(stderr, "%s\n", __FUNCTION__); + + intelFlush(ctx); + + if (do_blit_readpixels + (ctx, x, y, width, height, format, type, pack, pixels)) + return; + + if (do_texture_readpixels + (ctx, x, y, width, height, format, type, pack, pixels)) + return; + + if (INTEL_DEBUG & DEBUG_PIXEL) + _mesa_printf("%s: fallback to swrast\n", __FUNCTION__); + + _swrast_ReadPixels(ctx, x, y, width, height, format, type, pack, pixels); +} diff --git a/src/mesa/drivers/dri/i915/intel_reg.h b/src/mesa/drivers/dri/i915/intel_reg.h deleted file mode 100644 index 1ec153266c..0000000000 --- a/src/mesa/drivers/dri/i915/intel_reg.h +++ /dev/null @@ -1,84 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef _INTEL_REG_H_ -#define _INTEL_REG_H_ - - - -#define CMD_3D (0x3<<29) - - -#define _3DPRIMITIVE ((0x3<<29)|(0x1f<<24)) -#define PRIM_INDIRECT (1<<23) -#define PRIM_INLINE (0<<23) -#define PRIM_INDIRECT_SEQUENTIAL (0<<17) -#define PRIM_INDIRECT_ELTS (1<<17) - -#define PRIM3D_TRILIST (0x0<<18) -#define PRIM3D_TRISTRIP (0x1<<18) -#define PRIM3D_TRISTRIP_RVRSE (0x2<<18) -#define PRIM3D_TRIFAN (0x3<<18) -#define PRIM3D_POLY (0x4<<18) -#define PRIM3D_LINELIST (0x5<<18) -#define PRIM3D_LINESTRIP (0x6<<18) -#define PRIM3D_RECTLIST (0x7<<18) -#define PRIM3D_POINTLIST (0x8<<18) -#define PRIM3D_DIB (0x9<<18) -#define PRIM3D_MASK (0x1f<<18) - -#define I915PACKCOLOR4444(r,g,b,a) \ - ((((a) & 0xf0) << 8) | (((r) & 0xf0) << 4) | ((g) & 0xf0) | ((b) >> 4)) - -#define I915PACKCOLOR1555(r,g,b,a) \ - ((((r) & 0xf8) << 7) | (((g) & 0xf8) << 2) | (((b) & 0xf8) >> 3) | \ - ((a) ? 0x8000 : 0)) - -#define I915PACKCOLOR565(r,g,b) \ - ((((r) & 0xf8) << 8) | (((g) & 0xfc) << 3) | (((b) & 0xf8) >> 3)) - -#define I915PACKCOLOR8888(r,g,b,a) \ - ((a<<24) | (r<<16) | (g<<8) | b) - - - - -#define BR00_BITBLT_CLIENT 0x40000000 -#define BR00_OP_COLOR_BLT 0x10000000 -#define BR00_OP_SRC_COPY_BLT 0x10C00000 -#define BR13_SOLID_PATTERN 0x80000000 - -#define XY_COLOR_BLT_CMD ((2<<29)|(0x50<<22)|0x4) -#define XY_COLOR_BLT_WRITE_ALPHA (1<<21) -#define XY_COLOR_BLT_WRITE_RGB (1<<20) - -#define XY_SRC_COPY_BLT_CMD ((2<<29)|(0x53<<22)|6) -#define XY_SRC_COPY_BLT_WRITE_ALPHA (1<<21) -#define XY_SRC_COPY_BLT_WRITE_RGB (1<<20) - -#endif diff --git a/src/mesa/drivers/dri/i915/intel_regions.c b/src/mesa/drivers/dri/i915/intel_regions.c new file mode 100644 index 0000000000..187ccf1776 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_regions.c @@ -0,0 +1,482 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Provide additional functionality on top of bufmgr buffers: + * - 2d semantics and blit operations + * - refcounting of buffers for multiple images in a buffer. + * - refcounting of buffer mappings. + * - some logic for moving the buffers to the best memory pools for + * given operations. + * + * Most of this is to make it easier to implement the fixed-layout + * mipmap tree required by intel hardware in the face of GL's + * programming interface where each image can be specifed in random + * order and it isn't clear what layout the tree should have until the + * last moment. + */ + +#include "intel_context.h" +#include "intel_regions.h" +#include "intel_blit.h" +#include "intel_buffer_objects.h" +#include "dri_bufmgr.h" +#include "intel_batchbuffer.h" + +#define FILE_DEBUG_FLAG DEBUG_REGION + +void +intel_region_idle(intelScreenPrivate *intelScreen, struct intel_region *region) +{ + DBG("%s\n", __FUNCTION__); + /* XXX: Using this function is likely bogus -- it ought to only have been + * used before a map, anyway, but leave this cheap implementation of it + * for now. + */ + if (region && region->buffer) { + /* Mapping it for read will ensure that any acceleration to the region + * would have landed already. + */ + dri_bo_map(region->buffer, GL_TRUE); + dri_bo_unmap(region->buffer); + } +} + +/* XXX: Thread safety? + */ +GLubyte * +intel_region_map(intelScreenPrivate *intelScreen, struct intel_region *region) +{ + DBG("%s\n", __FUNCTION__); + if (!region->map_refcount++) { + if (region->pbo) + intel_region_cow(intelScreen, region); + + dri_bo_map(region->buffer, GL_TRUE); + region->map = region->buffer->virtual; + } + + return region->map; +} + +void +intel_region_unmap(intelScreenPrivate *intelScreen, struct intel_region *region) +{ + DBG("%s\n", __FUNCTION__); + if (!--region->map_refcount) { + dri_bo_unmap(region->buffer); + region->map = NULL; + } +} + +struct intel_region * +intel_region_alloc(intelScreenPrivate *intelScreen, + GLuint cpp, GLuint pitch, GLuint height) +{ + struct intel_region *region = calloc(sizeof(*region), 1); + + DBG("%s\n", __FUNCTION__); + + region->cpp = cpp; + region->pitch = pitch; + region->height = height; /* needed? */ + region->refcount = 1; + + region->buffer = dri_bo_alloc(intelScreen->bufmgr, "region", + pitch * cpp * height, 64, DRM_BO_FLAG_MEM_TT); + return region; +} + +void +intel_region_reference(struct intel_region **dst, struct intel_region *src) +{ + assert(*dst == NULL); + if (src) { + src->refcount++; + *dst = src; + } +} + +void +intel_region_release(struct intel_region **region) +{ + if (!*region) + return; + + DBG("%s %d\n", __FUNCTION__, (*region)->refcount - 1); + + ASSERT((*region)->refcount > 0); + (*region)->refcount--; + + if ((*region)->refcount == 0) { + assert((*region)->map_refcount == 0); + + if ((*region)->pbo) + (*region)->pbo->region = NULL; + (*region)->pbo = NULL; + dri_bo_unreference((*region)->buffer); + free(*region); + } + *region = NULL; +} + + +struct intel_region * +intel_region_create_static(intelScreenPrivate *intelScreen, + GLuint mem_type, + unsigned int bo_handle, + GLuint offset, + void *virtual, + GLuint cpp, GLuint pitch, GLuint height) +{ + struct intel_region *region = calloc(sizeof(*region), 1); + DBG("%s\n", __FUNCTION__); + + region->cpp = cpp; + region->pitch = pitch; + region->height = height; /* needed? */ + region->refcount = 1; + + if (intelScreen->ttm) { + assert(bo_handle != -1); + region->buffer = intel_ttm_bo_create_from_handle(intelScreen->bufmgr, + "static region", + bo_handle); + } else { + region->buffer = dri_bo_alloc_static(intelScreen->bufmgr, + "static region", + offset, pitch * cpp * height, + virtual, + DRM_BO_FLAG_MEM_TT); + } + + return region; +} + + + +void +intel_region_update_static(intelScreenPrivate *intelScreen, + struct intel_region *region, + GLuint mem_type, + unsigned int bo_handle, + GLuint offset, + void *virtual, + GLuint cpp, GLuint pitch, GLuint height) +{ + DBG("%s\n", __FUNCTION__); + + region->cpp = cpp; + region->pitch = pitch; + region->height = height; /* needed? */ + + /* + * We use a "shared" buffer type to indicate buffers created and + * shared by others. + */ + + dri_bo_unreference(region->buffer); + if (intelScreen->ttm) { + assert(bo_handle != -1); + region->buffer = intel_ttm_bo_create_from_handle(intelScreen->bufmgr, + "static region", + bo_handle); + } else { + region->buffer = dri_bo_alloc_static(intelScreen->bufmgr, + "static region", + offset, pitch * cpp * height, + virtual, + DRM_BO_FLAG_MEM_TT); + } +} + + + +/* + * XXX Move this into core Mesa? + */ +static void +_mesa_copy_rect(GLubyte * dst, + GLuint cpp, + GLuint dst_pitch, + GLuint dst_x, + GLuint dst_y, + GLuint width, + GLuint height, + const GLubyte * src, + GLuint src_pitch, GLuint src_x, GLuint src_y) +{ + GLuint i; + + dst_pitch *= cpp; + src_pitch *= cpp; + dst += dst_x * cpp; + src += src_x * cpp; + dst += dst_y * dst_pitch; + src += src_y * dst_pitch; + width *= cpp; + + if (width == dst_pitch && width == src_pitch) + memcpy(dst, src, height * width); + else { + for (i = 0; i < height; i++) { + memcpy(dst, src, width); + dst += dst_pitch; + src += src_pitch; + } + } +} + + +/* Upload data to a rectangular sub-region. Lots of choices how to do this: + * + * - memcpy by span to current destination + * - upload data as new buffer and blit + * + * Currently always memcpy. + */ +void +intel_region_data(intelScreenPrivate *intelScreen, + struct intel_region *dst, + GLuint dst_offset, + GLuint dstx, GLuint dsty, + const void *src, GLuint src_pitch, + GLuint srcx, GLuint srcy, GLuint width, GLuint height) +{ + struct intel_context *intel = intelScreenContext(intelScreen); + + DBG("%s\n", __FUNCTION__); + + if (intel == NULL) + return; + + if (dst->pbo) { + if (dstx == 0 && + dsty == 0 && width == dst->pitch && height == dst->height) + intel_region_release_pbo(intelScreen, dst); + else + intel_region_cow(intelScreen, dst); + } + + + LOCK_HARDWARE(intel); + + _mesa_copy_rect(intel_region_map(intelScreen, dst) + dst_offset, + dst->cpp, + dst->pitch, + dstx, dsty, width, height, src, src_pitch, srcx, srcy); + + intel_region_unmap(intelScreen, dst); + + UNLOCK_HARDWARE(intel); + +} + +/* Copy rectangular sub-regions. Need better logic about when to + * push buffers into AGP - will currently do so whenever possible. + */ +void +intel_region_copy(intelScreenPrivate *intelScreen, + struct intel_region *dst, + GLuint dst_offset, + GLuint dstx, GLuint dsty, + struct intel_region *src, + GLuint src_offset, + GLuint srcx, GLuint srcy, GLuint width, GLuint height) +{ + struct intel_context *intel = intelScreenContext(intelScreen); + + DBG("%s\n", __FUNCTION__); + + if (intel == NULL) + return; + + if (dst->pbo) { + if (dstx == 0 && + dsty == 0 && width == dst->pitch && height == dst->height) + intel_region_release_pbo(intelScreen, dst); + else + intel_region_cow(intelScreen, dst); + } + + assert(src->cpp == dst->cpp); + + intelEmitCopyBlit(intel, + dst->cpp, + src->pitch, src->buffer, src_offset, + dst->pitch, dst->buffer, dst_offset, + srcx, srcy, dstx, dsty, width, height, + GL_COPY); +} + +/* Fill a rectangular sub-region. Need better logic about when to + * push buffers into AGP - will currently do so whenever possible. + */ +void +intel_region_fill(intelScreenPrivate *intelScreen, + struct intel_region *dst, + GLuint dst_offset, + GLuint dstx, GLuint dsty, + GLuint width, GLuint height, GLuint color) +{ + struct intel_context *intel = intelScreenContext(intelScreen); + + DBG("%s\n", __FUNCTION__); + + if (intel == NULL) + return; + + if (dst->pbo) { + if (dstx == 0 && + dsty == 0 && width == dst->pitch && height == dst->height) + intel_region_release_pbo(intelScreen, dst); + else + intel_region_cow(intelScreen, dst); + } + + intelEmitFillBlit(intel, + dst->cpp, + dst->pitch, dst->buffer, dst_offset, + dstx, dsty, width, height, color); +} + +/* Attach to a pbo, discarding our data. Effectively zero-copy upload + * the pbo's data. + */ +void +intel_region_attach_pbo(intelScreenPrivate *intelScreen, + struct intel_region *region, + struct intel_buffer_object *pbo) +{ + if (region->pbo == pbo) + return; + + /* If there is already a pbo attached, break the cow tie now. + * Don't call intel_region_release_pbo() as that would + * unnecessarily allocate a new buffer we would have to immediately + * discard. + */ + if (region->pbo) { + region->pbo->region = NULL; + region->pbo = NULL; + } + + if (region->buffer) { + dri_bo_unreference(region->buffer); + region->buffer = NULL; + } + + region->pbo = pbo; + region->pbo->region = region; + dri_bo_reference(pbo->buffer); + region->buffer = pbo->buffer; +} + + +/* Break the COW tie to the pbo and allocate a new buffer. + * The pbo gets to keep the data. + */ +void +intel_region_release_pbo(intelScreenPrivate *intelScreen, + struct intel_region *region) +{ + assert(region->buffer == region->pbo->buffer); + region->pbo->region = NULL; + region->pbo = NULL; + dri_bo_unreference(region->buffer); + region->buffer = NULL; + + region->buffer = dri_bo_alloc(intelScreen->bufmgr, "region", + region->pitch * region->cpp * region->height, + 64, DRM_BO_FLAG_MEM_TT); +} + +/* Break the COW tie to the pbo. Both the pbo and the region end up + * with a copy of the data. + */ +void +intel_region_cow(intelScreenPrivate *intelScreen, struct intel_region *region) +{ + struct intel_context *intel = intelScreenContext(intelScreen); + struct intel_buffer_object *pbo = region->pbo; + + if (intel == NULL) + return; + + intel_region_release_pbo(intelScreen, region); + + assert(region->cpp * region->pitch * region->height == pbo->Base.Size); + + DBG("%s (%d bytes)\n", __FUNCTION__, pbo->Base.Size); + + /* Now blit from the texture buffer to the new buffer: + */ + + intel_batchbuffer_flush(intel->batch); + + if (!intel->locked) { + LOCK_HARDWARE(intel); + intelEmitCopyBlit(intel, + region->cpp, + region->pitch, + region->buffer, 0, + region->pitch, + pbo->buffer, 0, + 0, 0, 0, 0, + region->pitch, region->height, + GL_COPY); + + intel_batchbuffer_flush(intel->batch); + UNLOCK_HARDWARE(intel); + } + else { + intelEmitCopyBlit(intel, + region->cpp, + region->pitch, + region->buffer, 0, + region->pitch, + pbo->buffer, 0, + 0, 0, 0, 0, + region->pitch, region->height, + GL_COPY); + + intel_batchbuffer_flush(intel->batch); + } +} + +dri_bo * +intel_region_buffer(intelScreenPrivate *intelScreen, + struct intel_region *region, GLuint flag) +{ + if (region->pbo) { + if (flag == INTEL_WRITE_PART) + intel_region_cow(intelScreen, region); + else if (flag == INTEL_WRITE_FULL) + intel_region_release_pbo(intelScreen, region); + } + + return region->buffer; +} diff --git a/src/mesa/drivers/dri/i915/intel_regions.h b/src/mesa/drivers/dri/i915/intel_regions.h new file mode 100644 index 0000000000..42d7b17711 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_regions.h @@ -0,0 +1,143 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_REGIONS_H +#define INTEL_REGIONS_H + +#include "mtypes.h" +#include "intel_screen.h" + +struct intel_context; +struct intel_buffer_object; + +/** + * A layer on top of the bufmgr buffers that adds a few useful things: + * + * - Refcounting for local buffer references. + * - Refcounting for buffer maps + * - Buffer dimensions - pitch and height. + * - Blitter commands for copying 2D regions between buffers. (really???) + */ +struct intel_region +{ + dri_bo *buffer; /**< buffer manager's buffer */ + GLuint refcount; /**< Reference count for region */ + GLuint cpp; /**< bytes per pixel */ + GLuint pitch; /**< in pixels */ + GLuint height; /**< in pixels */ + GLubyte *map; /**< only non-NULL when region is actually mapped */ + GLuint map_refcount; /**< Reference count for mapping */ + + GLuint draw_offset; /**< Offset of drawing address within the region */ + + struct intel_buffer_object *pbo; /* zero-copy uploads */ +}; + + +/* Allocate a refcounted region. Pointers to regions should only be + * copied by calling intel_reference_region(). + */ +struct intel_region *intel_region_alloc(intelScreenPrivate *intelScreen, + GLuint cpp, + GLuint pitch, GLuint height); + +void intel_region_reference(struct intel_region **dst, + struct intel_region *src); + +void intel_region_release(struct intel_region **ib); + +extern struct intel_region +*intel_region_create_static(intelScreenPrivate *intelScreen, + GLuint mem_type, + unsigned int bo_handle, + GLuint offset, + void *virtual, + GLuint cpp, + GLuint pitch, GLuint height); +extern void +intel_region_update_static(intelScreenPrivate *intelScreen, + struct intel_region *region, + GLuint mem_type, + unsigned int bo_handle, + GLuint offset, + void *virtual, + GLuint cpp, GLuint pitch, GLuint height); + + +void intel_region_idle(intelScreenPrivate *intelScreen, + struct intel_region *ib); + +/* Map/unmap regions. This is refcounted also: + */ +GLubyte *intel_region_map(intelScreenPrivate *intelScreen, + struct intel_region *ib); + +void intel_region_unmap(intelScreenPrivate *intelScreen, struct intel_region *ib); + + +/* Upload data to a rectangular sub-region + */ +void intel_region_data(intelScreenPrivate *intelScreen, + struct intel_region *dest, + GLuint dest_offset, + GLuint destx, GLuint desty, + const void *src, GLuint src_stride, + GLuint srcx, GLuint srcy, GLuint width, GLuint height); + +/* Copy rectangular sub-regions + */ +void intel_region_copy(intelScreenPrivate *intelScreen, + struct intel_region *dest, + GLuint dest_offset, + GLuint destx, GLuint desty, + struct intel_region *src, + GLuint src_offset, + GLuint srcx, GLuint srcy, GLuint width, GLuint height); + +/* Fill a rectangular sub-region + */ +void intel_region_fill(intelScreenPrivate *intelScreen, + struct intel_region *dest, + GLuint dest_offset, + GLuint destx, GLuint desty, + GLuint width, GLuint height, GLuint color); + +/* Helpers for zerocopy uploads, particularly texture image uploads: + */ +void intel_region_attach_pbo(intelScreenPrivate *intelScreen, + struct intel_region *region, + struct intel_buffer_object *pbo); +void intel_region_release_pbo(intelScreenPrivate *intelScreen, + struct intel_region *region); +void intel_region_cow(intelScreenPrivate *intelScreen, + struct intel_region *region); + +dri_bo *intel_region_buffer(intelScreenPrivate *intelScreen, + struct intel_region *region, + GLuint flag); + +#endif diff --git a/src/mesa/drivers/dri/i915/intel_render.c b/src/mesa/drivers/dri/i915/intel_render.c index d9438ba0fd..c8b6d308d9 100644 --- a/src/mesa/drivers/dri/i915/intel_render.c +++ b/src/mesa/drivers/dri/i915/intel_render.c @@ -51,14 +51,14 @@ * dma buffers. Use strip/fan hardware primitives where possible. * Try to simulate missing primitives with indexed vertices. */ -#define HAVE_POINTS 0 /* Has it, but can't use because subpixel has to - * be adjusted for points on the INTEL/I845G - */ +#define HAVE_POINTS 0 /* Has it, but can't use because subpixel has to + * be adjusted for points on the INTEL/I845G + */ #define HAVE_LINES 1 #define HAVE_LINE_STRIPS 1 #define HAVE_TRIANGLES 1 #define HAVE_TRI_STRIPS 1 -#define HAVE_TRI_STRIP_1 0 /* has it, template can't use it yet */ +#define HAVE_TRI_STRIP_1 0 /* has it, template can't use it yet */ #define HAVE_TRI_FANS 1 #define HAVE_POLYGONS 1 #define HAVE_QUADS 0 @@ -66,7 +66,7 @@ #define HAVE_ELTS 0 -static GLuint hw_prim[GL_POLYGON+1] = { +static GLuint hw_prim[GL_POLYGON + 1] = { 0, PRIM3D_LINELIST, PRIM3D_LINESTRIP, @@ -79,7 +79,7 @@ static GLuint hw_prim[GL_POLYGON+1] = { PRIM3D_POLY }; -static const GLenum reduced_prim[GL_POLYGON+1] = { +static const GLenum reduced_prim[GL_POLYGON + 1] = { GL_POINTS, GL_LINES, GL_LINES, @@ -92,58 +92,61 @@ static const GLenum reduced_prim[GL_POLYGON+1] = { GL_TRIANGLES }; -static const int scale_prim[GL_POLYGON+1] = { - 0, /* fallback case */ +static const int scale_prim[GL_POLYGON + 1] = { + 0, /* fallback case */ 1, 2, 2, 1, 3, 3, - 0, /* fallback case */ - 0, /* fallback case */ + 0, /* fallback case */ + 0, /* fallback case */ 3 }; -static void intelDmaPrimitive( intelContextPtr intel, GLenum prim ) +static void +intelDmaPrimitive(struct intel_context *intel, GLenum prim) { - if (0) fprintf(stderr, "%s %s\n", __FUNCTION__, _mesa_lookup_enum_by_nr(prim)); + if (0) + fprintf(stderr, "%s %s\n", __FUNCTION__, _mesa_lookup_enum_by_nr(prim)); INTEL_FIREVERTICES(intel); - intel->vtbl.reduced_primitive_state( intel, reduced_prim[prim] ); - intelStartInlinePrimitive( intel, hw_prim[prim] ); + intel->vtbl.reduced_primitive_state(intel, reduced_prim[prim]); + intelStartInlinePrimitive(intel, hw_prim[prim], INTEL_BATCH_CLIPRECTS); } -#define LOCAL_VARS intelContextPtr intel = INTEL_CONTEXT(ctx) +#define LOCAL_VARS struct intel_context *intel = intel_context(ctx) #define INIT( prim ) \ do { \ intelDmaPrimitive( intel, prim ); \ } while (0) -#define FLUSH() INTEL_FIREVERTICES( intel ) + +#define FLUSH() INTEL_FIREVERTICES(intel) #define GET_SUBSEQUENT_VB_MAX_VERTS() \ - (((intel->alloc.size / 2) - 1500) / (intel->vertex_size*4)) + ((intel->batch->size - 1500) / (intel->vertex_size*4)) #define GET_CURRENT_VB_MAX_VERTS() GET_SUBSEQUENT_VB_MAX_VERTS() #define ALLOC_VERTS( nr ) \ intelExtendInlinePrimitive( intel, (nr) * intel->vertex_size ) - + #define EMIT_VERTS( ctx, j, nr, buf ) \ - _tnl_emit_vertices_to_buffer(ctx, j, (j)+(nr), buf ) + _tnl_emit_vertices_to_buffer(ctx, j, (j)+(nr), buf ) #define TAG(x) intel_##x #include "tnl_dd/t_dd_dmatmp.h" - - + + /**********************************************************************/ /* Render pipeline stage */ /**********************************************************************/ /* Heuristic to choose between the two render paths: */ -static GLboolean choose_render( intelContextPtr intel, - struct vertex_buffer *VB ) +static GLboolean +choose_render(struct intel_context *intel, struct vertex_buffer *VB) { int vertsz = intel->vertex_size; int cost_render = 0; @@ -153,20 +156,20 @@ static GLboolean choose_render( intelContextPtr intel, int nr_rverts = 0; int rprim = intel->reduced_primitive; int i = 0; - - for (i = 0 ; i < VB->PrimitiveCount ; i++) { + + for (i = 0; i < VB->PrimitiveCount; i++) { GLuint prim = VB->Primitive[i].mode; GLuint length = VB->Primitive[i].count; if (!length) - continue; + continue; nr_prims++; nr_rverts += length * scale_prim[prim & PRIM_MODE_MASK]; if (reduced_prim[prim & PRIM_MODE_MASK] != rprim) { - nr_rprims++; - rprim = reduced_prim[prim & PRIM_MODE_MASK]; + nr_rprims++; + rprim = reduced_prim[prim & PRIM_MODE_MASK]; } } @@ -177,64 +180,65 @@ static GLboolean choose_render( intelContextPtr intel, /* One point for every 1024 dwords (4k) of dma: */ - cost_render += (vertsz * i) / 1024; - cost_fallback += (vertsz * nr_rverts) / 1024; + cost_render += (vertsz * i) / 1024; + cost_fallback += (vertsz * nr_rverts) / 1024; if (0) fprintf(stderr, "cost render: %d fallback: %d\n", - cost_render, cost_fallback); + cost_render, cost_fallback); - if (cost_render > cost_fallback) + if (cost_render > cost_fallback) return GL_FALSE; return GL_TRUE; } -static GLboolean intel_run_render( GLcontext *ctx, - struct tnl_pipeline_stage *stage ) +static GLboolean +intel_run_render(GLcontext * ctx, struct tnl_pipeline_stage *stage) { - intelContextPtr intel = INTEL_CONTEXT(ctx); + struct intel_context *intel = intel_context(ctx); TNLcontext *tnl = TNL_CONTEXT(ctx); struct vertex_buffer *VB = &tnl->vb; GLuint i; + intel->vtbl.render_prevalidate( intel ); + /* Don't handle clipping or indexed vertices. */ - if (intel->RenderIndex != 0 || - !intel_validate_render( ctx, VB ) || - !choose_render( intel, VB )) { + if (intel->RenderIndex != 0 || + !intel_validate_render(ctx, VB) || !choose_render(intel, VB)) { return GL_TRUE; } tnl->clipspace.new_inputs |= VERT_BIT_POS; - tnl->Driver.Render.Start( ctx ); - - for (i = 0 ; i < VB->PrimitiveCount ; i++) - { + tnl->Driver.Render.Start(ctx); + + for (i = 0; i < VB->PrimitiveCount; i++) { GLuint prim = VB->Primitive[i].mode; GLuint start = VB->Primitive[i].start; GLuint length = VB->Primitive[i].count; if (!length) - continue; + continue; - intel_render_tab_verts[prim & PRIM_MODE_MASK]( ctx, start, start + length, - prim ); + intel_render_tab_verts[prim & PRIM_MODE_MASK] (ctx, start, + start + length, prim); } - - tnl->Driver.Render.Finish( ctx ); - return GL_FALSE; /* finished the pipe */ + tnl->Driver.Render.Finish(ctx); + + INTEL_FIREVERTICES(intel); + + return GL_FALSE; /* finished the pipe */ } -const struct tnl_pipeline_stage _intel_render_stage = -{ +const struct tnl_pipeline_stage _intel_render_stage = { "intel render", NULL, NULL, NULL, NULL, - intel_run_render /* run */ + intel_run_render /* run */ }; diff --git a/src/mesa/drivers/dri/i915/intel_rotate.c b/src/mesa/drivers/dri/i915/intel_rotate.c index a77640ee54..12d98c4ad2 100644 --- a/src/mesa/drivers/dri/i915/intel_rotate.c +++ b/src/mesa/drivers/dri/i915/intel_rotate.c @@ -15,11 +15,14 @@ void matrix23Set(struct matrix23 *m, - int m00, int m01, int m02, - int m10, int m11, int m12) + int m00, int m01, int m02, int m10, int m11, int m12) { - m->m00 = m00; m->m01 = m01; m->m02 = m02; - m->m10 = m10; m->m11 = m11; m->m12 = m12; + m->m00 = m00; + m->m01 = m01; + m->m02 = m02; + m->m10 = m10; + m->m11 = m11; + m->m12 = m12; } @@ -66,9 +69,9 @@ matrix23TransformDistance(const struct matrix23 *m, int *xDist, int *yDist) *yDist = (y1 - y0) + (y2 - y0); if (*xDist < 0) - *xDist = -*xDist; + *xDist = -*xDist; if (*yDist < 0) - *yDist = -*yDist; + *yDist = -*yDist; } @@ -76,7 +79,8 @@ matrix23TransformDistance(const struct matrix23 *m, int *xDist, int *yDist) * Transform the rect defined by (x, y, w, h) by m. */ void -matrix23TransformRect(const struct matrix23 *m, int *x, int *y, int *w, int *h) +matrix23TransformRect(const struct matrix23 *m, int *x, int *y, int *w, + int *h) { int x0 = *x, y0 = *y; int x1 = *x + *w, y1 = *y; @@ -108,16 +112,16 @@ matrix23Rotate(struct matrix23 *m, int width, int height, int angle) matrix23Set(m, 1, 0, 0, 0, 1, 0); break; case 90: - matrix23Set(m, 0, 1, 0, -1, 0, width); + matrix23Set(m, 0, 1, 0, -1, 0, width); break; case 180: - matrix23Set(m, -1, 0, width, 0, -1, height); + matrix23Set(m, -1, 0, width, 0, -1, height); break; case 270: - matrix23Set(m, 0, -1, height, 1, 0, 0); + matrix23Set(m, 0, -1, height, 1, 0, 0); break; default: - /*abort()*/; + /*abort() */ ; } } @@ -129,16 +133,24 @@ void matrix23Flip(struct matrix23 *m, int width, int height, int xflip, int yflip) { if (xflip) { - m->m00 = -1; m->m01 = 0; m->m02 = width - 1; + m->m00 = -1; + m->m01 = 0; + m->m02 = width - 1; } else { - m->m00 = 1; m->m01 = 0; m->m02 = 0; + m->m00 = 1; + m->m01 = 0; + m->m02 = 0; } if (yflip) { - m->m10 = 0; m->m11 = -1; m->m12 = height - 1; + m->m10 = 0; + m->m11 = -1; + m->m12 = height - 1; } else { - m->m10 = 0; m->m11 = 1; m->m12 = 0; + m->m10 = 0; + m->m11 = 1; + m->m12 = 0; } } @@ -169,14 +181,18 @@ main(int argc, char *argv[]) { int width = 500, height = 400; int rot; - int fx = 0, fy = 0; /* flip x and/or y ? */ + int fx = 0, fy = 0; /* flip x and/or y ? */ int coords[4][2]; /* four corner coords to test with */ - coords[0][0] = 0; coords[0][1] = 0; - coords[1][0] = width-1; coords[1][1] = 0; - coords[2][0] = width-1; coords[2][1] = height-1; - coords[3][0] = 0; coords[3][1] = height-1; + coords[0][0] = 0; + coords[0][1] = 0; + coords[1][0] = width - 1; + coords[1][1] = 0; + coords[2][0] = width - 1; + coords[2][1] = height - 1; + coords[3][0] = 0; + coords[3][1] = height - 1; for (rot = 0; rot < 360; rot += 90) { diff --git a/src/mesa/drivers/dri/i915/intel_rotate.h b/src/mesa/drivers/dri/i915/intel_rotate.h index 0da45d20ce..9c8802ca47 100644 --- a/src/mesa/drivers/dri/i915/intel_rotate.h +++ b/src/mesa/drivers/dri/i915/intel_rotate.h @@ -11,11 +11,9 @@ struct matrix23 extern void matrix23Set(struct matrix23 *m, - int m00, int m01, int m02, - int m10, int m11, int m12); + int m00, int m01, int m02, int m10, int m11, int m12); -extern void -matrix23TransformCoordi(const struct matrix23 *m, int *x, int *y); +extern void matrix23TransformCoordi(const struct matrix23 *m, int *x, int *y); extern void matrix23TransformCoordf(const struct matrix23 *m, float *x, float *y); diff --git a/src/mesa/drivers/dri/i915/intel_screen.c b/src/mesa/drivers/dri/i915/intel_screen.c index ca8610b496..8be5d910a0 100644 --- a/src/mesa/drivers/dri/i915/intel_screen.c +++ b/src/mesa/drivers/dri/i915/intel_screen.c @@ -38,46 +38,51 @@ #include "intel_screen.h" +#include "intel_buffers.h" #include "intel_tex.h" #include "intel_span.h" #include "intel_tris.h" #include "intel_ioctl.h" +#include "intel_fbo.h" #include "i830_dri.h" +#include "dri_bufmgr.h" +#include "intel_regions.h" +#include "intel_batchbuffer.h" + +#include "intel_bufmgr_ttm.h" PUBLIC const char __driConfigOptions[] = -DRI_CONF_BEGIN - DRI_CONF_SECTION_PERFORMANCE - DRI_CONF_FTHROTTLE_MODE(DRI_CONF_FTHROTTLE_IRQS) - DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_DEF_INTERVAL_0) - DRI_CONF_SECTION_END - DRI_CONF_SECTION_QUALITY - DRI_CONF_FORCE_S3TC_ENABLE(false) - DRI_CONF_ALLOW_LARGE_TEXTURES(1) - DRI_CONF_SECTION_END -DRI_CONF_END; -const GLuint __driNConfigOptions = 4; + DRI_CONF_BEGIN DRI_CONF_SECTION_PERFORMANCE + DRI_CONF_FTHROTTLE_MODE(DRI_CONF_FTHROTTLE_IRQS) + DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_DEF_INTERVAL_0) + DRI_CONF_SECTION_END DRI_CONF_SECTION_QUALITY + DRI_CONF_FORCE_S3TC_ENABLE(false) + DRI_CONF_ALLOW_LARGE_TEXTURES(1) + DRI_CONF_SECTION_END DRI_CONF_END; + const GLuint __driNConfigOptions = 4; #ifdef USE_NEW_INTERFACE -static PFNGLXCREATECONTEXTMODES create_context_modes = NULL; -#endif /*USE_NEW_INTERFACE*/ + static PFNGLXCREATECONTEXTMODES create_context_modes = NULL; +#endif /*USE_NEW_INTERFACE */ -extern const struct dri_extension card_extensions[]; + extern const struct dri_extension card_extensions[]; + extern const struct dri_extension ttm_extensions[]; /** * Map all the memory regions described by the screen. * \return GL_TRUE if success, GL_FALSE if error. */ GLboolean -intelMapScreenRegions(__DRIscreenPrivate *sPriv) +intelMapScreenRegions(__DRIscreenPrivate * sPriv) { - intelScreenPrivate *intelScreen = (intelScreenPrivate *)sPriv->private; + intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; if (intelScreen->front.handle) { if (drmMap(sPriv->fd, intelScreen->front.handle, intelScreen->front.size, - (drmAddress *)&intelScreen->front.map) != 0) { + (drmAddress *) & intelScreen->front.map) != 0) { _mesa_problem(NULL, "drmMap(frontbuffer) failed!"); return GL_FALSE; } @@ -86,42 +91,187 @@ intelMapScreenRegions(__DRIscreenPrivate *sPriv) _mesa_warning(NULL, "no front buffer handle in intelMapScreenRegions!"); } + if (0) + _mesa_printf("Back 0x%08x ", intelScreen->back.handle); if (drmMap(sPriv->fd, intelScreen->back.handle, intelScreen->back.size, - (drmAddress *)&intelScreen->back.map) != 0) { + (drmAddress *) & intelScreen->back.map) != 0) { intelUnmapScreenRegions(intelScreen); return GL_FALSE; } + if (intelScreen->third.handle) { + if (0) + _mesa_printf("Third 0x%08x ", intelScreen->third.handle); + if (drmMap(sPriv->fd, + intelScreen->third.handle, + intelScreen->third.size, + (drmAddress *) & intelScreen->third.map) != 0) { + intelUnmapScreenRegions(intelScreen); + return GL_FALSE; + } + } + + if (0) + _mesa_printf("Depth 0x%08x ", intelScreen->depth.handle); if (drmMap(sPriv->fd, intelScreen->depth.handle, intelScreen->depth.size, - (drmAddress *)&intelScreen->depth.map) != 0) { + (drmAddress *) & intelScreen->depth.map) != 0) { intelUnmapScreenRegions(intelScreen); return GL_FALSE; } - if (drmMap(sPriv->fd, - intelScreen->tex.handle, - intelScreen->tex.size, - (drmAddress *)&intelScreen->tex.map) != 0) { - intelUnmapScreenRegions(intelScreen); - return GL_FALSE; + if (0) + _mesa_printf("TEX 0x%08x ", intelScreen->tex.handle); + if (intelScreen->tex.size != 0) { + if (drmMap(sPriv->fd, + intelScreen->tex.handle, + intelScreen->tex.size, + (drmAddress *) & intelScreen->tex.map) != 0) { + intelUnmapScreenRegions(intelScreen); + return GL_FALSE; + } } if (0) - printf("Mappings: front: %p back: %p depth: %p tex: %p\n", - intelScreen->front.map, - intelScreen->back.map, - intelScreen->depth.map, - intelScreen->tex.map); + printf("Mappings: front: %p back: %p third: %p depth: %p tex: %p\n", + intelScreen->front.map, + intelScreen->back.map, intelScreen->third.map, + intelScreen->depth.map, intelScreen->tex.map); return GL_TRUE; } +/** Driver-specific fence emit implementation for the fake memory manager. */ +static unsigned int +intel_fence_emit(void *private) +{ + intelScreenPrivate *intelScreen = (intelScreenPrivate *)private; + unsigned int fence; + + /* XXX: Need to emit a flush, if we haven't already (at least with the + * current batchbuffer implementation, we have). + */ + + fence = intelEmitIrqLocked(intelScreen); + + return fence; +} + +/** Driver-specific fence wait implementation for the fake memory manager. */ +static int +intel_fence_wait(void *private, unsigned int cookie) +{ + intelScreenPrivate *intelScreen = (intelScreenPrivate *)private; + + intelWaitIrq(intelScreen, cookie); + + return 0; +} + +static struct intel_region * +intel_recreate_static(intelScreenPrivate *intelScreen, + struct intel_region *region, + intelRegion *region_desc, + GLuint mem_type) +{ + if (region) { + intel_region_update_static(intelScreen, region, mem_type, + region_desc->bo_handle, region_desc->offset, + region_desc->map, intelScreen->cpp, + region_desc->pitch / intelScreen->cpp, + intelScreen->height); + } else { + region = intel_region_create_static(intelScreen, mem_type, + region_desc->bo_handle, + region_desc->offset, + region_desc->map, intelScreen->cpp, + region_desc->pitch / intelScreen->cpp, + intelScreen->height); + } + + assert(region->buffer != NULL); + + return region; +} + + +/* Create intel_region structs to describe the static front,back,depth + * buffers created by the xserver. + * + * Although FBO's mean we now no longer use these as render targets in + * all circumstances, they won't go away until the back and depth + * buffers become private, and the front and rotated buffers will + * remain even then. + * + * Note that these don't allocate video memory, just describe + * allocations alread made by the X server. + */ +static void +intel_recreate_static_regions(intelScreenPrivate *intelScreen) +{ + intelScreen->front_region = + intel_recreate_static(intelScreen, + intelScreen->front_region, + &intelScreen->front, + DRM_BO_FLAG_MEM_TT); + + /* The rotated region is only used for old DDXes that didn't handle rotation +\ * on their own. + */ + if (intelScreen->driScrnPriv->ddx_version.minor < 8) { + intelScreen->rotated_region = + intel_recreate_static(intelScreen, + intelScreen->rotated_region, + &intelScreen->rotated, + DRM_BO_FLAG_MEM_TT); + } + + intelScreen->back_region = + intel_recreate_static(intelScreen, + intelScreen->back_region, + &intelScreen->back, + DRM_BO_FLAG_MEM_TT); + + if (intelScreen->third.handle) { + intelScreen->third_region = + intel_recreate_static(intelScreen, + intelScreen->third_region, + &intelScreen->third, + DRM_BO_FLAG_MEM_TT); + } + + /* Still assumes front.cpp == depth.cpp. We can kill this when we move to + * private buffers. + */ + intelScreen->depth_region = + intel_recreate_static(intelScreen, + intelScreen->depth_region, + &intelScreen->depth, + DRM_BO_FLAG_MEM_TT); +} + +/** + * Use the information in the sarea to update the screen parameters + * related to screen rotation. Needs to be called locked. + */ +void +intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea) +{ + intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; + + intelUnmapScreenRegions(intelScreen); + intelUpdateScreenFromSAREA(intelScreen, sarea); + if (!intelMapScreenRegions(sPriv)) { + fprintf(stderr, "ERROR Remapping screen regions!!!\n"); + } + intel_recreate_static_regions(intelScreen); +} + void -intelUnmapScreenRegions(intelScreenPrivate *intelScreen) +intelUnmapScreenRegions(intelScreenPrivate * intelScreen) { #define REALLY_UNMAP 1 if (intelScreen->front.map) { @@ -138,6 +288,13 @@ intelUnmapScreenRegions(intelScreenPrivate *intelScreen) #endif intelScreen->back.map = NULL; } + if (intelScreen->third.map) { +#if REALLY_UNMAP + if (drmUnmap(intelScreen->third.map, intelScreen->third.size) != 0) + printf("drmUnmap third failed!\n"); +#endif + intelScreen->third.map = NULL; + } if (intelScreen->depth.map) { #if REALLY_UNMAP drmUnmap(intelScreen->depth.map, intelScreen->depth.size); @@ -154,9 +311,8 @@ intelUnmapScreenRegions(intelScreenPrivate *intelScreen) static void -intelPrintDRIInfo(intelScreenPrivate *intelScreen, - __DRIscreenPrivate *sPriv, - I830DRIPtr gDRIPriv) +intelPrintDRIInfo(intelScreenPrivate * intelScreen, + __DRIscreenPrivate * sPriv, I830DRIPtr gDRIPriv) { fprintf(stderr, "*** Front size: 0x%x offset: 0x%x pitch: %d\n", intelScreen->front.size, intelScreen->front.offset, @@ -177,9 +333,10 @@ intelPrintDRIInfo(intelScreenPrivate *intelScreen, static void -intelPrintSAREA(const drmI830Sarea *sarea) +intelPrintSAREA(const drmI830Sarea * sarea) { - fprintf(stderr, "SAREA: sarea width %d height %d\n", sarea->width, sarea->height); + fprintf(stderr, "SAREA: sarea width %d height %d\n", sarea->width, + sarea->height); fprintf(stderr, "SAREA: pitch: %d\n", sarea->pitch); fprintf(stderr, "SAREA: front offset: 0x%08x size: 0x%x handle: 0x%x\n", @@ -193,8 +350,7 @@ intelPrintSAREA(const drmI830Sarea *sarea) sarea->depth_offset, sarea->depth_size, (unsigned) sarea->depth_handle); fprintf(stderr, "SAREA: tex offset: 0x%08x size: 0x%x handle: 0x%x\n", - sarea->tex_offset, sarea->tex_size, - (unsigned) sarea->tex_handle); + sarea->tex_offset, sarea->tex_size, (unsigned) sarea->tex_handle); fprintf(stderr, "SAREA: rotation: %d\n", sarea->rotation); fprintf(stderr, "SAREA: rotated offset: 0x%08x size: 0x%x\n", @@ -208,8 +364,8 @@ intelPrintSAREA(const drmI830Sarea *sarea) * information in the SAREA. This function updates those parameters. */ void -intelUpdateScreenFromSAREA(intelScreenPrivate *intelScreen, - drmI830Sarea *sarea) +intelUpdateScreenFromSAREA(intelScreenPrivate * intelScreen, + drmI830Sarea * sarea) { intelScreen->width = sarea->width; intelScreen->height = sarea->height; @@ -223,12 +379,31 @@ intelUpdateScreenFromSAREA(intelScreenPrivate *intelScreen, intelScreen->back.pitch = sarea->pitch * intelScreen->cpp; intelScreen->back.handle = sarea->back_handle; intelScreen->back.size = sarea->back_size; - + + if (intelScreen->driScrnPriv->ddx_version.minor >= 8) { + intelScreen->third.offset = sarea->third_offset; + intelScreen->third.pitch = sarea->pitch * intelScreen->cpp; + intelScreen->third.handle = sarea->third_handle; + intelScreen->third.size = sarea->third_size; + } + intelScreen->depth.offset = sarea->depth_offset; intelScreen->depth.pitch = sarea->pitch * intelScreen->cpp; intelScreen->depth.handle = sarea->depth_handle; intelScreen->depth.size = sarea->depth_size; + if (intelScreen->driScrnPriv->ddx_version.minor >= 9) { + intelScreen->front.bo_handle = sarea->front_bo_handle; + intelScreen->back.bo_handle = sarea->back_bo_handle; + intelScreen->third.bo_handle = sarea->third_bo_handle; + intelScreen->depth.bo_handle = sarea->depth_bo_handle; + } else { + intelScreen->front.bo_handle = -1; + intelScreen->back.bo_handle = -1; + intelScreen->third.bo_handle = -1; + intelScreen->depth.bo_handle = -1; + } + intelScreen->tex.offset = sarea->tex_offset; intelScreen->logTextureGranularity = sarea->log_tex_granularity; intelScreen->tex.handle = sarea->tex_handle; @@ -247,60 +422,86 @@ intelUpdateScreenFromSAREA(intelScreenPrivate *intelScreen, intelPrintSAREA(sarea); } +static const __DRItexOffsetExtension intelTexOffsetExtension = { + { __DRI_TEX_OFFSET }, + intelSetTexOffset, +}; + +static const __DRIextension *intelExtensions[] = { + &driReadDrawableExtension, + &driCopySubBufferExtension.base, + &driSwapControlExtension.base, + &driFrameTrackingExtension.base, + &driMediaStreamCounterExtension.base, + &intelTexOffsetExtension.base, + NULL +}; + static GLboolean intelInitDriver(__DRIscreenPrivate *sPriv) { intelScreenPrivate *intelScreen; - I830DRIPtr gDRIPriv = (I830DRIPtr)sPriv->pDevPriv; + I830DRIPtr gDRIPriv = (I830DRIPtr) sPriv->pDevPriv; drmI830Sarea *sarea; - PFNGLXSCRENABLEEXTENSIONPROC glx_enable_extension = - (PFNGLXSCRENABLEEXTENSIONPROC) (*dri_interface->getProcAddress("glxEnableExtension")); - void * const psc = sPriv->psc->screenConfigs; if (sPriv->devPrivSize != sizeof(I830DRIRec)) { - fprintf(stderr,"\nERROR! sizeof(I830DRIRec) does not match passed size from device driver\n"); + fprintf(stderr, + "\nERROR! sizeof(I830DRIRec) does not match passed size from device driver\n"); return GL_FALSE; } /* Allocate the private area */ - intelScreen = (intelScreenPrivate *)CALLOC(sizeof(intelScreenPrivate)); + intelScreen = (intelScreenPrivate *) CALLOC(sizeof(intelScreenPrivate)); if (!intelScreen) { - fprintf(stderr,"\nERROR! Allocating private area failed\n"); + fprintf(stderr, "\nERROR! Allocating private area failed\n"); return GL_FALSE; } /* parse information in __driConfigOptions */ - driParseOptionInfo (&intelScreen->optionCache, - __driConfigOptions, __driNConfigOptions); + driParseOptionInfo(&intelScreen->optionCache, + __driConfigOptions, __driNConfigOptions); intelScreen->driScrnPriv = sPriv; - sPriv->private = (void *)intelScreen; + sPriv->private = (void *) intelScreen; intelScreen->sarea_priv_offset = gDRIPriv->sarea_priv_offset; sarea = (drmI830Sarea *) - (((GLubyte *)sPriv->pSAREA)+intelScreen->sarea_priv_offset); + (((GLubyte *) sPriv->pSAREA) + intelScreen->sarea_priv_offset); intelScreen->deviceID = gDRIPriv->deviceID; + if (intelScreen->deviceID == PCI_CHIP_I865_G) + intelScreen->maxBatchSize = 4096; + else + intelScreen->maxBatchSize = BATCH_SZ; + intelScreen->mem = gDRIPriv->mem; intelScreen->cpp = gDRIPriv->cpp; switch (gDRIPriv->bitsPerPixel) { - case 15: intelScreen->fbFormat = DV_PF_555; break; - case 16: intelScreen->fbFormat = DV_PF_565; break; - case 32: intelScreen->fbFormat = DV_PF_8888; break; + case 16: + intelScreen->fbFormat = DV_PF_565; + break; + case 32: + intelScreen->fbFormat = DV_PF_8888; + break; + default: + exit(1); + break; } - - intelUpdateScreenFromSAREA(intelScreen, sarea); - if (0) - intelPrintDRIInfo(intelScreen, sPriv, gDRIPriv); + intelUpdateScreenFromSAREA(intelScreen, sarea); if (!intelMapScreenRegions(sPriv)) { - fprintf(stderr,"\nERROR! mapping regions\n"); + fprintf(stderr, "\nERROR! mapping regions\n"); _mesa_free(intelScreen); sPriv->private = NULL; return GL_FALSE; } - intelScreen->drmMinor = sPriv->drmMinor; + intelScreen->sarea_priv_offset = gDRIPriv->sarea_priv_offset; + + if (0) + intelPrintDRIInfo(intelScreen, sPriv, gDRIPriv); + + intelScreen->drmMinor = sPriv->drm_version.minor; /* Determine if IRQs are active? */ { @@ -310,11 +511,11 @@ static GLboolean intelInitDriver(__DRIscreenPrivate *sPriv) gp.param = I830_PARAM_IRQ_ACTIVE; gp.value = &intelScreen->irq_active; - ret = drmCommandWriteRead( sPriv->fd, DRM_I830_GETPARAM, - &gp, sizeof(gp)); + ret = drmCommandWriteRead(sPriv->fd, DRM_I830_GETPARAM, + &gp, sizeof(gp)); if (ret) { - fprintf(stderr, "drmI830GetParam: %d\n", ret); - return GL_FALSE; + fprintf(stderr, "drmI830GetParam: %d\n", ret); + return GL_FALSE; } } @@ -326,128 +527,178 @@ static GLboolean intelInitDriver(__DRIscreenPrivate *sPriv) gp.param = I830_PARAM_ALLOW_BATCHBUFFER; gp.value = &intelScreen->allow_batchbuffer; - ret = drmCommandWriteRead( sPriv->fd, DRM_I830_GETPARAM, - &gp, sizeof(gp)); + ret = drmCommandWriteRead(sPriv->fd, DRM_I830_GETPARAM, + &gp, sizeof(gp)); if (ret) { - fprintf(stderr, "drmI830GetParam: (%d) %d\n", gp.param, ret); - return GL_FALSE; + fprintf(stderr, "drmI830GetParam: (%d) %d\n", gp.param, ret); + return GL_FALSE; } } - if (glx_enable_extension != NULL) { - (*glx_enable_extension)( psc, "GLX_SGI_swap_control" ); - (*glx_enable_extension)( psc, "GLX_SGI_video_sync" ); - (*glx_enable_extension)( psc, "GLX_MESA_swap_control" ); - (*glx_enable_extension)( psc, "GLX_MESA_swap_frame_usage" ); - (*glx_enable_extension)( psc, "GLX_SGI_make_current_read" ); - (*glx_enable_extension)( psc, "GLX_MESA_allocate_memory" ); - (*glx_enable_extension)( psc, "GLX_MESA_copy_sub_buffer" ); + sPriv->extensions = intelExtensions; + + /* If we've got a new enough DDX that's initializing TTM and giving us + * object handles for the shared buffers, use that. + */ + intelScreen->ttm = GL_FALSE; + if (getenv("INTEL_NO_TTM") == NULL && + intelScreen->driScrnPriv->ddx_version.minor >= 9 && + intelScreen->drmMinor >= 11 && + intelScreen->front.bo_handle != -1) { + intelScreen->bufmgr = intel_bufmgr_ttm_init(sPriv->fd, + DRM_FENCE_TYPE_EXE, + DRM_FENCE_TYPE_EXE | + DRM_I915_FENCE_TYPE_RW, + BATCH_SZ); + if (intelScreen->bufmgr != NULL) + intelScreen->ttm = GL_TRUE; } - - sPriv->psc->allocateMemory = (void *) intelAllocateMemoryMESA; - sPriv->psc->freeMemory = (void *) intelFreeMemoryMESA; - sPriv->psc->memoryOffset = (void *) intelGetMemoryOffsetMESA; + /* Otherwise, use the classic buffer manager. */ + if (intelScreen->bufmgr == NULL) { + if (intelScreen->tex.size == 0) { + fprintf(stderr, "[%s:%u] Error initializing buffer manager.\n", + __func__, __LINE__); + return GL_FALSE; + } + fprintf(stderr, "[%s:%u] Failed to init TTM buffer manager, falling back" + " to classic.\n", __func__, __LINE__); + intelScreen->bufmgr = dri_bufmgr_fake_init(intelScreen->tex.offset, + intelScreen->tex.map, + intelScreen->tex.size, + intel_fence_emit, + intel_fence_wait, + intelScreen); + } + + intel_recreate_static_regions(intelScreen); return GL_TRUE; } - - -static void intelDestroyScreen(__DRIscreenPrivate *sPriv) + + +static void +intelDestroyScreen(__DRIscreenPrivate * sPriv) { - intelScreenPrivate *intelScreen = (intelScreenPrivate *)sPriv->private; + intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; intelUnmapScreenRegions(intelScreen); - driDestroyOptionInfo (&intelScreen->optionCache); - + dri_bufmgr_destroy(intelScreen->bufmgr); FREE(intelScreen); sPriv->private = NULL; } -static GLboolean intelCreateBuffer( __DRIscreenPrivate *driScrnPriv, - __DRIdrawablePrivate *driDrawPriv, - const __GLcontextModes *mesaVis, - GLboolean isPixmap ) +/** + * This is called when we need to set up GL rendering to a new X window. + */ +static GLboolean +intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, + __DRIdrawablePrivate * driDrawPriv, + const __GLcontextModes * mesaVis, GLboolean isPixmap) { intelScreenPrivate *screen = (intelScreenPrivate *) driScrnPriv->private; if (isPixmap) { - return GL_FALSE; /* not implemented */ - } else { - GLboolean swStencil = (mesaVis->stencilBits > 0 && - mesaVis->depthBits != 24); + return GL_FALSE; /* not implemented */ + } + else { + GLboolean swStencil = (mesaVis->stencilBits > 0 && + mesaVis->depthBits != 24); + GLenum rgbFormat = (mesaVis->redBits == 5 ? GL_RGB5 : GL_RGBA8); - struct gl_framebuffer *fb = _mesa_create_framebuffer(mesaVis); + struct intel_framebuffer *intel_fb = CALLOC_STRUCT(intel_framebuffer); + + if (!intel_fb) + return GL_FALSE; + _mesa_initialize_framebuffer(&intel_fb->Base, mesaVis); + + /* setup the hardware-based renderbuffers */ { - driRenderbuffer *frontRb - = driNewRenderbuffer(GL_RGBA, - screen->front.map, - screen->cpp, - screen->front.offset, screen->front.pitch, - driDrawPriv); - intelSetSpanFunctions(frontRb, mesaVis); - _mesa_add_renderbuffer(fb, BUFFER_FRONT_LEFT, &frontRb->Base); + intel_fb->color_rb[0] + = intel_create_renderbuffer(rgbFormat, + screen->width, screen->height, + screen->front.offset, + screen->front.pitch, + screen->cpp, + screen->front.map); + intel_set_span_functions(&intel_fb->color_rb[0]->Base); + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_FRONT_LEFT, + &intel_fb->color_rb[0]->Base); } if (mesaVis->doubleBufferMode) { - driRenderbuffer *backRb - = driNewRenderbuffer(GL_RGBA, - screen->back.map, - screen->cpp, - screen->back.offset, screen->back.pitch, - driDrawPriv); - intelSetSpanFunctions(backRb, mesaVis); - _mesa_add_renderbuffer(fb, BUFFER_BACK_LEFT, &backRb->Base); + intel_fb->color_rb[1] + = intel_create_renderbuffer(rgbFormat, + screen->width, screen->height, + screen->back.offset, + screen->back.pitch, + screen->cpp, + screen->back.map); + intel_set_span_functions(&intel_fb->color_rb[1]->Base); + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT, + &intel_fb->color_rb[1]->Base); + + if (screen->third.handle) { + struct gl_renderbuffer *tmp_rb = NULL; + + intel_fb->color_rb[2] + = intel_create_renderbuffer(rgbFormat, + screen->width, screen->height, + screen->third.offset, + screen->third.pitch, + screen->cpp, + screen->third.map); + intel_set_span_functions(&intel_fb->color_rb[2]->Base); + _mesa_reference_renderbuffer(&tmp_rb, &intel_fb->color_rb[2]->Base); + } } - if (mesaVis->depthBits == 16) { - driRenderbuffer *depthRb - = driNewRenderbuffer(GL_DEPTH_COMPONENT16, - screen->depth.map, - screen->cpp, - screen->depth.offset, screen->depth.pitch, - driDrawPriv); - intelSetSpanFunctions(depthRb, mesaVis); - _mesa_add_renderbuffer(fb, BUFFER_DEPTH, &depthRb->Base); - } - else if (mesaVis->depthBits == 24) { - driRenderbuffer *depthRb - = driNewRenderbuffer(GL_DEPTH_COMPONENT24, - screen->depth.map, - screen->cpp, - screen->depth.offset, screen->depth.pitch, - driDrawPriv); - intelSetSpanFunctions(depthRb, mesaVis); - _mesa_add_renderbuffer(fb, BUFFER_DEPTH, &depthRb->Base); + if (mesaVis->depthBits == 24 && mesaVis->stencilBits == 8) { + /* combined depth/stencil buffer */ + struct intel_renderbuffer *depthStencilRb + = intel_create_renderbuffer(GL_DEPTH24_STENCIL8_EXT, + screen->width, screen->height, + screen->depth.offset, + screen->depth.pitch, + screen->cpp, /* 4! */ + screen->depth.map); + intel_set_span_functions(&depthStencilRb->Base); + /* note: bind RB to two attachment points */ + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, + &depthStencilRb->Base); + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_STENCIL, + &depthStencilRb->Base); } - - if (mesaVis->stencilBits > 0 && !swStencil) { - driRenderbuffer *stencilRb - = driNewRenderbuffer(GL_STENCIL_INDEX8_EXT, - screen->depth.map, - screen->cpp, - screen->depth.offset, screen->depth.pitch, - driDrawPriv); - intelSetSpanFunctions(stencilRb, mesaVis); - _mesa_add_renderbuffer(fb, BUFFER_STENCIL, &stencilRb->Base); + else if (mesaVis->depthBits == 16) { + /* just 16-bit depth buffer, no hw stencil */ + struct intel_renderbuffer *depthRb + = intel_create_renderbuffer(GL_DEPTH_COMPONENT16, + screen->width, screen->height, + screen->depth.offset, + screen->depth.pitch, + screen->cpp, /* 2! */ + screen->depth.map); + intel_set_span_functions(&depthRb->Base); + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, &depthRb->Base); } - _mesa_add_soft_renderbuffers(fb, - GL_FALSE, /* color */ - GL_FALSE, /* depth */ - swStencil, - mesaVis->accumRedBits > 0, - GL_FALSE, /* alpha */ - GL_FALSE /* aux */); - driDrawPriv->driverPrivate = (void *) fb; + /* now add any/all software-based renderbuffers we may need */ + _mesa_add_soft_renderbuffers(&intel_fb->Base, + GL_FALSE, /* never sw color */ + GL_FALSE, /* never sw depth */ + swStencil, mesaVis->accumRedBits > 0, + GL_FALSE, /* never sw alpha */ + GL_FALSE /* never sw aux */ ); + driDrawPriv->driverPrivate = (void *) intel_fb; - return (driDrawPriv->driverPrivate != NULL); + return GL_TRUE; } } -static void intelDestroyBuffer(__DRIdrawablePrivate *driDrawPriv) +static void +intelDestroyBuffer(__DRIdrawablePrivate * driDrawPriv) { _mesa_unreference_framebuffer((GLframebuffer **)(&(driDrawPriv->driverPrivate))); } @@ -457,24 +708,23 @@ static void intelDestroyBuffer(__DRIdrawablePrivate *driDrawPriv) * Get information about previous buffer swaps. */ static int -intelGetSwapInfo( __DRIdrawablePrivate *dPriv, __DRIswapInfo * sInfo ) +intelGetSwapInfo(__DRIdrawablePrivate * dPriv, __DRIswapInfo * sInfo) { - intelContextPtr intel; + struct intel_framebuffer *intel_fb; - if ( (dPriv == NULL) || (dPriv->driContextPriv == NULL) - || (dPriv->driContextPriv->driverPrivate == NULL) - || (sInfo == NULL) ) { + if ((dPriv == NULL) || (dPriv->driverPrivate == NULL) + || (sInfo == NULL)) { return -1; } - intel = dPriv->driContextPriv->driverPrivate; - sInfo->swap_count = intel->swap_count; - sInfo->swap_ust = intel->swap_ust; - sInfo->swap_missed_count = intel->swap_missed_count; + intel_fb = dPriv->driverPrivate; + sInfo->swap_count = intel_fb->swap_count; + sInfo->swap_ust = intel_fb->swap_ust; + sInfo->swap_missed_count = intel_fb->swap_missed_count; sInfo->swap_missed_usage = (sInfo->swap_missed_count != 0) - ? driCalculateSwapUsage( dPriv, 0, intel->swap_missed_ust ) - : 0.0; + ? driCalculateSwapUsage(dPriv, 0, intel_fb->swap_missed_ust) + : 0.0; return 0; } @@ -484,31 +734,33 @@ intelGetSwapInfo( __DRIdrawablePrivate *dPriv, __DRIswapInfo * sInfo ) * init-designated function to register chipids and createcontext * functions. */ -extern GLboolean i830CreateContext( const __GLcontextModes *mesaVis, - __DRIcontextPrivate *driContextPriv, - void *sharedContextPrivate); +extern GLboolean i830CreateContext(const __GLcontextModes * mesaVis, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate); -extern GLboolean i915CreateContext( const __GLcontextModes *mesaVis, - __DRIcontextPrivate *driContextPriv, - void *sharedContextPrivate); +extern GLboolean i915CreateContext(const __GLcontextModes * mesaVis, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate); -static GLboolean intelCreateContext( const __GLcontextModes *mesaVis, - __DRIcontextPrivate *driContextPriv, - void *sharedContextPrivate) +static GLboolean +intelCreateContext(const __GLcontextModes * mesaVis, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate) { __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; - intelScreenPrivate *intelScreen = (intelScreenPrivate *)sPriv->private; + intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; switch (intelScreen->deviceID) { + /* Don't deal with i830 until texture work complete: + */ case PCI_CHIP_845_G: case PCI_CHIP_I830_M: case PCI_CHIP_I855_GM: case PCI_CHIP_I865_G: - return i830CreateContext( mesaVis, driContextPriv, - sharedContextPrivate ); + return i830CreateContext(mesaVis, driContextPriv, sharedContextPrivate); case PCI_CHIP_I915_G: case PCI_CHIP_I915_GM: @@ -518,9 +770,8 @@ static GLboolean intelCreateContext( const __GLcontextModes *mesaVis, case PCI_CHIP_G33_G: case PCI_CHIP_Q35_G: case PCI_CHIP_Q33_G: - return i915CreateContext( mesaVis, driContextPriv, - sharedContextPrivate ); - + return i915CreateContext(mesaVis, driContextPriv, sharedContextPrivate); + default: fprintf(stderr, "Unrecognized deviceID %x\n", intelScreen->deviceID); return GL_FALSE; @@ -529,30 +780,30 @@ static GLboolean intelCreateContext( const __GLcontextModes *mesaVis, static const struct __DriverAPIRec intelAPI = { - .InitDriver = intelInitDriver, - .DestroyScreen = intelDestroyScreen, - .CreateContext = intelCreateContext, - .DestroyContext = intelDestroyContext, - .CreateBuffer = intelCreateBuffer, - .DestroyBuffer = intelDestroyBuffer, - .SwapBuffers = intelSwapBuffers, - .MakeCurrent = intelMakeCurrent, - .UnbindContext = intelUnbindContext, - .GetSwapInfo = intelGetSwapInfo, - .GetMSC = driGetMSC32, - .WaitForMSC = driWaitForMSC32, - .WaitForSBC = NULL, - .SwapBuffersMSC = NULL, - .CopySubBuffer = intelCopySubBuffer + .DestroyScreen = intelDestroyScreen, + .CreateContext = intelCreateContext, + .DestroyContext = intelDestroyContext, + .CreateBuffer = intelCreateBuffer, + .DestroyBuffer = intelDestroyBuffer, + .SwapBuffers = intelSwapBuffers, + .MakeCurrent = intelMakeCurrent, + .UnbindContext = intelUnbindContext, + .GetSwapInfo = intelGetSwapInfo, + .GetMSC = driGetMSC32, + .WaitForMSC = driWaitForMSC32, + .WaitForSBC = NULL, + .SwapBuffersMSC = NULL, + .CopySubBuffer = intelCopySubBuffer, + .setTexOffset = intelSetTexOffset, }; static __GLcontextModes * -intelFillInModes( unsigned pixel_bits, unsigned depth_bits, - unsigned stencil_bits, GLboolean have_back_buffer ) +intelFillInModes(unsigned pixel_bits, unsigned depth_bits, + unsigned stencil_bits, GLboolean have_back_buffer) { - __GLcontextModes * modes; - __GLcontextModes * m; + __GLcontextModes *modes; + __GLcontextModes *m; unsigned num_modes; unsigned depth_buffer_factor; unsigned back_buffer_factor; @@ -580,46 +831,51 @@ intelFillInModes( unsigned pixel_bits, unsigned depth_bits, */ stencil_bits_array[0] = 0; stencil_bits_array[1] = 0; + if (depth_bits == 24) + stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits; + stencil_bits_array[2] = (stencil_bits == 0) ? 