diff options
Diffstat (limited to 'src/mesa/drivers/dri/i915')
64 files changed, 7250 insertions, 13172 deletions
diff --git a/src/mesa/drivers/dri/i915/Makefile b/src/mesa/drivers/dri/i915/Makefile index 805abf75e0..fd77980ebf 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 \ @@ -24,18 +14,45 @@ 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_bitmap.c \ + intel_pixel_copy.c \ + intel_pixel_draw.c \ + intel_pixel_read.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 C_SOURCES = \ $(COMMON_SOURCES) \ @@ -43,8 +60,14 @@ C_SOURCES = \ ASM_SOURCES = - +DRIVER_DEFINES = -I../intel -I../intel/server -DI915 \ + $(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..116b52511f 100644 --- a/src/mesa/drivers/dri/i915/i830_context.c +++ b/src/mesa/drivers/dri/i915/i830_context.c @@ -32,93 +32,81 @@ #include "tnl/tnl.h" #include "tnl/t_vertex.h" #include "tnl/t_context.h" +#include "tnl/t_pipeline.h" #include "utils.h" +#include "intel_span.h" +#include "intel_pixel.h" +#include "intel_tris.h" /*************************************** * Mesa's Driver Functions ***************************************/ -static const struct dri_extension i830_extensions[] = +static void +i830InitDriverFunctions(struct dd_function_table *functions) { - { "GL_ARB_texture_env_crossbar", NULL }, - { NULL, NULL } -}; - - -static void i830InitDriverFunctions( struct dd_function_table *functions ) -{ - intelInitDriverFunctions( functions ); - i830InitStateFuncs( functions ); - i830InitTextureFuncs( functions ); + intelInitDriverFunctions(functions); + i830InitStateFuncs(functions); + i830InitTextureFuncs(functions); } +extern const struct tnl_pipeline_stage *intel_pipeline[]; -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; } + /* Initialize swrast, tnl driver tables: */ + intelInitSpanFuncs(ctx); + intelInitTriFuncs(ctx); + + /* Install the customized pipeline: */ + _tnl_destroy_pipeline(ctx); + _tnl_install_pipeline(ctx, intel_pipeline); + + if (intel->no_rast) + FALLBACK(intel, INTEL_FALLBACK_USER, 1); + intel->ctx.Const.MaxTextureUnits = I830_TEX_UNITS; 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) ); - - intel->verts = TNL_CONTEXT(ctx)->clipspace.vertex_buf; + ctx->Const.MaxTextureLevels = 12; + ctx->Const.Max3DTextureLevels = 9; + ctx->Const.MaxCubeTextureLevels = 11; + ctx->Const.MaxTextureRectSize = (1 << 11); + ctx->Const.MaxTextureUnits = I830_TEX_UNITS; - driInitExtensions( ctx, i830_extensions, GL_FALSE ); + _tnl_init_vertices(ctx, ctx->Const.MaxArrayLockSize + 12, + 18 * sizeof(GLfloat)); - i830InitState( i830 ); + intel->verts = TNL_CONTEXT(ctx)->clipspace.vertex_buf; + 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..a298c1407d 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 +154,56 @@ 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, - GLuint srcBuf); +/*====================================================================== + * 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..d210c2d08e 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) @@ -494,10 +494,6 @@ #define VFT1_TEX0_FMT(x) (x) #define VFT1_TEX0_MASK 3 #define VFT1_TEX1_SHIFT 2 -#define TEXCOORDFMT_2D 0 -#define TEXCOORDFMT_3D 1 -#define TEXCOORDFMT_4D 2 -#define TEXCOORDFMT_1D 3 /*New stuff picked up along the way */ @@ -554,8 +550,8 @@ #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 +559,9 @@ #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 +569,17 @@ #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) @@ -634,8 +631,4 @@ #define ENABLE_TEX_STREAM_MAP_IDX (1<<3) #define TEX_STREAM_MAP_IDX(x) (x) - -#define MI_FLUSH ((0<<29)|(4<<23)) -#define FLUSH_MAP_CACHE (1<<0) - #endif diff --git a/src/mesa/drivers/dri/i915/i830_state.c b/src/mesa/drivers/dri/i915/i830_state.c index f7980201f9..e44a7dfca1 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; - - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "[%s] x(%d) y(%d) w(%d) h(%d)\n", __FUNCTION__, - x, y, w, h); + DBG("%s %d,%d %dx%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); - - if (INTEL_DEBUG&DEBUG_DRI) - fprintf(stderr, "%s\n", __FUNCTION__); + struct i830_context *i830 = i830_context(ctx); - 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; @@ -739,18 +760,18 @@ 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 ); + if (i830->intel.ctx.Visual.rgbBits == 16) + 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; @@ -846,206 +876,172 @@ 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; - /* Zero all state */ memset(&i830->state, 0, sizeof(i830->state)); /* 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)); - - - 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); + TEXBIND_SET2(TEXCOORDSRC_VTXSET_2) | + TEXBIND_SET1(TEXCOORDSRC_VTXSET_1) | + TEXBIND_SET0(TEXCOORDSRC_VTXSET_0)); - 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.Stipple[I830_STPREG_ST0] = _3DSTATE_STIPPLE; i830->state.Buffer[I830_DESTREG_DV0] = _3DSTATE_DST_BUF_VARS_CMD; - - 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); - 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); - break; - } - 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 +1066,21 @@ 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..79b0fcf596 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 +71,32 @@ 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..4e9b022eae 100644 --- a/src/mesa/drivers/dri/i915/i830_texstate.c +++ b/src/mesa/drivers/dri/i915/i830_texstate.c @@ -25,459 +25,327 @@ * **************************************************************************/ -#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_unit *tUnit = &ctx->Texture.Unit[unit]; + struct gl_texture_object *tObj = tUnit->_Current; + struct intel_texture_object *intelObj = intel_texture_object(tObj); + struct gl_texture_image *firstImage; + GLuint *state = i830->state.Tex[unit], format, pitch; + GLint lodbias; + 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); - - 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; + 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); - 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; } - } - - return GL_TRUE; -} + 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; + } + } + lodbias = (int) ((tUnit->LodBias + tObj->LodBias) * 16.0); + if (lodbias < -64) + lodbias = -64; + if (lodbias > 63) + lodbias = 63; + + state[I830_TEXREG_TM0S3] = ((lodbias << TM0S3_LOD_BIAS_SHIFT) & + TM0S3_LOD_BIAS_MASK); +#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..40a50ff772 100644 --- a/src/mesa/drivers/dri/i915/i830_vtbl.c +++ b/src/mesa/drivers/dri/i915/i830_vtbl.c @@ -25,17 +25,20 @@ * **************************************************************************/ +#include "glapi.h" #include "i830_context.h" #include "i830_reg.h" - #include "intel_batchbuffer.h" - +#include "intel_regions.h" +#include "intel_tris.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 +62,16 @@ 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 +79,7 @@ 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 +89,215 @@ 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, IGNORE_CLIPRECTS); OUT_BATCH(_3DSTATE_DFLT_DIFFUSE_CMD); OUT_BATCH(0); @@ -282,37 +310,35 @@ 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 +350,13 @@ 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); @@ -349,45 +367,46 @@ 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(); } #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); - -static GLuint get_state_size( struct i830_hw_state *state ) + intel_batchbuffer_data(intel->batch, state, size, IGNORE_CLIPRECTS ) + +static GLuint +get_dirty(struct i830_hw_state *state) +{ + return state->active & ~state->emitted; +} + +static GLuint +get_state_size(struct i830_hw_state *state) { - GLuint dirty = state->active & ~state->emitted; + GLuint dirty = get_dirty(state); GLuint sz = 0; GLuint i; 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 +415,312 @@ 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; + int i, count; + GLuint dirty; + GET_CURRENT_CONTEXT(ctx); BATCH_LOCALS; + dri_bo *aper_array[3 + I830_TEX_UNITS]; + int aper_count; + + /* We don't hold the lock at this point, so want to make sure that + * there won't be a buffer wrap between the state emits and the primitive + * emit header. + * + * It might be better to talk about explicit places where + * scheduling is allowed, rather than assume that it is whenever a + * batchbuffer fills up. + * + * Set the space as LOOP_CLIPRECTS now, since that's what our primitives + * will be emitted under. + */ + intel_batchbuffer_require_space(intel->batch, + get_state_size(state) + INTEL_PRIM_EMIT_SIZE, + LOOP_CLIPRECTS); + count = 0; + again: + aper_count = 0; + dirty = get_dirty(state); + + aper_array[aper_count++] = intel->batch->buf; + if (dirty & I830_UPLOAD_BUFFERS) { + aper_array[aper_count++] = state->draw_region->buffer; + aper_array[aper_count++] = state->depth_region->buffer; + } - if (intel->batch.space < get_state_size(state)) { - intelFlushBatch(intel, GL_TRUE); - dirty = state->active & ~state->emitted; - counter = intel->batch.counter; + for (i = 0; i < I830_TEX_UNITS; i++) + if (dirty & I830_UPLOAD_TEX(i)) { + if (state->tex_buffer[i]) { + aper_array[aper_count++] = state->tex_buffer[i]; + } + } + + if (dri_bufmgr_check_aperture_space(aper_array, aper_count)) { + if (count == 0) { + count++; + intel_batchbuffer_flush(intel->batch); + goto again; + } else { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "i830 emit state"); + assert(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; + assert(get_dirty(state) == 0); + 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(intel, 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, IGNORE_CLIPRECTS); + OUT_BATCH(state->Buffer[I830_DESTREG_CBUFADDR0]); + OUT_BATCH(state->Buffer[I830_DESTREG_CBUFADDR1]); + OUT_RELOC(state->draw_region->buffer, + I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER, + 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, + I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER, + 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(intel, 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, IGNORE_CLIPRECTS); + OUT_BATCH(state->Tex[i][I830_TEXREG_TM0LI]); + + if (state->tex_buffer[i]) { + OUT_RELOC(state->tex_buffer[i], + I915_GEM_DOMAIN_SAMPLER, 0, + 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(intel, state->TexBlend[i], state->TexBlendWordsUsed[i] * 4); } } - state->emitted |= dirty; - intel->batch.last_emit_state = counter; - assert(counter == intel->batch.counter); + intel->batch->dirty_state &= ~dirty; + assert(get_dirty(state) == 0); + assert((intel->batch->dirty_state & (1<<1)) == 0); } -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_regions[], + struct intel_region *depth_region, + GLuint num_regions) { - 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_regions[0], 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_new_batch(struct intel_context *intel) { - I830_CONTEXT(intel)->state.emitted = 0; + struct i830_context *i830 = i830_context(&intel->ctx); + i830->state.emitted = 0; + + /* Check that we didn't just wrap our batchbuffer at a bad time. */ + assert(!intel->no_batch_wrap); } -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; + assert(!get_dirty(state)); +} +static void +i830_note_unlock( struct intel_context *intel ) +{ + /* nothing */ +} -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.new_batch = i830_new_batch; 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; + i830->intel.vtbl.note_unlock = i830_note_unlock; + i830->intel.vtbl.finish_batch = intel_finish_vb; } diff --git a/src/mesa/drivers/dri/i915/i915_context.c b/src/mesa/drivers/dri/i915/i915_context.c index 2bc1cae9c3..940b02513d 100644 --- a/src/mesa/drivers/dri/i915/i915_context.c +++ b/src/mesa/drivers/dri/i915/i915_context.c @@ -36,151 +36,153 @@ #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" +#include "intel_tris.h" +#include "intel_span.h" +#include "intel_pixel.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_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; } +extern const struct tnl_pipeline_stage *intel_pipeline[]; -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; } + /* Initialize swrast, tnl driver tables: */ + intelInitSpanFuncs(ctx); + intelInitTriFuncs(ctx); + + /* Install the customized pipeline: */ + _tnl_destroy_pipeline(ctx); + _tnl_install_pipeline(ctx, intel_pipeline); + + if (intel->no_rast) + FALLBACK(intel, INTEL_FALLBACK_USER, 1); + ctx->Const.MaxTextureUnits = I915_TEX_UNITS; 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..c6958dd8d4 100644 --- a/src/mesa/drivers/dri/i915/i915_context.h +++ b/src/mesa/drivers/dri/i915/i915_context.h @@ -29,6 +29,7 @@ #define I915CONTEXT_INC #include "intel_context.h" +#include "i915_reg.h" #define I915_FALLBACK_TEXTURE 0x1000 #define I915_FALLBACK_COLORMASK 0x2000 @@ -46,6 +47,7 @@ #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 +57,8 @@ */ #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 +89,6 @@ #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 @@ -97,23 +96,34 @@ #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)) #define I915_PROGRAM_SIZE 192 +#define I915_MAX_INSN (I915_MAX_TEX_INSN+I915_MAX_ALU_INSN) /* 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; @@ -135,52 +145,40 @@ 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. + */ + /* Track which R registers are "live" for each instruction. + * A register is live between the time it's written to and the last time + * it's read. */ + GLuint usedRegs[I915_MAX_INSN]; + /* Helpers for i915_fragprog.c: */ 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 +186,68 @@ 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 +259,13 @@ 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 +294,58 @@ 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/intel_span.h b/src/mesa/drivers/dri/i915/i915_debug.h index 2d4f8589d0..0643a8c631 100644 --- a/src/mesa/drivers/dri/i915/intel_span.h +++ b/src/mesa/drivers/dri/i915/i915_debug.h @@ -1,8 +1,8 @@ /************************************************************************** * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * 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 @@ -25,17 +25,31 @@ * **************************************************************************/ -#ifndef _INTEL_SPAN_H -#define _INTEL_SPAN_H +/* 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; +}; + + -#include "drirenderbuffer.h" +extern void i915_disassemble_program(const unsigned *program, unsigned sz); +extern void i915_print_ureg(const char *msg, unsigned ureg); -extern void intelInitSpanFuncs( GLcontext *ctx ); -extern void intelSpanRenderFinish( GLcontext *ctx ); -extern void intelSpanRenderStart( GLcontext *ctx ); +void +i915_dump_batchbuffer( unsigned *start, + unsigned *end ); -extern void -intelSetSpanFunctions(driRenderbuffer *rb, const GLvisual *vis); #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..1876218ca0 100644 --- a/src/mesa/drivers/dri/i915/i915_fragprog.c +++ b/src/mesa/drivers/dri/i915/i915_fragprog.c @@ -29,40 +29,56 @@ #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" - - +static const GLfloat sin_quad_constants[2][4] = { + { + 2.0, + -1.0, + .5, + .75 + }, + { + 4.0, + -4.0, + 1.0 / (2.0 * M_PI), + .2225 + } +}; -/* 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 +86,152 @@ 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, ZERO, ZERO, ONE); + 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; } } @@ -211,7 +243,7 @@ do { \ GLuint coord = src_vector( p, &inst->SrcReg[0], program); \ /* Texel lookup */ \ \ - i915_emit_texld( p, \ + i915_emit_texld( p, get_live_regs(p, inst), \ get_result_vector( p, inst ), \ get_result_flags( inst ), \ sampler, \ @@ -234,6 +266,43 @@ do { \ #define EMIT_2ARG_ARITH( OP ) EMIT_ARITH( OP, 2 ) #define EMIT_3ARG_ARITH( OP ) EMIT_ARITH( OP, 3 ) +/* + * TODO: consider moving this into core + */ +static void calc_live_regs( struct i915_fragment_program *p ) +{ + const struct gl_fragment_program *program = p->ctx->FragmentProgram._Current; + GLuint regsUsed = 0xffff0000; + GLint i; + + for (i = program->Base.NumInstructions - 1; i >= 0; i--) { + struct prog_instruction *inst = &program->Base.Instructions[i]; + int opArgs = _mesa_num_inst_src_regs(inst->Opcode); + int a; + + /* Register is written to: unmark as live for this and preceeding ops */ + if (inst->DstReg.File == PROGRAM_TEMPORARY) + regsUsed &= ~(1 << inst->DstReg.Index); + + for (a = 0; a < opArgs; a++) { + /* Register is read from: mark as live for this and preceeding ops */ + if (inst->SrcReg[a].File == PROGRAM_TEMPORARY) + regsUsed |= 1 << inst->SrcReg[a].Index; + } + + p->usedRegs[i] = regsUsed; + } +} + +static GLuint get_live_regs( struct i915_fragment_program *p, + const struct prog_instruction *inst ) +{ + const struct gl_fragment_program *program = p->ctx->FragmentProgram._Current; + GLuint nr = inst - program->Base.Instructions; + + return p->usedRegs[nr]; +} + /* Possible concerns: * @@ -246,9 +315,11 @@ 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); */ @@ -258,510 +329,551 @@ 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; } + if (program->Base.NumInstructions > I915_MAX_INSN) { + i915_program_error( p, "Exceeded max instructions" ); + return; + } + + /* Not always needed: + */ + calc_live_regs(p); + while (1) { GLuint src0, src1, src2, flags; - GLuint tmp = 0; + GLuint tmp = 0, consts0 = 0, consts1 = 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_COS: - src0 = src_vector( p, &inst->SrcReg[0], program); - tmp = i915_get_utemp( p ); + case OPCODE_ADD: + EMIT_2ARG_ARITH(A0_ADD); + break; - i915_emit_arith( p, - A0_MUL, - tmp, A0_DEST_CHANNEL_X, 0, - src0, - i915_emit_const1f(p, 1.0/(M_PI * 2)), - 0); + 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; - i915_emit_arith( p, - A0_MOD, + case OPCODE_COS: + src0 = src_vector(p, &inst->SrcReg[0], program); + tmp = i915_get_utemp(p); + consts0 = i915_emit_const4fv(p, sin_quad_constants[0]); + consts1 = i915_emit_const4fv(p, sin_quad_constants[1]); + + /* Reduce range from repeating about [-pi,pi] to [-1,1] */ + i915_emit_arith(p, + A0_MAD, + tmp, A0_DEST_CHANNEL_X, 0, + src0, + swizzle(consts1, Z, ZERO, ZERO, ZERO), /* 1/(2pi) */ + swizzle(consts0, W, ZERO, ZERO, ZERO)); /* .75 */ + + i915_emit_arith(p, A0_FRC, tmp, A0_DEST_CHANNEL_X, 0, tmp, 0, 0); + + i915_emit_arith(p, + A0_MAD, tmp, A0_DEST_CHANNEL_X, 