diff options
Diffstat (limited to 'src/mesa/drivers')
116 files changed, 3999 insertions, 1570 deletions
diff --git a/src/mesa/drivers/common/driverfuncs.c b/src/mesa/drivers/common/driverfuncs.c index e5aab02a42..4ca0e7bcc3 100644 --- a/src/mesa/drivers/common/driverfuncs.c +++ b/src/mesa/drivers/common/driverfuncs.c @@ -183,19 +183,6 @@ _mesa_init_driver_functions(struct dd_function_table *driver) driver->TexParameter = NULL; driver->Viewport = NULL; - /* vertex arrays */ - driver->VertexPointer = NULL; - driver->NormalPointer = NULL; - driver->ColorPointer = NULL; - driver->FogCoordPointer = NULL; - driver->IndexPointer = NULL; - driver->SecondaryColorPointer = NULL; - driver->TexCoordPointer = NULL; - driver->EdgeFlagPointer = NULL; - driver->VertexAttribPointer = NULL; - driver->LockArraysEXT = NULL; - driver->UnlockArraysEXT = NULL; - /* state queries */ driver->GetBooleanv = NULL; driver->GetDoublev = NULL; diff --git a/src/mesa/drivers/dri/Makefile.template b/src/mesa/drivers/dri/Makefile.template index 1ce9315530..39d25ce3f4 100644 --- a/src/mesa/drivers/dri/Makefile.template +++ b/src/mesa/drivers/dri/Makefile.template @@ -60,9 +60,13 @@ SHARED_INCLUDES = \ ##### TARGETS ##### -default: symlinks subdirs depend $(LIBNAME) $(TOP)/$(LIB_DIR)/$(LIBNAME) +default: subdirs lib +.PHONY: lib +lib: symlinks subdirs depend + @$(MAKE) $(LIBNAME) $(TOP)/$(LIB_DIR)/$(LIBNAME) + $(LIBNAME): $(OBJECTS) $(MESA_MODULES) $(EXTRA_MODULES) $(WINOBJ) Makefile \ $(TOP)/src/mesa/drivers/dri/Makefile.template $(MKLIB) -o $@ -noprefix -linker '$(CC)' -ldflags '$(LDFLAGS)' \ diff --git a/src/mesa/drivers/dri/common/spantmp2.h b/src/mesa/drivers/dri/common/spantmp2.h index 89c815722f..95f97414a9 100644 --- a/src/mesa/drivers/dri/common/spantmp2.h +++ b/src/mesa/drivers/dri/common/spantmp2.h @@ -82,6 +82,72 @@ rgba[3] = 0xff; \ } while (0) +#elif (SPANTMP_PIXEL_FMT == GL_RGB) && (SPANTMP_PIXEL_TYPE == GL_UNSIGNED_SHORT_5_6_5_REV) + +/** + ** GL_RGB, GL_UNSIGNED_SHORT_5_6_5_REV + **/ + +#ifndef GET_VALUE +#ifndef GET_PTR +#define GET_PTR(_x, _y) (buf + (_x) * 2 + (_y) * pitch) +#endif + +#define GET_VALUE(_x, _y) *(volatile GLushort *)(GET_PTR(_x, _y)) +#define PUT_VALUE(_x, _y, _v) *(volatile GLushort *)(GET_PTR(_x, _y)) = (_v) +#endif /* GET_VALUE */ + +#define INIT_MONO_PIXEL(p, color) \ + p = PACK_COLOR_565_REV( color[0], color[1], color[2] ) + +#define WRITE_RGBA( _x, _y, r, g, b, a ) \ + PUT_VALUE(_x, _y, PACK_COLOR_565_REV( r, g, b )) + +#define WRITE_PIXEL( _x, _y, p ) PUT_VALUE(_x, _y, p) + +#define READ_RGBA( rgba, _x, _y ) \ + do { \ + GLushort p = GET_VALUE(_x, _y); \ + p = p << 8 | p >> 8; \ + rgba[0] = ((p >> 8) & 0xf8) * 255 / 0xf8; \ + rgba[1] = ((p >> 3) & 0xfc) * 255 / 0xfc; \ + rgba[2] = ((p << 3) & 0xf8) * 255 / 0xf8; \ + rgba[3] = 0xff; \ + } while (0) + +#elif (SPANTMP_PIXEL_FMT == GL_BGRA) && (SPANTMP_PIXEL_TYPE == GL_UNSIGNED_SHORT_4_4_4_4) + +/** + ** GL_BGRA, GL_UNSIGNED_SHORT_4_4_4_4 + **/ + +#ifndef GET_VALUE +#ifndef GET_PTR +#define GET_PTR(_x, _y) (buf + (_x) * 2 + (_y) * pitch) +#endif + +#define GET_VALUE(_x, _y) *(volatile GLushort *)(GET_PTR(_x, _y)) +#define PUT_VALUE(_x, _y, _v) *(volatile GLushort *)(GET_PTR(_x, _y)) = (_v) +#endif /* GET_VALUE */ + +#define INIT_MONO_PIXEL(p, color) \ + p = PACK_COLOR_4444_REV(color[3], color[0], color[1], color[2]) + +#define WRITE_RGBA( _x, _y, r, g, b, a ) \ + PUT_VALUE(_x, _y, PACK_COLOR_4444_REV(a, r, g, b)) \ + +#define WRITE_PIXEL( _x, _y, p ) PUT_VALUE(_x, _y, p) + +#define READ_RGBA( rgba, _x, _y ) \ + do { \ + GLushort p = GET_VALUE(_x, _y); \ + rgba[0] = ((p >> 0) & 0xf) * 0x11; \ + rgba[1] = ((p >> 12) & 0xf) * 0x11; \ + rgba[2] = ((p >> 4) & 0xf) * 0x11; \ + rgba[3] = ((p >> 8) & 0xf) * 0x11; \ + } while (0) + + #elif (SPANTMP_PIXEL_FMT == GL_BGRA) && (SPANTMP_PIXEL_TYPE == GL_UNSIGNED_SHORT_4_4_4_4_REV) /** @@ -147,6 +213,39 @@ rgba[3] = ((p >> 15) & 0x1) * 0xff; \ } while (0) +#elif (SPANTMP_PIXEL_FMT == GL_BGRA) && (SPANTMP_PIXEL_TYPE == GL_UNSIGNED_SHORT_1_5_5_5) + +/** + ** GL_BGRA, GL_UNSIGNED_SHORT_1_5_5_5 + **/ + +#ifndef GET_VALUE +#ifndef GET_PTR +#define GET_PTR(_x, _y) (buf + (_x) * 2 + (_y) * pitch) +#endif + +#define GET_VALUE(_x, _y) *(volatile GLushort *)(GET_PTR(_x, _y)) +#define PUT_VALUE(_x, _y, _v) *(volatile GLushort *)(GET_PTR(_x, _y)) = (_v) +#endif /* GET_VALUE */ + +#define INIT_MONO_PIXEL(p, color) \ + p = PACK_COLOR_1555_REV(color[3], color[0], color[1], color[2]) + +#define WRITE_RGBA( _x, _y, r, g, b, a ) \ + PUT_VALUE(_x, _y, PACK_COLOR_1555_REV(a, r, g, b)) \ + +#define WRITE_PIXEL( _x, _y, p ) PUT_VALUE(_x, _y, p) + +#define READ_RGBA( rgba, _x, _y ) \ + do { \ + GLushort p = GET_VALUE(_x, _y); \ + p = p << 8 | p >> 8; \ + rgba[0] = ((p >> 7) & 0xf8) * 255 / 0xf8; \ + rgba[1] = ((p >> 2) & 0xf8) * 255 / 0xf8; \ + rgba[2] = ((p << 3) & 0xf8) * 255 / 0xf8; \ + rgba[3] = ((p >> 15) & 0x1) * 0xff; \ + } while (0) + #elif (SPANTMP_PIXEL_FMT == GL_BGRA) && (SPANTMP_PIXEL_TYPE == GL_UNSIGNED_INT_8_8_8_8_REV) /** @@ -202,6 +301,61 @@ } while (0) # endif +#elif (SPANTMP_PIXEL_FMT == GL_BGRA) && (SPANTMP_PIXEL_TYPE == GL_UNSIGNED_INT_8_8_8_8) + +/** + ** GL_BGRA, GL_UNSIGNED_INT_8_8_8_8 + **/ + +#ifndef GET_VALUE +#ifndef GET_PTR +#define GET_PTR(_x, _y) ( buf + (_x) * 4 + (_y) * pitch) +#endif + +#define GET_VALUE(_x, _y) *(volatile GLuint *)(GET_PTR(_x, _y)) +#define PUT_VALUE(_x, _y, _v) *(volatile GLuint *)(GET_PTR(_x, _y)) = (_v) +#endif /* GET_VALUE */ + +# define INIT_MONO_PIXEL(p, color) \ + p = PACK_COLOR_8888(color[2], color[1], color[0], color[3]) + +# define WRITE_RGBA(_x, _y, r, g, b, a) \ + PUT_VALUE(_x, _y, ((r << 8) | \ + (g << 16) | \ + (b << 24) | \ + (a << 0))) + +#define WRITE_PIXEL(_x, _y, p) PUT_VALUE(_x, _y, p) + +# if defined( USE_X86_ASM ) +# define READ_RGBA(rgba, _x, _y) \ + do { \ + GLuint p = GET_VALUE(_x, _y); \ + __asm__ __volatile__( "rorl $8, %0" \ + : "=r" (p) : "0" (p) ); \ + ((GLuint *)rgba)[0] = p; \ + } while (0) +# elif defined( MESA_BIG_ENDIAN ) + /* On PowerPC with GCC 3.4.2 the shift madness below becomes a single + * rotlwi instruction. It also produces good code on SPARC. + */ +# define READ_RGBA( rgba, _x, _y ) \ + do { \ + GLuint p = CPU_TO_LE32(GET_VALUE(_x, _y)); \ + GLuint t = p; \ + *((uint32_t *) rgba) = (t >> 24) | (p << 8); \ + } while (0) +# else +# define READ_RGBA( rgba, _x, _y ) \ + do { \ + GLuint p = GET_VALUE(_x, _y); \ + rgba[0] = (p >> 8) & 0xff; \ + rgba[1] = (p >> 16) & 0xff; \ + rgba[2] = (p >> 24) & 0xff; \ + rgba[3] = (p >> 0) & 0xff; \ + } while (0) +# endif + #else #error SPANTMP_PIXEL_FMT must be set to a valid value! #endif diff --git a/src/mesa/drivers/dri/ffb/ffb_vbtmp.h b/src/mesa/drivers/dri/ffb/ffb_vbtmp.h index 0495d0e276..c548ef3ad5 100644 --- a/src/mesa/drivers/dri/ffb/ffb_vbtmp.h +++ b/src/mesa/drivers/dri/ffb/ffb_vbtmp.h @@ -38,11 +38,11 @@ static void TAG(emit)(GLcontext *ctx, GLuint start, GLuint end) #endif #if (IND & (FFB_VB_RGBA_BIT)) - col0 = VB->ColorPtr[0]->data; - col0_stride = VB->ColorPtr[0]->stride; + col0 = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col0_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; #if (IND & (FFB_VB_TWOSIDE_BIT)) - col1 = VB->ColorPtr[1]->data; - col1_stride = VB->ColorPtr[1]->stride; + col1 = VB->BackfaceColorPtr->data; + col1_stride = VB->BackfaceColorPtr->stride; #endif #endif diff --git a/src/mesa/drivers/dri/gamma/gamma_render.c b/src/mesa/drivers/dri/gamma/gamma_render.c index 1b9fd169f4..a03a93d132 100644 --- a/src/mesa/drivers/dri/gamma/gamma_render.c +++ b/src/mesa/drivers/dri/gamma/gamma_render.c @@ -53,13 +53,13 @@ static void gamma_emit( GLcontext *ctx, GLuint start, GLuint end) GLfloat (*tc0)[4] = 0; GLuint tc0_size = 0; - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; if (ctx->Texture.Unit[0]._ReallyEnabled) { - tc0_stride = VB->TexCoordPtr[0]->stride; - tc0 = VB->TexCoordPtr[0]->data; - tc0_size = VB->TexCoordPtr[0]->size; + tc0_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0]->stride; + tc0 = VB->AttribPtr[_TNL_ATTRIB_TEX0]->data; + tc0_size = VB->AttribPtr[_TNL_ATTRIB_TEX0]->size; coord = VB->ClipPtr->data; coord_stride = VB->ClipPtr->stride; } else { diff --git a/src/mesa/drivers/dri/i915/i830_texstate.c b/src/mesa/drivers/dri/i915/i830_texstate.c index f4bbb53b86..c62281d341 100644 --- a/src/mesa/drivers/dri/i915/i830_texstate.c +++ b/src/mesa/drivers/dri/i915/i830_texstate.c @@ -27,6 +27,7 @@ #include "main/mtypes.h" #include "main/enums.h" +#include "main/colormac.h" #include "intel_mipmap_tree.h" #include "intel_tex.h" @@ -311,11 +312,10 @@ i830_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3) CLAMPED_FLOAT_TO_UBYTE(border[2], tObj->BorderColor[2]); CLAMPED_FLOAT_TO_UBYTE(border[3], tObj->BorderColor[3]); - state[I830_TEXREG_TM0S4] = INTEL_PACKCOLOR8888(border[0], - border[1], - border[2], - border[3]); - + state[I830_TEXREG_TM0S4] = PACK_COLOR_8888(border[3], + border[0], + border[1], + border[2]); I830_ACTIVESTATE(i830, I830_UPLOAD_TEX(unit), GL_TRUE); /* memcmp was already disabled, but definitely won't work as the diff --git a/src/mesa/drivers/dri/i915/i830_vtbl.c b/src/mesa/drivers/dri/i915/i830_vtbl.c index a6f554701e..c05c7759ac 100644 --- a/src/mesa/drivers/dri/i915/i830_vtbl.c +++ b/src/mesa/drivers/dri/i915/i830_vtbl.c @@ -126,7 +126,7 @@ i830_render_start(struct intel_context *intel) for (i = 0; i < I830_TEX_UNITS; i++) { if (RENDERINPUTS_TEST(index_bitset, _TNL_ATTRIB_TEX(i))) { - GLuint sz = VB->TexCoordPtr[i]->size; + GLuint sz = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->size; GLuint emit; GLuint mcs = (i830->state.Tex[i][I830_TEXREG_MCS] & ~TEXCOORDTYPE_MASK); @@ -714,9 +714,6 @@ i830_new_batch(struct intel_context *intel) { 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 diff --git a/src/mesa/drivers/dri/i915/i915_fragprog.c b/src/mesa/drivers/dri/i915/i915_fragprog.c index d9c61446f5..9e4d318036 100644 --- a/src/mesa/drivers/dri/i915/i915_fragprog.c +++ b/src/mesa/drivers/dri/i915/i915_fragprog.c @@ -1301,7 +1301,7 @@ i915ValidateFragmentProgram(struct i915_context *i915) for (i = 0; i < p->ctx->Const.MaxTextureCoordUnits; i++) { if (inputsRead & FRAG_BIT_TEX(i)) { - int sz = VB->TexCoordPtr[i]->size; + int sz = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->size; s2 &= ~S2_TEXCOORD_FMT(i, S2_TEXCOORD_FMT0_MASK); s2 |= S2_TEXCOORD_FMT(i, SZ_TO_HW(sz)); diff --git a/src/mesa/drivers/dri/i915/i915_texstate.c b/src/mesa/drivers/dri/i915/i915_texstate.c index d6689af53f..1bacd51aec 100644 --- a/src/mesa/drivers/dri/i915/i915_texstate.c +++ b/src/mesa/drivers/dri/i915/i915_texstate.c @@ -28,6 +28,7 @@ #include "main/mtypes.h" #include "main/enums.h" #include "main/macros.h" +#include "main/colormac.h" #include "intel_mipmap_tree.h" #include "intel_tex.h" @@ -363,15 +364,15 @@ i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3) * R channel, while the hardware uses A. Spam R into all the channels * for safety. */ - state[I915_TEXREG_SS4] = INTEL_PACKCOLOR8888(border[0], - border[0], - border[0], - border[0]); + state[I915_TEXREG_SS4] = PACK_COLOR_8888(border[0], + border[0], + border[0], + border[0]); } else { - state[I915_TEXREG_SS4] = INTEL_PACKCOLOR8888(border[0], - border[1], - border[2], - border[3]); + state[I915_TEXREG_SS4] = PACK_COLOR_8888(border[3], + border[0], + border[1], + border[2]); } diff --git a/src/mesa/drivers/dri/i915/i915_vtbl.c b/src/mesa/drivers/dri/i915/i915_vtbl.c index 77ba8d5581..3e7b5101cc 100644 --- a/src/mesa/drivers/dri/i915/i915_vtbl.c +++ b/src/mesa/drivers/dri/i915/i915_vtbl.c @@ -667,9 +667,6 @@ i915_new_batch(struct intel_context *intel) * difficulties associated with them (physical address requirements). */ i915->state.emitted = 0; - - /* Check that we didn't just wrap our batchbuffer at a bad time. */ - assert(!intel->no_batch_wrap); } static void diff --git a/src/mesa/drivers/dri/i915/intel_tris.c b/src/mesa/drivers/dri/i915/intel_tris.c index bc527aae47..8a3ab39bc2 100644 --- a/src/mesa/drivers/dri/i915/intel_tris.c +++ b/src/mesa/drivers/dri/i915/intel_tris.c @@ -1250,81 +1250,6 @@ union fi 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; - unsigned int saved_vertex_size = intel->vertex_size; - - 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; - - 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); -} - - - /**********************************************************************/ /* Initialization. */ /**********************************************************************/ @@ -1333,7 +1258,6 @@ intel_meta_draw_quad(struct intel_context *intel, void intelInitTriFuncs(GLcontext * ctx) { - struct intel_context *intel = intel_context(ctx); TNLcontext *tnl = TNL_CONTEXT(ctx); static int firsttime = 1; @@ -1350,6 +1274,4 @@ 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/i965/brw_clip.c b/src/mesa/drivers/dri/i965/brw_clip.c index f45dcf8282..dbd10a5297 100644 --- a/src/mesa/drivers/dri/i965/brw_clip.c +++ b/src/mesa/drivers/dri/i965/brw_clip.c @@ -78,7 +78,7 @@ static void compile_clip_prog( struct brw_context *brw, delta = REG_SIZE; for (i = 0; i < VERT_RESULT_MAX; i++) - if (c.key.attrs & (1<<i)) { + if (c.key.attrs & BITFIELD64_BIT(i)) { c.offset[i] = delta; delta += ATTR_SIZE; } diff --git a/src/mesa/drivers/dri/i965/brw_clip.h b/src/mesa/drivers/dri/i965/brw_clip.h index dc550ac793..1c6825510a 100644 --- a/src/mesa/drivers/dri/i965/brw_clip.h +++ b/src/mesa/drivers/dri/i965/brw_clip.h @@ -42,7 +42,7 @@ * up polygon offset and flatshading at this point: */ struct brw_clip_prog_key { - GLuint attrs:32; + GLbitfield64 attrs; GLuint primitive:4; GLuint nr_userclip:3; GLuint do_flat_shading:1; diff --git a/src/mesa/drivers/dri/i965/brw_context.h b/src/mesa/drivers/dri/i965/brw_context.h index 87e8a6aad4..e73e21433c 100644 --- a/src/mesa/drivers/dri/i965/brw_context.h +++ b/src/mesa/drivers/dri/i965/brw_context.h @@ -172,8 +172,8 @@ struct brw_fragment_program { GLuint id; /**< serial no. to identify frag progs, never re-used */ GLboolean isGLSL; /**< really, any IF/LOOP/CONT/BREAK instructions */ - dri_bo *const_buffer; /** Program constant buffer/surface */ GLboolean use_const_buffer; + dri_bo *const_buffer; /** Program constant buffer/surface */ /** for debugging, which texture units are referenced */ GLbitfield tex_units_used; @@ -231,7 +231,7 @@ struct brw_vs_prog_data { GLuint curb_read_length; GLuint urb_read_length; GLuint total_grf; - GLuint outputs_written; + GLbitfield64 outputs_written; GLuint nr_params; /**< number of float params/constants */ GLuint inputs_read; @@ -438,7 +438,6 @@ struct brw_context GLuint primitive; GLboolean emit_state_always; - GLboolean no_batch_wrap; struct { struct brw_state_flags dirty; @@ -742,9 +741,5 @@ brw_fragment_program_const(const struct gl_fragment_program *p) return (const struct brw_fragment_program *) p; } - - -#define DO_SETUP_BITS ((1<<(FRAG_ATTRIB_MAX)) - 1) - #endif diff --git a/src/mesa/drivers/dri/i965/brw_draw.c b/src/mesa/drivers/dri/i965/brw_draw.c index 8bcb6083f7..7ad860898f 100644 --- a/src/mesa/drivers/dri/i965/brw_draw.c +++ b/src/mesa/drivers/dri/i965/brw_draw.c @@ -145,7 +145,7 @@ static void brw_emit_prim(struct brw_context *brw, prim_packet.base_vert_location = prim->basevertex; /* Can't wrap here, since we rely on the validated state. */ - brw->no_batch_wrap = GL_TRUE; + intel->no_batch_wrap = GL_TRUE; /* If we're set to always flush, do it before and after the primitive emit. * We want to catch both missed flushes that hurt instruction/state cache @@ -163,7 +163,7 @@ static void brw_emit_prim(struct brw_context *brw, intel_batchbuffer_emit_mi_flush(intel->batch); } - brw->no_batch_wrap = GL_FALSE; + intel->no_batch_wrap = GL_FALSE; } static void brw_merge_inputs( struct brw_context *brw, diff --git a/src/mesa/drivers/dri/i965/brw_draw_upload.c b/src/mesa/drivers/dri/i965/brw_draw_upload.c index 271a88dae0..7c796dae93 100644 --- a/src/mesa/drivers/dri/i965/brw_draw_upload.c +++ b/src/mesa/drivers/dri/i965/brw_draw_upload.c @@ -243,14 +243,6 @@ static void wrap_buffers( struct brw_context *brw, dri_bo_unreference(brw->vb.upload.bo); brw->vb.upload.bo = dri_bo_alloc(brw->intel.bufmgr, "temporary VBO", size, 1); - - /* Set the internal VBO\ to no-backing-store. We only use them as a - * temporary within a brw_try_draw_prims while the lock is held. - */ - /* DON'T DO THIS AS IF WE HAVE TO RE-ORG MEMORY WE NEED SOMEWHERE WITH - FAKE TO PUSH THIS STUFF */ -// if (!brw->intel.ttm) -// dri_bo_fake_disable_backing_store(brw->vb.upload.bo, NULL, NULL); } static void get_space( struct brw_context *brw, diff --git a/src/mesa/drivers/dri/i965/brw_gs.h b/src/mesa/drivers/dri/i965/brw_gs.h index e0cf07256b..010c1c2352 100644 --- a/src/mesa/drivers/dri/i965/brw_gs.h +++ b/src/mesa/drivers/dri/i965/brw_gs.h @@ -40,7 +40,7 @@ #define MAX_GS_VERTS (4) struct brw_gs_prog_key { - GLuint attrs:32; + GLbitfield64 attrs; GLuint primitive:4; GLuint hint_gs_always:1; GLuint pv_first:1; diff --git a/src/mesa/drivers/dri/i965/brw_sf.c b/src/mesa/drivers/dri/i965/brw_sf.c index f350cbd74e..968890f7fb 100644 --- a/src/mesa/drivers/dri/i965/brw_sf.c +++ b/src/mesa/drivers/dri/i965/brw_sf.c @@ -61,7 +61,7 @@ static void compile_sf_prog( struct brw_context *brw, c.key = *key; c.nr_attrs = brw_count_bits(c.key.attrs); c.nr_attr_regs = (c.nr_attrs+1)/2; - c.nr_setup_attrs = brw_count_bits(c.key.attrs & DO_SETUP_BITS); + c.nr_setup_attrs = brw_count_bits(c.key.attrs); c.nr_setup_regs = (c.nr_setup_attrs+1)/2; c.prog_data.urb_read_length = c.nr_attr_regs; @@ -70,7 +70,7 @@ static void compile_sf_prog( struct brw_context *brw, /* Construct map from attribute number to position in the vertex. */ for (i = idx = 0; i < VERT_RESULT_MAX; i++) - if (c.key.attrs & (1<<i)) { + if (c.key.attrs & BITFIELD64_BIT(i)) { c.attr_to_idx[i] = idx; c.idx_to_attr[idx] = i; if (i >= VERT_RESULT_TEX0 && i <= VERT_RESULT_TEX7) { @@ -147,7 +147,7 @@ static void upload_sf_prog(struct brw_context *brw) * edgeflag testing here, it is already done in the clip * program. */ - if (key.attrs & (1<<VERT_RESULT_EDGE)) + if (key.attrs & BITFIELD64_BIT(VERT_RESULT_EDGE)) key.primitive = SF_UNFILLED_TRIS; else key.primitive = SF_TRIANGLES; diff --git a/src/mesa/drivers/dri/i965/brw_sf.h b/src/mesa/drivers/dri/i965/brw_sf.h index e835229a1d..0ba731fac9 100644 --- a/src/mesa/drivers/dri/i965/brw_sf.h +++ b/src/mesa/drivers/dri/i965/brw_sf.h @@ -45,7 +45,7 @@ #define SF_UNFILLED_TRIS 3 struct brw_sf_prog_key { - GLuint attrs:32; + GLbitfield64 attrs; GLuint primitive:2; GLuint do_twoside_color:1; GLuint do_flat_shading:1; diff --git a/src/mesa/drivers/dri/i965/brw_sf_emit.c b/src/mesa/drivers/dri/i965/brw_sf_emit.c index 561fcd501b..3eae41ee74 100644 --- a/src/mesa/drivers/dri/i965/brw_sf_emit.c +++ b/src/mesa/drivers/dri/i965/brw_sf_emit.c @@ -56,7 +56,7 @@ static struct brw_reg get_vert_attr(struct brw_sf_compile *c, static GLboolean have_attr(struct brw_sf_compile *c, GLuint attr) { - return (c->key.attrs & (1<<attr)) ? 1 : 0; + return (c->key.attrs & BITFIELD64_BIT(attr)) ? 1 : 0; } /*********************************************************************** @@ -122,8 +122,8 @@ static void do_twoside_color( struct brw_sf_compile *c ) * Flat shading */ -#define VERT_RESULT_COLOR_BITS ((1<<VERT_RESULT_COL0) | \ - (1<<VERT_RESULT_COL1)) +#define VERT_RESULT_COLOR_BITS (BITFIELD64_BIT(VERT_RESULT_COL0) | \ + BITFIELD64_BIT(VERT_RESULT_COL1)) static void copy_colors( struct brw_sf_compile *c, struct brw_reg dst, @@ -312,8 +312,8 @@ static GLboolean calculate_masks( struct brw_sf_compile *c, GLushort *pc_linear) { GLboolean is_last_attr = (reg == c->nr_setup_regs - 1); - GLuint persp_mask; - GLuint linear_mask; + GLbitfield64 persp_mask; + GLbitfield64 linear_mask; if (c->key.do_flat_shading || c->key.linear_color) persp_mask = c->key.attrs & ~(FRAG_BIT_WPOS | @@ -331,10 +331,10 @@ static GLboolean calculate_masks( struct brw_sf_compile *c, *pc_linear = 0; *pc = 0xf; - if (persp_mask & (1 << c->idx_to_attr[reg*2])) + if (persp_mask & BITFIELD64_BIT(c->idx_to_attr[reg*2])) *pc_persp = 0xf; - if (linear_mask & (1 << c->idx_to_attr[reg*2])) + if (linear_mask & BITFIELD64_BIT(c->idx_to_attr[reg*2])) *pc_linear = 0xf; /* Maybe only processs one attribute on the final round: @@ -342,10 +342,10 @@ static GLboolean calculate_masks( struct brw_sf_compile *c, if (reg*2+1 < c->nr_setup_attrs) { *pc |= 0xf0; - if (persp_mask & (1 << c->idx_to_attr[reg*2+1])) + if (persp_mask & BITFIELD64_BIT(c->idx_to_attr[reg*2+1])) *pc_persp |= 0xf0; - if (linear_mask & (1 << c->idx_to_attr[reg*2+1])) + if (linear_mask & BITFIELD64_BIT(c->idx_to_attr[reg*2+1])) *pc_linear |= 0xf0; } diff --git a/src/mesa/drivers/dri/i965/brw_state.h b/src/mesa/drivers/dri/i965/brw_state.h index ab6f158080..b129b1f1c3 100644 --- a/src/mesa/drivers/dri/i965/brw_state.h +++ b/src/mesa/drivers/dri/i965/brw_state.h @@ -112,6 +112,7 @@ void brw_validate_state(struct brw_context *brw); void brw_upload_state(struct brw_context *brw); void brw_init_state(struct brw_context *brw); void brw_destroy_state(struct brw_context *brw); +void brw_clear_validated_bos(struct brw_context *brw); /*********************************************************************** * brw_state_cache.c diff --git a/src/mesa/drivers/dri/i965/brw_state_upload.c b/src/mesa/drivers/dri/i965/brw_state_upload.c index f4283bda1b..af8dfb4c15 100644 --- a/src/mesa/drivers/dri/i965/brw_state_upload.c +++ b/src/mesa/drivers/dri/i965/brw_state_upload.c @@ -143,7 +143,7 @@ static void xor_states( struct brw_state_flags *result, result->cache = a->cache ^ b->cache; } -static void +void brw_clear_validated_bos(struct brw_context *brw) { int i; diff --git a/src/mesa/drivers/dri/i965/brw_util.c b/src/mesa/drivers/dri/i965/brw_util.c index ce21aa4869..bba9249d1b 100644 --- a/src/mesa/drivers/dri/i965/brw_util.c +++ b/src/mesa/drivers/dri/i965/brw_util.c @@ -35,7 +35,7 @@ #include "brw_util.h" #include "brw_defines.h" -GLuint brw_count_bits( GLuint val ) +GLuint brw_count_bits(uint64_t val) { GLuint i; for (i = 0; val ; val >>= 1) diff --git a/src/mesa/drivers/dri/i965/brw_util.h b/src/mesa/drivers/dri/i965/brw_util.h index 33e7cd87e4..04f3175d3e 100644 --- a/src/mesa/drivers/dri/i965/brw_util.h +++ b/src/mesa/drivers/dri/i965/brw_util.h @@ -35,7 +35,7 @@ #include "main/mtypes.h" -extern GLuint brw_count_bits( GLuint val ); +extern GLuint brw_count_bits(uint64_t val); extern GLuint brw_parameter_list_state_flags(struct gl_program_parameter_list *paramList); extern GLuint brw_translate_blend_factor( GLenum factor ); extern GLuint brw_translate_blend_equation( GLenum mode ); diff --git a/src/mesa/drivers/dri/i965/brw_vs.c b/src/mesa/drivers/dri/i965/brw_vs.c index f0c79efbd9..fd055e225e 100644 --- a/src/mesa/drivers/dri/i965/brw_vs.c +++ b/src/mesa/drivers/dri/i965/brw_vs.c @@ -56,7 +56,7 @@ static void do_vs_prog( struct brw_context *brw, c.prog_data.inputs_read = vp->program.Base.InputsRead; if (c.key.copy_edgeflag) { - c.prog_data.outputs_written |= 1<<VERT_RESULT_EDGE; + c.prog_data.outputs_written |= BITFIELD64_BIT(VERT_RESULT_EDGE); c.prog_data.inputs_read |= 1<<VERT_ATTRIB_EDGEFLAG; } diff --git a/src/mesa/drivers/dri/i965/brw_vs_emit.c b/src/mesa/drivers/dri/i965/brw_vs_emit.c index f7b0726636..00efd3443d 100644 --- a/src/mesa/drivers/dri/i965/brw_vs_emit.c +++ b/src/mesa/drivers/dri/i965/brw_vs_emit.c @@ -147,7 +147,7 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c ) mrf = 4; for (i = 0; i < VERT_RESULT_MAX; i++) { - if (c->prog_data.outputs_written & (1 << i)) { + if (c->prog_data.outputs_written & BITFIELD64_BIT(i)) { c->nr_outputs++; assert(i < Elements(c->regs[PROGRAM_OUTPUT])); if (i == VERT_RESULT_HPOS) { @@ -1124,7 +1124,7 @@ static void emit_vertex_write( struct brw_vs_compile *c) /* Update the header for point size, user clipping flags, and -ve rhw * workaround. */ - if ((c->prog_data.outputs_written & (1<<VERT_RESULT_PSIZ)) || + if ((c->prog_data.outputs_written & BITFIELD64_BIT(VERT_RESULT_PSIZ)) || c->key.nr_userclip || BRW_IS_965(p->brw)) { struct brw_reg header1 = retype(get_tmp(c), BRW_REGISTER_TYPE_UD); @@ -1134,7 +1134,7 @@ static void emit_vertex_write( struct brw_vs_compile *c) brw_set_access_mode(p, BRW_ALIGN_16); - if (c->prog_data.outputs_written & (1<<VERT_RESULT_PSIZ)) { + if (c->prog_data.outputs_written & BITFIELD64_BIT(VERT_RESULT_PSIZ)) { struct brw_reg psiz = c->regs[PROGRAM_OUTPUT][VERT_RESULT_PSIZ]; brw_MUL(p, brw_writemask(header1, WRITEMASK_W), brw_swizzle1(psiz, 0), brw_imm_f(1<<11)); brw_AND(p, brw_writemask(header1, WRITEMASK_W), header1, brw_imm_ud(0x7ff<<8)); @@ -1224,7 +1224,7 @@ static void emit_vertex_write( struct brw_vs_compile *c) */ GLuint i, mrf = 0; for (i = c->first_overflow_output; i < VERT_RESULT_MAX; i++) { - if (c->prog_data.outputs_written & (1 << i)) { + if (c->prog_data.outputs_written & BITFIELD64_BIT(i)) { /* move from GRF to MRF */ brw_MOV(p, brw_message_reg(4+mrf), c->regs[PROGRAM_OUTPUT][i]); mrf++; diff --git a/src/mesa/drivers/dri/i965/brw_vtbl.c b/src/mesa/drivers/dri/i965/brw_vtbl.c index 114e6bd018..72749b3859 100644 --- a/src/mesa/drivers/dri/i965/brw_vtbl.c +++ b/src/mesa/drivers/dri/i965/brw_vtbl.c @@ -46,7 +46,7 @@ #include "brw_state.h" #include "brw_fallback.h" #include "brw_vs.h" - +#include "brw_wm.h" static void dri_bo_release(dri_bo **bo) @@ -66,8 +66,14 @@ static void brw_destroy_context( struct intel_context *intel ) brw_destroy_state(brw); brw_draw_destroy( brw ); - - _mesa_free(brw->wm.compile_data); + brw_clear_validated_bos(brw); + if (brw->wm.compile_data) { + _mesa_free(brw->wm.compile_data->instruction); + _mesa_free(brw->wm.compile_data->vreg); + _mesa_free(brw->wm.compile_data->refs); + _mesa_free(brw->wm.compile_data->prog_instructions); + _mesa_free(brw->wm.compile_data); + } for (i = 0; i < brw->state.nr_color_regions; i++) intel_region_release(&brw->state.color_regions[i]); @@ -144,9 +150,6 @@ static void brw_new_batch( struct intel_context *intel ) { struct brw_context *brw = brw_context(&intel->ctx); - /* Check that we didn't just wrap our batchbuffer at a bad time. */ - assert(!brw->no_batch_wrap); - brw->curbe.need_new_bo = GL_TRUE; /* Mark all context state as needing to be re-emitted. diff --git a/src/mesa/drivers/dri/i965/brw_wm.c b/src/mesa/drivers/dri/i965/brw_wm.c index d8971321f3..6895f64410 100644 --- a/src/mesa/drivers/dri/i965/brw_wm.c +++ b/src/mesa/drivers/dri/i965/brw_wm.c @@ -157,7 +157,6 @@ static void do_wm_prog( struct brw_context *brw, sizeof(*c->prog_instructions)); c->vreg = _mesa_calloc(BRW_WM_MAX_VREG * sizeof(*c->vreg)); c->refs = _mesa_calloc(BRW_WM_MAX_REF * sizeof(*c->refs)); - c->vreg = _mesa_calloc(BRW_WM_MAX_VREG * sizeof(*c->vreg)); } else { void *instruction = c->instruction; void *prog_instructions = c->prog_instructions; @@ -231,7 +230,7 @@ static void brw_wm_populate_key( struct brw_context *brw, ctx->Color.AlphaEnabled) lookup |= IZ_PS_KILL_ALPHATEST_BIT; - if (fp->program.Base.OutputsWritten & (1<<FRAG_RESULT_DEPTH)) + if (fp->program.Base.OutputsWritten & BITFIELD64_BIT(FRAG_RESULT_DEPTH)) lookup |= IZ_PS_COMPUTES_DEPTH_BIT; /* _NEW_DEPTH */ @@ -347,7 +346,7 @@ static void brw_wm_populate_key( struct brw_context *brw, key->nr_color_regions = brw->state.nr_color_regions; /* CACHE_NEW_VS_PROG */ - key->vp_outputs_written = brw->vs.prog_data->outputs_written & DO_SETUP_BITS; + key->vp_outputs_written = brw->vs.prog_data->outputs_written; /* The unique fragment program ID */ key->program_string_id = fp->id; diff --git a/src/mesa/drivers/dri/i965/brw_wm.h b/src/mesa/drivers/dri/i965/brw_wm.h index b3c05eb0ad..b9b987ea70 100644 --- a/src/mesa/drivers/dri/i965/brw_wm.h +++ b/src/mesa/drivers/dri/i965/brw_wm.h @@ -76,10 +76,10 @@ struct brw_wm_prog_key { GLushort tex_swizzles[BRW_MAX_TEX_UNIT]; - GLuint program_string_id:32; GLushort origin_x, origin_y; GLushort drawable_height; - GLuint vp_outputs_written; + GLbitfield64 vp_outputs_written; + GLuint program_string_id:32; }; diff --git a/src/mesa/drivers/dri/i965/brw_wm_fp.c b/src/mesa/drivers/dri/i965/brw_wm_fp.c index 1c4f62ba48..3737faf26f 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_fp.c +++ b/src/mesa/drivers/dri/i965/brw_wm_fp.c @@ -138,7 +138,6 @@ static struct prog_dst_register dst_reg(GLuint file, GLuint idx) reg.CondMask = COND_TR; reg.CondSwizzle = 0; reg.CondSrc = 0; - reg.pad = 0; return reg; } @@ -986,7 +985,7 @@ static void emit_render_target_writes( struct brw_wm_compile *c ) } else { /* if gl_FragData[0] is written, use it, else use gl_FragColor */ - if (c->fp->program.Base.OutputsWritten & (1 << FRAG_RESULT_DATA0)) + if (c->fp->program.Base.OutputsWritten & BITFIELD64_BIT(FRAG_RESULT_DATA0)) outcolor = src_reg(PROGRAM_OUTPUT, FRAG_RESULT_DATA0); else outcolor = src_reg(PROGRAM_OUTPUT, FRAG_RESULT_COLOR); diff --git a/src/mesa/drivers/dri/i965/brw_wm_glsl.c b/src/mesa/drivers/dri/i965/brw_wm_glsl.c index 3ab446164c..e8c2cb66ec 