8 : stencil_bits; depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 3 : 1; - back_buffer_factor = (have_back_buffer) ? 3 : 1; + back_buffer_factor = (have_back_buffer) ? 3 : 1; num_modes = depth_buffer_factor * back_buffer_factor * 4; - if ( pixel_bits == 16 ) { - fb_format = GL_RGB; - fb_type = GL_UNSIGNED_SHORT_5_6_5; - } - else { - fb_format = GL_BGRA; - fb_type = GL_UNSIGNED_INT_8_8_8_8_REV; - } + if (pixel_bits == 16) { + fb_format = GL_RGB; + fb_type = GL_UNSIGNED_SHORT_5_6_5; + } + else { + fb_format = GL_BGRA; + fb_type = GL_UNSIGNED_INT_8_8_8_8_REV; + } - modes = (*dri_interface->createContextModes)( num_modes, sizeof( __GLcontextModes ) ); + modes = + (*dri_interface->createContextModes) (num_modes, + sizeof(__GLcontextModes)); m = modes; - if ( ! driFillInModes( & m, fb_format, fb_type, - depth_bits_array, stencil_bits_array, depth_buffer_factor, - back_buffer_modes, back_buffer_factor, - GLX_TRUE_COLOR ) ) { - fprintf( stderr, "[%s:%u] Error creating FBConfig!\n", - __func__, __LINE__ ); - return NULL; - } - if ( ! driFillInModes( & m, fb_format, fb_type, - depth_bits_array, stencil_bits_array, depth_buffer_factor, - back_buffer_modes, back_buffer_factor, - GLX_DIRECT_COLOR ) ) { - fprintf( stderr, "[%s:%u] Error creating FBConfig!\n", - __func__, __LINE__ ); - return NULL; + if (!driFillInModes(&m, fb_format, fb_type, + depth_bits_array, stencil_bits_array, + depth_buffer_factor, back_buffer_modes, + back_buffer_factor, GLX_TRUE_COLOR)) { + fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, + __LINE__); + return NULL; + } + if (!driFillInModes(&m, fb_format, fb_type, + depth_bits_array, stencil_bits_array, + depth_buffer_factor, back_buffer_modes, + back_buffer_factor, GLX_DIRECT_COLOR)) { + fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, + __LINE__); + return NULL; } /* Mark the visual as slow if there are "fake" stencil bits. */ - for ( m = modes ; m != NULL ; m = m->next ) { - if ( (m->stencilBits != 0) && (m->stencilBits != stencil_bits) ) { - m->visualRating = GLX_SLOW_CONFIG; + for (m = modes; m != NULL; m = m->next) { + if ((m->stencilBits != 0) && (m->stencilBits != stencil_bits)) { + m->visualRating = GLX_SLOW_CONFIG; } } @@ -628,63 +884,61 @@ intelFillInModes( unsigned pixel_bits, unsigned depth_bits, /** - * This is the bootstrap function for the driver. libGL supplies all of the - * requisite information about the system, and the driver initializes itself. - * This routine also fills in the linked list pointed to by \c driver_modes - * with the \c __GLcontextModes that the driver can support for windows or - * pbuffers. + * This is the driver specific part of the createNewScreen entry point. * - * \return A pointer to a \c __DRIscreenPrivate on success, or \c NULL on - * failure. + * \todo maybe fold this into intelInitDriver + * + * \return the __GLcontextModes supported by this driver */ -PUBLIC -void * __driCreateNewScreen_20050727( __DRInativeDisplay *dpy, int scrn, __DRIscreen *psc, - const __GLcontextModes * modes, - const __DRIversion * ddx_version, - const __DRIversion * dri_version, - const __DRIversion * drm_version, - const __DRIframebuffer * frame_buffer, - drmAddress pSAREA, int fd, - int internal_api_version, - const __DRIinterfaceMethods * interface, - __GLcontextModes ** driver_modes ) - +PUBLIC __GLcontextModes *__driDriverInitScreen(__DRIscreenPrivate *psp) { - __DRIscreenPrivate *psp; static const __DRIversion ddx_expected = { 1, 5, 0 }; static const __DRIversion dri_expected = { 4, 0, 0 }; - static const __DRIversion drm_expected = { 1, 4, 0 }; + static const __DRIversion drm_expected = { 1, 5, 0 }; + I830DRIPtr dri_priv = (I830DRIPtr) psp->pDevPriv; - dri_interface = interface; + psp->DriverAPI = intelAPI; - if ( ! driCheckDriDdxDrmVersions2( "i915", - dri_version, & dri_expected, - ddx_version, & ddx_expected, - drm_version, & drm_expected ) ) { + if (!driCheckDriDdxDrmVersions2("i915", + &psp->dri_version, &dri_expected, + &psp->ddx_version, &ddx_expected, + &psp->drm_version, &drm_expected)) { return NULL; } - psp = __driUtilCreateNewScreen(dpy, scrn, psc, NULL, - ddx_version, dri_version, drm_version, - frame_buffer, pSAREA, fd, - internal_api_version, &intelAPI); - if ( psp != NULL ) { - I830DRIPtr dri_priv = (I830DRIPtr) psp->pDevPriv; - *driver_modes = intelFillInModes( dri_priv->cpp * 8, - (dri_priv->cpp == 2) ? 16 : 24, - (dri_priv->cpp == 2) ? 0 : 8, - 1 ); - - /* Calling driInitExtensions here, with a NULL context pointer, does not actually - * enable the extensions. It just makes sure that all the dispatch offsets for all - * the extensions that *might* be enables are known. This is needed because the - * dispatch offsets need to be known when _mesa_context_create is called, but we can't - * enable the extensions until we have a context pointer. - * - * Hello chicken. Hello egg. How are you two today? - */ - driInitExtensions( NULL, card_extensions, GL_FALSE ); - } + /* Calling driInitExtensions here, with a NULL context pointer, + * does not actually enable the extensions. It just makes sure + * that all the dispatch offsets for all the extensions that + * *might* be enables are known. This is needed because the + * dispatch offsets need to be known when _mesa_context_create is + * called, but we can't enable the extensions until we have a + * context pointer. + * + * Hello chicken. Hello egg. How are you two today? + */ + driInitExtensions(NULL, card_extensions, GL_FALSE); + driInitExtensions(NULL, ttm_extensions, GL_FALSE); - return (void *) psp; + if (!intelInitDriver(psp)) + return NULL; + + return intelFillInModes(dri_priv->cpp * 8, + (dri_priv->cpp == 2) ? 16 : 24, + (dri_priv->cpp == 2) ? 0 : 8, 1); } + +struct intel_context *intelScreenContext(intelScreenPrivate *intelScreen) +{ + /* + * This should probably change to have the screen allocate a dummy + * context at screen creation. For now just use the current context. + */ + + GET_CURRENT_CONTEXT(ctx); + if (ctx == NULL) { + _mesa_problem(NULL, "No current context in intelScreenContext\n"); + return NULL; + } + return intel_context(ctx); +} + diff --git a/src/mesa/drivers/dri/i915/intel_screen.h b/src/mesa/drivers/dri/i915/intel_screen.h index 24cfd9bf8b..aa0ef2c509 100644 --- a/src/mesa/drivers/dri/i915/intel_screen.h +++ b/src/mesa/drivers/dri/i915/intel_screen.h @@ -29,39 +29,51 @@ #define _INTEL_INIT_H_ #include <sys/time.h> -#include "xmlconfig.h" #include "dri_util.h" #include "intel_rotate.h" #include "i830_common.h" +#include "xmlconfig.h" +#include "dri_bufmgr.h" - -/* This roughly corresponds to a gl_renderbuffer (Mesa 6.4) */ -typedef struct { +/* XXX: change name or eliminate to avoid conflict with "struct + * intel_region"!!! + */ +typedef struct +{ drm_handle_t handle; - drmSize size; /* region size in bytes */ - char *map; /* memory map */ - int offset; /* from start of video mem, in bytes */ - int pitch; /* row stride, in bytes */ + drmSize size; /* region size in bytes */ + char *map; /* memory map */ + int offset; /* from start of video mem, in bytes */ + int pitch; /* row stride, in bytes */ + unsigned int bo_handle; /* buffer object id if available, or -1 */ } intelRegion; -typedef struct +typedef struct { intelRegion front; intelRegion back; + intelRegion third; intelRegion rotated; intelRegion depth; intelRegion tex; - + + struct intel_region *front_region; + struct intel_region *back_region; + struct intel_region *third_region; + struct intel_region *depth_region; + struct intel_region *rotated_region; + int deviceID; int width; int height; - int mem; /* unused */ - - int cpp; /* for front and back buffers */ - int fbFormat; + int mem; /* unused */ + + int cpp; /* for front and back buffers */ +/* int bitsPerPixel; */ + int fbFormat; /* XXX FBO: this is obsolete - remove after i830 updates */ int logTextureGranularity; - + __DRIscreenPrivate *driScrnPriv; unsigned int sarea_priv_offset; @@ -72,41 +84,51 @@ typedef struct struct matrix23 rotMatrix; - int current_rotation; /* 0, 90, 180 or 270 */ + int current_rotation; /* 0, 90, 180 or 270 */ int rotatedWidth, rotatedHeight; /** * Configuration cache with default values for all contexts */ driOptionCache optionCache; + + dri_bufmgr *bufmgr; + unsigned int maxBatchSize; + + /** + * This value indicates that the kernel memory manager is being used + * instead of the fake client-side memory manager. + */ + GLboolean ttm; } intelScreenPrivate; -extern GLboolean -intelMapScreenRegions(__DRIscreenPrivate *sPriv); -extern void -intelUnmapScreenRegions(intelScreenPrivate *intelScreen); +extern GLboolean intelMapScreenRegions(__DRIscreenPrivate * sPriv); -extern void -intelUpdateScreenFromSAREA(intelScreenPrivate *intelScreen, - drmI830Sarea *sarea); +extern void intelUnmapScreenRegions(intelScreenPrivate * intelScreen); extern void -intelDestroyContext(__DRIcontextPrivate *driContextPriv); +intelUpdateScreenFromSAREA(intelScreenPrivate * intelScreen, + drmI830Sarea * sarea); -extern GLboolean -intelUnbindContext(__DRIcontextPrivate *driContextPriv); +extern void intelDestroyContext(__DRIcontextPrivate * driContextPriv); + +extern GLboolean intelUnbindContext(__DRIcontextPrivate * driContextPriv); extern GLboolean -intelMakeCurrent(__DRIcontextPrivate *driContextPriv, - __DRIdrawablePrivate *driDrawPriv, - __DRIdrawablePrivate *driReadPriv); +intelMakeCurrent(__DRIcontextPrivate * driContextPriv, + __DRIdrawablePrivate * driDrawPriv, + __DRIdrawablePrivate * driReadPriv); + +extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv); extern void -intelSwapBuffers(__DRIdrawablePrivate *dPriv); +intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h); + +extern struct intel_context *intelScreenContext(intelScreenPrivate *intelScreen); extern void -intelCopySubBuffer( __DRIdrawablePrivate *dPriv, int x, int y, int w, int h ); +intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea); #endif diff --git a/src/mesa/drivers/dri/i915/intel_span.c b/src/mesa/drivers/dri/i915/intel_span.c index c3ffc4b2ac..d1f8ef06be 100644 --- a/src/mesa/drivers/dri/i915/intel_span.c +++ b/src/mesa/drivers/dri/i915/intel_span.c @@ -30,229 +30,380 @@ #include "mtypes.h" #include "colormac.h" +#include "intel_fbo.h" #include "intel_screen.h" - #include "intel_span.h" +#include "intel_regions.h" #include "intel_ioctl.h" +#include "intel_tex.h" + #include "swrast/swrast.h" +/* + break intelWriteRGBASpan_ARGB8888 +*/ +#undef DBG #define DBG 0 -#define LOCAL_VARS \ - intelContextPtr intel = INTEL_CONTEXT(ctx); \ - __DRIdrawablePrivate *dPriv = intel->driDrawable; \ - driRenderbuffer *drb = (driRenderbuffer *) rb; \ - GLuint pitch = drb->pitch; \ - GLuint height = dPriv->h; \ - char *buf = (char *) drb->Base.Data + \ - dPriv->x * drb->cpp + \ - dPriv->y * pitch; \ - GLushort p; \ - (void) buf; (void) p - -#define LOCAL_DEPTH_VARS \ - intelContextPtr intel = INTEL_CONTEXT(ctx); \ - __DRIdrawablePrivate *dPriv = intel->driDrawable; \ - driRenderbuffer *drb = (driRenderbuffer *) rb; \ - GLuint pitch = drb->pitch; \ - GLuint height = dPriv->h; \ - char *buf = (char *) drb->Base.Data + \ - dPriv->x * drb->cpp + \ - dPriv->y * pitch - -#define LOCAL_STENCIL_VARS LOCAL_DEPTH_VARS - -#define INIT_MONO_PIXEL(p,color)\ - p = INTEL_PACKCOLOR565(color[0],color[1],color[2]) - -#define Y_FLIP(_y) (height - _y - 1) +#define LOCAL_VARS \ + struct intel_context *intel = intel_context(ctx); \ + struct intel_renderbuffer *irb = intel_renderbuffer(rb); \ + const GLint yScale = irb->RenderToTexture ? 1 : -1; \ + const GLint yBias = irb->RenderToTexture ? 0 : irb->Base.Height - 1; \ + GLubyte *buf = (GLubyte *) irb->pfMap \ + + (intel->drawY * irb->pfPitch + intel->drawX) * irb->region->cpp;\ + GLuint p; \ + assert(irb->pfMap);\ + (void) p; + +/* XXX FBO: this is identical to the macro in spantmp2.h except we get + * the cliprect info from the context, not the driDrawable. + * Move this into spantmp2.h someday. + */ +#define HW_CLIPLOOP() \ + do { \ + int _nc = intel->numClipRects; \ + while ( _nc-- ) { \ + int minx = intel->pClipRects[_nc].x1 - intel->drawX; \ + int miny = intel->pClipRects[_nc].y1 - intel->drawY; \ + int maxx = intel->pClipRects[_nc].x2 - intel->drawX; \ + int maxy = intel->pClipRects[_nc].y2 - intel->drawY; + + + + +#define Y_FLIP(_y) ((_y) * yScale + yBias) #define HW_LOCK() #define HW_UNLOCK() -/* 16 bit, 565 rgb color spanline and pixel functions +/* 16 bit, RGB565 color spanline and pixel functions */ -#define WRITE_RGBA( _x, _y, r, g, b, a ) \ - *(GLushort *)(buf + _x*2 + _y*pitch) = ( (((int)r & 0xf8) << 8) | \ - (((int)g & 0xfc) << 3) | \ - (((int)b & 0xf8) >> 3)) -#define WRITE_PIXEL( _x, _y, p ) \ - *(GLushort *)(buf + _x*2 + _y*pitch) = p - -#define READ_RGBA( rgba, _x, _y ) \ -do { \ - GLushort p = *(GLushort *)(buf + _x*2 + _y*pitch); \ - rgba[0] = (((p >> 11) & 0x1f) * 255) / 31; \ - rgba[1] = (((p >> 5) & 0x3f) * 255) / 63; \ - rgba[2] = (((p >> 0) & 0x1f) * 255) / 31; \ - rgba[3] = 255; \ -} while(0) - -#define TAG(x) intel##x##_565 -#include "spantmp.h" - -/* 15 bit, 555 rgb color spanline and pixel functions - */ -#define WRITE_RGBA( _x, _y, r, g, b, a ) \ - *(GLushort *)(buf + _x*2 + _y*pitch) = (((r & 0xf8) << 7) | \ - ((g & 0xf8) << 3) | \ - ((b & 0xf8) >> 3)) - -#define WRITE_PIXEL( _x, _y, p ) \ - *(GLushort *)(buf + _x*2 + _y*pitch) = p - -#define READ_RGBA( rgba, _x, _y ) \ -do { \ - GLushort p = *(GLushort *)(buf + _x*2 + _y*pitch); \ - rgba[0] = (p >> 7) & 0xf8; \ - rgba[1] = (p >> 3) & 0xf8; \ - rgba[2] = (p << 3) & 0xf8; \ - rgba[3] = 255; \ -} while(0) - -#define TAG(x) intel##x##_555 -#include "spantmp.h" - -/* 16 bit depthbuffer functions. +#define SPANTMP_PIXEL_FMT GL_RGB +#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_SHORT_5_6_5 + +#define TAG(x) intel##x##_RGB565 +#define TAG2(x,y) intel##x##_RGB565##y +#define GET_PTR(X,Y) (buf + ((Y) * irb->pfPitch + (X)) * 2) +#include "spantmp2.h" + +/* 32 bit, ARGB8888 color spanline and pixel functions */ -#define WRITE_DEPTH( _x, _y, d ) \ - *(GLushort *)(buf + (_x)*2 + (_y)*pitch) = d; +#define SPANTMP_PIXEL_FMT GL_BGRA +#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV -#define READ_DEPTH( d, _x, _y ) \ - d = *(GLushort *)(buf + (_x)*2 + (_y)*pitch); +#define TAG(x) intel##x##_ARGB8888 +#define TAG2(x,y) intel##x##_ARGB8888##y +#define GET_PTR(X,Y) (buf + ((Y) * irb->pfPitch + (X)) * 4) +#include "spantmp2.h" -#define TAG(x) intel##x##_z16 -#include "depthtmp.h" +#define LOCAL_DEPTH_VARS \ + struct intel_context *intel = intel_context(ctx); \ + struct intel_renderbuffer *irb = intel_renderbuffer(rb); \ + const GLuint pitch = irb->pfPitch/***XXX region->pitch*/; /* in pixels */ \ + const GLint yScale = irb->RenderToTexture ? 1 : -1; \ + const GLint yBias = irb->RenderToTexture ? 0 : irb->Base.Height - 1; \ + char *buf = (char *) irb->pfMap/*XXX use region->map*/ + \ + (intel->drawY * pitch + intel->drawX) * irb->region->cpp; -#undef LOCAL_VARS -#define LOCAL_VARS \ - intelContextPtr intel = INTEL_CONTEXT(ctx); \ - __DRIdrawablePrivate *dPriv = intel->driDrawable; \ - driRenderbuffer *drb = (driRenderbuffer *) rb; \ - GLuint pitch = drb->pitch; \ - GLuint height = dPriv->h; \ - char *buf = (char *)drb->Base.Data + \ - dPriv->x * drb->cpp + \ - dPriv->y * pitch; \ - GLuint p; \ - (void) buf; (void) p - -#undef INIT_MONO_PIXEL -#define INIT_MONO_PIXEL(p,color)\ - p = INTEL_PACKCOLOR8888(color[0],color[1],color[2],color[3]) - -/* 32 bit, 8888 argb color spanline and pixel functions - */ -#define WRITE_RGBA(_x, _y, r, g, b, a) \ - *(GLuint *)(buf + _x*4 + _y*pitch) = ((r << 16) | \ - (g << 8) | \ - (b << 0) | \ - (a << 24) ) +#define LOCAL_STENCIL_VARS LOCAL_DEPTH_VARS -#define WRITE_PIXEL(_x, _y, p) \ - *(GLuint *)(buf + _x*4 + _y*pitch) = p +/** + ** 16-bit depthbuffer functions. + **/ +#define WRITE_DEPTH( _x, _y, d ) \ + ((GLushort *)buf)[(_x) + (_y) * pitch] = d; +#define READ_DEPTH( d, _x, _y ) \ + d = ((GLushort *)buf)[(_x) + (_y) * pitch]; -#define READ_RGBA(rgba, _x, _y) \ - do { \ - GLuint p = *(GLuint *)(buf + _x*4 + _y*pitch); \ - rgba[0] = (p >> 16) & 0xff; \ - rgba[1] = (p >> 8) & 0xff; \ - rgba[2] = (p >> 0) & 0xff; \ - rgba[3] = (p >> 24) & 0xff; \ - } while (0) -#define TAG(x) intel##x##_8888 -#include "spantmp.h" +#define TAG(x) intel##x##_z16 +#include "depthtmp.h" -/* 24/8 bit interleaved depth/stencil functions - */ -#define WRITE_DEPTH( _x, _y, d ) { \ - GLuint tmp = *(GLuint *)(buf + (_x)*4 + (_y)*pitch); \ - tmp &= 0xff000000; \ - tmp |= (d) & 0xffffff; \ - *(GLuint *)(buf + (_x)*4 + (_y)*pitch) = tmp; \ +/** + ** 24/8-bit interleaved depth/stencil functions + ** Note: we're actually reading back combined depth+stencil values. + ** The wrappers in main/depthstencil.c are used to extract the depth + ** and stencil values. + **/ +/* Change ZZZS -> SZZZ */ +#define WRITE_DEPTH( _x, _y, d ) { \ + GLuint tmp = ((d) >> 8) | ((d) << 24); \ + ((GLuint *)buf)[(_x) + (_y) * pitch] = tmp; \ } -#define READ_DEPTH( d, _x, _y ) \ - d = *(GLuint *)(buf + (_x)*4 + (_y)*pitch) & 0xffffff; - +/* Change SZZZ -> ZZZS */ +#define READ_DEPTH( d, _x, _y ) { \ + GLuint tmp = ((GLuint *)buf)[(_x) + (_y) * pitch]; \ + d = (tmp << 8) | (tmp >> 24); \ +} #define TAG(x) intel##x##_z24_s8 #include "depthtmp.h" -#define WRITE_STENCIL( _x, _y, d ) { \ - GLuint tmp = *(GLuint *)(buf + (_x)*4 + (_y)*pitch); \ - tmp &= 0xffffff; \ - tmp |= ((d)<<24); \ - *(GLuint *)(buf + (_x)*4 + (_y)*pitch) = tmp; \ + +/** + ** 8-bit stencil function (XXX FBO: This is obsolete) + **/ +#define WRITE_STENCIL( _x, _y, d ) { \ + GLuint tmp = ((GLuint *)buf)[(_x) + (_y) * pitch]; \ + tmp &= 0xffffff; \ + tmp |= ((d) << 24); \ + ((GLuint *) buf)[(_x) + (_y) * pitch] = tmp; \ } -#define READ_STENCIL( d, _x, _y ) \ - d = *(GLuint *)(buf + (_x)*4 + (_y)*pitch) >> 24; +#define READ_STENCIL( d, _x, _y ) \ + d = ((GLuint *)buf)[(_x) + (_y) * pitch] >> 24; #define TAG(x) intel##x##_z24_s8 #include "stenciltmp.h" -/* Move locking out to get reasonable span performance. + +/** + * Map or unmap all the renderbuffers which we may need during + * software rendering. + * XXX in the future, we could probably convey extra information to + * reduce the number of mappings needed. I.e. if doing a glReadPixels + * from the depth buffer, we really only need one mapping. + * + * XXX Rewrite this function someday. + * We can probably just loop over all the renderbuffer attachments, + * map/unmap all of them, and not worry about the _ColorDrawBuffers + * _ColorReadBuffer, _DepthBuffer or _StencilBuffer fields. */ -void intelSpanRenderStart( GLcontext *ctx ) +static void +intel_map_unmap_buffers(struct intel_context *intel, GLboolean map) { - intelContextPtr intel = INTEL_CONTEXT(ctx); + GLcontext *ctx = &intel->ctx; + GLuint i, j; + struct intel_renderbuffer *irb; + + /* color draw buffers */ + for (i = 0; i < ctx->Const.MaxDrawBuffers; i++) { + for (j = 0; j < ctx->DrawBuffer->_NumColorDrawBuffers[i]; j++) { + struct gl_renderbuffer *rb = + ctx->DrawBuffer->_ColorDrawBuffers[i][j]; + irb = intel_renderbuffer(rb); + if (irb) { + /* this is a user-created intel_renderbuffer */ + if (irb->region) { + if (map) + intel_region_map(intel->intelScreen, irb->region); + else + intel_region_unmap(intel->intelScreen, irb->region); + irb->pfMap = irb->region->map; + irb->pfPitch = irb->region->pitch; + } + } + } + } + + /* check for render to textures */ + for (i = 0; i < BUFFER_COUNT; i++) { + struct gl_renderbuffer_attachment *att = + ctx->DrawBuffer->Attachment + i; + struct gl_texture_object *tex = att->Texture; + if (tex) { + /* render to texture */ + ASSERT(att->Renderbuffer); + if (map) { + struct gl_texture_image *texImg; + texImg = tex->Image[att->CubeMapFace][att->TextureLevel]; + intel_tex_map_images(intel, intel_texture_object(tex)); + } + else { + intel_tex_unmap_images(intel, intel_texture_object(tex)); + } + } + } + + /* color read buffers */ + irb = intel_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer); + if (irb && irb->region) { + if (map) + intel_region_map(intel->intelScreen, irb->region); + else + intel_region_unmap(intel->intelScreen, irb->region); + irb->pfMap = irb->region->map; + irb->pfPitch = irb->region->pitch; + } - intelFlush(&intel->ctx); + /* Account for front/back color page flipping. + * The span routines use the pfMap and pfPitch fields which will + * swap the front/back region map/pitch if we're page flipped. + * Do this after mapping, above, so the map field is valid. + */ +#if 0 + if (map && ctx->DrawBuffer->Name == 0) { + struct intel_renderbuffer *irbFront + = intel_get_renderbuffer(ctx->DrawBuffer, BUFFER_FRONT_LEFT); + struct intel_renderbuffer *irbBack + = intel_get_renderbuffer(ctx->DrawBuffer, BUFFER_BACK_LEFT); + if (irbBack) { + /* double buffered */ + if (intel->sarea->pf_current_page == 0) { + irbFront->pfMap = irbFront->region->map; + irbFront->pfPitch = irbFront->region->pitch; + irbBack->pfMap = irbBack->region->map; + irbBack->pfPitch = irbBack->region->pitch; + } + else { + irbFront->pfMap = irbBack->region->map; + irbFront->pfPitch = irbBack->region->pitch; + irbBack->pfMap = irbFront->region->map; + irbBack->pfPitch = irbFront->region->pitch; + } + } + } +#endif + + /* depth buffer (Note wrapper!) */ + if (ctx->DrawBuffer->_DepthBuffer) { + irb = intel_renderbuffer(ctx->DrawBuffer->_DepthBuffer->Wrapped); + if (irb && irb->region && irb->Base.Name != 0) { + if (map) { + intel_region_map(intel->intelScreen, irb->region); + irb->pfMap = irb->region->map; + irb->pfPitch = irb->region->pitch; + } + else { + intel_region_unmap(intel->intelScreen, irb->region); + irb->pfMap = NULL; + irb->pfPitch = 0; + } + } + } + + /* stencil buffer (Note wrapper!) */ + if (ctx->DrawBuffer->_StencilBuffer) { + irb = intel_renderbuffer(ctx->DrawBuffer->_StencilBuffer->Wrapped); + if (irb && irb->region && irb->Base.Name != 0) { + if (map) { + intel_region_map(intel->intelScreen, irb->region); + irb->pfMap = irb->region->map; + irb->pfPitch = irb->region->pitch; + } + else { + intel_region_unmap(intel->intelScreen, irb->region); + irb->pfMap = NULL; + irb->pfPitch = 0; + } + } + } +} + + + +/** + * Prepare for softare rendering. Map current read/draw framebuffers' + * renderbuffes and all currently bound texture objects. + * + * Old note: Moved locking out to get reasonable span performance. + */ +void +intelSpanRenderStart(GLcontext * ctx) +{ + struct intel_context *intel = intel_context(ctx); + GLuint i; + + intelFinish(&intel->ctx); LOCK_HARDWARE(intel); - intelWaitForIdle(intel); + +#if 0 + /* Just map the framebuffer and all textures. Bufmgr code will + * take care of waiting on the necessary fences: + */ + intel_region_map(intel->intelScreen, intel->front_region); + intel_region_map(intel->intelScreen, intel->back_region); + intel_region_map(intel->intelScreen, intel->intelScreen->depth_region); +#endif + + for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) { + if (ctx->Texture.Unit[i]._ReallyEnabled) { + struct gl_texture_object *texObj = ctx->Texture.Unit[i]._Current; + intel_tex_map_images(intel, intel_texture_object(texObj)); + } + } + + intel_map_unmap_buffers(intel, GL_TRUE); } -void intelSpanRenderFinish( GLcontext *ctx ) +/** + * Called when done softare rendering. Unmap the buffers we mapped in + * the above function. + */ +void +intelSpanRenderFinish(GLcontext * ctx) { - intelContextPtr intel = INTEL_CONTEXT( ctx ); - _swrast_flush( ctx ); - UNLOCK_HARDWARE( intel ); + struct intel_context *intel = intel_context(ctx); + GLuint i; + + _swrast_flush(ctx); + + /* Now unmap the framebuffer: + */ +#if 0 + intel_region_unmap(intel, intel->front_region); + intel_region_unmap(intel, intel->back_region); + intel_region_unmap(intel, intel->intelScreen->depth_region); +#endif + + for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) { + if (ctx->Texture.Unit[i]._ReallyEnabled) { + struct gl_texture_object *texObj = ctx->Texture.Unit[i]._Current; + intel_tex_unmap_images(intel, intel_texture_object(texObj)); + } + } + + intel_map_unmap_buffers(intel, GL_FALSE); + + UNLOCK_HARDWARE(intel); } -void intelInitSpanFuncs( GLcontext *ctx ) + +void +intelInitSpanFuncs(GLcontext * ctx) { struct swrast_device_driver *swdd = _swrast_GetDeviceDriverReference(ctx); swdd->SpanRenderStart = intelSpanRenderStart; - swdd->SpanRenderFinish = intelSpanRenderFinish; + swdd->SpanRenderFinish = intelSpanRenderFinish; } /** - * Plug in the Get/Put routines for the given driRenderbuffer. + * Plug in appropriate span read/write functions for the given renderbuffer. + * These are used for the software fallbacks. */ void -intelSetSpanFunctions(driRenderbuffer *drb, const GLvisual *vis) +intel_set_span_functions(struct gl_renderbuffer *rb) { - if (drb->Base.InternalFormat == GL_RGBA) { - if (vis->redBits == 5 && vis->greenBits == 5 && vis->blueBits == 5) { - intelInitPointers_555(&drb->Base); - } - else if (vis->redBits == 5 && vis->greenBits == 6 && vis->blueBits == 5) { - intelInitPointers_565(&drb->Base); - } - else { - assert(vis->redBits == 8); - assert(vis->greenBits == 8); - assert(vis->blueBits == 8); - intelInitPointers_8888(&drb->Base); - } + if (rb->_ActualFormat == GL_RGB5) { + /* 565 RGB */ + intelInitPointers_RGB565(rb); + } + else if (rb->_ActualFormat == GL_RGBA8) { + /* 8888 RGBA */ + intelInitPointers_ARGB8888(rb); + } + else if (rb->_ActualFormat == GL_DEPTH_COMPONENT16) { + intelInitDepthPointers_z16(rb); } - else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT16) { - intelInitDepthPointers_z16(&drb->Base); + else if (rb->_ActualFormat == GL_DEPTH_COMPONENT24 || /* XXX FBO remove */ + rb->_ActualFormat == GL_DEPTH24_STENCIL8_EXT) { + intelInitDepthPointers_z24_s8(rb); } - else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT24) { - intelInitDepthPointers_z24_s8(&drb->Base); + else if (rb->_ActualFormat == GL_STENCIL_INDEX8_EXT) { /* XXX FBO remove */ + intelInitStencilPointers_z24_s8(rb); } - else if (drb->Base.InternalFormat == GL_STENCIL_INDEX8_EXT) { - intelInitStencilPointers_z24_s8(&drb->Base); + else { + _mesa_problem(NULL, + "Unexpected _ActualFormat in intelSetSpanFunctions"); } } diff --git a/src/mesa/drivers/dri/i915/intel_span.h b/src/mesa/drivers/dri/i915/intel_span.h index 2d4f8589d0..5201f6d6c6 100644 --- a/src/mesa/drivers/dri/i915/intel_span.h +++ b/src/mesa/drivers/dri/i915/intel_span.h @@ -28,14 +28,11 @@ #ifndef _INTEL_SPAN_H #define _INTEL_SPAN_H -#include "drirenderbuffer.h" +extern void intelInitSpanFuncs(GLcontext * ctx); -extern void intelInitSpanFuncs( GLcontext *ctx ); +extern void intelSpanRenderFinish(GLcontext * ctx); +extern void intelSpanRenderStart(GLcontext * ctx); -extern void intelSpanRenderFinish( GLcontext *ctx ); -extern void intelSpanRenderStart( GLcontext *ctx ); - -extern void -intelSetSpanFunctions(driRenderbuffer *rb, const GLvisual *vis); +extern void intel_set_span_functions(struct gl_renderbuffer *rb); #endif diff --git a/src/mesa/drivers/dri/i915/intel_state.c b/src/mesa/drivers/dri/i915/intel_state.c index e5988a5ed6..271511037e 100644 --- a/src/mesa/drivers/dri/i915/intel_state.c +++ b/src/mesa/drivers/dri/i915/intel_state.c @@ -30,252 +30,240 @@ #include "context.h" #include "macros.h" #include "enums.h" +#include "colormac.h" #include "dd.h" #include "intel_screen.h" #include "intel_context.h" +#include "intel_fbo.h" +#include "intel_regions.h" #include "swrast/swrast.h" -int intel_translate_compare_func( GLenum func ) +int +intel_translate_compare_func(GLenum func) { - switch(func) { - case GL_NEVER: - return COMPAREFUNC_NEVER; - case GL_LESS: - return COMPAREFUNC_LESS; - case GL_LEQUAL: - return COMPAREFUNC_LEQUAL; - case GL_GREATER: - return COMPAREFUNC_GREATER; - case GL_GEQUAL: - return COMPAREFUNC_GEQUAL; - case GL_NOTEQUAL: - return COMPAREFUNC_NOTEQUAL; - case GL_EQUAL: - return COMPAREFUNC_EQUAL; - case GL_ALWAYS: - return COMPAREFUNC_ALWAYS; + switch (func) { + case GL_NEVER: + return COMPAREFUNC_NEVER; + case GL_LESS: + return COMPAREFUNC_LESS; + case GL_LEQUAL: + return COMPAREFUNC_LEQUAL; + case GL_GREATER: + return COMPAREFUNC_GREATER; + case GL_GEQUAL: + return COMPAREFUNC_GEQUAL; + case GL_NOTEQUAL: + return COMPAREFUNC_NOTEQUAL; + case GL_EQUAL: + return COMPAREFUNC_EQUAL; + case GL_ALWAYS: + return COMPAREFUNC_ALWAYS; } fprintf(stderr, "Unknown value in %s: %x\n", __FUNCTION__, func); - return COMPAREFUNC_ALWAYS; + return COMPAREFUNC_ALWAYS; } -int intel_translate_stencil_op( GLenum op ) +int +intel_translate_stencil_op(GLenum op) { - switch(op) { - case GL_KEEP: - return STENCILOP_KEEP; - case GL_ZERO: - return STENCILOP_ZERO; - case GL_REPLACE: - return STENCILOP_REPLACE; - case GL_INCR: + switch (op) { + case GL_KEEP: + return STENCILOP_KEEP; + case GL_ZERO: + return STENCILOP_ZERO; + case GL_REPLACE: + return STENCILOP_REPLACE; + case GL_INCR: return STENCILOP_INCRSAT; - case GL_DECR: + case GL_DECR: return STENCILOP_DECRSAT; case GL_INCR_WRAP: - return STENCILOP_INCR; + return STENCILOP_INCR; case GL_DECR_WRAP: - return STENCILOP_DECR; - case GL_INVERT: - return STENCILOP_INVERT; - default: + return STENCILOP_DECR; + case GL_INVERT: + return STENCILOP_INVERT; + default: return STENCILOP_ZERO; } } -int intel_translate_blend_factor( GLenum factor ) +int +intel_translate_blend_factor(GLenum factor) { - switch(factor) { - case GL_ZERO: - return BLENDFACT_ZERO; - case GL_SRC_ALPHA: - return BLENDFACT_SRC_ALPHA; - case GL_ONE: - return BLENDFACT_ONE; - case GL_SRC_COLOR: - return BLENDFACT_SRC_COLR; - case GL_ONE_MINUS_SRC_COLOR: - return BLENDFACT_INV_SRC_COLR; - case GL_DST_COLOR: - return BLENDFACT_DST_COLR; - case GL_ONE_MINUS_DST_COLOR: - return BLENDFACT_INV_DST_COLR; + switch (factor) { + case GL_ZERO: + return BLENDFACT_ZERO; + case GL_SRC_ALPHA: + return BLENDFACT_SRC_ALPHA; + case GL_ONE: + return BLENDFACT_ONE; + case GL_SRC_COLOR: + return BLENDFACT_SRC_COLR; + case GL_ONE_MINUS_SRC_COLOR: + return BLENDFACT_INV_SRC_COLR; + case GL_DST_COLOR: + return BLENDFACT_DST_COLR; + case GL_ONE_MINUS_DST_COLOR: + return BLENDFACT_INV_DST_COLR; case GL_ONE_MINUS_SRC_ALPHA: - return BLENDFACT_INV_SRC_ALPHA; - case GL_DST_ALPHA: - return BLENDFACT_DST_ALPHA; + return BLENDFACT_INV_SRC_ALPHA; + case GL_DST_ALPHA: + return BLENDFACT_DST_ALPHA; case GL_ONE_MINUS_DST_ALPHA: - return BLENDFACT_INV_DST_ALPHA; - case GL_SRC_ALPHA_SATURATE: + return BLENDFACT_INV_DST_ALPHA; + case GL_SRC_ALPHA_SATURATE: return BLENDFACT_SRC_ALPHA_SATURATE; case GL_CONSTANT_COLOR: - return BLENDFACT_CONST_COLOR; + return BLENDFACT_CONST_COLOR; case GL_ONE_MINUS_CONSTANT_COLOR: return BLENDFACT_INV_CONST_COLOR; case GL_CONSTANT_ALPHA: - return BLENDFACT_CONST_ALPHA; + return BLENDFACT_CONST_ALPHA; case GL_ONE_MINUS_CONSTANT_ALPHA: return BLENDFACT_INV_CONST_ALPHA; } - + fprintf(stderr, "Unknown value in %s: %x\n", __FUNCTION__, factor); return BLENDFACT_ZERO; } -int intel_translate_logic_op( GLenum opcode ) +int +intel_translate_logic_op(GLenum opcode) { - switch(opcode) { - case GL_CLEAR: - return LOGICOP_CLEAR; - case GL_AND: - return LOGICOP_AND; - case GL_AND_REVERSE: - return LOGICOP_AND_RVRSE; - case GL_COPY: - return LOGICOP_COPY; - case GL_COPY_INVERTED: - return LOGICOP_COPY_INV; - case GL_AND_INVERTED: - return LOGICOP_AND_INV; - case GL_NOOP: - return LOGICOP_NOOP; - case GL_XOR: - return LOGICOP_XOR; - case GL_OR: - return LOGICOP_OR; - case GL_OR_INVERTED: - return LOGICOP_OR_INV; - case GL_NOR: - return LOGICOP_NOR; - case GL_EQUIV: - return LOGICOP_EQUIV; - case GL_INVERT: - return LOGICOP_INV; - case GL_OR_REVERSE: - return LOGICOP_OR_RVRSE; - case GL_NAND: - return LOGICOP_NAND; - case GL_SET: - return LOGICOP_SET; - default: + switch (opcode) { + case GL_CLEAR: + return LOGICOP_CLEAR; + case GL_AND: + return LOGICOP_AND; + case GL_AND_REVERSE: + return LOGICOP_AND_RVRSE; + case GL_COPY: + return LOGICOP_COPY; + case GL_COPY_INVERTED: + return LOGICOP_COPY_INV; + case GL_AND_INVERTED: + return LOGICOP_AND_INV; + case GL_NOOP: + return LOGICOP_NOOP; + case GL_XOR: + return LOGICOP_XOR; + case GL_OR: + return LOGICOP_OR; + case GL_OR_INVERTED: + return LOGICOP_OR_INV; + case GL_NOR: + return LOGICOP_NOR; + case GL_EQUIV: + return LOGICOP_EQUIV; + case GL_INVERT: + return LOGICOP_INV; + case GL_OR_REVERSE: + return LOGICOP_OR_RVRSE; + case GL_NAND: + return LOGICOP_NAND; + case GL_SET: return LOGICOP_SET; - } -} - -static void intelDrawBuffer(GLcontext *ctx, GLenum mode ) -{ - intelContextPtr intel = INTEL_CONTEXT(ctx); - int front = 0; - - if (!ctx->DrawBuffer) - return; - - switch ( ctx->DrawBuffer->_ColorDrawBufferMask[0] ) { - case BUFFER_BIT_FRONT_LEFT: - front = 1; - FALLBACK( intel, INTEL_FALLBACK_DRAW_BUFFER, GL_FALSE ); - break; - case BUFFER_BIT_BACK_LEFT: - front = 0; - FALLBACK( intel, INTEL_FALLBACK_DRAW_BUFFER, GL_FALSE ); - break; default: - FALLBACK( intel, INTEL_FALLBACK_DRAW_BUFFER, GL_TRUE ); - return; - } - - if ( intel->sarea->pf_current_page == 1 ) - front ^= 1; - - intelSetFrontClipRects( intel ); - - if (front) { - intel->drawRegion = &intel->intelScreen->front; - intel->readRegion = &intel->intelScreen->front; - } else { - intel->drawRegion = &intel->intelScreen->back; - intel->readRegion = &intel->intelScreen->back; + return LOGICOP_SET; } - - intel->vtbl.set_color_region( intel, intel->drawRegion ); -} - -static void intelReadBuffer( GLcontext *ctx, GLenum mode ) -{ - /* nothing, until we implement h/w glRead/CopyPixels or CopyTexImage */ } -static void intelClearColor(GLcontext *ctx, const GLfloat color[4]) +static void +intelClearColor(GLcontext * ctx, const GLfloat color[4]) { - intelContextPtr intel = INTEL_CONTEXT(ctx); - intelScreenPrivate *screen = intel->intelScreen; + struct intel_context *intel = intel_context(ctx); + GLubyte clear[4]; - CLAMPED_FLOAT_TO_UBYTE(intel->clear_red, color[0]); - CLAMPED_FLOAT_TO_UBYTE(intel->clear_green, color[1]); - CLAMPED_FLOAT_TO_UBYTE(intel->clear_blue, color[2]); - CLAMPED_FLOAT_TO_UBYTE(intel->clear_alpha, color[3]); + CLAMPED_FLOAT_TO_UBYTE(clear[0], color[0]); + CLAMPED_FLOAT_TO_UBYTE(clear[1], color[1]); + CLAMPED_FLOAT_TO_UBYTE(clear[2], color[2]); + CLAMPED_FLOAT_TO_UBYTE(clear[3], color[3]); - intel->ClearColor = INTEL_PACKCOLOR(screen->fbFormat, - intel->clear_red, - intel->clear_green, - intel->clear_blue, - intel->clear_alpha); + /* compute both 32 and 16-bit clear values */ + intel->ClearColor8888 = INTEL_PACKCOLOR8888(clear[0], clear[1], + clear[2], clear[3]); + intel->ClearColor565 = INTEL_PACKCOLOR565(clear[0], clear[1], clear[2]); } -static void intelCalcViewport( GLcontext *ctx ) +/** + * Update the viewport transformation matrix. Depends on: + * - viewport pos/size + * - depthrange + * - window pos/size or FBO size + */ +static void +intelCalcViewport(GLcontext * ctx) { - intelContextPtr intel = INTEL_CONTEXT(ctx); + struct intel_context *intel = intel_context(ctx); const GLfloat *v = ctx->Viewport._WindowMap.m; + const GLfloat depthScale = 1.0F / ctx->DrawBuffer->_DepthMaxF; GLfloat *m = intel->ViewportMatrix.m; - GLint h = 0; + GLfloat yScale, yBias; - if (intel->driDrawable) - h = intel->driDrawable->h + SUBPIXEL_Y; + if (ctx->DrawBuffer->Name) { + /* User created FBO */ + struct intel_renderbuffer *irb + = intel_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0][0]); + if (irb && !irb->RenderToTexture) { + /* y=0=top */ + yScale = -1.0; + yBias = irb->Base.Height; + } + else { + /* y=0=bottom */ + yScale = 1.0; + yBias = 0.0; + } + } + else { + /* window buffer, y=0=top */ + yScale = -1.0; + yBias = (intel->driDrawable) ? intel->driDrawable->h : 0.0F; + } + + m[MAT_SX] = v[MAT_SX]; + m[MAT_TX] = v[MAT_TX] + SUBPIXEL_X; - /* See also intel_translate_vertex. SUBPIXEL adjustments can be done - * via state vars, too. - */ - m[MAT_SX] = v[MAT_SX]; - m[MAT_TX] = v[MAT_TX] + SUBPIXEL_X; - m[MAT_SY] = - v[MAT_SY]; - m[MAT_TY] = - v[MAT_TY] + h; - m[MAT_SZ] = v[MAT_SZ] * intel->depth_scale; - m[MAT_TZ] = v[MAT_TZ] * intel->depth_scale; + m[MAT_SY] = v[MAT_SY] * yScale; + m[MAT_TY] = v[MAT_TY] * yScale + yBias + SUBPIXEL_Y; + + m[MAT_SZ] = v[MAT_SZ] * depthScale; + m[MAT_TZ] = v[MAT_TZ] * depthScale; } -static void intelViewport( GLcontext *ctx, - GLint x, GLint y, - GLsizei width, GLsizei height ) +static void +intelViewport(GLcontext * ctx, + GLint x, GLint y, GLsizei width, GLsizei height) { - intelCalcViewport( ctx ); + intelCalcViewport(ctx); } -static void intelDepthRange( GLcontext *ctx, - GLclampd nearval, GLclampd farval ) +static void +intelDepthRange(GLcontext * ctx, GLclampd nearval, GLclampd farval) { - intelCalcViewport( ctx ); + intelCalcViewport(ctx); } /* Fallback to swrast for select and feedback. */ -static void intelRenderMode( GLcontext *ctx, GLenum mode ) +static void +intelRenderMode(GLcontext * ctx, GLenum mode) { - intelContextPtr intel = INTEL_CONTEXT(ctx); - FALLBACK( intel, INTEL_FALLBACK_RENDERMODE, (mode != GL_RENDER) ); + struct intel_context *intel = intel_context(ctx); + FALLBACK(intel, INTEL_FALLBACK_RENDERMODE, (mode != GL_RENDER)); } -void intelInitStateFuncs( struct dd_function_table *functions ) +void +intelInitStateFuncs(struct dd_function_table *functions) { - functions->DrawBuffer = intelDrawBuffer; - functions->ReadBuffer = intelReadBuffer; functions->RenderMode = intelRenderMode; functions->Viewport = intelViewport; functions->DepthRange = intelDepthRange; functions->ClearColor = intelClearColor; } - diff --git a/src/mesa/drivers/dri/i915/intel_structs.h b/src/mesa/drivers/dri/i915/intel_structs.h new file mode 100644 index 0000000000..522e3bd92c --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_structs.h @@ -0,0 +1,132 @@ +#ifndef INTEL_STRUCTS_H +#define INTEL_STRUCTS_H + +struct br0 { + GLuint length:8; + GLuint pad0:3; + GLuint dst_tiled:1; + GLuint pad1:8; + GLuint write_rgb:1; + GLuint write_alpha:1; + GLuint opcode:7; + GLuint client:3; +}; + + +struct br13 { + GLint dest_pitch:16; + GLuint rop:8; + GLuint color_depth:2; + GLuint pad1:3; + GLuint mono_source_transparency:1; + GLuint clipping_enable:1; + GLuint pad0:1; +}; + + + +/* This is an attempt to move some of the 2D interaction in this + * driver to using structs for packets rather than a bunch of #defines + * and dwords. + */ +struct xy_color_blit { + struct br0 br0; + struct br13 br13; + + struct { + GLuint dest_x1:16; + GLuint dest_y1:16; + } dw2; + + struct { + GLuint dest_x2:16; + GLuint dest_y2:16; + } dw3; + + GLuint dest_base_addr; + GLuint color; +}; + +struct xy_src_copy_blit { + struct br0 br0; + struct br13 br13; + + struct { + GLuint dest_x1:16; + GLuint dest_y1:16; + } dw2; + + struct { + GLuint dest_x2:16; + GLuint dest_y2:16; + } dw3; + + GLuint dest_base_addr; + + struct { + GLuint src_x1:16; + GLuint src_y1:16; + } dw5; + + struct { + GLint src_pitch:16; + GLuint pad:16; + } dw6; + + GLuint src_base_addr; +}; + +struct xy_setup_blit { + struct br0 br0; + struct br13 br13; + + struct { + GLuint clip_x1:16; + GLuint clip_y1:16; + } dw2; + + struct { + GLuint clip_x2:16; + GLuint clip_y2:16; + } dw3; + + GLuint dest_base_addr; + GLuint background_color; + GLuint foreground_color; + GLuint pattern_base_addr; +}; + + +struct xy_text_immediate_blit { + struct { + GLuint length:8; + GLuint pad2:3; + GLuint dst_tiled:1; + GLuint pad1:4; + GLuint byte_packed:1; + GLuint pad0:5; + GLuint opcode:7; + GLuint client:3; + } dw0; + + struct { + GLuint dest_x1:16; + GLuint dest_y1:16; + } dw1; + + struct { + GLuint dest_x2:16; + GLuint dest_y2:16; + } dw2; + + /* Src bitmap data follows as inline dwords. + */ +}; + + +#define CLIENT_2D 0x2 +#define OPCODE_XY_SETUP_BLT 0x1 +#define OPCODE_XY_COLOR_BLT 0x50 +#define OPCODE_XY_TEXT_IMMEDIATE_BLT 0x31 + +#endif diff --git a/src/mesa/drivers/dri/i915/intel_tex.c b/src/mesa/drivers/dri/i915/intel_tex.c index 5bd280652a..b08dee43bc 100644 --- a/src/mesa/drivers/dri/i915/intel_tex.c +++ b/src/mesa/drivers/dri/i915/intel_tex.c @@ -1,877 +1,192 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "glheader.h" -#include "mtypes.h" -#include "imports.h" -#include "macros.h" -#include "simple_list.h" -#include "enums.h" -#include "image.h" -#include "texstore.h" -#include "texformat.h" -#include "teximage.h" -#include "texmem.h" #include "texobj.h" -#include "swrast/swrast.h" - -#include "mm.h" - -#include "intel_screen.h" -#include "intel_batchbuffer.h" #include "intel_context.h" +#include "intel_mipmap_tree.h" #include "intel_tex.h" -#include "intel_ioctl.h" - +#define FILE_DEBUG_FLAG DEBUG_TEXTURE static GLboolean -intelValidateClientStorage( intelContextPtr intel, GLenum target, - GLint internalFormat, - GLint srcWidth, GLint srcHeight, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) - +intelIsTextureResident(GLcontext * ctx, struct gl_texture_object *texObj) { - GLcontext *ctx = &intel->ctx; - int texelBytes; - - if (0) - fprintf(stderr, "intformat %s format %s type %s\n", - _mesa_lookup_enum_by_nr( internalFormat ), - _mesa_lookup_enum_by_nr( format ), - _mesa_lookup_enum_by_nr( type )); - - if (!ctx->Unpack.ClientStorage) - return 0; - - if (ctx->_ImageTransferState || - texImage->IsCompressed || - texObj->GenerateMipmap) - return 0; - - - /* This list is incomplete - */ - switch ( internalFormat ) { - case GL_RGBA: - if ( format == GL_BGRA && type == GL_UNSIGNED_INT_8_8_8_8_REV ) { - texImage->TexFormat = &_mesa_texformat_argb8888; - texelBytes = 4; - } - else - return 0; - break; - - case GL_RGB: - if ( format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5 ) { - texImage->TexFormat = &_mesa_texformat_rgb565; - texelBytes = 2; - } - else - return 0; - break; - - case GL_YCBCR_MESA: - if ( format == GL_YCBCR_MESA && - type == GL_UNSIGNED_SHORT_8_8_REV_APPLE ) { - texImage->TexFormat = &_mesa_texformat_ycbcr_rev; - texelBytes = 2; - } - else if ( format == GL_YCBCR_MESA && - (type == GL_UNSIGNED_SHORT_8_8_APPLE || - type == GL_UNSIGNED_BYTE)) { - texImage->TexFormat = &_mesa_texformat_ycbcr; - texelBytes = 2; - } - else - return 0; - break; - - - default: - return 0; - } +#if 0 + struct intel_context *intel = intel_context(ctx); + struct intel_texture_object *intelObj = intel_texture_object(texObj); - /* Could deal with these packing issues, but currently don't: - */ - if (packing->SkipPixels || - packing->SkipRows || - packing->SwapBytes || - packing->LsbFirst) { - return 0; - } - - { - GLint srcRowStride = _mesa_image_row_stride(packing, srcWidth, - format, type); - - - if (0) - fprintf(stderr, "%s: srcRowStride %d/%x\n", - __FUNCTION__, srcRowStride, srcRowStride); - - /* Could check this later in upload, pitch restrictions could be - * relaxed, but would need to store the image pitch somewhere, - * as packing details might change before image is uploaded: - */ - if (!intelIsAgpMemory( intel, pixels, srcHeight * srcRowStride ) || - (srcRowStride & 63)) - return 0; - - - /* Have validated that _mesa_transfer_teximage would be a straight - * memcpy at this point. NOTE: future calls to TexSubImage will - * overwrite the client data. This is explicitly mentioned in the - * extension spec. - */ - texImage->Data = (void *)pixels; - texImage->IsClientData = GL_TRUE; - texImage->RowStride = srcRowStride / texelBytes; - return 1; - } + return + intelObj->mt && + intelObj->mt->region && + intel_is_region_resident(intel, intelObj->mt->region); +#endif + return 1; } - - -static void intelTexImage1D( GLcontext *ctx, GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint border, - GLenum format, GLenum type, const GLvoid *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ) -{ - driTextureObject * t = (driTextureObject *) texObj->DriverData; - assert(t); - intelFlush( ctx ); - driSwapOutTextureObject( t ); - texImage->IsClientData = GL_FALSE; - - _mesa_store_teximage1d( ctx, target, level, internalFormat, - width, border, format, type, - pixels, packing, texObj, texImage ); - - t->dirty_images[0] |= (1 << level); +static struct gl_texture_image * +intelNewTextureImage(GLcontext * ctx) +{ + DBG("%s\n", __FUNCTION__); + (void) ctx; + return (struct gl_texture_image *) CALLOC_STRUCT(intel_texture_image); } -static void intelTexSubImage1D( GLcontext *ctx, - GLenum target, - GLint level, - GLint xoffset, - GLsizei width, - GLenum format, GLenum type, - const GLvoid *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ) + +static struct gl_texture_object * +intelNewTextureObject(GLcontext * ctx, GLuint name, GLenum target) { - driTextureObject * t = (driTextureObject *) texObj->DriverData; + struct intel_texture_object *obj = CALLOC_STRUCT(intel_texture_object); - assert(t); - intelFlush( ctx ); - driSwapOutTextureObject( t ); + DBG("%s\n", __FUNCTION__); + _mesa_initialize_texture_object(&obj->base, name, target); - _mesa_store_texsubimage1d(ctx, target, level, xoffset, width, - format, type, pixels, packing, texObj, - texImage); + return &obj->base; } - -/* Handles 2D, CUBE, RECT: - */ -static void intelTexImage2D( GLcontext *ctx, GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint border, - GLenum format, GLenum type, const GLvoid *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ) +static void +intelDeleteTextureObject(GLcontext *ctx, + struct gl_texture_object *texObj) { - driTextureObject * t = (driTextureObject *) texObj->DriverData; - GLuint face; - - /* which cube face or ordinary 2D image */ - switch (target) { - case GL_TEXTURE_CUBE_MAP_POSITIVE_X: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_X: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Y: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Z: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z: - face = (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X; - ASSERT(face < 6); - break; - default: - face = 0; - } + struct intel_context *intel = intel_context(ctx); + struct intel_texture_object *intelObj = intel_texture_object(texObj); - assert(t); - intelFlush( ctx ); - driSwapOutTextureObject( t ); - texImage->IsClientData = GL_FALSE; - - if (intelValidateClientStorage( INTEL_CONTEXT(ctx), target, - internalFormat, - width, height, - format, type, pixels, - packing, texObj, texImage)) { - if (INTEL_DEBUG & DEBUG_TEXTURE) - fprintf(stderr, "%s: Using client storage\n", __FUNCTION__); - } - else { - _mesa_store_teximage2d( ctx, target, level, internalFormat, - width, height, border, format, type, - pixels, packing, texObj, texImage ); + if (intelObj->mt) + intel_miptree_release(intel, &intelObj->mt); - t->dirty_images[face] |= (1 << level); - } + _mesa_delete_texture_object(ctx, texObj); } -static void intelTexSubImage2D( GLcontext *ctx, - GLenum target, - GLint level, - GLint xoffset, GLint yoffset, - GLsizei width, GLsizei height, - GLenum format, GLenum type, - const GLvoid *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ) -{ - driTextureObject * t = (driTextureObject *) texObj->DriverData; - GLuint face; - - /* which cube face or ordinary 2D image */ - switch (target) { - case GL_TEXTURE_CUBE_MAP_POSITIVE_X: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_X: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Y: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Z: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z: - face = (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X; - ASSERT(face < 6); - break; - default: - face = 0; - } - - if (texImage->IsClientData && - (char *)pixels == (char *)texImage->Data + - ((xoffset + yoffset * texImage->RowStride) * - texImage->TexFormat->TexelBytes)) { - /* Notification only - no upload required */ - } - else { - assert( t ); /* this _should_ be true */ - intelFlush( ctx ); - driSwapOutTextureObject( t ); +static void +intelFreeTextureImageData(GLcontext * ctx, struct gl_texture_image *texImage) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_texture_image *intelImage = intel_texture_image(texImage); - _mesa_store_texsubimage2d(ctx, target, level, xoffset, yoffset, width, - height, format, type, pixels, packing, texObj, - texImage); + DBG("%s\n", __FUNCTION__); - t->dirty_images[face] |= (1 << level); + if (intelImage->mt) { + intel_miptree_release(intel, &intelImage->mt); } -} -static void intelCompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint border, - GLsizei imageSize, const GLvoid *data, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ) -{ - driTextureObject * t = (driTextureObject *) texObj->DriverData; - GLuint face; - - /* which cube face or ordinary 2D image */ - switch (target) { - case GL_TEXTURE_CUBE_MAP_POSITIVE_X: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_X: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Y: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Z: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z: - face = (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X; - ASSERT(face < 6); - break; - default: - face = 0; + if (texImage->Data) { + free(texImage->Data); + texImage->Data = NULL; } - - assert(t); - intelFlush( ctx ); - - driSwapOutTextureObject( t ); - texImage->IsClientData = GL_FALSE; - - if (INTEL_DEBUG & DEBUG_TEXTURE) - fprintf(stderr, "%s: Using normal storage\n", __FUNCTION__); - - _mesa_store_compressed_teximage2d(ctx, target, level, internalFormat, width, - height, border, imageSize, data, texObj, texImage); - - t->dirty_images[face] |= (1 << level); } -static void intelCompressedTexSubImage2D( GLcontext *ctx, GLenum target, GLint level, - GLint xoffset, GLint yoffset, - GLsizei width, GLsizei height, - GLenum format, - GLsizei imageSize, const GLvoid *data, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ) +/* The system memcpy (at least on ubuntu 5.10) has problems copying + * to agp (writecombined) memory from a