0, - tmp, - 0, 0 ); + tmp, + swizzle(consts0, X, ZERO, ZERO, ZERO), /* 2 */ + swizzle(consts0, Y, ZERO, ZERO, ZERO)); /* -1 */ - /* By choosing different taylor constants, could get rid of this mul: + /* Compute COS with the same calculation used for SIN, but a + * different source range has been mapped to [-1,1] this time. */ - i915_emit_arith( p, - A0_MUL, - tmp, A0_DEST_CHANNEL_X, 0, - tmp, - i915_emit_const1f(p, (M_PI * 2)), + + /* tmp.y = abs(tmp.x); {x, abs(x), 0, 0} */ + i915_emit_arith(p, + A0_MAX, + tmp, A0_DEST_CHANNEL_Y, 0, + swizzle(tmp, ZERO, X, ZERO, ZERO), + negate(swizzle(tmp, ZERO, X, ZERO, ZERO), 0, 1, 0, 0), 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, + /* tmp.y = tmp.y * tmp.x; {x, x * abs(x), 0, 0} */ + 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); + tmp, A0_DEST_CHANNEL_Y, 0, + swizzle(tmp, ZERO, X, ZERO, ZERO), + tmp, + 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); + /* tmp.x = tmp.xy DP sin_quad_constants[2].xy */ + i915_emit_arith(p, + A0_DP3, + tmp, A0_DEST_CHANNEL_X, 0, + tmp, + swizzle(consts1, X, Y, ZERO, ZERO), + 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]; + /* tmp.x now contains a first approximation (y). Now, weight it + * against tmp.y**2 to get closer. */ - 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 ), + i915_emit_arith(p, + A0_MAX, + tmp, A0_DEST_CHANNEL_Y, 0, + swizzle(tmp, ZERO, X, ZERO, ZERO), + negate(swizzle(tmp, ZERO, X, ZERO, ZERO), 0, 1, 0, 0), 0); - break; - case OPCODE_EX2: - src0 = src_vector( p, &inst->SrcReg[0], program); + /* tmp.y = tmp.x * tmp.y - tmp.x; {y, y * abs(y) - y, 0, 0} */ + i915_emit_arith(p, + A0_MAD, + tmp, A0_DEST_CHANNEL_Y, 0, + swizzle(tmp, ZERO, X, ZERO, ZERO), + swizzle(tmp, ZERO, Y, ZERO, ZERO), + negate(swizzle(tmp, ZERO, X, ZERO, ZERO), 0, 1, 0, 0)); + + /* result = .2225 * tmp.y + tmp.x =.2225(y * abs(y) - y) + y= */ + i915_emit_arith(p, + A0_MAD, + get_result_vector(p, inst), + get_result_flags(inst), 0, + swizzle(consts1, W, W, W, W), + swizzle(tmp, Y, Y, Y, Y), + swizzle(tmp, X, X, X, X)); + 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_EXP, - get_result_vector( p, inst ), - get_result_flags( inst ), 0, - swizzle(src0,X,X,X,X), 0, 0); - break; + 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_FLR: - EMIT_1ARG_ARITH( A0_FLR ); - break; + case OPCODE_EX2: + src0 = src_vector(p, &inst->SrcReg[0], program); - case OPCODE_FRC: - EMIT_1ARG_ARITH( A0_FRC ); - break; + 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 ); + src0 = src_vector(p, &inst->SrcReg[0], program); + tmp = i915_get_utemp(p); + + i915_emit_texld(p, get_live_regs(p, inst), + 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)); - i915_emit_arith( p, A0_MAD, - get_result_vector( p, inst ), - flags, 0, - negate(src2, 1,1,1,1), src0, tmp ); - break; + 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); + EMIT_2ARG_ARITH(A0_MAX); + break; - i915_emit_arith( p, - A0_MUL, - tmp, A0_DEST_CHANNEL_X, 0, - tmp, src1, 0); + 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; - i915_emit_arith( p, - A0_EXP, - get_result_vector( p, inst ), - flags, 0, - swizzle(tmp,X,X,X,X), 0, 0); + case OPCODE_MUL: + EMIT_2ARG_ARITH(A0_MUL); + break; - 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); - case OPCODE_RCP: - src0 = src_vector( p, &inst->SrcReg[0], program); + /* 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_RCP, - get_result_vector( p, inst ), - get_result_flags( inst ), 0, - swizzle(src0,X,X,X,X), 0, 0); - break; + i915_emit_arith(p, A0_MUL, tmp, A0_DEST_CHANNEL_X, 0, tmp, src1, 0); - case OPCODE_RSQ: - src0 = src_vector( p, &inst->SrcReg[0], program); + 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; - 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); + 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; - 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_SGE: + EMIT_2ARG_ARITH(A0_SGE); + break; case OPCODE_SIN: - src0 = src_vector( p, &inst->SrcReg[0], program); - tmp = i915_get_utemp( p ); + src0 = src_vector(p, &inst->SrcReg[0], program); + tmp = i915_get_utemp(p); + consts0 = i915_emit_const4fv(p, sin_quad_constants[0]); + consts1 = i915_emit_const4fv(p, sin_quad_constants[1]); + + /* Reduce range from repeating about [-pi,pi] to [-1,1] */ + i915_emit_arith(p, + A0_MAD, + tmp, A0_DEST_CHANNEL_X, 0, + src0, + swizzle(consts1, Z, ZERO, ZERO, ZERO), /* 1/(2pi) */ + swizzle(consts0, Z, ZERO, ZERO, ZERO)); /* .5 */ + + i915_emit_arith(p, A0_FRC, tmp, A0_DEST_CHANNEL_X, 0, tmp, 0, 0); + + i915_emit_arith(p, + A0_MAD, + tmp, A0_DEST_CHANNEL_X, 0, + tmp, + swizzle(consts0, X, ZERO, ZERO, ZERO), /* 2 */ + swizzle(consts0, Y, ZERO, ZERO, ZERO)); /* -1 */ - i915_emit_arith( p, + /* Compute sin using a quadratic and quartic. It gives continuity + * that repeating the Taylor series lacks every 2*pi, and has + * reduced error. + * + * The idea was described at: + * http://www.devmaster.net/forums/showthread.php?t=5784 + */ + + /* tmp.y = abs(tmp.x); {x, abs(x), 0, 0} */ + i915_emit_arith(p, + A0_MAX, + tmp, A0_DEST_CHANNEL_Y, 0, + swizzle(tmp, ZERO, X, ZERO, ZERO), + negate(swizzle(tmp, ZERO, X, ZERO, ZERO), 0, 1, 0, 0), + 0); + + /* tmp.y = tmp.y * tmp.x; {x, x * abs(x), 0, 0} */ + i915_emit_arith(p, A0_MUL, - tmp, A0_DEST_CHANNEL_X, 0, - src0, - i915_emit_const1f(p, 1.0/(M_PI * 2)), + tmp, A0_DEST_CHANNEL_Y, 0, + swizzle(tmp, ZERO, X, ZERO, ZERO), + tmp, 0); - i915_emit_arith( p, - A0_MOD, - tmp, A0_DEST_CHANNEL_X, 0, - tmp, - 0, 0 ); + /* tmp.x = tmp.xy DP sin_quad_constants[2].xy */ + i915_emit_arith(p, + A0_DP3, + tmp, A0_DEST_CHANNEL_X, 0, + tmp, + swizzle(consts1, X, Y, ZERO, ZERO), + 0); - /* By choosing different taylor constants, could get rid of this mul: + /* tmp.x now contains a first approximation (y). Now, weight it + * against tmp.y**2 to get closer. */ - i915_emit_arith( p, - A0_MUL, - tmp, A0_DEST_CHANNEL_X, 0, - tmp, - i915_emit_const1f(p, (M_PI * 2)), + i915_emit_arith(p, + A0_MAX, + tmp, A0_DEST_CHANNEL_Y, 0, + swizzle(tmp, ZERO, X, ZERO, ZERO), + negate(swizzle(tmp, ZERO, X, ZERO, ZERO), 0, 1, 0, 0), 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); + /* tmp.y = tmp.x * tmp.y - tmp.x; {y, y * abs(y) - y, 0, 0} */ + i915_emit_arith(p, + A0_MAD, + tmp, A0_DEST_CHANNEL_Y, 0, + swizzle(tmp, ZERO, X, ZERO, ZERO), + swizzle(tmp, ZERO, Y, ZERO, ZERO), + negate(swizzle(tmp, ZERO, X, ZERO, ZERO), 0, 1, 0, 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); + /* result = .2225 * tmp.y + tmp.x =.2225(y * abs(y) - y) + y= */ + i915_emit_arith(p, + A0_MAD, + get_result_vector(p, inst), + get_result_flags(inst), 0, + swizzle(consts1, W, W, W, W), + swizzle(tmp, Y, Y, Y, Y), + swizzle(tmp, X, X, X, X)); - 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; + 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 +881,22 @@ 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 +904,12 @@ 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 +918,54 @@ 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 +974,70 @@ 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 +1054,28 @@ 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 +1084,8 @@ 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 +1094,31 @@ 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 +1126,66 @@ 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; - EMIT_ATTR( _TNL_ATTRIB_TEX0+i, EMIT_SZ(sz), 0, sz * 4 ); + 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); } 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); + /* 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 ); - } + 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..73aa63451e 100644 --- a/src/mesa/drivers/dri/i915/i915_metaops.c +++ b/src/mesa/drivers/dri/i915/i915_metaops.c @@ -34,128 +34,169 @@ #include "intel_screen.h" #include "intel_batchbuffer.h" #include "intel_ioctl.h" -#include "intel_rotate.h" +#include "intel_regions.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 +253,64 @@ 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 +323,186 @@ 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..f79d00ddcf 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 - @@ -164,57 +170,67 @@ 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; } +static GLuint get_free_rreg (struct i915_fragment_program *p, + GLuint live_regs) +{ + int bit = ffs(~live_regs); + if (!bit) { + i915_program_error(p, "Can't find free R reg"); + return UREG_BAD; + } + return UREG(REG_TYPE_R, bit - 1); +} + GLuint i915_emit_texld( struct i915_fragment_program *p, + GLuint live_regs, GLuint dest, GLuint destmask, GLuint sampler, GLuint coord, GLuint op ) { - if (coord != UREG(GET_UREG_TYPE(coord), GET_UREG_NR(coord))) { - /* No real way to work around this in the general case - need to - * allocate and declare a new temporary register (a utemp won't - * do). Will fallback for now. - */ - i915_program_error(p, "Can't (yet) swizzle TEX arguments"); - return 0; - } - - /* Don't worry about saturate as we only support + if (coord != UREG(GET_UREG_TYPE(coord), GET_UREG_NR(coord))) { + /* With the help of the "needed registers" table created earlier, pick + * a register we can MOV the swizzled TC to (since TEX doesn't support + * swizzled sources) */ + GLuint swizCoord = get_free_rreg(p, live_regs); + if (swizCoord == UREG_BAD) + return 0; + + i915_emit_arith( p, A0_MOV, swizCoord, A0_DEST_CHANNEL_ALL, 0, coord, 0, 0 ); + coord = swizCoord; + } + + /* Don't worry about saturate as we only support texture formats + * that are always in the 0..1 range. */ if (destmask != A0_DEST_CHANNEL_ALL) { GLuint tmp = i915_get_utemp(p); - i915_emit_texld( p, tmp, A0_DEST_CHANNEL_ALL, sampler, coord, op ); + i915_emit_texld( p, 0, tmp, A0_DEST_CHANNEL_ALL, sampler, coord, op ); i915_emit_arith( p, A0_MOV, dest, destmask, 0, tmp, 0, 0 ); return dest; } @@ -239,24 +255,28 @@ 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 +285,35 @@ 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 +324,28 @@ 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 +355,36 @@ 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 +395,31 @@ 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 +432,17 @@ 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 +472,24 @@ 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 +498,28 @@ 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..14a3f08801 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 +84,76 @@ /* 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 live_regs, + 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..8891e11c6f 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)) @@ -313,120 +325,21 @@ #define SCISSOR_RECT_0_YMAX(x) ((x)<<16) #define SCISSOR_RECT_0_XMAX(x) (x) -/* p189 */ -#define _3DSTATE_LOAD_STATE_IMMEDIATE_1 ((0x3<<29)|(0x1d<<24)|(0x04<<16)) -#define I1_LOAD_S(n) (1<<(4+n)) - -#define S0_VB_OFFSET_MASK 0xffffffc -#define S0_AUTO_CACHE_INV_DISABLE (1<<0) - -#define S1_VERTEX_WIDTH_SHIFT 24 -#define S1_VERTEX_WIDTH_MASK (0x3f<<24) -#define S1_VERTEX_PITCH_SHIFT 16 -#define S1_VERTEX_PITCH_MASK (0x3f<<16) - -#define TEXCOORDFMT_2D 0x0 -#define TEXCOORDFMT_3D 0x1 -#define TEXCOORDFMT_4D 0x2 -#define TEXCOORDFMT_1D 0x3 -#define TEXCOORDFMT_2D_16 0x4 -#define TEXCOORDFMT_4D_16 0x5 -#define TEXCOORDFMT_NOT_PRESENT 0xf -#define S2_TEXCOORD_FMT0_MASK 0xf -#define S2_TEXCOORD_FMT1_SHIFT 4 -#define S2_TEXCOORD_FMT(unit, type) ((type)<<(unit*4)) -#define S2_TEXCOORD_NONE (~0) - -/* S3 not interesting */ - -#define S4_POINT_WIDTH_SHIFT 23 -#define S4_POINT_WIDTH_MASK (0x1ff<<23) -#define S4_LINE_WIDTH_SHIFT 19 -#define S4_LINE_WIDTH_ONE (0x2<<19) -#define S4_LINE_WIDTH_MASK (0xf<<19) -#define S4_FLATSHADE_ALPHA (1<<18) -#define S4_FLATSHADE_FOG (1<<17) -#define S4_FLATSHADE_SPECULAR (1<<16) -#define S4_FLATSHADE_COLOR (1<<15) -#define S4_CULLMODE_BOTH (0<<13) -#define S4_CULLMODE_NONE (1<<13) -#define S4_CULLMODE_CW (2<<13) -#define S4_CULLMODE_CCW (3<<13) -#define S4_CULLMODE_MASK (3<<13) -#define S4_VFMT_POINT_WIDTH (1<<12) -#define S4_VFMT_SPEC_FOG (1<<11) -#define S4_VFMT_COLOR (1<<10) -#define S4_VFMT_DEPTH_OFFSET (1<<9) -#define S4_VFMT_XYZ (1<<6) -#define S4_VFMT_XYZW (2<<6) -#define S4_VFMT_XY (3<<6) -#define S4_VFMT_XYW (4<<6) -#define S4_VFMT_XYZW_MASK (7<<6) -#define S4_FORCE_DEFAULT_DIFFUSE (1<<5) -#define S4_FORCE_DEFAULT_SPECULAR (1<<4) -#define S4_LOCAL_DEPTH_OFFSET_ENABLE (1<<3) -#define S4_VFMT_FOG_PARAM (1<<2) -#define S4_SPRITE_POINT_ENABLE (1<<1) -#define S4_LINE_ANTIALIAS_ENABLE (1<<0) - -#define S4_VFMT_MASK (S4_VFMT_POINT_WIDTH | \ - S4_VFMT_SPEC_FOG | \ - S4_VFMT_COLOR | \ - S4_VFMT_DEPTH_OFFSET | \ - S4_VFMT_XYZW_MASK | \ - S4_VFMT_FOG_PARAM) - - -#define S5_WRITEDISABLE_ALPHA (1<<31) -#define S5_WRITEDISABLE_RED (1<<30) -#define S5_WRITEDISABLE_GREEN (1<<29) -#define S5_WRITEDISABLE_BLUE (1<<28) -#define S5_WRITEDISABLE_MASK (0xf<<28) -#define S5_FORCE_DEFAULT_POINT_SIZE (1<<27) -#define S5_LAST_PIXEL_ENABLE (1<<26) -#define S5_GLOBAL_DEPTH_OFFSET_ENABLE (1<<25) -#define S5_FOG_ENABLE (1<<24) -#define S5_STENCIL_REF_SHIFT 16 -#define S5_STENCIL_REF_MASK (0xff<<16) -#define S5_STENCIL_TEST_FUNC_SHIFT 13 -#define S5_STENCIL_TEST_FUNC_MASK (0x7<<13) -#define S5_STENCIL_FAIL_SHIFT 10 -#define S5_STENCIL_FAIL_MASK (0x7<<10) -#define S5_STENCIL_PASS_Z_FAIL_SHIFT 7 -#define S5_STENCIL_PASS_Z_FAIL_MASK (0x7<<7) -#define S5_STENCIL_PASS_Z_PASS_SHIFT 4 -#define S5_STENCIL_PASS_Z_PASS_MASK (0x7<<4) -#define S5_STENCIL_WRITE_ENABLE (1<<3) -#define S5_STENCIL_TEST_ENABLE (1<<2) -#define S5_COLOR_DITHER_ENABLE (1<<1) -#define S5_LOGICOP_ENABLE (1<<0) - - -#define S6_ALPHA_TEST_ENABLE (1<<31) -#define S6_ALPHA_TEST_FUNC_SHIFT 28 -#define S6_ALPHA_TEST_FUNC_MASK (0x7<<28) -#define S6_ALPHA_REF_SHIFT 20 -#define S6_ALPHA_REF_MASK (0xff<<20) -#define S6_DEPTH_TEST_ENABLE (1<<19) -#define S6_DEPTH_TEST_FUNC_SHIFT 16 -#define S6_DEPTH_TEST_FUNC_MASK (0x7<<16) -#define S6_CBUF_BLEND_ENABLE (1<<15) -#define S6_CBUF_BLEND_FUNC_SHIFT 12 -#define S6_CBUF_BLEND_FUNC_MASK (0x7<<12) -#define S6_CBUF_SRC_BLEND_FACT_SHIFT 8 -#define S6_CBUF_SRC_BLEND_FACT_MASK (0xf<<8) -#define S6_CBUF_DST_BLEND_FACT_SHIFT 4 -#define S6_CBUF_DST_BLEND_FACT_MASK (0xf<<4) -#define S6_DEPTH_WRITE_ENABLE (1<<3) -#define S6_COLOR_WRITE_ENABLE (1<<2) -#define S6_TRISTRIP_PV_SHIFT 0 -#define S6_TRISTRIP_PV_MASK (0x3<<0) - -#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 +348,7 @@ #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 +371,7 @@ #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 +383,33 @@ */ #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 +426,34 @@ #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 +512,23 @@ /* 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 +540,18 @@ */ #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 +608,12 @@ #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 +622,7 @@ #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 +638,11 @@ #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 +664,7 @@ #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 +692,7 @@ #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 +739,8 @@ #define ST1_ENABLE (1<<16) #define ST1_MASK (0xffff) - -#define MI_FLUSH ((0<<29)|(4<<23)) -#define FLUSH_MAP_CACHE (1<<0) -#define FLUSH_RENDER_CACHE (1<<1) - +#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)) #endif diff --git a/src/mesa/drivers/dri/i915/i915_state.c b/src/mesa/drivers/dri/i915/i915_state.c index 1c4ec74755..c814f8d532 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); - const GLubyte *m = mask; + struct i915_context *i915 = I915_CONTEXT(ctx); + const GLubyte *m; 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 +323,32 @@ 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; - } + /* Use the already unpacked stipple data from the context rather than the + * uninterpreted mask passed in. + */ + mask = (const GLubyte *)ctx->PolygonStipple; + m = mask; + + 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 +369,54 @@ 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 +424,14 @@ 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 +439,12 @@ 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 +455,16 @@ 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 +474,15 @@ 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 +497,24 @@ 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 +522,55 @@ 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 +581,7 @@ 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 +594,49 @@ 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 +658,38 @@ 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 +697,8 @@ 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 +711,26 @@ 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 +742,8 @@ 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 (ctx->Visual.rgbBits == 16) + FALLBACK(&i915->intel, I915_FALLBACK_LOGICOP, state); break; case GL_FRAGMENT_PROGRAM_ARB: @@ -714,37 +752,37 @@ 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 +793,25 @@ 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 +820,20 @@ 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: @@ -798,10 +842,9 @@ 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; - /* Zero all state */ memset(&i915->state, 0, sizeof(i915->state)); @@ -811,39 +854,35 @@ 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 (i915->intel.ctx.Visual.rgbBits == 16) + 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; } @@ -858,79 +897,50 @@ 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; } - { - 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 */ - - - /* 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; - - i915->state.Buffer[I915_DESTREG_DV0] = _3DSTATE_DST_BUF_VARS_CMD; - /* color/depth pixel format */ - 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; - 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; - } - /* 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 +967,15 @@ 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..386617a1ff 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 +71,10 @@ 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..b5085f49d4 --- /dev/null +++ b/src/mesa/drivers/dri/i915/i915_tex_layout.c @@ -0,0 +1,477 @@ +/************************************************************************** + * + * 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. + * + **************************************************************************/ + +/** @file i915_tex_layout.c + * Code to layout images in a mipmap tree for i830M-GM915 and G945 and beyond. + */ + +#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] = { + [FACE_POS_X] = {0, 0}, + [FACE_POS_Y] = {1, 0}, + [FACE_POS_Z] = {1, 1}, + [FACE_NEG_X] = {0, 2}, + [FACE_NEG_Y] = {1, 2}, + [FACE_NEG_Z] = {1, 3}, +}; + + +static GLint step_offsets[6][2] = { + [FACE_POS_X] = {0, 2}, + [FACE_POS_Y] = {-1, 2}, + [FACE_POS_Z] = {-1, 1}, + [FACE_NEG_X] = {0, 2}, + [FACE_NEG_Y] = {-1, 2}, + [FACE_NEG_Z] = {-1, 1}, +}; + +/** + * Cube texture map layout for i830M-GM915. + * + * Hardware layout looks like: + * + * +-------+-------+ + * | | | + * | | | + * | | | + * | +x | +y | + * | | | + * | | | + * | | | + * | | | + * +---+---+-------+ + * | | | | + * | +x| +y| | + * | | | | + * | | | | + * +-+-+---+ +z | + * | | | | | + * +-+-+ +z| | + * | | | | + * +-+-+---+-------+ + * | | | + * | | | + * | | | + * | -x | -y | + * | | | + * | | | + * | | | + * | | | + * +---+---+-------+ + * | | | | + * | -x| -y| | + * | | | | + * | | | | + * +-+-+---+ -z | + * | | | | | + * +-+-+ -z| | + * | | | | + * +-+---+-------+ + * + */ +static void +i915_miptree_layout_cube(struct intel_context *intel, + struct intel_mipmap_tree * mt) +{ + const GLuint dim = mt->width0; + GLuint face; + GLuint lvlWidth = mt->width0, lvlHeight = mt->height0; + GLint level; + + assert(lvlWidth == lvlHeight); /* cubemap images are square */ + + /* double pitch for cube layouts */ + mt->pitch = intel_miptree_pitch_align (intel, mt, dim * 2); + 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; + } + } +} + +static void +i915_miptree_layout_3d(struct intel_context *intel, + struct intel_mipmap_tree * mt) +{ + GLuint width = mt->width0; + GLuint height = mt->height0; + GLuint depth = mt->depth0; + GLuint stack_height = 0; + GLint level; + + /* Calculate the size of a single slice. */ + mt->pitch = intel_miptree_pitch_align (intel, mt, mt->width0); + + /* 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; +} + +static void +i915_miptree_layout_2d(struct intel_context *intel, + struct intel_mipmap_tree * mt) +{ + GLuint width = mt->width0; + GLuint height = mt->height0; + GLuint img_height; + GLint level; + + mt->pitch = intel_miptree_pitch_align (intel, mt, mt->width0); + 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); + } +} + +GLboolean +i915_miptree_layout(struct intel_context *intel, struct intel_mipmap_tree * mt) +{ + switch (mt->target) { + case GL_TEXTURE_CUBE_MAP: + i915_miptree_layout_cube(intel, mt); + break; + case GL_TEXTURE_3D: + i915_miptree_layout_3d(intel, mt); + break; + case GL_TEXTURE_1D: + case GL_TEXTURE_2D: + case GL_TEXTURE_RECTANGLE_ARB: + i915_miptree_layout_2d(intel, mt); + break; + default: + _mesa_problem(NULL, "Unexpected tex target in i915_miptree_layout()"); + 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; +} + + +/** + * Cube texture map layout for GM945 and later. + * + * The hardware layout looks like the 830-915 layout, except for the small + * sizes. A zoomed in view of the layout for 945 is: + * + * +-------+-------+ + * | 8x8 | 8x8 | + * | | | + * | | | + * | +x | +y | + * | | | + * | | | + * | | | + * | | | + * +---+---+-------+ + * |4x4| | 8x8 | + * | +x| | | + * | | | | + * | | | | + * +---+ | +z | + * |4x4| | | + * | +y| | | + * | | | | + * +---+ +-------+ + * + * ... + * + * +-------+-------+ + * | 8x8 | 8x8 | + * | | | + * | | | + * | -x | -y | + * | | | + * | | | + * | | | + * | | | + * +---+---+-------+ + * |4x4| | 8x8 | + * | -x| | | + * | | | | + * | | | | + * +---+ | -z | + * |4x4| | | + * | -y| | | + * | | | | + * +---+ +---+---+---+---+---+---+---+---+---+ + * |4x4| |4x4| |2x2| |2x2| |2x2| |2x2| + * | +z| | -z| | +x| | +y| | +z| | -x| ... + * | | | | | | | | | | | | + * +---+ +---+ +---+ +---+ +---+ +---+ + * + * The bottom row continues with the remaining 2x2 then the 1x1 mip contents + * in order, with each of them aligned to a 4x4 block boundary. Thus, for + * 32x32 cube maps and smaller, the bottom row layout is going to dictate the + * pitch of the tree. For a tree with 4x4 images, the pitch is at least + * 14 * 8 = 112 texels, for 2x2 it is at least 12 * 8 texels, and for 1x1 + * it is 6 * 8 texels. + */ + +static void +i945_miptree_layout_cube(struct intel_context *intel, + struct intel_mipmap_tree * mt) +{ + const GLuint dim = mt->width0; + GLuint face; + GLuint lvlWidth = mt->width0, lvlHeight = mt->height0; + GLint level; + + 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 = intel_miptree_pitch_align (intel, mt, dim * 2); + else + mt->pitch = intel_miptree_pitch_align (intel, mt, 14 * 8); + + if (dim >= 4) + mt->total_height = dim * 4 + 4; + else + mt->total_height = 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; + } + } + } +} + +static void +i945_miptree_layout_3d(struct intel_context *intel, + struct intel_mipmap_tree * mt) +{ + 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 = intel_miptree_pitch_align (intel, mt, mt->width0); + 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++) { + GLint x = 0; + GLint y = 0; + GLint q, j; + + intel_miptree_set_level_info(mt, level, depth, + 0, mt->total_height, + width, height, depth); + + for (q = 0; q < depth;) { + for (j = 0; j < pack_x_nr && q < depth; 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); + } +} + +GLboolean +i945_miptree_layout(struct intel_context *intel, struct intel_mipmap_tree * mt) +{ + switch (mt->target) { + case GL_TEXTURE_CUBE_MAP: + i945_miptree_layout_cube(intel, mt); + break; + case GL_TEXTURE_3D: + i945_miptree_layout_3d(intel, mt); + break; + case GL_TEXTURE_1D: + case GL_TEXTURE_2D: + case GL_TEXTURE_RECTANGLE_ARB: + i945_miptree_layout_2d(intel, mt); + break; + default: + _mesa_problem(NULL, "Unexpected tex target in i945_miptree_layout()"); + 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; +} 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..69c1e0755e 100644 --- a/src/mesa/drivers/dri/i915/i915_texstate.c +++ b/src/mesa/drivers/dri/i915/i915_texstate.c @@ -25,902 +25,351 @@ * **************************************************************************/ -#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 ) +static GLuint +translate_texture_format(GLuint mesa_format, GLenum DepthMode) { - 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 ) -{ - 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; - + if (DepthMode == GL_ALPHA) + return (MAPSURF_16BIT | MT_16BIT_A16); + else if (DepthMode == GL_INTENSITY) + return (MAPSURF_16BIT | MT_16BIT_I16); + else + 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_xI824); 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_unit *tUnit = &ctx->Texture.Unit[unit]; + struct gl_texture_object *tObj = tUnit->_Current; + struct intel_texture_object *intelObj = intel_texture_object(tObj); + struct gl_texture_image *firstImage; + GLuint *state = i915->state.Tex[unit], format, pitch; + GLint lodbias; - 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); - - 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)); + if (!intelObj->imageOverride && !intel_finalize_mipmap_tree(intel, unit)) + return GL_FALSE; - state[I915_TEXREG_SS3] |= (unit<<SS3_TEXTUREMAP_INDEX_SHIFT); + /* 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_SS4] = t->Setup[I915_TEXREG_SS4]; + if (intelObj->imageOverride) { + i915->state.tex_buffer[unit] = NULL; + i915->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; + } - 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)); + 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); + + format = translate_texture_format(firstImage->TexFormat->MesaFormat, + tObj->DepthMode); + pitch = intelObj->mt->pitch * intelObj->mt->cpp; } -} - - -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; + state[I915_TEXREG_MS3] = + (((firstImage->Height - 1) << MS3_HEIGHT_SHIFT) | + ((firstImage->Width - 1) << MS3_WIDTH_SHIFT) | format | + MS3_USE_FENCE_REGS); - 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; -} + lodbias = (int) ((tUnit->LodBias + tObj->LodBias) * 16.0); + if (lodbias < -256) + lodbias = -256; + if (lodbias > 255) + lodbias = 255; + state[I915_TEXREG_SS2] = ((lodbias << SS2_LOD_BIAS_SHIFT) & + SS2_LOD_BIAS_MASK); -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 (tObj->Target == GL_TEXTURE_1D) + return GL_FALSE; - 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_shadow_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]; - - ss3 |= SS3_NORMALIZED_COORDS; + { + GLenum ws = tObj->WrapS; + GLenum wt = tObj->WrapT; + GLenum wr = tObj->WrapR; - 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..19f2206285 100644 --- a/src/mesa/drivers/dri/i915/i915_vtbl.c +++ b/src/mesa/drivers/dri/i915/i915_vtbl.c @@ -37,115 +37,148 @@ #include "tnl/t_vertex.h" #include "intel_batchbuffer.h" +#include "intel_tex.h" +#include "intel_regions.h" +#include "intel_tris.h" #include "i915_reg.h" #include "i915_context.h" -static void i915_render_start( intelContextPtr intel ) +#include "glapi.h" + +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 ); - } + if (!intel->Fallback) + 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, IGNORE_CLIPRECTS); 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 +191,27 @@ 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 +219,23 @@ 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, IGNORE_CLIPRECTS ) -static GLuint get_dirty( struct i915_hw_state *state ) +static GLuint +get_dirty(struct i915_hw_state *state) { GLuint dirty; @@ -227,94 +246,172 @@ 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; } - /* 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; + int i, count, aper_count; + GLuint dirty; + dri_bo *aper_array[3 + I915_TEX_UNITS]; + GET_CURRENT_CONTEXT(ctx); 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 between the state emits and the primitive + * emit header. + * + * It might be better to talk about explicit places where + * scheduling is allowed, rather than assume that it is whenever a + * batchbuffer fills up. + * + * Set the space as LOOP_CLIPRECTS now, since that's what our primitives + * will be emitted under. + */ + intel_batchbuffer_require_space(intel->batch, + get_state_size(state) + INTEL_PRIM_EMIT_SIZE, + LOOP_CLIPRECTS); + count = 0; + again: + aper_count = 0; + dirty = get_dirty(state); + + aper_array[aper_count++] = intel->batch->buf; + if (dirty & I915_UPLOAD_BUFFERS) { + aper_array[aper_count++] = state->draw_region->buffer; + if (state->depth_region) + aper_array[aper_count++] = state->depth_region->buffer; + } + + if (dirty & I915_UPLOAD_TEX_ALL) { + for (i = 0; i < I915_TEX_UNITS; i++) { + if (dirty & I915_UPLOAD_TEX(i)) { + if (state->tex_buffer[i]) { + aper_array[aper_count++] = state->tex_buffer[i]; + } + } + } } - if (VERBOSE) + if (dri_bufmgr_check_aperture_space(aper_array, aper_count)) { + if (count == 0) { + count++; + intel_batchbuffer_flush(intel->batch); + goto again; + } else { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "i915 emit state"); + assert(0); + } + } + + /* work out list of buffers to emit */ + + /* Do this here as we may have flushed the batchbuffer above, + * causing more state to be dirty! + */ + dirty = get_dirty(state); + state->emitted |= dirty; + assert(get_dirty(state) == 0); + + 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, IGNORE_CLIPRECTS); + OUT_BATCH(state->Buffer[I915_DESTREG_CBUFADDR0]); + OUT_BATCH(state->Buffer[I915_DESTREG_CBUFADDR1]); + OUT_RELOC(state->draw_region->buffer, + I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER, + 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, + I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER, + 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 +420,219 @@ 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, IGNORE_CLIPRECTS); + 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], + I915_GEM_DOMAIN_SAMPLER, 0, + 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, IGNORE_CLIPRECTS); + 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 (state->ProgramSize) { + if (INTEL_DEBUG & DEBUG_STATE) + fprintf(stderr, "I915_UPLOAD_PROGRAM:\n"); - 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 ); + assert((state->Program[0] & 0x1ff) + 2 == 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); + intel->batch->dirty_state &= ~dirty; + assert(get_dirty(state) == 0); + assert((intel->batch->dirty_state & (1<<1)) == 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_regions[], + struct intel_region *depth_region, + GLuint num_regions) { - 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_regions[0], 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_new_batch(struct intel_context *intel) { - i915ContextPtr i915 = I915_CONTEXT(intel); + struct i915_context *i915 = i915_context(&intel->ctx); - 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; + /* Mark all state as needing to be emitted when starting a new batchbuffer. + * Using hardware contexts would be an alternative, but they have some + * difficulties associated with them (physical address requirements). + */ + i915->state.emitted = 0; - 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; + /* Check that we didn't just wrap our batchbuffer at a bad time. */ + assert(!intel->no_batch_wrap); } - -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; + struct i915_context *i915 = i915_context(&intel->ctx); + struct i915_hw_state *state = i915->current; + GLuint dirty = get_dirty(state); + assert(!dirty); +} - BEGIN_BATCH(2); - OUT_BATCH( MI_FLUSH | FLUSH_MAP_CACHE | FLUSH_RENDER_CACHE ); - OUT_BATCH( 0 ); - ADVANCE_BATCH(); +static void +i915_note_unlock( struct intel_context *intel ) +{ + /* nothing */ } -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.new_batch = i915_new_batch; 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; + i915->intel.vtbl.note_unlock = i915_note_unlock; + i915->intel.vtbl.finish_batch = intel_finish_vb; } - diff --git a/src/mesa/drivers/dri/i915/intel_batchbuffer.c b/src/mesa/drivers/dri/i915/intel_batchbuffer.c index 803b41b256..d38cdf31cc 100644..120000 --- a/src/mesa/drivers/dri/i915/intel_batchbuffer.c +++ b/src/mesa/drivers/dri/i915/intel_batchbuffer.c @@ -1,829 +1 @@ -/************************************************************************** - * - * 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 "vblank.h" - -#include "intel_reg.h" -#include "intel_batchbuffer.h" -#include "intel_context.h" - - - - -/* ================================================================ - * Performance monitoring functions - */ - -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)); -} - -static void intel_draw_performance_boxes( intelContextPtr intel ) -{ - /* 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 ); - - - /* 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; - } -#endif - - intel->perf_boxes = 0; -} - - - - - - -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 ) -{ - 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; - - 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; - } - - 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; -} - - -/* Emit a primitive referencing vertices in a vertex buffer. - */ -void intelStartInlinePrimitive( intelContextPtr intel, GLuint prim ) -{ - 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(); -} - - -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. - */ -GLuint *intelEmitInlinePrimitiveLocked(intelContextPtr intel, - int primitive, - int dwords, - int vertex_size ) -{ - 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 ); - - 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; - } - - if (used < 8) - goto do_discard; - } - - /* Emit 3D_PRIMITIVE commands: - */ - 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->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 ); - - ADVANCE_BATCH(); - } - } - intelFlushBatchLocked( intel, GL_TRUE, GL_TRUE, GL_TRUE ); - UNLOCK_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; - } -} - - - - -void intelEmitFillBlitLocked( intelContextPtr intel, - GLuint cpp, - GLshort dst_pitch, /* in bytes */ - GLuint dst_offset, - GLshort x, GLshort y, - GLshort w, GLshort h, - GLuint color ) -{ - GLuint BR13, CMD; - BATCH_LOCALS; - - 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: - 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; - } - - if (dst_y2 < dst_y || - dst_x2 < dst_x) { - return; - } - - 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(); -} - - - -void intelClearWithBlit(GLcontext *ctx, GLbitfield buffers, GLboolean allFoo, - GLint cx1Foo, GLint cy1Foo, GLint cwFoo, GLint chFoo) -{ - 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 ); -} - - - - -void intelDestroyBatchBuffer( GLcontext *ctx ) -{ - 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; - } - - memset(&intel->batch, 0, sizeof(intel->batch)); -} - - -void intelInitBatchBuffer( GLcontext *ctx ) -{ - 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); -} +../intel/intel_batchbuffer.c
\ No newline at end of file diff --git a/src/mesa/drivers/dri/i915/intel_batchbuffer.h b/src/mesa/drivers/dri/i915/intel_batchbuffer.h deleted file mode 100644 index 577d07137f..0000000000 --- a/src/mesa/drivers/dri/i915/intel_batchbuffer.h +++ /dev/null @@ -1,126 +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_BATCHBUFFER_H -#define INTEL_BATCHBUFFER_H - -#include "intel_context.h" -#include "intel_ioctl.h" - - -#define BATCH_LOCALS GLubyte *batch_ptr; - -/* #define VERBOSE 0 */ -#ifndef VERBOSE -extern int VERBOSE; -#endif - - -#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) - -#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 ) -{ - GLuint sz = dwords * sizeof(GLuint); - GLuint *ptr; - - 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; - - return ptr; -} - - - -#endif diff --git a/src/mesa/drivers/dri/i915/intel_blit.c b/src/mesa/drivers/dri/i915/intel_blit.c new file mode 120000 index 0000000000..dd6c8d17c2 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_blit.c @@ -0,0 +1 @@ +../intel/intel_blit.c
\ No newline at end of file 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 120000 index 0000000000..e06dd3c8d3 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_buffer_objects.c @@ -0,0 +1 @@ +../intel/intel_buffer_objects.c
\ No newline at end of file diff --git a/src/mesa/drivers/dri/i915/intel_buffers.c b/src/mesa/drivers/dri/i915/intel_buffers.c new file mode 120000 index 0000000000..c86daa49f4 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_buffers.c @@ -0,0 +1 @@ +../intel/intel_buffers.c