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_glsl.c +++ b/src/mesa/drivers/dri/i965/brw_wm_glsl.c @@ -371,7 +371,7 @@ static void prealloc_reg(struct brw_wm_compile *c) for (j = 0; j < 4; j++) set_reg(c, PROGRAM_PAYLOAD, fp_input, j, reg); } - if (c->key.vp_outputs_written & (1 << i)) { + if (c->key.vp_outputs_written & BITFIELD64_BIT(i)) { reg_index += 2; } } diff --git a/src/mesa/drivers/dri/i965/brw_wm_pass2.c b/src/mesa/drivers/dri/i965/brw_wm_pass2.c index 6faea018fb..31303febf0 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_pass2.c +++ b/src/mesa/drivers/dri/i965/brw_wm_pass2.c @@ -82,8 +82,8 @@ static void init_registers( struct brw_wm_compile *c ) for (j = 0; j < c->nr_creg; j++) prealloc_reg(c, &c->creg[j], i++); - for (j = 0; j < FRAG_ATTRIB_MAX; j++) { - if (c->key.vp_outputs_written & (1<<j)) { + for (j = 0; j < VERT_RESULT_MAX; j++) { + if (c->key.vp_outputs_written & BITFIELD64_BIT(j)) { int fp_index; if (j >= VERT_RESULT_VAR0) diff --git a/src/mesa/drivers/dri/i965/brw_wm_state.c b/src/mesa/drivers/dri/i965/brw_wm_state.c index 361f91292b..f89ed9bce7 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_state.c +++ b/src/mesa/drivers/dri/i965/brw_wm_state.c @@ -106,7 +106,7 @@ wm_unit_populate_key(struct brw_context *brw, struct brw_wm_unit_key *key) /* as far as we can tell */ key->computes_depth = - (fp->Base.OutputsWritten & (1 << FRAG_RESULT_DEPTH)) != 0; + (fp->Base.OutputsWritten & BITFIELD64_BIT(FRAG_RESULT_DEPTH)) != 0; /* BRW_NEW_DEPTH_BUFFER * Override for NULL depthbuffer case, required by the Pixel Shader Computed * Depth field. diff --git a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c index e2f0a383e7..47035cc6fc 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c +++ b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c @@ -86,6 +86,9 @@ static GLuint translate_tex_format( gl_format mesa_format, case MESA_FORMAT_AL88: return BRW_SURFACEFORMAT_L8A8_UNORM; + case MESA_FORMAT_AL1616: + return BRW_SURFACEFORMAT_L16A16_UNORM; + case MESA_FORMAT_RGB888: assert(0); /* not supported for sampling */ return BRW_SURFACEFORMAT_R8G8B8_UNORM; diff --git a/src/mesa/drivers/dri/intel/intel_batchbuffer.c b/src/mesa/drivers/dri/intel/intel_batchbuffer.c index ca6e2fa5b1..2eae9b66d8 100644 --- a/src/mesa/drivers/dri/intel/intel_batchbuffer.c +++ b/src/mesa/drivers/dri/intel/intel_batchbuffer.c @@ -80,7 +80,7 @@ intel_batchbuffer_reset(struct intel_batchbuffer *batch) batch->buf = NULL; } - if (!batch->buffer && intel->ttm == GL_TRUE) + if (!batch->buffer) batch->buffer = malloc (intel->maxBatchSize); batch->buf = dri_bo_alloc(intel->bufmgr, "batchbuffer", @@ -212,7 +212,7 @@ _intel_batchbuffer_flush(struct intel_batchbuffer *batch, const char *file, batch->reserved_space = 0; /* Emit a flush if the bufmgr doesn't do it for us. */ - if (intel->always_flush_cache || !intel->ttm) { + if (intel->always_flush_cache) { intel_batchbuffer_emit_mi_flush(batch); used = batch->ptr - batch->map; } @@ -244,6 +244,9 @@ _intel_batchbuffer_flush(struct intel_batchbuffer *batch, const char *file, if (intel->vtbl.finish_batch) intel->vtbl.finish_batch(intel); + /* Check that we didn't just wrap our batchbuffer at a bad time. */ + assert(!intel->no_batch_wrap); + batch->reserved_space = BATCH_RESERVED; /* TODO: Just pass the relocation list and dma buffer up to the diff --git a/src/mesa/drivers/dri/intel/intel_blit.c b/src/mesa/drivers/dri/intel/intel_blit.c index 817223da41..f14854602b 100644 --- a/src/mesa/drivers/dri/intel/intel_blit.c +++ b/src/mesa/drivers/dri/intel/intel_blit.c @@ -499,10 +499,11 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask) switch (irb->texformat) { case MESA_FORMAT_ARGB8888: case MESA_FORMAT_XRGB8888: - clearVal = intel->ClearColor8888; + clearVal = PACK_COLOR_8888(clear[3], clear[0], + clear[1], clear[2]); break; case MESA_FORMAT_RGB565: - clearVal = intel->ClearColor565; + clearVal = PACK_COLOR_565(clear[0], clear[1], clear[2]); break; case MESA_FORMAT_ARGB4444: clearVal = PACK_COLOR_4444(clear[3], clear[0], diff --git a/src/mesa/drivers/dri/intel/intel_context.c b/src/mesa/drivers/dri/intel/intel_context.c index 2aeca6b81b..1434ae530b 100644 --- a/src/mesa/drivers/dri/intel/intel_context.c +++ b/src/mesa/drivers/dri/intel/intel_context.c @@ -176,9 +176,7 @@ intelGetString(GLcontext * ctx, GLenum name) break; } - (void) driGetRendererString(buffer, chipset, - (intel->ttm) ? DRIVER_DATE_GEM : DRIVER_DATE, - 0); + (void) driGetRendererString(buffer, chipset, DRIVER_DATE_GEM, 0); return (GLubyte *) buffer; default: @@ -601,6 +599,7 @@ intelInitContext(struct intel_context *intel, __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; int fthrottle_mode; + int bo_reuse_mode; if (!_mesa_initialize_context(&intel->ctx, mesaVis, shareCtx, functions, (void *) intel)) { @@ -635,18 +634,14 @@ intelInitContext(struct intel_context *intel, intel->maxBatchSize = BATCH_SZ; intel->bufmgr = intelScreen->bufmgr; - intel->ttm = intelScreen->ttm; - if (intel->ttm) { - int bo_reuse_mode; - bo_reuse_mode = driQueryOptioni(&intel->optionCache, "bo_reuse"); - switch (bo_reuse_mode) { - case DRI_CONF_BO_REUSE_DISABLED: - break; - case DRI_CONF_BO_REUSE_ALL: - intel_bufmgr_gem_enable_reuse(intel->bufmgr); - break; - } + bo_reuse_mode = driQueryOptioni(&intel->optionCache, "bo_reuse"); + switch (bo_reuse_mode) { + case DRI_CONF_BO_REUSE_DISABLED: + break; + case DRI_CONF_BO_REUSE_ALL: + intel_bufmgr_gem_enable_reuse(intel->bufmgr); + break; } /* This doesn't yet catch all non-conformant rendering, but it's a @@ -733,12 +728,6 @@ intelInitContext(struct intel_context *intel, intel->RenderIndex = ~0; fthrottle_mode = driQueryOptioni(&intel->optionCache, "fthrottle_mode"); - 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); if (intel->gen >= 4 && !intel->intelScreen->irq_active) { _mesa_printf("IRQs not active. Exiting\n"); @@ -1058,21 +1047,6 @@ intelContendedLock(struct intel_context *intel, GLuint flags) sarea->ctxOwner = me; } - /* If the last consumer of the texture memory wasn't us, notify the fake - * bufmgr and record the new owner. We should have the memory shared - * between contexts of a single fake bufmgr, but this will at least make - * things correct for now. - */ - if (!intel->ttm && sarea->texAge != intel->hHWContext) { - sarea->texAge = intel->hHWContext; - intel_bufmgr_fake_contended_lock_take(intel->bufmgr); - if (INTEL_DEBUG & DEBUG_BATCH) - intel_decode_context_reset(); - if (INTEL_DEBUG & DEBUG_BUFMGR) - fprintf(stderr, "Lost Textures: sarea->texAge %x hw context %x\n", - sarea->ctxOwner, intel->hHWContext); - } - /* Drawable changed? */ if (dPriv && intel->lastStamp != dPriv->lastStamp) { diff --git a/src/mesa/drivers/dri/intel/intel_context.h b/src/mesa/drivers/dri/intel/intel_context.h index eb7be7ddd0..481202c971 100644 --- a/src/mesa/drivers/dri/intel/intel_context.h +++ b/src/mesa/drivers/dri/intel/intel_context.h @@ -135,14 +135,6 @@ struct intel_context struct intel_region * draw_region, struct intel_region * depth_region); - void (*meta_draw_quad)(struct intel_context *intel, - GLfloat x0, GLfloat x1, - GLfloat y0, GLfloat y1, - GLfloat z, - GLuint color, /* ARGB32 */ - GLfloat s0, GLfloat s1, - GLfloat t0, GLfloat t1); - void (*meta_color_mask) (struct intel_context * intel, GLboolean); void (*meta_stencil_replace) (struct intel_context * intel, @@ -189,12 +181,6 @@ struct intel_context struct intel_region *back_region; struct intel_region *depth_region; - /** - * This value indicates that the kernel memory manager is being used - * instead of the fake client-side memory manager. - */ - GLboolean ttm; - struct intel_batchbuffer *batch; drm_intel_bo *first_post_swapbuffers_batch; GLboolean no_batch_wrap; @@ -217,10 +203,6 @@ struct intel_context char *prevLockFile; int prevLockLine; - GLuint ClearColor565; - GLuint ClearColor8888; - - /* Offsets of fields within the current vertex: */ GLuint coloroffset; @@ -237,6 +219,7 @@ struct intel_context GLboolean hw_stipple; GLboolean depth_buffer_is_float; GLboolean no_rast; + GLboolean no_hw; GLboolean always_flush_batch; GLboolean always_flush_cache; @@ -302,13 +285,6 @@ struct intel_context GLboolean use_early_z; drm_clip_rect_t fboRect; /**< cliprect for FBO rendering */ - int perf_boxes; - - GLuint do_usleeps; - int do_irqs; - GLuint irqsEmitted; - - GLboolean scissor; drm_clip_rect_t draw_rect; drm_clip_rect_t scissor_rect; @@ -325,8 +301,6 @@ struct intel_context GLuint lastStamp; - GLboolean no_hw; - /** * Configuration cache */ @@ -374,29 +348,6 @@ do { \ } 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: * XXX Put this in src/mesa/main/imports.h ??? diff --git a/src/mesa/drivers/dri/intel/intel_extensions.c b/src/mesa/drivers/dri/intel/intel_extensions.c index 1682e115cc..f5fe543b5d 100644 --- a/src/mesa/drivers/dri/intel/intel_extensions.c +++ b/src/mesa/drivers/dri/intel/intel_extensions.c @@ -79,6 +79,7 @@ static const struct dri_extension card_extensions[] = { { "GL_ARB_half_float_pixel", NULL }, { "GL_ARB_map_buffer_range", GL_ARB_map_buffer_range_functions }, { "GL_ARB_multitexture", NULL }, + { "GL_ARB_pixel_buffer_object", NULL }, { "GL_ARB_point_parameters", GL_ARB_point_parameters_functions }, { "GL_ARB_point_sprite", NULL }, { "GL_ARB_shader_objects", GL_ARB_shader_objects_functions }, @@ -104,6 +105,8 @@ static const struct dri_extension card_extensions[] = { { "GL_EXT_blend_logic_op", NULL }, { "GL_EXT_blend_subtract", NULL }, { "GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions }, + { "GL_EXT_framebuffer_blit", GL_EXT_framebuffer_blit_functions }, + { "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions }, { "GL_EXT_fog_coord", GL_EXT_fog_coord_functions }, { "GL_EXT_gpu_program_parameters", GL_EXT_gpu_program_parameters_functions }, { "GL_EXT_packed_depth_stencil", NULL }, @@ -176,14 +179,6 @@ static const struct dri_extension arb_oq_extensions[] = { { NULL, NULL } }; - -static const struct dri_extension ttm_extensions[] = { - { "GL_ARB_pixel_buffer_object", NULL }, - { "GL_EXT_framebuffer_blit", GL_EXT_framebuffer_blit_functions }, - { "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions }, - { NULL, NULL } -}; - static const struct dri_extension fragment_shader_extensions[] = { { "GL_ARB_fragment_shader", NULL }, { NULL, NULL } @@ -202,9 +197,6 @@ intelInitExtensions(GLcontext *ctx) */ driInitExtensions(ctx, card_extensions, GL_FALSE); - if (intel->ttm) - driInitExtensions(ctx, ttm_extensions, GL_FALSE); - if (IS_965(intel->intelScreen->deviceID)) driInitExtensions(ctx, brw_extensions, GL_FALSE); diff --git a/src/mesa/drivers/dri/intel/intel_mipmap_tree.c b/src/mesa/drivers/dri/intel/intel_mipmap_tree.c index abb3024bfb..6a565f80cf 100644 --- a/src/mesa/drivers/dri/intel/intel_mipmap_tree.c +++ b/src/mesa/drivers/dri/intel/intel_mipmap_tree.c @@ -224,16 +224,12 @@ int intel_miptree_pitch_align (struct intel_context *intel, if (!mt->compressed) { int pitch_align; - if (intel->ttm) { - /* XXX: Align pitch to multiple of 64 bytes for now to allow - * render-to-texture to work in all cases. This should probably be - * replaced at some point by some scheme to only do this when really - * necessary. - */ - pitch_align = 64; - } else { - pitch_align = 4; - } + /* XXX: Align pitch to multiple of 64 bytes for now to allow + * render-to-texture to work in all cases. This should probably be + * replaced at some point by some scheme to only do this when really + * necessary. + */ + pitch_align = 64; if (tiling == I915_TILING_X) pitch_align = 512; diff --git a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c index 99330b6ddf..9572b67326 100644 --- a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c +++ b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c @@ -228,10 +228,9 @@ do_blit_bitmap( GLcontext *ctx, UNCLAMPED_FLOAT_TO_UBYTE(ubcolor[3], tmpColor[3]); if (dst->cpp == 2) - color = INTEL_PACKCOLOR565(ubcolor[0], ubcolor[1], ubcolor[2]); + color = PACK_COLOR_565(ubcolor[0], ubcolor[1], ubcolor[2]); else - color = INTEL_PACKCOLOR8888(ubcolor[0], ubcolor[1], - ubcolor[2], ubcolor[3]); + color = PACK_COLOR_8888(ubcolor[3], ubcolor[0], ubcolor[1], ubcolor[2]); if (!intel_check_blit_fragment_ops(ctx, tmpColor[3] == 1.0F)) return GL_FALSE; diff --git a/src/mesa/drivers/dri/intel/intel_regions.c b/src/mesa/drivers/dri/intel/intel_regions.c index 80975163d4..d6b9dc4446 100644 --- a/src/mesa/drivers/dri/intel/intel_regions.c +++ b/src/mesa/drivers/dri/intel/intel_regions.c @@ -542,55 +542,18 @@ intel_recreate_static(struct intel_context *intel, region->buffer = NULL; } - if (intel->ttm) { - assert(region_desc->bo_handle != -1); - region->buffer = intel_bo_gem_create_from_name(intel->bufmgr, - name, - region_desc->bo_handle); - - ret = dri_bo_get_tiling(region->buffer, ®ion->tiling, - ®ion->bit_6_swizzle); - if (ret != 0) { - fprintf(stderr, "Couldn't get tiling of buffer %d (%s): %s\n", - region_desc->bo_handle, name, strerror(-ret)); - intel_region_release(®ion); - return NULL; - } - } else { - if (region->classic_map != NULL) { - drmUnmap(region->classic_map, - region->pitch * region->cpp * region->height); - region->classic_map = NULL; - } - ret = drmMap(intel->driFd, region_desc->handle, - region->pitch * region->cpp * region->height, - ®ion->classic_map); - if (ret != 0) { - fprintf(stderr, "Failed to drmMap %s buffer\n", name); - free(region); - return NULL; - } - - region->buffer = intel_bo_fake_alloc_static(intel->bufmgr, + assert(region_desc->bo_handle != -1); + region->buffer = intel_bo_gem_create_from_name(intel->bufmgr, name, - region_desc->offset, - region->pitch * region->cpp * - region->height, - region->classic_map); - - /* The sarea just gives us a boolean for whether it's tiled or not, - * instead of which tiling mode it is. Guess. - */ - if (region_desc->tiled) { - if (intel->gen >= 4 && region_desc == &intelScreen->depth) - region->tiling = I915_TILING_Y; - else - region->tiling = I915_TILING_X; - } else { - region->tiling = I915_TILING_NONE; - } + region_desc->bo_handle); - region->bit_6_swizzle = I915_BIT_6_SWIZZLE_NONE; + ret = dri_bo_get_tiling(region->buffer, ®ion->tiling, + ®ion->bit_6_swizzle); + if (ret != 0) { + fprintf(stderr, "Couldn't get tiling of buffer %d (%s): %s\n", + region_desc->bo_handle, name, strerror(-ret)); + intel_region_release(®ion); + return NULL; } assert(region->buffer != NULL); diff --git a/src/mesa/drivers/dri/intel/intel_screen.c b/src/mesa/drivers/dri/intel/intel_screen.c index 789135b49f..2c5a884a9b 100644 --- a/src/mesa/drivers/dri/intel/intel_screen.c +++ b/src/mesa/drivers/dri/intel/intel_screen.c @@ -605,7 +605,6 @@ intelFillInModes(__DRIscreenPrivate *psp, static GLboolean intel_init_bufmgr(intelScreenPrivate *intelScreen) { - GLboolean gem_disable = getenv("INTEL_NO_GEM") != NULL; int gem_kernel = 0; GLboolean gem_supported; struct drm_i915_getparam gp; @@ -622,43 +621,24 @@ intel_init_bufmgr(intelScreenPrivate *intelScreen) /* If we've got a new enough DDX that's initializing GEM and giving us * object handles for the shared buffers, use that. */ - intelScreen->ttm = GL_FALSE; if (intelScreen->driScrnPriv->dri2.enabled) gem_supported = GL_TRUE; else if (intelScreen->driScrnPriv->ddx_version.minor >= 9 && gem_kernel && intelScreen->front.bo_handle != -1) gem_supported = GL_TRUE; - else - gem_supported = GL_FALSE; - - if (!gem_disable && gem_supported) { - intelScreen->bufmgr = intel_bufmgr_gem_init(spriv->fd, BATCH_SZ); - if (intelScreen->bufmgr != NULL) - intelScreen->ttm = GL_TRUE; + else { + fprintf(stderr, "[%s:%u] Error initializing GEM.\n", + __func__, __LINE__); + return GL_FALSE; } + + intelScreen->bufmgr = intel_bufmgr_gem_init(spriv->fd, BATCH_SZ); /* Otherwise, use the classic buffer manager. */ if (intelScreen->bufmgr == NULL) { - if (gem_disable) { - _mesa_warning(NULL, "GEM disabled. Using classic."); - } else { - _mesa_warning(NULL, - "Failed to initialize GEM. Falling back to classic."); - } - - if (intelScreen->tex.size == 0) { - fprintf(stderr, "[%s:%u] Error initializing buffer manager.\n", - __func__, __LINE__); - return GL_FALSE; - } - - intelScreen->bufmgr = - intel_bufmgr_fake_init(spriv->fd, - intelScreen->tex.offset, - intelScreen->tex.map, - intelScreen->tex.size, - (unsigned int * volatile) - &intelScreen->sarea->last_dispatch); + fprintf(stderr, "[%s:%u] Error initializing buffer manager.\n", + __func__, __LINE__); + return GL_FALSE; } if (intel_get_param(spriv, I915_PARAM_NUM_FENCES_AVAIL, &num_fences)) diff --git a/src/mesa/drivers/dri/intel/intel_screen.h b/src/mesa/drivers/dri/intel/intel_screen.h index a9b9e109a6..14ca0903b6 100644 --- a/src/mesa/drivers/dri/intel/intel_screen.h +++ b/src/mesa/drivers/dri/intel/intel_screen.h @@ -77,7 +77,6 @@ typedef struct GLboolean no_hw; GLboolean no_vbo; - int ttm; dri_bufmgr *bufmgr; GLboolean kernel_exec_fencing; diff --git a/src/mesa/drivers/dri/intel/intel_span.c b/src/mesa/drivers/dri/intel/intel_span.c index 3607c7dded..2c89a66a95 100644 --- a/src/mesa/drivers/dri/intel/intel_span.c +++ b/src/mesa/drivers/dri/intel/intel_span.c @@ -613,15 +613,7 @@ intel_set_span_functions(struct intel_context *intel, struct gl_renderbuffer *rb) { struct intel_renderbuffer *irb = (struct intel_renderbuffer *) rb; - uint32_t tiling; - - /* If in GEM mode, we need to do the tile address swizzling ourselves, - * instead of the fence registers handling it. - */ - if (intel->ttm) - tiling = irb->region->tiling; - else - tiling = I915_TILING_NONE; + uint32_t tiling = irb->region->tiling; if (intel->intelScreen->kernel_exec_fencing) { switch (irb->texformat) { @@ -673,6 +665,9 @@ intel_set_span_functions(struct intel_context *intel, return; } + /* If in GEM mode, we need to do the tile address swizzling ourselves, + * instead of the fence registers handling it. + */ switch (irb->texformat) { case MESA_FORMAT_RGB565: switch (tiling) { diff --git a/src/mesa/drivers/dri/intel/intel_state.c b/src/mesa/drivers/dri/intel/intel_state.c index 4ee742377d..aefae53eb2 100644 --- a/src/mesa/drivers/dri/intel/intel_state.c +++ b/src/mesa/drivers/dri/intel/intel_state.c @@ -196,25 +196,6 @@ intel_translate_logic_op(GLenum opcode) } } - -static void -intelClearColor(GLcontext *ctx, const GLfloat color[4]) -{ - struct intel_context *intel = intel_context(ctx); - GLubyte clear[4]; - - 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]); - - /* 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]); -} - - /* Fallback to swrast for select and feedback. */ static void @@ -229,5 +210,4 @@ void intelInitStateFuncs(struct dd_function_table *functions) { functions->RenderMode = intelRenderMode; - functions->ClearColor = intelClearColor; } diff --git a/src/mesa/drivers/dri/intel/intel_tex_copy.c b/src/mesa/drivers/dri/intel/intel_tex_copy.c index 4b5fe7be9f..767d04d2f4 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_copy.c +++ b/src/mesa/drivers/dri/intel/intel_tex_copy.c @@ -221,6 +221,8 @@ intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, return; fail: + if (INTEL_DEBUG & DEBUG_FALLBACKS) + fprintf(stderr, "%s - fallback to swrast\n", __FUNCTION__); _mesa_meta_CopyTexImage1D(ctx, target, level, internalFormat, x, y, width, border); } @@ -268,6 +270,8 @@ intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, return; fail: + if (INTEL_DEBUG & DEBUG_FALLBACKS) + fprintf(stderr, "%s - fallback to swrast\n", __FUNCTION__); _mesa_meta_CopyTexImage2D(ctx, target, level, internalFormat, x, y, width, height, border); } @@ -292,6 +296,8 @@ intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, if (!do_copy_texsubimage(intel_context(ctx), target, intel_texture_image(texImage), internalFormat, xoffset, 0, x, y, width, 1)) { + if (INTEL_DEBUG & DEBUG_FALLBACKS) + fprintf(stderr, "%s - fallback to swrast\n", __FUNCTION__); _mesa_meta_CopyTexSubImage1D(ctx, target, level, xoffset, x, y, width); } } @@ -317,8 +323,8 @@ intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, internalFormat, xoffset, yoffset, x, y, width, height)) { - DBG("%s - fallback to _mesa_meta_CopyTexSubImage2D\n", __FUNCTION__); - + if (INTEL_DEBUG & DEBUG_FALLBACKS) + fprintf(stderr, "%s - fallback to swrast\n", __FUNCTION__); _mesa_meta_CopyTexSubImage2D(ctx, target, level, xoffset, yoffset, x, y, width, height); } diff --git a/src/mesa/drivers/dri/intel/intel_tex_format.c b/src/mesa/drivers/dri/intel/intel_tex_format.c index f37a545c7f..bfa3dba1f5 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_format.c +++ b/src/mesa/drivers/dri/intel/intel_tex_format.c @@ -95,14 +95,20 @@ intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat, case GL_COMPRESSED_LUMINANCE: return MESA_FORMAT_L8; + case GL_LUMINANCE12_ALPHA4: + case GL_LUMINANCE12_ALPHA12: + case GL_LUMINANCE16_ALPHA16: +#ifndef I915 + return MESA_FORMAT_AL1616; +#else + /* FALLTHROUGH */ +#endif + 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_FORMAT_AL88; diff --git a/src/mesa/drivers/dri/mach64/mach64_native_vb.c b/src/mesa/drivers/dri/mach64/mach64_native_vb.c index 99f1a14e17..816682ec5f 100644 --- a/src/mesa/drivers/dri/mach64/mach64_native_vb.c +++ b/src/mesa/drivers/dri/mach64/mach64_native_vb.c @@ -207,19 +207,19 @@ INTERP_QUALIFIER void TAG(interp_extras)( GLcontext *ctx, LOCALVARS struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; - if (VB->ColorPtr[1]) { - assert(VB->ColorPtr[1]->stride == 4 * sizeof(GLfloat)); + if (VB->BackfaceColorPtr) { + assert(VB->BackfaceColorPtr->stride == 4 * sizeof(GLfloat)); INTERP_4F( t, - GET_COLOR(VB->ColorPtr[1], dst), - GET_COLOR(VB->ColorPtr[1], out), - GET_COLOR(VB->ColorPtr[1], in) ); + GET_COLOR(VB->BackfaceColorPtr, dst), + GET_COLOR(VB->BackfaceColorPtr, out), + GET_COLOR(VB->BackfaceColorPtr, in) ); - if (VB->SecondaryColorPtr[1]) { + if (VB->BackfaceSecondaryColorPtr) { INTERP_3F( t, - GET_COLOR(VB->SecondaryColorPtr[1], dst), - GET_COLOR(VB->SecondaryColorPtr[1], out), - GET_COLOR(VB->SecondaryColorPtr[1], in) ); + GET_COLOR(VB->BackfaceSecondaryColorPtr, dst), + GET_COLOR(VB->BackfaceSecondaryColorPtr, out), + GET_COLOR(VB->BackfaceSecondaryColorPtr, in) ); } } @@ -236,13 +236,13 @@ INTERP_QUALIFIER void TAG(copy_pv_extras)( GLcontext *ctx, LOCALVARS struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; - if (VB->ColorPtr[1]) { - COPY_4FV( GET_COLOR(VB->ColorPtr[1], dst), - GET_COLOR(VB->ColorPtr[1], src) ); + if (VB->BackfaceColorPtr) { + COPY_4FV( GET_COLOR(VB->BackfaceColorPtr, dst), + GET_COLOR(VB->BackfaceColorPtr, src) ); - if (VB->SecondaryColorPtr[1]) { - COPY_4FV( GET_COLOR(VB->SecondaryColorPtr[1], dst), - GET_COLOR(VB->SecondaryColorPtr[1], src) ); + if (VB->BackfaceSecondaryColorPtr) { + COPY_4FV( GET_COLOR(VB->BackfaceSecondaryColorPtr, dst), + GET_COLOR(VB->BackfaceSecondaryColorPtr, src) ); } } diff --git a/src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h b/src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h index 684f2acc89..6e5fa3520e 100644 --- a/src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h +++ b/src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h @@ -103,10 +103,10 @@ static void TAG(emit)( GLcontext *ctx, #if DO_TEX1 { const GLuint t1 = GET_TEXSOURCE(1); - tc1 = VB->TexCoordPtr[t1]->data; - tc1_stride = VB->TexCoordPtr[t1]->stride; + tc1 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->data; + tc1_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->stride; #if DO_PTEX - tc1_size = VB->TexCoordPtr[t1]->size; + tc1_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->size; #endif } #endif @@ -114,18 +114,18 @@ static void TAG(emit)( GLcontext *ctx, #if DO_TEX0 { const GLuint t0 = GET_TEXSOURCE(0); - tc0 = VB->TexCoordPtr[t0]->data; - tc0_stride = VB->TexCoordPtr[t0]->stride; + tc0 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->data; + tc0_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->stride; #if DO_PTEX - tc0_size = VB->TexCoordPtr[t0]->size; + tc0_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->size; #endif } #endif #if DO_SPEC - if (VB->SecondaryColorPtr[0]) { - spec = VB->SecondaryColorPtr[0]->data; - spec_stride = VB->SecondaryColorPtr[0]->stride; + if (VB->AttribPtr[_TNL_ATTRIB_COLOR1]) { + spec = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->data; + spec_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->stride; } else { spec = (GLfloat (*)[4])ctx->Current.Attrib[VERT_ATTRIB_COLOR1]; spec_stride = 0; @@ -133,9 +133,9 @@ static void TAG(emit)( GLcontext *ctx, #endif #if DO_FOG - if (VB->FogCoordPtr) { - fog = VB->FogCoordPtr->data; - fog_stride = VB->FogCoordPtr->stride; + if (VB->AttribPtr[_TNL_ATTRIB_FOG]) { + fog = VB->AttribPtr[_TNL_ATTRIB_FOG]->data; + fog_stride = VB->AttribPtr[_TNL_ATTRIB_FOG]->stride; } else { static GLfloat tmp[4] = {0, 0, 0, 0}; fog = &tmp; @@ -144,8 +144,8 @@ static void TAG(emit)( GLcontext *ctx, #endif #if DO_RGBA - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; #endif coord = VB->NdcPtr->data; @@ -319,8 +319,8 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) /* Force 'missing' texcoords to something valid. */ - if (DO_TEX1 && VB->TexCoordPtr[0] == 0) - VB->TexCoordPtr[0] = VB->TexCoordPtr[1]; + if (DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0] = VB->AttribPtr[_TNL_ATTRIB_TEX1]; if (DO_PTEX) return GL_TRUE; @@ -328,12 +328,12 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) /* No hardware support for projective texture. Can fake it for * TEX0 only. */ - if ((DO_TEX1 && VB->TexCoordPtr[GET_TEXSOURCE(1)]->size == 4)) { + if ((DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(1)]->size == 4)) { PTEX_FALLBACK(); return GL_FALSE; } - if (DO_TEX0 && VB->TexCoordPtr[GET_TEXSOURCE(0)]->size == 4) { + if (DO_TEX0 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(0)]->size == 4) { if (DO_TEX1) { PTEX_FALLBACK(); } diff --git a/src/mesa/drivers/dri/mach64/mach64_vbtmp.h b/src/mesa/drivers/dri/mach64/mach64_vbtmp.h index 938804af9e..60bfab8f6d 100644 --- a/src/mesa/drivers/dri/mach64/mach64_vbtmp.h +++ b/src/mesa/drivers/dri/mach64/mach64_vbtmp.h @@ -156,53 +156,53 @@ static void TAG(emit)( GLcontext *ctx, if (DO_TEX3) { const GLuint t3 = GET_TEXSOURCE(3); - tc3 = VB->TexCoordPtr[t3]->data; - tc3_stride = VB->TexCoordPtr[t3]->stride; + tc3 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t3]->data; + tc3_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t3]->stride; if (DO_PTEX) - tc3_size = VB->TexCoordPtr[t3]->size; + tc3_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t3]->size; } if (DO_TEX2) { const GLuint t2 = GET_TEXSOURCE(2); - tc2 = VB->TexCoordPtr[t2]->data; - tc2_stride = VB->TexCoordPtr[t2]->stride; + tc2 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->data; + tc2_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->stride; if (DO_PTEX) - tc2_size = VB->TexCoordPtr[t2]->size; + tc2_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->size; } if (DO_TEX1) { const GLuint t1 = GET_TEXSOURCE(1); - tc1 = VB->TexCoordPtr[t1]->data; - tc1_stride = VB->TexCoordPtr[t1]->stride; + tc1 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->data; + tc1_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->stride; if (DO_PTEX) - tc1_size = VB->TexCoordPtr[t1]->size; + tc1_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->size; } if (DO_TEX0) { const GLuint t0 = GET_TEXSOURCE(0); - tc0_stride = VB->TexCoordPtr[t0]->stride; - tc0 = VB->TexCoordPtr[t0]->data; + tc0_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->stride; + tc0 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->data; if (DO_PTEX) - tc0_size = VB->TexCoordPtr[t0]->size; + tc0_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->size; } if (DO_RGBA) { - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; } if (DO_SPEC) { - spec = VB->SecondaryColorPtr[0]->data; - spec_stride = VB->SecondaryColorPtr[0]->stride; + spec = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->data; + spec_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->stride; } else { spec = (GLfloat (*)[4])ctx->Current.Attrib[VERT_ATTRIB_COLOR1]; spec_stride = 0; } if (DO_FOG) { - if (VB->FogCoordPtr) { - fog = VB->FogCoordPtr->data; - fog_stride = VB->FogCoordPtr->stride; + if (VB->AttribPtr[_TNL_ATTRIB_FOG]) { + fog = VB->AttribPtr[_TNL_ATTRIB_FOG]->data; + fog_stride = VB->AttribPtr[_TNL_ATTRIB_FOG]->stride; } else { static GLfloat tmp[4] = {0, 0, 0, 0}; fog = &tmp; @@ -384,8 +384,8 @@ static void TAG(emit)( GLcontext *ctx, GLuint start, GLuint end, ASSERT(stride == 4); - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; /* Pack what's left into a 4-dword vertex. Color is in a different * place, and there is no 'w' coordinate. @@ -432,8 +432,8 @@ static void TAG(emit)( GLcontext *ctx, GLuint start, GLuint end, GLfloat *v = (GLfloat *)dest; int i; - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; if (start) STRIDE_4F(col, col_stride * start); @@ -473,22 +473,22 