source which isn't 64-byte + * aligned - there is a 4x performance falloff. + * + * The x86 __memcpy is immune to this but is slightly slower + * (10%-ish) than the system memcpy. + * + * The sse_memcpy seems to have a slight cliff at 64/32 bytes, but + * isn't much faster than x86_memcpy for agp copies. + * + * TODO: switch dynamically. + */ +static void * +do_memcpy(void *dest, const void *src, size_t n) { - driTextureObject * t = (driTextureObject *) texObj->DriverData; - GLuint face; - - - /* which cube face or ordinary 2D image */ - switch (target) { - case GL_TEXTURE_CUBE_MAP_POSITIVE_X: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_X: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Y: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Z: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z: - face = (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X; - ASSERT(face < 6); - break; - default: - face = 0; + if ((((unsigned) src) & 63) || (((unsigned) dest) & 63)) { + return __memcpy(dest, src, n); } - - assert( t ); /* this _should_ be true */ - intelFlush( ctx ); - driSwapOutTextureObject( t ); - - _mesa_store_compressed_texsubimage2d(ctx, target, level, xoffset, yoffset, width, - height, format, imageSize, data, texObj, texImage); - - t->dirty_images[face] |= (1 << level); + else + return memcpy(dest, src, n); } -static void intelTexImage3D( GLcontext *ctx, GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint depth, - GLint border, - GLenum format, GLenum type, const GLvoid *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ) -{ - driTextureObject * t = (driTextureObject *) texObj->DriverData; - - assert(t); - driSwapOutTextureObject( t ); - texImage->IsClientData = GL_FALSE; - - _mesa_store_teximage3d(ctx, target, level, internalFormat, - width, height, depth, border, - format, type, pixels, - &ctx->Unpack, texObj, texImage); - - t->dirty_images[0] |= (1 << level); -} - +#if DO_DEBUG -static void -intelTexSubImage3D( GLcontext *ctx, GLenum target, GLint level, - GLint xoffset, GLint yoffset, GLint zoffset, - GLsizei width, GLsizei height, GLsizei depth, - GLenum format, GLenum type, - const GLvoid *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ) +#ifndef __x86_64__ +static unsigned +fastrdtsc(void) { - driTextureObject * t = (driTextureObject *) texObj->DriverData; - - assert( t ); /* this _should_ be true */ - driSwapOutTextureObject( t ); + unsigned eax; + __asm__ volatile ("\t" + "pushl %%ebx\n\t" + "cpuid\n\t" ".byte 0x0f, 0x31\n\t" + "popl %%ebx\n":"=a" (eax) + :"0"(0) + :"ecx", "edx", "cc"); - _mesa_store_texsubimage3d(ctx, target, level, xoffset, yoffset, zoffset, - width, height, depth, - format, type, pixels, packing, texObj, texImage); - - t->dirty_images[0] |= (1 << level); + return eax; } - - - - -static void intelDeleteTexture( GLcontext *ctx, struct gl_texture_object *tObj ) +#else +static unsigned +fastrdtsc(void) { - driTextureObject * t = (driTextureObject *) tObj->DriverData; + unsigned eax; + __asm__ volatile ("\t" "cpuid\n\t" ".byte 0x0f, 0x31\n\t":"=a" (eax) + :"0"(0) + :"ecx", "edx", "ebx", "cc"); - if ( t != NULL ) { - intelFlush( ctx ); - driDestroyTextureObject( t ); - } - - /* Free mipmap images and the texture object itself */ - _mesa_delete_texture_object(ctx, tObj); + return eax; } +#endif - -static const struct gl_texture_format * -intelChooseTextureFormat( GLcontext *ctx, GLint internalFormat, - GLenum format, GLenum type ) +static unsigned +time_diff(unsigned t, unsigned t2) { - intelContextPtr intel = INTEL_CONTEXT( ctx ); - const GLboolean do32bpt = ( intel->intelScreen->cpp == 4 && - intel->intelScreen->tex.size > 4*1024*1024); - - switch ( internalFormat ) { - case 4: - case GL_RGBA: - case GL_COMPRESSED_RGBA: - if ( format == GL_BGRA ) { - if ( type == GL_UNSIGNED_INT_8_8_8_8_REV ) { - return &_mesa_texformat_argb8888; - } - else if ( type == GL_UNSIGNED_SHORT_4_4_4_4_REV ) { - return &_mesa_texformat_argb4444; - } - else if ( type == GL_UNSIGNED_SHORT_1_5_5_5_REV ) { - return &_mesa_texformat_argb1555; - } - } - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; - - case 3: - case GL_RGB: - case GL_COMPRESSED_RGB: - if ( format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5 ) { - return &_mesa_texformat_rgb565; - } - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_rgb565; - - case GL_RGBA8: - case GL_RGB10_A2: - case GL_RGBA12: - case GL_RGBA16: - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; - - case GL_RGBA4: - case GL_RGBA2: - return &_mesa_texformat_argb4444; - - case GL_RGB5_A1: - return &_mesa_texformat_argb1555; - - case GL_RGB8: - case GL_RGB10: - case GL_RGB12: - case GL_RGB16: - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_rgb565; - - case GL_RGB5: - case GL_RGB4: - case GL_R3_G3_B2: - return &_mesa_texformat_rgb565; - - case GL_ALPHA: - case GL_ALPHA4: - case GL_ALPHA8: - case GL_ALPHA12: - case GL_ALPHA16: - case GL_COMPRESSED_ALPHA: - return &_mesa_texformat_a8; - - case 1: - case GL_LUMINANCE: - case GL_LUMINANCE4: - case GL_LUMINANCE8: - case GL_LUMINANCE12: - case GL_LUMINANCE16: - case GL_COMPRESSED_LUMINANCE: - return &_mesa_texformat_l8; - - case 2: - case GL_LUMINANCE_ALPHA: - case GL_LUMINANCE4_ALPHA4: - case GL_LUMINANCE6_ALPHA2: - case GL_LUMINANCE8_ALPHA8: - case GL_LUMINANCE12_ALPHA4: - case GL_LUMINANCE12_ALPHA12: - case GL_LUMINANCE16_ALPHA16: - case GL_COMPRESSED_LUMINANCE_ALPHA: - return &_mesa_texformat_al88; - - case GL_INTENSITY: - case GL_INTENSITY4: - case GL_INTENSITY8: - case GL_INTENSITY12: - case GL_INTENSITY16: - case GL_COMPRESSED_INTENSITY: - return &_mesa_texformat_i8; - - case GL_YCBCR_MESA: - if (type == GL_UNSIGNED_SHORT_8_8_MESA || - type == GL_UNSIGNED_BYTE) - return &_mesa_texformat_ycbcr; - else - return &_mesa_texformat_ycbcr_rev; - - case GL_COMPRESSED_RGB_FXT1_3DFX: - return &_mesa_texformat_rgb_fxt1; - case GL_COMPRESSED_RGBA_FXT1_3DFX: - return &_mesa_texformat_rgba_fxt1; - - case GL_RGB_S3TC: - case GL_RGB4_S3TC: - case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: - return &_mesa_texformat_rgb_dxt1; - - case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: - return &_mesa_texformat_rgba_dxt1; - - case GL_RGBA_S3TC: - case GL_RGBA4_S3TC: - case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: - return &_mesa_texformat_rgba_dxt3; - - case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: - return &_mesa_texformat_rgba_dxt5; - - case GL_DEPTH_COMPONENT: - case GL_DEPTH_COMPONENT16: - case GL_DEPTH_COMPONENT24: - case GL_DEPTH_COMPONENT32: - return &_mesa_texformat_z16; - - default: - fprintf(stderr, "unexpected texture format %s in %s\n", - _mesa_lookup_enum_by_nr(internalFormat), - __FUNCTION__); - return NULL; - } - - return NULL; /* never get here */ + return ((t < t2) ? t2 - t : 0xFFFFFFFFU - (t - t2 - 1)); } - -void intelDestroyTexObj(intelContextPtr intel, intelTextureObjectPtr t) +static void * +timed_memcpy(void *dest, const void *src, size_t n) { - unsigned i; + void *ret; + unsigned t1, t2; + double rate; - if ( intel == NULL ) - return; + if ((((unsigned) src) & 63) || (((unsigned) dest) & 63)) + _mesa_printf("Warning - non-aligned texture copy!\n"); - if ( t->age > intel->dirtyAge ) - intel->dirtyAge = t->age; + t1 = fastrdtsc(); + ret = do_memcpy(dest, src, n); + t2 = fastrdtsc(); - for ( i = 0 ; i < MAX_TEXTURE_UNITS ; i++ ) { - if ( t == intel->CurrentTexObj[ i ] ) - intel->CurrentTexObj[ i ] = NULL; - } + rate = time_diff(t1, t2); + rate /= (double) n; + _mesa_printf("timed_memcpy: %u %u --> %f clocks/byte\n", t1, t2, rate); + return ret; } +#endif /* DO_DEBUG */ - -/* Upload an image from mesa's internal copy. Image may be 1D, 2D or - * 3D. Cubemaps are expanded elsewhere. - */ -static void intelUploadTexImage( intelContextPtr intel, - intelTextureObjectPtr t, - const struct gl_texture_image *image, - const GLuint offset ) +void +intelInitTextureFuncs(struct dd_function_table *functions) { + functions->ChooseTextureFormat = intelChooseTextureFormat; + functions->TexImage1D = intelTexImage1D; + functions->TexImage2D = intelTexImage2D; + functions->TexImage3D = intelTexImage3D; + functions->TexSubImage1D = intelTexSubImage1D; + functions->TexSubImage2D = intelTexSubImage2D; + functions->TexSubImage3D = intelTexSubImage3D; + functions->CopyTexImage1D = intelCopyTexImage1D; + functions->CopyTexImage2D = intelCopyTexImage2D; + functions->CopyTexSubImage1D = intelCopyTexSubImage1D; + functions->CopyTexSubImage2D = intelCopyTexSubImage2D; + functions->GetTexImage = intelGetTexImage; - if (!image || !image->Data) - return; - - if (image->Depth == 1 && image->IsClientData) { - if (INTEL_DEBUG & DEBUG_TEXTURE) - fprintf(stderr, "Blit uploading\n"); - - /* Do it with a blit. - */ - intelEmitCopyBlitLocked( intel, - image->TexFormat->TexelBytes, - image->RowStride, /* ? */ - intelGetMemoryOffsetMESA( NULL, 0, image->Data ), - t->Pitch / image->TexFormat->TexelBytes, - intelGetMemoryOffsetMESA( NULL, 0, t->BufAddr + offset ), - 0, 0, - 0, 0, - image->Width, - image->Height); - } - else if (image->IsCompressed) { - GLuint row_len = 0; - GLubyte *dst = (GLubyte *)(t->BufAddr + offset); - GLubyte *src = (GLubyte *)image->Data; - GLuint j; - - /* must always copy whole blocks (8/16 bytes) */ - switch (image->InternalFormat) { - case GL_COMPRESSED_RGB_FXT1_3DFX: - case GL_COMPRESSED_RGBA_FXT1_3DFX: - case GL_RGB_S3TC: - case GL_RGB4_S3TC: - case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: - case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: - row_len = (image->Width * 2 + 7) & ~7; - break; - case GL_RGBA_S3TC: - case GL_RGBA4_S3TC: - case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: - case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: - row_len = (image->Width * 4 + 15) & ~15; - break; - default: - fprintf(stderr,"Internal Compressed format not supported %d\n", image->InternalFormat); - break; - } - - if (INTEL_DEBUG & DEBUG_TEXTURE) - fprintf(stderr, - "Upload image %dx%dx%d offset %xm row_len %x " - "pitch %x depth_pitch %x\n", - image->Width, image->Height, image->Depth, offset, - row_len, t->Pitch, t->depth_pitch); - - if (row_len) { - for (j = 0 ; j < (image->Height + 3)/4 ; j++, dst += (t->Pitch)) { - __memcpy(dst, src, row_len ); - src += row_len; - } - } - } - /* Time for another vtbl entry: - */ - else if (intel->intelScreen->deviceID == PCI_CHIP_I945_G || - intel->intelScreen->deviceID == PCI_CHIP_I945_GM || - intel->intelScreen->deviceID == PCI_CHIP_I945_GME || - intel->intelScreen->deviceID == PCI_CHIP_G33_G || - intel->intelScreen->deviceID == PCI_CHIP_Q33_G || - intel->intelScreen->deviceID == PCI_CHIP_Q35_G) { - GLuint row_len = image->Width * image->TexFormat->TexelBytes; - GLubyte *dst = (GLubyte *)(t->BufAddr + offset); - GLubyte *src = (GLubyte *)image->Data; - GLuint d, j; - - if (INTEL_DEBUG & DEBUG_TEXTURE) - fprintf(stderr, - "Upload image %dx%dx%d offset %xm row_len %x " - "pitch %x depth_pitch %x\n", - image->Width, image->Height, image->Depth, offset, - row_len, t->Pitch, t->depth_pitch); - - if (row_len == t->Pitch) { - memcpy( dst, src, row_len * image->Height * image->Depth ); - } - else { - GLuint x = 0, y = 0; - - for (d = 0 ; d < image->Depth ; d++) { - GLubyte *dst0 = dst + x + y * t->Pitch; - - for (j = 0 ; j < image->Height ; j++) { - __memcpy(dst0, src, row_len ); - src += row_len; - dst0 += t->Pitch; - } - - x += MIN2(4, row_len); /* Guess: 4 byte minimum alignment */ - if (x > t->Pitch) { - x = 0; - y += image->Height; - } - } - } - - } - else { - GLuint row_len = image->Width * image->TexFormat->TexelBytes; - GLubyte *dst = (GLubyte *)(t->BufAddr + offset); - GLubyte *src = (GLubyte *)image->Data; - GLuint d, j; - - if (INTEL_DEBUG & DEBUG_TEXTURE) - fprintf(stderr, - "Upload image %dx%dx%d offset %xm row_len %x " - "pitch %x depth_pitch %x\n", - image->Width, image->Height, image->Depth, offset, - row_len, t->Pitch, t->depth_pitch); - - if (row_len == t->Pitch) { - for (d = 0; d < image->Depth; d++) { - memcpy( dst, src, t->Pitch * image->Height ); - dst += t->depth_pitch; - src += row_len * image->Height; - } - } - else { - for (d = 0 ; d < image->Depth ; d++) { - for (j = 0 ; j < image->Height ; j++) { - __memcpy(dst, src, row_len ); - src += row_len; - dst += t->Pitch; - } - - dst += t->depth_pitch - (t->Pitch * image->Height); - } - } - } -} - - - -int intelUploadTexImages( intelContextPtr intel, - intelTextureObjectPtr t, - GLuint face) -{ - const int numLevels = t->base.lastLevel - t->base.firstLevel + 1; - const struct gl_texture_image *firstImage = t->image[face][t->base.firstLevel].image; - int pitch = firstImage->RowStride * firstImage->TexFormat->TexelBytes; - - /* Can we texture out of the existing client data? */ - if ( numLevels == 1 && - firstImage->IsClientData && - (pitch & 3) == 0) { - - if (INTEL_DEBUG & DEBUG_TEXTURE) - fprintf(stderr, "AGP texturing from client memory\n"); - - t->TextureOffset = intelAgpOffsetFromVirtual( intel, firstImage->Data ); - t->BufAddr = 0; - t->dirty = ~0; - return GL_TRUE; - } - else { - if (INTEL_DEBUG & DEBUG_TEXTURE) - fprintf(stderr, "Uploading client data to agp\n"); - - INTEL_FIREVERTICES( intel ); - LOCK_HARDWARE( intel ); - - if ( t->base.memBlock == NULL ) { - int heap; - - heap = driAllocateTexture( intel->texture_heaps, intel->nr_heaps, - (driTextureObject *) t ); - if ( heap == -1 ) { - UNLOCK_HARDWARE( intel ); - return GL_FALSE; - } - - /* Set the base offset of the texture image */ - t->BufAddr = (GLubyte *) (intel->intelScreen->tex.map + - t->base.memBlock->ofs); - t->TextureOffset = intel->intelScreen->tex.offset + t->base.memBlock->ofs; - t->dirty = ~0; - } - - - /* Let the world know we've used this memory recently. - */ - driUpdateTextureLRU( (driTextureObject *) t ); - - - /* Upload any images that are new */ - if (t->base.dirty_images[face]) { - int i; - - intelWaitForIdle( intel ); - - for (i = 0 ; i < numLevels ; i++) { - int level = i + t->base.firstLevel; - - if (t->base.dirty_images[face] & (1<<level)) { - - const struct gl_texture_image *image = t->image[face][i].image; - GLuint offset = t->image[face][i].offset; - - if (INTEL_DEBUG & DEBUG_TEXTURE) - fprintf(stderr, "upload level %d, offset %x\n", - level, offset); - - intelUploadTexImage( intel, t, image, offset ); - } - } - t->base.dirty_images[face] = 0; - intel->perf_boxes |= I830_BOX_TEXTURE_LOAD; - } - - UNLOCK_HARDWARE( intel ); - return GL_TRUE; - } -} + /* compressed texture functions */ + functions->CompressedTexImage2D = intelCompressedTexImage2D; + functions->GetCompressedTexImage = intelGetCompressedTexImage; -/** - * Allocate a new texture object. - * Called via ctx->Driver.NewTextureObject. - * Note: this function will be called during context creation to - * allocate the default texture objects. - * Note: we could use containment here to 'derive' the driver-specific - * texture object from the core mesa gl_texture_object. Not done at this time. - */ -static struct gl_texture_object * -intelNewTextureObject( GLcontext *ctx, GLuint name, GLenum target ) -{ - struct gl_texture_object *obj = _mesa_new_texture_object(ctx, name, target); - INTEL_CONTEXT(ctx)->vtbl.alloc_tex_obj( obj ); - return obj; -} + functions->NewTextureObject = intelNewTextureObject; + functions->NewTextureImage = intelNewTextureImage; + functions->DeleteTexture = intelDeleteTextureObject; + functions->FreeTexImageData = intelFreeTextureImageData; + functions->UpdateTexturePalette = 0; + functions->IsTextureResident = intelIsTextureResident; - -void intelInitTextureFuncs( struct dd_function_table *functions ) -{ - functions->NewTextureObject = intelNewTextureObject; - functions->ChooseTextureFormat = intelChooseTextureFormat; - functions->TexImage1D = intelTexImage1D; - functions->TexImage2D = intelTexImage2D; - functions->TexImage3D = intelTexImage3D; - functions->TexSubImage1D = intelTexSubImage1D; - functions->TexSubImage2D = intelTexSubImage2D; - functions->TexSubImage3D = intelTexSubImage3D; - functions->CopyTexImage1D = _swrast_copy_teximage1d; - functions->CopyTexImage2D = _swrast_copy_teximage2d; - functions->CopyTexSubImage1D = _swrast_copy_texsubimage1d; - functions->CopyTexSubImage2D = _swrast_copy_texsubimage2d; - functions->CopyTexSubImage3D = _swrast_copy_texsubimage3d; - functions->DeleteTexture = intelDeleteTexture; - functions->UpdateTexturePalette = NULL; - functions->IsTextureResident = driIsTextureResident; - functions->TestProxyTexImage = _mesa_test_proxy_teximage; - functions->DeleteTexture = intelDeleteTexture; - functions->CompressedTexImage2D = intelCompressedTexImage2D; - functions->CompressedTexSubImage2D = intelCompressedTexSubImage2D; +#if DO_DEBUG + if (INTEL_DEBUG & DEBUG_BUFMGR) + functions->TextureMemCpy = timed_memcpy; + else +#endif + functions->TextureMemCpy = do_memcpy; } diff --git a/src/mesa/drivers/dri/i915/intel_tex.h b/src/mesa/drivers/dri/i915/intel_tex.h index 9b7e550232..b77d7a1d8a 100644 --- a/src/mesa/drivers/dri/i915/intel_tex.h +++ b/src/mesa/drivers/dri/i915/intel_tex.h @@ -33,13 +33,119 @@ #include "texmem.h" -void intelInitTextureFuncs( struct dd_function_table *functions ); +void intelInitTextureFuncs(struct dd_function_table *functions); -void intelDestroyTexObj( intelContextPtr intel, intelTextureObjectPtr t ); -int intelUploadTexImages( intelContextPtr intel, intelTextureObjectPtr t, - GLuint face ); +const struct gl_texture_format *intelChooseTextureFormat(GLcontext * ctx, + GLint internalFormat, + GLenum format, + GLenum type); + + +void intelTexImage3D(GLcontext * ctx, + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint depth, + GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); + +void intelTexSubImage3D(GLcontext * ctx, + GLenum target, + GLint level, + GLint xoffset, GLint yoffset, GLint zoffset, + GLsizei width, GLsizei height, GLsizei depth, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); + +void intelTexImage2D(GLcontext * ctx, + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); + +void intelTexSubImage2D(GLcontext * ctx, + GLenum target, + GLint level, + GLint xoffset, GLint yoffset, + GLsizei width, GLsizei height, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); + +void intelTexImage1D(GLcontext * ctx, + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); + +void intelTexSubImage1D(GLcontext * ctx, + GLenum target, + GLint level, + GLint xoffset, + GLsizei width, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); + +void intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, + GLenum internalFormat, + GLint x, GLint y, GLsizei width, GLint border); + +void intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, + GLenum internalFormat, + GLint x, GLint y, GLsizei width, GLsizei height, + GLint border); + +void intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, + GLint xoffset, GLint x, GLint y, GLsizei width); + +void intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, + GLint xoffset, GLint yoffset, + GLint x, GLint y, GLsizei width, GLsizei height); + +void intelGetTexImage(GLcontext * ctx, GLenum target, GLint level, + GLenum format, GLenum type, GLvoid * pixels, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); + +void intelCompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint border, + GLsizei imageSize, const GLvoid *data, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage ); + +void intelGetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level, + GLvoid *pixels, + const struct gl_texture_object *texObj, + const struct gl_texture_image *texImage); + +void intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname, + unsigned long long offset, GLint depth, GLuint pitch); + +GLuint intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit); + +void intel_tex_map_images(struct intel_context *intel, + struct intel_texture_object *intelObj); + +void intel_tex_unmap_images(struct intel_context *intel, + struct intel_texture_object *intelObj); + +int intel_compressed_num_bytes(GLuint mesaFormat); -GLboolean -intel_driReinitTextureHeap( driTexHeap *heap, - unsigned size ); #endif diff --git a/src/mesa/drivers/dri/i915/intel_tex_copy.c b/src/mesa/drivers/dri/i915/intel_tex_copy.c new file mode 100644 index 0000000000..b85a25642a --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_tex_copy.c @@ -0,0 +1,302 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "mtypes.h" +#include "enums.h" +#include "image.h" +#include "teximage.h" +#include "swrast/swrast.h" + +#include "intel_screen.h" +#include "intel_context.h" +#include "intel_batchbuffer.h" +#include "intel_buffers.h" +#include "intel_mipmap_tree.h" +#include "intel_regions.h" +#include "intel_fbo.h" +#include "intel_tex.h" +#include "intel_blit.h" +#include "intel_pixel.h" + +#define FILE_DEBUG_FLAG DEBUG_TEXTURE + +/** + * Get the intel_region which is the source for any glCopyTex[Sub]Image call. + * + * Do the best we can using the blitter. A future project is to use + * the texture engine and fragment programs for these copies. + */ +static const struct intel_region * +get_teximage_source(struct intel_context *intel, GLenum internalFormat) +{ + struct intel_renderbuffer *irb; + + DBG("%s %s\n", __FUNCTION__, + _mesa_lookup_enum_by_nr(internalFormat)); + + switch (internalFormat) { + case GL_DEPTH_COMPONENT: + case GL_DEPTH_COMPONENT16_ARB: + irb = intel_get_renderbuffer(intel->ctx.ReadBuffer, BUFFER_DEPTH); + if (irb && irb->region && irb->region->cpp == 2) + return irb->region; + return NULL; + case GL_DEPTH24_STENCIL8_EXT: + case GL_DEPTH_STENCIL_EXT: + irb = intel_get_renderbuffer(intel->ctx.ReadBuffer, BUFFER_DEPTH); + if (irb && irb->region && irb->region->cpp == 4) + return irb->region; + return NULL; + case GL_RGBA: + case GL_RGBA8: + return intel_readbuf_region(intel); + case GL_RGB: + if (intel->intelScreen->cpp == 2) + return intel_readbuf_region(intel); + return NULL; + default: + return NULL; + } +} + + +static GLboolean +do_copy_texsubimage(struct intel_context *intel, + struct intel_texture_image *intelImage, + GLenum internalFormat, + GLint dstx, GLint dsty, + GLint x, GLint y, GLsizei width, GLsizei height) +{ + GLcontext *ctx = &intel->ctx; + const struct intel_region *src = + get_teximage_source(intel, internalFormat); + + if (!intelImage->mt || !src) { + DBG("%s fail %p %p\n", __FUNCTION__, intelImage->mt, src); + return GL_FALSE; + } + + intelFlush(ctx); + LOCK_HARDWARE(intel); + { + GLuint image_offset = intel_miptree_image_offset(intelImage->mt, + intelImage->face, + intelImage->level); + const GLint orig_x = x; + const GLint orig_y = y; + const struct gl_framebuffer *fb = ctx->DrawBuffer; + + if (_mesa_clip_to_region(fb->_Xmin, fb->_Ymin, fb->_Xmax, fb->_Ymax, + &x, &y, &width, &height)) { + /* Update dst for clipped src. Need to also clip the source rect. + */ + dstx += x - orig_x; + dsty += y - orig_y; + + if (ctx->ReadBuffer->Name == 0) { + /* reading from a window, adjust x, y */ + __DRIdrawablePrivate *dPriv = intel->driDrawable; + GLuint window_y; + /* window_y = position of window on screen if y=0=bottom */ + window_y = intel->intelScreen->height - (dPriv->y + dPriv->h); + y = window_y + y; + x += dPriv->x; + } + else { + /* reading from a FBO */ + /* invert Y */ + y = ctx->ReadBuffer->Height - y - 1; + } + + + /* A bit of fiddling to get the blitter to work with -ve + * pitches. But we get a nice inverted blit this way, so it's + * worth it: + */ + intelEmitCopyBlit(intel, + intelImage->mt->cpp, + -src->pitch, + src->buffer, + src->height * src->pitch * src->cpp, + intelImage->mt->pitch, + intelImage->mt->region->buffer, + image_offset, + x, y + height, dstx, dsty, width, height, + GL_COPY); /* ? */ + + intel_batchbuffer_flush(intel->batch); + } + } + + + UNLOCK_HARDWARE(intel); + +#if 0 + /* GL_SGIS_generate_mipmap -- this can be accelerated now. + * XXX Add a ctx->Driver.GenerateMipmaps() function? + */ + if (level == texObj->BaseLevel && texObj->GenerateMipmap) { + intel_generate_mipmap(ctx, target, + &ctx->Texture.Unit[ctx->Texture.CurrentUnit], + texObj); + } +#endif + + return GL_TRUE; +} + + + + + +void +intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, + GLenum internalFormat, + GLint x, GLint y, GLsizei width, GLint border) +{ + struct gl_texture_unit *texUnit = + &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + struct gl_texture_object *texObj = + _mesa_select_tex_object(ctx, texUnit, target); + struct gl_texture_image *texImage = + _mesa_select_tex_image(ctx, texObj, target, level); + + if (border) + goto fail; + + /* Setup or redefine the texture object, mipmap tree and texture + * image. Don't populate yet. + */ + ctx->Driver.TexImage1D(ctx, target, level, internalFormat, + width, border, + GL_RGBA, CHAN_TYPE, NULL, + &ctx->DefaultPacking, texObj, texImage); + + if (!do_copy_texsubimage(intel_context(ctx), + intel_texture_image(texImage), + internalFormat, 0, 0, x, y, width, 1)) + goto fail; + + return; + + fail: + _swrast_copy_teximage1d(ctx, target, level, internalFormat, x, y, + width, border); +} + +void +intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, + GLenum internalFormat, + GLint x, GLint y, GLsizei width, GLsizei height, + GLint border) +{ + struct gl_texture_unit *texUnit = + &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + struct gl_texture_object *texObj = + _mesa_select_tex_object(ctx, texUnit, target); + struct gl_texture_image *texImage = + _mesa_select_tex_image(ctx, texObj, target, level); + + if (border) + goto fail; + + /* Setup or redefine the texture object, mipmap tree and texture + * image. Don't populate yet. + */ + ctx->Driver.TexImage2D(ctx, target, level, internalFormat, + width, height, border, + GL_RGBA, CHAN_TYPE, NULL, + &ctx->DefaultPacking, texObj, texImage); + + + if (!do_copy_texsubimage(intel_context(ctx), + intel_texture_image(texImage), + internalFormat, 0, 0, x, y, width, height)) + goto fail; + + return; + + fail: + _swrast_copy_teximage2d(ctx, target, level, internalFormat, x, y, + width, height, border); +} + + +void +intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, + GLint xoffset, GLint x, GLint y, GLsizei width) +{ + struct gl_texture_unit *texUnit = + &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + struct gl_texture_object *texObj = + _mesa_select_tex_object(ctx, texUnit, target); + struct gl_texture_image *texImage = + _mesa_select_tex_image(ctx, texObj, target, level); + GLenum internalFormat = texImage->InternalFormat; + + /* XXX need to check <border> as in above function? */ + + /* Need to check texture is compatible with source format. + */ + + if (!do_copy_texsubimage(intel_context(ctx), + intel_texture_image(texImage), + internalFormat, xoffset, 0, x, y, width, 1)) { + _swrast_copy_texsubimage1d(ctx, target, level, xoffset, x, y, width); + } +} + + + +void +intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, + GLint xoffset, GLint yoffset, + GLint x, GLint y, GLsizei width, GLsizei height) +{ + struct gl_texture_unit *texUnit = + &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + struct gl_texture_object *texObj = + _mesa_select_tex_object(ctx, texUnit, target); + struct gl_texture_image *texImage = + _mesa_select_tex_image(ctx, texObj, target, level); + GLenum internalFormat = texImage->InternalFormat; + + + /* Need to check texture is compatible with source format. + */ + + if (!do_copy_texsubimage(intel_context(ctx), + intel_texture_image(texImage), + internalFormat, + xoffset, yoffset, x, y, width, height)) { + + DBG("%s - fallback to swrast\n", __FUNCTION__); + + _swrast_copy_texsubimage2d(ctx, target, level, + xoffset, yoffset, x, y, width, height); + } +} diff --git a/src/mesa/drivers/dri/i915/intel_tex_format.c b/src/mesa/drivers/dri/i915/intel_tex_format.c new file mode 100644 index 0000000000..6e058dff69 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_tex_format.c @@ -0,0 +1,172 @@ +#include "intel_context.h" +#include "intel_tex.h" +#include "texformat.h" +#include "enums.h" + +/* It works out that this function is fine for all the supported + * hardware. However, there is still a need to map the formats onto + * hardware descriptors. + */ +/* Note that the i915 can actually support many more formats than + * these if we take the step of simply swizzling the colors + * immediately after sampling... + */ +const struct gl_texture_format * +intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat, + GLenum format, GLenum type) +{ + struct intel_context *intel = intel_context(ctx); + const GLboolean do32bpt = (intel->intelScreen->cpp == 4); + + switch (internalFormat) { + case 4: + case GL_RGBA: + case GL_COMPRESSED_RGBA: + if (format == GL_BGRA) { + if (type == GL_UNSIGNED_BYTE || type == GL_UNSIGNED_INT_8_8_8_8_REV) { + return &_mesa_texformat_argb8888; + } + else if (type == GL_UNSIGNED_SHORT_4_4_4_4_REV) { + return &_mesa_texformat_argb4444; + } + else if (type == GL_UNSIGNED_SHORT_1_5_5_5_REV) { + return &_mesa_texformat_argb1555; + } + } + return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; + + case 3: + case GL_RGB: + case GL_COMPRESSED_RGB: + if (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) { + return &_mesa_texformat_rgb565; + } + return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_rgb565; + + case GL_RGBA8: + case GL_RGB10_A2: + case GL_RGBA12: + case GL_RGBA16: + return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; + + case GL_RGBA4: + case GL_RGBA2: + return &_mesa_texformat_argb4444; + + case GL_RGB5_A1: + return &_mesa_texformat_argb1555; + + case GL_RGB8: + case GL_RGB10: + case GL_RGB12: + case GL_RGB16: + return &_mesa_texformat_argb8888; + + case GL_RGB5: + case GL_RGB4: + case GL_R3_G3_B2: + return &_mesa_texformat_rgb565; + + case GL_ALPHA: + case GL_ALPHA4: + case GL_ALPHA8: + case GL_ALPHA12: + case GL_ALPHA16: + case GL_COMPRESSED_ALPHA: + return &_mesa_texformat_a8; + + case 1: + case GL_LUMINANCE: + case GL_LUMINANCE4: + case GL_LUMINANCE8: + case GL_LUMINANCE12: + case GL_LUMINANCE16: + case GL_COMPRESSED_LUMINANCE: + return &_mesa_texformat_l8; + + case 2: + case GL_LUMINANCE_ALPHA: + case GL_LUMINANCE4_ALPHA4: + case GL_LUMINANCE6_ALPHA2: + case GL_LUMINANCE8_ALPHA8: + case GL_LUMINANCE12_ALPHA4: + case GL_LUMINANCE12_ALPHA12: + case GL_LUMINANCE16_ALPHA16: + case GL_COMPRESSED_LUMINANCE_ALPHA: + return &_mesa_texformat_al88; + + case GL_INTENSITY: + case GL_INTENSITY4: + case GL_INTENSITY8: + case GL_INTENSITY12: + case GL_INTENSITY16: + case GL_COMPRESSED_INTENSITY: + return &_mesa_texformat_i8; + + case GL_YCBCR_MESA: + if (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE) + return &_mesa_texformat_ycbcr; + else + return &_mesa_texformat_ycbcr_rev; + + case GL_COMPRESSED_RGB_FXT1_3DFX: + return &_mesa_texformat_rgb_fxt1; + case GL_COMPRESSED_RGBA_FXT1_3DFX: + return &_mesa_texformat_rgba_fxt1; + + case GL_RGB_S3TC: + case GL_RGB4_S3TC: + case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: + return &_mesa_texformat_rgb_dxt1; + + case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: + return &_mesa_texformat_rgba_dxt1; + + case GL_RGBA_S3TC: + case GL_RGBA4_S3TC: + case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: + return &_mesa_texformat_rgba_dxt3; + + case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: + return &_mesa_texformat_rgba_dxt5; + + case GL_DEPTH_COMPONENT: + case GL_DEPTH_COMPONENT16: + case GL_DEPTH_COMPONENT24: + case GL_DEPTH_COMPONENT32: + return &_mesa_texformat_z16; + + case GL_DEPTH_STENCIL_EXT: + case GL_DEPTH24_STENCIL8_EXT: + return &_mesa_texformat_z24_s8; + + default: + fprintf(stderr, "unexpected texture format %s in %s\n", + _mesa_lookup_enum_by_nr(internalFormat), __FUNCTION__); + return NULL; + } + + return NULL; /* never get here */ +} + +int intel_compressed_num_bytes(GLuint mesaFormat) +{ + int bytes = 0; + switch(mesaFormat) { + + case MESA_FORMAT_RGB_FXT1: + case MESA_FORMAT_RGBA_FXT1: + case MESA_FORMAT_RGB_DXT1: + case MESA_FORMAT_RGBA_DXT1: + bytes = 2; + break; + + case MESA_FORMAT_RGBA_DXT3: + case MESA_FORMAT_RGBA_DXT5: + bytes = 4; + default: + break; + } + + return bytes; +} diff --git a/src/mesa/drivers/dri/i915/intel_tex_image.c b/src/mesa/drivers/dri/i915/intel_tex_image.c new file mode 100644 index 0000000000..197cf35ebe --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_tex_image.c @@ -0,0 +1,690 @@ + +#include <stdlib.h> +#include <stdio.h> + +#include "glheader.h" +#include "macros.h" +#include "mtypes.h" +#include "enums.h" +#include "colortab.h" +#include "convolve.h" +#include "context.h" +#include "simple_list.h" +#include "texcompress.h" +#include "texformat.h" +#include "texobj.h" +#include "texstore.h" + +#include "intel_context.h" +#include "intel_mipmap_tree.h" +#include "intel_buffer_objects.h" +#include "intel_batchbuffer.h" +#include "intel_tex.h" +#include "intel_ioctl.h" +#include "intel_blit.h" + +#define FILE_DEBUG_FLAG DEBUG_TEXTURE + +/* Functions to store texture images. Where possible, mipmap_tree's + * will be created or further instantiated with image data, otherwise + * images will be stored in malloc'd memory. A validation step is + * required to pull those images into a mipmap tree, or otherwise + * decide a fallback is required. + */ + + +static int +logbase2(int n) +{ + GLint i = 1; + GLint log2 = 0; + + while (n > i) { + i *= 2; + log2++; + } + + return log2; +} + + +/* Otherwise, store it in memory if (Border != 0) or (any dimension == + * 1). + * + * Otherwise, if max_level >= level >= min_level, create tree with + * space for textures from min_level down to max_level. + * + * Otherwise, create tree with space for textures from (level + * 0)..(1x1). Consider pruning this tree at a validation if the + * saving is worth it. + */ +static void +guess_and_alloc_mipmap_tree(struct intel_context *intel, + struct intel_texture_object *intelObj, + struct intel_texture_image *intelImage) +{ + GLuint firstLevel; + GLuint lastLevel; + GLuint width = intelImage->base.Width; + GLuint height = intelImage->base.Height; + GLuint depth = intelImage->base.Depth; + GLuint l2width, l2height, l2depth; + GLuint i, comp_byte = 0; + + DBG("%s\n", __FUNCTION__); + + if (intelImage->base.Border) + return; + + if (intelImage->level > intelObj->base.BaseLevel && + (intelImage->base.Width == 1 || + (intelObj->base.Target != GL_TEXTURE_1D && + intelImage->base.Height == 1) || + (intelObj->base.Target == GL_TEXTURE_3D && + intelImage->base.Depth == 1))) + return; + + /* If this image disrespects BaseLevel, allocate from level zero. + * Usually BaseLevel == 0, so it's unlikely to happen. + */ + if (intelImage->level < intelObj->base.BaseLevel) + firstLevel = 0; + else + firstLevel = intelObj->base.BaseLevel; + + + /* Figure out image dimensions at start level. + */ + for (i = intelImage->level; i > firstLevel; i--) { + width <<= 1; + if (height != 1) + height <<= 1; + if (depth != 1) + depth <<= 1; + } + + /* Guess a reasonable value for lastLevel. This is probably going + * to be wrong fairly often and might mean that we have to look at + * resizable buffers, or require that buffers implement lazy + * pagetable arrangements. + */ + if ((intelObj->base.MinFilter == GL_NEAREST || + intelObj->base.MinFilter == GL_LINEAR) && + intelImage->level == firstLevel) { + lastLevel = firstLevel; + } + else { + l2width = logbase2(width); + l2height = logbase2(height); + l2depth = logbase2(depth); + lastLevel = firstLevel + MAX2(MAX2(l2width, l2height), l2depth); + } + + assert(!intelObj->mt); + if (intelImage->base.IsCompressed) + comp_byte = intel_compressed_num_bytes(intelImage->base.TexFormat->MesaFormat); + intelObj->mt = intel_miptree_create(intel, + intelObj->base.Target, + intelImage->base.InternalFormat, + firstLevel, + lastLevel, + width, + height, + depth, + intelImage->base.TexFormat->TexelBytes, + comp_byte); + + DBG("%s - success\n", __FUNCTION__); +} + + + + +static GLuint +target_to_face(GLenum target) +{ + switch (target) { + case GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB: + case GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB: + case GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB: + return ((GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X); + default: + return 0; + } +} + +/* There are actually quite a few combinations this will work for, + * more than what I've listed here. + */ +static GLboolean +check_pbo_format(GLint internalFormat, + GLenum format, GLenum type, + const struct gl_texture_format *mesa_format) +{ + switch (internalFormat) { + case 4: + case GL_RGBA: + return (format == GL_BGRA && + (type == GL_UNSIGNED_BYTE || + type == GL_UNSIGNED_INT_8_8_8_8_REV) && + mesa_format == &_mesa_texformat_argb8888); + case 3: + case GL_RGB: + return (format == GL_RGB && + type == GL_UNSIGNED_SHORT_5_6_5 && + mesa_format == &_mesa_texformat_rgb565); + case GL_YCBCR_MESA: + return (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE); + default: + return GL_FALSE; + } +} + + +/* XXX: Do this for TexSubImage also: + */ +static GLboolean +try_pbo_upload(struct intel_context *intel, + struct intel_texture_image *intelImage, + const struct gl_pixelstore_attrib *unpack, + GLint internalFormat, + GLint width, GLint height, + GLenum format, GLenum type, const void *pixels) +{ + struct intel_buffer_object *pbo = intel_buffer_object(unpack->BufferObj); + GLuint src_offset, src_stride; + GLuint dst_offset, dst_stride; + + if (!pbo || + intel->ctx._ImageTransferState || + unpack->SkipPixels || unpack->SkipRows) { + _mesa_printf("%s: failure 1\n", __FUNCTION__); + return GL_FALSE; + } + + src_offset = (GLuint) pixels; + + if (unpack->RowLength > 0) + src_stride = unpack->RowLength; + else + src_stride = width; + + dst_offset = intel_miptree_image_offset(intelImage->mt, + intelImage->face, + intelImage->level); + + dst_stride = intelImage->mt->pitch; + + intelFlush(&intel->ctx); + LOCK_HARDWARE(intel); + { + dri_bo *src_buffer = intel_bufferobj_buffer(intel, pbo, INTEL_READ); + dri_bo *dst_buffer = intel_region_buffer(intel->intelScreen, + intelImage->mt->region, + INTEL_WRITE_FULL); + + + intelEmitCopyBlit(intel, + intelImage->mt->cpp, + src_stride, src_buffer, src_offset, + dst_stride, dst_buffer, dst_offset, + 0, 0, 0, 0, width, height, + GL_COPY); + + intel_batchbuffer_flush(intel->batch); + } + UNLOCK_HARDWARE(intel); + + return GL_TRUE; +} + + + +static GLboolean +try_pbo_zcopy(struct intel_context *intel, + struct intel_texture_image *intelImage, + const struct gl_pixelstore_attrib *unpack, + GLint internalFormat, + GLint width, GLint height, + GLenum format, GLenum type, const void *pixels) +{ + struct intel_buffer_object *pbo = intel_buffer_object(unpack->BufferObj); + GLuint src_offset, src_stride; + GLuint dst_offset, dst_stride; + + if (!pbo || + intel->ctx._ImageTransferState || + unpack->SkipPixels || unpack->SkipRows) { + _mesa_printf("%s: failure 1\n", __FUNCTION__); + return GL_FALSE; + } + + src_offset = (GLuint) pixels; + + if (unpack->RowLength > 0) + src_stride = unpack->RowLength; + else + src_stride = width; + + dst_offset = intel_miptree_image_offset(intelImage->mt, + intelImage->face, + intelImage->level); + + dst_stride = intelImage->mt->pitch; + + if (src_stride != dst_stride || dst_offset != 0 || src_offset != 0) { + _mesa_printf("%s: failure 2\n", __FUNCTION__); + return GL_FALSE; + } + + intel_region_attach_pbo(intel->intelScreen, intelImage->mt->region, pbo); + + return GL_TRUE; +} + + + + + + +static void +intelTexImage(GLcontext * ctx, + GLint dims, + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint depth, + GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *unpack, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage, GLsizei imageSize, int compressed) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_texture_object *intelObj = intel_texture_object(texObj); + struct intel_texture_image *intelImage = intel_texture_image(texImage); + GLint postConvWidth = width; + GLint postConvHeight = height; + GLint texelBytes, sizeInBytes; + GLuint dstRowStride; + + + DBG("%s target %s level %d %dx%dx%d border %d\n", __FUNCTION__, + _mesa_lookup_enum_by_nr(target), level, width, height, depth, border); + + intelFlush(ctx); + + intelImage->face = target_to_face(target); + intelImage->level = level; + + if (ctx->_ImageTransferState & IMAGE_CONVOLUTION_BIT) { + _mesa_adjust_image_for_convolution(ctx, dims, &postConvWidth, + &postConvHeight); + } + + /* choose the texture format */ + texImage->TexFormat = intelChooseTextureFormat(ctx, internalFormat, + format, type); + + _mesa_set_fetch_functions(texImage, dims); + + if (texImage->TexFormat->TexelBytes == 0) { + /* must be a compressed format */ + texelBytes = 0; + texImage->IsCompressed = GL_TRUE; + texImage->CompressedSize = + ctx->Driver.CompressedTextureSize(ctx, texImage->Width, + texImage->Height, texImage->Depth, + texImage->TexFormat->MesaFormat); + } else { + texelBytes = texImage->TexFormat->TexelBytes; + + /* Minimum pitch of 32 bytes */ + if (postConvWidth * texelBytes < 32) { + postConvWidth = 32 / texelBytes; + texImage->RowStride = postConvWidth; + } + + assert(texImage->RowStride == postConvWidth); + } + + /* Release the reference to a potentially orphaned buffer. + * Release any old malloced memory. + */ + if (intelImage->mt) { + intel_miptree_release(intel, &intelImage->mt); + assert(!texImage->Data); + } + else if (texImage->Data) { + _mesa_align_free(texImage->Data); + } + + /* If this is the only texture image in the tree, could call + * bmBufferData with NULL data to free the old block and avoid + * waiting on any outstanding fences. + */ + if (intelObj->mt && + intelObj->mt->first_level == level && + intelObj->mt->last_level == level && + intelObj->mt->target != GL_TEXTURE_CUBE_MAP_ARB && + !intel_miptree_match_image(intelObj->mt, &intelImage->base, + intelImage->face, intelImage->level)) { + + DBG("release it\n"); + intel_miptree_release(intel, &intelObj->mt); + assert(!intelObj->mt); + } + + if (!intelObj->mt) { + guess_and_alloc_mipmap_tree(intel, intelObj, intelImage); + if (!intelObj->mt) { + DBG("guess_and_alloc_mipmap_tree: failed\n"); + } + } + + assert(!intelImage->mt); + + if (intelObj->mt && + intel_miptree_match_image(intelObj->mt, &intelImage->base, + intelImage->face, intelImage->level)) { + + intel_miptree_reference(&intelImage->mt, intelObj->mt); + assert(intelImage->mt); + } + + if (!intelImage->mt) + DBG("XXX: Image did not fit into tree - storing in local memory!\n"); + + /* PBO fastpaths: + */ + if (dims <= 2 && + intelImage->mt && + intel_buffer_object(unpack->BufferObj) && + check_pbo_format(internalFormat, format, + type, intelImage->base.TexFormat)) { + + DBG("trying pbo upload\n"); + + /* Attempt to texture directly from PBO data (zero copy upload). + * + * Currently disable as it can lead to worse as well as better + * performance (in particular when intel_region_cow() is + * required). + */ + if (intelObj->mt == intelImage->mt && + intelObj->mt->first_level == level && + intelObj->mt->last_level == level) { + + if (try_pbo_zcopy(intel, intelImage, unpack, + internalFormat, + width, height, format, type, pixels)) { + + DBG("pbo zcopy upload succeeded\n"); + return; + } + } + + + /* Otherwise, attempt to use the blitter for PBO image uploads. + */ + if (try_pbo_upload(intel, intelImage, unpack, + internalFormat, + width, height, format, type, pixels)) { + DBG("pbo upload succeeded\n"); + return; + } + + DBG("pbo upload failed\n"); + } + + + + /* intelCopyTexImage calls this function with pixels == NULL, with + * the expectation that the mipmap tree will be set up but nothing + * more will be done. This is where those calls return: + */ + if (compressed) { + pixels = _mesa_validate_pbo_compressed_teximage(ctx, imageSize, pixels, + unpack, + "glCompressedTexImage"); + } else { + pixels = _mesa_validate_pbo_teximage(ctx, dims, width, height, 1, + format, type, + pixels, unpack, "glTexImage"); + } + if (!pixels) + return; + + + if (intelImage->mt) + intel_region_idle(intel->intelScreen, intelImage->mt->region); + + LOCK_HARDWARE(intel); + + if (intelImage->mt) { + texImage->Data = intel_miptree_image_map(intel, + intelImage->mt, + intelImage->face, + intelImage->level, + &dstRowStride, + intelImage->base.ImageOffsets); + } + else { + /* Allocate regular memory and store the image there temporarily. */ + if (texImage->IsCompressed) { + sizeInBytes = texImage->CompressedSize; + dstRowStride = + _mesa_compressed_row_stride(texImage->TexFormat->MesaFormat, width); + assert(dims != 3); + } + else { + dstRowStride = postConvWidth * texelBytes; + sizeInBytes = depth * dstRowStride * postConvHeight; + } + + texImage->Data = malloc(sizeInBytes); + } + + DBG("Upload image %dx%dx%d row_len %x " + "pitch %x\n", + width, height, depth, width * texelBytes, dstRowStride); + + /* Copy data. Would like to know when it's ok for us to eg. use + * the blitter to copy. Or, use the hardware to do the format + * conversion and copy: + */ + if (compressed) { + memcpy(texImage->Data, pixels, imageSize); + } else if (!texImage->TexFormat->StoreImage(ctx, dims, + texImage->_BaseFormat, + texImage->TexFormat, + texImage->Data, 0, 0, 0, /* dstX/Y/Zoffset */ + dstRowStride, + texImage->ImageOffsets, + width, height, depth, + format, type, pixels, unpack)) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); + } + + _mesa_unmap_teximage_pbo(ctx, unpack); + + if (intelImage->mt) { + intel_miptree_image_unmap(intel, intelImage->mt); + texImage->Data = NULL; + } + + UNLOCK_HARDWARE(intel); + +#if 0 + /* GL_SGIS_generate_mipmap -- this can be accelerated now. + */ + if (level == texObj->BaseLevel && texObj->GenerateMipmap) { + intel_generate_mipmap(ctx, target, + &ctx->Texture.Unit[ctx->Texture.CurrentUnit], + texObj); + } +#endif +} + +void +intelTexImage3D(GLcontext * ctx, + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint depth, + GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *unpack, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + intelTexImage(ctx, 3, target, level, + internalFormat, width, height, depth, border, + format, type, pixels, unpack, texObj, texImage, 0, 0); +} + + +void +intelTexImage2D(GLcontext * ctx, + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *unpack, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + intelTexImage(ctx, 2, target, level, + internalFormat, width, height, 1, border, + format, type, pixels, unpack, texObj, texImage, 0, 0); +} + +void +intelTexImage1D(GLcontext * ctx, + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *unpack, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + intelTexImage(ctx, 1, target, level, + internalFormat, width, 1, 1, border, + format, type, pixels, unpack, texObj, texImage, 0, 0); +} + +void intelCompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint border, + GLsizei imageSize, const GLvoid *data, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage ) +{ + intelTexImage(ctx, 2, target, level, + internalFormat, width, height, 1, border, + 0, 0, data, &ctx->Unpack, texObj, texImage, imageSize, 1); +} + +/** + * Need to map texture image into memory before copying image data, + * then unmap it. + */ +static void +intel_get_tex_image(GLcontext * ctx, GLenum target, GLint level, + GLenum format, GLenum type, GLvoid * pixels, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage, int compressed) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_texture_image *intelImage = intel_texture_image(texImage); + + /* Map */ + if (intelImage->mt) { + /* Image is stored in hardware format in a buffer managed by the + * kernel. Need to explicitly map and unmap it. + */ + intelImage->base.Data = + intel_miptree_image_map(intel, + intelImage->mt, + intelImage->face, + intelImage->level, + &intelImage->base.RowStride, + intelImage->base.ImageOffsets); + intelImage->base.RowStride /= intelImage->mt->cpp; + } + else { + /* Otherwise, the image should actually be stored in + * intelImage->base.Data. This is pretty confusing for + * everybody, I'd much prefer to separate the two functions of + * texImage->Data - storage for texture images in main memory + * and access (ie mappings) of images. In other words, we'd + * create a new texImage->Map field and leave Data simply for + * storage. + */ + assert(intelImage->base.Data); + } + + + if (compressed) { + _mesa_get_compressed_teximage(ctx, target, level, pixels, + texObj, texImage); + } else { + _mesa_get_teximage(ctx, target, level, format, type, pixels, + texObj, texImage); + } + + + /* Unmap */ + if (intelImage->mt) { + intel_miptree_image_unmap(intel, intelImage->mt); + intelImage->base.Data = NULL; + } +} + +void +intelGetTexImage(GLcontext * ctx, GLenum target, GLint level, + GLenum format, GLenum type, GLvoid * pixels, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + intel_get_tex_image(ctx, target, level, format, type, pixels, + texObj, texImage, 0); + + +} + +void +intelGetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level, + GLvoid *pixels, + const struct gl_texture_object *texObj, + const struct gl_texture_image *texImage) +{ + intel_get_tex_image(ctx, target, level, 0, 0, pixels, + texObj, texImage, 1); + +} + +void +intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname, + unsigned long long offset, GLint depth, GLuint pitch) +{ + struct intel_context *intel = (struct intel_context*) + ((__DRIcontextPrivate*)pDRICtx->private)->driverPrivate; + struct gl_texture_object *tObj = _mesa_lookup_texture(&intel->ctx, texname); + struct intel_texture_object *intelObj = intel_texture_object(tObj); + + if (!intelObj) + return; + + if (intelObj->mt) + intel_miptree_release(intel, &intelObj->mt); + + intelObj->imageOverride = GL_TRUE; + intelObj->depthOverride = depth; + intelObj->pitchOverride = pitch; + + if (offset) + intelObj->textureOffset = offset; +} diff --git a/src/mesa/drivers/dri/i915/intel_tex_layout.c b/src/mesa/drivers/dri/i915/intel_tex_layout.c new file mode 120000 index 0000000000..fe61b44194 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_tex_layout.c @@ -0,0 +1 @@ +../intel/intel_tex_layout.c