\ No newline at end of file diff --git a/src/mesa/drivers/dri/i915/intel_context.c b/src/mesa/drivers/dri/i915/intel_context.c index 11c23f24a1..27a1cbb255 100644..120000 --- a/src/mesa/drivers/dri/i915/intel_context.c +++ b/src/mesa/drivers/dri/i915/intel_context.c @@ -1,776 +1 @@ -/************************************************************************** - * - * 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 "context.h" -#include "matrix.h" -#include "simple_list.h" -#include "extensions.h" -#include "framebuffer.h" -#include "imports.h" -#include "points.h" - -#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" - -#include "drivers/common/driverfuncs.h" - -#include "intel_screen.h" - -#include "i830_dri.h" -#include "i830_common.h" - -#include "intel_tex.h" -#include "intel_span.h" -#include "intel_tris.h" -#include "intel_ioctl.h" -#include "intel_batchbuffer.h" - -#include "vblank.h" -#include "utils.h" -#include "xmlpool.h" /* for symbolic values of enum-type options */ -#ifndef INTEL_DEBUG -int INTEL_DEBUG = (0); -#endif - -#define need_GL_ARB_multisample -#define need_GL_ARB_point_parameters -#define need_GL_ARB_texture_compression -#define need_GL_ARB_vertex_buffer_object -#define need_GL_ARB_vertex_program -#define need_GL_ARB_window_pos -#define need_GL_EXT_blend_color -#define need_GL_EXT_blend_equation_separate -#define need_GL_EXT_blend_func_separate -#define need_GL_EXT_blend_minmax -#define need_GL_EXT_cull_vertex -#define need_GL_EXT_fog_coord -#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 - -/*************************************** - * Mesa's Driver Functions - ***************************************/ - -#define DRIVER_DATE "20061017" - -const GLubyte *intelGetString( GLcontext *ctx, GLenum name ) -{ - const char * chipset; - static char buffer[128]; - - switch (name) { - case GL_VENDOR: - return (GLubyte *)"Tungsten Graphics, Inc"; - break; - - case GL_RENDERER: - switch (INTEL_CONTEXT(ctx)->intelScreen->deviceID) { - case PCI_CHIP_845_G: - chipset = "Intel(R) 845G"; break; - case PCI_CHIP_I830_M: - chipset = "Intel(R) 830M"; break; - case PCI_CHIP_I855_GM: - chipset = "Intel(R) 852GM/855GM"; break; - case PCI_CHIP_I865_G: - chipset = "Intel(R) 865G"; break; - case PCI_CHIP_I915_G: - chipset = "Intel(R) 915G"; break; - case PCI_CHIP_I915_GM: - chipset = "Intel(R) 915GM"; break; - case PCI_CHIP_I945_G: - chipset = "Intel(R) 945G"; break; - case PCI_CHIP_I945_GM: - chipset = "Intel(R) 945GM"; break; - case PCI_CHIP_I945_GME: - chipset = "Intel(R) 945GME"; break; - case PCI_CHIP_G33_G: - chipset = "Intel(R) G33"; break; - case PCI_CHIP_Q35_G: - chipset = "Intel(R) Q35"; break; - case PCI_CHIP_Q33_G: - chipset = "Intel(R) Q33"; break; - default: - chipset = "Unknown Intel Chipset"; break; - } - - (void) driGetRendererString( buffer, chipset, DRIVER_DATE, 0 ); - return (GLubyte *) buffer; - - default: - return NULL; - } -} - - -/** - * 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 } -}; - -extern const struct tnl_pipeline_stage _intel_render_stage; - -static const struct tnl_pipeline_stage *intel_pipeline[] = { - &_tnl_vertex_transform_stage, - &_tnl_vertex_cull_stage, - &_tnl_normal_transform_stage, - &_tnl_lighting_stage, - &_tnl_fog_coordinate_stage, - &_tnl_texgen_stage, - &_tnl_texture_transform_stage, - &_tnl_point_attenuation_stage, - &_tnl_vertex_program_stage, -#if 1 - &_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 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; -} - - -void intelInitDriverFunctions( struct dd_function_table *functions ) -{ - _mesa_init_driver_functions( functions ); - - functions->Clear = intelClear; - functions->Flush = intelglFlush; - functions->Finish = intelFinish; - functions->GetString = intelGetString; - functions->UpdateState = intelInvalidateState; - - intelInitTextureFuncs( functions ); - intelInitPixelFuncs( functions ); - intelInitStateFuncs( functions ); -} - -static void intel_emit_invarient_state( GLcontext *ctx ) -{ -} - - - -GLboolean intelInitContext( intelContextPtr 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; - drmI830Sarea *saPriv = (drmI830Sarea *) - (((GLubyte *)sPriv->pSAREA)+intelScreen->sarea_priv_offset); - int fthrottle_mode; - - if (!_mesa_initialize_context(&intel->ctx, - mesaVis, shareCtx, - functions, - (void*) intel)) - return GL_FALSE; - - driContextPriv->driverPrivate = intel; - intel->intelScreen = intelScreen; - intel->driScreen = sPriv; - intel->sarea = saPriv; - - - (void) memset( intel->texture_heaps, 0, sizeof( intel->texture_heaps ) ); - make_empty_list( & intel->swapped ); - - driParseConfigFiles (&intel->optionCache, &intelScreen->optionCache, - intel->driScreen->myNum, "i915"); - - ctx->Const.MaxTextureMaxAnisotropy = 2.0; - - ctx->Const.MinLineWidth = 1.0; - ctx->Const.MinLineWidthAA = 1.0; - ctx->Const.MaxLineWidth = 3.0; - ctx->Const.MaxLineWidthAA = 3.0; - ctx->Const.LineWidthGranularity = 1.0; - - ctx->Const.MinPointSize = 1.0; - ctx->Const.MinPointSizeAA = 1.0; - ctx->Const.MaxPointSize = 255.0; - ctx->Const.MaxPointSizeAA = 3.0; - ctx->Const.PointSizeGranularity = 1.0; - - /* reinitialize the context point state. - * It depend on constants in __GLcontextRec::Const - */ - _mesa_init_point(ctx); - - /* Initialize the software rasterizer and helper modules. */ - _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 ); - - /* Configure swrast to match hardware characteristics: */ - _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->hw_stencil = mesaVis->stencilBits && mesaVis->depthBits == 24; - intel->hw_stipple = 1; - - 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; - 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; - break; - default: - assert(0); - break; - } - - /* Initialize swrast, tnl driver tables: */ - 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); - - intel->do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS); - - intel->vblank_flags = (intel->intelScreen->irq_active != 0) - ? driGetDefaultVBlankFlags(&intel->optionCache) : VBLANK_FLAG_NO_IRQ; - - (*dri_interface->getUST)(&intel->swap_ust); - _math_matrix_ctr (&intel->ViewportMatrix); - - driInitExtensions( ctx, card_extensions, GL_TRUE ); - - if (intel->ctx.Mesa_DXTn) { - _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" ); - } - -/* 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; -#endif - - if (getenv("INTEL_NO_RAST") || - getenv("INTEL_NO_RAST")) { - fprintf(stderr, "disabling 3D rasterization\n"); - FALLBACK(intel, INTEL_FALLBACK_USER, 1); - } - - return GL_TRUE; -} - -void intelDestroyContext(__DRIcontextPrivate *driContextPriv) -{ - intelContextPtr intel = (intelContextPtr) driContextPriv->driverPrivate; - - assert(intel); /* should never be null */ - if (intel) { - GLboolean release_texture_heaps; - - INTEL_FIREVERTICES( 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); - - _swrast_DestroyContext (&intel->ctx); - intel->Fallback = 0; /* don't call _swrast_Flush later */ - - intelDestroyBatchBuffer(&intel->ctx); - - - 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 ) ); - } - - /* free the Mesa context */ - _mesa_destroy_context(&intel->ctx); - } -} - -void intelSetFrontClipRects( intelContextPtr intel ) -{ - __DRIdrawablePrivate *dPriv = intel->driDrawable; - - if (!dPriv) return; - - intel->numClipRects = dPriv->numClipRects; - intel->pClipRects = dPriv->pClipRects; - intel->drawX = dPriv->x; - intel->drawY = dPriv->y; -} - - -void intelSetBackClipRects( intelContextPtr intel ) -{ - __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; - } - } - } -} - - -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 (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 */ - } - - /* 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; - } - - if (flags != intel->vblank_flags) { - intel->vblank_flags = flags; - driGetCurrentVBlank(dPriv, intel->vblank_flags, &intel->vbl_seq); - } - } else { - intel->vblank_flags &= ~VBLANK_FLAG_SECONDARY; - } - - 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 { - _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 ) -{ - __DRIdrawablePrivate *dPriv = intel->driDrawable; - __DRIscreenPrivate *sPriv = intel->driScreen; - intelScreenPrivate *intelScreen = (intelScreenPrivate *)sPriv->private; - drmI830Sarea * sarea = intel->sarea; - unsigned i; - - drmGetLock(intel->driFd, intel->hHWContext, flags); - - /* If the window moved, may need to set a new cliprect now. - * - * NOTE: This releases and regains the hw lock, so all state - * checking must be done *after* this call: - */ - 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 (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 ); - - intel->lastStamp = 0; /* force window update */ - - /* Release batch buffer - */ - intelDestroyBatchBuffer(&intel->ctx); - intelInitBatchBuffer(&intel->ctx); - intel->prim.flush = intel_emit_invarient_state; - - /* Still need to reset the global LRU? - */ - intel_driReinitTextureHeap( intel->texture_heaps[0], intel->intelScreen->tex.size ); - } - - /* 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. - */ - for ( i = 0 ; i < intel->nr_heaps ; i++ ) { - DRI_AGE_TEXTURES( intel->texture_heaps[ i ] ); - } -} - - -void intelSwapBuffers( __DRIdrawablePrivate *dPriv ) -{ - 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__); - } -} - -void intelCopySubBuffer( __DRIdrawablePrivate *dPriv, - int x, int y, int w, int h ) -{ - 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/intel_context.c
\ No newline at end of file diff --git a/src/mesa/drivers/dri/i915/intel_context.h b/src/mesa/drivers/dri/i915/intel_context.h deleted file mode 100644 index 3b50107d73..0000000000 --- a/src/mesa/drivers/dri/i915/intel_context.h +++ /dev/null @@ -1,563 +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 INTELCONTEXT_INC -#define INTELCONTEXT_INC - - - -#include "mtypes.h" -#include "drm.h" -#include "mm.h" -#include "texmem.h" -#include "vblank.h" - -#include "intel_screen.h" -#include "i915_drm.h" -#include "i830_common.h" -#include "tnl/t_vertex.h" - -#define TAG(x) intel##x -#include "tnl_dd/t_dd_vertex.h" -#undef TAG - -#define DV_PF_555 (1<<8) -#define DV_PF_565 (2<<8) -#define DV_PF_8888 (3<<8) - -#define INTEL_CONTEXT(ctx) ((intelContextPtr)(ctx)) - -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 *); - -#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_RENDERMODE 0x20 - -extern void intelFallback( intelContextPtr intel, GLuint bit, GLboolean mode ); -#define FALLBACK( intel, bit, mode ) intelFallback( intel, bit, mode ) - - -#define INTEL_TEX_MAXLEVELS 10 - - -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 intel_context -{ - GLcontext ctx; /* the parent class */ - - struct { - void (*destroy)( intelContextPtr intel ); - void (*emit_state)( intelContextPtr intel ); - void (*lost_hardware)( intelContextPtr intel ); - void (*update_texture_state)( intelContextPtr intel ); - - 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 ); - - GLboolean (*check_vertex_size)( intelContextPtr intel, GLuint expected ); - - void (*clear_with_tris)( intelContextPtr intel, GLbitfield mask, - GLboolean all, - GLint cx, GLint cy, GLint cw, GLint ch); - - void (*rotate_window)( intelContextPtr intel, - __DRIdrawablePrivate *dPriv, GLuint srcBuf); - - intelTextureObjectPtr (*alloc_tex_obj)( struct gl_texture_object *tObj ); - - } vtbl; - - 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 { - GLuint primitive; - GLubyte *start_ptr; - void (*flush)( GLcontext * ); - } prim; - - GLboolean locked; - - GLubyte clear_red; - GLubyte clear_green; - GLubyte clear_blue; - GLubyte clear_alpha; - GLuint ClearColor; - GLuint ClearDepth; - - 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; - - 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]; - - /* State for intelvb.c and inteltris.c. - */ - GLuint RenderIndex; - GLmatrix ViewportMatrix; - GLenum render_primitive; - GLenum reduced_primitive; - GLuint vertex_size; - unsigned char *verts; /* points to tnl->clipspace.vertex_buf */ - - - /* Fallback rasterization functions - */ - intel_point_func draw_point; - intel_line_func draw_line; - intel_tri_func draw_tri; - - /* Drawing buffer state - */ - 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 */ - drm_clip_rect_t *pClipRects; - - 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; - int driFd; - - __DRIdrawablePrivate *driDrawable; - __DRIscreenPrivate *driScreen; - intelScreenPrivate *intelScreen; - drmI830Sarea *sarea; - - /** - * 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; - - GLuint swap_scheduled; -}; - - -#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. - */ -#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 - - - - -/* 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) - - -#define SUBPIXEL_X 0.125 -#define SUBPIXEL_Y 0.125 - -#define INTEL_FIREVERTICES(intel) \ -do { \ - if ((intel)->prim.flush) \ - (intel)->prim.flush(&(intel)->ctx); \ -} while (0) - -/* ================================================================ - * Color packing: - */ - -#define INTEL_PACKCOLOR4444(r,g,b,a) \ - ((((a) & 0xf0) << 8) | (((r) & 0xf0) << 4) | ((g) & 0xf0) | ((b) >> 4)) - -#define INTEL_PACKCOLOR1555(r,g,b,a) \ - ((((r) & 0xf8) << 7) | (((g) & 0xf8) << 2) | (((b) & 0xf8) >> 3) | \ - ((a) ? 0x8000 : 0)) - -#define INTEL_PACKCOLOR565(r,g,b) \ - ((((r) & 0xf8) << 8) | (((g) & 0xfc) << 3) | (((b) & 0xf8) >> 3)) - -#define INTEL_PACKCOLOR8888(r,g,b,a) \ - ((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: - */ -#if defined(i386) || defined(__i386__) -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"); - return (to); -} -#else -#define __memcpy(a,b,c) memcpy(a,b,c) -#endif - - - -/* ================================================================ - * Debugging: - */ -#define DO_DEBUG 1 -#if DO_DEBUG -extern int INTEL_DEBUG; -#else -#define INTEL_DEBUG 0 -#endif - -#define DEBUG_TEXTURE 0x1 -#define DEBUG_STATE 0x2 -#define DEBUG_IOCTL 0x4 -#define DEBUG_PRIMS 0x8 -#define DEBUG_VERTS 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 PCI_CHIP_845_G 0x2562 -#define PCI_CHIP_I830_M 0x3577 -#define PCI_CHIP_I855_GM 0x3582 -#define PCI_CHIP_I865_G 0x2572 -#define PCI_CHIP_I915_G 0x2582 -#define PCI_CHIP_I915_GM 0x2592 -#define PCI_CHIP_I945_G 0x2772 -#define PCI_CHIP_I945_GM 0x27A2 -#define PCI_CHIP_I945_GME 0x27AE -#define PCI_CHIP_G33_G 0x29C2 -#define PCI_CHIP_Q35_G 0x29B2 -#define PCI_CHIP_Q33_G 0x29D2 - - -/* ================================================================ - * intel_context.c: - */ - -extern void intelInitDriverFunctions( 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(intelContextPtr intel, GLuint flags); -extern void intelSetBackClipRects(intelContextPtr intel); -extern void intelSetFrontClipRects(intelContextPtr intel); -extern void intelWindowMoved( intelContextPtr intel ); - -extern const GLubyte *intelGetString( GLcontext *ctx, GLenum name ); - - -/* ================================================================ - * intel_state.c: - */ -extern void intelInitStateFuncs( struct dd_function_table *functions ); - -#define COMPAREFUNC_ALWAYS 0 -#define COMPAREFUNC_NEVER 0x1 -#define COMPAREFUNC_LESS 0x2 -#define COMPAREFUNC_EQUAL 0x3 -#define COMPAREFUNC_LEQUAL 0x4 -#define COMPAREFUNC_GREATER 0x5 -#define COMPAREFUNC_NOTEQUAL 0x6 -#define COMPAREFUNC_GEQUAL 0x7 - -#define STENCILOP_KEEP 0 -#define STENCILOP_ZERO 0x1 -#define STENCILOP_REPLACE 0x2 -#define STENCILOP_INCRSAT 0x3 -#define STENCILOP_DECRSAT 0x4 -#define STENCILOP_INCR 0x5 -#define STENCILOP_DECR 0x6 -#define STENCILOP_INVERT 0x7 - -#define LOGICOP_CLEAR 0 -#define LOGICOP_NOR 0x1 -#define LOGICOP_AND_INV 0x2 -#define LOGICOP_COPY_INV 0x3 -#define LOGICOP_AND_RVRSE 0x4 -#define LOGICOP_INV 0x5 -#define LOGICOP_XOR 0x6 -#define LOGICOP_NAND 0x7 -#define LOGICOP_AND 0x8 -#define LOGICOP_EQUIV 0x9 -#define LOGICOP_NOOP 0xa -#define LOGICOP_OR_INV 0xb -#define LOGICOP_COPY 0xc -#define LOGICOP_OR_RVRSE 0xd -#define LOGICOP_OR 0xe -#define LOGICOP_SET 0xf - -#define BLENDFACT_ZERO 0x01 -#define BLENDFACT_ONE 0x02 -#define BLENDFACT_SRC_COLR 0x03 -#define BLENDFACT_INV_SRC_COLR 0x04 -#define BLENDFACT_SRC_ALPHA 0x05 -#define BLENDFACT_INV_SRC_ALPHA 0x06 -#define BLENDFACT_DST_ALPHA 0x07 -#define BLENDFACT_INV_DST_ALPHA 0x08 -#define BLENDFACT_DST_COLR 0x09 -#define BLENDFACT_INV_DST_COLR 0x0a -#define BLENDFACT_SRC_ALPHA_SATURATE 0x0b -#define BLENDFACT_CONST_COLOR 0x0c -#define BLENDFACT_INV_CONST_COLOR 0x0d -#define BLENDFACT_CONST_ALPHA 0x0e -#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 ); - - -/* ================================================================ - * intel_ioctl.c: - */ -extern void intel_dump_batchbuffer( long offset, - int *ptr, - int count ); - - -/* ================================================================ - * intel_pixel.c: - */ -extern void intelInitPixelFuncs( struct dd_function_table *functions ); - - - -#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 120000 index 0000000000..4ac4ae690a --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_depthstencil.c @@ -0,0 +1 @@ +../intel/intel_depthstencil.c
\ No newline at end of file diff --git a/src/mesa/drivers/dri/i915/intel_fbo.c b/src/mesa/drivers/dri/i915/intel_fbo.c new file mode 120000 index 0000000000..a19f86dcc5 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_fbo.c @@ -0,0 +1 @@ +../intel/intel_fbo.c
\ No newline at end of file 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 120000 index 0000000000..242fed0b6a --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_mipmap_tree.c @@ -0,0 +1 @@ +../intel/intel_mipmap_tree.c