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) /* Force 'missing' texcoords to something valid. */ - if (DO_TEX3 && VB->TexCoordPtr[2] == 0) - VB->TexCoordPtr[2] = VB->TexCoordPtr[3]; + if (DO_TEX3 && VB->AttribPtr[_TNL_ATTRIB_TEX2] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX2] = VB->AttribPtr[_TNL_ATTRIB_TEX3]; - if (DO_TEX2 && VB->TexCoordPtr[1] == 0) - VB->TexCoordPtr[1] = VB->TexCoordPtr[2]; + if (DO_TEX2 && VB->AttribPtr[_TNL_ATTRIB_TEX1] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX1] = VB->AttribPtr[_TNL_ATTRIB_TEX2]; - if (DO_TEX1 && VB->TexCoordPtr[0] == 0) - VB->TexCoordPtr[0] = VB->TexCoordPtr[1]; + if (DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0] = VB->AttribPtr[_TNL_ATTRIB_TEX1]; if (DO_PTEX) return GL_TRUE; - if ((DO_TEX3 && VB->TexCoordPtr[GET_TEXSOURCE(3)]->size == 4) || - (DO_TEX2 && VB->TexCoordPtr[GET_TEXSOURCE(2)]->size == 4) || - (DO_TEX1 && VB->TexCoordPtr[GET_TEXSOURCE(1)]->size == 4) || - (DO_TEX0 && VB->TexCoordPtr[GET_TEXSOURCE(0)]->size == 4)) + if ((DO_TEX3 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(3)]->size == 4) || + (DO_TEX2 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(2)]->size == 4) || + (DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(1)]->size == 4) || + (DO_TEX0 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(0)]->size == 4)) return GL_FALSE; return GL_TRUE; @@ -501,14 +501,14 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) /* Force 'missing' texcoords to something valid. */ - if (DO_TEX3 && VB->TexCoordPtr[2] == 0) - VB->TexCoordPtr[2] = VB->TexCoordPtr[3]; + if (DO_TEX3 && VB->AttribPtr[_TNL_ATTRIB_TEX2] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX2] = VB->AttribPtr[_TNL_ATTRIB_TEX3]; - if (DO_TEX2 && VB->TexCoordPtr[1] == 0) - VB->TexCoordPtr[1] = VB->TexCoordPtr[2]; + if (DO_TEX2 && VB->AttribPtr[_TNL_ATTRIB_TEX1] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX1] = VB->AttribPtr[_TNL_ATTRIB_TEX2]; - if (DO_TEX1 && VB->TexCoordPtr[0] == 0) - VB->TexCoordPtr[0] = VB->TexCoordPtr[1]; + if (DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0] = VB->AttribPtr[_TNL_ATTRIB_TEX1]; if (DO_PTEX) return GL_TRUE; @@ -516,14 +516,14 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) /* No hardware support for projective texture. Can fake it for * TEX0 only. */ - if ((DO_TEX3 && VB->TexCoordPtr[GET_TEXSOURCE(3)]->size == 4) || - (DO_TEX2 && VB->TexCoordPtr[GET_TEXSOURCE(2)]->size == 4) || - (DO_TEX1 && VB->TexCoordPtr[GET_TEXSOURCE(1)]->size == 4)) { + if ((DO_TEX3 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(3)]->size == 4) || + (DO_TEX2 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(2)]->size == 4) || + (DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(1)]->size == 4)) { PTEX_FALLBACK(); return GL_FALSE; } - if (DO_TEX0 && VB->TexCoordPtr[GET_TEXSOURCE(0)]->size == 4) { + if (DO_TEX0 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(0)]->size == 4) { if (DO_TEX1 || DO_TEX2 || DO_TEX3) { PTEX_FALLBACK(); } diff --git a/src/mesa/drivers/dri/r128/r128_tris.c b/src/mesa/drivers/dri/r128/r128_tris.c index 5b91271d74..448e34e047 100644 --- a/src/mesa/drivers/dri/r128/r128_tris.c +++ b/src/mesa/drivers/dri/r128/r128_tris.c @@ -650,12 +650,12 @@ static void r128RenderStart( GLcontext *ctx ) } if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX(rmesa->tmu_source[0]) )) { - if ( VB->TexCoordPtr[rmesa->tmu_source[0]]->size > 2 ) + if ( VB->AttribPtr[_TNL_ATTRIB_TEX0 + rmesa->tmu_source[0]]->size > 2 ) fallback_projtex = GL_TRUE; EMIT_ATTR( _TNL_ATTRIB_TEX0, EMIT_2F, R128_CCE_VC_FRMT_S_T, 8 ); } if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX(rmesa->tmu_source[1]) )) { - if ( VB->TexCoordPtr[rmesa->tmu_source[1]]->size > 2 ) + if ( VB->AttribPtr[_TNL_ATTRIB_TEX0 + rmesa->tmu_source[1]]->size > 2 ) fallback_projtex = GL_TRUE; EMIT_ATTR( _TNL_ATTRIB_TEX1, EMIT_2F, R128_CCE_VC_FRMT_S2_T2, 8 ); } diff --git a/src/mesa/drivers/dri/r200/r200_maos_arrays.c b/src/mesa/drivers/dri/r200/r200_maos_arrays.c index 383a0c4b0d..249c0bbc11 100644 --- a/src/mesa/drivers/dri/r200/r200_maos_arrays.c +++ b/src/mesa/drivers/dri/r200/r200_maos_arrays.c @@ -90,12 +90,14 @@ static void r200_emit_vecfog(GLcontext *ctx, struct radeon_aos *aos, aos->components = size; aos->count = count; + radeon_bo_map(aos->bo, 1); out = (uint32_t*)((char*)aos->bo->ptr + aos->offset); for (i = 0; i < count; i++) { out[0] = r200ComputeFogBlendFactor( ctx, *(GLfloat *)data ); out++; data += stride; } + radeon_bo_unmap(aos->bo); } /* Emit any changed arrays to new GART memory, re-emit a packet to diff --git a/src/mesa/drivers/dri/r200/r200_state_init.c b/src/mesa/drivers/dri/r200/r200_state_init.c index 68bfeea701..e53fd72290 100644 --- a/src/mesa/drivers/dri/r200/r200_state_init.c +++ b/src/mesa/drivers/dri/r200/r200_state_init.c @@ -640,7 +640,7 @@ static void tex_emit(GLcontext *ctx, struct radeon_state_atom *atom) OUT_BATCH_TABLE(atom->cmd, 10); if (t && t->mt && !t->image_override) { - OUT_BATCH_RELOC(t->tile_bits, t->mt->bo, 0, + OUT_BATCH_RELOC(t->tile_bits, t->mt->bo, get_base_teximage_offset(t), RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0); } else if (!t) { /* workaround for old CS mechanism */ diff --git a/src/mesa/drivers/dri/r200/r200_swtcl.c b/src/mesa/drivers/dri/r200/r200_swtcl.c index 240fb45078..4596912ddc 100644 --- a/src/mesa/drivers/dri/r200/r200_swtcl.c +++ b/src/mesa/drivers/dri/r200/r200_swtcl.c @@ -168,7 +168,7 @@ static void r200SetVertexFormat( GLcontext *ctx ) for (i = 0; i < ctx->Const.MaxTextureUnits; i++) { if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX(i) )) { - GLuint sz = VB->TexCoordPtr[i]->size; + GLuint sz = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->size; fmt_1 |= sz << (3 * i); EMIT_ATTR( _TNL_ATTRIB_TEX0+i, EMIT_1F + sz - 1, 0 ); @@ -297,7 +297,7 @@ void r200_swtcl_flush(GLcontext *ctx, uint32_t current_offset) radeonEmitState(&rmesa->radeon); r200EmitVertexAOS( rmesa, rmesa->radeon.swtcl.vertex_size, - first_elem(&rmesa->radeon.dma.reserved)->bo, + rmesa->radeon.swtcl.bo, current_offset); diff --git a/src/mesa/drivers/dri/r200/r200_tcl.c b/src/mesa/drivers/dri/r200/r200_tcl.c index c702910ef2..e7d48a7f29 100644 --- a/src/mesa/drivers/dri/r200/r200_tcl.c +++ b/src/mesa/drivers/dri/r200/r200_tcl.c @@ -509,25 +509,26 @@ static GLboolean r200_run_tcl_render( GLcontext *ctx, prog to a not enabled output however, so just don't mess with it. We only need to change compsel. */ GLuint out_compsel = 0; - GLuint vp_out = rmesa->curr_vp_hw->mesa_program.Base.OutputsWritten; + const GLbitfield64 vp_out = + rmesa->curr_vp_hw->mesa_program.Base.OutputsWritten; vimap_rev = &rmesa->curr_vp_hw->inputmap_rev[0]; - assert(vp_out & (1 << VERT_RESULT_HPOS)); + assert(vp_out & BITFIELD64_BIT(VERT_RESULT_HPOS)); out_compsel = R200_OUTPUT_XYZW; - if (vp_out & (1 << VERT_RESULT_COL0)) { + if (vp_out & BITFIELD64_BIT(VERT_RESULT_COL0)) { out_compsel |= R200_OUTPUT_COLOR_0; } - if (vp_out & (1 << VERT_RESULT_COL1)) { + if (vp_out & BITFIELD64_BIT(VERT_RESULT_COL1)) { out_compsel |= R200_OUTPUT_COLOR_1; } - if (vp_out & (1 << VERT_RESULT_FOGC)) { + if (vp_out & BITFIELD64_BIT(VERT_RESULT_FOGC)) { out_compsel |= R200_OUTPUT_DISCRETE_FOG; } - if (vp_out & (1 << VERT_RESULT_PSIZ)) { + if (vp_out & BITFIELD64_BIT(VERT_RESULT_PSIZ)) { out_compsel |= R200_OUTPUT_PT_SIZE; } for (i = VERT_RESULT_TEX0; i < VERT_RESULT_TEX6; i++) { - if (vp_out & (1 << i)) { + if (vp_out & BITFIELD64_BIT(i)) { out_compsel |= R200_OUTPUT_TEX_0 << (i - VERT_RESULT_TEX0); } } diff --git a/src/mesa/drivers/dri/r200/r200_tex.c b/src/mesa/drivers/dri/r200/r200_tex.c index 5a21a8b9c5..a417721553 100644 --- a/src/mesa/drivers/dri/r200/r200_tex.c +++ b/src/mesa/drivers/dri/r200/r200_tex.c @@ -385,16 +385,7 @@ static void r200TexParameter( GLcontext *ctx, GLenum target, case GL_TEXTURE_MAX_LEVEL: case GL_TEXTURE_MIN_LOD: case GL_TEXTURE_MAX_LOD: - /* This isn't the most efficient solution but there doesn't appear to - * be a nice alternative. Since there's no LOD clamping, - * we just have to rely on loading the right subset of mipmap levels - * to simulate a clamped LOD. - */ - if (t->mt) { - radeon_miptree_unreference(t->mt); - t->mt = 0; - t->validated = GL_FALSE; - } + t->validated = GL_FALSE; break; default: @@ -413,7 +404,7 @@ static void r200DeleteTexture(GLcontext * ctx, struct gl_texture_object *texObj) (void *)texObj, _mesa_lookup_enum_by_nr(texObj->Target)); } - + if (rmesa) { int i; radeon_firevertices(&rmesa->radeon); @@ -425,11 +416,9 @@ static void r200DeleteTexture(GLcontext * ctx, struct gl_texture_object *texObj) } } } - - if (t->mt) { - radeon_miptree_unreference(t->mt); - t->mt = 0; - } + + radeon_miptree_unreference(&t->mt); + _mesa_delete_texture_object(ctx, texObj); } diff --git a/src/mesa/drivers/dri/r200/r200_texstate.c b/src/mesa/drivers/dri/r200/r200_texstate.c index 7d0afa1add..7782404a79 100644 --- a/src/mesa/drivers/dri/r200/r200_texstate.c +++ b/src/mesa/drivers/dri/r200/r200_texstate.c @@ -824,14 +824,10 @@ void r200SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_fo radeon_bo_unref(rImage->bo); rImage->bo = NULL; } - if (t->mt) { - radeon_miptree_unreference(t->mt); - t->mt = NULL; - } - if (rImage->mt) { - radeon_miptree_unreference(rImage->mt); - rImage->mt = NULL; - } + + radeon_miptree_unreference(&t->mt); + radeon_miptree_unreference(&rImage->mt); + _mesa_init_teximage_fields(radeon->glCtx, target, texImage, rb->base.Width, rb->base.Height, 1, 0, rb->cpp); texImage->RowStride = rb->pitch / rb->cpp; @@ -1423,10 +1419,9 @@ void set_re_cntl_d3d( GLcontext *ctx, int unit, GLboolean use_d3d ) */ static void setup_hardware_state(r200ContextPtr rmesa, radeonTexObj *t) { - int firstlevel = t->mt ? t->mt->firstLevel : 0; - const struct gl_texture_image *firstImage = t->base.Image[0][firstlevel]; + const struct gl_texture_image *firstImage = t->base.Image[0][t->minLod]; GLint log2Width, log2Height, log2Depth, texelBytes; - + if ( t->bo ) { return; } @@ -1454,9 +1449,9 @@ static void setup_hardware_state(r200ContextPtr rmesa, radeonTexObj *t) return; } } - + t->pp_txfilter &= ~R200_MAX_MIP_LEVEL_MASK; - t->pp_txfilter |= (t->mt->lastLevel - t->mt->firstLevel) << R200_MAX_MIP_LEVEL_SHIFT; + t->pp_txfilter |= (t->maxLod - t->minLod) << R200_MAX_MIP_LEVEL_SHIFT; t->pp_txformat &= ~(R200_TXFORMAT_WIDTH_MASK | R200_TXFORMAT_HEIGHT_MASK | diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_program.h b/src/mesa/drivers/dri/r300/compiler/radeon_program.h index 33db3ea0ff..03592884eb 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_program.h +++ b/src/mesa/drivers/dri/r300/compiler/radeon_program.h @@ -191,7 +191,7 @@ struct rc_src_register lmul_swizzle(unsigned int swizzle, struct rc_src_register static inline void reset_srcreg(struct rc_src_register* reg) { - memset(reg, 0, sizeof(reg)); + memset(reg, 0, sizeof(struct rc_src_register)); reg->Swizzle = RC_SWIZZLE_XYZW; } diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c b/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c index 0326d25233..ced66af1eb 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c @@ -560,23 +560,23 @@ static void sincos_constants(struct radeon_compiler* c, unsigned int *constants) * MAD dest, tmp.y, weight, tmp.x */ static void sin_approx( - struct radeon_compiler* c, struct rc_instruction * before, + struct radeon_compiler* c, struct rc_instruction * inst, struct rc_dst_register dst, struct rc_src_register src, const unsigned int* constants) { unsigned int tempreg = rc_find_free_temporary(c); - emit2(c, before, RC_OPCODE_MUL, 0, dstregtmpmask(tempreg, RC_MASK_XY), + emit2(c, inst->Prev, RC_OPCODE_MUL, 0, dstregtmpmask(tempreg, RC_MASK_XY), swizzle(src, RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X), srcreg(RC_FILE_CONSTANT, constants[0])); - emit3(c, before, RC_OPCODE_MAD, 0, dstregtmpmask(tempreg, RC_MASK_X), + emit3(c, inst->Prev, RC_OPCODE_MAD, 0, dstregtmpmask(tempreg, RC_MASK_X), swizzle(srcreg(RC_FILE_TEMPORARY, tempreg), RC_SWIZZLE_Y, RC_SWIZZLE_Y, RC_SWIZZLE_Y, RC_SWIZZLE_Y), absolute(swizzle(src, RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X)), swizzle(srcreg(RC_FILE_TEMPORARY, tempreg), RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X)); - emit3(c, before, RC_OPCODE_MAD, 0, dstregtmpmask(tempreg, RC_MASK_Y), + emit3(c, inst->Prev, RC_OPCODE_MAD, 0, dstregtmpmask(tempreg, RC_MASK_Y), swizzle(srcreg(RC_FILE_TEMPORARY, tempreg), RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X), absolute(swizzle(srcreg(RC_FILE_TEMPORARY, tempreg), RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X)), negate(swizzle(srcreg(RC_FILE_TEMPORARY, tempreg), RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X))); - emit3(c, before, RC_OPCODE_MAD, 0, dst, + emit3(c, inst->Prev, RC_OPCODE_MAD, 0, dst, swizzle(srcreg(RC_FILE_TEMPORARY, tempreg), RC_SWIZZLE_Y, RC_SWIZZLE_Y, RC_SWIZZLE_Y, RC_SWIZZLE_Y), swizzle(srcreg(RC_FILE_CONSTANT, constants[0]), RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W), swizzle(srcreg(RC_FILE_TEMPORARY, tempreg), RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X)); diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.c b/src/mesa/drivers/dri/r300/r300_cmdbuf.c index 1e2a54f634..ad8db6e68e 100644 --- a/src/mesa/drivers/dri/r300/r300_cmdbuf.c +++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.c @@ -46,14 +46,12 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r300_context.h" #include "r300_ioctl.h" -#include "radeon_reg.h" #include "r300_reg.h" #include "r300_cmdbuf.h" #include "r300_emit.h" #include "radeon_bocs_wrapper.h" #include "radeon_mipmap_tree.h" #include "r300_state.h" -#include "radeon_reg.h" #include "radeon_queryobj.h" /** # of dwords reserved for additional instructions that may need to be written @@ -171,7 +169,7 @@ static void emit_tex_offsets(GLcontext *ctx, struct radeon_state_atom * atom) if (t && !t->image_override) { BEGIN_BATCH_NO_AUTOSTATE(4); OUT_BATCH_REGSEQ(R300_TX_OFFSET_0 + (i * 4), 1); - OUT_BATCH_RELOC(t->tile_bits, t->mt->bo, 0, + OUT_BATCH_RELOC(t->tile_bits, t->mt->bo, get_base_teximage_offset(t), RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0); END_BATCH(); } else if (!t) { @@ -281,12 +279,27 @@ static void emit_cb_offset(GLcontext *ctx, struct radeon_state_atom * atom) cbpitch |= R300_COLOR_FORMAT_ARGB8888; else switch (rrb->base.Format) { case MESA_FORMAT_RGB565: + assert(_mesa_little_endian()); + cbpitch |= R300_COLOR_FORMAT_RGB565; + break; + case MESA_FORMAT_RGB565_REV: + assert(!_mesa_little_endian()); cbpitch |= R300_COLOR_FORMAT_RGB565; break; case MESA_FORMAT_ARGB4444: + assert(_mesa_little_endian()); + cbpitch |= R300_COLOR_FORMAT_ARGB4444; + break; + case MESA_FORMAT_ARGB4444_REV: + assert(!_mesa_little_endian()); cbpitch |= R300_COLOR_FORMAT_ARGB4444; break; case MESA_FORMAT_ARGB1555: + assert(_mesa_little_endian()); + cbpitch |= R300_COLOR_FORMAT_ARGB1555; + break; + case MESA_FORMAT_ARGB1555_REV: + assert(!_mesa_little_endian()); cbpitch |= R300_COLOR_FORMAT_ARGB1555; break; default: diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c index 6f66e970e4..5f07b95634 100644 --- a/src/mesa/drivers/dri/r300/r300_context.c +++ b/src/mesa/drivers/dri/r300/r300_context.c @@ -439,11 +439,11 @@ static void r300InitGLExtensions(GLcontext *ctx) if (r300->options.stencil_two_side_disabled) _mesa_disable_extension(ctx, "GL_EXT_stencil_two_side"); - if (r300->options.s3tc_force_enabled) { + if (r300->options.s3tc_force_disabled) { + _mesa_disable_extension(ctx, "GL_EXT_texture_compression_s3tc"); + } else if (ctx->Mesa_DXTn || r300->options.s3tc_force_enabled) { _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc"); _mesa_enable_extension(ctx, "GL_S3_s3tc"); - } else if (r300->options.s3tc_force_disabled) { - _mesa_disable_extension(ctx, "GL_EXT_texture_compression_s3tc"); } if (!r300->radeon.radeonScreen->drmSupportsOcclusionQueries) { diff --git a/src/mesa/drivers/dri/r300/r300_draw.c b/src/mesa/drivers/dri/r300/r300_draw.c index e9968f9ffe..3dcd986e22 100644 --- a/src/mesa/drivers/dri/r300/r300_draw.c +++ b/src/mesa/drivers/dri/r300/r300_draw.c @@ -100,7 +100,7 @@ static void r300FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer GLubyte *in = (GLubyte *)src_ptr; radeonAllocDmaRegion(&r300->radeon, &r300->ind_buf.bo, &r300->ind_buf.bo_offset, size, 4); - + radeon_bo_map(r300->ind_buf.bo, 1); assert(r300->ind_buf.bo->ptr != NULL); out = (GLuint *)ADD_POINTERS(r300->ind_buf.bo->ptr, r300->ind_buf.bo_offset); @@ -111,7 +111,7 @@ static void r300FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer if (i < mesa_ind_buf->count) { *out++ = in[i]; } - + radeon_bo_unmap(r300->ind_buf.bo); #if MESA_BIG_ENDIAN } else { /* if (mesa_ind_buf->type == GL_UNSIGNED_SHORT) */ GLushort *in = (GLushort *)src_ptr; @@ -120,6 +120,7 @@ static void r300FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer radeonAllocDmaRegion(&r300->radeon, &r300->ind_buf.bo, &r300->ind_buf.bo_offset, size, 4); + radeon_bo_map(r300->ind_buf.bo, 1); assert(r300->ind_buf.bo->ptr != NULL); out = (GLuint *)ADD_POINTERS(r300->ind_buf.bo->ptr, r300->ind_buf.bo_offset); @@ -130,6 +131,7 @@ static void r300FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer if (i < mesa_ind_buf->count) { *out++ = in[i]; } + radeon_bo_unmap(r300->ind_buf.bo); #endif } @@ -173,10 +175,12 @@ static void r300SetupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer radeonAllocDmaRegion(&r300->radeon, &r300->ind_buf.bo, &r300->ind_buf.bo_offset, size, 4); + radeon_bo_map(r300->ind_buf.bo, 1); assert(r300->ind_buf.bo->ptr != NULL); dst_ptr = ADD_POINTERS(r300->ind_buf.bo->ptr, r300->ind_buf.bo_offset); _mesa_memcpy(dst_ptr, src_ptr, size); + radeon_bo_unmap(r300->ind_buf.bo); r300->ind_buf.is_32bit = (mesa_ind_buf->type == GL_UNSIGNED_INT); r300->ind_buf.count = mesa_ind_buf->count; @@ -242,6 +246,7 @@ static void r300ConvertAttrib(GLcontext *ctx, int count, const struct gl_client_ } radeonAllocDmaRegion(&r300->radeon, &attr->bo, &attr->bo_offset, sizeof(GLfloat) * input->Size * count, 32); + radeon_bo_map(attr->bo, 1); dst_ptr = (GLfloat *)ADD_POINTERS(attr->bo->ptr, attr->bo_offset); radeon_print(RADEON_FALLBACKS, RADEON_IMPORTANT, @@ -280,6 +285,7 @@ static void r300ConvertAttrib(GLcontext *ctx, int count, const struct gl_client_ break; } + radeon_bo_unmap(attr->bo); if (mapped_named_bo) { ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER, input->BufferObj); } @@ -294,6 +300,8 @@ static void r300AlignDataToDword(GLcontext *ctx, const struct gl_client_array *i radeonAllocDmaRegion(&r300->radeon, &attr->bo, &attr->bo_offset, size, 32); + radeon_bo_map(attr->bo, 1); + if (!input->BufferObj->Pointer) { ctx->Driver.MapBuffer(ctx, GL_ARRAY_BUFFER, GL_READ_ONLY_ARB, input->BufferObj); mapped_named_bo = GL_TRUE; @@ -317,6 +325,7 @@ static void r300AlignDataToDword(GLcontext *ctx, const struct gl_client_array *i ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER, input->BufferObj); } + radeon_bo_unmap(attr->bo); attr->stride = dst_stride; } @@ -527,6 +536,7 @@ static void r300AllocDmaRegions(GLcontext *ctx, const struct gl_client_array *in } radeonAllocDmaRegion(&r300->radeon, &vbuf->attribs[index].bo, &vbuf->attribs[index].bo_offset, size, 32); + radeon_bo_map(vbuf->attribs[index].bo, 1); assert(vbuf->attribs[index].bo->ptr != NULL); dst = (uint32_t *)ADD_POINTERS(vbuf->attribs[index].bo->ptr, vbuf->attribs[index].bo_offset); switch (vbuf->attribs[index].dwords) { @@ -536,6 +546,7 @@ static void r300AllocDmaRegions(GLcontext *ctx, const struct gl_client_array *in case 4: radeonEmitVec16(dst, input[i]->Ptr, input[i]->StrideB, local_count); break; default: assert(0); break; } + radeon_bo_unmap(vbuf->attribs[index].bo); } } diff --git a/src/mesa/drivers/dri/r300/r300_emit.h b/src/mesa/drivers/dri/r300/r300_emit.h index 8e57e354d1..a456d8867c 100644 --- a/src/mesa/drivers/dri/r300/r300_emit.h +++ b/src/mesa/drivers/dri/r300/r300_emit.h @@ -42,7 +42,6 @@ #include "main/glheader.h" #include "r300_context.h" #include "r300_cmdbuf.h" -#include "radeon_reg.h" static INLINE uint32_t cmdpacket0(struct radeon_screen *rscrn, int reg, int count) diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c index 3cd38753b8..4ae593cbe7 100644 --- a/src/mesa/drivers/dri/r300/r300_render.c +++ b/src/mesa/drivers/dri/r300/r300_render.c @@ -67,8 +67,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #include "vbo/vbo_split.h" #include "tnl/tnl.h" #include "tnl/t_vp_build.h" -#include "radeon_reg.h" -#include "radeon_macros.h" #include "r300_context.h" #include "r300_ioctl.h" #include "r300_state.h" diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c index ac20c08e20..1fd32d497b 100644 --- a/src/mesa/drivers/dri/r300/r300_state.c +++ b/src/mesa/drivers/dri/r300/r300_state.c @@ -1741,7 +1741,8 @@ static void r300Enable(GLcontext * ctx, GLenum cap, GLboolean state) r300SetPolygonOffsetState(ctx, state); break; case GL_SCISSOR_TEST: - radeon_firevertices(&rmesa->radeon); + if (!rmesa->radeon.radeonScreen->kernel_mm) + radeon_firevertices(&rmesa->radeon); rmesa->radeon.state.scissor.enabled = state; radeonUpdateScissor( ctx ); break; diff --git a/src/mesa/drivers/dri/r300/r300_swtcl.c b/src/mesa/drivers/dri/r300/r300_swtcl.c index ee2c71e1a7..383c8a274b 100644 --- a/src/mesa/drivers/dri/r300/r300_swtcl.c +++ b/src/mesa/drivers/dri/r300/r300_swtcl.c @@ -124,7 +124,7 @@ void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead, GLuint *_ } if (ctx->Light.Enabled && ctx->Light.Model.TwoSide) { - VB->AttribPtr[VERT_ATTRIB_GENERIC0] = VB->ColorPtr[1]; + VB->AttribPtr[VERT_ATTRIB_GENERIC0] = VB->BackfaceColorPtr; OutputsWritten |= 1 << VERT_RESULT_BFC0; #if MESA_LITTLE_ENDIAN EMIT_ATTR( _TNL_ATTRIB_GENERIC0, EMIT_4UB_4F_RGBA ); @@ -134,7 +134,7 @@ void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead, GLuint *_ ADD_ATTR(VERT_ATTRIB_GENERIC0, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR2, SWIZZLE_XYZW, MASK_XYZW, 1); #endif if (fp_reads & FRAG_BIT_COL1) { - VB->AttribPtr[VERT_ATTRIB_GENERIC1] = VB->SecondaryColorPtr[1]; + VB->AttribPtr[VERT_ATTRIB_GENERIC1] = VB->BackfaceSecondaryColorPtr; GLuint swiz = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ONE); OutputsWritten |= 1 << VERT_RESULT_BFC1; #if MESA_LITTLE_ENDIAN @@ -159,7 +159,7 @@ void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead, GLuint *_ int tex_id = rmesa->selected_fp->wpos_attr - FRAG_ATTRIB_TEX0; VB->AttribPtr[VERT_ATTRIB_TEX0 + tex_id] = VB->AttribPtr[VERT_ATTRIB_POS]; - VB->TexCoordPtr[tex_id] = VB->AttribPtr[VERT_ATTRIB_POS]; + VB->AttribPtr[_TNL_ATTRIB_TEX0 + tex_id] = VB->AttribPtr[VERT_ATTRIB_POS]; RENDERINPUTS_SET(tnl->render_inputs_bitset, _TNL_ATTRIB_TEX0 + tex_id); } @@ -167,7 +167,7 @@ void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead, GLuint *_ int tex_id = rmesa->selected_fp->fog_attr - FRAG_ATTRIB_TEX0; VB->AttribPtr[VERT_ATTRIB_TEX0 + tex_id] = VB->AttribPtr[VERT_ATTRIB_FOG]; - VB->TexCoordPtr[tex_id] = VB->AttribPtr[VERT_ATTRIB_FOG]; + VB->AttribPtr[_TNL_ATTRIB_TEX0 + tex_id] = VB->AttribPtr[VERT_ATTRIB_FOG]; RENDERINPUTS_SET(tnl->render_inputs_bitset, _TNL_ATTRIB_TEX0 + tex_id); } @@ -180,7 +180,7 @@ void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead, GLuint *_ GLuint swiz, format, hw_format; for (i = 0; i < ctx->Const.MaxTextureUnits; i++) { if (fp_reads & FRAG_BIT_TEX(i)) { - switch (VB->TexCoordPtr[i]->size) { + switch (VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->size) { case 1: format = EMIT_1F; hw_format = R300_DATA_TYPE_FLOAT_1; @@ -665,11 +665,11 @@ void r300_swtcl_flush(GLcontext *ctx, uint32_t current_offset) r300EmitCacheFlush(rmesa); radeonEmitState(&rmesa->radeon); - r300_emit_scissor(ctx); + r300_emit_scissor(ctx); r300EmitVertexAOS(rmesa, - rmesa->radeon.swtcl.vertex_size, - first_elem(&rmesa->radeon.dma.reserved)->bo, - current_offset); + rmesa->radeon.swtcl.vertex_size, + rmesa->radeon.swtcl.bo, + current_offset); r300EmitVbufPrim(rmesa, rmesa->radeon.swtcl.hw_primitive, diff --git a/src/mesa/drivers/dri/r300/r300_tex.c b/src/mesa/drivers/dri/r300/r300_tex.c index 27b78a912f..726b3ff98e 100644 --- a/src/mesa/drivers/dri/r300/r300_tex.c +++ b/src/mesa/drivers/dri/r300/r300_tex.c @@ -223,16 +223,7 @@ static void r300TexParameter(GLcontext * ctx, GLenum target, case GL_TEXTURE_MAX_LEVEL: case GL_TEXTURE_MIN_LOD: case GL_TEXTURE_MAX_LOD: - /* This isn't the most efficient solution but there doesn't appear to - * be a nice alternative. Since there's no LOD clamping, - * we just have to rely on loading the right subset of mipmap levels - * to simulate a clamped LOD. - */ - if (t->mt) { - radeon_miptree_unreference(t->mt); - t->mt = 0; - t->validated = GL_FALSE; - } + t->validated = GL_FALSE; break; case GL_DEPTH_TEXTURE_MODE: @@ -270,7 +261,11 @@ static void r300DeleteTexture(GLcontext * ctx, struct gl_texture_object *texObj) if (rmesa) { int i; - radeon_firevertices(&rmesa->radeon); + struct radeon_bo *bo; + bo = !t->mt ? t->bo : t->mt->bo; + if (bo && radeon_bo_is_referenced_by_cs(bo, rmesa->radeon.cmdbuf.cs)) { + radeon_firevertices(&rmesa->radeon); + } for(i = 0; i < R300_MAX_TEXTURE_UNITS; ++i) if (rmesa->hw.textures[i] == t) @@ -282,10 +277,8 @@ static void r300DeleteTexture(GLcontext * ctx, struct gl_texture_object *texObj) t->bo = NULL; } - if (t->mt) { - radeon_miptree_unreference(t->mt); - t->mt = 0; - } + radeon_miptree_unreference(&t->mt); + _mesa_delete_texture_object(ctx, texObj); } diff --git a/src/mesa/drivers/dri/r300/r300_texstate.c b/src/mesa/drivers/dri/r300/r300_texstate.c index 44ca24daf8..e6f2c0c1a7 100644 --- a/src/mesa/drivers/dri/r300/r300_texstate.c +++ b/src/mesa/drivers/dri/r300/r300_texstate.c @@ -83,6 +83,7 @@ static const struct tx_table { _ASSIGN(ARGB8888, R300_EASY_TX_FORMAT(W, Z, Y, X, W8Z8Y8X8)), _ASSIGN(ARGB8888_REV, R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8)), #endif + _ASSIGN(XRGB8888, R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8)), _ASSIGN(RGB888, R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8)), _ASSIGN(RGB565, R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z5Y6X5)), _ASSIGN(RGB565_REV, R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z5Y6X5)), @@ -202,9 +203,7 @@ void r300SetDepthTexMode(struct gl_texture_object *tObj) static void setup_hardware_state(r300ContextPtr rmesa, radeonTexObj *t) { const struct gl_texture_image *firstImage; - int firstlevel = t->mt ? t->mt->firstLevel : 0; - - firstImage = t->base.Image[0][firstlevel]; + firstImage = t->base.Image[0][t->minLod]; if (!t->image_override && VALID_FORMAT(firstImage->TexFormat)) { @@ -227,7 +226,7 @@ static void setup_hardware_state(r300ContextPtr rmesa, radeonTexObj *t) t->pp_txsize = (((R300_TX_WIDTHMASK_MASK & ((firstImage->Width - 1) << R300_TX_WIDTHMASK_SHIFT))) | ((R300_TX_HEIGHTMASK_MASK & ((firstImage->Height - 1) << R300_TX_HEIGHTMASK_SHIFT))) | ((R300_TX_DEPTHMASK_MASK & ((firstImage->DepthLog2) << R300_TX_DEPTHMASK_SHIFT))) - | ((R300_TX_MAX_MIP_LEVEL_MASK & ((t->mt->lastLevel - t->mt->firstLevel) << R300_TX_MAX_MIP_LEVEL_SHIFT)))); + | ((R300_TX_MAX_MIP_LEVEL_MASK & ((t->maxLod - t->minLod) << R300_TX_MAX_MIP_LEVEL_SHIFT)))); t->tile_bits = 0; @@ -238,7 +237,7 @@ static void setup_hardware_state(r300ContextPtr rmesa, radeonTexObj *t) if (t->base.Target == GL_TEXTURE_RECTANGLE_NV) { - unsigned int align = (64 / t->mt->bpp) - 1; + unsigned int align = (64 / _mesa_get_format_bytes(firstImage->TexFormat)) - 1; t->pp_txsize |= R300_TX_SIZE_TXPITCH_EN; if (!t->image_override) t->pp_txpitch = ((firstImage->Width + align) & ~align) - 1; @@ -437,14 +436,10 @@ void r300SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_fo radeon_bo_unref(rImage->bo); rImage->bo = NULL; } - if (t->mt) { - radeon_miptree_unreference(t->mt); - t->mt = NULL; - } - if (rImage->mt) { - radeon_miptree_unreference(rImage->mt); - rImage->mt = NULL; - } + + radeon_miptree_unreference(&t->mt); + radeon_miptree_unreference(&rImage->mt); + _mesa_init_teximage_fields(radeon->glCtx, target, texImage, rb->base.Width, rb->base.Height, 1, 0, rb->cpp); texImage->RowStride = rb->pitch / rb->cpp; diff --git a/src/mesa/drivers/dri/r600/r600_context.c b/src/mesa/drivers/dri/r600/r600_context.c index dbd233729c..25314eff56 100644 --- a/src/mesa/drivers/dri/r600/r600_context.c +++ b/src/mesa/drivers/dri/r600/r600_context.c @@ -74,6 +74,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "utils.h" #include "xmlpool.h" /* for symbolic values of enum-type options */ +//#define R600_ENABLE_GLSL_TEST 1 + #define need_GL_VERSION_2_0 #define need_GL_ARB_occlusion_query #define need_GL_ARB_point_parameters @@ -109,6 +111,7 @@ static const struct dri_extension card_extensions[] = { {"GL_ARB_texture_env_crossbar", NULL}, {"GL_ARB_texture_env_dot3", NULL}, {"GL_ARB_texture_mirrored_repeat", NULL}, + {"GL_ARB_texture_non_power_of_two", NULL}, {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions}, {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions}, {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions}, @@ -155,7 +158,11 @@ static const struct dri_extension mm_extensions[] = { * functions added by GL_ATI_separate_stencil. */ static const struct dri_extension gl_20_extension[] = { +#ifdef R600_ENABLE_GLSL_TEST + {"GL_ARB_shading_language_100", GL_VERSION_2_0_functions }, +#else {"GL_VERSION_2_0", GL_VERSION_2_0_functions }, +#endif /* R600_ENABLE_GLSL_TEST */ }; static const struct tnl_pipeline_stage *r600_pipeline[] = { @@ -308,6 +315,26 @@ static void r600InitGLExtensions(GLcontext *ctx) if (r600->radeon.radeonScreen->kernel_mm) driInitExtensions(ctx, mm_extensions, GL_FALSE); +#ifdef R600_ENABLE_GLSL_TEST + driInitExtensions(ctx, gl_20_extension, GL_TRUE); + //_mesa_enable_2_0_extensions(ctx); + //1.5 + ctx->Extensions.ARB_occlusion_query = GL_TRUE; + ctx->Extensions.ARB_vertex_buffer_object = GL_TRUE; + ctx->Extensions.EXT_shadow_funcs = GL_TRUE; + //2.0 + ctx->Extensions.ARB_draw_buffers = GL_TRUE; + ctx->Extensions.ARB_point_sprite = GL_TRUE; + ctx->Extensions.ARB_shader_objects = GL_TRUE; + ctx->Extensions.ARB_vertex_shader = GL_TRUE; + ctx->Extensions.ARB_fragment_shader = GL_TRUE; + ctx->Extensions.EXT_blend_equation_separate = GL_TRUE; + ctx->Extensions.ATI_separate_stencil = GL_TRUE; + + /* glsl compiler has problem if this is not GL_TRUE */ + ctx->Shader.EmitCondCodes = GL_TRUE; +#endif /* R600_ENABLE_GLSL_TEST */ + if (driQueryOptionb (&r600->radeon.optionCache, "disable_stencil_two_side")) _mesa_disable_extension(ctx, "GL_EXT_stencil_two_side"); diff --git a/src/mesa/drivers/dri/r600/r600_tex.c b/src/mesa/drivers/dri/r600/r600_tex.c index 20965bb3c8..9d83a64e22 100644 --- a/src/mesa/drivers/dri/r600/r600_tex.c +++ b/src/mesa/drivers/dri/r600/r600_tex.c @@ -312,16 +312,7 @@ static void r600TexParameter(GLcontext * ctx, GLenum target, case GL_TEXTURE_MAX_LEVEL: case GL_TEXTURE_MIN_LOD: case GL_TEXTURE_MAX_LOD: - /* This isn't the most efficient solution but there doesn't appear to - * be a nice alternative. Since there's no LOD clamping, - * we just have to rely on loading the right subset of mipmap levels - * to simulate a clamped LOD. - */ - if (t->mt) { - radeon_miptree_unreference(t->mt); - t->mt = 0; - t->validated = GL_FALSE; - } + t->validated = GL_FALSE; break; case GL_DEPTH_TEXTURE_MODE: @@ -369,10 +360,8 @@ static void r600DeleteTexture(GLcontext * ctx, struct gl_texture_object *texObj) t->bo = NULL; } - if (t->mt) { - radeon_miptree_unreference(t->mt); - t->mt = 0; - } + radeon_miptree_unreference(&t->mt); + _mesa_delete_texture_object(ctx, texObj); } diff --git a/src/mesa/drivers/dri/r600/r600_texstate.c b/src/mesa/drivers/dri/r600/r600_texstate.c index 27c8354923..4ec315b78c 100644 --- a/src/mesa/drivers/dri/r600/r600_texstate.c +++ b/src/mesa/drivers/dri/r600/r600_texstate.c @@ -649,7 +649,6 @@ static void setup_hardware_state(context_t *rmesa, struct gl_texture_object *tex { radeonTexObj *t = radeon_tex_obj(texObj); const struct gl_texture_image *firstImage; - int firstlevel = t->mt ? t->mt->firstLevel : 0; GLuint uTexelPitch, row_align; if (rmesa->radeon.radeonScreen->driScreen->dri2.enabled && @@ -657,7 +656,7 @@ static void setup_hardware_state(context_t *rmesa, struct gl_texture_object *tex t->bo) return; - firstImage = t->base.Image[0][firstlevel]; + firstImage = t->base.Image[0][t->minLod]; if (!t->image_override) { if (!r600GetTexFormat(texObj, firstImage->TexFormat)) { @@ -692,7 +691,8 @@ static void setup_hardware_state(context_t *rmesa, struct gl_texture_object *tex } row_align = rmesa->radeon.texture_row_align - 1; - uTexelPitch = ((firstImage->Width * t->mt->bpp + row_align) & ~row_align) / t->mt->bpp; + uTexelPitch = (_mesa_format_row_stride(firstImage->TexFormat, firstImage->Width) + row_align) & ~row_align; + uTexelPitch = uTexelPitch / _mesa_get_format_bytes(firstImage->TexFormat); uTexelPitch = (uTexelPitch + R700_TEXEL_PITCH_ALIGNMENT_MASK) & ~R700_TEXEL_PITCH_ALIGNMENT_MASK; @@ -706,10 +706,10 @@ static void setup_hardware_state(context_t *rmesa, struct gl_texture_object *tex SETfield(t->SQ_TEX_RESOURCE1, firstImage->Height - 1, TEX_HEIGHT_shift, TEX_HEIGHT_mask); - if ((t->mt->lastLevel - t->mt->firstLevel) > 0) { - t->SQ_TEX_RESOURCE3 = t->mt->levels[0].size / 256; - SETfield(t->SQ_TEX_RESOURCE4, t->mt->firstLevel, BASE_LEVEL_shift, BASE_LEVEL_mask); - SETfield(t->SQ_TEX_RESOURCE5, t->mt->lastLevel, LAST_LEVEL_shift, LAST_LEVEL_mask); + if ((t->maxLod - t->minLod) > 0) { + t->SQ_TEX_RESOURCE3 = t->mt->levels[t->minLod].size / 256; + SETfield(t->SQ_TEX_RESOURCE4, 0, BASE_LEVEL_shift, BASE_LEVEL_mask); + SETfield(t->SQ_TEX_RESOURCE5, t->maxLod - t->minLod, LAST_LEVEL_shift, LAST_LEVEL_mask); } } @@ -808,9 +808,8 @@ void r600SetTexOffset(__DRIcontext * pDRICtx, GLint texname, struct gl_texture_object *tObj = _mesa_lookup_texture(rmesa->radeon.glCtx, texname); radeonTexObjPtr t = radeon_tex_obj(tObj); - int firstlevel = t->mt ? t->mt->firstLevel : 0; const struct gl_texture_image *firstImage; - uint32_t pitch_val, size, row_align, bpp; + uint32_t pitch_val, size, row_align; if (!tObj) return; @@ -820,13 +819,9 @@ void r600SetTexOffset(__DRIcontext * pDRICtx, GLint texname, if (!offset) return; - bpp = depth / 8; - if (bpp == 3) - bpp = 4; - - firstImage = t->base.Image[0][firstlevel]; + firstImage = t->base.Image[0][t->minLod]; row_align = rmesa->radeon.texture_row_align - 1; - size = ((firstImage->Width * bpp + row_align) & ~row_align) * firstImage->Height; + size = ((_mesa_format_row_stride(firstImage->TexFormat, firstImage->Width) + row_align) & ~row_align) * firstImage->Height; if (t->bo) { radeon_bo_unref(t->bo); t->bo = NULL; @@ -949,14 +944,10 @@ void r600SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_fo radeon_bo_unref(rImage->bo); rImage->bo = NULL; } - if (t->mt) { - radeon_miptree_unreference(t->mt); - t->mt = NULL; - } - if (rImage->mt) { - radeon_miptree_unreference(rImage->mt); - rImage->mt = NULL; - } + + radeon_miptree_unreference(&t->mt); + radeon_miptree_unreference(&rImage->mt); + _mesa_init_teximage_fields(radeon->glCtx, target, texImage, rb->base.Width, rb->base.Height, 1, 0, rb->cpp); texImage->RowStride = rb->pitch / rb->cpp; diff --git a/src/mesa/drivers/dri/r600/r700_assembler.c b/src/mesa/drivers/dri/r600/r700_assembler.c index e0d7d4fa6b..ba97d3e073 100644 --- a/src/mesa/drivers/dri/r600/r700_assembler.c +++ b/src/mesa/drivers/dri/r600/r700_assembler.c @@ -38,6 +38,9 @@ #include "r700_assembler.h" +#define USE_CF_FOR_CONTINUE_BREAK 1 +#define USE_CF_FOR_POP_AFTER 1 + BITS addrmode_PVSDST(PVSDST * pPVSDST) { return pPVSDST->addrmode0 | ((BITS)pPVSDST->addrmode1 << 1); @@ -337,12 +340,17 @@ unsigned int r700GetNumOperands(r700_AssemblerBase* pAsm) switch (pAsm->D.dst.opcode) { case SQ_OP2_INST_ADD: + case SQ_OP2_INST_KILLE: case SQ_OP2_INST_KILLGT: + case SQ_OP2_INST_KILLGE: + case SQ_OP2_INST_KILLNE: case SQ_OP2_INST_MUL: case SQ_OP2_INST_MAX: case SQ_OP2_INST_MIN: //case SQ_OP2_INST_MAX_DX10: //case SQ_OP2_INST_MIN_DX10: + case SQ_OP2_INST_SETE: + case SQ_OP2_INST_SETNE: case SQ_OP2_INST_SETGT: case SQ_OP2_INST_SETGE: case SQ_OP2_INST_PRED_SETE: @@ -358,6 +366,7 @@ unsigned int r700GetNumOperands(r700_AssemblerBase* pAsm) case SQ_OP2_INST_MOVA_FLOOR: case SQ_OP2_INST_FRACT: case SQ_OP2_INST_FLOOR: + case SQ_OP2_INST_TRUNC: case SQ_OP2_INST_EXP_IEEE: case SQ_OP2_INST_LOG_CLAMPED: case SQ_OP2_INST_LOG_IEEE: @@ -383,98 +392,115 @@ int Init_r700_AssemblerBase(SHADER_PIPE_TYPE spt, r700_AssemblerBase* pAsm, R700 pAsm->pR700Shader = pShader; pAsm->currentShaderType = spt; - pAsm->cf_last_export_ptr = NULL; + pAsm->cf_last_export_ptr = NULL; + + pAsm->cf_current_export_clause_ptr = NULL; + pAsm->cf_current_alu_clause_ptr = NULL; + pAsm->cf_current_tex_clause_ptr = NULL; + pAsm->cf_current_vtx_clause_ptr = NULL; + pAsm->cf_current_cf_clause_ptr = NULL; - pAsm->cf_current_export_clause_ptr = NULL; - pAsm->cf_current_alu_clause_ptr = NULL; - pAsm->cf_current_tex_clause_ptr = NULL; - pAsm->cf_current_vtx_clause_ptr = NULL; - pAsm->cf_current_cf_clause_ptr = NULL; + // No clause has been created yet + pAsm->cf_current_clause_type = CF_EMPTY_CLAUSE; - // No clause has been created yet - pAsm->cf_current_clause_type = CF_EMPTY_CLAUSE; + pAsm->number_of_colorandz_exports = 0; + pAsm->number_of_exports = 0; + pAsm->number_of_export_opcodes = 0; - pAsm->number_of_colorandz_exports = 0; - pAsm->number_of_exports = 0; - pAsm->number_of_export_opcodes = 0; + pAsm->alu_x_opcode = 0; + pAsm->D2.bits = 0; - pAsm->D.bits = 0; - pAsm->S[0].bits = 0; - pAsm->S[1].bits = 0; - pAsm->S[2].bits = 0; + pAsm->D.bits = 0; + pAsm->S[0].bits = 0; + pAsm->S[1].bits = 0; + pAsm->S[2].bits = 0; - pAsm->uLastPosUpdate = 0; + pAsm->uLastPosUpdate = 0; - *(BITS *) &pAsm->fp_stOutFmt0 = 0; + *(BITS *) &pAsm->fp_stOutFmt0 = 0; - pAsm->uIIns = 0; - pAsm->uOIns = 0; - pAsm->number_used_registers = 0; - pAsm->uUsedConsts = 256; + pAsm->uIIns = 0; + pAsm->uOIns = 0; + pAsm->number_used_registers = 0; + pAsm->uUsedConsts = 256; - // Fragment programs - pAsm->uBoolConsts = 0; - pAsm->uIntConsts = 0; - pAsm->uInsts = 0; - pAsm->uConsts = 0; + // Fragment programs + pAsm->uBoolConsts = 0; + pAsm->uIntConsts = 0; + pAsm->uInsts = 0; + pAsm->uConsts = 0; - pAsm->FCSP = 0; - pAsm->fc_stack[0].type = FC_NONE; + pAsm->FCSP = 0; + pAsm->fc_stack[0].type = FC_NONE; - pAsm->branch_depth = 0; - pAsm->max_branch_depth = 0; + pAsm->aArgSubst[0] = + pAsm->aArgSubst[1] = + pAsm->aArgSubst[2] = + pAsm->aArgSubst[3] = (-1); - pAsm->aArgSubst[0] = - pAsm->aArgSubst[1] = - pAsm->aArgSubst[2] = - pAsm->aArgSubst[3] = (-1); + pAsm->uOutputs = 0; - pAsm->uOutputs = 0; + for (i=0; i<NUMBER_OF_OUTPUT_COLORS; i++) + { + pAsm->color_export_register_number[i] = (-1); + } - for (i=0; i<NUMBER_OF_OUTPUT_COLORS; i++) - { - pAsm->color_export_register_number[i] = (-1); - } + pAsm->depth_export_register_number = (-1); + pAsm->stencil_export_register_number = (-1); + pAsm->coverage_to_mask_export_register_number = (-1); + pAsm->mask_export_register_number = (-1); - pAsm->depth_export_register_number = (-1); - pAsm->stencil_export_register_number = (-1); - pAsm->coverage_to_mask_export_register_number = (-1); - pAsm->mask_export_register_number = (-1); + pAsm->starting_export_register_number = 0; + pAsm->starting_vfetch_register_number = 0; + pAsm->starting_temp_register_number = 0; + pAsm->uFirstHelpReg = 0; - pAsm->starting_export_register_number = 0; - pAsm->starting_vfetch_register_number = 0; - pAsm->starting_temp_register_number = 0; - pAsm->uFirstHelpReg = 0; + pAsm->input_position_is_used = GL_FALSE; + pAsm->input_normal_is_used = GL_FALSE; + for (i=0; i<NUMBER_OF_INPUT_COLORS; i++) + { + pAsm->input_color_is_used[ i ] = GL_FALSE; + } - pAsm->input_position_is_used = GL_FALSE; - pAsm->input_normal_is_used = GL_FALSE; + for (i=0; i<NUMBER_OF_TEXTURE_UNITS; i++) + { + pAsm->input_texture_unit_is_used[ i ] = GL_FALSE; + } + for (i=0; i<VERT_ATTRIB_MAX; i++) + { + pAsm->vfetch_instruction_ptr_array[ i ] = NULL; + } - for (i=0; i<NUMBER_OF_INPUT_COLORS; i++) - { - pAsm->input_color_is_used[ i ] = GL_FALSE; - } + pAsm->number_of_inputs = 0; - for (i=0; i<NUMBER_OF_TEXTURE_UNITS; i++) - { - pAsm->input_texture_unit_is_used[ i ] = GL_FALSE; - } + pAsm->is_tex = GL_FALSE; + pAsm->need_tex_barrier = GL_FALSE; - for (i=0; i<VERT_ATTRIB_MAX; i++) - { - pAsm->vfetch_instruction_ptr_array[ i ] = NULL; - } + pAsm->subs = NULL; + pAsm->unSubArraySize = 0; + pAsm->unSubArrayPointer = 0; + pAsm->callers = NULL; + pAsm->unCallerArraySize = 0; + pAsm->unCallerArrayPointer = 0; + + pAsm->CALLSP = 0; + pAsm->CALLSTACK[0].FCSP_BeforeEntry = 0; + pAsm->CALLSTACK[0].plstCFInstructions_local + = &(pAsm->pR700Shader->lstCFInstructions); - pAsm->number_of_inputs = 0; + pAsm->CALLSTACK[0].max = 0; + pAsm->CALLSTACK[0].current = 0; - pAsm->is_tex = GL_FALSE; - pAsm->need_tex_barrier = GL_FALSE; + SetActiveCFlist(pAsm->pR700Shader, pAsm->CALLSTACK[0].plstCFInstructions_local); - return 0; + pAsm->unCFflags = 0; + + return 0; } GLboolean IsTex(gl_inst_opcode Opcode) @@ -592,6 +618,31 @@ int check_current_clause(r700_AssemblerBase* pAsm, return GL_TRUE; } +GLboolean add_cf_instruction(r700_AssemblerBase* pAsm) +{ + if(GL_FALSE == check_current_clause(pAsm, CF_OTHER_CLAUSE)) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr = + (R700ControlFlowGenericClause*) CALLOC_STRUCT(R700ControlFlowGenericClause); + + if (pAsm->cf_current_cf_clause_ptr != NULL) + { + Init_R700ControlFlowGenericClause(pAsm->cf_current_cf_clause_ptr); + AddCFInstruction( pAsm->pR700Shader, + (R700ControlFlowInstruction *)pAsm->cf_current_cf_clause_ptr ); + } + else + { + radeon_error("Could not allocate a new VFetch CF instruction.\n"); + return GL_FALSE; + } + + return GL_TRUE; +} + GLboolean add_vfetch_instruction(r700_AssemblerBase* pAsm, R700VertexInstruction* vertex_instruction_ptr) { @@ -1153,6 +1204,7 @@ GLboolean assemble_src(r700_AssemblerBase *pAsm, case PROGRAM_LOCAL_PARAM: case PROGRAM_ENV_PARAM: case PROGRAM_STATE_VAR: + case PROGRAM_UNIFORM: if (1 == pILInst->SrcReg[src].RelAddr) { setaddrmode_PVSSRC(&(pAsm->S[fld].src), ADDR_RELATIVE_A0); @@ -1179,7 +1231,7 @@ GLboolean assemble_src(r700_AssemblerBase *pAsm, } break; default: - radeon_error("Invalid source argument type\n"); + radeon_error("Invalid source argument type : %d \n", pILInst->SrcReg[src].File); return GL_FALSE; } } @@ -1315,7 +1367,7 @@ GLboolean tex_src(r700_AssemblerBase *pAsm) case FRAG_ATTRIB_TEX0: case FRAG_ATTRIB_TEX1: case FRAG_ATTRIB_TEX2: - case FRAG_ATTRIB_TEX3: + case FRAG_ATTRIB_TEX3: case FRAG_ATTRIB_TEX4: case FRAG_ATTRIB_TEX5: case FRAG_ATTRIB_TEX6: @@ -1331,10 +1383,17 @@ GLboolean tex_src(r700_AssemblerBase *pAsm) case FRAG_ATTRIB_PNTC: fprintf(stderr, "FRAG_ATTRIB_PNTC unsupported\n"); break; - case FRAG_ATTRIB_VAR0: - fprintf(stderr, "FRAG_ATTRIB_VAR0 unsupported\n"); - break; } + + if( (pILInst->SrcReg[0].Index >= FRAG_ATTRIB_VAR0) || + (pILInst->SrcReg[0].Index < FRAG_ATTRIB_MAX) ) + { + bValidTexCoord = GL_TRUE; + pAsm->S[0].src.reg = + pAsm->uiFP_AttributeMap[pILInst->SrcReg[0].Index]; + pAsm->S[0].src.rtype = SRC_REG_INPUT; + } + break; } } @@ -1517,6 +1576,10 @@ GLboolean assemble_alu_src(R700ALUInstruction* alu_instruction_ptr, { src_sel = pSource->reg + CFILE_REGISTER_OFFSET; } + else if (pSource->rtype == SRC_REC_LITERAL) + { + src_sel = SQ_ALU_SRC_LITERAL; + } else { radeon_error("Source (%d) register type (%d) not one of TEMP, INPUT, or CONSTANT.\n", @@ -1606,7 +1669,8 @@ GLboolean add_alu_instruction(r700_AssemblerBase* pAsm, return GL_FALSE; } - if ( pAsm->cf_current_alu_clause_ptr == NULL || + if ( pAsm->alu_x_opcode != 0 || + pAsm->cf_current_alu_clause_ptr == NULL || ( (pAsm->cf_current_alu_clause_ptr != NULL) && (pAsm->cf_current_alu_clause_ptr->m_Word1.f.count >= (GetCFMaxInstructions(pAsm->cf_current_alu_clause_ptr->m_ShaderInstType)-contiguous_slots_needed-1) ) ) ) @@ -1636,9 +1700,17 @@ GLboolean add_alu_instruction(r700_AssemblerBase* pAsm, pAsm->cf_current_alu_clause_ptr->m_Word1.f.kcache_addr0 = 0x0; pAsm->cf_current_alu_clause_ptr->m_Word1.f.kcache_addr1 = 0x0; - //cf_current_alu_clause_ptr->m_Word1.f.count = number_of_scalar_operations - 1; pAsm->cf_current_alu_clause_ptr->m_Word1.f.count = 0x0; - pAsm->cf_current_alu_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_ALU; + + if(pAsm->alu_x_opcode != 0) + { + pAsm->cf_current_alu_clause_ptr->m_Word1.f.cf_inst = pAsm->alu_x_opcode; + pAsm->alu_x_opcode = 0; + } + else + { + pAsm->cf_current_alu_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_ALU; + } pAsm->cf_current_alu_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; @@ -2358,6 +2430,506 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) return GL_TRUE; } +GLboolean assemble_alu_instruction2(r700_AssemblerBase *pAsm) +{ + GLuint number_of_scalar_operations; + GLboolean is_single_scalar_operation; + GLuint scalar_channel_index; + + PVSSRC * pcurrent_source; + int current_source_index; + GLuint contiguous_slots_needed; + + GLuint uNumSrc = r700GetNumOperands(pAsm); + + GLboolean bSplitInst = GL_FALSE; + + if (1 == pAsm->D.dst.math) + { + is_single_scalar_operation = GL_TRUE; + number_of_scalar_operations = 1; + } + else + { + is_single_scalar_operation = GL_FALSE; + number_of_scalar_operations = 4; + } + + contiguous_slots_needed = 0; + + if(GL_TRUE == is_reduction_opcode(&(pAsm->D)) ) + { + contiguous_slots_needed = 4; + } + + initialize(pAsm); + + for (scalar_channel_index=0; + scalar_channel_index < number_of_scalar_operations; + scalar_channel_index++) + { + R700ALUInstruction* alu_instruction_ptr = (R700ALUInstruction*) CALLOC_STRUCT(R700ALUInstruction); + if (alu_instruction_ptr == NULL) + { + return GL_FALSE; + } + Init_R700ALUInstruction(alu_instruction_ptr); + + //src 0 + current_source_index = 0; + pcurrent_source = &(pAsm->S[0].src); + + if (GL_FALSE == assemble_alu_src(alu_instruction_ptr, + current_source_index, + pcurrent_source, + scalar_channel_index) ) + { + return GL_FALSE; + } + + if (uNumSrc > 1) + { + // Process source 1 + current_source_index = 1; + pcurrent_source = &(pAsm->S[current_source_index].src); + + if (GL_FALSE == assemble_alu_src(alu_instruction_ptr, + current_source_index, + pcurrent_source, + scalar_channel_index) ) + { + return GL_FALSE; + } + } + + //other bits + alu_instruction_ptr->m_Word0.f.index_mode = SQ_INDEX_LOOP; + + if( (is_single_scalar_operation == GL_TRUE) + || (GL_TRUE == bSplitInst) ) + { + alu_instruction_ptr->m_Word0.f.last = 1; + } + else + { + alu_instruction_ptr->m_Word0.f.last = (scalar_channel_index == 3) ? 1 : 0; + } + + alu_instruction_ptr->m_Word0.f.pred_sel = (pAsm->D.dst.pred_inv > 0) ? 1 : 0; + if(1 == pAsm->D.dst.predicated) + { + alu_instruction_ptr->m_Word1_OP2.f.update_pred = 0x1; + alu_instruction_ptr->m_Word1_OP2.f.update_execute_mask = 0x1; + } + else + { + alu_instruction_ptr->m_Word1_OP2.f.update_pred = 0x0; + alu_instruction_ptr->m_Word1_OP2.f.update_execute_mask = 0x0; + } + + // dst + if( (pAsm->D.dst.rtype == DST_REG_TEMPORARY) || + (pAsm->D.dst.rtype == DST_REG_OUT) ) + { + alu_instruction_ptr->m_Word1.f.dst_gpr = pAsm->D.dst.reg; + } + else + { + radeon_error("Only temp destination registers supported for ALU dest regs.\n"); + return GL_FALSE; + } + + alu_instruction_ptr->m_Word1.f.dst_rel = SQ_ABSOLUTE; //D.rtype + + if ( is_single_scalar_operation == GL_TRUE ) + { + // Override scalar_channel_index since only one scalar value will be written + if(pAsm->D.dst.writex) + { + scalar_channel_index = 0; + } + else if(pAsm->D.dst.writey) + { + scalar_channel_index = 1; + } + else if(pAsm->D.dst.writez) + { + scalar_channel_index = 2; + } + else if(pAsm->D.dst.writew) + { + scalar_channel_index = 3; + } + } + + alu_instruction_ptr->m_Word1.f.dst_chan = scalar_channel_index; + + alu_instruction_ptr->m_Word1.f.clamp = pAsm->D2.dst2.SaturateMode; + + if (pAsm->D.dst.op3) + { + //op3 + + alu_instruction_ptr->m_Word1_OP3.f.alu_inst = pAsm->D.dst.opcode; + + //There's 3rd src for op3 + current_source_index = 2; + pcurrent_source = &(pAsm->S[current_source_index].src); + + if ( GL_FALSE == assemble_alu_src(alu_instruction_ptr, + current_source_index, + pcurrent_source, + scalar_channel_index) ) + { + return GL_FALSE; + } + } + else + { + //op2 + if (pAsm->bR6xx) + { + alu_instruction_ptr->m_Word1_OP2.f6.alu_inst = pAsm->D.dst.opcode; + + alu_instruction_ptr->m_Word1_OP2.f6.src0_abs = 0x0; + alu_instruction_ptr->m_Word1_OP2.f6.src1_abs = 0x0; + + //alu_instruction_ptr->m_Word1_OP2.f6.update_execute_mask = 0x0; + //alu_instruction_ptr->m_Word1_OP2.f6.update_pred = 0x0; + switch (scalar_channel_index) + { + case 0: + alu_instruction_ptr->m_Word1_OP2.f6.write_mask = pAsm->D.dst.writex; + break; + case 1: + alu_instruction_ptr->m_Word1_OP2.f6.write_mask = pAsm->D.dst.writey; + break; + case 2: + alu_instruction_ptr->m_Word1_OP2.f6.write_mask = pAsm->D.dst.writez; + break; + case 3: + alu_instruction_ptr->m_Word1_OP2.f6.write_mask = pAsm->D.dst.writew; + break; + default: + alu_instruction_ptr->m_Word1_OP2.f6.write_mask = 1; //SQ_SEL_MASK; + break; + } + alu_instruction_ptr->m_Word1_OP2.f6.omod = SQ_ALU_OMOD_OFF; + } + else + { + alu_instruction_ptr->m_Word1_OP2.f.alu_inst = pAsm->D.dst.opcode; + + alu_instruction_ptr->m_Word1_OP2.f.src0_abs = 0x0; + alu_instruction_ptr->m_Word1_OP2.f.src1_abs = 0x0; + + //alu_instruction_ptr->m_Word1_OP2.f.update_execute_mask = 0x0; + //alu_instruction_ptr->m_Word1_OP2.f.update_pred = 0x0; + switch (scalar_channel_index) + { + case 0: + alu_instruction_ptr->m_Word1_OP2.f.write_mask = pAsm->D.dst.writex; + break; + case 1: + alu_instruction_ptr->m_Word1_OP2.f.write_mask = pAsm->D.dst.writey; + break; + case 2: + alu_instruction_ptr->m_Word1_OP2.f.write_mask = pAsm->D.dst.writez; + break; + case 3: + alu_instruction_ptr->m_Word1_OP2.f.write_mask = pAsm->D.dst.writew; + break; + default: + alu_instruction_ptr->m_Word1_OP2.f.write_mask = 1; //SQ_SEL_MASK; + break; + } + alu_instruction_ptr->m_Word1_OP2.f.omod = SQ_ALU_OMOD_OFF; + } + } + + if(GL_FALSE == add_alu_instruction(pAsm, alu_instruction_ptr, contiguous_slots_needed) ) + { + return GL_FALSE; + } + + /* + * Judge the type of current instruction, is it vector or scalar + * instruction. + */ + if (is_single_scalar_operation) + { + if(GL_FALSE == check_scalar(pAsm, alu_instruction_ptr) ) + { + return GL_FALSE; + } + } + else + { + if(GL_FALSE == check_vector(pAsm, alu_instruction_ptr) ) + { + return 1; + } + } + + contiguous_slots_needed = 0; + } + + return GL_TRUE; +} + +GLboolean assemble_alu_instruction_literal(r700_AssemblerBase *pAsm, GLfloat * pLiteral) +{ + R700ALUInstruction * alu_instruction_ptr; + R700ALUInstructionHalfLiteral * alu_instruction_ptr_hl; + R700ALUInstructionFullLiteral * alu_instruction_ptr_fl; + + GLuint number_of_scalar_operations; + GLboolean is_single_scalar_operation; + GLuint scalar_channel_index; + + GLuint contiguous_slots_needed; + GLuint lastInstruction; + GLuint not_masked[4]; + + GLuint uNumSrc = r700GetNumOperands(pAsm); + + GLboolean bSplitInst = GL_FALSE; + + number_of_scalar_operations = 0; + contiguous_slots_needed = 0; + + if(1 == pAsm->D.dst.writew) + { + lastInstruction = 3; + number_of_scalar_operations++; + not_masked[3] = 1; + } + else + { + not_masked[3] = 0; + } + if(1 == pAsm->D.dst.writez) + { + lastInstruction = 2; + number_of_scalar_operations++; + not_masked[2] = 1; + } + else + { + not_masked[2] = 0; + } + if(1 == pAsm->D.dst.writey) + { + lastInstruction = 1; + number_of_scalar_operations++; + not_masked[1] = 1; + } + else + { + not_masked[1] = 0; + } + if(1 == pAsm->D.dst.writex) + { + lastInstruction = 0; + number_of_scalar_operations++; + not_masked[0] = 1; + } + else + { + not_masked[0] = 0; + } + + if(GL_TRUE == is_reduction_opcode(&(pAsm->D)) ) + { + contiguous_slots_needed = 4; + } + else + { + contiguous_slots_needed = number_of_scalar_operations; + } + + if(1 == pAsm->D2.dst2.literal) + { + contiguous_slots_needed += 1; + } + else if(2 == pAsm->D2.dst2.literal) + { + contiguous_slots_needed += 2; + } + + initialize(pAsm); + + for (scalar_channel_index=0; scalar_channel_index < 4; scalar_channel_index++) + { + if(0 == not_masked[scalar_channel_index]) + { + continue; + } + + if(scalar_channel_index == lastInstruction) + { + switch (pAsm->D2.dst2.literal) + { + case 0: + alu_instruction_ptr = (R700ALUInstruction*) CALLOC_STRUCT(R700ALUInstruction); + if (alu_instruction_ptr == NULL) + { + return GL_FALSE; + } + Init_R700ALUInstruction(alu_instruction_ptr); + break; + case 1: + alu_instruction_ptr_hl = (R700ALUInstructionHalfLiteral*) CALLOC_STRUCT(R700ALUInstructionHalfLiteral); + if (alu_instruction_ptr_hl == NULL) + { + return GL_FALSE; + } + Init_R700ALUInstructionHalfLiteral(alu_instruction_ptr_hl, pLiteral[0], pLiteral[1]); + alu_instruction_ptr = (R700ALUInstruction*)alu_instruction_ptr_hl; + break; + case 2: + alu_instruction_ptr_fl = (R700ALUInstructionFullLiteral*) CALLOC_STRUCT(R700ALUInstructionFullLiteral); + if (alu_instruction_ptr_fl == NULL) + { + return GL_FALSE; + } + Init_R700ALUInstructionFullLiteral(alu_instruction_ptr_fl, pLiteral[0], pLiteral[1], pLiteral[2], pLiteral[3]); + alu_instruction_ptr = (R700ALUInstruction*)alu_instruction_ptr_fl; + break; + default: + break; + }; + } + else + { + alu_instruction_ptr = (R700ALUInstruction*) CALLOC_STRUCT(R700ALUInstruction); + if (alu_instruction_ptr == NULL) + { + return GL_FALSE; + } + Init_R700ALUInstruction(alu_instruction_ptr); + } + + //src 0 + if (GL_FALSE == assemble_alu_src(alu_instruction_ptr, + 0, + &(pAsm->S[0].src), + scalar_channel_index) ) + { + return GL_FALSE; + } + + if (uNumSrc > 1) + { + // Process source 1 + if (GL_FALSE == assemble_alu_src(alu_instruction_ptr, + 1, + &(pAsm->S[1].src), + scalar_channel_index) ) + { + return GL_FALSE; + } + } + + //other bits + alu_instruction_ptr->m_Word0.f.index_mode = SQ_INDEX_LOOP; + + if(scalar_channel_index == lastInstruction) + { + alu_instruction_ptr->m_Word0.f.last = 1; + } + + alu_instruction_ptr->m_Word0.f.pred_sel = 0x0; + if(1 == pAsm->D.dst.predicated) + { + alu_instruction_ptr->m_Word1_OP2.f.update_pred = 0x1; + alu_instruction_ptr->m_Word1_OP2.f.update_execute_mask = 0x1; + } + else + { + alu_instruction_ptr->m_Word1_OP2.f.update_pred = 0; + alu_instruction_ptr->m_Word1_OP2.f.update_execute_mask = 0; + } + + // dst + if( (pAsm->D.dst.rtype == DST_REG_TEMPORARY) || + (pAsm->D.dst.rtype == DST_REG_OUT) ) + { + alu_instruction_ptr->m_Word1.f.dst_gpr = pAsm->D.dst.reg; + } + else + { + radeon_error("Only temp destination registers supported for ALU dest regs.