\ No newline at end of file diff --git a/src/mesa/drivers/dri/i915/intel_tex_subimage.c b/src/mesa/drivers/dri/i915/intel_tex_subimage.c new file mode 100644 index 0000000000..3935787806 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_tex_subimage.c @@ -0,0 +1,182 @@ + +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "mtypes.h" +#include "texobj.h" +#include "texstore.h" +#include "enums.h" + +#include "intel_context.h" +#include "intel_tex.h" +#include "intel_mipmap_tree.h" + +#define FILE_DEBUG_FLAG DEBUG_TEXTURE + +static void +intelTexSubimage(GLcontext * ctx, + GLint dims, + GLenum target, GLint level, + GLint xoffset, GLint yoffset, GLint zoffset, + GLint width, GLint height, GLint depth, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_texture_image *intelImage = intel_texture_image(texImage); + GLuint dstRowStride; + + DBG("%s target %s level %d offset %d,%d %dx%d\n", __FUNCTION__, + _mesa_lookup_enum_by_nr(target), + level, xoffset, yoffset, width, height); + + intelFlush(ctx); + + pixels = + _mesa_validate_pbo_teximage(ctx, dims, width, height, depth, format, + type, pixels, packing, "glTexSubImage2D"); + if (!pixels) + return; + + if (intelImage->mt) + intel_region_idle(intel->intelScreen, intelImage->mt->region); + + LOCK_HARDWARE(intel); + + /* Map buffer if necessary. Need to lock to prevent other contexts + * from uploading the buffer under us. + */ + if (intelImage->mt) + texImage->Data = intel_miptree_image_map(intel, + intelImage->mt, + intelImage->face, + intelImage->level, + &dstRowStride, + texImage->ImageOffsets); + + assert(dstRowStride); + + if (!texImage->TexFormat->StoreImage(ctx, dims, texImage->_BaseFormat, + texImage->TexFormat, + texImage->Data, + xoffset, yoffset, zoffset, + dstRowStride, + texImage->ImageOffsets, + width, height, depth, + format, type, pixels, packing)) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "intelTexSubImage"); + } + +#if 0 + /* GL_SGIS_generate_mipmap */ + if (level == texObj->BaseLevel && texObj->GenerateMipmap) { + _mesa_generate_mipmap(ctx, target, + &ctx->Texture.Unit[ctx->Texture.CurrentUnit], + texObj); + } +#endif + + _mesa_unmap_teximage_pbo(ctx, packing); + + if (intelImage->mt) { + intel_miptree_image_unmap(intel, intelImage->mt); + texImage->Data = NULL; + } + + UNLOCK_HARDWARE(intel); +} + + + + + +void +intelTexSubImage3D(GLcontext * ctx, + GLenum target, + GLint level, + GLint xoffset, GLint yoffset, GLint zoffset, + GLsizei width, GLsizei height, GLsizei depth, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + + intelTexSubimage(ctx, 3, + target, level, + xoffset, yoffset, zoffset, + width, height, depth, + format, type, pixels, packing, texObj, texImage); + +} + + + +void +intelTexSubImage2D(GLcontext * ctx, + GLenum target, + GLint level, + GLint xoffset, GLint yoffset, + GLsizei width, GLsizei height, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + + intelTexSubimage(ctx, 2, + target, level, + xoffset, yoffset, 0, + width, height, 1, + format, type, pixels, packing, texObj, texImage); + +} + + +void +intelTexSubImage1D(GLcontext * ctx, + GLenum target, + GLint level, + GLint xoffset, + GLsizei width, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + intelTexSubimage(ctx, 1, + target, level, + xoffset, 0, 0, + width, 1, 1, + format, type, pixels, packing, texObj, texImage); + +} diff --git a/src/mesa/drivers/dri/i915/intel_tex_validate.c b/src/mesa/drivers/dri/i915/intel_tex_validate.c new file mode 100644 index 0000000000..af18c26d55 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_tex_validate.c @@ -0,0 +1,272 @@ +#include "mtypes.h" +#include "macros.h" + +#include "intel_context.h" +#include "intel_batchbuffer.h" +#include "intel_mipmap_tree.h" +#include "intel_tex.h" + +#define FILE_DEBUG_FLAG DEBUG_TEXTURE + +/** + * Compute which mipmap levels that really need to be sent to the hardware. + * This depends on the base image size, GL_TEXTURE_MIN_LOD, + * GL_TEXTURE_MAX_LOD, GL_TEXTURE_BASE_LEVEL, and GL_TEXTURE_MAX_LEVEL. + */ +static void +intel_calculate_first_last_level(struct intel_texture_object *intelObj) +{ + struct gl_texture_object *tObj = &intelObj->base; + const struct gl_texture_image *const baseImage = + tObj->Image[0][tObj->BaseLevel]; + + /* These must be signed values. MinLod and MaxLod can be negative numbers, + * and having firstLevel and lastLevel as signed prevents the need for + * extra sign checks. + */ + int firstLevel; + int lastLevel; + + /* Yes, this looks overly complicated, but it's all needed. + */ + switch (tObj->Target) { + case GL_TEXTURE_1D: + case GL_TEXTURE_2D: + case GL_TEXTURE_3D: + case GL_TEXTURE_CUBE_MAP: + if (tObj->MinFilter == GL_NEAREST || tObj->MinFilter == GL_LINEAR) { + /* GL_NEAREST and GL_LINEAR only care about GL_TEXTURE_BASE_LEVEL. + */ + firstLevel = lastLevel = tObj->BaseLevel; + } + else { + firstLevel = tObj->BaseLevel + (GLint) (tObj->MinLod + 0.5); + firstLevel = MAX2(firstLevel, tObj->BaseLevel); + lastLevel = tObj->BaseLevel + (GLint) (tObj->MaxLod + 0.5); + lastLevel = MAX2(lastLevel, tObj->BaseLevel); + lastLevel = MIN2(lastLevel, tObj->BaseLevel + baseImage->MaxLog2); + lastLevel = MIN2(lastLevel, tObj->MaxLevel); + lastLevel = MAX2(firstLevel, lastLevel); /* need at least one level */ + } + break; + case GL_TEXTURE_RECTANGLE_NV: + case GL_TEXTURE_4D_SGIS: + firstLevel = lastLevel = 0; + break; + default: + return; + } + + /* save these values */ + intelObj->firstLevel = firstLevel; + intelObj->lastLevel = lastLevel; +} + +static void +copy_image_data_to_tree(struct intel_context *intel, + struct intel_texture_object *intelObj, + struct intel_texture_image *intelImage) +{ + if (intelImage->mt) { + /* Copy potentially with the blitter: + */ + intel_miptree_image_copy(intel, + intelObj->mt, + intelImage->face, + intelImage->level, intelImage->mt); + + intel_miptree_release(intel, &intelImage->mt); + } + else { + assert(intelImage->base.Data != NULL); + + /* More straightforward upload. + */ + intel_miptree_image_data(intel, + intelObj->mt, + intelImage->face, + intelImage->level, + intelImage->base.Data, + intelImage->base.RowStride, + intelImage->base.RowStride * + intelImage->base.Height); + _mesa_align_free(intelImage->base.Data); + intelImage->base.Data = NULL; + } + + intel_miptree_reference(&intelImage->mt, intelObj->mt); +} + + +/* + */ +GLuint +intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit) +{ + struct gl_texture_object *tObj = intel->ctx.Texture.Unit[unit]._Current; + struct intel_texture_object *intelObj = intel_texture_object(tObj); + int comp_byte = 0; + int cpp; + + GLuint face, i; + GLuint nr_faces = 0; + struct intel_texture_image *firstImage; + + GLboolean need_flush = GL_FALSE; + + /* We know/require this is true by now: + */ + assert(intelObj->base._Complete); + + /* What levels must the tree include at a minimum? + */ + intel_calculate_first_last_level(intelObj); + firstImage = + intel_texture_image(intelObj->base.Image[0][intelObj->firstLevel]); + + /* Fallback case: + */ + if (firstImage->base.Border) { + if (intelObj->mt) { + intel_miptree_release(intel, &intelObj->mt); + } + return GL_FALSE; + } + + + /* If both firstImage and intelObj have a tree which can contain + * all active images, favour firstImage. Note that because of the + * completeness requirement, we know that the image dimensions + * will match. + */ + if (firstImage->mt && + firstImage->mt != intelObj->mt && + firstImage->mt->first_level <= intelObj->firstLevel && + firstImage->mt->last_level >= intelObj->lastLevel) { + + if (intelObj->mt) + intel_miptree_release(intel, &intelObj->mt); + + intel_miptree_reference(&intelObj->mt, firstImage->mt); + } + + if (firstImage->base.IsCompressed) { + comp_byte = intel_compressed_num_bytes(firstImage->base.TexFormat->MesaFormat); + cpp = comp_byte; + } + else cpp = firstImage->base.TexFormat->TexelBytes; + + /* Check tree can hold all active levels. Check tree matches + * target, imageFormat, etc. + * + * XXX: For some layouts (eg i945?), the test might have to be + * first_level == firstLevel, as the tree isn't valid except at the + * original start level. Hope to get around this by + * programming minLod, maxLod, baseLevel into the hardware and + * leaving the tree alone. + */ + if (intelObj->mt && + (intelObj->mt->target != intelObj->base.Target || + intelObj->mt->internal_format != firstImage->base.InternalFormat || + intelObj->mt->first_level != intelObj->firstLevel || + intelObj->mt->last_level != intelObj->lastLevel || + intelObj->mt->width0 != firstImage->base.Width || + intelObj->mt->height0 != firstImage->base.Height || + intelObj->mt->depth0 != firstImage->base.Depth || + intelObj->mt->cpp != cpp || + intelObj->mt->compressed != firstImage->base.IsCompressed)) { + intel_miptree_release(intel, &intelObj->mt); + } + + + /* May need to create a new tree: + */ + if (!intelObj->mt) { + intelObj->mt = intel_miptree_create(intel, + intelObj->base.Target, + firstImage->base.InternalFormat, + intelObj->firstLevel, + intelObj->lastLevel, + firstImage->base.Width, + firstImage->base.Height, + firstImage->base.Depth, + cpp, + comp_byte); + } + + /* Pull in any images not in the object's tree: + */ + nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; + for (face = 0; face < nr_faces; face++) { + for (i = intelObj->firstLevel; i <= intelObj->lastLevel; i++) { + struct intel_texture_image *intelImage = + intel_texture_image(intelObj->base.Image[face][i]); + + /* Need to import images in main memory or held in other trees. + */ + if (intelObj->mt != intelImage->mt) { + copy_image_data_to_tree(intel, intelObj, intelImage); + need_flush = GL_TRUE; + } + } + } + + if (need_flush) + intel_batchbuffer_flush(intel->batch); + + return GL_TRUE; +} + + + +void +intel_tex_map_images(struct intel_context *intel, + struct intel_texture_object *intelObj) +{ + GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; + GLuint face, i; + + DBG("%s\n", __FUNCTION__); + + for (face = 0; face < nr_faces; face++) { + for (i = intelObj->firstLevel; i <= intelObj->lastLevel; i++) { + struct intel_texture_image *intelImage = + intel_texture_image(intelObj->base.Image[face][i]); + + if (intelImage->mt) { + intelImage->base.Data = + intel_miptree_image_map(intel, + intelImage->mt, + intelImage->face, + intelImage->level, + &intelImage->base.RowStride, + intelImage->base.ImageOffsets); + /* convert stride to texels, not bytes */ + intelImage->base.RowStride /= intelImage->mt->cpp; +/* intelImage->base.ImageStride /= intelImage->mt->cpp; */ + } + } + } +} + + + +void +intel_tex_unmap_images(struct intel_context *intel, + struct intel_texture_object *intelObj) +{ + GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; + GLuint face, i; + + for (face = 0; face < nr_faces; face++) { + for (i = intelObj->firstLevel; i <= intelObj->lastLevel; i++) { + struct intel_texture_image *intelImage = + intel_texture_image(intelObj->base.Image[face][i]); + + if (intelImage->mt) { + intel_miptree_image_unmap(intel, intelImage->mt); + intelImage->base.Data = NULL; + } + } + } +} diff --git a/src/mesa/drivers/dri/i915/intel_texmem.c b/src/mesa/drivers/dri/i915/intel_texmem.c deleted file mode 100644 index 09beec95b8..0000000000 --- a/src/mesa/drivers/dri/i915/intel_texmem.c +++ /dev/null @@ -1,72 +0,0 @@ -#include "texmem.h" -#include "simple_list.h" -#include "imports.h" -#include "macros.h" - -#include "intel_tex.h" - -static GLuint -driLog2( GLuint n ) -{ - GLuint log2; - - for ( log2 = 1 ; n > 1 ; log2++ ) { - n >>= 1; - } - - return log2; -} - -static void calculate_heap_size( driTexHeap * heap, unsigned size, - unsigned nr_regions, unsigned alignmentShift ) -{ - unsigned l; - - l = driLog2( (size - 1) / nr_regions ); - if ( l < alignmentShift ) - { - l = alignmentShift; - } - - heap->logGranularity = l; - heap->size = size & ~((1L << l) - 1); -} - - -GLboolean -intel_driReinitTextureHeap( driTexHeap *heap, - unsigned size ) -{ - driTextureObject *t, *tmp; - - /* Kick out everything: - */ - foreach_s ( t, tmp, & heap->texture_objects ) { - if ( t->tObj != NULL ) { - driSwapOutTextureObject( t ); - } - else { - driDestroyTextureObject( t ); - } - } - - /* Destroy the memory manager: - */ - mmDestroy( heap->memory_heap ); - - /* Recreate the memory manager: - */ - calculate_heap_size(heap, size, heap->nrRegions, heap->alignmentShift); - heap->memory_heap = mmInit( 0, heap->size ); - if ( heap->memory_heap == NULL ) { - fprintf(stderr, "driReinitTextureHeap: couldn't recreate memory heap\n"); - FREE( heap ); - return GL_FALSE; - } - - make_empty_list( & heap->texture_objects ); - - return GL_TRUE; -} - - diff --git a/src/mesa/drivers/dri/i915/intel_tris.c b/src/mesa/drivers/dri/i915/intel_tris.c index b2787ee60a..5fe3d4561f 100644 --- a/src/mesa/drivers/dri/i915/intel_tris.c +++ b/src/mesa/drivers/dri/i915/intel_tris.c @@ -29,6 +29,8 @@ #include "context.h" #include "macros.h" #include "enums.h" +#include "texobj.h" +#include "state.h" #include "dd.h" #include "swrast/swrast.h" @@ -38,19 +40,121 @@ #include "tnl/t_vertex.h" #include "intel_screen.h" +#include "intel_context.h" #include "intel_tris.h" #include "intel_batchbuffer.h" +#include "intel_buffers.h" #include "intel_reg.h" #include "intel_span.h" +#include "intel_tex.h" -/* XXX we shouldn't include these headers in this file, but we need them - * for fallbackStrings, below. +static void intelRenderPrimitive(GLcontext * ctx, GLenum prim); +static void intelRasterPrimitive(GLcontext * ctx, GLenum rprim, + GLuint hwprim); + +/* */ -#include "i830_context.h" -#include "i915_context.h" +static void +intel_flush_inline_primitive(struct intel_context *intel) +{ + GLuint used = intel->batch->ptr - intel->prim.start_ptr; + + assert(intel->prim.primitive != ~0); + +/* _mesa_printf("/\n"); */ + + if (used < 8) + goto do_discard; + + *(int *) intel->prim.start_ptr = (_3DPRIMITIVE | + intel->prim.primitive | (used / 4 - 2)); + + goto finished; + + do_discard: + intel->batch->ptr -= used; + + finished: + intel->prim.primitive = ~0; + intel->prim.start_ptr = 0; + intel->prim.flush = 0; +} + + +/* Emit a primitive referencing vertices in a vertex buffer. + */ +void +intelStartInlinePrimitive(struct intel_context *intel, + GLuint prim, GLuint batch_flags) +{ + BATCH_LOCALS; + + intel->vtbl.emit_state(intel); + + /* Need to make sure at the very least that we don't wrap + * batchbuffers in BEGIN_BATCH below, otherwise the primitive will + * be emitted to a batchbuffer missing the required full-state + * preamble. + */ + if (intel_batchbuffer_space(intel->batch) < 100) { + intel_batchbuffer_flush(intel->batch); + intel->vtbl.emit_state(intel); + } + +/* _mesa_printf("%s *", __progname); */ + + intel_wait_flips(intel, batch_flags); + + /* Emit a slot which will be filled with the inline primitive + * command later. + */ + BEGIN_BATCH(2, batch_flags); + OUT_BATCH(0); + + intel->prim.start_ptr = intel->batch->ptr; + intel->prim.primitive = prim; + intel->prim.flush = intel_flush_inline_primitive; + + OUT_BATCH(0); + ADVANCE_BATCH(); + +/* _mesa_printf(">"); */ +} + + +void +intelWrapInlinePrimitive(struct intel_context *intel) +{ + GLuint prim = intel->prim.primitive; + GLuint batchflags = intel->batch->flags; + + intel_flush_inline_primitive(intel); + intel_batchbuffer_flush(intel->batch); + intelStartInlinePrimitive(intel, prim, batchflags); /* ??? */ +} + +GLuint * +intelExtendInlinePrimitive(struct intel_context *intel, GLuint dwords) +{ + GLuint sz = dwords * sizeof(GLuint); + GLuint *ptr; + + assert(intel->prim.flush == intel_flush_inline_primitive); + + if (intel_batchbuffer_space(intel->batch) < sz) + intelWrapInlinePrimitive(intel); + +/* _mesa_printf("."); */ + + intel->vtbl.assert_not_dirty(intel); + + ptr = (GLuint *) intel->batch->ptr; + intel->batch->ptr += sz; + + return ptr; +} + -static void intelRenderPrimitive( GLcontext *ctx, GLenum prim ); -static void intelRasterPrimitive( GLcontext *ctx, GLenum rprim, GLuint hwprim ); /*********************************************************************** * Emit primitives as inline vertices * @@ -69,75 +173,81 @@ do { \ #else #define COPY_DWORDS( j, vb, vertsize, v ) \ do { \ - if (0) fprintf(stderr, "\n"); \ for ( j = 0 ; j < vertsize ; j++ ) { \ - if (0) fprintf(stderr, " -- v(%d): %x/%f\n",j, \ - ((GLuint *)v)[j], \ - ((GLfloat *)v)[j]); \ vb[j] = ((GLuint *)v)[j]; \ } \ vb += vertsize; \ } while (0) #endif -static void __inline__ intel_draw_quad( intelContextPtr intel, - intelVertexPtr v0, - intelVertexPtr v1, - intelVertexPtr v2, - intelVertexPtr v3 ) +static void +intel_draw_quad(struct intel_context *intel, + intelVertexPtr v0, + intelVertexPtr v1, intelVertexPtr v2, intelVertexPtr v3) { GLuint vertsize = intel->vertex_size; - GLuint *vb = intelExtendInlinePrimitive( intel, 6 * vertsize ); + GLuint *vb = intelExtendInlinePrimitive(intel, 6 * vertsize); int j; - COPY_DWORDS( j, vb, vertsize, v0 ); - COPY_DWORDS( j, vb, vertsize, v1 ); - COPY_DWORDS( j, vb, vertsize, v3 ); - COPY_DWORDS( j, vb, vertsize, v1 ); - COPY_DWORDS( j, vb, vertsize, v2 ); - COPY_DWORDS( j, vb, vertsize, v3 ); + COPY_DWORDS(j, vb, vertsize, v0); + COPY_DWORDS(j, vb, vertsize, v1); + + /* If smooth shading, draw like a trifan which gives better + * rasterization. Otherwise draw as two triangles with provoking + * vertex in third position as required for flat shading. + */ + if (intel->ctx.Light.ShadeModel == GL_FLAT) { + COPY_DWORDS(j, vb, vertsize, v3); + COPY_DWORDS(j, vb, vertsize, v1); + } + else { + COPY_DWORDS(j, vb, vertsize, v2); + COPY_DWORDS(j, vb, vertsize, v0); + } + + COPY_DWORDS(j, vb, vertsize, v2); + COPY_DWORDS(j, vb, vertsize, v3); } -static void __inline__ intel_draw_triangle( intelContextPtr intel, - intelVertexPtr v0, - intelVertexPtr v1, - intelVertexPtr v2 ) +static void +intel_draw_triangle(struct intel_context *intel, + intelVertexPtr v0, intelVertexPtr v1, intelVertexPtr v2) { GLuint vertsize = intel->vertex_size; - GLuint *vb = intelExtendInlinePrimitive( intel, 3 * vertsize ); + GLuint *vb = intelExtendInlinePrimitive(intel, 3 * vertsize); int j; - - COPY_DWORDS( j, vb, vertsize, v0 ); - COPY_DWORDS( j, vb, vertsize, v1 ); - COPY_DWORDS( j, vb, vertsize, v2 ); + + COPY_DWORDS(j, vb, vertsize, v0); + COPY_DWORDS(j, vb, vertsize, v1); + COPY_DWORDS(j, vb, vertsize, v2); } -static __inline__ void intel_draw_line( intelContextPtr intel, - intelVertexPtr v0, - intelVertexPtr v1 ) +static void +intel_draw_line(struct intel_context *intel, + intelVertexPtr v0, intelVertexPtr v1) { GLuint vertsize = intel->vertex_size; - GLuint *vb = intelExtendInlinePrimitive( intel, 2 * vertsize ); + GLuint *vb = intelExtendInlinePrimitive(intel, 2 * vertsize); int j; - COPY_DWORDS( j, vb, vertsize, v0 ); - COPY_DWORDS( j, vb, vertsize, v1 ); + COPY_DWORDS(j, vb, vertsize, v0); + COPY_DWORDS(j, vb, vertsize, v1); } -static __inline__ void intel_draw_point( intelContextPtr intel, - intelVertexPtr v0 ) +static void +intel_draw_point(struct intel_context *intel, intelVertexPtr v0) { GLuint vertsize = intel->vertex_size; - GLuint *vb = intelExtendInlinePrimitive( intel, vertsize ); + GLuint *vb = intelExtendInlinePrimitive(intel, vertsize); int j; /* Adjust for sub pixel position -- still required for conform. */ - *(float *)&vb[0] = v0->v.x - 0.125; - *(float *)&vb[1] = v0->v.y - 0.125; - for (j = 2 ; j < vertsize ; j++) - vb[j] = v0->ui[j]; + *(float *) &vb[0] = v0->v.x - 0.125; + *(float *) &vb[1] = v0->v.y - 0.125; + for (j = 2; j < vertsize; j++) + vb[j] = v0->ui[j]; } @@ -146,13 +256,17 @@ static __inline__ void intel_draw_point( intelContextPtr intel, * Fixup for ARB_point_parameters * ***********************************************************************/ -static void intel_atten_point( intelContextPtr intel, intelVertexPtr v0 ) +/* Currently not working - VERT_ATTRIB_POINTSIZE isn't correctly + * represented in the fragment program InputsRead field. + */ +static void +intel_atten_point(struct intel_context *intel, intelVertexPtr v0) { GLcontext *ctx = &intel->ctx; GLfloat psz[4], col[4], restore_psz, restore_alpha; - _tnl_get_attr( ctx, v0, _TNL_ATTRIB_POINTSIZE, psz ); - _tnl_get_attr( ctx, v0, _TNL_ATTRIB_COLOR0, col ); + _tnl_get_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz); + _tnl_get_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col); restore_psz = psz[0]; restore_alpha = col[3]; @@ -170,19 +284,19 @@ static void intel_atten_point( intelContextPtr intel, intelVertexPtr v0 ) psz[0] = 1.0; if (restore_psz != psz[0] || restore_alpha != col[3]) { - _tnl_set_attr( ctx, v0, _TNL_ATTRIB_POINTSIZE, psz); - _tnl_set_attr( ctx, v0, _TNL_ATTRIB_COLOR0, col); - - intel_draw_point( intel, v0 ); + _tnl_set_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz); + _tnl_set_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col); + + intel_draw_point(intel, v0); psz[0] = restore_psz; col[3] = restore_alpha; - _tnl_set_attr( ctx, v0, _TNL_ATTRIB_POINTSIZE, psz); - _tnl_set_attr( ctx, v0, _TNL_ATTRIB_COLOR0, col); + _tnl_set_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz); + _tnl_set_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col); } else - intel_draw_point( intel, v0 ); + intel_draw_point(intel, v0); } @@ -195,45 +309,44 @@ static void intel_atten_point( intelContextPtr intel, intelVertexPtr v0 ) -static void intel_wpos_triangle( intelContextPtr intel, - intelVertexPtr v0, - intelVertexPtr v1, - intelVertexPtr v2 ) +static void +intel_wpos_triangle(struct intel_context *intel, + intelVertexPtr v0, intelVertexPtr v1, intelVertexPtr v2) { GLuint offset = intel->wpos_offset; GLuint size = intel->wpos_size; - - __memcpy( ((char *)v0) + offset, v0, size ); - __memcpy( ((char *)v1) + offset, v1, size ); - __memcpy( ((char *)v2) + offset, v2, size ); - intel_draw_triangle( intel, v0, v1, v2 ); + __memcpy(((char *) v0) + offset, v0, size); + __memcpy(((char *) v1) + offset, v1, size); + __memcpy(((char *) v2) + offset, v2, size); + + intel_draw_triangle(intel, v0, v1, v2); } -static void intel_wpos_line( intelContextPtr intel, - intelVertexPtr v0, - intelVertexPtr v1 ) +static void +intel_wpos_line(struct intel_context *intel, + intelVertexPtr v0, intelVertexPtr v1) { GLuint offset = intel->wpos_offset; GLuint size = intel->wpos_size; - __memcpy( ((char *)v0) + offset, v0, size ); - __memcpy( ((char *)v1) + offset, v1, size ); + __memcpy(((char *) v0) + offset, v0, size); + __memcpy(((char *) v1) + offset, v1, size); - intel_draw_line( intel, v0, v1 ); + intel_draw_line(intel, v0, v1); } -static void intel_wpos_point( intelContextPtr intel, - intelVertexPtr v0 ) +static void +intel_wpos_point(struct intel_context *intel, intelVertexPtr v0) { GLuint offset = intel->wpos_offset; GLuint size = intel->wpos_size; - __memcpy( ((char *)v0) + offset, v0, size ); + __memcpy(((char *) v0) + offset, v0, size); - intel_draw_point( intel, v0 ); + intel_draw_point(intel, v0); } @@ -290,11 +403,12 @@ do { \ #define INTEL_MAX_TRIFUNC 0x10 -static struct { - tnl_points_func points; - tnl_line_func line; - tnl_triangle_func triangle; - tnl_quad_func quad; +static struct +{ + tnl_points_func points; + tnl_line_func line; + tnl_triangle_func triangle; + tnl_quad_func quad; } rast_tab[INTEL_MAX_TRIFUNC]; @@ -355,7 +469,7 @@ do { \ #define VERT_RESTORE_SPEC( idx ) if (specoffset) v[idx]->ui[specoffset] = spec[idx] #define LOCAL_VARS(n) \ - intelContextPtr intel = INTEL_CONTEXT(ctx); \ + struct intel_context *intel = intel_context(ctx); \ GLuint color[n], spec[n]; \ GLuint coloroffset = intel->coloroffset; \ GLboolean specoffset = intel->specoffset; \ @@ -366,7 +480,7 @@ do { \ * Helpers for rendering unfilled primitives * ***********************************************************************/ -static const GLuint hw_prim[GL_POLYGON+1] = { +static const GLuint hw_prim[GL_POLYGON + 1] = { PRIM3D_POINTLIST, PRIM3D_LINELIST, PRIM3D_LINELIST, @@ -456,7 +570,8 @@ static const GLuint hw_prim[GL_POLYGON+1] = { #include "tnl_dd/t_dd_tritmp.h" -static void init_rast_tab( void ) +static void +init_rast_tab(void) { init(); init_offset(); @@ -487,10 +602,8 @@ static void init_rast_tab( void ) * primitives. */ static void -intel_fallback_tri( intelContextPtr intel, - intelVertex *v0, - intelVertex *v1, - intelVertex *v2 ) +intel_fallback_tri(struct intel_context *intel, + intelVertex * v0, intelVertex * v1, intelVertex * v2) { GLcontext *ctx = &intel->ctx; SWvertex v[3]; @@ -498,19 +611,20 @@ intel_fallback_tri( intelContextPtr intel, if (0) fprintf(stderr, "\n%s\n", __FUNCTION__); - _swsetup_Translate( ctx, v0, &v[0] ); - _swsetup_Translate( ctx, v1, &v[1] ); - _swsetup_Translate( ctx, v2, &v[2] ); - intelSpanRenderStart( ctx ); - _swrast_Triangle( ctx, &v[0], &v[1], &v[2] ); - intelSpanRenderFinish( ctx ); + INTEL_FIREVERTICES(intel); + + _swsetup_Translate(ctx, v0, &v[0]); + _swsetup_Translate(ctx, v1, &v[1]); + _swsetup_Translate(ctx, v2, &v[2]); + intelSpanRenderStart(ctx); + _swrast_Triangle(ctx, &v[0], &v[1], &v[2]); + intelSpanRenderFinish(ctx); } static void -intel_fallback_line( intelContextPtr intel, - intelVertex *v0, - intelVertex *v1 ) +intel_fallback_line(struct intel_context *intel, + intelVertex * v0, intelVertex * v1) { GLcontext *ctx = &intel->ctx; SWvertex v[2]; @@ -518,17 +632,18 @@ intel_fallback_line( intelContextPtr intel, if (0) fprintf(stderr, "\n%s\n", __FUNCTION__); - _swsetup_Translate( ctx, v0, &v[0] ); - _swsetup_Translate( ctx, v1, &v[1] ); - intelSpanRenderStart( ctx ); - _swrast_Line( ctx, &v[0], &v[1] ); - intelSpanRenderFinish( ctx ); -} + INTEL_FIREVERTICES(intel); + _swsetup_Translate(ctx, v0, &v[0]); + _swsetup_Translate(ctx, v1, &v[1]); + intelSpanRenderStart(ctx); + _swrast_Line(ctx, &v[0], &v[1]); + intelSpanRenderFinish(ctx); +} static void -intel_fallback_point( intelContextPtr intel, - intelVertex *v0 ) +intel_fallback_point(struct intel_context *intel, + intelVertex * v0) { GLcontext *ctx = &intel->ctx; SWvertex v[1]; @@ -536,12 +651,13 @@ intel_fallback_point( intelContextPtr intel, if (0) fprintf(stderr, "\n%s\n", __FUNCTION__); - _swsetup_Translate( ctx, v0, &v[0] ); - intelSpanRenderStart( ctx ); - _swrast_Point( ctx, &v[0] ); - intelSpanRenderFinish( ctx ); -} + INTEL_FIREVERTICES(intel); + _swsetup_Translate(ctx, v0, &v[0]); + intelSpanRenderStart(ctx); + _swrast_Point(ctx, &v[0]); + intelSpanRenderFinish(ctx); +} /**********************************************************************/ @@ -558,7 +674,7 @@ intel_fallback_point( intelContextPtr intel, #define INIT(x) intelRenderPrimitive( ctx, x ) #undef LOCAL_VARS #define LOCAL_VARS \ - intelContextPtr intel = INTEL_CONTEXT(ctx); \ + struct intel_context *intel = intel_context(ctx); \ GLubyte *vertptr = (GLubyte *)intel->verts; \ const GLuint vertsize = intel->vertex_size; \ const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts; \ @@ -581,10 +697,10 @@ intel_fallback_point( intelContextPtr intel, -static void intelRenderClippedPoly( GLcontext *ctx, const GLuint *elts, - GLuint n ) +static void +intelRenderClippedPoly(GLcontext * ctx, const GLuint * elts, GLuint n) { - intelContextPtr intel = INTEL_CONTEXT(ctx); + struct intel_context *intel = intel_context(ctx); TNLcontext *tnl = TNL_CONTEXT(ctx); struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; GLuint prim = intel->render_primitive; @@ -593,39 +709,40 @@ static void intelRenderClippedPoly( GLcontext *ctx, const GLuint *elts, */ { GLuint *tmp = VB->Elts; - VB->Elts = (GLuint *)elts; - tnl->Driver.Render.PrimTabElts[GL_POLYGON]( ctx, 0, n, - PRIM_BEGIN|PRIM_END ); + VB->Elts = (GLuint *) elts; + tnl->Driver.Render.PrimTabElts[GL_POLYGON] (ctx, 0, n, + PRIM_BEGIN | PRIM_END); VB->Elts = tmp; } /* Restore the render primitive */ if (prim != GL_POLYGON) - tnl->Driver.Render.PrimitiveNotify( ctx, prim ); + tnl->Driver.Render.PrimitiveNotify(ctx, prim); } -static void intelRenderClippedLine( GLcontext *ctx, GLuint ii, GLuint jj ) +static void +intelRenderClippedLine(GLcontext * ctx, GLuint ii, GLuint jj) { TNLcontext *tnl = TNL_CONTEXT(ctx); - tnl->Driver.Render.Line( ctx, ii, jj ); + tnl->Driver.Render.Line(ctx, ii, jj); } -static void intelFastRenderClippedPoly( GLcontext *ctx, const GLuint *elts, - GLuint n ) +static void +intelFastRenderClippedPoly(GLcontext * ctx, const GLuint * elts, GLuint n) { - intelContextPtr intel = INTEL_CONTEXT( ctx ); + struct intel_context *intel = intel_context(ctx); const