\ No newline at end of file diff --git a/src/mesa/drivers/dri/i915/intel_pixel.c b/src/mesa/drivers/dri/i915/intel_pixel.c index c3030d42b0..d733c5e874 100644..120000 --- a/src/mesa/drivers/dri/i915/intel_pixel.c +++ b/src/mesa/drivers/dri/i915/intel_pixel.c @@ -1,524 +1 @@ -/************************************************************************** - * - * 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 "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; -} - - -/** - * 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. - */ -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 ) -{ - 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 (!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. - */ - 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); -} - - - - -static void do_draw_pix( GLcontext *ctx, - GLint x, GLint y, GLsizei width, GLsizei height, - GLint pitch, - const void *pixels, - GLuint dest ) -{ - 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 (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 ); -} - - - -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 ) -{ - 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 ( intelIsAgpMemory(intel, pixels, size) ) - { - do_draw_pix( ctx, x, y, width, height, pitch, pixels, dest ); - 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); - - 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 -} - - - - -void intelInitPixelFuncs( struct dd_function_table *functions ) -{ - functions->CopyPixels = intelCopyPixels; - if (!getenv("INTEL_NO_BLITS")) { - functions->ReadPixels = intelReadPixels; - functions->DrawPixels = intelDrawPixels; - } -} +../intel/intel_pixel.c
\ No newline at end of file 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 120000 index 0000000000..9085c7b039 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_pixel_bitmap.c @@ -0,0 +1 @@ +../intel/intel_pixel_bitmap.c
\ No newline at end of file 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 120000 index 0000000000..ee43360590 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_pixel_copy.c @@ -0,0 +1 @@ +../intel/intel_pixel_copy.c
\ No newline at end of file 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 120000 index 0000000000..8431a24edf --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_pixel_draw.c @@ -0,0 +1 @@ +../intel/intel_pixel_draw.c
\ No newline at end of file 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..d009590a4b --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_pixel_read.c @@ -0,0 +1,307 @@ +/************************************************************************** + * + * 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; + + 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, src->tiling, + rowLength, dst_buffer, dst_offset, GL_FALSE, + rect.x1, + rect.y1, + rect.x1 - src_rect.x1, + rect.y2 - src_rect.y2, + rect.x2 - rect.x1, rect.y2 - rect.y1, + GL_COPY); + } + } + UNLOCK_HARDWARE(intel); + + 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 120000 index 0000000000..89b2f15c10 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_regions.c @@ -0,0 +1 @@ +../intel/intel_regions.c
\ No newline at end of file diff --git a/src/mesa/drivers/dri/i915/intel_render.c b/src/mesa/drivers/dri/i915/intel_render.c index d9438ba0fd..838d450378 100644 --- a/src/mesa/drivers/dri/i915/intel_render.c +++ b/src/mesa/drivers/dri/i915/intel_render.c @@ -39,6 +39,7 @@ #include "tnl/t_context.h" #include "tnl/t_vertex.h" +#include "tnl/t_pipeline.h" #include "intel_screen.h" #include "intel_context.h" @@ -51,14 +52,14 @@ * 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 +67,7 @@ #define HAVE_ELTS 0 -static GLuint hw_prim[GL_POLYGON+1] = { +static uint32_t hw_prim[GL_POLYGON + 1] = { 0, PRIM3D_LINELIST, PRIM3D_LINESTRIP, @@ -79,7 +80,7 @@ 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 +93,60 @@ 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]); + intel_set_prim(intel, hw_prim[prim]); } -#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 GET_SUBSEQUENT_VB_MAX_VERTS() \ - (((intel->alloc.size / 2) - 1500) / (intel->vertex_size*4)) -#define GET_CURRENT_VB_MAX_VERTS() GET_SUBSEQUENT_VB_MAX_VERTS() +#define FLUSH() INTEL_FIREVERTICES(intel) + +#define GET_SUBSEQUENT_VB_MAX_VERTS() (INTEL_VB_SIZE / (intel->vertex_size * 4)) +#define GET_CURRENT_VB_MAX_VERTS() \ + ((INTEL_VB_SIZE - intel->prim.current_offset) / (intel->vertex_size * 4)) + +#define ALLOC_VERTS(nr) intel_get_prim_space(intel, nr) -#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 +180,82 @@ 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++) - { - GLuint prim = VB->Primitive[i].mode; + tnl->Driver.Render.Start(ctx); + + for (i = 0; i < VB->PrimitiveCount; i++) { + GLuint prim = _tnl_translate_prim(&VB->Primitive[i]); 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 = -{ +static const struct tnl_pipeline_stage _intel_render_stage = { "intel render", NULL, NULL, NULL, NULL, - intel_run_render /* run */ + intel_run_render /* run */ +}; + +const struct tnl_pipeline_stage *intel_pipeline[] = { + &_tnl_vertex_transform_stage, + &_tnl_vertex_cull_stage, + &_tnl_normal_transform_stage, + &_tnl_lighting_stage, + &_tnl_fog_coordinate_stage, + &_tnl_texgen_stage, + &_tnl_texture_transform_stage, + &_tnl_point_attenuation_stage, + &_tnl_vertex_program_stage, +#if 1 + &_intel_render_stage, /* ADD: unclipped rastersetup-to-dma */ +#endif + &_tnl_render_stage, + 0, }; diff --git a/src/mesa/drivers/dri/i915/intel_rotate.c b/src/mesa/drivers/dri/i915/intel_rotate.c deleted file mode 100644 index a77640ee54..0000000000 --- a/src/mesa/drivers/dri/i915/intel_rotate.c +++ /dev/null @@ -1,221 +0,0 @@ - -/** - * Routines for simple 2D->2D transformations for rotated, flipped screens. - * - * XXX This code is not intel-specific. Move it into a common/utility - * someday. - */ - -#include "intel_rotate.h" - -#define MIN2(A, B) ( ((A) < (B)) ? (A) : (B) ) - -#define ABS(A) ( ((A) < 0) ? -(A) : (A) ) - - -void -matrix23Set(struct matrix23 *m, - 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; -} - - -/* - * Transform (x,y) coordinate by the given matrix. - */ -void -matrix23TransformCoordf(const struct matrix23 *m, float *x, float *y) -{ - const float x0 = *x; - const float y0 = *y; - - *x = m->m00 * x0 + m->m01 * y0 + m->m02; - *y = m->m10 * x0 + m->m11 * y0 + m->m12; -} - - -void -matrix23TransformCoordi(const struct matrix23 *m, int *x, int *y) -{ - const int x0 = *x; - const int y0 = *y; - - *x = m->m00 * x0 + m->m01 * y0 + m->m02; - *y = m->m10 * x0 + m->m11 * y0 + m->m12; -} - - -/* - * Transform a width and height by the given matrix. - * XXX this could be optimized quite a bit. - */ -void -matrix23TransformDistance(const struct matrix23 *m, int *xDist, int *yDist) -{ - int x0 = 0, y0 = 0; - int x1 = *xDist, y1 = 0; - int x2 = 0, y2 = *yDist; - matrix23TransformCoordi(m, &x0, &y0); - matrix23TransformCoordi(m, &x1, &y1); - matrix23TransformCoordi(m, &x2, &y2); - - *xDist = (x1 - x0) + (x2 - x0); - *yDist = (y1 - y0) + (y2 - y0); - - if (*xDist < 0) - *xDist = -*xDist; - if (*yDist < 0) - *yDist = -*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) -{ - int x0 = *x, y0 = *y; - int x1 = *x + *w, y1 = *y; - int x2 = *x + *w, y2 = *y + *h; - int x3 = *x, y3 = *y + *h; - matrix23TransformCoordi(m, &x0, &y0); - matrix23TransformCoordi(m, &x1, &y1); - matrix23TransformCoordi(m, &x2, &y2); - matrix23TransformCoordi(m, &x3, &y3); - *w = ABS(x1 - x0) + ABS(x2 - x1); - /**w = ABS(*w);*/ - *h = ABS(y1 - y0) + ABS(y2 - y1); - /**h = ABS(*h);*/ - *x = MIN2(x0, x1); - *x = MIN2(*x, x2); - *y = MIN2(y0, y1); - *y = MIN2(*y, y2); -} - - -/* - * Make rotation matrix for width X height screen. - */ -void -matrix23Rotate(struct matrix23 *m, int width, int height, int angle) -{ - switch (angle) { - case 0: - matrix23Set(m, 1, 0, 0, 0, 1, 0); - break; - case 90: - matrix23Set(m, 0, 1, 0, -1, 0, width); - break; - case 180: - matrix23Set(m, -1, 0, width, 0, -1, height); - break; - case 270: - matrix23Set(m, 0, -1, height, 1, 0, 0); - break; - default: - /*abort()*/; - } -} - - -/* - * Make flip/reflection matrix for width X height screen. - */ -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; - } - else { - m->m00 = 1; m->m01 = 0; m->m02 = 0; - } - if (yflip) { - m->m10 = 0; m->m11 = -1; m->m12 = height - 1; - } - else { - m->m10 = 0; m->m11 = 1; m->m12 = 0; - } -} - - -/* - * result = a * b - */ -void -matrix23Multiply(struct matrix23 *result, - const struct matrix23 *a, const struct matrix23 *b) -{ - result->m00 = a->m00 * b->m00 + a->m01 * b->m10; - result->m01 = a->m00 * b->m01 + a->m01 * b->m11; - result->m02 = a->m00 * b->m02 + a->m01 * b->m12 + a->m02; - - result->m10 = a->m10 * b->m00 + a->m11 * b->m10; - result->m11 = a->m10 * b->m01 + a->m11 * b->m11; - result->m12 = a->m10 * b->m02 + a->m11 * b->m12 + a->m12; -} - - -#if 000 - -#include <stdio.h> - -int -main(int argc, char *argv[]) -{ - int width = 500, height = 400; - int rot; - 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; - - - for (rot = 0; rot < 360; rot += 90) { - struct matrix23 rotate, flip, m; - int i; - - printf("Rot %d, xFlip %d, yFlip %d:\n", rot, fx, fy); - - /* make transformation matrix 'm' */ - matrix23Rotate(&rotate, width, height, rot); - matrix23Flip(&flip, width, height, fx, fy); - matrix23Multiply(&m, &rotate, &flip); - - /* xform four coords */ - for (i = 0; i < 4; i++) { - int x = coords[i][0]; - int y = coords[i][1]; - matrix23TransformCoordi(&m, &x, &y); - printf(" %d, %d -> %d %d\n", coords[i][0], coords[i][1], x, y); - } - - /* xform width, height */ - { - int x = width; - int y = height; - matrix23TransformDistance(&m, &x, &y); - printf(" %d x %d -> %d x %d\n", width, height, x, y); - } - - /* xform rect */ - { - int x = 50, y = 10, w = 200, h = 100; - matrix23TransformRect(&m, &x, &y, &w, &h); - printf(" %d,%d %d x %d -> %d, %d %d x %d\n", 50, 10, 200, 100, - x, y, w, h); - } - - } - - return 0; -} -#endif diff --git a/src/mesa/drivers/dri/i915/intel_rotate.h b/src/mesa/drivers/dri/i915/intel_rotate.h deleted file mode 100644 index 0da45d20ce..0000000000 --- a/src/mesa/drivers/dri/i915/intel_rotate.h +++ /dev/null @@ -1,41 +0,0 @@ -#ifndef INTEL_ROTATE_H -#define INTEL_ROTATE_H 1 - -struct matrix23 -{ - int m00, m01, m02; - int m10, m11, m12; -}; - - - -extern void -matrix23Set(struct matrix23 *m, - 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 -matrix23TransformCoordf(const struct matrix23 *m, float *x, float *y); - -extern void -matrix23TransformDistance(const struct matrix23 *m, int *xDist, int *yDist); - -extern void -matrix23TransformRect(const struct matrix23 *m, - int *x, int *y, int *w, int *h); - -extern void -matrix23Rotate(struct matrix23 *m, int width, int height, int angle); - -extern void -matrix23Flip(struct matrix23 *m, int width, int height, int xflip, int yflip); - -extern void -matrix23Multiply(struct matrix23 *result, - const struct matrix23 *a, const struct matrix23 *b); - - -#endif /* INTEL_ROTATE_H */ diff --git a/src/mesa/drivers/dri/i915/intel_screen.h b/src/mesa/drivers/dri/i915/intel_screen.h deleted file mode 100644 index 24cfd9bf8b..0000000000 --- a/src/mesa/drivers/dri/i915/intel_screen.h +++ /dev/null @@ -1,112 +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_INIT_H_ -#define _INTEL_INIT_H_ - -#include <sys/time.h> -#include "xmlconfig.h" -#include "dri_util.h" -#include "intel_rotate.h" -#include "i830_common.h" - - -/* This roughly corresponds to a gl_renderbuffer (Mesa 6.4) */ -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 */ -} intelRegion; - -typedef struct -{ - intelRegion front; - intelRegion back; - intelRegion rotated; - intelRegion depth; - intelRegion tex; - - int deviceID; - int width; - int height; - int mem; /* unused */ - - int cpp; /* for front and back buffers */ - int fbFormat; - - int logTextureGranularity; - - __DRIscreenPrivate *driScrnPriv; - unsigned int sarea_priv_offset; - - int drmMinor; - - int irq_active; - int allow_batchbuffer; - - struct matrix23 rotMatrix; - - int current_rotation; /* 0, 90, 180 or 270 */ - int rotatedWidth, rotatedHeight; - - /** - * Configuration cache with default values for all contexts - */ - driOptionCache optionCache; -} intelScreenPrivate; - - -extern GLboolean -intelMapScreenRegions(__DRIscreenPrivate *sPriv); - -extern void -intelUnmapScreenRegions(intelScreenPrivate *intelScreen); - -extern void -intelUpdateScreenFromSAREA(intelScreenPrivate *intelScreen, - drmI830Sarea *sarea); - -extern void -intelDestroyContext(__DRIcontextPrivate *driContextPriv); - -extern GLboolean -intelUnbindContext(__DRIcontextPrivate *driContextPriv); - -extern GLboolean -intelMakeCurrent(__DRIcontextPrivate *driContextPriv, - __DRIdrawablePrivate *driDrawPriv, - __DRIdrawablePrivate *driReadPriv); - -extern void -intelSwapBuffers(__DRIdrawablePrivate *dPriv); - -extern void -intelCopySubBuffer( __DRIdrawablePrivate *dPriv, int x, int y, int w, int h ); - -#endif diff --git a/src/mesa/drivers/dri/i915/intel_span.c b/src/mesa/drivers/dri/i915/intel_span.c index c3ffc4b2ac..05e5e8e583 100644..120000 --- a/src/mesa/drivers/dri/i915/intel_span.c +++ b/src/mesa/drivers/dri/i915/intel_span.c @@ -1,258 +1 @@ -/************************************************************************** - * - * 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 "macros.h" -#include "mtypes.h" -#include "colormac.h" - -#include "intel_screen.h" - -#include "intel_span.h" -#include "intel_ioctl.h" -#include "swrast/swrast.h" - - -#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 HW_LOCK() - -#define HW_UNLOCK() - -/* 16 bit, 565 rgb 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 WRITE_DEPTH( _x, _y, d ) \ - *(GLushort *)(buf + (_x)*2 + (_y)*pitch) = d; - -#define READ_DEPTH( d, _x, _y ) \ - d = *(GLushort *)(buf + (_x)*2 + (_y)*pitch); - - -#define TAG(x) intel##x##_z16 -#include "depthtmp.h" - - -#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 WRITE_PIXEL(_x, _y, p) \ - *(GLuint *)(buf + _x*4 + _y*pitch) = p - - -#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" - - -/* 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; \ -} - -#define READ_DEPTH( d, _x, _y ) \ - d = *(GLuint *)(buf + (_x)*4 + (_y)*pitch) & 0xffffff; - - -#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; \ -} - -#define READ_STENCIL( d, _x, _y ) \ - d = *(GLuint *)(buf + (_x)*4 + (_y)*pitch) >> 24; - -#define TAG(x) intel##x##_z24_s8 -#include "stenciltmp.h" - - -/* Move locking out to get reasonable span performance. - */ -void intelSpanRenderStart( GLcontext *ctx ) -{ - intelContextPtr intel = INTEL_CONTEXT(ctx); - - intelFlush(&intel->ctx); - LOCK_HARDWARE(intel); - intelWaitForIdle(intel); -} - -void intelSpanRenderFinish( GLcontext *ctx ) -{ - intelContextPtr intel = INTEL_CONTEXT( ctx ); - _swrast_flush( ctx ); - UNLOCK_HARDWARE( intel ); -} - -void intelInitSpanFuncs( GLcontext *ctx ) -{ - struct swrast_device_driver *swdd = _swrast_GetDeviceDriverReference(ctx); - swdd->SpanRenderStart = intelSpanRenderStart; - swdd->SpanRenderFinish = intelSpanRenderFinish; -} - - -/** - * Plug in the Get/Put routines for the given driRenderbuffer. - */ -void -intelSetSpanFunctions(driRenderbuffer *drb, const GLvisual *vis) -{ - 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); - } - } - else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT16) { - intelInitDepthPointers_z16(&drb->Base); - } - else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT24) { - intelInitDepthPointers_z24_s8(&drb->Base); - } - else if (drb->Base.InternalFormat == GL_STENCIL_INDEX8_EXT) { - intelInitStencilPointers_z24_s8(&drb->Base); - } -} +../intel/intel_span.c
\ No newline at end of file diff --git a/src/mesa/drivers/dri/i915/intel_state.c b/src/mesa/drivers/dri/i915/intel_state.c index e5988a5ed6..d1ca11dec4 100644 --- a/src/mesa/drivers/dri/i915/intel_state.c +++ b/src/mesa/drivers/dri/i915/intel_state.c @@ -30,252 +30,266 @@ #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_shadow_compare_func( GLenum func ) { switch(func) { case GL_NEVER: - return COMPAREFUNC_NEVER; + return COMPAREFUNC_ALWAYS; case GL_LESS: - return COMPAREFUNC_LESS; + return COMPAREFUNC_LEQUAL; case GL_LEQUAL: - return COMPAREFUNC_LEQUAL; + return COMPAREFUNC_LESS; case GL_GREATER: - return COMPAREFUNC_GREATER; + return COMPAREFUNC_GEQUAL; case GL_GEQUAL: - return COMPAREFUNC_GEQUAL; + return COMPAREFUNC_GREATER; case GL_NOTEQUAL: - return COMPAREFUNC_NOTEQUAL; - case GL_EQUAL: return COMPAREFUNC_EQUAL; + case GL_EQUAL: + return COMPAREFUNC_NOTEQUAL; case GL_ALWAYS: - return COMPAREFUNC_ALWAYS; + return COMPAREFUNC_NEVER; + } + + fprintf(stderr, "Unknown value in %s: %x\n", __FUNCTION__, func); + return COMPAREFUNC_NEVER; +} + +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; } 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 (ctx->DrawBuffer->Name) { + /* User created FBO */ + struct intel_renderbuffer *irb + = intel_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[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; + } - if (intel->driDrawable) - h = intel->driDrawable->h + SUBPIXEL_Y; + m[MAT_SX] = v[MAT_SX]; + m[MAT_TX] = v[MAT_TX]; - /* 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; + + 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..d77ce749a3 100644..120000 --- a/src/mesa/drivers/dri/i915/intel_tex.c +++ b/src/mesa/drivers/dri/i915/intel_tex.c @@ -1,877 +1 @@ -/************************************************************************** - * - * 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_tex.h" -#include "intel_ioctl.h" - - - -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) - -{ - 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; - } - - /* 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; - } -} - - - -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 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 ) -{ - driTextureObject * t = (driTextureObject *) texObj->DriverData; - - assert(t); - intelFlush( ctx ); - driSwapOutTextureObject( t ); - - _mesa_store_texsubimage1d(ctx, target, level, xoffset, width, - format, type, pixels, packing, texObj, - texImage); -} - - -/* 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 ) -{ - 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; - } - - 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 ); - - t->dirty_images[face] |= (1 << level); - } -} - -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 ); - - _mesa_store_texsubimage2d(ctx, target, level, xoffset, yoffset, width, - height, format, type, pixels, packing, texObj, - texImage); - - t->dirty_images[face] |= (1 << level); - } -} - -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; - } - - 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 ) -{ - 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; - } - - 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); -} - - -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); -} - - -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 ) -{ - driTextureObject * t = (driTextureObject *) texObj->DriverData; - - assert( t ); /* this _should_ be true */ - driSwapOutTextureObject( t ); - - _mesa_store_texsubimage3d(ctx, target, level, xoffset, yoffset, zoffset, - width, height, depth, - format, type, pixels, packing, texObj, texImage); - - t->dirty_images[0] |= (1 << level); -} - - - - -static void intelDeleteTexture( GLcontext *ctx, struct gl_texture_object *tObj ) -{ - driTextureObject * t = (driTextureObject *) tObj->DriverData; - - if ( t != NULL ) { - intelFlush( ctx ); - driDestroyTextureObject( t ); - } - - /* Free mipmap images and the texture object itself */ - _mesa_delete_texture_object(ctx, tObj); -} - - -static const struct gl_texture_format * -intelChooseTextureFormat( GLcontext *ctx, GLint internalFormat, - GLenum format, GLenum type ) -{ - 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 */ -} - - - -void intelDestroyTexObj(intelContextPtr intel, intelTextureObjectPtr t) -{ - unsigned i; - - if ( intel == NULL ) - return; - - if ( t->age > intel->dirtyAge ) - intel->dirtyAge = t->age; - - for ( i = 0 ; i < MAX_TEXTURE_UNITS ; i++ ) { - if ( t == intel->CurrentTexObj[ i ] ) - intel->CurrentTexObj[ i ] = NULL; - } -} - - - -/* 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 ) -{ - - 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; - } -} - -/** - * 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; -} - - -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; -} +../intel/intel_tex.c