\n"); + return GL_FALSE; + } + + alu_instruction_ptr->m_Word1.f.dst_rel = SQ_ABSOLUTE; //D.rtype + + alu_instruction_ptr->m_Word1.f.dst_chan = scalar_channel_index; + + alu_instruction_ptr->m_Word1.f.clamp = pAsm->D2.dst2.SaturateMode; + + if (pAsm->D.dst.op3) + { + //op3 + alu_instruction_ptr->m_Word1_OP3.f.alu_inst = pAsm->D.dst.opcode; + + //There's 3rd src for op3 + if ( GL_FALSE == assemble_alu_src(alu_instruction_ptr, + 2, + &(pAsm->S[2].src), + scalar_channel_index) ) + { + return GL_FALSE; + } + } + else + { + //op2 + if (pAsm->bR6xx) + { + alu_instruction_ptr->m_Word1_OP2.f6.alu_inst = pAsm->D.dst.opcode; + alu_instruction_ptr->m_Word1_OP2.f6.src0_abs = 0x0; + alu_instruction_ptr->m_Word1_OP2.f6.src1_abs = 0x0; + alu_instruction_ptr->m_Word1_OP2.f6.write_mask = 1; + alu_instruction_ptr->m_Word1_OP2.f6.omod = SQ_ALU_OMOD_OFF; + } + else + { + alu_instruction_ptr->m_Word1_OP2.f.alu_inst = pAsm->D.dst.opcode; + alu_instruction_ptr->m_Word1_OP2.f.src0_abs = 0x0; + alu_instruction_ptr->m_Word1_OP2.f.src1_abs = 0x0; + alu_instruction_ptr->m_Word1_OP2.f.write_mask = 1; + alu_instruction_ptr->m_Word1_OP2.f.omod = SQ_ALU_OMOD_OFF; + } + } + + if(GL_FALSE == add_alu_instruction(pAsm, alu_instruction_ptr, contiguous_slots_needed) ) + { + return GL_FALSE; + } + + if (1 == number_of_scalar_operations) + { + if(GL_FALSE == check_scalar(pAsm, alu_instruction_ptr) ) + { + return GL_FALSE; + } + } + else + { + if(GL_FALSE == check_vector(pAsm, alu_instruction_ptr) ) + { + return GL_FALSE; + } + } + + contiguous_slots_needed -= 2; + } + + return GL_TRUE; +} + GLboolean next_ins(r700_AssemblerBase *pAsm) { struct prog_instruction *pILInst = &(pAsm->pILInst[pAsm->uiCurInst]); @@ -2403,6 +2975,70 @@ GLboolean next_ins(r700_AssemblerBase *pAsm) //reset for next inst. pAsm->D.bits = 0; + pAsm->D2.bits = 0; + pAsm->S[0].bits = 0; + pAsm->S[1].bits = 0; + pAsm->S[2].bits = 0; + pAsm->is_tex = GL_FALSE; + pAsm->need_tex_barrier = GL_FALSE; + + return GL_TRUE; +} + +GLboolean next_ins2(r700_AssemblerBase *pAsm) +{ + struct prog_instruction *pILInst = &(pAsm->pILInst[pAsm->uiCurInst]); + + //ALU + if( GL_FALSE == assemble_alu_instruction2(pAsm) ) + { + radeon_error("Error assembling ALU instruction\n"); + return GL_FALSE; + } + + if(pAsm->D.dst.rtype == DST_REG_OUT) + { + if(pAsm->D.dst.op3) + { + // There is no mask for OP3 instructions, so all channels are written + pAsm->pucOutMask[pAsm->D.dst.reg - pAsm->starting_export_register_number] = 0xF; + } + else + { + pAsm->pucOutMask[pAsm->D.dst.reg - pAsm->starting_export_register_number] + |= (unsigned char)pAsm->pILInst[pAsm->uiCurInst].DstReg.WriteMask; + } + } + + //reset for next inst. + pAsm->D.bits = 0; + pAsm->D2.bits = 0; + pAsm->S[0].bits = 0; + pAsm->S[1].bits = 0; + pAsm->S[2].bits = 0; + pAsm->is_tex = GL_FALSE; + pAsm->need_tex_barrier = GL_FALSE; + + pAsm->D2.bits = 0; + + return GL_TRUE; +} + +/* not work yet */ +GLboolean next_ins_literal(r700_AssemblerBase *pAsm, GLfloat * pLiteral) +{ + struct prog_instruction *pILInst = &(pAsm->pILInst[pAsm->uiCurInst]); + + //ALU + if( GL_FALSE == assemble_alu_instruction_literal(pAsm, pLiteral) ) + { + radeon_error("Error assembling ALU instruction\n"); + return GL_FALSE; + } + + //reset for next inst. + pAsm->D.bits = 0; + pAsm->D2.bits = 0; pAsm->S[0].bits = 0; pAsm->S[1].bits = 0; pAsm->S[2].bits = 0; @@ -2910,13 +3546,12 @@ GLboolean assemble_FRC(r700_AssemblerBase *pAsm) return GL_TRUE; } -GLboolean assemble_KIL(r700_AssemblerBase *pAsm) -{ - /* TODO: doc says KILL has to be last(end) ALU clause */ - - checkop1(pAsm); +GLboolean assemble_KIL(r700_AssemblerBase *pAsm, GLuint opcode) +{ + checkop2(pAsm); - pAsm->D.dst.opcode = SQ_OP2_INST_KILLGT; + pAsm->D.dst.opcode = opcode; + pAsm->D.dst.math = 1; setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); pAsm->D.dst.rtype = DST_REG_TEMPORARY; @@ -2926,24 +3561,24 @@ GLboolean assemble_KIL(r700_AssemblerBase *pAsm) pAsm->D.dst.writez = 0; pAsm->D.dst.writew = 0; - setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); - pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; - pAsm->S[0].src.reg = 0; - - setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_0); - noneg_PVSSRC(&(pAsm->S[0].src)); + if( GL_FALSE == assemble_src(pAsm, 0, -1) ) + { + return GL_FALSE; + } - if ( GL_FALSE == assemble_src(pAsm, 0, 1) ) + if( GL_FALSE == assemble_src(pAsm, 1, -1) ) { return GL_FALSE; } - if ( GL_FALSE == next_ins(pAsm) ) + if ( GL_FALSE == next_ins2(pAsm) ) { return GL_FALSE; } + /* Doc says KILL has to be last(end) ALU clause */ pAsm->pR700Shader->killIsUsed = GL_TRUE; + pAsm->alu_x_opcode = SQ_CF_INST_ALU; return GL_TRUE; } @@ -3816,6 +4451,74 @@ GLboolean assemble_SCS(r700_AssemblerBase *pAsm) return GL_TRUE; } + +GLboolean assemble_LOGIC(r700_AssemblerBase *pAsm, BITS opcode) +{ + if( GL_FALSE == checkop2(pAsm) ) + { + return GL_FALSE; + } + + pAsm->D.dst.opcode = opcode; + pAsm->D.dst.math = 1; + + if( GL_FALSE == assemble_dst(pAsm) ) + { + return GL_FALSE; + } + + if( GL_FALSE == assemble_src(pAsm, 0, -1) ) + { + return GL_FALSE; + } + + if( GL_FALSE == assemble_src(pAsm, 1, -1) ) + { + return GL_FALSE; + } + + if( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } + + return GL_TRUE; +} + +GLboolean assemble_LOGIC_PRED(r700_AssemblerBase *pAsm, BITS opcode) +{ + if( GL_FALSE == checkop2(pAsm) ) + { + return GL_FALSE; + } + + pAsm->D.dst.opcode = opcode; + pAsm->D.dst.math = 1; + pAsm->D.dst.predicated = 1; + pAsm->D2.dst2.SaturateMode = pAsm->pILInst[pAsm->uiCurInst].SaturateMode; + + if( GL_FALSE == assemble_dst(pAsm) ) + { + return GL_FALSE; + } + + if( GL_FALSE == assemble_src(pAsm, 0, -1) ) + { + return GL_FALSE; + } + + if( GL_FALSE == assemble_src(pAsm, 1, -1) ) + { + return GL_FALSE; + } + + if( GL_FALSE == next_ins2(pAsm) ) + { + return GL_FALSE; + } + + return GL_TRUE; +} GLboolean assemble_SGE(r700_AssemblerBase *pAsm) { @@ -4273,27 +4976,876 @@ GLboolean assemble_EXPORT(r700_AssemblerBase *pAsm) return GL_TRUE; } -GLboolean assemble_IF(r700_AssemblerBase *pAsm) +static inline void decreaseCurrent(r700_AssemblerBase *pAsm, GLuint uReason) +{ + switch (uReason) + { + case FC_PUSH_VPM: + pAsm->CALLSTACK[pAsm->CALLSP].current--; + break; + case FC_PUSH_WQM: + pAsm->CALLSTACK[pAsm->CALLSP].current -= 4; + break; + case FC_LOOP: + pAsm->CALLSTACK[pAsm->CALLSP].current -= 4; + break; + case FC_REP: + /* TODO : for 16 vp asic, should -= 2; */ + pAsm->CALLSTACK[pAsm->CALLSP].current -= 1; + break; + }; +} + +static inline void checkStackDepth(r700_AssemblerBase *pAsm, GLuint uReason, GLboolean bCheckMaxOnly) +{ + if(GL_TRUE == bCheckMaxOnly) + { + switch (uReason) + { + case FC_PUSH_VPM: + if((pAsm->CALLSTACK[pAsm->CALLSP].current + 1) + > pAsm->CALLSTACK[pAsm->CALLSP].max) + { + pAsm->CALLSTACK[pAsm->CALLSP].max = + pAsm->CALLSTACK[pAsm->CALLSP].current + 1; + } + break; + case FC_PUSH_WQM: + if((pAsm->CALLSTACK[pAsm->CALLSP].current + 4) + > pAsm->CALLSTACK[pAsm->CALLSP].max) + { + pAsm->CALLSTACK[pAsm->CALLSP].max = + pAsm->CALLSTACK[pAsm->CALLSP].current + 4; + } + break; + } + return; + } + + switch (uReason) + { + case FC_PUSH_VPM: + pAsm->CALLSTACK[pAsm->CALLSP].current++; + break; + case FC_PUSH_WQM: + pAsm->CALLSTACK[pAsm->CALLSP].current += 4; + break; + case FC_LOOP: + pAsm->CALLSTACK[pAsm->CALLSP].current += 4; + break; + case FC_REP: + /* TODO : for 16 vp asic, should += 2; */ + pAsm->CALLSTACK[pAsm->CALLSP].current += 1; + break; + }; + + if(pAsm->CALLSTACK[pAsm->CALLSP].current + > pAsm->CALLSTACK[pAsm->CALLSP].max) + { + pAsm->CALLSTACK[pAsm->CALLSP].max = + pAsm->CALLSTACK[pAsm->CALLSP].current; + } +} + +GLboolean jumpToOffest(r700_AssemblerBase *pAsm, GLuint pops, GLint offset) +{ + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = pops; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_JUMP; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->cf_current_cf_clause_ptr->m_Word0.f.addr = pAsm->cf_current_cf_clause_ptr->m_uIndex + offset; + + return GL_TRUE; +} + +GLboolean pops(r700_AssemblerBase *pAsm, GLuint pops) +{ + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = pops; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_POP; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + pAsm->cf_current_cf_clause_ptr->m_Word0.f.addr = pAsm->cf_current_cf_clause_ptr->m_uIndex + 1; + + return GL_TRUE; +} + +GLboolean assemble_IF(r700_AssemblerBase *pAsm, GLboolean bHasElse) +{ + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + if(GL_TRUE != bHasElse) + { + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + } + else + { + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 0; + } + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_JUMP; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->FCSP++; + pAsm->fc_stack[pAsm->FCSP].type = FC_IF; + pAsm->fc_stack[pAsm->FCSP].mid = NULL; + pAsm->fc_stack[pAsm->FCSP].midLen= 0; + pAsm->fc_stack[pAsm->FCSP].first = pAsm->cf_current_cf_clause_ptr; + +#ifndef USE_CF_FOR_POP_AFTER + if(GL_TRUE != bHasElse) + { + pAsm->alu_x_opcode = SQ_CF_INST_ALU_POP_AFTER; + } +#endif /* USE_CF_FOR_POP_AFTER */ + + checkStackDepth(pAsm, FC_PUSH_VPM, GL_FALSE); + + return GL_TRUE; +} + +GLboolean assemble_ELSE(r700_AssemblerBase *pAsm) { + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; /// + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_ELSE; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->fc_stack[pAsm->FCSP].mid = (R700ControlFlowGenericClause **)_mesa_realloc( (void *)pAsm->fc_stack[pAsm->FCSP].mid, + 0, + sizeof(R700ControlFlowGenericClause *) ); + pAsm->fc_stack[pAsm->FCSP].mid[0] = pAsm->cf_current_cf_clause_ptr; + //pAsm->fc_stack[pAsm->FCSP].unNumMid = 1; + +#ifndef USE_CF_FOR_POP_AFTER + pAsm->alu_x_opcode = SQ_CF_INST_ALU_POP_AFTER; +#endif /* USE_CF_FOR_POP_AFTER */ + + pAsm->fc_stack[pAsm->FCSP].first->m_Word0.f.addr = pAsm->pR700Shader->plstCFInstructions_active->uNumOfNode - 1; + return GL_TRUE; } GLboolean assemble_ENDIF(r700_AssemblerBase *pAsm) { +#ifdef USE_CF_FOR_POP_AFTER + pops(pAsm, 1); +#endif /* USE_CF_FOR_POP_AFTER */ + + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + + if(NULL == pAsm->fc_stack[pAsm->FCSP].mid) + { + /* no else in between */ + pAsm->fc_stack[pAsm->FCSP].first->m_Word0.f.addr = pAsm->pR700Shader->plstCFInstructions_active->uNumOfNode; + } + else + { + pAsm->fc_stack[pAsm->FCSP].mid[0]->m_Word0.f.addr = pAsm->pR700Shader->plstCFInstructions_active->uNumOfNode; + } + + if(NULL != pAsm->fc_stack[pAsm->FCSP].mid) + { + FREE(pAsm->fc_stack[pAsm->FCSP].mid); + } + + if(pAsm->fc_stack[pAsm->FCSP].type != FC_IF) + { + radeon_error("if/endif in shader code are not paired. \n"); + return GL_FALSE; + } + + pAsm->FCSP--; + + decreaseCurrent(pAsm, FC_PUSH_VPM); + + return GL_TRUE; +} + +GLboolean assemble_BGNLOOP(r700_AssemblerBase *pAsm) +{ + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_LOOP_START_NO_AL; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->FCSP++; + pAsm->fc_stack[pAsm->FCSP].type = FC_LOOP; + pAsm->fc_stack[pAsm->FCSP].mid = NULL; + pAsm->fc_stack[pAsm->FCSP].unNumMid = 0; + pAsm->fc_stack[pAsm->FCSP].midLen = 0; + pAsm->fc_stack[pAsm->FCSP].first = pAsm->cf_current_cf_clause_ptr; + + checkStackDepth(pAsm, FC_LOOP, GL_FALSE); + + return GL_TRUE; +} + +GLboolean assemble_BRK(r700_AssemblerBase *pAsm) +{ +#ifdef USE_CF_FOR_CONTINUE_BREAK + unsigned int unFCSP; + for(unFCSP=pAsm->FCSP; unFCSP>0; unFCSP--) + { + if(FC_LOOP == pAsm->fc_stack[unFCSP].type) + { + break; + } + } + if(0 == FC_LOOP) + { + radeon_error("Break is not inside loop/endloop pair.\n"); + return GL_FALSE; + } + + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_LOOP_BREAK; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->fc_stack[unFCSP].mid = (R700ControlFlowGenericClause **)_mesa_realloc( + (void *)pAsm->fc_stack[unFCSP].mid, + sizeof(R700ControlFlowGenericClause *) * pAsm->fc_stack[unFCSP].unNumMid, + sizeof(R700ControlFlowGenericClause *) * (pAsm->fc_stack[unFCSP].unNumMid + 1) ); + pAsm->fc_stack[unFCSP].mid[pAsm->fc_stack[unFCSP].unNumMid] = pAsm->cf_current_cf_clause_ptr; + pAsm->fc_stack[unFCSP].unNumMid++; + + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_POP; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + pAsm->cf_current_cf_clause_ptr->m_Word0.f.addr = pAsm->cf_current_cf_clause_ptr->m_uIndex + 1; + + checkStackDepth(pAsm, FC_PUSH_VPM, GL_TRUE); + +#endif //USE_CF_FOR_CONTINUE_BREAK + return GL_TRUE; +} + +GLboolean assemble_CONT(r700_AssemblerBase *pAsm) +{ +#ifdef USE_CF_FOR_CONTINUE_BREAK + unsigned int unFCSP; + for(unFCSP=pAsm->FCSP; unFCSP>0; unFCSP--) + { + if(FC_LOOP == pAsm->fc_stack[unFCSP].type) + { + break; + } + } + if(0 == FC_LOOP) + { + radeon_error("Continue is not inside loop/endloop pair.\n"); + return GL_FALSE; + } + + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_LOOP_CONTINUE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->fc_stack[unFCSP].mid = (R700ControlFlowGenericClause **)_mesa_realloc( + (void *)pAsm->fc_stack[unFCSP].mid, + sizeof(R700ControlFlowGenericClause *) * pAsm->fc_stack[unFCSP].unNumMid, + sizeof(R700ControlFlowGenericClause *) * (pAsm->fc_stack[unFCSP].unNumMid + 1) ); + pAsm->fc_stack[unFCSP].mid[pAsm->fc_stack[unFCSP].unNumMid] = pAsm->cf_current_cf_clause_ptr; + pAsm->fc_stack[unFCSP].unNumMid++; + + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_POP; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + pAsm->cf_current_cf_clause_ptr->m_Word0.f.addr = pAsm->cf_current_cf_clause_ptr->m_uIndex + 1; + + checkStackDepth(pAsm, FC_PUSH_VPM, GL_TRUE); + +#endif /* USE_CF_FOR_CONTINUE_BREAK */ + + return GL_TRUE; +} + +GLboolean assemble_ENDLOOP(r700_AssemblerBase *pAsm) +{ + GLuint i; + + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_LOOP_END; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->cf_current_cf_clause_ptr->m_Word0.f.addr = pAsm->fc_stack[pAsm->FCSP].first->m_uIndex + 1; + pAsm->fc_stack[pAsm->FCSP].first->m_Word0.f.addr = pAsm->cf_current_cf_clause_ptr->m_uIndex + 1; + +#ifdef USE_CF_FOR_CONTINUE_BREAK + for(i=0; i<pAsm->fc_stack[pAsm->FCSP].unNumMid; i++) + { + pAsm->fc_stack[pAsm->FCSP].mid[i]->m_Word0.f.addr = pAsm->cf_current_cf_clause_ptr->m_uIndex; + } + if(NULL != pAsm->fc_stack[pAsm->FCSP].mid) + { + FREE(pAsm->fc_stack[pAsm->FCSP].mid); + } +#endif + + if(pAsm->fc_stack[pAsm->FCSP].type != FC_LOOP) + { + radeon_error("loop/endloop in shader code are not paired. \n"); + return GL_FALSE; + } + + GLuint unFCSP; + GLuint unIF = 0; + if((pAsm->unCFflags & HAS_CURRENT_LOOPRET) > 0) + { + for(unFCSP=(pAsm->FCSP-1); unFCSP>pAsm->CALLSTACK[pAsm->CALLSP].FCSP_BeforeEntry; unFCSP--) + { + if(FC_LOOP == pAsm->fc_stack[unFCSP].type) + { + breakLoopOnFlag(pAsm, unFCSP); + break; + } + else if(FC_IF == pAsm->fc_stack[unFCSP].type) + { + unIF++; + } + } + if(unFCSP <= pAsm->CALLSTACK[pAsm->CALLSP].FCSP_BeforeEntry) + { +#ifdef USE_CF_FOR_POP_AFTER + returnOnFlag(pAsm, unIF); +#else + returnOnFlag(pAsm, 0); +#endif /* USE_CF_FOR_POP_AFTER */ + pAsm->unCFflags &= ~HAS_CURRENT_LOOPRET; + } + } + + pAsm->FCSP--; + + decreaseCurrent(pAsm, FC_LOOP); + + return GL_TRUE; +} + +void add_return_inst(r700_AssemblerBase *pAsm) +{ + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + //pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_RETURN; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; +} + +GLboolean assemble_BGNSUB(r700_AssemblerBase *pAsm, GLint nILindex) +{ + /* Put in sub */ + if( (pAsm->unSubArrayPointer + 1) > pAsm->unSubArraySize ) + { + pAsm->subs = (SUB_OFFSET*)_mesa_realloc( (void *)pAsm->subs, + sizeof(SUB_OFFSET) * pAsm->unSubArraySize, + sizeof(SUB_OFFSET) * (pAsm->unSubArraySize + 10) ); + if(NULL == pAsm->subs) + { + return GL_FALSE; + } + pAsm->unSubArraySize += 10; + } + + pAsm->subs[pAsm->unSubArrayPointer].subIL_Offset = nILindex; + pAsm->subs[pAsm->unSubArrayPointer].lstCFInstructions_local.pHead=NULL; + pAsm->subs[pAsm->unSubArrayPointer].lstCFInstructions_local.pTail=NULL; + pAsm->subs[pAsm->unSubArrayPointer].lstCFInstructions_local.uNumOfNode=0; + + pAsm->CALLSP++; + pAsm->CALLSTACK[pAsm->CALLSP].subDescIndex = pAsm->unSubArrayPointer; + pAsm->CALLSTACK[pAsm->CALLSP].FCSP_BeforeEntry = pAsm->FCSP; + pAsm->CALLSTACK[pAsm->CALLSP].plstCFInstructions_local + = &(pAsm->subs[pAsm->unSubArrayPointer].lstCFInstructions_local); + pAsm->CALLSTACK[pAsm->CALLSP].max = 0; + pAsm->CALLSTACK[pAsm->CALLSP].current = 0; + SetActiveCFlist(pAsm->pR700Shader, + pAsm->CALLSTACK[pAsm->CALLSP].plstCFInstructions_local); + + pAsm->unSubArrayPointer++; + + /* start sub */ + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + + pAsm->FCSP++; + pAsm->fc_stack[pAsm->FCSP].type = FC_REP; + + checkStackDepth(pAsm, FC_REP, GL_FALSE); + + return GL_TRUE; +} + +GLboolean assemble_ENDSUB(r700_AssemblerBase *pAsm) +{ + if(pAsm->fc_stack[pAsm->FCSP].type != FC_REP) + { + radeon_error("BGNSUB/ENDSUB in shader code are not paired. \n"); + return GL_FALSE; + } + + /* copy max to sub structure */ + pAsm->subs[pAsm->CALLSTACK[pAsm->CALLSP].subDescIndex].unStackDepthMax + = pAsm->CALLSTACK[pAsm->CALLSP].max; + + decreaseCurrent(pAsm, FC_REP); + + pAsm->CALLSP--; + SetActiveCFlist(pAsm->pR700Shader, + pAsm->CALLSTACK[pAsm->CALLSP].plstCFInstructions_local); + + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + + pAsm->FCSP--; + + return GL_TRUE; +} + +GLboolean assemble_RET(r700_AssemblerBase *pAsm) +{ + GLuint unIF = 0; + + if(pAsm->CALLSP > 0) + { /* in sub */ + GLuint unFCSP; + for(unFCSP=pAsm->FCSP; unFCSP>pAsm->CALLSTACK[pAsm->CALLSP].FCSP_BeforeEntry; unFCSP--) + { + if(FC_LOOP == pAsm->fc_stack[unFCSP].type) + { + setRetInLoopFlag(pAsm, SQ_SEL_1); + breakLoopOnFlag(pAsm, unFCSP); + pAsm->unCFflags |= LOOPRET_FLAGS; + + return GL_TRUE; + } + else if(FC_IF == pAsm->fc_stack[unFCSP].type) + { + unIF++; + } + } + } + +#ifdef USE_CF_FOR_POP_AFTER + if(unIF > 0) + { + pops(pAsm, unIF); + } +#endif /* USE_CF_FOR_POP_AFTER */ + + add_return_inst(pAsm); + + return GL_TRUE; +} + +GLboolean assemble_CAL(r700_AssemblerBase *pAsm, + GLint nILindex, + GLuint uiNumberInsts, + struct prog_instruction *pILInst) +{ + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.call_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_CALL; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + /* Put in caller */ + if( (pAsm->unCallerArrayPointer + 1) > pAsm->unCallerArraySize ) + { + pAsm->callers = (CALLER_POINTER*)_mesa_realloc( (void *)pAsm->callers, + sizeof(CALLER_POINTER) * pAsm->unCallerArraySize, + sizeof(CALLER_POINTER) * (pAsm->unCallerArraySize + 10) ); + if(NULL == pAsm->callers) + { + return GL_FALSE; + } + pAsm->unCallerArraySize += 10; + } + + pAsm->callers[pAsm->unCallerArrayPointer].subIL_Offset = nILindex; + pAsm->callers[pAsm->unCallerArrayPointer].cf_ptr = pAsm->cf_current_cf_clause_ptr; + + pAsm->unCallerArrayPointer++; + + int j; + GLuint max; + GLuint unSubID; + GLboolean bRet; + for(j=0; j<pAsm->unSubArrayPointer; j++) + { + if(nILindex == pAsm->subs[j].subIL_Offset) + { /* compiled before */ + + max = pAsm->subs[j].unStackDepthMax + + pAsm->CALLSTACK[pAsm->CALLSP].current; + if(max > pAsm->CALLSTACK[pAsm->CALLSP].max) + { + pAsm->CALLSTACK[pAsm->CALLSP].max = max; + } + + pAsm->callers[pAsm->unCallerArrayPointer - 1].subDescIndex = j; + return GL_TRUE; + } + } + + pAsm->callers[pAsm->unCallerArrayPointer - 1].subDescIndex = pAsm->unSubArrayPointer; + unSubID = pAsm->unSubArrayPointer; + + bRet = AssembleInstr(nILindex, uiNumberInsts, pILInst, pAsm); + + if(GL_TRUE == bRet) + { + max = pAsm->subs[unSubID].unStackDepthMax + + pAsm->CALLSTACK[pAsm->CALLSP].current; + if(max > pAsm->CALLSTACK[pAsm->CALLSP].max) + { + pAsm->CALLSTACK[pAsm->CALLSP].max = max; + } + } + + return bRet; +} + +GLboolean setRetInLoopFlag(r700_AssemblerBase *pAsm, GLuint flagValue) +{ + GLfloat fLiteral[2] = {0.1, 0.0}; + + pAsm->D.dst.opcode = SQ_OP2_INST_MOV; + pAsm->D.dst.op3 = 0; + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = pAsm->flag_reg_index; + pAsm->D.dst.writex = 1; + pAsm->D.dst.writey = 0; + pAsm->D.dst.writez = 0; + pAsm->D.dst.writew = 0; + pAsm->D2.dst2.literal = 1; + pAsm->D2.dst2.SaturateMode = SATURATE_OFF; + pAsm->D.dst.predicated = 0; + /* in reloc where dislink flag init inst, only one slot alu inst is handled. */ + pAsm->D.dst.math = 1; /* TODO : not math really, but one channel op, more generic alu assembler needed */ +#if 0 + pAsm->S[0].src.rtype = SRC_REC_LITERAL; + //pAsm->S[0].src.reg = 0; + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + noneg_PVSSRC(&(pAsm->S[0].src)); + pAsm->S[0].src.swizzlex = SQ_SEL_X; + pAsm->S[0].src.swizzley = SQ_SEL_Y; + pAsm->S[0].src.swizzlez = SQ_SEL_Z; + pAsm->S[0].src.swizzlew = SQ_SEL_W; + + if( GL_FALSE == next_ins_literal(pAsm, &(fLiteral[0])) ) + { + return GL_FALSE; + } +#else + pAsm->S[0].src.rtype = DST_REG_TEMPORARY; + pAsm->S[0].src.reg = 0; + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + noneg_PVSSRC(&(pAsm->S[0].src)); + pAsm->S[0].src.swizzlex = flagValue; + pAsm->S[0].src.swizzley = flagValue; + pAsm->S[0].src.swizzlez = flagValue; + pAsm->S[0].src.swizzlew = flagValue; + + if( GL_FALSE == next_ins2(pAsm) ) + { + return GL_FALSE; + } +#endif + return GL_TRUE; } -GLboolean AssembleInstr(GLuint uiNumberInsts, +GLboolean testFlag(r700_AssemblerBase *pAsm) +{ + GLfloat fLiteral[2] = {0.1, 0.0}; + + //Test flag + GLuint tmp = gethelpr(pAsm); + pAsm->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; + + pAsm->D.dst.opcode = SQ_OP2_INST_PRED_SETE; + pAsm->D.dst.math = 1; + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = tmp; + pAsm->D.dst.writex = 1; + pAsm->D.dst.writey = 0; + pAsm->D.dst.writez = 0; + pAsm->D.dst.writew = 0; + pAsm->D2.dst2.literal = 1; + pAsm->D2.dst2.SaturateMode = SATURATE_OFF; + pAsm->D.dst.predicated = 1; + + pAsm->S[0].src.rtype = DST_REG_TEMPORARY; + pAsm->S[0].src.reg = pAsm->flag_reg_index; + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + noneg_PVSSRC(&(pAsm->S[0].src)); + pAsm->S[0].src.swizzlex = SQ_SEL_X; + pAsm->S[0].src.swizzley = SQ_SEL_Y; + pAsm->S[0].src.swizzlez = SQ_SEL_Z; + pAsm->S[0].src.swizzlew = SQ_SEL_W; +#if 0 + pAsm->S[1].src.rtype = SRC_REC_LITERAL; + //pAsm->S[1].src.reg = 0; + setaddrmode_PVSSRC(&(pAsm->S[1].src), ADDR_ABSOLUTE); + noneg_PVSSRC(&(pAsm->S[1].src)); + pAsm->S[1].src.swizzlex = SQ_SEL_X; + pAsm->S[1].src.swizzley = SQ_SEL_Y; + pAsm->S[1].src.swizzlez = SQ_SEL_Z; + pAsm->S[1].src.swizzlew = SQ_SEL_W; + + if( GL_FALSE == next_ins_literal(pAsm, &(fLiteral[0])) ) + { + return GL_FALSE; + } +#else + pAsm->S[1].src.rtype = DST_REG_TEMPORARY; + pAsm->S[1].src.reg = 0; + setaddrmode_PVSSRC(&(pAsm->S[1].src), ADDR_ABSOLUTE); + noneg_PVSSRC(&(pAsm->S[1].src)); + pAsm->S[1].src.swizzlex = SQ_SEL_1; + pAsm->S[1].src.swizzley = SQ_SEL_1; + pAsm->S[1].src.swizzlez = SQ_SEL_1; + pAsm->S[1].src.swizzlew = SQ_SEL_1; + + if( GL_FALSE == next_ins2(pAsm) ) + { + return GL_FALSE; + } +#endif + + checkStackDepth(pAsm, FC_PUSH_VPM, GL_TRUE); + + return GL_TRUE; +} + +GLboolean returnOnFlag(r700_AssemblerBase *pAsm, GLuint unIF) +{ + testFlag(pAsm); + jumpToOffest(pAsm, 1, 4); + setRetInLoopFlag(pAsm, SQ_SEL_0); + pops(pAsm, unIF + 1); + add_return_inst(pAsm); + + return GL_TRUE; +} + +GLboolean breakLoopOnFlag(r700_AssemblerBase *pAsm, GLuint unFCSP) +{ + testFlag(pAsm); + + //break + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_LOOP_BREAK; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->fc_stack[unFCSP].mid = (R700ControlFlowGenericClause **)_mesa_realloc( + (void *)pAsm->fc_stack[unFCSP].mid, + sizeof(R700ControlFlowGenericClause *) * pAsm->fc_stack[unFCSP].unNumMid, + sizeof(R700ControlFlowGenericClause *) * (pAsm->fc_stack[unFCSP].unNumMid + 1) ); + pAsm->fc_stack[unFCSP].mid[pAsm->fc_stack[unFCSP].unNumMid] = pAsm->cf_current_cf_clause_ptr; + pAsm->fc_stack[unFCSP].unNumMid++; + + pops(pAsm, 1); + + return GL_TRUE; +} + +GLboolean AssembleInstr(GLuint uiFirstInst, + GLuint uiNumberInsts, struct prog_instruction *pILInst, r700_AssemblerBase *pR700AsmCode) { GLuint i; pR700AsmCode->pILInst = pILInst; - for(i=0; i<uiNumberInsts; i++) + for(i=uiFirstInst; i<uiNumberInsts; i++) { pR700AsmCode->uiCurInst = i; +#ifndef USE_CF_FOR_CONTINUE_BREAK + if(OPCODE_BRK == pILInst[i+1].Opcode) + { + switch(pILInst[i].Opcode) + { + case OPCODE_SLE: + pILInst[i].Opcode = OPCODE_SGT; + break; + case OPCODE_SLT: + pILInst[i].Opcode = OPCODE_SGE; + break; + case OPCODE_SGE: + pILInst[i].Opcode = OPCODE_SLT; + break; + case OPCODE_SGT: + pILInst[i].Opcode = OPCODE_SLE; + break; + case OPCODE_SEQ: + pILInst[i].Opcode = OPCODE_SNE; + break; + case OPCODE_SNE: + pILInst[i].Opcode = OPCODE_SEQ; + break; + default: + break; + } + } +#endif + switch (pILInst[i].Opcode) { case OPCODE_ABS: @@ -4350,7 +5902,8 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, if ( GL_FALSE == assemble_FLR(pR700AsmCode) ) return GL_FALSE; break; - //case OP_FLR_INT: + //case OP_FLR_INT: ; + // if ( GL_FALSE == assemble_FLR_INT() ) // return GL_FALSE; // break; @@ -4361,8 +5914,10 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, break; case OPCODE_KIL: - if ( GL_FALSE == assemble_KIL(pR700AsmCode) ) - return GL_FALSE; + case OPCODE_KIL_NV: + /* done at OPCODE_SE/SGT...etc. */ + /* if ( GL_FALSE == assemble_KIL(pR700AsmCode) ) + return GL_FALSE; */ break; case OPCODE_LG2: if ( GL_FALSE == assemble_LG2(pR700AsmCode) ) @@ -4422,16 +5977,340 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, case OPCODE_SCS: if ( GL_FALSE == assemble_SCS(pR700AsmCode) ) return GL_FALSE; - break; + break; + + case OPCODE_SEQ: + if(OPCODE_IF == pILInst[i+1].Opcode) + { + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETE) ) + { + return GL_FALSE; + } + } + else if(OPCODE_BRK == pILInst[i+1].Opcode) + { +#ifdef USE_CF_FOR_CONTINUE_BREAK + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; +#else + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_BREAK; +#endif + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETE) ) + { + return GL_FALSE; + } + } + else if(OPCODE_CONT == pILInst[i+1].Opcode) + { +#ifdef USE_CF_FOR_CONTINUE_BREAK + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; +#else + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_CONTINUE; +#endif + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETE) ) + { + return GL_FALSE; + } + } + else if((OPCODE_KIL == pILInst[i+1].Opcode)||(OPCODE_KIL_NV == pILInst[i+1].Opcode)) + { + if ( GL_FALSE == assemble_KIL(pR700AsmCode, SQ_OP2_INST_KILLE) ) + { + return GL_FALSE; + } + } + else + { + if ( GL_FALSE == assemble_LOGIC(pR700AsmCode, SQ_OP2_INST_SETE) ) + { + return GL_FALSE; + } + } + break; + + case OPCODE_SGT: + if(OPCODE_IF == pILInst[i+1].Opcode) + { + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETGT) ) + { + return GL_FALSE; + } + } + else if(OPCODE_BRK == pILInst[i+1].Opcode) + { +#ifdef USE_CF_FOR_CONTINUE_BREAK + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; +#else + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_BREAK; +#endif + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETGT) ) + { + return GL_FALSE; + } + } + else if(OPCODE_CONT == pILInst[i+1].Opcode) + { +#ifdef USE_CF_FOR_CONTINUE_BREAK + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; +#else + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_CONTINUE; +#endif + + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETGT) ) + { + return GL_FALSE; + } + } + else if((OPCODE_KIL == pILInst[i+1].Opcode)||(OPCODE_KIL_NV == pILInst[i+1].Opcode)) + { + if ( GL_FALSE == assemble_KIL(pR700AsmCode, SQ_OP2_INST_KILLGT) ) + { + return GL_FALSE; + } + } + else + { + if ( GL_FALSE == assemble_LOGIC(pR700AsmCode, SQ_OP2_INST_SETGT) ) + { + return GL_FALSE; + } + } + break; case OPCODE_SGE: - if ( GL_FALSE == assemble_SGE(pR700AsmCode) ) - return GL_FALSE; - break; + if(OPCODE_IF == pILInst[i+1].Opcode) + { + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETGE) ) + { + return GL_FALSE; + } + } + else if(OPCODE_BRK == pILInst[i+1].Opcode) + { +#ifdef USE_CF_FOR_CONTINUE_BREAK + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; +#else + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_BREAK; +#endif + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETGE) ) + { + return GL_FALSE; + } + } + else if(OPCODE_CONT == pILInst[i+1].Opcode) + { +#ifdef USE_CF_FOR_CONTINUE_BREAK + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; +#else + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_CONTINUE; +#endif + + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETGE) ) + { + return GL_FALSE; + } + } + else if((OPCODE_KIL == pILInst[i+1].Opcode)||(OPCODE_KIL_NV == pILInst[i+1].Opcode)) + { + if ( GL_FALSE == assemble_KIL(pR700AsmCode, SQ_OP2_INST_KILLGE) ) + { + return GL_FALSE; + } + } + else + { + if ( GL_FALSE == assemble_SGE(pR700AsmCode) ) + { + return GL_FALSE; + } + } + break; + + /* NO LT, LE, TODO : use GE => LE, GT => LT : reverse 2 src order would be simpliest. Or use SQ_CF_COND_FALSE for SQ_CF_COND_ACTIVE.