GLuint vertsize = intel->vertex_size; - GLuint *vb = intelExtendInlinePrimitive( intel, (n-2) * 3 * vertsize ); - GLubyte *vertptr = (GLubyte *)intel->verts; - const GLuint *start = (const GLuint *)V(elts[0]); - int i,j; - - for (i = 2 ; i < n ; i++) { - COPY_DWORDS( j, vb, vertsize, V(elts[i-1]) ); - COPY_DWORDS( j, vb, vertsize, V(elts[i]) ); - COPY_DWORDS( j, vb, vertsize, start ); + GLuint *vb = intelExtendInlinePrimitive(intel, (n - 2) * 3 * vertsize); + GLubyte *vertptr = (GLubyte *) intel->verts; + const GLuint *start = (const GLuint *) V(elts[0]); + int i, j; + + for (i = 2; i < n; i++) { + COPY_DWORDS(j, vb, vertsize, V(elts[i - 1])); + COPY_DWORDS(j, vb, vertsize, V(elts[i])); + COPY_DWORDS(j, vb, vertsize, start); } } @@ -636,68 +753,75 @@ static void intelFastRenderClippedPoly( GLcontext *ctx, const GLuint *elts, -#define POINT_FALLBACK (0) -#define LINE_FALLBACK (DD_LINE_STIPPLE) -#define TRI_FALLBACK (0) -#define ANY_FALLBACK_FLAGS (POINT_FALLBACK|LINE_FALLBACK|TRI_FALLBACK|\ - DD_TRI_STIPPLE|DD_POINT_ATTEN) -#define ANY_RASTER_FLAGS (DD_TRI_LIGHT_TWOSIDE|DD_TRI_OFFSET|DD_TRI_UNFILLED) +#define ANY_FALLBACK_FLAGS (DD_LINE_STIPPLE | DD_TRI_STIPPLE | DD_POINT_ATTEN | DD_POINT_SMOOTH | DD_TRI_SMOOTH) +#define ANY_RASTER_FLAGS (DD_TRI_LIGHT_TWOSIDE | DD_TRI_OFFSET | DD_TRI_UNFILLED) -void intelChooseRenderState(GLcontext *ctx) +void +intelChooseRenderState(GLcontext * ctx) { TNLcontext *tnl = TNL_CONTEXT(ctx); - intelContextPtr intel = INTEL_CONTEXT(ctx); + struct intel_context *intel = intel_context(ctx); GLuint flags = ctx->_TriangleCaps; const struct gl_fragment_program *fprog = ctx->FragmentProgram._Current; GLboolean have_wpos = (fprog && (fprog->Base.InputsRead & FRAG_BIT_WPOS)); GLuint index = 0; if (INTEL_DEBUG & DEBUG_STATE) - fprintf(stderr,"\n%s\n",__FUNCTION__); + fprintf(stderr, "\n%s\n", __FUNCTION__); - if ((flags & (ANY_FALLBACK_FLAGS|ANY_RASTER_FLAGS)) || have_wpos) { + if ((flags & (ANY_FALLBACK_FLAGS | ANY_RASTER_FLAGS)) || have_wpos) { if (flags & ANY_RASTER_FLAGS) { - if (flags & DD_TRI_LIGHT_TWOSIDE) index |= INTEL_TWOSIDE_BIT; - if (flags & DD_TRI_OFFSET) index |= INTEL_OFFSET_BIT; - if (flags & DD_TRI_UNFILLED) index |= INTEL_UNFILLED_BIT; + if (flags & DD_TRI_LIGHT_TWOSIDE) + index |= INTEL_TWOSIDE_BIT; + if (flags & DD_TRI_OFFSET) + index |= INTEL_OFFSET_BIT; + if (flags & DD_TRI_UNFILLED) + index |= INTEL_UNFILLED_BIT; } if (have_wpos) { - intel->draw_point = intel_wpos_point; - intel->draw_line = intel_wpos_line; - intel->draw_tri = intel_wpos_triangle; + intel->draw_point = intel_wpos_point; + intel->draw_line = intel_wpos_line; + intel->draw_tri = intel_wpos_triangle; - /* Make sure these get called: - */ - index |= INTEL_FALLBACK_BIT; + /* Make sure these get called: + */ + index |= INTEL_FALLBACK_BIT; } else { - intel->draw_point = intel_draw_point; - intel->draw_line = intel_draw_line; - intel->draw_tri = intel_draw_triangle; + intel->draw_point = intel_draw_point; + intel->draw_line = intel_draw_line; + intel->draw_tri = intel_draw_triangle; } /* Hook in fallbacks for specific primitives. */ - if (flags & ANY_FALLBACK_FLAGS) - { - if (flags & POINT_FALLBACK) - intel->draw_point = intel_fallback_point; - - if (flags & LINE_FALLBACK) - intel->draw_line = intel_fallback_line; - - if (flags & TRI_FALLBACK) - intel->draw_tri = intel_fallback_tri; - - if ((flags & DD_TRI_STIPPLE) && !intel->hw_stipple) - intel->draw_tri = intel_fallback_tri; - - if (flags & DD_POINT_ATTEN) - intel->draw_point = intel_atten_point; - - index |= INTEL_FALLBACK_BIT; + if (flags & ANY_FALLBACK_FLAGS) { + if (flags & DD_LINE_STIPPLE) + intel->draw_line = intel_fallback_line; + + if ((flags & DD_TRI_STIPPLE) && !intel->hw_stipple) + intel->draw_tri = intel_fallback_tri; + + if (flags & DD_TRI_SMOOTH) { + if (intel->strict_conformance) + intel->draw_tri = intel_fallback_tri; + } + + if (flags & DD_POINT_ATTEN) { + if (0) + intel->draw_point = intel_atten_point; + else + intel->draw_point = intel_fallback_point; + } + + if (flags & DD_POINT_SMOOTH) { + if (intel->strict_conformance) + intel->draw_point = intel_fallback_point; + } + + index |= INTEL_FALLBACK_BIT; } } @@ -710,20 +834,21 @@ void intelChooseRenderState(GLcontext *ctx) tnl->Driver.Render.Quad = rast_tab[index].quad; if (index == 0) { - tnl->Driver.Render.PrimTabVerts = intel_render_tab_verts; - tnl->Driver.Render.PrimTabElts = intel_render_tab_elts; - tnl->Driver.Render.ClippedLine = line; /* from tritmp.h */ - tnl->Driver.Render.ClippedPolygon = intelFastRenderClippedPoly; - } else { - tnl->Driver.Render.PrimTabVerts = _tnl_render_tab_verts; - tnl->Driver.Render.PrimTabElts = _tnl_render_tab_elts; - tnl->Driver.Render.ClippedLine = intelRenderClippedLine; - tnl->Driver.Render.ClippedPolygon = intelRenderClippedPoly; + tnl->Driver.Render.PrimTabVerts = intel_render_tab_verts; + tnl->Driver.Render.PrimTabElts = intel_render_tab_elts; + tnl->Driver.Render.ClippedLine = line; /* from tritmp.h */ + tnl->Driver.Render.ClippedPolygon = intelFastRenderClippedPoly; + } + else { + tnl->Driver.Render.PrimTabVerts = _tnl_render_tab_verts; + tnl->Driver.Render.PrimTabElts = _tnl_render_tab_elts; + tnl->Driver.Render.ClippedLine = intelRenderClippedLine; + tnl->Driver.Render.ClippedPolygon = intelRenderClippedPoly; } } } -static const GLenum reduced_prim[GL_POLYGON+1] = { +static const GLenum reduced_prim[GL_POLYGON + 1] = { GL_POINTS, GL_LINES, GL_LINES, @@ -744,35 +869,52 @@ static const GLenum reduced_prim[GL_POLYGON+1] = { -static void intelRunPipeline( GLcontext *ctx ) +static void +intelRunPipeline(GLcontext * ctx) { - intelContextPtr intel = INTEL_CONTEXT(ctx); + struct intel_context *intel = intel_context(ctx); + + _mesa_lock_context_textures(ctx); + + if (ctx->NewState) + _mesa_update_state_locked(ctx); if (intel->NewGLState) { if (intel->NewGLState & _NEW_TEXTURE) { - intel->vtbl.update_texture_state( intel ); + intel->vtbl.update_texture_state(intel); } if (!intel->Fallback) { - if (intel->NewGLState & _INTEL_NEW_RENDERSTATE) - intelChooseRenderState( ctx ); + if (intel->NewGLState & _INTEL_NEW_RENDERSTATE) + intelChooseRenderState(ctx); } intel->NewGLState = 0; } - _tnl_run_pipeline( ctx ); + _tnl_run_pipeline(ctx); + + _mesa_unlock_context_textures(ctx); } -static void intelRenderStart( GLcontext *ctx ) +static void +intelRenderStart(GLcontext * ctx) { - INTEL_CONTEXT(ctx)->vtbl.render_start( INTEL_CONTEXT(ctx) ); + struct intel_context *intel = intel_context(ctx); + + intel->vtbl.render_start(intel_context(ctx)); + intel->vtbl.emit_state(intel); } -static void intelRenderFinish( GLcontext *ctx ) +static void +intelRenderFinish(GLcontext * ctx) { - if (INTEL_CONTEXT(ctx)->RenderIndex & INTEL_FALLBACK_BIT) - _swrast_flush( ctx ); + struct intel_context *intel = intel_context(ctx); + + if (intel->RenderIndex & INTEL_FALLBACK_BIT) + _swrast_flush(ctx); + + INTEL_FIREVERTICES(intel); } @@ -781,28 +923,33 @@ static void intelRenderFinish( GLcontext *ctx ) /* System to flush dma and emit state changes based on the rasterized * primitive. */ -static void intelRasterPrimitive( GLcontext *ctx, GLenum rprim, GLuint hwprim ) +static void +intelRasterPrimitive(GLcontext * ctx, GLenum rprim, GLuint hwprim) { - intelContextPtr intel = INTEL_CONTEXT(ctx); + struct intel_context *intel = intel_context(ctx); if (0) - fprintf(stderr, "%s %s %x\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(rprim), hwprim); + fprintf(stderr, "%s %s %x\n", __FUNCTION__, + _mesa_lookup_enum_by_nr(rprim), hwprim); + + intel->vtbl.reduced_primitive_state(intel, rprim); - intel->vtbl.reduced_primitive_state( intel, rprim ); - /* Start a new primitive. Arrange to have it flushed later on. */ - if (hwprim != intel->prim.primitive) - intelStartInlinePrimitive( intel, hwprim ); + if (hwprim != intel->prim.primitive) { + INTEL_FIREVERTICES(intel); + + intelStartInlinePrimitive(intel, hwprim, INTEL_BATCH_CLIPRECTS); + } } -/* - */ -static void intelRenderPrimitive( GLcontext *ctx, GLenum prim ) + /* + */ +static void +intelRenderPrimitive(GLcontext * ctx, GLenum prim) { - intelContextPtr intel = INTEL_CONTEXT(ctx); + struct intel_context *intel = intel_context(ctx); if (0) fprintf(stderr, "%s %s\n", __FUNCTION__, _mesa_lookup_enum_by_nr(prim)); @@ -817,63 +964,54 @@ static void intelRenderPrimitive( GLcontext *ctx, GLenum prim ) * lower level functions in that case, potentially pingponging the * state: */ - if (reduced_prim[prim] == GL_TRIANGLES && + if (reduced_prim[prim] == GL_TRIANGLES && (ctx->_TriangleCaps & DD_TRI_UNFILLED)) return; /* Set some primitive-dependent state and Start? a new primitive. */ - intelRasterPrimitive( ctx, reduced_prim[prim], hw_prim[prim] ); + intelRasterPrimitive(ctx, reduced_prim[prim], hw_prim[prim]); } -/**********************************************************************/ -/* Transition to/from hardware rasterization. */ -/**********************************************************************/ - -static struct { - GLuint bit; - const char *str; -} fallbackStrings[] = { - { INTEL_FALLBACK_DRAW_BUFFER, "Draw buffer" }, - { INTEL_FALLBACK_READ_BUFFER, "Read buffer" }, - { INTEL_FALLBACK_USER, "User" }, - { INTEL_FALLBACK_NO_BATCHBUFFER, "No Batchbuffer" }, - { INTEL_FALLBACK_NO_TEXMEM, "No Texmem" }, - { INTEL_FALLBACK_RENDERMODE, "Rendermode" }, - - { I830_FALLBACK_TEXTURE, "i830 texture" }, - { I830_FALLBACK_COLORMASK, "i830 colormask" }, - { I830_FALLBACK_STENCIL, "i830 stencil" }, - { I830_FALLBACK_STIPPLE, "i830 stipple" }, - { I830_FALLBACK_LOGICOP, "i830 logicop" }, - - { I915_FALLBACK_TEXTURE, "i915 texture" }, - { I915_FALLBACK_COLORMASK, "i915 colormask" }, - { I915_FALLBACK_STENCIL, "i915 stencil" }, - { I915_FALLBACK_STIPPLE, "i915 stipple" }, - { I915_FALLBACK_PROGRAM, "i915 program" }, - { I915_FALLBACK_LOGICOP, "i915 logicop" }, - { I915_FALLBACK_POLYGON_SMOOTH, "i915 polygon smooth" }, - { I915_FALLBACK_POINT_SMOOTH, "i915 point smooth" }, - - { 0, NULL } + /**********************************************************************/ + /* Transition to/from hardware rasterization. */ + /**********************************************************************/ + +static char *fallbackStrings[] = { + [0] = "Draw buffer", + [1] = "Read buffer", + [2] = "Depth buffer", + [3] = "Stencil buffer", + [4] = "User disable", + [5] = "Render mode", + + [12] = "Texture", + [13] = "Color mask", + [14] = "Stencil", + [15] = "Stipple", + [16] = "Program", + [17] = "Logic op", + [18] = "Smooth polygon", + [19] = "Smooth point", }; -static const char * +static char * getFallbackString(GLuint bit) { - int i; - for (i = 0; fallbackStrings[i].bit; i++) { - if (fallbackStrings[i].bit == bit) - return fallbackStrings[i].str; + int i = 0; + while (bit > 1) { + i++; + bit >>= 1; } - return "unknown fallback bit"; + return fallbackStrings[i]; } -void intelFallback( intelContextPtr intel, GLuint bit, GLboolean mode ) + +void +intelFallback(struct intel_context *intel, GLuint bit, GLboolean mode) { GLcontext *ctx = &intel->ctx; TNLcontext *tnl = TNL_CONTEXT(ctx); @@ -883,20 +1021,19 @@ void intelFallback( intelContextPtr intel, GLuint bit, GLboolean mode ) intel->Fallback |= bit; if (oldfallback == 0) { intelFlush(ctx); - if (INTEL_DEBUG & DEBUG_FALLBACKS) - fprintf(stderr, "ENTER FALLBACK 0x%x: %s\n", + if (INTEL_DEBUG & DEBUG_FALLBACKS) + fprintf(stderr, "ENTER FALLBACK %x: %s\n", bit, getFallbackString(bit)); - _swsetup_Wakeup( ctx ); + _swsetup_Wakeup(ctx); intel->RenderIndex = ~0; } } else { intel->Fallback &= ~bit; if (oldfallback == bit) { - _swrast_flush( ctx ); - if (INTEL_DEBUG & DEBUG_FALLBACKS) - fprintf(stderr, "LEAVE FALLBACK 0x%x: %s\n", - bit, getFallbackString(bit)); + _swrast_flush(ctx); + if (INTEL_DEBUG & DEBUG_FALLBACKS) + fprintf(stderr, "LEAVE FALLBACK %s\n", getFallbackString(bit)); tnl->Driver.Render.Start = intelRenderStart; tnl->Driver.Render.PrimitiveNotify = intelRenderPrimitive; tnl->Driver.Render.Finish = intelRenderFinish; @@ -904,18 +1041,87 @@ void intelFallback( intelContextPtr intel, GLuint bit, GLboolean mode ) tnl->Driver.Render.CopyPV = _tnl_copy_pv; tnl->Driver.Render.Interp = _tnl_interp; - _tnl_invalidate_vertex_state( ctx, ~0 ); - _tnl_invalidate_vertices( ctx, ~0 ); - _tnl_install_attrs( ctx, - intel->vertex_attrs, - intel->vertex_attr_count, - intel->ViewportMatrix.m, 0 ); + _tnl_invalidate_vertex_state(ctx, ~0); + _tnl_invalidate_vertices(ctx, ~0); + _tnl_install_attrs(ctx, + intel->vertex_attrs, + intel->vertex_attr_count, + intel->ViewportMatrix.m, 0); intel->NewGLState |= _INTEL_NEW_RENDERSTATE; } } } +union fi +{ + GLfloat f; + GLint i; +}; + + +/**********************************************************************/ +/* Used only with the metaops callbacks. */ +/**********************************************************************/ +void +intel_meta_draw_poly(struct intel_context *intel, + GLuint n, + GLfloat xy[][2], + GLfloat z, GLuint color, GLfloat tex[][2]) +{ + union fi *vb; + GLint i; + + /* All 3d primitives should be emitted with INTEL_BATCH_CLIPRECTS, + * otherwise the drawing origin (DR4) might not be set correctly. + */ + intelStartInlinePrimitive(intel, PRIM3D_TRIFAN, INTEL_BATCH_CLIPRECTS); + vb = (union fi *) intelExtendInlinePrimitive(intel, n * 6); + + for (i = 0; i < n; i++) { + vb[0].f = xy[i][0]; + vb[1].f = xy[i][1]; + vb[2].f = z; + vb[3].i = color; + vb[4].f = tex[i][0]; + vb[5].f = tex[i][1]; + vb += 6; + } + + INTEL_FIREVERTICES(intel); +} + +void +intel_meta_draw_quad(struct intel_context *intel, + GLfloat x0, GLfloat x1, + GLfloat y0, GLfloat y1, + GLfloat z, + GLuint color, + GLfloat s0, GLfloat s1, GLfloat t0, GLfloat t1) +{ + GLfloat xy[4][2]; + GLfloat tex[4][2]; + + xy[0][0] = x0; + xy[0][1] = y0; + xy[1][0] = x1; + xy[1][1] = y0; + xy[2][0] = x1; + xy[2][1] = y1; + xy[3][0] = x0; + xy[3][1] = y1; + + tex[0][0] = s0; + tex[0][1] = t0; + tex[1][0] = s1; + tex[1][1] = t0; + tex[2][0] = s1; + tex[2][1] = t1; + tex[3][0] = s0; + tex[3][1] = t1; + + intel_meta_draw_poly(intel, 4, xy, z, color, tex); +} @@ -924,7 +1130,8 @@ void intelFallback( intelContextPtr intel, GLuint bit, GLboolean mode ) /**********************************************************************/ -void intelInitTriFuncs( GLcontext *ctx ) +void +intelInitTriFuncs(GLcontext * ctx) { TNLcontext *tnl = TNL_CONTEXT(ctx); static int firsttime = 1; diff --git a/src/mesa/drivers/dri/i915/intel_tris.h b/src/mesa/drivers/dri/i915/intel_tris.h index d7e382fdb3..b7bae8cd3b 100644 --- a/src/mesa/drivers/dri/i915/intel_tris.h +++ b/src/mesa/drivers/dri/i915/intel_tris.h @@ -30,6 +30,8 @@ #include "mtypes.h" + + #define _INTEL_NEW_RENDERSTATE (_DD_NEW_LINE_STIPPLE | \ _DD_NEW_TRI_UNFILLED | \ _DD_NEW_TRI_LIGHT_TWOSIDE | \ @@ -38,9 +40,30 @@ _NEW_PROGRAM | \ _NEW_POLYGONSTIPPLE) -extern void intelInitTriFuncs( GLcontext *ctx ); +extern void intelInitTriFuncs(GLcontext * ctx); + +extern void intelChooseRenderState(GLcontext * ctx); + +extern void intelStartInlinePrimitive(struct intel_context *intel, + GLuint prim, GLuint flags); +extern void intelWrapInlinePrimitive(struct intel_context *intel); + +GLuint *intelExtendInlinePrimitive(struct intel_context *intel, + GLuint dwords); + + +void intel_meta_draw_quad(struct intel_context *intel, + GLfloat x0, GLfloat x1, + GLfloat y0, GLfloat y1, + GLfloat z, + GLuint color, + GLfloat s0, GLfloat s1, GLfloat t0, GLfloat t1); + +void intel_meta_draw_poly(struct intel_context *intel, + GLuint n, + GLfloat xy[][2], + GLfloat z, GLuint color, GLfloat tex[][2]); + -extern void intelPrintRenderState( const char *msg, GLuint state ); -extern void intelChooseRenderState( GLcontext *ctx ); #endif diff --git a/src/mesa/drivers/dri/i915/server/i830_common.h b/src/mesa/drivers/dri/i915/server/i830_common.h index fb6ceaa52d..f1fd3939ab 100644 --- a/src/mesa/drivers/dri/i915/server/i830_common.h +++ b/src/mesa/drivers/dri/i915/server/i830_common.h @@ -52,6 +52,9 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #define DRM_I830_INIT_HEAP 0x0a #define DRM_I830_CMDBUFFER 0x0b #define DRM_I830_DESTROY_HEAP 0x0c +#define DRM_I830_SET_VBLANK_PIPE 0x0d +#define DRM_I830_GET_VBLANK_PIPE 0x0e +#define DRM_I830_MMIO 0x10 typedef struct { enum { @@ -83,6 +86,7 @@ typedef struct { int last_enqueue; /* last time a buffer was enqueued */ int last_dispatch; /* age of the most recently dispatched buffer */ int ctxOwner; /* last context to upload state */ + /** Last context that used the buffer manager. */ int texAge; int pf_enabled; /* is pageflipping allowed? */ int pf_active; @@ -119,14 +123,29 @@ typedef struct { unsigned int rotated_tiled; unsigned int rotated2_tiled; - int pipeA_x; - int pipeA_y; - int pipeA_w; - int pipeA_h; - int pipeB_x; - int pipeB_y; - int pipeB_w; - int pipeB_h; + int planeA_x; + int planeA_y; + int planeA_w; + int planeA_h; + int planeB_x; + int planeB_y; + int planeB_w; + int planeB_h; + + /* Triple buffering */ + drm_handle_t third_handle; + int third_offset; + int third_size; + unsigned int third_tiled; + + /* buffer object handles for the static buffers. May change + * over the lifetime of the client, though it doesn't in our current + * implementation. + */ + unsigned int front_bo_handle; + unsigned int back_bo_handle; + unsigned int third_bo_handle; + unsigned int depth_bo_handle; } drmI830Sarea; /* Flags for perf_boxes @@ -208,5 +227,30 @@ typedef struct { int region; } drmI830MemDestroyHeap; +#define DRM_I830_VBLANK_PIPE_A 1 +#define DRM_I830_VBLANK_PIPE_B 2 + +typedef struct { + int pipe; +} drmI830VBlankPipe; + +#define MMIO_READ 0 +#define MMIO_WRITE 1 + +#define MMIO_REGS_IA_PRIMATIVES_COUNT 0 +#define MMIO_REGS_IA_VERTICES_COUNT 1 +#define MMIO_REGS_VS_INVOCATION_COUNT 2 +#define MMIO_REGS_GS_PRIMITIVES_COUNT 3 +#define MMIO_REGS_GS_INVOCATION_COUNT 4 +#define MMIO_REGS_CL_PRIMITIVES_COUNT 5 +#define MMIO_REGS_CL_INVOCATION_COUNT 6 +#define MMIO_REGS_PS_INVOCATION_COUNT 7 +#define MMIO_REGS_PS_DEPTH_COUNT 8 + +typedef struct { + unsigned int read_write:1; + unsigned int reg:31; + void __user *data; +} drmI830MMIO; #endif /* _I830_DRM_H_ */ diff --git a/src/mesa/drivers/dri/i915/server/i830_dri.h b/src/mesa/drivers/dri/i915/server/i830_dri.h index 6c9a709021..c2a3af8cbf 100644 --- a/src/mesa/drivers/dri/i915/server/i830_dri.h +++ b/src/mesa/drivers/dri/i915/server/i830_dri.h @@ -9,8 +9,8 @@ #define I830_MAX_DRAWABLES 256 #define I830_MAJOR_VERSION 1 -#define I830_MINOR_VERSION 3 -#define I830_PATCHLEVEL 0 +#define I830_MINOR_VERSION 7 +#define I830_PATCHLEVEL 2 #define I830_REG_SIZE 0x80000 @@ -18,20 +18,20 @@ typedef struct _I830DRIRec { drm_handle_t regs; drmSize regsSize; - drmSize backbufferSize; - drm_handle_t backbuffer; + drmSize unused1; /* backbufferSize */ + drm_handle_t unused2; /* backbuffer */ - drmSize depthbufferSize; - drm_handle_t depthbuffer; + drmSize unused3; /* depthbufferSize */ + drm_handle_t unused4; /* depthbuffer */ - drmSize rotatedSize; - drm_handle_t rotatedbuffer; + drmSize unused5; /* rotatedSize */ + drm_handle_t unused6; /* rotatedbuffer */ - drm_handle_t textures; - int textureSize; + drm_handle_t unused7; /* textures */ + int unused8; /* textureSize */ - drm_handle_t agp_buffers; - drmSize agp_buf_size; + drm_handle_t unused9; /* agp_buffers */ + drmSize unused10; /* agp_buf_size */ int deviceID; int width; @@ -40,20 +40,10 @@ typedef struct _I830DRIRec { int cpp; int bitsPerPixel; - int fbOffset; - int fbStride; - - int backOffset; - int backPitch; - - int depthOffset; - int depthPitch; - - int rotatedOffset; - int rotatedPitch; - - int logTextureGranularity; - int textureOffset; + int unused11[8]; /* was front/back/depth/rotated offset/pitch */ + + int unused12; /* logTextureGranularity */ + int unused13; /* textureOffset */ int irq; int sarea_priv_offset; diff --git a/src/mesa/drivers/dri/i915/server/intel.h b/src/mesa/drivers/dri/i915/server/intel.h index d7858a20c8..6ea72499c1 100644 --- a/src/mesa/drivers/dri/i915/server/intel.h +++ b/src/mesa/drivers/dri/i915/server/intel.h @@ -75,6 +75,9 @@ #define I830_GMCH_CTRL 0x52 +#define I830_GMCH_MEM_MASK 0x1 +#define I830_GMCH_MEM_64M 0x1 +#define I830_GMCH_MEM_128M 0 #define I830_GMCH_GMS_MASK 0x70 #define I830_GMCH_GMS_DISABLED 0x00 @@ -141,7 +144,7 @@ typedef struct _I830Rec { unsigned char *MMIOBase; unsigned char *FbBase; int cpp; - + uint32_t aper_size; unsigned int bios_version; /* These are set in PreInit and never changed. */ diff --git a/src/mesa/drivers/dri/i915/server/intel_dri.c b/src/mesa/drivers/dri/i915/server/intel_dri.c index b6946b75d2..e49c4214ad 100644 --- a/src/mesa/drivers/dri/i915/server/intel_dri.c +++ b/src/mesa/drivers/dri/i915/server/intel_dri.c @@ -292,15 +292,43 @@ static void I830SetupMemoryTiling(const DRIDriverContext *ctx, I830Rec *pI830) static int I830DetectMemory(const DRIDriverContext *ctx, I830Rec *pI830) { - struct pci_device host_bridge; + struct pci_device host_bridge, ig_dev; uint32_t gmch_ctrl; int memsize = 0; int range; - + uint32_t aper_size; + uint32_t membase2 = 0; + memset(&host_bridge, 0, sizeof(host_bridge)); + memset(&ig_dev, 0, sizeof(ig_dev)); + + ig_dev.dev = 2; pci_device_cfg_read_u32(&host_bridge, &gmch_ctrl, I830_GMCH_CTRL); - + + if (IS_I830(pI830) || IS_845G(pI830)) { + if ((gmch_ctrl & I830_GMCH_MEM_MASK) == I830_GMCH_MEM_128M) { + aper_size = 0x80000000; + } else { + aper_size = 0x40000000; + } + } else { + if (IS_I9XX(pI830)) { + int ret; + ret = pci_device_cfg_read_u32(&ig_dev, &membase2, 0x18); + if (membase2 & 0x08000000) + aper_size = 0x8000000; + else + aper_size = 0x10000000; + + fprintf(stderr,"aper size is %08X %08x %d\n", aper_size, membase2, ret); + } else + aper_size = 0x8000000; + } + + pI830->aper_size = aper_size; + + /* We need to reduce the stolen size, by the GTT and the popup. * The GTT varying according the the FbMapSize and the popup is 4KB */ range = (ctx->shared.fbSize / (1024*1024)) + 4; @@ -577,7 +605,8 @@ I830AllocateMemory(const DRIDriverContext *ctx, I830Rec *pI830) fprintf(stderr,"unable to allocate context buffer %ld\n", ret); return FALSE; } - + +#if 0 memset(&(pI830->TexMem), 0, sizeof(pI830->TexMem)); pI830->TexMem.Key = -1; @@ -588,6 +617,7 @@ I830AllocateMemory(const DRIDriverContext *ctx, I830Rec *pI830) fprintf(stderr,"unable to allocate texture memory %ld\n", ret); return FALSE; } +#endif return TRUE; } @@ -605,12 +635,29 @@ I830BindMemory(const DRIDriverContext *ctx, I830Rec *pI830) return FALSE; if (!BindAgpRange(ctx, &pI830->ContextMem)) return FALSE; +#if 0 if (!BindAgpRange(ctx, &pI830->TexMem)) return FALSE; - +#endif return TRUE; } +static void SetupDRIMM(const DRIDriverContext *ctx, I830Rec *pI830) +{ + unsigned long aperEnd = ROUND_DOWN_TO(pI830->aper_size, GTT_PAGE_SIZE) / GTT_PAGE_SIZE; + unsigned long aperStart = ROUND_TO(pI830->aper_size - KB(32768), GTT_PAGE_SIZE) / GTT_PAGE_SIZE; + + fprintf(stderr, "aper size is %08X\n", ctx->shared.fbSize); + if (drmMMInit(ctx->drmFD, aperStart, aperEnd - aperStart, DRM_BO_MEM_TT)) { + fprintf(stderr, + "DRM MM Initialization Failed\n"); + } else { + fprintf(stderr, + "DRM MM Initialized at offset 0x%lx length %d page\n", aperStart, aperEnd-aperStart); + } + +} + static Bool I830CleanupDma(const DRIDriverContext *ctx) { @@ -810,6 +857,7 @@ I830DRIMapScreenRegions(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sar fprintf(stderr, "[drm] Depth Buffer = 0x%08x\n", sarea->depth_handle); +#if 0 if (drmAddMap(ctx->drmFD, (drm_handle_t)sarea->tex_offset, sarea->tex_size, DRM_AGP, 0, @@ -820,7 +868,7 @@ I830DRIMapScreenRegions(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sar } fprintf(stderr, "[drm] textures = 0x%08x\n", sarea->tex_handle); - +#endif return TRUE; } @@ -848,29 +896,6 @@ I830DRIUnmapScreenRegions(const DRIDriverContext *ctx, I830Rec *pI830, drmI830Sa } } -static void -I830InitTextureHeap(const DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) -{ - /* Start up the simple memory manager for agp space */ - drmI830MemInitHeap drmHeap; - drmHeap.region = I830_MEM_REGION_AGP; - drmHeap.start = 0; - drmHeap.size = sarea->tex_size; - - if (drmCommandWrite(ctx->drmFD, DRM_I830_INIT_HEAP, - &drmHeap, sizeof(drmHeap))) { - fprintf(stderr, - "[drm] Failed to initialized agp heap manager\n"); - } else { - fprintf(stderr, - "[drm] Initialized kernel agp heap manager, %d\n", - sarea->tex_size); - - I830SetParam(ctx, I830_SETPARAM_TEX_LRU_LOG_GRANULARITY, - sarea->log_tex_granularity); - } -} - static Bool I830DRIDoMappings(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) { @@ -892,7 +917,7 @@ I830DRIDoMappings(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) /* init to zero to be safe */ I830DRIMapScreenRegions(ctx, pI830, sarea); - I830InitTextureHeap(ctx, pI830, sarea); + SetupDRIMM(ctx, pI830); if (ctx->pciDevice != PCI_CHIP_845_G && ctx->pciDevice != PCI_CHIP_I830_M) { @@ -1084,6 +1109,10 @@ I830ScreenInit(DRIDriverContext *ctx, I830Rec *pI830) return FALSE; } + pSAREAPriv->rotated_offset = -1; + pSAREAPriv->rotated_size = 0; + pSAREAPriv->rotated_pitch = ctx->shared.virtualWidth; + pSAREAPriv->front_offset = pI830->FrontBuffer.Start; pSAREAPriv->front_size = pI830->FrontBuffer.Size; pSAREAPriv->width = ctx->shared.virtualWidth; @@ -1095,8 +1124,10 @@ I830ScreenInit(DRIDriverContext *ctx, I830Rec *pI830) pSAREAPriv->back_size = pI830->BackBuffer.Size; pSAREAPriv->depth_offset = pI830->DepthBuffer.Start; pSAREAPriv->depth_size = pI830->DepthBuffer.Size; +#if 0 pSAREAPriv->tex_offset = pI830->TexMem.Start; pSAREAPriv->tex_size = pI830->TexMem.Size; +#endif pSAREAPriv->log_tex_granularity = pI830->TexGranularity; ctx->driverClientMsg = malloc(sizeof(I830DRIRec)); @@ -1108,14 +1139,6 @@ I830ScreenInit(DRIDriverContext *ctx, I830Rec *pI830) pI830DRI->height = ctx->shared.virtualHeight; pI830DRI->mem = ctx->shared.fbSize; pI830DRI->cpp = ctx->cpp; - pI830DRI->backOffset = pI830->BackBuffer.Start; - pI830DRI->backPitch = pI830->BackBuffer.Pitch; - - pI830DRI->depthOffset = pI830->DepthBuffer.Start; - pI830DRI->depthPitch = pI830->DepthBuffer.Pitch; - - pI830DRI->fbOffset = pI830->FrontBuffer.Start; - pI830DRI->fbStride = pI830->FrontBuffer.Pitch; pI830DRI->bitsPerPixel = ctx->bpp; pI830DRI->sarea_priv_offset = sizeof(drm_sarea_t); |