\ No newline at end of file diff --git a/src/mesa/drivers/dri/i915/intel_tex.h b/src/mesa/drivers/dri/i915/intel_tex.h deleted file mode 100644 index 9b7e550232..0000000000 --- a/src/mesa/drivers/dri/i915/intel_tex.h +++ /dev/null @@ -1,45 +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 INTELTEX_INC -#define INTELTEX_INC - -#include "mtypes.h" -#include "intel_context.h" -#include "texmem.h" - - -void intelInitTextureFuncs( struct dd_function_table *functions ); - -void intelDestroyTexObj( intelContextPtr intel, intelTextureObjectPtr t ); -int intelUploadTexImages( intelContextPtr intel, intelTextureObjectPtr t, - GLuint face ); - -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 120000 index 0000000000..87196c5d1e --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_tex_copy.c @@ -0,0 +1 @@ +../intel/intel_tex_copy.c
\ No newline at end of file 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 120000 index 0000000000..3415f75470 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_tex_format.c @@ -0,0 +1 @@ +../intel/intel_tex_format.c
\ No newline at end of file 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 120000 index 0000000000..567abe4974 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_tex_image.c @@ -0,0 +1 @@ +../intel/intel_tex_image.c
\ No newline at end of file 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 120000 index 0000000000..b3a8a3d7ca --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_tex_subimage.c @@ -0,0 +1 @@ +../intel/intel_tex_subimage.c
\ No newline at end of file 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 120000 index 0000000000..41a75674c2 --- /dev/null +++ b/src/mesa/drivers/dri/i915/intel_tex_validate.c @@ -0,0 +1 @@ +../intel/intel_tex_validate.c
\ No newline at end of file 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..5f4b852212 100644 --- a/src/mesa/drivers/dri/i915/intel_tris.c +++ b/src/mesa/drivers/dri/i915/intel_tris.c @@ -25,10 +25,18 @@ * **************************************************************************/ +/** @file intel_tris.c + * + * This file contains functions for managing the vertex buffer and emitting + * primitives into it. + */ + #include "glheader.h" #include "context.h" #include "macros.h" #include "enums.h" +#include "texobj.h" +#include "state.h" #include "dd.h" #include "swrast/swrast.h" @@ -38,19 +46,191 @@ #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" +#include "intel_chipset.h" +#include "i830_context.h" +#include "i830_reg.h" + +static void intelRenderPrimitive(GLcontext * ctx, GLenum prim); +static void intelRasterPrimitive(GLcontext * ctx, GLenum rprim, + GLuint hwprim); + +/** Sets the primitive type for a primitive sequence, flushing as needed. */ +void intel_set_prim(struct intel_context *intel, uint32_t prim) +{ + if (prim != intel->prim.primitive) { + INTEL_FIREVERTICES(intel); + intel->prim.primitive = prim; + } +} + +/** Returns mapped VB space for the given number of vertices */ +uint32_t *intel_get_prim_space(struct intel_context *intel, unsigned int count) +{ + uint32_t *addr; + + /* Check for space in the existing VB */ + if (intel->prim.vb_bo == NULL || + (intel->prim.current_offset + + count * intel->vertex_size * 4) > INTEL_VB_SIZE || + (intel->prim.count + count) >= (1 << 16)) { + /* Flush existing prim if any */ + INTEL_FIREVERTICES(intel); + + intel_finish_vb(intel); + + /* Start a new VB */ + if (intel->prim.vb == NULL) + intel->prim.vb = malloc(INTEL_VB_SIZE); + intel->prim.vb_bo = dri_bo_alloc(intel->bufmgr, "vb", + INTEL_VB_SIZE, 4); + intel->prim.start_offset = 0; + intel->prim.current_offset = 0; + } + + intel->prim.flush = intel_flush_prim; + + addr = (uint32_t *)(intel->prim.vb + intel->prim.current_offset); + intel->prim.current_offset += intel->vertex_size * 4 * count; + intel->prim.count += count; + + return addr; +} + +/** Dispatches the accumulated primitive to the batchbuffer. */ +void intel_flush_prim(struct intel_context *intel) +{ + BATCH_LOCALS; + dri_bo *aper_array[2]; + dri_bo *vb_bo; + + /* Must be called after an intel_start_prim. */ + assert(intel->prim.primitive != ~0); + + if (intel->prim.count == 0) + return; + + /* Keep a reference on the BO as it may get finished as we start the + * batch emit. + */ + vb_bo = intel->prim.vb_bo; + dri_bo_reference(vb_bo); + + intel_wait_flips(intel); + + intel->vtbl.emit_state(intel); + + aper_array[0] = intel->batch->buf; + aper_array[1] = vb_bo; + if (dri_bufmgr_check_aperture_space(aper_array, 2)) { + intel_batchbuffer_flush(intel->batch); + intel->vtbl.emit_state(intel); + } + + /* Ensure that we don't start a new batch for the following emit, which + * depends on the state just emitted. emit_state should be making sure we + * have the space for this. + */ + intel->no_batch_wrap = GL_TRUE; + + /* Check that we actually emitted the state into this batch, using the + * UPLOAD_CTX bit as the signal. + */ + assert((intel->batch->dirty_state & (1<<1)) == 0); + +#if 0 + printf("emitting %d..%d=%d vertices size %d\n", intel->prim.start_offset, + intel->prim.current_offset, intel->prim.count, + intel->vertex_size * 4); +#endif -/* XXX we shouldn't include these headers in this file, but we need them - * for fallbackStrings, below. + if (IS_9XX(intel->intelScreen->deviceID)) { + BEGIN_BATCH(5, LOOP_CLIPRECTS); + OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | + I1_LOAD_S(0) | I1_LOAD_S(1) | 1); + assert((intel->prim.start_offset & !S0_VB_OFFSET_MASK) == 0); + OUT_RELOC(vb_bo, I915_GEM_DOMAIN_VERTEX, 0, + intel->prim.start_offset); + OUT_BATCH((intel->vertex_size << S1_VERTEX_WIDTH_SHIFT) | + (intel->vertex_size << S1_VERTEX_PITCH_SHIFT)); + + OUT_BATCH(_3DPRIMITIVE | + PRIM_INDIRECT | + PRIM_INDIRECT_SEQUENTIAL | + intel->prim.primitive | + intel->prim.count); + OUT_BATCH(0); /* Beginning vertex index */ + ADVANCE_BATCH(); + } else { + struct i830_context *i830 = i830_context(&intel->ctx); + + BEGIN_BATCH(5, LOOP_CLIPRECTS); + OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | + I1_LOAD_S(0) | I1_LOAD_S(2) | 1); + /* S0 */ + assert((intel->prim.start_offset & !S0_VB_OFFSET_MASK_830) == 0); + OUT_RELOC(vb_bo, I915_GEM_DOMAIN_VERTEX, 0, + intel->prim.start_offset | + (intel->vertex_size << S0_VB_PITCH_SHIFT_830) | + S0_VB_ENABLE_830); + /* S1 + * This is somewhat unfortunate -- VB width is tied up with + * vertex format data that we've already uploaded through + * _3DSTATE_VFT[01]_CMD. We may want to replace emits of VFT state with + * STATE_IMMEDIATE_1 like this to avoid duplication. + */ + OUT_BATCH((i830->state.Ctx[I830_CTXREG_VF] & VFT0_TEX_COUNT_MASK) >> + VFT0_TEX_COUNT_SHIFT << S2_TEX_COUNT_SHIFT_830 | + (i830->state.Ctx[I830_CTXREG_VF2] << 16) | + intel->vertex_size << S2_VERTEX_0_WIDTH_SHIFT_830); + + OUT_BATCH(_3DPRIMITIVE | + PRIM_INDIRECT | + PRIM_INDIRECT_SEQUENTIAL | + intel->prim.primitive | + intel->prim.count); + OUT_BATCH(0); /* Beginning vertex index */ + ADVANCE_BATCH(); + } + + intel->no_batch_wrap = GL_FALSE; + + intel->prim.flush = NULL; + intel->prim.start_offset = intel->prim.current_offset; + if (!IS_9XX(intel->intelScreen->deviceID)) + intel->prim.start_offset = ALIGN(intel->prim.start_offset, 128); + intel->prim.count = 0; + + dri_bo_unreference(vb_bo); +} + +/** + * Uploads the locally-accumulated VB into the buffer object. + * + * This avoids us thrashing the cachelines in and out as the buffer gets + * filled, dispatched, then reused as the hardware completes rendering from it, + * and also lets us clflush less if we dispatch with a partially-filled VB. + * + * This is called normally from get_space when we're finishing a BO, but also + * at batch flush time so that we don't try accessing the contents of a + * just-dispatched buffer. */ -#include "i830_context.h" -#include "i915_context.h" +void intel_finish_vb(struct intel_context *intel) +{ + if (intel->prim.vb_bo == NULL) + return; -static void intelRenderPrimitive( GLcontext *ctx, GLenum prim ); -static void intelRasterPrimitive( GLcontext *ctx, GLenum rprim, GLuint hwprim ); + dri_bo_subdata(intel->prim.vb_bo, 0, intel->prim.start_offset, + intel->prim.vb); + dri_bo_unreference(intel->prim.vb_bo); + intel->prim.vb_bo = NULL; +} /*********************************************************************** * Emit primitives as inline vertices * @@ -69,75 +249,81 @@ 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 = intel_get_prim_space(intel, 6); 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 = intel_get_prim_space(intel, 3); 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 = intel_get_prim_space(intel, 2); 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 = intel_get_prim_space(intel, 1); 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; + *(float *) &vb[1] = v0->v.y; + for (j = 2; j < vertsize; j++) + vb[j] = v0->ui[j]; } @@ -146,13 +332,17 @@ 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 +360,19 @@ 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 +385,59 @@ 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 ); + GLfloat *v0_wpos = (GLfloat *)((char *)v0 + offset); + GLfloat *v1_wpos = (GLfloat *)((char *)v1 + offset); + GLfloat *v2_wpos = (GLfloat *)((char *)v2 + offset); + + __memcpy(v0_wpos, v0, size); + __memcpy(v1_wpos, v1, size); + __memcpy(v2_wpos, v2, size); + + v0_wpos[1] = -v0_wpos[1] + intel->driDrawable->h; + v1_wpos[1] = -v1_wpos[1] + intel->driDrawable->h; + v2_wpos[1] = -v2_wpos[1] + intel->driDrawable->h; + - intel_draw_triangle( intel, v0, v1, v2 ); + 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; + GLfloat *v0_wpos = (GLfloat *)((char *)v0 + offset); + GLfloat *v1_wpos = (GLfloat *)((char *)v1 + offset); + + __memcpy(v0_wpos, v0, size); + __memcpy(v1_wpos, v1, size); - __memcpy( ((char *)v0) + offset, v0, size ); - __memcpy( ((char *)v1) + offset, v1, size ); + v0_wpos[1] = -v0_wpos[1] + intel->driDrawable->h; + v1_wpos[1] = -v1_wpos[1] + intel->driDrawable->h; - 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; + GLfloat *v0_wpos = (GLfloat *)((char *)v0 + offset); - __memcpy( ((char *)v0) + offset, v0, size ); + __memcpy(v0_wpos, v0, size); + v0_wpos[1] = -v0_wpos[1] + intel->driDrawable->h; - intel_draw_point( intel, v0 ); + intel_draw_point(intel, v0); } @@ -290,11 +494,12 @@ 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]; @@ -355,10 +560,10 @@ do { \ #define VERT_RESTORE_SPEC( idx ) if (specoffset) v[idx]->ui[specoffset] = spec[idx] #define LOCAL_VARS(n) \ - intelContextPtr intel = INTEL_CONTEXT(ctx); \ - GLuint color[n], spec[n]; \ - GLuint coloroffset = intel->coloroffset; \ - GLboolean specoffset = intel->specoffset; \ + struct intel_context *intel = intel_context(ctx); \ + GLuint color[n] = { 0, }, spec[n] = { 0, }; \ + GLuint coloroffset = intel->coloroffset; \ + GLboolean specoffset = intel->specoffset; \ (void) color; (void) spec; (void) coloroffset; (void) specoffset; @@ -366,7 +571,7 @@ 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 +661,8 @@ 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 +693,8 @@ 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 +702,20 @@ 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 +723,18 @@ 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 +742,13 @@ 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 +765,7 @@ 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 +788,10 @@ 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 +800,40 @@ 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 = intel_get_prim_space(intel, (n - 2) * 3); + 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 +844,75 @@ 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 +925,21 @@ 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 +960,52 @@ 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 +1014,33 @@ 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); + + intel_set_prim(intel, hwprim); + } } -/* - */ -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 +1055,54 @@ 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 +1112,19 @@ 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 +1132,99 @@ 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. */ +/**********************************************************************/ +static 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; + GLboolean was_locked = intel->locked; + unsigned int saved_vertex_size = intel->vertex_size; + + if (!was_locked) + LOCK_HARDWARE(intel); + + intel->vertex_size = 6; + + /* All 3d primitives should be emitted with LOOP_CLIPRECTS, + * otherwise the drawing origin (DR4) might not be set correctly. + */ + intel_set_prim(intel, PRIM3D_TRIFAN); + vb = (union fi *) intel_get_prim_space(intel, n); + + 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); + + intel->vertex_size = saved_vertex_size; + + if (!was_locked) + UNLOCK_HARDWARE(intel); +} + +static 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); +} @@ -924,8 +1233,10 @@ void intelFallback( intelContextPtr intel, GLuint bit, GLboolean mode ) /**********************************************************************/ -void intelInitTriFuncs( GLcontext *ctx ) +void +intelInitTriFuncs(GLcontext * ctx) { + struct intel_context *intel = intel_context(ctx); TNLcontext *tnl = TNL_CONTEXT(ctx); static int firsttime = 1; @@ -942,4 +1253,6 @@ void intelInitTriFuncs( GLcontext *ctx ) tnl->Driver.Render.BuildVertices = _tnl_build_vertices; tnl->Driver.Render.CopyPV = _tnl_copy_pv; tnl->Driver.Render.Interp = _tnl_interp; + + intel->vtbl.meta_draw_quad = intel_meta_draw_quad; } diff --git a/src/mesa/drivers/dri/i915/intel_tris.h b/src/mesa/drivers/dri/i915/intel_tris.h index d7e382fdb3..0e08986221 100644 --- a/src/mesa/drivers/dri/i915/intel_tris.h +++ b/src/mesa/drivers/dri/i915/intel_tris.h @@ -30,6 +30,10 @@ #include "mtypes.h" +#define INTEL_VB_SIZE (32 * 1024) +/** 3 dwords of state_immediate and 2 of 3dprim, in intel_flush_prim */ +#define INTEL_PRIM_EMIT_SIZE (5 * 4) + #define _INTEL_NEW_RENDERSTATE (_DD_NEW_LINE_STIPPLE | \ _DD_NEW_TRI_UNFILLED | \ _DD_NEW_TRI_LIGHT_TWOSIDE | \ @@ -38,9 +42,13 @@ _NEW_PROGRAM | \ _NEW_POLYGONSTIPPLE) -extern void intelInitTriFuncs( GLcontext *ctx ); +extern void intelInitTriFuncs(GLcontext * ctx); + +extern void intelChooseRenderState(GLcontext * ctx); -extern void intelPrintRenderState( const char *msg, GLuint state ); -extern void intelChooseRenderState( GLcontext *ctx ); +void intel_set_prim(struct intel_context *intel, uint32_t prim); +GLuint *intel_get_prim_space(struct intel_context *intel, unsigned int count); +void intel_flush_prim(struct intel_context *intel); +void intel_finish_vb(struct intel_context *intel); #endif diff --git a/src/mesa/drivers/dri/i915/server/intel.h b/src/mesa/drivers/dri/i915/server/intel.h deleted file mode 100644 index d7858a20c8..0000000000 --- a/src/mesa/drivers/dri/i915/server/intel.h +++ /dev/null @@ -1,328 +0,0 @@ -#ifndef _INTEL_H_ -#define _INTEL_H_ - -#include "xf86drm.h" /* drm_handle_t, etc */ - -/* Intel */ -#ifndef PCI_CHIP_I810 -#define PCI_CHIP_I810 0x7121 -#define PCI_CHIP_I810_DC100 0x7123 -#define PCI_CHIP_I810_E 0x7125 -#define PCI_CHIP_I815 0x1132 -#define PCI_CHIP_I810_BRIDGE 0x7120 -#define PCI_CHIP_I810_DC100_BRIDGE 0x7122 -#define PCI_CHIP_I810_E_BRIDGE 0x7124 -#define PCI_CHIP_I815_BRIDGE 0x1130 -#endif - -#define PCI_CHIP_845_G 0x2562 -#define PCI_CHIP_I830_M 0x3577 - -#ifndef PCI_CHIP_I855_GM -#define PCI_CHIP_I855_GM 0x3582 -#define PCI_CHIP_I855_GM_BRIDGE 0x3580 -#endif - -#ifndef PCI_CHIP_I865_G -#define PCI_CHIP_I865_G 0x2572 -#define PCI_CHIP_I865_G_BRIDGE 0x2570 -#endif - -#ifndef PCI_CHIP_I915_G -#define PCI_CHIP_I915_G 0x2582 -#define PCI_CHIP_I915_G_BRIDGE 0x2580 -#endif - -#ifndef PCI_CHIP_I915_GM -#define PCI_CHIP_I915_GM 0x2592 -#define PCI_CHIP_I915_GM_BRIDGE 0x2590 -#endif - -#ifndef PCI_CHIP_E7221_G -#define PCI_CHIP_E7221_G 0x258A -/* Same as I915_G_BRIDGE */ -#define PCI_CHIP_E7221_G_BRIDGE 0x2580 -#endif - -#ifndef PCI_CHIP_I945_G -#define PCI_CHIP_I945_G 0x2772 -#define PCI_CHIP_I945_G_BRIDGE 0x2770 -#endif - -#ifndef PCI_CHIP_I945_GM -#define PCI_CHIP_I945_GM 0x27A2 -#define PCI_CHIP_I945_GM_BRIDGE 0x27A0 -#endif - -#define IS_I810(pI810) (pI810->Chipset == PCI_CHIP_I810 || \ - pI810->Chipset == PCI_CHIP_I810_DC100 || \ - pI810->Chipset == PCI_CHIP_I810_E) -#define IS_I815(pI810) (pI810->Chipset == PCI_CHIP_I815) -#define IS_I830(pI810) (pI810->Chipset == PCI_CHIP_I830_M) -#define IS_845G(pI810) (pI810->Chipset == PCI_CHIP_845_G) -#define IS_I85X(pI810) (pI810->Chipset == PCI_CHIP_I855_GM) -#define IS_I852(pI810) (pI810->Chipset == PCI_CHIP_I855_GM && (pI810->variant == I852_GM || pI810->variant == I852_GME)) -#define IS_I855(pI810) (pI810->Chipset == PCI_CHIP_I855_GM && (pI810->variant == I855_GM || pI810->variant == I855_GME)) -#define IS_I865G(pI810) (pI810->Chipset == PCI_CHIP_I865_G) - -#define IS_I915G(pI810) (pI810->Chipset == PCI_CHIP_I915_G || pI810->Chipset == PCI_CHIP_E7221_G) -#define IS_I915GM(pI810) (pI810->Chipset == PCI_CHIP_I915_GM) -#define IS_I945G(pI810) (pI810->Chipset == PCI_CHIP_I945_G) -#define IS_I945GM(pI810) (pI810->Chipset == PCI_CHIP_I945_GM) -#define IS_I9XX(pI810) (IS_I915G(pI810) || IS_I915GM(pI810) || IS_I945G(pI810) || IS_I945GM(pI810)) - -#define IS_MOBILE(pI810) (IS_I830(pI810) || IS_I85X(pI810) || IS_I915GM(pI810) || IS_I945GM(pI810)) - -#define I830_GMCH_CTRL 0x52 - - -#define I830_GMCH_GMS_MASK 0x70 -#define I830_GMCH_GMS_DISABLED 0x00 -#define I830_GMCH_GMS_LOCAL 0x10 -#define I830_GMCH_GMS_STOLEN_512 0x20 -#define I830_GMCH_GMS_STOLEN_1024 0x30 -#define I830_GMCH_GMS_STOLEN_8192 0x40 - -#define I855_GMCH_GMS_MASK (0x7 << 4) -#define I855_GMCH_GMS_DISABLED 0x00 -#define I855_GMCH_GMS_STOLEN_1M (0x1 << 4) -#define I855_GMCH_GMS_STOLEN_4M (0x2 << 4) -#define I855_GMCH_GMS_STOLEN_8M (0x3 << 4) -#define I855_GMCH_GMS_STOLEN_16M (0x4 << 4) -#define I855_GMCH_GMS_STOLEN_32M (0x5 << 4) -#define I915G_GMCH_GMS_STOLEN_48M (0x6 << 4) -#define I915G_GMCH_GMS_STOLEN_64M (0x7 << 4) - -typedef unsigned char Bool; -#define TRUE 1 -#define FALSE 0 - -#define PIPE_NONE 0<<0 -#define PIPE_CRT 1<<0 -#define PIPE_TV 1<<1 -#define PIPE_DFP 1<<2 -#define PIPE_LFP 1<<3 -#define PIPE_CRT2 1<<4 -#define PIPE_TV2 1<<5 -#define PIPE_DFP2 1<<6 -#define PIPE_LFP2 1<<7 - -typedef struct _I830MemPool *I830MemPoolPtr; -typedef struct _I830MemRange *I830MemRangePtr; -typedef struct _I830MemRange { - long Start; - long End; - long Size; - unsigned long Physical; - unsigned long Offset; /* Offset of AGP-allocated portion */ - unsigned long Alignment; - drm_handle_t Key; - unsigned long Pitch; // add pitch - I830MemPoolPtr Pool; -} I830MemRange; - -typedef struct _I830MemPool { - I830MemRange Total; - I830MemRange Free; - I830MemRange Fixed; - I830MemRange Allocated; -} I830MemPool; - -typedef struct { - int tail_mask; - I830MemRange mem; - unsigned char *virtual_start; - int head; - int tail; - int space; -} I830RingBuffer; - -typedef struct _I830Rec { - unsigned char *MMIOBase; - unsigned char *FbBase; - int cpp; - - unsigned int bios_version; - - /* These are set in PreInit and never changed. */ - long FbMapSize; - long TotalVideoRam; - I830MemRange StolenMemory; /* pre-allocated memory */ - long BIOSMemorySize; /* min stolen pool size */ - int BIOSMemSizeLoc; - - /* These change according to what has been allocated. */ - long FreeMemory; - I830MemRange MemoryAperture; - I830MemPool StolenPool; - long allocatedMemory; - - /* Regions allocated either from the above pools, or from agpgart. */ - /* for single and dual head configurations */ - I830MemRange FrontBuffer; - I830MemRange FrontBuffer2; - I830MemRange Scratch; - I830MemRange Scratch2; - - I830RingBuffer *LpRing; - - I830MemRange BackBuffer; - I830MemRange DepthBuffer; - I830MemRange TexMem; - int TexGranularity; - I830MemRange ContextMem; - int drmMinor; - Bool have3DWindows; - - Bool NeedRingBufferLow; - Bool allowPageFlip; - Bool disableTiling; - - int Chipset; - unsigned long LinearAddr; - unsigned long MMIOAddr; - - drmSize registerSize; /**< \brief MMIO register map size */ - drm_handle_t registerHandle; /**< \brief MMIO register map handle */ - // IOADDRESS ioBase; - int irq; /**< \brief IRQ number */ - int GttBound; - - drm_handle_t ring_map; - unsigned int Fence[8]; - -} I830Rec; - -/* - * 12288 is set as the maximum, chosen because it is enough for - * 1920x1440@32bpp with a 2048 pixel line pitch with some to spare. - */ -#define I830_MAXIMUM_VBIOS_MEM 12288 -#define I830_DEFAULT_VIDEOMEM_2D (MB(32) / 1024) -#define I830_DEFAULT_VIDEOMEM_3D (MB(64) / 1024) - -/* Flags for memory allocation function */ -#define FROM_ANYWHERE 0x00000000 -#define FROM_POOL_ONLY 0x00000001 -#define FROM_NEW_ONLY 0x00000002 -#define FROM_MASK 0x0000000f - -#define ALLOCATE_AT_TOP 0x00000010 -#define ALLOCATE_AT_BOTTOM 0x00000020 -#define FORCE_GAPS 0x00000040 - -#define NEED_PHYSICAL_ADDR 0x00000100 -#define ALIGN_BOTH_ENDS 0x00000200 -#define FORCE_LOW 0x00000400 - -#define ALLOC_NO_TILING 0x00001000 -#define ALLOC_INITIAL 0x00002000 - -#define ALLOCATE_DRY_RUN 0x80000000 - -/* Chipset registers for VIDEO BIOS memory RW access */ -#define _855_DRAM_RW_CONTROL 0x58 -#define _845_DRAM_RW_CONTROL 0x90 -#define DRAM_WRITE 0x33330000 - -#define KB(x) ((x) * 1024) -#define MB(x) ((x) * KB(1024)) - -#define GTT_PAGE_SIZE KB(4) -#define ROUND_TO(x, y) (((x) + (y) - 1) / (y) * (y)) -#define ROUND_DOWN_TO(x, y) ((x) / (y) * (y)) -#define ROUND_TO_PAGE(x) ROUND_TO((x), GTT_PAGE_SIZE) -#define ROUND_TO_MB(x) ROUND_TO((x), MB(1)) -#define PRIMARY_RINGBUFFER_SIZE KB(128) - - -/* Ring buffer registers, p277, overview p19 - */ -#define LP_RING 0x2030 -#define HP_RING 0x2040 - -#define RING_TAIL 0x00 -#define TAIL_ADDR 0x000FFFF8 -#define I830_TAIL_MASK 0x001FFFF8 - -#define RING_HEAD 0x04 -#define HEAD_WRAP_COUNT 0xFFE00000 -#define HEAD_WRAP_ONE 0x00200000 -#define HEAD_ADDR 0x001FFFFC -#define I830_HEAD_MASK 0x001FFFFC - -#define RING_START 0x08 -#define START_ADDR 0x03FFFFF8 -#define I830_RING_START_MASK 0xFFFFF000 - -#define RING_LEN 0x0C -#define RING_NR_PAGES 0x001FF000 -#define I830_RING_NR_PAGES 0x001FF000 -#define RING_REPORT_MASK 0x00000006 -#define RING_REPORT_64K 0x00000002 -#define RING_REPORT_128K 0x00000004 -#define RING_NO_REPORT 0x00000000 -#define RING_VALID_MASK 0x00000001 -#define RING_VALID 0x00000001 -#define RING_INVALID 0x00000000 - - -/* Fence/Tiling ranges [0..7] - */ -#define FENCE 0x2000 -#define FENCE_NR 8 - -#define I915G_FENCE_START_MASK 0x0ff00000 - -#define I830_FENCE_START_MASK 0x07f80000 - -#define FENCE_START_MASK 0x03F80000 -#define FENCE_X_MAJOR 0x00000000 -#define FENCE_Y_MAJOR 0x00001000 -#define FENCE_SIZE_MASK 0x00000700 -#define FENCE_SIZE_512K 0x00000000 -#define FENCE_SIZE_1M 0x00000100 -#define FENCE_SIZE_2M 0x00000200 -#define FENCE_SIZE_4M 0x00000300 -#define FENCE_SIZE_8M 0x00000400 -#define FENCE_SIZE_16M 0x00000500 -#define FENCE_SIZE_32M 0x00000600 -#define FENCE_SIZE_64M 0x00000700 -#define I915G_FENCE_SIZE_1M 0x00000000 -#define I915G_FENCE_SIZE_2M 0x00000100 -#define I915G_FENCE_SIZE_4M 0x00000200 -#define I915G_FENCE_SIZE_8M 0x00000300 -#define I915G_FENCE_SIZE_16M 0x00000400 -#define I915G_FENCE_SIZE_32M 0x00000500 -#define I915G_FENCE_SIZE_64M 0x00000600 -#define I915G_FENCE_SIZE_128M 0x00000700 -#define FENCE_PITCH_1 0x00000000 -#define FENCE_PITCH_2 0x00000010 -#define FENCE_PITCH_4 0x00000020 -#define FENCE_PITCH_8 0x00000030 -#define FENCE_PITCH_16 0x00000040 -#define FENCE_PITCH_32 0x00000050 -#define FENCE_PITCH_64 0x00000060 -#define FENCE_VALID 0x00000001 - -#include <mmio.h> - -# define MMIO_IN8(base, offset) \ - *(volatile unsigned char *)(((unsigned char*)(base)) + (offset)) -# define MMIO_IN32(base, offset) \ - read_MMIO_LE32(base, offset) -# define MMIO_OUT8(base, offset, val) \ - *(volatile unsigned char *)(((unsigned char*)(base)) + (offset)) = (val) -# define MMIO_OUT32(base, offset, val) \ - *(volatile unsigned int *)(void *)(((unsigned char*)(base)) + (offset)) = CPU_TO_LE32(val) - - - /* Memory mapped register access macros */ -#define INREG8(addr) MMIO_IN8(MMIO, addr) -#define INREG(addr) MMIO_IN32(MMIO, addr) -#define OUTREG8(addr, val) MMIO_OUT8(MMIO, addr, val) -#define OUTREG(addr, val) MMIO_OUT32(MMIO, addr, val) - -#define DSPABASE 0x70184 - -#endif diff --git a/src/mesa/drivers/dri/i915/server/intel_dri.c b/src/mesa/drivers/dri/i915/server/intel_dri.c index b6946b75d2..effdd26448 100644..120000 --- a/src/mesa/drivers/dri/i915/server/intel_dri.c +++ b/src/mesa/drivers/dri/i915/server/intel_dri.c @@ -1,1283 +1 @@ -/** - * \file server/intel_dri.c - * \brief File to perform the device-specific initialization tasks typically - * done in the X server. - * - * Here they are converted to run in the client (or perhaps a standalone - * process), and to work with the frame buffer device rather than the X - * server infrastructure. - * - * Copyright (C) 2006 Dave Airlie (airlied@linux.ie) - - 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 THE COPYRIGHT HOLDERS AND/OR THEIR 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 <stdlib.h> -#include <string.h> -#include <errno.h> -#include <unistd.h> - -#include "driver.h" -#include "drm.h" - -#include "intel.h" -#include "i830_dri.h" - -#include "memops.h" -#include "pciaccess.h" - -static size_t drm_page_size; -static int nextTile = 0; -#define xf86DrvMsg(...) do {} while(0) - -static const int pitches[] = { - 128 * 8, - 128 * 16, - 128 * 32, - 128 * 64, - 0 -}; - -static Bool I830DRIDoMappings(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea); - -static unsigned long -GetBestTileAlignment(unsigned long size) -{ - unsigned long i; - - for (i = KB(512); i < size; i <<= 1) - ; - - if (i > MB(64)) - i = MB(64); - - return i; -} - -static void SetFenceRegs(const DRIDriverContext *ctx, I830Rec *pI830) -{ - int i; - unsigned char *MMIO = ctx->MMIOAddress; - - for (i = 0; i < 8; i++) { - OUTREG(FENCE + i * 4, pI830->Fence[i]); - // if (I810_DEBUG & DEBUG_VERBOSE_VGA) - fprintf(stderr,"Fence Register : %x\n", pI830->Fence[i]); - } -} - -/* Tiled memory is good... really, really good... - * - * Need to make it less likely that we miss out on this - probably - * need to move the frontbuffer away from the 'guarenteed' alignment - * of the first memory segment, or perhaps allocate a discontigous - * framebuffer to get more alignment 'sweet spots'. - */ -static void -SetFence(const DRIDriverContext *ctx, I830Rec *pI830, - int nr, unsigned int start, unsigned int pitch, - unsigned int size) -{ - unsigned int val; - unsigned int fence_mask = 0; - unsigned int fence_pitch; - - if (nr < 0 || nr > 7) { - fprintf(stderr, - "SetFence: fence %d out of range\n",nr); - return; - } - - pI830->Fence[nr] = 0; - - if (IS_I9XX(pI830)) - fence_mask = ~I915G_FENCE_START_MASK; - else - fence_mask = ~I830_FENCE_START_MASK; - - if (start & fence_mask) { - fprintf(stderr, - "SetFence: %d: start (0x%08x) is not %s aligned\n", - nr, start, (IS_I9XX(pI830)) ? "1MB" : "512k"); - return; - } - - if (start % size) { - fprintf(stderr, - "SetFence: %d: start (0x%08x) is not size (%dk) aligned\n", - nr, start, size / 1024); - return; - } - - if (pitch & 127) { - fprintf(stderr, - "SetFence: %d: pitch (%d) not a multiple of 128 bytes\n", - nr, pitch); - return; - } - - val = (start | FENCE_X_MAJOR | FENCE_VALID); - - if (IS_I9XX(pI830)) { - switch (size) { - case MB(1): - val |= I915G_FENCE_SIZE_1M; - break; - case MB(2): - val |= I915G_FENCE_SIZE_2M; - break; - case MB(4): - val |= I915G_FENCE_SIZE_4M; - break; - case MB(8): - val |= I915G_FENCE_SIZE_8M; - break; - case MB(16): - val |= I915G_FENCE_SIZE_16M; - break; - case MB(32): - val |= I915G_FENCE_SIZE_32M; - break; - case MB(64): - val |= I915G_FENCE_SIZE_64M; - break; - default: - fprintf(stderr, - "SetFence: %d: illegal size (%d kByte)\n", nr, size / 1024); - return; - } - } else { - switch (size) { - case KB(512): - val |= FENCE_SIZE_512K; - break; - case MB(1): - val |= FENCE_SIZE_1M; - break; - case MB(2): - val |= FENCE_SIZE_2M; - break; - case MB(4): - val |= FENCE_SIZE_4M; - break; - case MB(8): - val |= FENCE_SIZE_8M; - break; - case MB(16): - val |= FENCE_SIZE_16M; - break; - case MB(32): - val |= FENCE_SIZE_32M; - break; - case MB(64): - val |= FENCE_SIZE_64M; - break; - default: - fprintf(stderr, - "SetFence: %d: illegal size (%d kByte)\n", nr, size / 1024); - return; - } - } - - if (IS_I9XX(pI830)) - fence_pitch = pitch / 512; - else - fence_pitch = pitch / 128; - - switch (fence_pitch) { - case 1: - val |= FENCE_PITCH_1; - break; - case 2: - val |= FENCE_PITCH_2; - break; - case 4: - val |= FENCE_PITCH_4; - break; - case 8: - val |= FENCE_PITCH_8; - break; - case 16: - val |= FENCE_PITCH_16; - break; - case 32: - val |= FENCE_PITCH_32; - break; - case 64: - val |= FENCE_PITCH_64; - break; - default: - fprintf(stderr, - "SetFence: %d: illegal pitch (%d)\n", nr, pitch); - return; - } - - pI830->Fence[nr] = val; -} - -static Bool -MakeTiles(const DRIDriverContext *ctx, I830Rec *pI830, I830MemRange *pMem) -{ - int pitch, ntiles, i; - - pitch = pMem->Pitch * ctx->cpp; - /* - * Simply try to break the region up into at most four pieces of size - * equal to the alignment. - */ - ntiles = ROUND_TO(pMem->Size, pMem->Alignment) / pMem->Alignment; - if (ntiles >= 4) { - return FALSE; - } - - for (i = 0; i < ntiles; i++, nextTile++) { - SetFence(ctx, pI830, nextTile, pMem->Start + i * pMem->Alignment, - pitch, pMem->Alignment); - } - return TRUE; -} - -static void I830SetupMemoryTiling(const DRIDriverContext *ctx, I830Rec *pI830) -{ - int i; - - /* Clear out */ - for (i = 0; i < 8; i++) - pI830->Fence[i] = 0; - - nextTile = 0; - - if (pI830->BackBuffer.Alignment >= KB(512)) { - if (MakeTiles(ctx, pI830, &(pI830->BackBuffer))) { - fprintf(stderr, - "Activating tiled memory for the back buffer.\n"); - } else { - fprintf(stderr, - "MakeTiles failed for the back buffer.\n"); - pI830->allowPageFlip = FALSE; - } - } - - if (pI830->DepthBuffer.Alignment >= KB(512)) { - if (MakeTiles(ctx, pI830, &(pI830->DepthBuffer))) { - fprintf(stderr, - "Activating tiled memory for the depth buffer.\n"); - } else { - fprintf(stderr, - "MakeTiles failed for the depth buffer.\n"); - } - } - - return; -} - -static int I830DetectMemory(const DRIDriverContext *ctx, I830Rec *pI830) -{ - struct pci_device host_bridge; - uint32_t gmch_ctrl; - int memsize = 0; - int range; - - memset(&host_bridge, 0, sizeof(host_bridge)); - - pci_device_cfg_read_u32(&host_bridge, &gmch_ctrl, I830_GMCH_CTRL); - - /* 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; - - if (IS_I85X(pI830) || IS_I865G(pI830) || IS_I9XX(pI830)) { - switch (gmch_ctrl & I830_GMCH_GMS_MASK) { - case I855_GMCH_GMS_STOLEN_1M: - memsize = MB(1) - KB(range); - break; - case I855_GMCH_GMS_STOLEN_4M: - memsize = MB(4) - KB(range); - break; - case I855_GMCH_GMS_STOLEN_8M: - memsize = MB(8) - KB(range); - break; - case I855_GMCH_GMS_STOLEN_16M: - memsize = MB(16) - KB(range); - break; - case I855_GMCH_GMS_STOLEN_32M: - memsize = MB(32) - KB(range); - break; - case I915G_GMCH_GMS_STOLEN_48M: - if (IS_I9XX(pI830)) - memsize = MB(48) - KB(range); - break; - case I915G_GMCH_GMS_STOLEN_64M: - if (IS_I9XX(pI830)) - memsize = MB(64) - KB(range); - break; - } - } else { - switch (gmch_ctrl & I830_GMCH_GMS_MASK) { - case I830_GMCH_GMS_STOLEN_512: - memsize = KB(512) - KB(range); - break; - case I830_GMCH_GMS_STOLEN_1024: - memsize = MB(1) - KB(range); - break; - case I830_GMCH_GMS_STOLEN_8192: - memsize = MB(8) - KB(range); - break; - case I830_GMCH_GMS_LOCAL: - memsize = 0; - xf86DrvMsg(pScrn->scrnIndex, X_WARNING, - "Local memory found, but won't be used.\n"); - break; - } - } - if (memsize > 0) { - fprintf(stderr, - "detected %d kB stolen memory.\n", memsize / 1024); - } else { - fprintf(stderr, - "no video memory detected.\n"); - } - return memsize; -} - -static int AgpInit(const DRIDriverContext *ctx, I830Rec *info) -{ - unsigned long mode = 0x4; - - if (drmAgpAcquire(ctx->drmFD) < 0) { - fprintf(stderr, "[gart] AGP not available\n"); - return 0; - } - - if (drmAgpEnable(ctx->drmFD, mode) < 0) { - fprintf(stderr, "[gart] AGP not enabled\n"); - drmAgpRelease(ctx->drmFD); - return 0; - } - else - fprintf(stderr, "[gart] AGP enabled at %dx\n", ctx->agpmode); - - return 1; -} - -/* - * Allocate memory from the given pool. Grow the pool if needed and if - * possible. - */ -static unsigned long -AllocFromPool(const DRIDriverContext *ctx, I830Rec *pI830, - I830MemRange *result, I830MemPool *pool, - long size, unsigned long alignment, int flags) -{ - long needed, start, end; - - if (!result || !pool || !size) - return 0; - - /* Calculate how much space is needed. */ - if (alignment <= GTT_PAGE_SIZE) - needed = size; - else { - start = ROUND_TO(pool->Free.Start, alignment); - end = ROUND_TO(start + size, alignment); - needed = end - pool->Free.Start; - } - if (needed > pool->Free.Size) { - return 0; - } - - result->Start = ROUND_TO(pool->Free.Start, alignment); - pool->Free.Start += needed; - result->End = pool->Free.Start; - - pool->Free.Size = pool->Free.End - pool->Free.Start; - result->Size = result->End - result->Start; - result->Pool = pool; - result->Alignment = alignment; - return needed; -} - -static unsigned long AllocFromAGP(const DRIDriverContext *ctx, I830Rec *pI830, long size, unsigned long alignment, I830MemRange *result) -{ - unsigned long start, end; - unsigned long newApStart, newApEnd; - int ret; - if (!result || !size) - return 0; - - if (!alignment) - alignment = 4; - - start = ROUND_TO(pI830->MemoryAperture.Start, alignment); - end = ROUND_TO(start + size, alignment); - newApStart = end; - newApEnd = pI830->MemoryAperture.End; - - ret=drmAgpAlloc(ctx->drmFD, size, 0, &(result->Physical), (drm_handle_t *)&(result->Key)); - - if (ret) - { - fprintf(stderr,"drmAgpAlloc failed %d\n", ret); - return 0; - } - pI830->allocatedMemory += size; - pI830->MemoryAperture.Start = newApStart; - pI830->MemoryAperture.End = newApEnd; - pI830->MemoryAperture.Size = newApEnd - newApStart; - // pI830->FreeMemory -= size; - result->Start = start; - result->End = start + size; - result->Size = size; - result->Offset = start; - result->Alignment = alignment; - result->Pool = NULL; - - return size; -} - -unsigned long -I830AllocVidMem(const DRIDriverContext *ctx, I830Rec *pI830, - I830MemRange *result, I830MemPool *pool, long size, - unsigned long alignment, int flags) -{ - unsigned long ret; - - if (!result) - return 0; - - /* Make sure these are initialised. */ - result->Size = 0; - result->Key = -1; - - if (!size) { - return 0; - } - - if (pool->Free.Size < size) { - ret = AllocFromAGP(ctx, pI830, size, alignment, result); - } - else { - ret = AllocFromPool(ctx, pI830, result, pool, size, alignment, flags); - if (ret == 0) - ret = AllocFromAGP(ctx, pI830, size, alignment, result); - } - return ret; -} - -static Bool BindAgpRange(const DRIDriverContext *ctx, I830MemRange *mem) -{ - if (!mem) - return FALSE; - - if (mem->Key == -1) - return TRUE; - - return !drmAgpBind(ctx->drmFD, mem->Key, mem->Offset); -} - -/* simple memory allocation routines needed */ -/* put ring buffer in low memory */ -/* need to allocate front, back, depth buffers aligned correctly, - allocate ring buffer, -*/ - -/* */ -static Bool -I830AllocateMemory(const DRIDriverContext *ctx, I830Rec *pI830) -{ - unsigned long size, ret; - unsigned long lines, lineSize, align; - - /* allocate ring buffer */ - memset(pI830->LpRing, 0, sizeof(I830RingBuffer)); - pI830->LpRing->mem.Key = -1; - - size = PRIMARY_RINGBUFFER_SIZE; - - ret = I830AllocVidMem(ctx, pI830, &pI830->LpRing->mem, &pI830->StolenPool, size, 0x1000, 0); - - if (ret != size) - { - fprintf(stderr,"unable to allocate ring buffer %ld\n", ret); - return FALSE; - } - - pI830->LpRing->tail_mask = pI830->LpRing->mem.Size - 1; - - - /* allocate front buffer */ - memset(&(pI830->FrontBuffer), 0, sizeof(pI830->FrontBuffer)); - pI830->FrontBuffer.Key = -1; - pI830->FrontBuffer.Pitch = ctx->shared.virtualWidth; - - align = KB(512); - - lineSize = ctx->shared.virtualWidth * ctx->cpp; - lines = (ctx->shared.virtualHeight + 15) / 16 * 16; - size = lineSize * lines; - size = ROUND_TO_PAGE(size); - - align = GetBestTileAlignment(size); - - ret = I830AllocVidMem(ctx, pI830, &pI830->FrontBuffer, &pI830->StolenPool, size, align, 0); - if (ret < size) - { - fprintf(stderr,"unable to allocate front buffer %ld\n", ret); - return FALSE; - } - - memset(&(pI830->BackBuffer), 0, sizeof(pI830->BackBuffer)); - pI830->BackBuffer.Key = -1; - pI830->BackBuffer.Pitch = ctx->shared.virtualWidth; - - ret = I830AllocVidMem(ctx, pI830, &pI830->BackBuffer, &pI830->StolenPool, size, align, 0); - if (ret < size) - { - fprintf(stderr,"unable to allocate back buffer %ld\n", ret); - return FALSE; - } - - memset(&(pI830->DepthBuffer), 0, sizeof(pI830->DepthBuffer)); - pI830->DepthBuffer.Key = -1; - pI830->DepthBuffer.Pitch = ctx->shared.virtualWidth; - - ret = I830AllocVidMem(ctx, pI830, &pI830->DepthBuffer, &pI830->StolenPool, size, align, 0); - if (ret < size) - { - fprintf(stderr,"unable to allocate depth buffer %ld\n", ret); - return FALSE; - } - - memset(&(pI830->ContextMem), 0, sizeof(pI830->ContextMem)); - pI830->ContextMem.Key = -1; - size = KB(32); - - ret = I830AllocVidMem(ctx, pI830, &pI830->ContextMem, &pI830->StolenPool, size, align, 0); - if (ret < size) - { - fprintf(stderr,"unable to allocate context buffer %ld\n", ret); - return FALSE; - } - - memset(&(pI830->TexMem), 0, sizeof(pI830->TexMem)); - pI830->TexMem.Key = -1; - - size = 32768 * 1024; - ret = AllocFromAGP(ctx, pI830, size, align, &pI830->TexMem); - if (ret < size) - { - fprintf(stderr,"unable to allocate texture memory %ld\n", ret); - return FALSE; - } - - return TRUE; -} - -static Bool -I830BindMemory(const DRIDriverContext *ctx, I830Rec *pI830) -{ - if (!BindAgpRange(ctx, &pI830->LpRing->mem)) - return FALSE; - if (!BindAgpRange(ctx, &pI830->FrontBuffer)) - return FALSE; - if (!BindAgpRange(ctx, &pI830->BackBuffer)) - return FALSE; - if (!BindAgpRange(ctx, &pI830->DepthBuffer)) - return FALSE; - if (!BindAgpRange(ctx, &pI830->ContextMem)) - return FALSE; - if (!BindAgpRange(ctx, &pI830->TexMem)) - return FALSE; - - return TRUE; -} - -static Bool -I830CleanupDma(const DRIDriverContext *ctx) -{ - drmI830Init info; - - memset(&info, 0, sizeof(drmI830Init)); - info.func = I830_CLEANUP_DMA; - - if (drmCommandWrite(ctx->drmFD, DRM_I830_INIT, - &info, sizeof(drmI830Init))) { - fprintf(stderr, "I830 Dma Cleanup Failed\n"); - return FALSE; - } - - return TRUE; -} - -static Bool -I830InitDma(const DRIDriverContext *ctx, I830Rec *pI830) -{ - I830RingBuffer *ring = pI830->LpRing; - drmI830Init info; - - memset(&info, 0, sizeof(drmI830Init)); - info.func = I830_INIT_DMA; - - info.ring_start = ring->mem.Start + pI830->LinearAddr; - info.ring_end = ring->mem.End + pI830->LinearAddr; - info.ring_size = ring->mem.Size; - - info.mmio_offset = (unsigned int)ctx->MMIOStart; - - info.sarea_priv_offset = sizeof(drm_sarea_t); - - info.front_offset = pI830->FrontBuffer.Start; - info.back_offset = pI830->BackBuffer.Start; - info.depth_offset = pI830->DepthBuffer.Start; - info.w = ctx->shared.virtualWidth; - info.h = ctx->shared.virtualHeight; - info.pitch = ctx->shared.virtualWidth; - info.back_pitch = pI830->BackBuffer.Pitch; - info.depth_pitch = pI830->DepthBuffer.Pitch; - info.cpp = ctx->cpp; - - if (drmCommandWrite(ctx->drmFD, DRM_I830_INIT, - &info, sizeof(drmI830Init))) { - fprintf(stderr, - "I830 Dma Initialization Failed\n"); - return FALSE; - } - - return TRUE; -} - -static int I830CheckDRMVersion( const DRIDriverContext *ctx, - I830Rec *pI830 ) -{ - drmVersionPtr version; - - version = drmGetVersion(ctx->drmFD); - - if (version) { - int req_minor, req_patch; - - req_minor = 4; - req_patch = 0; - - if (version->version_major != 1 || - version->version_minor < req_minor || - (version->version_minor == req_minor && - version->version_patchlevel < req_patch)) { - /* Incompatible drm version */ - fprintf(stderr, - "[dri] I830DRIScreenInit failed because of a version " - "mismatch.\n" - "[dri] i915.o kernel module version is %d.%d.%d " - "but version 1.%d.%d or newer is needed.\n" - "[dri] Disabling DRI.\n", - version->version_major, - version->version_minor, - version->version_patchlevel, - req_minor, - req_patch); - drmFreeVersion(version); - return 0; - } - - pI830->drmMinor = version->version_minor; - drmFreeVersion(version); - } - return 1; -} - -static void -I830SetRingRegs(const DRIDriverContext *ctx, I830Rec *pI830) -{ - unsigned int itemp; - unsigned char *MMIO = ctx->MMIOAddress; - - OUTREG(LP_RING + RING_LEN, 0); - OUTREG(LP_RING + RING_TAIL, 0); - OUTREG(LP_RING + RING_HEAD, 0); - - if ((long)(pI830->LpRing->mem.Start & I830_RING_START_MASK) != - pI830->LpRing->mem.Start) { - fprintf(stderr, - "I830SetRingRegs: Ring buffer start (%lx) violates its " - "mask (%x)\n", pI830->LpRing->mem.Start, I830_RING_START_MASK); - } - /* Don't care about the old value. Reserved bits must be zero anyway. */ - itemp = pI830->LpRing->mem.Start & I830_RING_START_MASK; - OUTREG(LP_RING + RING_START, itemp); - - if (((pI830->LpRing->mem.Size - 4096) & I830_RING_NR_PAGES) != - pI830->LpRing->mem.Size - 4096) { - fprintf(stderr, - "I830SetRingRegs: Ring buffer size - 4096 (%lx) violates its " - "mask (%x)\n", pI830->LpRing->mem.Size - 4096, - I830_RING_NR_PAGES); - } - /* Don't care about the old value. Reserved bits must be zero anyway. */ - itemp = (pI830->LpRing->mem.Size - 4096) & I830_RING_NR_PAGES; - itemp |= (RING_NO_REPORT | RING_VALID); - OUTREG(LP_RING + RING_LEN, itemp); - - pI830->LpRing->head = INREG(LP_RING + RING_HEAD) & I830_HEAD_MASK; - pI830->LpRing->tail = INREG(LP_RING + RING_TAIL); - pI830->LpRing->space = pI830->LpRing->head - (pI830->LpRing->tail + 8); - if (pI830->LpRing->space < 0) - pI830->LpRing->space += pI830->LpRing->mem.Size; - - SetFenceRegs(ctx, pI830); - - /* RESET THE DISPLAY PIPE TO POINT TO THE FRONTBUFFER - hacky - hacky hacky */ - OUTREG(DSPABASE, pI830->FrontBuffer.Start + pI830->LinearAddr); - -} - -static Bool -I830SetParam(const DRIDriverContext *ctx, int param, int value) -{ - drmI830SetParam sp; - - memset(&sp, 0, sizeof(sp)); - sp.param = param; - sp.value = value; - - if (drmCommandWrite(ctx->drmFD, DRM_I830_SETPARAM, &sp, sizeof(sp))) { - fprintf(stderr, "I830 SetParam Failed\n"); - return FALSE; - } - - return TRUE; -} - -static Bool -I830DRIMapScreenRegions(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) -{ - fprintf(stderr, - "[drm] Mapping front buffer\n"); - - if (drmAddMap(ctx->drmFD, - (drm_handle_t)(sarea->front_offset + pI830->LinearAddr), - sarea->front_size, - DRM_FRAME_BUFFER, /*DRM_AGP,*/ - 0, - &sarea->front_handle) < 0) { - fprintf(stderr, - "[drm] drmAddMap(front_handle) failed. Disabling DRI\n"); - return FALSE; - } - ctx->shared.hFrameBuffer = sarea->front_handle; - ctx->shared.fbSize = sarea->front_size; - fprintf(stderr, "[drm] Front Buffer = 0x%08x\n", - sarea->front_handle); - - if (drmAddMap(ctx->drmFD, - (drm_handle_t)(sarea->back_offset), - sarea->back_size, DRM_AGP, 0, - &sarea->back_handle) < 0) { - fprintf(stderr, - "[drm] drmAddMap(back_handle) failed. Disabling DRI\n"); - return FALSE; - } - fprintf(stderr, "[drm] Back Buffer = 0x%08x\n", - sarea->back_handle); - - if (drmAddMap(ctx->drmFD, - (drm_handle_t)sarea->depth_offset, - sarea->depth_size, DRM_AGP, 0, - &sarea->depth_handle) < 0) { - fprintf(stderr, - "[drm] drmAddMap(depth_handle) failed. Disabling DRI\n"); - return FALSE; - } - fprintf(stderr, "[drm] Depth Buffer = 0x%08x\n", - sarea->depth_handle); - - if (drmAddMap(ctx->drmFD, - (drm_handle_t)sarea->tex_offset, - sarea->tex_size, DRM_AGP, 0, - &sarea->tex_handle) < 0) { - fprintf(stderr, - "[drm] drmAddMap(tex_handle) failed. Disabling DRI\n"); - return FALSE; - } - fprintf(stderr, "[drm] textures = 0x%08x\n", - sarea->tex_handle); - - return TRUE; -} - - -static void -I830DRIUnmapScreenRegions(const DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) -{ -#if 1 - if (sarea->front_handle) { - drmRmMap(ctx->drmFD, sarea->front_handle); - sarea->front_handle = 0; - } -#endif - if (sarea->back_handle) { - drmRmMap(ctx->drmFD, sarea->back_handle); - sarea->back_handle = 0; - } - if (sarea->depth_handle) { - drmRmMap(ctx->drmFD, sarea->depth_handle); - sarea->depth_handle = 0; - } - if (sarea->tex_handle) { - drmRmMap(ctx->drmFD, sarea->tex_handle); - sarea->tex_handle = 0; - } -} - -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) -{ - if (drmAddMap(ctx->drmFD, - (drm_handle_t)pI830->LpRing->mem.Start, - pI830->LpRing->mem.Size, DRM_AGP, 0, - &pI830->ring_map) < 0) { - fprintf(stderr, - "[drm] drmAddMap(ring_map) failed. Disabling DRI\n"); - return FALSE; - } - fprintf(stderr, "[drm] ring buffer = 0x%08x\n", - pI830->ring_map); - - if (I830InitDma(ctx, pI830) == FALSE) { - return FALSE; - } - - /* init to zero to be safe */ - - I830DRIMapScreenRegions(ctx, pI830, sarea); - I830InitTextureHeap(ctx, pI830, sarea); - - if (ctx->pciDevice != PCI_CHIP_845_G && - ctx->pciDevice != PCI_CHIP_I830_M) { - I830SetParam(ctx, I830_SETPARAM_USE_MI_BATCHBUFFER_START, 1 ); - } - - /* Okay now initialize the dma engine */ - { - pI830->irq = drmGetInterruptFromBusID(ctx->drmFD, - ctx->pciBus, - ctx->pciDevice, - ctx->pciFunc); - - if (drmCtlInstHandler(ctx->drmFD, pI830->irq)) { - fprintf(stderr, - "[drm] failure adding irq handler\n"); - pI830->irq = 0; - return FALSE; - } - else - fprintf(stderr, - "[drm] dma control initialized, using IRQ %d\n", - pI830->irq); - } - - fprintf(stderr, "[dri] visual configs initialized\n"); - - return TRUE; -} - -static Bool -I830ClearScreen(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) -{ - /* need to drmMap front and back buffers and zero them */ - drmAddress map_addr; - int ret; - - ret = drmMap(ctx->drmFD, - sarea->front_handle, - sarea->front_size, - &map_addr); - - if (ret) - { - fprintf(stderr, "Unable to map front buffer\n"); - return FALSE; - } - - drimemsetio((char *)map_addr, - 0, - sarea->front_size); - drmUnmap(map_addr, sarea->front_size); - - - ret = drmMap(ctx->drmFD, - sarea->back_handle, - sarea->back_size, - &map_addr); - - if (ret) - { - fprintf(stderr, "Unable to map back buffer\n"); - return FALSE; - } - - drimemsetio((char *)map_addr, - 0, - sarea->back_size); - drmUnmap(map_addr, sarea->back_size); - - return TRUE; -} - -static Bool -I830ScreenInit(DRIDriverContext *ctx, I830Rec *pI830) - -{ - I830DRIPtr pI830DRI; - drmI830Sarea *pSAREAPriv; - int err; - - drm_page_size = getpagesize(); - - pI830->registerSize = ctx->MMIOSize; - /* This is a hack for now. We have to have more than a 4k page here - * because of the size of the state. However, the state should be - * in a per-context mapping. This will be added in the Mesa 3.5 port - * of the I830 driver. - */ - ctx->shared.SAREASize = SAREA_MAX; - - /* Note that drmOpen will try to load the kernel module, if needed. */ - ctx->drmFD = drmOpen("i915", NULL ); - if (ctx->drmFD < 0) { - fprintf(stderr, "[drm] drmOpen failed\n"); - return 0; - } - - if ((err = drmSetBusid(ctx->drmFD, ctx->pciBusID)) < 0) { - fprintf(stderr, "[drm] drmSetBusid failed (%d, %s), %s\n", - ctx->drmFD, ctx->pciBusID, strerror(-err)); - return 0; - } - - if (drmAddMap( ctx->drmFD, - 0, - ctx->shared.SAREASize, - DRM_SHM, - DRM_CONTAINS_LOCK, - &ctx->shared.hSAREA) < 0) - { - fprintf(stderr, "[drm] drmAddMap failed\n"); - return 0; - } - - fprintf(stderr, "[drm] added %d byte SAREA at 0x%08x\n", - ctx->shared.SAREASize, ctx->shared.hSAREA); - - if (drmMap( ctx->drmFD, - ctx->shared.hSAREA, - ctx->shared.SAREASize, - (drmAddressPtr)(&ctx->pSAREA)) < 0) - { - fprintf(stderr, "[drm] drmMap failed\n"); - return 0; - - } - - memset(ctx->pSAREA, 0, ctx->shared.SAREASize); - fprintf(stderr, "[drm] mapped SAREA 0x%08x to %p, size %d\n", - ctx->shared.hSAREA, ctx->pSAREA, ctx->shared.SAREASize); - - - if (drmAddMap(ctx->drmFD, - ctx->MMIOStart, - ctx->MMIOSize, - DRM_REGISTERS, - DRM_READ_ONLY, - &pI830->registerHandle) < 0) { - fprintf(stderr, "[drm] drmAddMap mmio failed\n"); - return 0; - } - fprintf(stderr, - "[drm] register handle = 0x%08x\n", pI830->registerHandle); - - - if (!I830CheckDRMVersion(ctx, pI830)) { - return FALSE; - } - - /* Create a 'server' context so we can grab the lock for - * initialization ioctls. - */ - if ((err = drmCreateContext(ctx->drmFD, &ctx->serverContext)) != 0) { - fprintf(stderr, "%s: drmCreateContext failed %d\n", __FUNCTION__, err); - return 0; - } - - DRM_LOCK(ctx->drmFD, ctx->pSAREA, ctx->serverContext, 0); - - /* Initialize the SAREA private data structure */ - pSAREAPriv = (drmI830Sarea *)(((char*)ctx->pSAREA) + - sizeof(drm_sarea_t)); - memset(pSAREAPriv, 0, sizeof(*pSAREAPriv)); - - pI830->StolenMemory.Size = I830DetectMemory(ctx, pI830); - pI830->StolenMemory.Start = 0; - pI830->StolenMemory.End = pI830->StolenMemory.Size; - - pI830->MemoryAperture.Start = pI830->StolenMemory.End; - pI830->MemoryAperture.End = KB(40000); - pI830->MemoryAperture.Size = pI830->MemoryAperture.End - pI830->MemoryAperture.Start; - - pI830->StolenPool.Fixed = pI830->StolenMemory; - pI830->StolenPool.Total = pI830->StolenMemory; - pI830->StolenPool.Free = pI830->StolenPool.Total; - pI830->FreeMemory = pI830->StolenPool.Total.Size; - - if (!AgpInit(ctx, pI830)) - return FALSE; - - if (I830AllocateMemory(ctx, pI830) == FALSE) - { - return FALSE; - } - - if (I830BindMemory(ctx, pI830) == FALSE) - { - return FALSE; - } - - pSAREAPriv->front_offset = pI830->FrontBuffer.Start; - pSAREAPriv->front_size = pI830->FrontBuffer.Size; - pSAREAPriv->width = ctx->shared.virtualWidth; - pSAREAPriv->height = ctx->shared.virtualHeight; - pSAREAPriv->pitch = ctx->shared.virtualWidth; - pSAREAPriv->virtualX = ctx->shared.virtualWidth; - pSAREAPriv->virtualY = ctx->shared.virtualHeight; - pSAREAPriv->back_offset = pI830->BackBuffer.Start; - pSAREAPriv->back_size = pI830->BackBuffer.Size; - pSAREAPriv->depth_offset = pI830->DepthBuffer.Start; - pSAREAPriv->depth_size = pI830->DepthBuffer.Size; - pSAREAPriv->tex_offset = pI830->TexMem.Start; - pSAREAPriv->tex_size = pI830->TexMem.Size; - pSAREAPriv->log_tex_granularity = pI830->TexGranularity; - - ctx->driverClientMsg = malloc(sizeof(I830DRIRec)); - ctx->driverClientMsgSize = sizeof(I830DRIRec); - pI830DRI = (I830DRIPtr)ctx->driverClientMsg; - pI830DRI->deviceID = pI830->Chipset; - pI830DRI->regsSize = I830_REG_SIZE; - pI830DRI->width = ctx->shared.virtualWidth; - 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); - - err = I830DRIDoMappings(ctx, pI830, pSAREAPriv); - if (err == FALSE) - return FALSE; - - I830SetupMemoryTiling(ctx, pI830); - - /* Quick hack to clear the front & back buffers. Could also use - * the clear ioctl to do this, but would need to setup hw state - * first. - */ - I830ClearScreen(ctx, pI830, pSAREAPriv); - - I830SetRingRegs(ctx, pI830); - - return TRUE; -} - - -/** - * \brief Validate the fbdev mode. - * - * \param ctx display handle. - * - * \return one on success, or zero on failure. - * - * Saves some registers and returns 1. - * - * \sa radeonValidateMode(). - */ -static int i830ValidateMode( const DRIDriverContext *ctx ) -{ - return 1; -} - -/** - * \brief Examine mode returned by fbdev. - * - * \param ctx display handle. - * - * \return one on success, or zero on failure. - * - * Restores registers that fbdev has clobbered and returns 1. - * - * \sa i810ValidateMode(). - */ -static int i830PostValidateMode( const DRIDriverContext *ctx ) -{ - I830Rec *pI830 = ctx->driverPrivate; - - I830SetRingRegs(ctx, pI830); - return 1; -} - - -/** - * \brief Initialize the framebuffer device mode - * - * \param ctx display handle. - * - * \return one on success, or zero on failure. - * - * Fills in \p info with some default values and some information from \p ctx - * and then calls I810ScreenInit() for the screen initialization. - * - * Before exiting clears the framebuffer memory accessing it directly. - */ -static int i830InitFBDev( DRIDriverContext *ctx ) -{ - I830Rec *pI830 = calloc(1, sizeof(I830Rec)); - int i; - - { - int dummy = ctx->shared.virtualWidth; - - switch (ctx->bpp / 8) { - case 1: dummy = (ctx->shared.virtualWidth + 127) & ~127; break; - case 2: dummy = (ctx->shared.virtualWidth + 31) & ~31; break; - case 3: - case 4: dummy = (ctx->shared.virtualWidth + 15) & ~15; break; - } - - ctx->shared.virtualWidth = dummy; - ctx->shared.Width = ctx->shared.virtualWidth; - } - - - for (i = 0; pitches[i] != 0; i++) { - if (pitches[i] >= ctx->shared.virtualWidth) { - ctx->shared.virtualWidth = pitches[i]; - break; - } - } - - ctx->driverPrivate = (void *)pI830; - - pI830->LpRing = calloc(1, sizeof(I830RingBuffer)); - pI830->Chipset = ctx->chipset; - pI830->LinearAddr = ctx->FBStart; - - if (!I830ScreenInit( ctx, pI830 )) - return 0; - - - return 1; -} - - -/** - * \brief The screen is being closed, so clean up any state and free any - * resources used by the DRI. - * - * \param ctx display handle. - * - * Unmaps the SAREA, closes the DRM device file descriptor and frees the driver - * private data. - */ -static void i830HaltFBDev( DRIDriverContext *ctx ) -{ - drmI830Sarea *pSAREAPriv; - I830Rec *pI830 = ctx->driverPrivate; - - if (pI830->irq) { - drmCtlUninstHandler(ctx->drmFD); - pI830->irq = 0; } - - I830CleanupDma(ctx); - - pSAREAPriv = (drmI830Sarea *)(((char*)ctx->pSAREA) + - sizeof(drm_sarea_t)); - - I830DRIUnmapScreenRegions(ctx, pI830, pSAREAPriv); - drmUnmap( ctx->pSAREA, ctx->shared.SAREASize ); - drmClose(ctx->drmFD); - - if (ctx->driverPrivate) { - free(ctx->driverPrivate); - ctx->driverPrivate = 0; - } -} - - -extern void i810NotifyFocus( int ); - -/** - * \brief Exported driver interface for Mini GLX. - * - * \sa DRIDriverRec. - */ -const struct DRIDriverRec __driDriver = { - i830ValidateMode, - i830PostValidateMode, - i830InitFBDev, - i830HaltFBDev, - NULL,//I830EngineShutdown, - NULL, //I830EngineRestore, -#ifndef _EMBEDDED - 0, -#else - i810NotifyFocus, -#endif -}; +../../intel/server/intel_dri.c
\ No newline at end of file |