*/ case OPCODE_SLT: - if ( GL_FALSE == assemble_SLT(pR700AsmCode) ) - return GL_FALSE; - break; + { + struct prog_src_register SrcRegSave[2]; + SrcRegSave[0] = pILInst[i].SrcReg[0]; + SrcRegSave[1] = pILInst[i].SrcReg[1]; + pILInst[i].SrcReg[0] = SrcRegSave[1]; + pILInst[i].SrcReg[1] = SrcRegSave[0]; + if(OPCODE_IF == pILInst[i+1].Opcode) + { + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETGT) ) + { + pILInst[i].SrcReg[0] = SrcRegSave[0]; + pILInst[i].SrcReg[1] = SrcRegSave[1]; + return GL_FALSE; + } + } + else if(OPCODE_BRK == pILInst[i+1].Opcode) + { +#ifdef USE_CF_FOR_CONTINUE_BREAK + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; +#else + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_BREAK; +#endif + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETGT) ) + { + pILInst[i].SrcReg[0] = SrcRegSave[0]; + pILInst[i].SrcReg[1] = SrcRegSave[1]; + return GL_FALSE; + } + } + else if(OPCODE_CONT == pILInst[i+1].Opcode) + { +#ifdef USE_CF_FOR_CONTINUE_BREAK + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; +#else + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_CONTINUE; +#endif + + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETGT) ) + { + pILInst[i].SrcReg[0] = SrcRegSave[0]; + pILInst[i].SrcReg[1] = SrcRegSave[1]; + return GL_FALSE; + } + } + else if((OPCODE_KIL == pILInst[i+1].Opcode)||(OPCODE_KIL_NV == pILInst[i+1].Opcode)) + { + if ( GL_FALSE == assemble_KIL(pR700AsmCode, SQ_OP2_INST_KILLGT) ) + { + return GL_FALSE; + } + } + else + { + if ( GL_FALSE == assemble_LOGIC(pR700AsmCode, SQ_OP2_INST_SETGT) ) + { + pILInst[i].SrcReg[0] = SrcRegSave[0]; + pILInst[i].SrcReg[1] = SrcRegSave[1]; + return GL_FALSE; + } + } + pILInst[i].SrcReg[0] = SrcRegSave[0]; + pILInst[i].SrcReg[1] = SrcRegSave[1]; + } + break; + + case OPCODE_SLE: + { + struct prog_src_register SrcRegSave[2]; + SrcRegSave[0] = pILInst[i].SrcReg[0]; + SrcRegSave[1] = pILInst[i].SrcReg[1]; + pILInst[i].SrcReg[0] = SrcRegSave[1]; + pILInst[i].SrcReg[1] = SrcRegSave[0]; + if(OPCODE_IF == pILInst[i+1].Opcode) + { + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETGE) ) + { + pILInst[i].SrcReg[0] = SrcRegSave[0]; + pILInst[i].SrcReg[1] = SrcRegSave[1]; + return GL_FALSE; + } + } + else if(OPCODE_BRK == pILInst[i+1].Opcode) + { +#ifdef USE_CF_FOR_CONTINUE_BREAK + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; +#else + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_BREAK; +#endif + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETGE) ) + { + pILInst[i].SrcReg[0] = SrcRegSave[0]; + pILInst[i].SrcReg[1] = SrcRegSave[1]; + return GL_FALSE; + } + } + else if(OPCODE_CONT == pILInst[i+1].Opcode) + { +#ifdef USE_CF_FOR_CONTINUE_BREAK + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; +#else + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_CONTINUE; +#endif + + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETGE) ) + { + pILInst[i].SrcReg[0] = SrcRegSave[0]; + pILInst[i].SrcReg[1] = SrcRegSave[1]; + return GL_FALSE; + } + } + else if((OPCODE_KIL == pILInst[i+1].Opcode)||(OPCODE_KIL_NV == pILInst[i+1].Opcode)) + { + if ( GL_FALSE == assemble_KIL(pR700AsmCode, SQ_OP2_INST_KILLGE) ) + { + return GL_FALSE; + } + } + else + { + if ( GL_FALSE == assemble_LOGIC(pR700AsmCode, SQ_OP2_INST_SETGE) ) + { + pILInst[i].SrcReg[0] = SrcRegSave[0]; + pILInst[i].SrcReg[1] = SrcRegSave[1]; + return GL_FALSE; + } + } + pILInst[i].SrcReg[0] = SrcRegSave[0]; + pILInst[i].SrcReg[1] = SrcRegSave[1]; + } + break; + + case OPCODE_SNE: + if(OPCODE_IF == pILInst[i+1].Opcode) + { + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETNE) ) + { + return GL_FALSE; + } + } + else if(OPCODE_BRK == pILInst[i+1].Opcode) + { +#ifdef USE_CF_FOR_CONTINUE_BREAK + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; +#else + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_BREAK; +#endif + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETNE) ) + { + return GL_FALSE; + } + } + else if(OPCODE_CONT == pILInst[i+1].Opcode) + { +#ifdef USE_CF_FOR_CONTINUE_BREAK + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; +#else + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_CONTINUE; +#endif + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETNE) ) + { + return GL_FALSE; + } + } + else if((OPCODE_KIL == pILInst[i+1].Opcode)||(OPCODE_KIL_NV == pILInst[i+1].Opcode)) + { + if ( GL_FALSE == assemble_KIL(pR700AsmCode, SQ_OP2_INST_KILLNE) ) + { + return GL_FALSE; + } + } + else + { + if ( GL_FALSE == assemble_LOGIC(pR700AsmCode, SQ_OP2_INST_SETNE) ) + { + return GL_FALSE; + } + } + break; //case OP_STP: // if ( GL_FALSE == assemble_STP(pR700AsmCode) ) @@ -4465,30 +6344,102 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, return GL_FALSE; break; + case OPCODE_TRUNC: + if ( GL_FALSE == assemble_math_function(pR700AsmCode, SQ_OP2_INST_TRUNC) ) + return GL_FALSE; + break; + case OPCODE_XPD: if ( GL_FALSE == assemble_XPD(pR700AsmCode) ) return GL_FALSE; break; case OPCODE_IF : - if ( GL_FALSE == assemble_IF(pR700AsmCode) ) - return GL_FALSE; + { + GLboolean bHasElse = GL_FALSE; + + if(pILInst[pILInst[i].BranchTarget - 1].Opcode == OPCODE_ELSE) + { + bHasElse = GL_TRUE; + } + + if ( GL_FALSE == assemble_IF(pR700AsmCode, bHasElse) ) + { + return GL_FALSE; + } + } break; + case OPCODE_ELSE : - radeon_error("Not yet implemented instruction OPCODE_ELSE \n"); - //if ( GL_FALSE == assemble_BAD("ELSE") ) + if ( GL_FALSE == assemble_ELSE(pR700AsmCode) ) return GL_FALSE; break; + case OPCODE_ENDIF: if ( GL_FALSE == assemble_ENDIF(pR700AsmCode) ) return GL_FALSE; break; + case OPCODE_BGNLOOP: + if( GL_FALSE == assemble_BGNLOOP(pR700AsmCode) ) + { + return GL_FALSE; + } + break; + + case OPCODE_BRK: + if( GL_FALSE == assemble_BRK(pR700AsmCode) ) + { + return GL_FALSE; + } + break; + + case OPCODE_CONT: + if( GL_FALSE == assemble_CONT(pR700AsmCode) ) + { + return GL_FALSE; + } + break; + + case OPCODE_ENDLOOP: + if( GL_FALSE == assemble_ENDLOOP(pR700AsmCode) ) + { + return GL_FALSE; + } + break; + + case OPCODE_BGNSUB: + if( GL_FALSE == assemble_BGNSUB(pR700AsmCode, i) ) + { + return GL_FALSE; + } + break; + + case OPCODE_RET: + if( GL_FALSE == assemble_RET(pR700AsmCode) ) + { + return GL_FALSE; + } + break; + + case OPCODE_CAL: + if( GL_FALSE == assemble_CAL(pR700AsmCode, + pILInst[i].BranchTarget, + uiNumberInsts, + pILInst) ) + { + return GL_FALSE; + } + break; + //case OPCODE_EXPORT: // if ( GL_FALSE == assemble_EXPORT() ) // return GL_FALSE; // break; + case OPCODE_ENDSUB: + return assemble_ENDSUB(pR700AsmCode); + case OPCODE_END: //pR700AsmCode->uiCurInst = i; //This is to remaind that if in later exoort there is depth/stencil @@ -4505,6 +6456,123 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, return GL_TRUE; } +GLboolean InitShaderProgram(r700_AssemblerBase * pAsm) +{ + setRetInLoopFlag(pAsm, SQ_SEL_0); + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + return GL_TRUE; +} + +GLboolean RelocProgram(r700_AssemblerBase * pAsm) +{ + GLuint i; + GLuint unCFoffset; + TypedShaderList * plstCFmain; + TypedShaderList * plstCFsub; + + R700ShaderInstruction * pInst; + R700ControlFlowGenericClause * pCFInst; + + plstCFmain = pAsm->CALLSTACK[0].plstCFInstructions_local; + + /* remove flags init if they are not used */ + if((pAsm->unCFflags & HAS_LOOPRET) == 0) + { + R700ControlFlowALUClause * pCF_ALU; + pInst = plstCFmain->pHead; + while(pInst) + { + if(SIT_CF_ALU == pInst->m_ShaderInstType) + { + pCF_ALU = (R700ControlFlowALUClause *)pInst; + if(0 == pCF_ALU->m_Word1.f.count) + { + pCF_ALU->m_Word1.f.cf_inst = SQ_CF_INST_NOP; + } + else + { + R700ALUInstruction * pALU = pCF_ALU->m_pLinkedALUInstruction; + + pALU->m_pLinkedALUClause = NULL; + pALU = (R700ALUInstruction *)(pALU->pNextInst); + pALU->m_pLinkedALUClause = pCF_ALU; + pCF_ALU->m_pLinkedALUInstruction = pALU; + + pCF_ALU->m_Word1.f.count--; + } + break; + } + pInst = pInst->pNextInst; + }; + } + + if(pAsm->CALLSTACK[0].max > 0) + { + pAsm->pR700Shader->uStackSize = ((pAsm->CALLSTACK[0].max + 3)>>2) + 2; + } + + if(0 == pAsm->unSubArrayPointer) + { + return GL_TRUE; + } + + unCFoffset = plstCFmain->uNumOfNode; + + /* Reloc subs */ + for(i=0; i<pAsm->unSubArrayPointer; i++) + { + pAsm->subs[i].unCFoffset = unCFoffset; + plstCFsub = &(pAsm->subs[i].lstCFInstructions_local); + + pInst = plstCFsub->pHead; + + /* reloc instructions */ + while(pInst) + { + if(SIT_CF_GENERIC == pInst->m_ShaderInstType) + { + pCFInst = (R700ControlFlowGenericClause *)pInst; + + switch (pCFInst->m_Word1.f.cf_inst) + { + case SQ_CF_INST_POP: + case SQ_CF_INST_JUMP: + case SQ_CF_INST_ELSE: + case SQ_CF_INST_LOOP_END: + case SQ_CF_INST_LOOP_START: + case SQ_CF_INST_LOOP_START_NO_AL: + case SQ_CF_INST_LOOP_CONTINUE: + case SQ_CF_INST_LOOP_BREAK: + pCFInst->m_Word0.f.addr += unCFoffset; + break; + default: + break; + } + } + + pInst->m_uIndex += unCFoffset; + + pInst = pInst->pNextInst; + }; + + /* Put sub into main */ + plstCFmain->pTail->pNextInst = plstCFsub->pHead; + plstCFmain->pTail = plstCFsub->pTail; + plstCFmain->uNumOfNode += plstCFsub->uNumOfNode; + + unCFoffset += plstCFsub->uNumOfNode; + } + + /* reloc callers */ + for(i=0; i<pAsm->unCallerArrayPointer; i++) + { + pAsm->callers[i].cf_ptr->m_Word0.f.addr + = pAsm->subs[pAsm->callers[i].subDescIndex].unCFoffset; + } + + return GL_TRUE; +} + GLboolean Process_Export(r700_AssemblerBase* pAsm, GLuint type, GLuint export_starting_index, @@ -4800,6 +6868,25 @@ GLboolean Process_Vertex_Exports(r700_AssemblerBase *pR700AsmCode, } } + for(i=VERT_RESULT_VAR0; i<VERT_RESULT_MAX; i++) + { + unBit = 1 << i; + if(OutputsWritten & unBit) + { + if( GL_FALSE == Process_Export(pR700AsmCode, + SQ_EXPORT_PARAM, + export_starting_index, + 1, + pR700AsmCode->ucVP_OutputMap[i], + GL_FALSE) ) + { + return GL_FALSE; + } + + export_starting_index++; + } + } + // At least one param should be exported if (export_count) { @@ -4833,6 +6920,16 @@ GLboolean Clean_Up_Assembler(r700_AssemblerBase *pR700AsmCode) { FREE(pR700AsmCode->pucOutMask); FREE(pR700AsmCode->pInstDeps); + + if(NULL != pR700AsmCode->subs) + { + FREE(pR700AsmCode->subs); + } + if(NULL != pR700AsmCode->callers) + { + FREE(pR700AsmCode->callers); + } + return GL_TRUE; } diff --git a/src/mesa/drivers/dri/r600/r700_assembler.h b/src/mesa/drivers/dri/r600/r700_assembler.h index c66db502a1..130fc89dae 100644 --- a/src/mesa/drivers/dri/r600/r700_assembler.h +++ b/src/mesa/drivers/dri/r600/r700_assembler.h @@ -72,7 +72,8 @@ typedef enum SrcRegisterType SRC_REG_INPUT = 1, SRC_REG_CONSTANT = 2, SRC_REG_ALT_TEMPORARY = 3, - NUMBER_OF_SRC_REG_TYPE = 4 + SRC_REC_LITERAL = 4, + NUMBER_OF_SRC_REG_TYPE = 5 } SrcRegisterType; typedef enum DstRegisterType @@ -111,6 +112,12 @@ typedef struct PVSDSTtag BITS addrmode1:1; //32 } PVSDST; +typedef struct PVSINSTtag +{ + BITS literal :2; + BITS SaturateMode :2; +} PVSINST; + typedef struct PVSSRCtag { BITS rtype:4; @@ -148,6 +155,7 @@ typedef union PVSDWORDtag { BITS bits; PVSDST dst; + PVSINST dst2; PVSSRC src; PVSMATH math; float f; @@ -251,6 +259,8 @@ enum FC_IF = 1, FC_LOOP = 2, FC_REP = 3, + FC_PUSH_VPM = 4, + FC_PUSH_WQM = 5, COND_NONE = 0, COND_BOOL = 1, @@ -263,22 +273,52 @@ enum typedef struct FC_LEVEL { - unsigned int first; ///< first fc instruction on level (if, rep, loop) - unsigned int* mid; ///< middle instructions - else or all breaks on this level - unsigned int midLen; - unsigned int type; - unsigned int cond; - unsigned int inv; - unsigned int bpush; ///< 1 if first instruction does branch stack push - int id; ///< id of bool or int variable + R700ControlFlowGenericClause * first; + R700ControlFlowGenericClause ** mid; + unsigned int unNumMid; + unsigned int midLen; + unsigned int type; + unsigned int cond; + unsigned int inv; + int id; ///< id of bool or int variable } FC_LEVEL; typedef struct VTX_FETCH_METHOD { - GLboolean bEnableMini; - GLuint mega_fetch_remainder; + GLboolean bEnableMini; + GLuint mega_fetch_remainder; } VTX_FETCH_METHOD; +typedef struct SUB_OFFSET +{ + GLint subIL_Offset; + GLuint unCFoffset; + GLuint unStackDepthMax; + TypedShaderList lstCFInstructions_local; +} SUB_OFFSET; + +typedef struct CALLER_POINTER +{ + GLint subIL_Offset; + GLint subDescIndex; + R700ControlFlowGenericClause* cf_ptr; +} CALLER_POINTER; + +#define SQ_MAX_CALL_DEPTH 0x00000020 + +typedef struct CALL_LEVEL +{ + unsigned int FCSP_BeforeEntry; + GLint subDescIndex; + GLushort current; + GLushort max; + TypedShaderList * plstCFInstructions_local; +} CALL_LEVEL; + +#define HAS_CURRENT_LOOPRET 0x1L +#define HAS_LOOPRET 0x2L +#define LOOPRET_FLAGS HAS_LOOPRET | HAS_CURRENT_LOOPRET + typedef struct r700_AssemblerBase { R700ControlFlowSXClause* cf_last_export_ptr; @@ -294,11 +334,14 @@ typedef struct r700_AssemblerBase // No clause has been created yet CF_CLAUSE_TYPE cf_current_clause_type; + BITS alu_x_opcode; + GLuint number_of_exports; GLuint number_of_colorandz_exports; GLuint number_of_export_opcodes; PVSDWORD D; + PVSDWORD D2; PVSDWORD S[3]; unsigned int uLastPosUpdate; @@ -310,6 +353,8 @@ typedef struct r700_AssemblerBase unsigned int number_used_registers; unsigned int uUsedConsts; + unsigned int flag_reg_index; + // Fragment programs unsigned int uiFP_AttributeMap[FRAG_ATTRIB_MAX]; unsigned int uiFP_OutputMap[FRAG_RESULT_MAX]; @@ -330,9 +375,6 @@ typedef struct r700_AssemblerBase unsigned int FCSP; FC_LEVEL fc_stack[32]; - unsigned int branch_depth; - unsigned int max_branch_depth; - //----------------------------------------------------------------------------------- // ArgSubst used in Assemble_Source() function //----------------------------------------------------------------------------------- @@ -378,6 +420,18 @@ typedef struct r700_AssemblerBase GLboolean is_tex; /* we inserted helper intructions and need barrier on next TEX ins */ GLboolean need_tex_barrier; + + SUB_OFFSET * subs; + GLuint unSubArraySize; + GLuint unSubArrayPointer; + CALLER_POINTER * callers; + GLuint unCallerArraySize; + GLuint unCallerArrayPointer; + unsigned int CALLSP; + CALL_LEVEL CALLSTACK[SQ_MAX_CALL_DEPTH]; + + GLuint unCFflags; + } r700_AssemblerBase; //Internal use @@ -446,6 +500,10 @@ GLboolean assemble_alu_src(R700ALUInstruction* alu_instruction_ptr, GLboolean add_alu_instruction(r700_AssemblerBase* pAsm, R700ALUInstruction* alu_instruction_ptr, GLuint contiguous_slots_needed); + +GLboolean add_cf_instruction(r700_AssemblerBase* pAsm); +void add_return_inst(r700_AssemblerBase *pAsm); + void get_src_properties(R700ALUInstruction* alu_instruction_ptr, int source_index, BITS* psrc_sel, @@ -467,6 +525,21 @@ GLboolean check_vector(r700_AssemblerBase* pAsm, R700ALUInstruction* alu_instruction_ptr); GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm); GLboolean next_ins(r700_AssemblerBase *pAsm); + +GLboolean next_ins2(r700_AssemblerBase *pAsm); +GLboolean assemble_alu_instruction2(r700_AssemblerBase *pAsm); + +/* TODO : merge next_ins/2/literal, assemble_alu_instruction/2/literal */ +GLboolean next_ins_literal(r700_AssemblerBase *pAsm, GLfloat * pLiteral); +GLboolean assemble_alu_instruction_literal(r700_AssemblerBase *pAsm, GLfloat * pLiteral); + +GLboolean pops(r700_AssemblerBase *pAsm, GLuint pops); +GLboolean jumpToOffest(r700_AssemblerBase *pAsm, GLuint pops, GLint offset); +GLboolean setRetInLoopFlag(r700_AssemblerBase *pAsm, GLuint flagValue); +GLboolean testFlag(r700_AssemblerBase *pAsm); +GLboolean breakLoopOnFlag(r700_AssemblerBase *pAsm, GLuint unFCSP); +GLboolean returnOnFlag(r700_AssemblerBase *pAsm, GLuint unIF); + GLboolean assemble_math_function(r700_AssemblerBase* pAsm, BITS opcode); GLboolean assemble_ABS(r700_AssemblerBase *pAsm); GLboolean assemble_ADD(r700_AssemblerBase *pAsm); @@ -481,7 +554,7 @@ GLboolean assemble_EXP(r700_AssemblerBase *pAsm); GLboolean assemble_FLR(r700_AssemblerBase *pAsm); GLboolean assemble_FLR_INT(r700_AssemblerBase *pAsm); GLboolean assemble_FRC(r700_AssemblerBase *pAsm); -GLboolean assemble_KIL(r700_AssemblerBase *pAsm); +GLboolean assemble_KIL(r700_AssemblerBase *pAsm, GLuint opcode); GLboolean assemble_LG2(r700_AssemblerBase *pAsm); GLboolean assemble_LRP(r700_AssemblerBase *pAsm); GLboolean assemble_LOG(r700_AssemblerBase *pAsm); @@ -497,14 +570,32 @@ GLboolean assemble_RSQ(r700_AssemblerBase *pAsm); GLboolean assemble_SIN(r700_AssemblerBase *pAsm); GLboolean assemble_SCS(r700_AssemblerBase *pAsm); GLboolean assemble_SGE(r700_AssemblerBase *pAsm); + +GLboolean assemble_LOGIC(r700_AssemblerBase *pAsm, BITS opcode); +GLboolean assemble_LOGIC_PRED(r700_AssemblerBase *pAsm, BITS opcode); + GLboolean assemble_SLT(r700_AssemblerBase *pAsm); GLboolean assemble_STP(r700_AssemblerBase *pAsm); GLboolean assemble_TEX(r700_AssemblerBase *pAsm); GLboolean assemble_XPD(r700_AssemblerBase *pAsm); GLboolean assemble_EXPORT(r700_AssemblerBase *pAsm); -GLboolean assemble_IF(r700_AssemblerBase *pAsm); +GLboolean assemble_IF(r700_AssemblerBase *pAsm, GLboolean bHasElse); +GLboolean assemble_ELSE(r700_AssemblerBase *pAsm); GLboolean assemble_ENDIF(r700_AssemblerBase *pAsm); +GLboolean assemble_BGNLOOP(r700_AssemblerBase *pAsm); +GLboolean assemble_BRK(r700_AssemblerBase *pAsm); +GLboolean assemble_COND(r700_AssemblerBase *pAsm); +GLboolean assemble_ENDLOOP(r700_AssemblerBase *pAsm); + +GLboolean assemble_BGNSUB(r700_AssemblerBase *pAsm, GLint nILindex); +GLboolean assemble_ENDSUB(r700_AssemblerBase *pAsm); +GLboolean assemble_RET(r700_AssemblerBase *pAsm); +GLboolean assemble_CAL(r700_AssemblerBase *pAsm, + GLint nILindex, + GLuint uiNumberInsts, + struct prog_instruction *pILInst); + GLboolean Process_Export(r700_AssemblerBase* pAsm, GLuint type, GLuint export_starting_index, @@ -516,12 +607,16 @@ GLboolean Move_Depth_Exports_To_Correct_Channels(r700_AssemblerBase *pAsm, //Interface -GLboolean AssembleInstr(GLuint uiNumberInsts, +GLboolean AssembleInstr(GLuint uiFirstInst, + GLuint uiNumberInsts, struct prog_instruction *pILInst, r700_AssemblerBase *pR700AsmCode); GLboolean Process_Fragment_Exports(r700_AssemblerBase *pR700AsmCode, GLbitfield OutputsWritten); GLboolean Process_Vertex_Exports(r700_AssemblerBase *pR700AsmCode, GLbitfield OutputsWritten); +GLboolean RelocProgram(r700_AssemblerBase * pAsm); +GLboolean InitShaderProgram(r700_AssemblerBase * pAsm); + int Init_r700_AssemblerBase(SHADER_PIPE_TYPE spt, r700_AssemblerBase* pAsm, R700_Shader* pShader); GLboolean Clean_Up_Assembler(r700_AssemblerBase *pR700AsmCode); diff --git a/src/mesa/drivers/dri/r600/r700_chip.c b/src/mesa/drivers/dri/r600/r700_chip.c index ec76fbcb6d..8126777bf4 100644 --- a/src/mesa/drivers/dri/r600/r700_chip.c +++ b/src/mesa/drivers/dri/r600/r700_chip.c @@ -54,11 +54,15 @@ static void r700SendTexState(GLcontext *ctx, struct radeon_state_atom *atom) for (i = 0; i < R700_TEXTURE_NUMBERUNITS; i++) { if (ctx->Texture.Unit[i]._ReallyEnabled) { radeonTexObj *t = r700->textures[i]; + uint32_t offset; if (t) { - if (!t->image_override) + if (!t->image_override) { bo = t->mt->bo; - else + offset = get_base_teximage_offset(t); + } else { bo = t->bo; + offset = 0; + } if (bo) { r700SyncSurf(context, bo, @@ -77,7 +81,7 @@ static void r700SendTexState(GLcontext *ctx, struct radeon_state_atom *atom) R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE6); R600_OUT_BATCH_RELOC(r700->textures[i]->SQ_TEX_RESOURCE2, bo, - 0, + offset, RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0); R600_OUT_BATCH_RELOC(r700->textures[i]->SQ_TEX_RESOURCE3, bo, @@ -442,68 +446,77 @@ static void r700SendRenderTargetState(GLcontext *ctx, struct radeon_state_atom * static void r700SendPSState(GLcontext *ctx, struct radeon_state_atom *atom) { - context_t *context = R700_CONTEXT(ctx); - R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); - struct radeon_bo * pbo; - BATCH_LOCALS(&context->radeon); - radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); + context_t *context = R700_CONTEXT(ctx); + R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); + struct radeon_bo * pbo; + BATCH_LOCALS(&context->radeon); + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); - pbo = (struct radeon_bo *)r700GetActiveFpShaderBo(GL_CONTEXT(context)); + pbo = (struct radeon_bo *)r700GetActiveFpShaderBo(GL_CONTEXT(context)); - if (!pbo) - return; + if (!pbo) + return; - r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit); + r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit); - BEGIN_BATCH_NO_AUTOSTATE(3 + 2); - R600_OUT_BATCH_REGSEQ(SQ_PGM_START_PS, 1); - R600_OUT_BATCH(r700->ps.SQ_PGM_START_PS.u32All); - R600_OUT_BATCH_RELOC(r700->ps.SQ_PGM_START_PS.u32All, - pbo, - r700->ps.SQ_PGM_START_PS.u32All, - RADEON_GEM_DOMAIN_GTT, 0, 0); - END_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(3 + 2); + R600_OUT_BATCH_REGSEQ(SQ_PGM_START_PS, 1); + R600_OUT_BATCH(r700->ps.SQ_PGM_START_PS.u32All); + R600_OUT_BATCH_RELOC(r700->ps.SQ_PGM_START_PS.u32All, + pbo, + r700->ps.SQ_PGM_START_PS.u32All, + RADEON_GEM_DOMAIN_GTT, 0, 0); + END_BATCH(); - BEGIN_BATCH_NO_AUTOSTATE(9); - R600_OUT_BATCH_REGVAL(SQ_PGM_RESOURCES_PS, r700->ps.SQ_PGM_RESOURCES_PS.u32All); - R600_OUT_BATCH_REGVAL(SQ_PGM_EXPORTS_PS, r700->ps.SQ_PGM_EXPORTS_PS.u32All); - R600_OUT_BATCH_REGVAL(SQ_PGM_CF_OFFSET_PS, r700->ps.SQ_PGM_CF_OFFSET_PS.u32All); - END_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(9); + R600_OUT_BATCH_REGVAL(SQ_PGM_RESOURCES_PS, r700->ps.SQ_PGM_RESOURCES_PS.u32All); + R600_OUT_BATCH_REGVAL(SQ_PGM_EXPORTS_PS, r700->ps.SQ_PGM_EXPORTS_PS.u32All); + R600_OUT_BATCH_REGVAL(SQ_PGM_CF_OFFSET_PS, r700->ps.SQ_PGM_CF_OFFSET_PS.u32All); + END_BATCH(); - COMMIT_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(3); + R600_OUT_BATCH_REGVAL(SQ_LOOP_CONST_0, 0x01000FFF); + END_BATCH(); + + COMMIT_BATCH(); } static void r700SendVSState(GLcontext *ctx, struct radeon_state_atom *atom) { - context_t *context = R700_CONTEXT(ctx); - R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); - struct radeon_bo * pbo; - BATCH_LOCALS(&context->radeon); - radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); + context_t *context = R700_CONTEXT(ctx); + R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); + struct radeon_bo * pbo; + BATCH_LOCALS(&context->radeon); + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); - pbo = (struct radeon_bo *)r700GetActiveVpShaderBo(GL_CONTEXT(context)); + pbo = (struct radeon_bo *)r700GetActiveVpShaderBo(GL_CONTEXT(context)); - if (!pbo) - return; + if (!pbo) + return; - r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit); + r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit); - BEGIN_BATCH_NO_AUTOSTATE(3 + 2); - R600_OUT_BATCH_REGSEQ(SQ_PGM_START_VS, 1); - R600_OUT_BATCH(r700->vs.SQ_PGM_START_VS.u32All); - R600_OUT_BATCH_RELOC(r700->vs.SQ_PGM_START_VS.u32All, - pbo, - r700->vs.SQ_PGM_START_VS.u32All, - RADEON_GEM_DOMAIN_GTT, 0, 0); - END_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(3 + 2); + R600_OUT_BATCH_REGSEQ(SQ_PGM_START_VS, 1); + R600_OUT_BATCH(r700->vs.SQ_PGM_START_VS.u32All); + R600_OUT_BATCH_RELOC(r700->vs.SQ_PGM_START_VS.u32All, + pbo, + r700->vs.SQ_PGM_START_VS.u32All, + RADEON_GEM_DOMAIN_GTT, 0, 0); + END_BATCH(); - BEGIN_BATCH_NO_AUTOSTATE(6); - R600_OUT_BATCH_REGVAL(SQ_PGM_RESOURCES_VS, r700->vs.SQ_PGM_RESOURCES_VS.u32All); - R600_OUT_BATCH_REGVAL(SQ_PGM_CF_OFFSET_VS, r700->vs.SQ_PGM_CF_OFFSET_VS.u32All); - END_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(6); + R600_OUT_BATCH_REGVAL(SQ_PGM_RESOURCES_VS, r700->vs.SQ_PGM_RESOURCES_VS.u32All); + R600_OUT_BATCH_REGVAL(SQ_PGM_CF_OFFSET_VS, r700->vs.SQ_PGM_CF_OFFSET_VS.u32All); + END_BATCH(); - COMMIT_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(3); + R600_OUT_BATCH_REGVAL((SQ_LOOP_CONST_0 + 32*4), 0x0100000F); + //R600_OUT_BATCH_REGVAL((SQ_LOOP_CONST_0 + (SQ_LOOP_CONST_vs<2)), 0x0100000F); + END_BATCH(); + + COMMIT_BATCH(); } static void r700SendFSState(GLcontext *ctx, struct radeon_state_atom *atom) @@ -1301,8 +1314,8 @@ void r600InitAtoms(context_t *context) ALLOC_STATE(spi, always, (59 + R700_MAX_SHADER_EXPORTS), r700SendSPIState); ALLOC_STATE(vpt, always, 16, r700SendViewportState); ALLOC_STATE(fs, always, 18, r700SendFSState); - ALLOC_STATE(vs, always, 18, r700SendVSState); - ALLOC_STATE(ps, always, 21, r700SendPSState); + ALLOC_STATE(vs, always, 21, r700SendVSState); + ALLOC_STATE(ps, always, 24, r700SendPSState); ALLOC_STATE(vs_consts, vs_consts, (2 + (R700_MAX_DX9_CONSTS * 4)), r700SendVSConsts); ALLOC_STATE(ps_consts, ps_consts, (2 + (R700_MAX_DX9_CONSTS * 4)), r700SendPSConsts); ALLOC_STATE(vtx, vtx, (6 + (VERT_ATTRIB_MAX * 18)), r700SendVTXState); diff --git a/src/mesa/drivers/dri/r600/r700_fragprog.c b/src/mesa/drivers/dri/r600/r700_fragprog.c index ccafd433bf..21ac46e7b8 100644 --- a/src/mesa/drivers/dri/r600/r700_fragprog.c +++ b/src/mesa/drivers/dri/r600/r700_fragprog.c @@ -73,11 +73,11 @@ void Map_Fragment_Program(r700_AssemblerBase *pAsm, pAsm->uiFP_AttributeMap[FRAG_ATTRIB_COL1] = pAsm->number_used_registers++; } - unBit = 1 << FRAG_ATTRIB_FOGC; - if(mesa_fp->Base.InputsRead & unBit) - { - pAsm->uiFP_AttributeMap[FRAG_ATTRIB_FOGC] = pAsm->number_used_registers++; - } + unBit = 1 << FRAG_ATTRIB_FOGC; + if(mesa_fp->Base.InputsRead & unBit) + { + pAsm->uiFP_AttributeMap[FRAG_ATTRIB_FOGC] = pAsm->number_used_registers++; + } for(i=0; i<8; i++) { @@ -88,6 +88,62 @@ void Map_Fragment_Program(r700_AssemblerBase *pAsm, } } +/* order has been taken care of */ +#if 1 + for(i=FRAG_ATTRIB_VAR0; i<FRAG_ATTRIB_MAX; i++) + { + unBit = 1 << i; + if(mesa_fp->Base.InputsRead & unBit) + { + pAsm->uiFP_AttributeMap[i] = pAsm->number_used_registers++; + } + } +#else + if( (mesa_fp->Base.InputsRead >> FRAG_ATTRIB_VAR0) > 0 ) + { + struct r700_vertex_program_cont *vpc = + (struct r700_vertex_program_cont *)ctx->VertexProgram._Current; + struct gl_program_parameter_list * VsVarying = vpc->mesa_program.Base.Varying; + struct gl_program_parameter_list * PsVarying = mesa_fp->Base.Varying; + struct gl_program_parameter * pVsParam; + struct gl_program_parameter * pPsParam; + GLuint j, k; + GLuint unMaxVarying = 0; + + for(i=0; i<VsVarying->NumParameters; i++) + { + pAsm->uiFP_AttributeMap[i + FRAG_ATTRIB_VAR0] = 0; + } + + for(i=FRAG_ATTRIB_VAR0; i<FRAG_ATTRIB_MAX; i++) + { + unBit = 1 << i; + if(mesa_fp->Base.InputsRead & unBit) + { + j = i - FRAG_ATTRIB_VAR0; + pPsParam = PsVarying->Parameters + j; + + for(k=0; k<VsVarying->NumParameters; k++) + { + pVsParam = VsVarying->Parameters + k; + + if( strcmp(pPsParam->Name, pVsParam->Name) == 0) + { + pAsm->uiFP_AttributeMap[i] = pAsm->number_used_registers + k; + if(k > unMaxVarying) + { + unMaxVarying = k; + } + break; + } + } + } + } + + pAsm->number_used_registers += unMaxVarying + 1; + } +#endif + /* Map temporary registers (GPRs) */ pAsm->starting_temp_register_number = pAsm->number_used_registers; @@ -127,6 +183,8 @@ void Map_Fragment_Program(r700_AssemblerBase *pAsm, pAsm->pucOutMask[ui] = 0x0; } + pAsm->flag_reg_index = pAsm->number_used_registers++; + pAsm->uFirstHelpReg = pAsm->number_used_registers; } @@ -247,8 +305,11 @@ GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp, { return GL_FALSE; } + + InitShaderProgram(&(fp->r700AsmCode)); - if( GL_FALSE == AssembleInstr(mesa_fp->Base.NumInstructions, + if( GL_FALSE == AssembleInstr(0, + mesa_fp->Base.NumInstructions, &(mesa_fp->Base.Instructions[0]), &(fp->r700AsmCode)) ) { @@ -260,6 +321,11 @@ GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp, return GL_FALSE; } + if( GL_FALSE == RelocProgram(&(fp->r700AsmCode)) ) + { + return GL_FALSE; + } + fp->r700Shader.nRegs = (fp->r700AsmCode.number_used_registers == 0) ? 0 : (fp->r700AsmCode.number_used_registers - 1); @@ -459,6 +525,22 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) } } + for(i=FRAG_ATTRIB_VAR0; i<FRAG_ATTRIB_MAX; i++) + { + unBit = 1 << i; + if(mesa_fp->Base.InputsRead & unBit) + { + ui = pAsm->uiFP_AttributeMap[i]; + SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, SEL_CENTROID_bit); + SETfield(r700->SPI_PS_INPUT_CNTL[ui].u32All, ui, + SEMANTIC_shift, SEMANTIC_mask); + if (r700->SPI_INTERP_CONTROL_0.u32All & FLAT_SHADE_ENA_bit) + SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); + else + CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); + } + } + exportCount = (r700->ps.SQ_PGM_EXPORTS_PS.u32All & EXPORT_MODE_mask) / (1 << EXPORT_MODE_shift); if (r700->CB_SHADER_CONTROL.u32All != ((1 << exportCount) - 1)) { diff --git a/src/mesa/drivers/dri/r600/r700_render.c b/src/mesa/drivers/dri/r600/r700_render.c index 47f89c91f8..eab27cbd84 100644 --- a/src/mesa/drivers/dri/r600/r700_render.c +++ b/src/mesa/drivers/dri/r600/r700_render.c @@ -526,6 +526,9 @@ static void r700ConvertAttrib(GLcontext *ctx, int count, radeonAllocDmaRegion(&context->radeon, &attr->bo, &attr->bo_offset, sizeof(GLfloat) * input->Size * count, 32); + + radeon_bo_map(attr->bo, 1); + dst_ptr = (GLfloat *)ADD_POINTERS(attr->bo->ptr, attr->bo_offset); assert(src_ptr != NULL); @@ -559,6 +562,8 @@ static void r700ConvertAttrib(GLcontext *ctx, int count, break; } + radeon_bo_unmap(attr->bo); + if (mapped_named_bo) { ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER, input->BufferObj); @@ -577,6 +582,8 @@ static void r700AlignDataToDword(GLcontext *ctx, radeonAllocDmaRegion(&context->radeon, &attr->bo, &attr->bo_offset, size, 32); + radeon_bo_map(attr->bo, 1); + if (!input->BufferObj->Pointer) { ctx->Driver.MapBuffer(ctx, GL_ARRAY_BUFFER, GL_READ_ONLY_ARB, input->BufferObj); @@ -596,6 +603,7 @@ static void r700AlignDataToDword(GLcontext *ctx, } } + radeon_bo_unmap(attr->bo); if (mapped_named_bo) { ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER, input->BufferObj); @@ -664,14 +672,18 @@ static void r700SetupStreams(GLcontext *ctx, const struct gl_client_array *input radeonAllocDmaRegion(&context->radeon, &context->stream_desc[index].bo, &context->stream_desc[index].bo_offset, size, 32); + + radeon_bo_map(context->stream_desc[index].bo, 1); assert(context->stream_desc[index].bo->ptr != NULL); + + dst = (uint32_t *)ADD_POINTERS(context->stream_desc[index].bo->ptr, context->stream_desc[index].bo_offset); switch (context->stream_desc[index].dwords) { case 1: - radeonEmitVec4(dst, input[i]->Ptr, input[i]->StrideB, local_count); + radeonEmitVec4(dst, input[i]->Ptr, input[i]->StrideB, local_count); break; case 2: radeonEmitVec8(dst, input[i]->Ptr, input[i]->StrideB, local_count); @@ -686,6 +698,7 @@ static void r700SetupStreams(GLcontext *ctx, const struct gl_client_array *input assert(0); break; } + radeon_bo_unmap(context->stream_desc[index].bo); } } @@ -757,6 +770,7 @@ static void r700FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer radeonAllocDmaRegion(&context->radeon, &context->ind_buf.bo, &context->ind_buf.bo_offset, size, 4); + radeon_bo_map(context->ind_buf.bo, 1); assert(context->ind_buf.bo->ptr != NULL); out = (GLuint *)ADD_POINTERS(context->ind_buf.bo->ptr, context->ind_buf.bo_offset); @@ -770,6 +784,7 @@ static void r700FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer *out++ = in[i]; } + radeon_bo_unmap(context->ind_buf.bo); #if MESA_BIG_ENDIAN } else @@ -780,6 +795,7 @@ static void r700FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer radeonAllocDmaRegion(&context->radeon, &context->ind_buf.bo, &context->ind_buf.bo_offset, size, 4); + radeon_bo_map(context->ind_buf.bo, 1); assert(context->ind_buf.bo->ptr != NULL); out = (GLuint *)ADD_POINTERS(context->ind_buf.bo->ptr, context->ind_buf.bo_offset); @@ -792,6 +808,7 @@ static void r700FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer { *out++ = in[i]; } + radeon_bo_unmap(context->ind_buf.bo); #endif } @@ -837,11 +854,13 @@ static void r700SetupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer radeonAllocDmaRegion(&context->radeon, &context->ind_buf.bo, &context->ind_buf.bo_offset, size, 4); + radeon_bo_map(context->ind_buf.bo, 1); assert(context->ind_buf.bo->ptr != NULL); dst_ptr = ADD_POINTERS(context->ind_buf.bo->ptr, context->ind_buf.bo_offset); _mesa_memcpy(dst_ptr, src_ptr, size); + radeon_bo_unmap(context->ind_buf.bo); context->ind_buf.is_32bit = (mesa_ind_buf->type == GL_UNSIGNED_INT); context->ind_buf.count = mesa_ind_buf->count; diff --git a/src/mesa/drivers/dri/r600/r700_shader.c b/src/mesa/drivers/dri/r600/r700_shader.c index 955ea4e4e1..2eed1acc2f 100644 --- a/src/mesa/drivers/dri/r600/r700_shader.c +++ b/src/mesa/drivers/dri/r600/r700_shader.c @@ -159,13 +159,18 @@ void Init_R700_Shader(R700_Shader * pShader) pShader->lstVTXInstructions.uNumOfNode=0; } +void SetActiveCFlist(R700_Shader *pShader, TypedShaderList * plstCF) +{ + pShader->plstCFInstructions_active = plstCF; +} + void AddCFInstruction(R700_Shader *pShader, R700ControlFlowInstruction *pCFInst) { R700ControlFlowSXClause* pSXClause; R700ControlFlowSMXClause* pSMXClause; - pCFInst->m_uIndex = pShader->lstCFInstructions.uNumOfNode; - AddInstToList(&(pShader->lstCFInstructions), + pCFInst->m_uIndex = pShader->plstCFInstructions_active->uNumOfNode; + AddInstToList(pShader->plstCFInstructions_active, (R700ShaderInstruction*)pCFInst); pShader->uShaderBinaryDWORDSize += GetInstructionSize(pCFInst->m_ShaderInstType); diff --git a/src/mesa/drivers/dri/r600/r700_shader.h b/src/mesa/drivers/dri/r600/r700_shader.h index c6a058617e..0599ffd901 100644 --- a/src/mesa/drivers/dri/r600/r700_shader.h +++ b/src/mesa/drivers/dri/r600/r700_shader.h @@ -109,6 +109,7 @@ typedef struct R700_Shader GLuint uStackSize; GLuint uMaxCallDepth; + TypedShaderList * plstCFInstructions_active; TypedShaderList lstCFInstructions; TypedShaderList lstALUInstructions; TypedShaderList lstTEXInstructions; @@ -132,13 +133,13 @@ void TakeInstOutFromList(TypedShaderList * plstCFInstructions, R700ShaderInstruc void ResolveLinks(R700_Shader *pShader); void Assemble(R700_Shader *pShader); - //Interface void Init_R700_Shader(R700_Shader * pShader); void AddCFInstruction(R700_Shader *pShader, R700ControlFlowInstruction *pCFInst); void AddVTXInstruction(R700_Shader *pShader, R700VertexInstruction *pVTXInst); void AddTEXInstruction(R700_Shader *pShader, R700TextureInstruction *pTEXInst); void AddALUInstruction(R700_Shader *pShader, R700ALUInstruction *pALUInst); +void SetActiveCFlist(R700_Shader *pShader, TypedShaderList * plstCF); void LoadProgram(R700_Shader *pShader); void UpdateShaderRegisters(R700_Shader *pShader); diff --git a/src/mesa/drivers/dri/r600/r700_state.c b/src/mesa/drivers/dri/r600/r700_state.c index 41000dc8ce..d7420678ff 100644 --- a/src/mesa/drivers/dri/r600/r700_state.c +++ b/src/mesa/drivers/dri/r600/r700_state.c @@ -202,9 +202,6 @@ static void r700SetDBRenderState(GLcontext * ctx) SETbit(r700->DB_SHADER_CONTROL.u32All, DUAL_EXPORT_ENABLE_bit); SETfield(r700->DB_SHADER_CONTROL.u32All, EARLY_Z_THEN_LATE_Z, Z_ORDER_shift, Z_ORDER_mask); - /* XXX not sure if this is required */ - if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770) - SETbit(r700->DB_RENDER_OVERRIDE.u32All, FORCE_SHADER_Z_ORDER_bit); /* XXX need to enable htile for hiz/s */ SETfield(r700->DB_RENDER_OVERRIDE.u32All, FORCE_DISABLE, FORCE_HIZ_ENABLE_shift, FORCE_HIZ_ENABLE_mask); SETfield(r700->DB_RENDER_OVERRIDE.u32All, FORCE_DISABLE, FORCE_HIS_ENABLE0_shift, FORCE_HIS_ENABLE0_mask); diff --git a/src/mesa/drivers/dri/r600/r700_vertprog.c b/src/mesa/drivers/dri/r600/r700_vertprog.c index ffc6068bd8..c8f72d588b 100644 --- a/src/mesa/drivers/dri/r600/r700_vertprog.c +++ b/src/mesa/drivers/dri/r600/r700_vertprog.c @@ -111,6 +111,15 @@ unsigned int Map_Vertex_Output(r700_AssemblerBase *pAsm, } } + for(i=VERT_RESULT_VAR0; i<VERT_RESULT_MAX; i++) + { + unBit = 1 << i; + if(mesa_vp->Base.OutputsWritten & unBit) + { + pAsm->ucVP_OutputMap[i] = unTotal++; + } + } + return (unTotal - unStart); } @@ -235,6 +244,8 @@ void Map_Vertex_Program(GLcontext *ctx, pAsm->number_used_registers += mesa_vp->Base.NumTemporaries; } + pAsm->flag_reg_index = pAsm->number_used_registers++; + pAsm->uFirstHelpReg = pAsm->number_used_registers; } @@ -324,7 +335,10 @@ struct r700_vertex_program* r700TranslateVertexShader(GLcontext *ctx, return NULL; } - if(GL_FALSE == AssembleInstr(vp->mesa_program->Base.NumInstructions, + InitShaderProgram(&(vp->r700AsmCode)); + + if(GL_FALSE == AssembleInstr(0, + vp->mesa_program->Base.NumInstructions, &(vp->mesa_program->Base.Instructions[0]), &(vp->r700AsmCode)) ) { @@ -336,6 +350,11 @@ struct r700_vertex_program* r700TranslateVertexShader(GLcontext *ctx, return NULL; } + if( GL_FALSE == RelocProgram(&(vp->r700AsmCode)) ) + { + return GL_FALSE; + } + vp->r700Shader.nRegs = (vp->r700AsmCode.number_used_registers == 0) ? 0 : (vp->r700AsmCode.number_used_registers - 1); diff --git a/src/mesa/drivers/dri/radeon/radeon_bo_drm.h b/src/mesa/drivers/dri/radeon/radeon_bo_drm.h index 7141371633..46e30b905a 100644 --- a/src/mesa/drivers/dri/radeon/radeon_bo_drm.h +++ b/src/mesa/drivers/dri/radeon/radeon_bo_drm.h @@ -39,6 +39,7 @@ #define RADEON_BO_FLAGS_MICRO_TILE 2 struct radeon_bo_manager; +struct radeon_cs; struct radeon_bo { uint32_t alignment; @@ -74,6 +75,7 @@ struct radeon_bo_funcs { int (*bo_get_tiling)(struct radeon_bo *bo, uint32_t *tiling_flags, uint32_t *pitch); int (*bo_is_busy)(struct radeon_bo *bo, uint32_t *domain); + int (*bo_is_referenced_by_cs)(struct radeon_bo *bo, struct radeon_cs *cs); }; struct radeon_bo_manager { @@ -199,6 +201,15 @@ static inline int radeon_bo_is_static(struct radeon_bo *bo) return 0; } +static inline int _radeon_bo_is_referenced_by_cs(struct radeon_bo *bo, + struct radeon_cs *cs, + const char *file, + const char *func, + unsigned line) +{ + return bo->cref > 1; +} + #define radeon_bo_open(bom, h, s, a, d, f)\ _radeon_bo_open(bom, h, s, a, d, f, __FILE__, __FUNCTION__, __LINE__) #define radeon_bo_ref(bo)\ @@ -215,5 +226,7 @@ static inline int radeon_bo_is_static(struct radeon_bo *bo) _radeon_bo_wait(bo, __FILE__, __func__, __LINE__) #define radeon_bo_is_busy(bo, domain) \ _radeon_bo_is_busy(bo, domain, __FILE__, __func__, __LINE__) +#define radeon_bo_is_referenced_by_cs(bo, cs) \ + _radeon_bo_is_referenced_by_cs(bo, cs, __FILE__, __FUNCTION__, __LINE__) #endif diff --git a/src/mesa/drivers/dri/radeon/radeon_buffer_objects.c b/src/mesa/drivers/dri/radeon/radeon_buffer_objects.c index 8fac5c6c51..99d3ec7005 100644 --- a/src/mesa/drivers/dri/radeon/radeon_buffer_objects.c +++ b/src/mesa/drivers/dri/radeon/radeon_buffer_objects.c @@ -136,8 +136,13 @@ radeonBufferSubData(GLcontext * ctx, const GLvoid * data, struct gl_buffer_object *obj) { + radeonContextPtr radeon = RADEON_CONTEXT(ctx); struct radeon_buffer_object *radeon_obj = get_radeon_buffer_object(obj); + if (radeon_bo_is_referenced_by_cs(radeon_obj->bo, radeon->cmdbuf.cs)) { + radeon_firevertices(radeon); + } + radeon_bo_map(radeon_obj->bo, GL_TRUE); _mesa_memcpy(radeon_obj->bo->ptr + offset, data, size); diff --git a/src/mesa/drivers/dri/radeon/radeon_common.c b/src/mesa/drivers/dri/radeon/radeon_common.c index 097ab7cf61..3b4366aa61 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common.c +++ b/src/mesa/drivers/dri/radeon/radeon_common.c @@ -257,7 +257,9 @@ void radeonScissor(GLcontext* ctx, GLint x, GLint y, GLsizei w, GLsizei h) radeonContextPtr radeon = RADEON_CONTEXT(ctx); if (ctx->Scissor.Enabled) { /* We don't pipeline cliprect changes */ - radeon_firevertices(radeon); + if (!radeon->radeonScreen->kernel_mm) { + radeon_firevertices(radeon); + } radeonUpdateScissor(ctx); } } @@ -1123,8 +1125,6 @@ void radeonFlush(GLcontext *ctx) if (radeon->dma.flush) radeon->dma.flush( ctx ); - radeonEmitState(radeon); - if (radeon->cmdbuf.cs->cdw) rcommonFlushCmdBuf(radeon, __FUNCTION__); @@ -1147,9 +1147,6 @@ void radeonFlush(GLcontext *ctx) } } } - - make_empty_list(&radeon->query.not_flushed_head); - } /* Make sure all commands have been sent to the hardware and have diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.c b/src/mesa/drivers/dri/radeon/radeon_common_context.c index fe99644907..71f70d724b 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common_context.c +++ b/src/mesa/drivers/dri/radeon/radeon_common_context.c @@ -262,10 +262,9 @@ GLboolean radeonInitContext(radeonContextPtr radeon, else radeon->texture_row_align = 32; radeon->texture_rect_row_align = 64; - radeon->texture_compressed_row_align = 64; + radeon->texture_compressed_row_align = 32; } - make_empty_list(&radeon->query.not_flushed_head); radeon_init_dma(radeon); return GL_TRUE; diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.h b/src/mesa/drivers/dri/radeon/radeon_common_context.h index 0309345393..ad953ddbb5 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common_context.h +++ b/src/mesa/drivers/dri/radeon/radeon_common_context.h @@ -208,6 +208,10 @@ struct radeon_tex_obj { * and so on. */ GLboolean validated; + /* Minimum LOD to be used during rendering */ + unsigned minLod; + /* Miximum LOD to be used during rendering */ + unsigned maxLod; GLuint override_offset; GLboolean image_override; /* Image overridden by GLX_EXT_tfp */ @@ -324,6 +328,7 @@ struct radeon_swtcl_info { GLuint vertex_attr_count; GLuint emit_prediction; + struct radeon_bo *bo; }; #define RADEON_MAX_AOS_ARRAYS 16 @@ -502,7 +507,6 @@ struct radeon_context { struct { struct radeon_query_object *current; - struct radeon_query_object not_flushed_head; struct radeon_state_atom queryobj; } query; diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.c b/src/mesa/drivers/dri/radeon/radeon_dma.c index c6edbae9a1..89a99974e2 100644 --- a/src/mesa/drivers/dri/radeon/radeon_dma.c +++ b/src/mesa/drivers/dri/radeon/radeon_dma.c @@ -151,6 +151,7 @@ void rcommon_emit_vector(GLcontext * ctx, struct radeon_aos *aos, aos->components = size; aos->count = count; + radeon_bo_map(aos->bo, 1); out = (uint32_t*)((char*)aos->bo->ptr + aos->offset); switch (size) { case 1: radeonEmitVec4(out, data, stride, count); break; @@ -161,6 +162,7 @@ void rcommon_emit_vector(GLcontext * ctx, struct radeon_aos *aos, assert(0); break; } + radeon_bo_unmap(aos->bo); } void radeon_init_dma(radeonContextPtr rmesa) @@ -183,10 +185,6 @@ void radeonRefillCurrentDmaRegion(radeonContextPtr rmesa, int size) __FUNCTION__, size, rmesa->dma.minimum_size); - /* unmap old reserved bo */ - if (!is_empty_list(&rmesa->dma.reserved)) - radeon_bo_unmap(first_elem(&rmesa->dma.reserved)->bo); - if (is_empty_list(&rmesa->dma.free) || last_elem(&rmesa->dma.free)->bo->size < size) { dma_bo = CALLOC_STRUCT(radeon_dma_bo); @@ -223,8 +221,6 @@ again_alloc: /* Cmd buff have been flushed in radeon_revalidate_bos */ goto again_alloc; } - - radeon_bo_map(first_elem(&rmesa->dma.reserved)->bo, 1); } /* Allocates a region from rmesa->dma.current. If there isn't enough @@ -281,7 +277,6 @@ void radeonFreeDmaRegions(radeonContextPtr rmesa) foreach_s(dma_bo, temp, &rmesa->dma.reserved) { remove_from_list(dma_bo); - radeon_bo_unmap(dma_bo->bo); radeon_bo_unref(dma_bo->bo); FREE(dma_bo); } @@ -367,9 +362,6 @@ void radeonReleaseDmaRegions(radeonContextPtr rmesa) insert_at_tail(&rmesa->dma.free, dma_bo); } - /* unmap the last dma region */ - if (!is_empty_list(&rmesa->dma.reserved)) - radeon_bo_unmap(first_elem(&rmesa->dma.reserved)->bo); /* move reserved to wait list */ foreach_s(dma_bo, temp, &rmesa->dma.reserved) { /* free objects that are too small to be used because of large request */ @@ -403,11 +395,12 @@ void rcommon_flush_last_swtcl_prim( GLcontext *ctx ) radeonContextPtr rmesa = RADEON_CONTEXT(ctx); struct radeon_dma *dma = &rmesa->dma; - if (RADEON_DEBUG & RADEON_IOCTL) fprintf(stderr, "%s\n", __FUNCTION__); dma->flush = NULL; + radeon_bo_unmap(rmesa->swtcl.bo); + if (!is_empty_list(&dma->reserved)) { GLuint current_offset = dma->current_used; @@ -422,6 +415,8 @@ void rcommon_flush_last_swtcl_prim( GLcontext *ctx ) } rmesa->swtcl.numverts = 0; } + radeon_bo_unref(rmesa->swtcl.bo); + rmesa->swtcl.bo = NULL; } /* Alloc space in the current dma region. */ @@ -432,6 +427,7 @@ rcommonAllocDmaLowVerts( radeonContextPtr rmesa, int nverts, int vsize ) void *head; if (RADEON_DEBUG & RADEON_IOCTL) fprintf(stderr, "%s\n", __FUNCTION__); + if(is_empty_list(&rmesa->dma.reserved) ||rmesa->dma.current_vertexptr + bytes > first_elem(&rmesa->dma.reserved)->bo->size) { if (rmesa->dma.flush) { @@ -455,7 +451,13 @@ rcommonAllocDmaLowVerts( radeonContextPtr rmesa, int nverts, int vsize ) rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 == rmesa->dma.current_vertexptr ); - head = (first_elem(&rmesa->dma.reserved)->bo->ptr + rmesa->dma.current_vertexptr); + if (!rmesa->swtcl.bo) { + rmesa->swtcl.bo = first_elem(&rmesa->dma.reserved)->bo; + radeon_bo_ref(rmesa->swtcl.bo); + radeon_bo_map(rmesa->swtcl.bo, 1); + } + + head = (rmesa->swtcl.bo->ptr + rmesa->dma.current_vertexptr); rmesa->dma.current_vertexptr += bytes; rmesa->swtcl.numverts += nverts; return head; diff --git a/src/mesa/drivers/dri/radeon/radeon_fbo.c b/src/mesa/drivers/dri/radeon/radeon_fbo.c index bf69cd9337..7ec641ff18 100644 --- a/src/mesa/drivers/dri/radeon/radeon_fbo.c +++ b/src/mesa/drivers/dri/radeon/radeon_fbo.c @@ -90,7 +90,7 @@ radeon_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, case GL_R3_G3_B2: case GL_RGB4: case GL_RGB5: - rb->Format = MESA_FORMAT_RGB565; + rb->Format = _dri_texformat_rgb565; rb->DataType = GL_UNSIGNED_BYTE; cpp = 2; break; @@ -99,7 +99,7 @@ radeon_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, case GL_RGB10: case GL_RGB12: case GL_RGB16: - rb->Format = MESA_FORMAT_ARGB8888; + rb->Format = _dri_texformat_argb8888; rb->DataType = GL_UNSIGNED_BYTE; cpp = 4; break; @@ -111,7 +111,7 @@ radeon_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, case GL_RGB10_A2: case GL_RGBA12: case GL_RGBA16: - rb->Format = MESA_FORMAT_ARGB8888; + rb->Format = _dri_texformat_argb8888; rb->DataType = GL_UNSIGNED_BYTE; cpp = 4; break; @@ -261,14 +261,32 @@ radeon_create_renderbuffer(gl_format format, __DRIdrawablePrivate *driDrawPriv) switch (format) { case MESA_FORMAT_RGB565: + assert(_mesa_little_endian()); + rrb->base.DataType = GL_UNSIGNED_BYTE; + rrb->base._BaseFormat = GL_RGB; + break; + case MESA_FORMAT_RGB565_REV: + assert(!_mesa_little_endian()); rrb->base.DataType = GL_UNSIGNED_BYTE; rrb->base._BaseFormat = GL_RGB; break; case MESA_FORMAT_XRGB8888: + assert(_mesa_little_endian()); + rrb->base.DataType = GL_UNSIGNED_BYTE; + rrb->base._BaseFormat = GL_RGB; + break; + case MESA_FORMAT_XRGB8888_REV: + assert(!_mesa_little_endian()); rrb->base.DataType = GL_UNSIGNED_BYTE; rrb->base._BaseFormat = GL_RGB; break; case MESA_FORMAT_ARGB8888: + assert(_mesa_little_endian()); + rrb->base.DataType = GL_UNSIGNED_BYTE; + rrb->base._BaseFormat = GL_RGBA; + break; + case MESA_FORMAT_ARGB8888_REV: + assert(!_mesa_little_endian()); rrb->base.DataType = GL_UNSIGNED_BYTE; rrb->base._BaseFormat = GL_RGBA; break; @@ -359,21 +377,21 @@ radeon_update_wrapper(GLcontext *ctx, struct radeon_renderbuffer *rrb, gl_format texFormat; restart: - if (texImage->TexFormat == MESA_FORMAT_ARGB8888) { + if (texImage->TexFormat == _dri_texformat_argb8888) { rrb->base.DataType = GL_UNSIGNED_BYTE; DBG("Render to RGBA8 texture OK\n"); } - else if (texImage->TexFormat == MESA_FORMAT_RGB565) { + else if (texImage->TexFormat == _dri_texformat_rgb565) { rrb->base.DataType = GL_UNSIGNED_BYTE; DBG("Render to RGB5 texture OK\n"); } - else if (texImage->TexFormat == MESA_FORMAT_ARGB1555) { + else if (texImage->TexFormat == _dri_texformat_argb1555) { rrb->base.DataType = GL_UNSIGNED_BYTE; DBG("Render to ARGB1555 texture OK\n"); } - else if (texImage->TexFormat == MESA_FORMAT_ARGB4444) { + else if (texImage->TexFormat == _dri_texformat_argb4444) { rrb->base.DataType = GL_UNSIGNED_BYTE; - DBG("Render to ARGB1555 texture OK\n"); + DBG("Render to ARGB4444 texture OK\n"); } else if (texImage->TexFormat == MESA_FORMAT_Z16) { rrb->base.DataType = GL_UNSIGNED_SHORT; diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c b/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c index 08e1c5d00d..d810e6080e 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c +++ b/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c @@ -76,12 +76,14 @@ static void emit_vecfog(GLcontext *ctx, struct radeon_aos *aos, /* Emit the data */ + radeon_bo_map(aos->bo, 1); out = (uint32_t*)((char*)aos->bo->ptr + aos->offset); for (i = 0; i < count; i++) { out[0] = radeonComputeFogBlendFactor( ctx, *(GLfloat *)data ); out++; data += stride; } + radeon_bo_unmap(aos->bo); } static void emit_s0_vec(uint32_t *out, GLvoid *data, int stride, int count) @@ -151,6 +153,7 @@ static void emit_tex_vector(GLcontext *ctx, struct radeon_aos *aos, /* Emit the data */ + radeon_bo_map(aos->bo, 1); out = (uint32_t*)((char*)aos->bo->ptr + aos->offset); switch (size) { case 1: @@ -170,6 +173,7 @@ static void emit_tex_vector(GLcontext *ctx, struct radeon_aos *aos, exit(1); break; } + radeon_bo_unmap(aos->bo); } @@ -196,12 +200,12 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (!rmesa->tcl.obj.buf) rcommon_emit_vector( ctx, &(rmesa->tcl.aos[nr]), - (char *)VB->ObjPtr->data, - VB->ObjPtr->size, - VB->ObjPtr->stride, + (char *)VB->AttribPtr[_TNL_ATTRIB_POS]->data, + VB->AttribPtr[_TNL_ATTRIB_POS]->size, + VB->AttribPtr[_TNL_ATTRIB_POS]->stride, count); - switch( VB->ObjPtr->size ) { + switch( VB->AttribPtr[_TNL_ATTRIB_POS]->size ) { case 4: vfmt |= RADEON_CP_VC_FRMT_W0; case 3: vfmt |= RADEON_CP_VC_FRMT_Z; case 2: vfmt |= RADEON_CP_VC_FRMT_XY; @@ -216,9 +220,9 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (!rmesa->tcl.norm.buf) rcommon_emit_vector( ctx, &(rmesa->tcl.aos[nr]), - (char *)VB->NormalPtr->data, + (char *)VB->AttribPtr[_TNL_ATTRIB_NORMAL]->data, 3, - VB->NormalPtr->stride, + VB->AttribPtr[_TNL_ATTRIB_NORMAL]->stride, count); vfmt |= RADEON_CP_VC_FRMT_N0; @@ -227,9 +231,9 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (inputs & VERT_BIT_COLOR0) { int emitsize; - if (VB->ColorPtr[0]->size == 4 && - (VB->ColorPtr[0]->stride != 0 || - VB->ColorPtr[0]->data[0][3] != 1.0)) { + if (VB->AttribPtr[_TNL_ATTRIB_COLOR0]->size == 4 && + (VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride != 0 || + VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data[0][3] != 1.0)) { vfmt |= RADEON_CP_VC_FRMT_FPCOLOR | RADEON_CP_VC_FRMT_FPALPHA; emitsize = 4; } @@ -242,9 +246,9 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (!rmesa->tcl.rgba.buf) rcommon_emit_vector( ctx, &(rmesa->tcl.aos[nr]), - (char *)VB->ColorPtr[0]->data, + (char *)VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data, emitsize, - VB->ColorPtr[0]->stride, + VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride, count); nr++; @@ -256,9 +260,9 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) rcommon_emit_vector( ctx, &(rmesa->tcl.aos[nr]), - (char *)VB->SecondaryColorPtr[0]->data, + (char *)VB->AttribPtr[_TNL_ATTRIB_COLOR1]->data, 3, - VB->SecondaryColorPtr[0]->stride, + VB->AttribPtr[_TNL_ATTRIB_COLOR1]->stride, count); } @@ -273,8 +277,8 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (!rmesa->tcl.fog.buf) emit_vecfog( ctx, &(rmesa->tcl.aos[nr]), - (char *)VB->FogCoordPtr->data, - VB->FogCoordPtr->stride, + (char *)VB->AttribPtr[_TNL_ATTRIB_FOG]->data, + VB->AttribPtr[_TNL_ATTRIB_FOG]->stride, count); vfmt |= RADEON_CP_VC_FRMT_FPFOG; @@ -290,24 +294,24 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (!rmesa->tcl.tex[unit].buf) emit_tex_vector( ctx, &(rmesa->tcl.aos[nr]), - (char *)VB->TexCoordPtr[unit]->data, - VB->TexCoordPtr[unit]->size, - VB->TexCoordPtr[unit]->stride, + (char *)VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->data, + VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size, + VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->stride, count ); nr++; vfmt |= RADEON_ST_BIT(unit); /* assume we need the 3rd coord if texgen is active for r/q OR at least 3 coords are submitted. This may not be 100% correct */ - if (VB->TexCoordPtr[unit]->size >= 3) { + if (VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size >= 3) { vtx |= RADEON_Q_BIT(unit); vfmt |= RADEON_Q_BIT(unit); } if ( (ctx->Texture.Unit[unit].TexGenEnabled & (R_BIT | Q_BIT)) ) vtx |= RADEON_Q_BIT(unit); - else if ((VB->TexCoordPtr[unit]->size >= 3) && + else if ((VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size >= 3) && ((ctx->Texture.Unit[unit]._ReallyEnabled & (TEXTURE_CUBE_BIT)) == 0)) { - GLuint swaptexmatcol = (VB->TexCoordPtr[unit]->size - 3); + GLuint swaptexmatcol = (VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size - 3); if (((rmesa->NeedTexMatrix >> unit) & 1) && (swaptexmatcol != ((rmesa->TexMatColSwap >> unit) & 1))) radeonUploadTexMatrix( rmesa, unit, swaptexmatcol ) ; diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h b/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h index 515783135d..d764ccb982 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h +++ b/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h @@ -56,18 +56,18 @@ static void TAG(emit)( GLcontext *ctx, radeon_print(RADEON_SWRENDER, RADEON_VERBOSE, "%s\n", __FUNCTION__); - coord = (GLuint (*)[4])VB->ObjPtr->data; - coord_stride = VB->ObjPtr->stride; + coord = (GLuint (*)[4])VB->AttribPtr[_TNL_ATTRIB_POS]->data; + coord_stride = VB->AttribPtr[_TNL_ATTRIB_POS]->stride; if (DO_TEX2) { - if (VB->TexCoordPtr[2]) { + if (VB->AttribPtr[_TNL_ATTRIB_TEX2]) { const GLuint t2 = GET_TEXSOURCE(2); - tc2 = (GLuint (*)[4])VB->TexCoordPtr[t2]->data; - tc2_stride = VB->TexCoordPtr[t2]->stride; - if (DO_PTEX && VB->TexCoordPtr[t2]->size < 3) { + tc2 = (GLuint (*)[4])VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->data; + tc2_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->stride; + if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->size < 3) { fill_tex |= (1<<2); } - else if (DO_PTEX && VB->TexCoordPtr[t2]->size < 4) { + else if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->size < 4) { rqcoordsnoswap |= (1<<2); } } else { @@ -77,14 +77,14 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_TEX1) { - if (VB->TexCoordPtr[1]) { + if (VB->AttribPtr[_TNL_ATTRIB_TEX1]) { const GLuint t1 = GET_TEXSOURCE(1); - tc1 = (GLuint (*)[4])VB->TexCoordPtr[t1]->data; - tc1_stride = VB->TexCoordPtr[t1]->stride; - if (DO_PTEX && VB->TexCoordPtr[t1]->size < 3) { + tc1 = (GLuint (*)[4])VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->data; + tc1_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->stride; + if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->size < 3) { fill_tex |= (1<<1); } - else if (DO_PTEX && VB->TexCoordPtr[t1]->size < 4) { + else if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->size < 4) { rqcoordsnoswap |= (1<<1); } } else { @@ -94,14 +94,14 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_TEX0) { - if (VB->TexCoordPtr[0]) { + if (VB->AttribPtr[_TNL_ATTRIB_TEX0]) { const GLuint t0 = GET_TEXSOURCE(0); - tc0_stride = VB->TexCoordPtr[t0]->stride; - tc0 = (GLuint (*)[4])VB->TexCoordPtr[t0]->data; - if (DO_PTEX && VB->TexCoordPtr[t0]->size < 3) { + tc0_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->stride; + tc0 = (GLuint (*)[4])VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->data; + if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->size < 3) { fill_tex |= (1<<0); } - else if (DO_PTEX && VB->TexCoordPtr[t0]->size < 4) { + else if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->size < 4) { rqcoordsnoswap |= (1<<0); } } else { @@ -112,9 +112,9 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_NORM) { - if (VB->NormalPtr) { - norm_stride = VB->NormalPtr->stride; - norm = (GLuint (*)[4])VB->NormalPtr->data; + if (VB->AttribPtr[_TNL_ATTRIB_NORMAL]) { + norm_stride = VB->AttribPtr[_TNL_ATTRIB_NORMAL]->stride; + norm = (GLuint (*)[4])VB->AttribPtr[_TNL_ATTRIB_NORMAL]->data; } else { norm_stride = 0; norm = (GLuint (*)[4])&ctx->Current.Attrib[VERT_ATTRIB_NORMAL]; @@ -122,9 +122,9 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_RGBA) { - if (VB->ColorPtr[0]) { - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + if (VB->AttribPtr[_TNL_ATTRIB_COLOR0]) { + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; } else { col = (GLfloat (*)[4])ctx->Current.Attrib[VERT_ATTRIB_COLOR0]; col_stride = 0; @@ -132,9 +132,9 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_SPEC_OR_FOG) { - if (VB->SecondaryColorPtr[0]) { - spec = VB->SecondaryColorPtr[0]->data; - spec_stride = VB->SecondaryColorPtr[0]->stride; + if (VB->AttribPtr[_TNL_ATTRIB_COLOR1]) { + spec = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->data; + spec_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->stride; } else { spec = (GLfloat (*)[4])ctx->Current.Attrib[VERT_ATTRIB_COLOR1]; spec_stride = 0; @@ -142,9 +142,9 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_SPEC_OR_FOG) { - if (VB->FogCoordPtr) { - fog = VB->FogCoordPtr->data; - fog_stride = VB->FogCoordPtr->stride; + if (VB->AttribPtr[_TNL_ATTRIB_FOG]) { + fog = VB->AttribPtr[_TNL_ATTRIB_FOG]->data; + fog_stride = VB->AttribPtr[_TNL_ATTRIB_FOG]->stride; } else { fog = (GLfloat (*)[4])ctx->Current.Attrib[VERT_ATTRIB_FOG]; fog_stride = 0; diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c index 78ec119302..98f96ff2a7 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c +++ b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c @@ -326,7 +326,7 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (1) { req |= RADEON_CP_VC_FRMT_Z; - if (VB->ObjPtr->size == 4) { + if (VB->AttribPtr[_TNL_ATTRIB_POS]->size == 4) { req |= RADEON_CP_VC_FRMT_W0; } } @@ -348,15 +348,15 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) req |= RADEON_ST_BIT(unit); /* assume we need the 3rd coord if texgen is active for r/q OR at least 3 coords are submitted. This may not be 100% correct */ - if (VB->TexCoordPtr[unit]->size >= 3) { + if (VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size >= 3) { req |= RADEON_Q_BIT(unit); vtx |= RADEON_Q_BIT(unit); } if ( (ctx->Texture.Unit[unit].TexGenEnabled & (R_BIT | Q_BIT)) ) vtx |= RADEON_Q_BIT(unit); - else if ((VB->TexCoordPtr[unit]->size >= 3) && + else if ((VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size >= 3) && ((ctx->Texture.Unit[unit]._ReallyEnabled & (TEXTURE_CUBE_BIT)) == 0)) { - GLuint swaptexmatcol = (VB->TexCoordPtr[unit]->size - 3); + GLuint swaptexmatcol = (VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size - 3); if (((rmesa->NeedTexMatrix >> unit) & 1) && (swaptexmatcol != ((rmesa->TexMatColSwap >> unit) & 1))) radeonUploadTexMatrix( rmesa, unit, swaptexmatcol ) ; @@ -390,19 +390,19 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) * this, add more vertex code (for obj-2, obj-3) or preferably move * to maos. */ - if (VB->ObjPtr->size < 3 || - (VB->ObjPtr->size == 3 && + if (VB->AttribPtr[_TNL_ATTRIB_POS]->size < 3 || + (VB->AttribPtr[_TNL_ATTRIB_POS]->size == 3 && (setup_tab[i].vertex_format & RADEON_CP_VC_FRMT_W0))) { _math_trans_4f( rmesa->tcl.ObjClean.data, - VB->ObjPtr->data, - VB->ObjPtr->stride, + VB->AttribPtr[_TNL_ATTRIB_POS]->data, + VB->AttribPtr[_TNL_ATTRIB_POS]->stride, GL_FLOAT, - VB->ObjPtr->size, + VB->AttribPtr[_TNL_ATTRIB_POS]->size, 0, VB->Count ); - switch (VB->ObjPtr->size) { + switch (VB->AttribPtr[_TNL_ATTRIB_POS]->size) { case 1: _mesa_vector4f_clean_elem(&rmesa->tcl.ObjClean, VB->Count, 1); case 2: @@ -416,14 +416,14 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) break; } - VB->ObjPtr = &rmesa->tcl.ObjClean; + VB->AttribPtr[_TNL_ATTRIB_POS] = &rmesa->tcl.ObjClean; } - + radeon_bo_map(rmesa->radeon.tcl.aos[0].bo, 1); setup_tab[i].emit( ctx, 0, VB->Count, rmesa->radeon.tcl.aos[0].bo->ptr + rmesa->radeon.tcl.aos[0].offset); - + radeon_bo_unmap(rmesa->radeon.tcl.aos[0].bo); // rmesa->radeon.tcl.aos[0].size = setup_tab[i].vertex_size; rmesa->radeon.tcl.aos[0].stride = setup_tab[i].vertex_size; rmesa->tcl.vertex_format = setup_tab[i].vertex_format; diff --git a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c index dadc72f4c1..f2f7b2a9fd 100644 --- a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c +++ b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c @@ -1,4 +1,5 @@ /* + * Copyright (C) 2009 Maciej Cencora. * Copyright (C) 2008 Nicolai Haehnle. * * All Rights Reserved. @@ -32,50 +33,52 @@ #include "main/simple_list.h" #include "main/texcompress.h" - -static GLuint radeon_compressed_texture_size(GLcontext *ctx, - GLsizei width, GLsizei height, GLsizei depth, - GLuint mesaFormat) +#include "main/teximage.h" +#include "main/texobj.h" +#include "radeon_texture.h" + +static unsigned get_aligned_compressed_row_stride( + gl_format format, + unsigned width, + unsigned minStride) { - GLuint size = _mesa_format_image_size(mesaFormat, width, height, depth); - - if (mesaFormat == MESA_FORMAT_RGB_DXT1 || - mesaFormat == MESA_FORMAT_RGBA_DXT1) { - if (width + 3 < 8) /* width one block */ - size = size * 4; - else if (width + 3 < 16) - size = size * 2; - } else { - /* DXT3/5, 16 bytes per block */ - // WARN_ONCE("DXT 3/5 suffers from multitexturing problems!\n"); - if (width + 3 < 8) - size = size * 2; + const unsigned blockSize = _mesa_get_format_bytes(format); + unsigned blockWidth, blockHeight, numXBlocks; + + _mesa_get_format_block_size(format, &blockWidth, &blockHeight); + numXBlocks = (width + blockWidth - 1) / blockWidth; + + while (numXBlocks * blockSize < minStride) + { + ++numXBlocks; } - return size; + return numXBlocks * blockSize; } +static unsigned get_compressed_image_size( + gl_format format, + unsigned rowStride, + unsigned height) +{ + unsigned blockWidth, blockHeight; + + _mesa_get_format_block_size(format, &blockWidth, &blockHeight); -static int radeon_compressed_num_bytes(GLuint mesaFormat) + return rowStride * ((height + blockHeight - 1) / blockHeight); +} + +static int find_next_power_of_two(GLuint value) { - int bytes = 0; - switch(mesaFormat) { - - case MESA_FORMAT_RGB_FXT1: - case MESA_FORMAT_RGBA_FXT1: - case MESA_FORMAT_RGB_DXT1: - case MESA_FORMAT_RGBA_DXT1: - bytes = 2; - break; - - case MESA_FORMAT_RGBA_DXT3: - case MESA_FORMAT_RGBA_DXT5: - bytes = 4; - default: - break; - } - - return bytes; + int i, tmp; + + i = 0; + tmp = value - 1; + while (tmp) { + tmp >>= 1; + i++; + } + return (1 << i); } /** @@ -90,28 +93,28 @@ static void compute_tex_image_offset(radeonContextPtr rmesa, radeon_mipmap_tree { radeon_mipmap_level *lvl = &mt->levels[level]; uint32_t row_align; + GLuint height; + + height = find_next_power_of_two(lvl->height); /* Find image size in bytes */ - if (mt->compressed) { - /* TODO: Is this correct? Need test cases for compressed textures! */ - row_align = rmesa->texture_compressed_row_align - 1; - lvl->rowstride = (lvl->width * mt->bpp + row_align) & ~row_align; - lvl->size = radeon_compressed_texture_size(mt->radeon->glCtx, - lvl->width, lvl->height, lvl->depth, mt->compressed); + if (_mesa_is_format_compressed(mt->mesaFormat)) { + lvl->rowstride = get_aligned_compressed_row_stride(mt->mesaFormat, lvl->width, rmesa->texture_compressed_row_align); + lvl->size = get_compressed_image_size(mt->mesaFormat, lvl->rowstride, height); } else if (mt->target == GL_TEXTURE_RECTANGLE_NV) { row_align = rmesa->texture_rect_row_align - 1; - lvl->rowstride = (lvl->width * mt->bpp + row_align) & ~row_align; - lvl->size = lvl->rowstride * lvl->height; + lvl->rowstride = (_mesa_format_row_stride(mt->mesaFormat, lvl->width) + row_align) & ~row_align; + lvl->size = lvl->rowstride * height; } else if (mt->tilebits & RADEON_TXO_MICRO_TILE) { /* tile pattern is 16 bytes x2. mipmaps stay 32 byte aligned, * though the actual offset may be different (if texture is less than * 32 bytes width) to the untiled case */ - lvl->rowstride = (lvl->width * mt->bpp * 2 + 31) & ~31; - lvl->size = lvl->rowstride * ((lvl->height + 1) / 2) * lvl->depth; + lvl->rowstride = (_mesa_format_row_stride(mt->mesaFormat, lvl->width) * 2 + 31) & ~31; + lvl->size = lvl->rowstride * ((height + 1) / 2) * lvl->depth; } else { row_align = rmesa->texture_row_align - 1; - lvl->rowstride = (lvl->width * mt->bpp + row_align) & ~row_align; - lvl->size = lvl->rowstride * lvl->height * lvl->depth; + lvl->rowstride = (_mesa_format_row_stride(mt->mesaFormat, lvl->width) + row_align) & ~row_align; + lvl->size = lvl->rowstride * height * lvl->depth; } assert(lvl->size > 0); @@ -123,7 +126,7 @@ static void compute_tex_image_offset(radeonContextPtr rmesa, radeon_mipmap_tree if (RADEON_DEBUG & RADEON_TEXTURE) fprintf(stderr, "level %d, face %d: rs:%d %dx%d at %d\n", - level, face, lvl->rowstride, lvl->width, lvl->height, lvl->faces[face].offset); + level, face, lvl->rowstride, lvl->width, height, lvl->faces[face].offset); } static GLuint minify(GLuint size, GLuint levels) @@ -138,17 +141,15 @@ static GLuint minify(GLuint size, GLuint levels) static void calculate_miptree_layout_r100(radeonContextPtr rmesa, radeon_mipmap_tree *mt) { GLuint curOffset; - GLuint numLevels; GLuint i; GLuint face; - numLevels = mt->lastLevel - mt->firstLevel + 1; - assert(numLevels <= rmesa->glCtx->Const.MaxTextureLevels); + assert(mt->numLevels <= rmesa->glCtx->Const.MaxTextureLevels); curOffset = 0; for(face = 0; face < mt->faces; face++) { - for(i = 0; i < numLevels; i++) { + for(i = 0; i < mt->numLevels; i++) { mt->levels[i].width = minify(mt->width0, i); mt->levels[i].height = minify(mt->height0, i); mt->levels[i].depth = minify(mt->depth0, i); @@ -163,14 +164,12 @@ static void calculate_miptree_layout_r100(radeonContextPtr rmesa, radeon_mipmap_ static void calculate_miptree_layout_r300(radeonContextPtr rmesa, radeon_mipmap_tree *mt) { GLuint curOffset; - GLuint numLevels; GLuint i; - numLevels = mt->lastLevel - mt->firstLevel + 1; - assert(numLevels <= rmesa->glCtx->Const.MaxTextureLevels); + assert(mt->numLevels <= rmesa->glCtx->Const.MaxTextureLevels); curOffset = 0; - for(i = 0; i < numLevels; i++) { + for(i = 0; i < mt->numLevels; i++) { GLuint face; mt->levels[i].width = minify(mt->width0, i); @@ -188,27 +187,22 @@ static void calculate_miptree_layout_r300(radeonContextPtr rmesa, radeon_mipmap_ /** * Create a new mipmap tree, calculate its layout and allocate memory. */ -radeon_mipmap_tree* radeon_miptree_create(radeonContextPtr rmesa, radeonTexObj *t, - GLenum target, GLenum internal_format, GLuint firstLevel, GLuint lastLevel, - GLuint width0, GLuint height0, GLuint depth0, - GLuint bpp, GLuint tilebits, GLuint compressed) +static radeon_mipmap_tree* radeon_miptree_create(radeonContextPtr rmesa, + GLenum target, gl_format mesaFormat, GLuint baseLevel, GLuint numLevels, + GLuint width0, GLuint height0, GLuint depth0, GLuint tilebits) { radeon_mipmap_tree *mt = CALLOC_STRUCT(_radeon_mipmap_tree); - mt->radeon = rmesa; - mt->internal_format = internal_format; + mt->mesaFormat = mesaFormat; mt->refcount = 1; - mt->t = t; mt->target = target; mt->faces = (target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; - mt->firstLevel = firstLevel; - mt->lastLevel = lastLevel; + mt->baseLevel = baseLevel; + mt->numLevels = numLevels; mt->width0 = width0; mt->height0 = height0; mt->depth0 = depth0; - mt->bpp = compressed ? radeon_compressed_num_bytes(compressed) : bpp; mt->tilebits = tilebits; - mt->compressed = compressed; if (rmesa->radeonScreen->chip_family >= CHIP_FAMILY_R300) calculate_miptree_layout_r300(rmesa, mt); @@ -223,53 +217,43 @@ radeon_mipmap_tree* radeon_miptree_create(radeonContextPtr rmesa, radeonTexObj * return mt; } -void radeon_miptree_reference(radeon_mipmap_tree *mt) +void radeon_miptree_reference(radeon_mipmap_tree *mt, radeon_mipmap_tree **ptr) { + assert(!*ptr); + mt->refcount++; assert(mt->refcount > 0); + + *ptr = mt; } -void radeon_miptree_unreference(radeon_mipmap_tree *mt) +void radeon_miptree_unreference(radeon_mipmap_tree **ptr) { + radeon_mipmap_tree *mt = *ptr; if (!mt) return; assert(mt->refcount > 0); + mt->refcount--; if (!mt->refcount) { radeon_bo_unref(mt->bo); free(mt); } -} + *ptr = 0; +} /** - * Calculate first and last mip levels for the given texture object, - * where the dimensions are taken from the given texture image at - * the given level. - * - * Note: level is the OpenGL level number, which is not necessarily the same - * as the first level that is actually present. - * - * The base level image of the given texture face must be non-null, - * or this will fail. + * Calculate min and max LOD for the given texture object. + * @param[in] tObj texture object whose LOD values to calculate + * @param[out] pminLod minimal LOD + * @param[out] pmaxLod maximal LOD */ -static void calculate_first_last_level(struct gl_texture_object *tObj, - GLuint *pfirstLevel, GLuint *plastLevel, - GLuint face, GLuint level) +static void calculate_min_max_lod(struct gl_texture_object *tObj, + unsigned *pminLod, unsigned *pmaxLod) { - const struct gl_texture_image * const baseImage = - tObj->Image[face][level]; - - assert(baseImage); - - /* These must be signed values. MinLod and MaxLod can be negative numbers, - * and having firstLevel and lastLevel as signed prevents the need for - * extra sign checks. - */ - int firstLevel; - int lastLevel; - + int minLod, maxLod; /* Yes, this looks overly complicated, but it's all needed. */ switch (tObj->Target) { @@ -280,55 +264,46 @@ static void calculate_first_last_level(struct gl_texture_object *tObj, if (tObj->MinFilter == GL_NEAREST || tObj->MinFilter == GL_LINEAR) { /* GL_NEAREST and GL_LINEAR only care about GL_TEXTURE_BASE_LEVEL. */ - firstLevel = lastLevel = tObj->BaseLevel; + minLod = maxLod = tObj->BaseLevel; } else { - firstLevel = tObj->BaseLevel + (GLint)(tObj->MinLod + 0.5); - firstLevel = MAX2(firstLevel, tObj->BaseLevel); - firstLevel = MIN2(firstLevel, level + baseImage->MaxLog2); - lastLevel = tObj->BaseLevel + (GLint)(tObj->MaxLod + 0.5); - lastLevel = MAX2(lastLevel, tObj->BaseLevel); - lastLevel = MIN2(lastLevel, level + baseImage->MaxLog2); - lastLevel = MIN2(lastLevel, tObj->MaxLevel); - lastLevel = MAX2(firstLevel, lastLevel); /* need at least one level */ + minLod = tObj->BaseLevel + (GLint)(tObj->MinLod); + minLod = MAX2(minLod, tObj->BaseLevel); + minLod = MIN2(minLod, tObj->MaxLevel); + maxLod = tObj->BaseLevel + (GLint)(tObj->MaxLod + 0.5); + maxLod = MIN2(maxLod, tObj->MaxLevel); + maxLod = MIN2(maxLod, tObj->Image[0][minLod]->MaxLog2 + minLod); + maxLod = MAX2(maxLod, minLod); /* need at least one level */ } break; case GL_TEXTURE_RECTANGLE_NV: case GL_TEXTURE_4D_SGIS: - firstLevel = lastLevel = 0; + minLod = maxLod = 0; break; default: return; } /* save these values */ - *pfirstLevel = firstLevel; - *plastLevel = lastLevel; + *pminLod = minLod; + *pmaxLod = maxLod; } - /** * Checks whether the given miptree can hold the given texture image at the * given face and level. */ GLboolean radeon_miptree_matches_image(radeon_mipmap_tree *mt, - struct gl_texture_image *texImage, GLuint face, GLuint level) + struct gl_texture_image *texImage, GLuint face, GLuint mtLevel) { - GLboolean isCompressed = _mesa_is_format_compressed(texImage->TexFormat); radeon_mipmap_level *lvl; - if (face >= mt->faces || level < mt->firstLevel || level > mt->lastLevel) - return GL_FALSE; - - if (texImage->InternalFormat != mt->internal_format || - isCompressed != mt->compressed) + if (face >= mt->faces || mtLevel > mt->numLevels) return GL_FALSE; - if (!isCompressed && - !mt->compressed && - _mesa_get_format_bytes(texImage->TexFormat) != mt->bpp) + if (texImage->TexFormat != mt->mesaFormat) return GL_FALSE; - lvl = &mt->levels[level - mt->firstLevel]; + lvl = &mt->levels[mtLevel]; if (lvl->width != texImage->Width || lvl->height != texImage->Height || lvl->depth != texImage->Depth) @@ -337,64 +312,72 @@ GLboolean radeon_miptree_matches_image(radeon_mipmap_tree *mt, return GL_TRUE; } - /** * Checks whether the given miptree has the right format to store the given texture object. */ -GLboolean radeon_miptree_matches_texture(radeon_mipmap_tree *mt, struct gl_texture_object *texObj) +static GLboolean radeon_miptree_matches_texture(radeon_mipmap_tree *mt, struct gl_texture_object *texObj) { struct gl_texture_image *firstImage; - GLuint compressed; - GLuint numfaces = 1; - GLuint firstLevel, lastLevel; - GLuint texelBytes; - - calculate_first_last_level(texObj, &firstLevel, &lastLevel, 0, texObj->BaseLevel); - if (texObj->Target == GL_TEXTURE_CUBE_MAP) - numfaces = 6; - - firstImage = texObj->Image[0][firstLevel]; - compressed = _mesa_is_format_compressed(firstImage->TexFormat) ? firstImage->TexFormat : 0; - texelBytes = _mesa_get_format_bytes(firstImage->TexFormat); - - return (mt->firstLevel == firstLevel && - mt->lastLevel == lastLevel && - mt->width0 == firstImage->Width && - mt->height0 == firstImage->Height && - mt->depth0 == firstImage->Depth && - mt->compressed == compressed && - (!mt->compressed ? (mt->bpp == texelBytes) : 1)); -} + unsigned numLevels; + radeon_mipmap_level *mtBaseLevel; + + if (texObj->BaseLevel < mt->baseLevel) + return GL_FALSE; + + mtBaseLevel = &mt->levels[texObj->BaseLevel - mt->baseLevel]; + firstImage = texObj->Image[0][texObj->BaseLevel]; + numLevels = MIN2(texObj->MaxLevel - texObj->BaseLevel + 1, firstImage->MaxLog2 + 1); + + if (RADEON_DEBUG & RADEON_TEXTURE) { + fprintf(stderr, "Checking if miptree %p matches texObj %p\n", mt, texObj); + fprintf(stderr, "target %d vs %d\n", mt->target, texObj->Target); + fprintf(stderr, "format %d vs %d\n", mt->mesaFormat, firstImage->TexFormat); + fprintf(stderr, "numLevels %d vs %d\n", mt->numLevels, numLevels); + fprintf(stderr, "width0 %d vs %d\n", mtBaseLevel->width, firstImage->Width); + fprintf(stderr, "height0 %d vs %d\n", mtBaseLevel->height, firstImage->Height); + fprintf(stderr, "depth0 %d vs %d\n", mtBaseLevel->depth, firstImage->Depth); + if (mt->target == texObj->Target && + mt->mesaFormat == firstImage->TexFormat && + mt->numLevels >= numLevels && + mtBaseLevel->width == firstImage->Width && + mtBaseLevel->height == firstImage->Height && + mtBaseLevel->depth == firstImage->Depth) { + fprintf(stderr, "MATCHED\n"); + } else { + fprintf(stderr, "NOT MATCHED\n"); + } + } + return (mt->target == texObj->Target && + mt->mesaFormat == firstImage->TexFormat && + mt->numLevels >= numLevels && + mtBaseLevel->width == firstImage->Width && + mtBaseLevel->height == firstImage->Height && + mtBaseLevel->depth == firstImage->Depth); +} /** - * Try to allocate a mipmap tree for the given texture that will fit the - * given image in the given position. + * Try to allocate a mipmap tree for the given texture object. + * @param[in] rmesa radeon context + * @param[in] t radeon texture object */ -void radeon_try_alloc_miptree(radeonContextPtr rmesa, radeonTexObj *t, - radeon_texture_image *image, GLuint face, GLuint level) +void radeon_try_alloc_miptree(radeonContextPtr rmesa, radeonTexObj *t) { - GLuint compressed = _mesa_is_format_compressed(image->base.TexFormat) ? image->base.TexFormat : 0; - GLuint numfaces = 1; - GLuint firstLevel, lastLevel; - GLuint texelBytes; + struct gl_texture_object *texObj = &t->base; + struct gl_texture_image *texImg = texObj->Image[0][texObj->BaseLevel]; + GLuint numLevels; assert(!t->mt); - calculate_first_last_level(&t->base, &firstLevel, &lastLevel, face, level); - if (t->base.Target == GL_TEXTURE_CUBE_MAP) - numfaces = 6; - - if (level != firstLevel || face >= numfaces) + if (!texImg) return; - texelBytes = _mesa_get_format_bytes(image->base.TexFormat); + numLevels = MIN2(texObj->MaxLevel - texObj->BaseLevel + 1, texImg->MaxLog2 + 1); - t->mt = radeon_miptree_create(rmesa, t, t->base.Target, - image->base.InternalFormat, - firstLevel, lastLevel, - image->base.Width, image->base.Height, image->base.Depth, - texelBytes, t->tile_bits, compressed); + t->mt = radeon_miptree_create(rmesa, t->base.Target, + texImg->TexFormat, texObj->BaseLevel, + numLevels, texImg->Width, texImg->Height, + texImg->Depth, t->tile_bits); } /* Although we use the image_offset[] array to store relative offsets @@ -406,21 +389,244 @@ void radeon_try_alloc_miptree(radeonContextPtr rmesa, radeonTexObj *t, void radeon_miptree_depth_offsets(radeon_mipmap_tree *mt, GLuint level, GLuint *offsets) { - if (mt->target != GL_TEXTURE_3D || mt->faces == 1) - offsets[0] = 0; - else { - int i; - for (i = 0; i < 6; i++) - offsets[i] = mt->levels[level].faces[i].offset; - } + if (mt->target != GL_TEXTURE_3D || mt->faces == 1) { + offsets[0] = 0; + } else { + int i; + for (i = 0; i < 6; i++) { + offsets[i] = mt->levels[level].faces[i].offset; + } + } } GLuint radeon_miptree_image_offset(radeon_mipmap_tree *mt, GLuint face, GLuint level) { - if (mt->target == GL_TEXTURE_CUBE_MAP_ARB) - return (mt->levels[level].faces[face].offset); - else - return mt->levels[level].faces[0].offset; + if (mt->target == GL_TEXTURE_CUBE_MAP_ARB) + return (mt->levels[level].faces[face].offset); + else + return mt->levels[level].faces[0].offset; +} + +/** + * Convert radeon miptree texture level to GL texture level + * @param[in] tObj texture object whom level is to be converted + * @param[in] level radeon miptree texture level + * @return GL texture level + */ +unsigned radeon_miptree_level_to_gl_level(struct gl_texture_object *tObj, unsigned level) +{ + return level + tObj->BaseLevel; +} + +/** + * Convert GL texture level to radeon miptree texture level + * @param[in] tObj texture object whom level is to be converted + * @param[in] level GL texture level + * @return radeon miptree texture level + */ +unsigned radeon_gl_level_to_miptree_level(struct gl_texture_object *tObj, unsigned level) +{ + return level - tObj->BaseLevel; +} + +/** + * Ensure that the given image is stored in the given miptree from now on. + */ +static void migrate_image_to_miptree(radeon_mipmap_tree *mt, + radeon_texture_image *image, + int face, int mtLevel) +{ + radeon_mipmap_level *dstlvl = &mt->levels[mtLevel]; + unsigned char *dest; + + assert(image->mt != mt); + assert(dstlvl->width == image->base.Width); + assert(dstlvl->height == image->base.Height); + assert(dstlvl->depth == image->base.Depth); + + radeon_bo_map(mt->bo, GL_TRUE); + dest = mt->bo->ptr + dstlvl->faces[face].offset; + + if (image->mt) { + /* Format etc. should match, so we really just need a memcpy(). + * In fact, that memcpy() could be done by the hardware in many + * cases, provided that we have a proper memory manager. + */ + assert(mt->mesaFormat == image->base.TexFormat); + + radeon_mipmap_level *srclvl = &image->mt->levels[image->mtlevel]; + + assert(srclvl->size == dstlvl->size); + assert(srclvl->rowstride == dstlvl->rowstride); + + radeon_bo_map(image->mt->bo, GL_FALSE); + + memcpy(dest, + image->mt->bo->ptr + srclvl->faces[face].offset, + dstlvl->size); + radeon_bo_unmap(image->mt->bo); + + radeon_miptree_unreference(&image->mt); + } else { + /* need to confirm this value is correct */ + if (_mesa_is_format_compressed(image->base.TexFormat)) { + unsigned size = _mesa_format_image_size(image->base.TexFormat, + image->base.Width, + image->base.Height, + image->base.Depth); + memcpy(dest, image->base.Data, size); + } else { + uint32_t srcrowstride; + uint32_t height; + + height = image->base.Height * image->base.Depth; + srcrowstride = image->base.Width * _mesa_get_format_bytes(image->base.TexFormat); + copy_rows(dest, dstlvl->rowstride, image->base.Data, srcrowstride, + height, srcrowstride); + } + + _mesa_free_texmemory(image->base.Data); + image->base.Data = 0; + } + + radeon_bo_unmap(mt->bo); + + radeon_miptree_reference(mt, &image->mt); + image->mtface = face; + image->mtlevel = mtLevel; +} + +/** + * Filter matching miptrees, and select one with the most of data. + * @param[in] texObj radeon texture object + * @param[in] firstLevel first texture level to check + * @param[in] lastLevel last texture level to check + */ +static radeon_mipmap_tree * get_biggest_matching_miptree(radeonTexObj *texObj, + unsigned firstLevel, + unsigned lastLevel) +{ + const unsigned numLevels = lastLevel - firstLevel + 1; + unsigned *mtSizes = calloc(numLevels, sizeof(unsigned)); + radeon_mipmap_tree **mts = calloc(numLevels, sizeof(radeon_mipmap_tree *)); + unsigned mtCount = 0; + unsigned maxMtIndex = 0; + radeon_mipmap_tree *tmp; + + for (unsigned level = firstLevel; level <= lastLevel; ++level) { + radeon_texture_image *img = get_radeon_texture_image(texObj->base.Image[0][level]); + unsigned found = 0; + // TODO: why this hack?? + if (!img) + break; + + if (!img->mt || !radeon_miptree_matches_texture(img->mt, &texObj->base)) + continue; + + for (int i = 0; i < mtCount; ++i) { + if (mts[i] == img->mt) { + found = 1; + mtSizes[i] += img->mt->levels[img->mtlevel].size; + break; + } + } + + if (!found) { + mtSizes[mtCount] += img->mt->levels[img->mtlevel].size; + mts[mtCount] = img->mt; + mtCount++; + } + } + + if (mtCount == 0) { + return NULL; + } + + for (int i = 1; i < mtCount; ++i) { + if (mtSizes[i] > mtSizes[maxMtIndex]) { + maxMtIndex = i; + } + } + + tmp = mts[maxMtIndex]; + free(mtSizes); + free(mts); + + return tmp; +} + +/** + * Validate texture mipmap tree. + * If individual images are stored in different mipmap trees + * use the mipmap tree that has the most of the correct data. + */ +int radeon_validate_texture_miptree(GLcontext * ctx, struct gl_texture_object *texObj) +{ + radeonContextPtr rmesa = RADEON_CONTEXT(ctx); + radeonTexObj *t = radeon_tex_obj(texObj); + + if (t->validated || t->image_override) { + return GL_TRUE; + } + + if (texObj->Image[0][texObj->BaseLevel]->Border > 0) + return GL_FALSE; + + _mesa_test_texobj_completeness(rmesa->glCtx, texObj); + if (!texObj->_Complete) { + return GL_FALSE; + } + + calculate_min_max_lod(&t->base, &t->minLod, &t->maxLod); + + if (RADEON_DEBUG & RADEON_TEXTURE) + fprintf(stderr, "%s: Validating texture %p now, minLod = %d, maxLod = %d\n", + __FUNCTION__, texObj ,t->minLod, t->maxLod); + + radeon_mipmap_tree *dst_miptree; + dst_miptree = get_biggest_matching_miptree(t, t->minLod, t->maxLod); + + if (!dst_miptree) { + radeon_miptree_unreference(&t->mt); + radeon_try_alloc_miptree(rmesa, t); + dst_miptree = t->mt; + } + + const unsigned faces = texObj->Target == GL_TEXTURE_CUBE_MAP ? 6 : 1; + unsigned face, level; + radeon_texture_image *img; + /* Validate only the levels that will actually be used during rendering */ + for (face = 0; face < faces; ++face) { + for (level = t->minLod; level <= t->maxLod; ++level) { + img = get_radeon_texture_image(texObj->Image[face][level]); + + if (RADEON_DEBUG & RADEON_TEXTURE) { + fprintf(stderr, "Checking image level %d, face %d, mt %p ... ", level, face, img->mt); + } + + if (img->mt != dst_miptree) { + if (RADEON_DEBUG & RADEON_TEXTURE) { + fprintf(stderr, "MIGRATING\n"); + } + migrate_image_to_miptree(dst_miptree, img, face, radeon_gl_level_to_miptree_level(texObj, level)); + } else if (RADEON_DEBUG & RADEON_TEXTURE) { + fprintf(stderr, "OK\n"); + } + } + } + + t->validated = GL_TRUE; + + return GL_TRUE; } + +uint32_t get_base_teximage_offset(radeonTexObj *texObj) +{ + if (!texObj->mt) { + return 0; + } else { + return radeon_miptree_image_offset(texObj->mt, 0, texObj->minLod); + } +}
\ No newline at end of file diff --git a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h index db28252da3..28b8485095 100644 --- a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h +++ b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h @@ -59,43 +59,38 @@ struct _radeon_mipmap_level { * changed. */ struct _radeon_mipmap_tree { - radeonContextPtr radeon; - radeonTexObj *t; struct radeon_bo *bo; GLuint refcount; GLuint totalsize; /** total size of the miptree, in bytes */ GLenum target; /** GL_TEXTURE_xxx */ - GLenum internal_format; + GLenum mesaFormat; /** MESA_FORMAT_xxx */ GLuint faces; /** # of faces: 6 for cubemaps, 1 otherwise */ - GLuint firstLevel; /** First mip level stored in this mipmap tree */ - GLuint lastLevel; /** Last mip level stored in this mipmap tree */ + GLuint baseLevel; /** gl_texture_object->baseLevel it was created for */ + GLuint numLevels; /** Number of mip levels stored in this mipmap tree */ GLuint width0; /** Width of firstLevel image */ GLuint height0; /** Height of firstLevel image */ GLuint depth0; /** Depth of firstLevel image */ - GLuint bpp; /** Bytes per texel */ GLuint tilebits; /** RADEON_TXO_xxx_TILE */ - GLuint compressed; /** MESA_FORMAT_xxx indicating a compressed format, or 0 if uncompressed */ radeon_mipmap_level levels[RADEON_MIPTREE_MAX_TEXTURE_LEVELS]; }; -radeon_mipmap_tree* radeon_miptree_create(radeonContextPtr rmesa, radeonTexObj *t, - GLenum target, GLenum internal_format, GLuint firstLevel, GLuint lastLevel, - GLuint width0, GLuint height0, GLuint depth0, - GLuint bpp, GLuint tilebits, GLuint compressed); -void radeon_miptree_reference(radeon_mipmap_tree *mt); -void radeon_miptree_unreference(radeon_mipmap_tree *mt); +void radeon_miptree_reference(radeon_mipmap_tree *mt, radeon_mipmap_tree **ptr); +void radeon_miptree_unreference(radeon_mipmap_tree **ptr); GLboolean radeon_miptree_matches_image(radeon_mipmap_tree *mt, struct gl_texture_image *texImage, GLuint face, GLuint level); -GLboolean radeon_miptree_matches_texture(radeon_mipmap_tree *mt, struct gl_texture_object *texObj); -void radeon_try_alloc_miptree(radeonContextPtr rmesa, radeonTexObj *t, - radeon_texture_image *texImage, GLuint face, GLuint level); +void radeon_try_alloc_miptree(radeonContextPtr rmesa, radeonTexObj *t); GLuint radeon_miptree_image_offset(radeon_mipmap_tree *mt, GLuint face, GLuint level); void radeon_miptree_depth_offsets(radeon_mipmap_tree *mt, GLuint level, GLuint *offsets); + +unsigned radeon_miptree_level_to_gl_level(struct gl_texture_object *tObj, unsigned level); +unsigned radeon_gl_level_to_miptree_level(struct gl_texture_object *tObj, unsigned level); + +uint32_t get_base_teximage_offset(radeonTexObj *texObj); #endif /* __RADEON_MIPMAP_TREE_H_ */ diff --git a/src/mesa/drivers/dri/radeon/radeon_queryobj.c b/src/mesa/drivers/dri/radeon/radeon_queryobj.c index 6539c36268..98117cdfc1 100644 --- a/src/mesa/drivers/dri/radeon/radeon_queryobj.c +++ b/src/mesa/drivers/dri/radeon/radeon_queryobj.c @@ -31,24 +31,11 @@ #include "main/imports.h" #include "main/simple_list.h" -static int radeonQueryIsFlushed(GLcontext *ctx, struct gl_query_object *q) -{ - radeonContextPtr radeon = RADEON_CONTEXT(ctx); - struct radeon_query_object *tmp, *query = (struct radeon_query_object *)q; - - foreach(tmp, &radeon->query.not_flushed_head) { - if (tmp == query) { - return 0; - } - } - - return 1; -} - static void radeonQueryGetResult(GLcontext *ctx, struct gl_query_object *q) { radeonContextPtr radeon = RADEON_CONTEXT(ctx); struct radeon_query_object *query = (struct radeon_query_object *)q; + uint32_t *result; int i; radeon_print(RADEON_STATE, RADEON_VERBOSE, @@ -56,6 +43,7 @@ static void radeonQueryGetResult(GLcontext *ctx, struct gl_query_object *q) __FUNCTION__, query->Base.Id, (int) query->Base.Result); radeon_bo_map(query->bo, GL_FALSE); + result = query->bo->ptr; query->Base.Result = 0; if (IS_R600_CLASS(radeon->radeonScreen)) { @@ -66,10 +54,11 @@ static void radeonQueryGetResult(GLcontext *ctx, struct gl_query_object *q) * hw writes zpass end counts to qwords 1, 3, 5, 7. * then we substract. MSB is the valid bit. */ - uint64_t *result = query->bo->ptr; - for (i = 0; i < 8; i += 2) { - uint64_t start = result[i]; - uint64_t end = result[i + 1]; + for (i = 0; i < 16; i += 4) { + uint64_t start = (uint64_t)LE32_TO_CPU(result[i]) | + (uint64_t)LE32_TO_CPU(result[i + 1]) << 32; + uint64_t end = (uint64_t)LE32_TO_CPU(result[i + 2]) | + (uint64_t)LE32_TO_CPU(result[i + 3]) << 32; if ((start & 0x8000000000000000) && (end & 0x8000000000000000)) { uint64_t query_count = end - start; query->Base.Result += query_count; @@ -79,10 +68,9 @@ static void radeonQueryGetResult(GLcontext *ctx, struct gl_query_object *q) "%d start: %lx, end: %lx %ld\n", i, start, end, end - start); } } else { - uint32_t *result = query->bo->ptr; for (i = 0; i < query->curr_offset/sizeof(uint32_t); ++i) { - query->Base.Result += result[i]; - radeon_print(RADEON_STATE, RADEON_TRACE, "result[%d] = %d\n", i, result[i]); + query->Base.Result += LE32_TO_CPU(result[i]); + radeon_print(RADEON_STATE, RADEON_TRACE, "result[%d] = %d\n", i, LE32_TO_CPU(result[i])); } } @@ -120,10 +108,11 @@ static void radeonDeleteQuery(GLcontext *ctx, struct gl_query_object *q) static void radeonWaitQuery(GLcontext *ctx, struct gl_query_object *q) { + radeonContextPtr radeon = RADEON_CONTEXT(ctx); struct radeon_query_object *query = (struct radeon_query_object *)q; /* If the cmdbuf with packets for this query hasn't been flushed yet, do it now */ - if (!radeonQueryIsFlushed(ctx, q)) + if (radeon_bo_is_referenced_by_cs(query->bo, radeon->cmdbuf.cs)) ctx->Driver.Flush(ctx); radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s: query id %d, bo %p, offset %d\n", __FUNCTION__, q->Id, query->bo, query->curr_offset); @@ -155,8 +144,6 @@ static void radeonBeginQuery(GLcontext *ctx, struct gl_query_object *q) radeon->query.queryobj.dirty = GL_TRUE; radeon->hw.is_dirty = GL_TRUE; - insert_at_tail(&radeon->query.not_flushed_head, query); - } void radeonEmitQueryEnd(GLcontext *ctx) @@ -204,7 +191,7 @@ static void radeonCheckQuery(GLcontext *ctx, struct gl_query_object *q) uint32_t domain; /* Need to perform a flush, as per ARB_occlusion_query spec */ - if (!radeonQueryIsFlushed(ctx, q)) { + if (radeon_bo_is_referenced_by_cs(query->bo, radeon->cmdbuf.cs)) { ctx->Driver.Flush(ctx); } diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.c b/src/mesa/drivers/dri/radeon/radeon_screen.c index 7a124a8be6..be2d8365ef 100644 --- a/src/mesa/drivers/dri/radeon/radeon_screen.c +++ b/src/mesa/drivers/dri/radeon/radeon_screen.c @@ -390,12 +390,14 @@ static int radeon_set_screen_flags(radeonScreenPtr screen, int device_id) screen->device_id = device_id; screen->chip_flags = 0; switch ( device_id ) { + case PCI_CHIP_RN50_515E: + case PCI_CHIP_RN50_5969: + return -1; + case PCI_CHIP_RADEON_LY: case PCI_CHIP_RADEON_LZ: case PCI_CHIP_RADEON_QY: case PCI_CHIP_RADEON_QZ: - case PCI_CHIP_RN50_515E: - case PCI_CHIP_RN50_5969: screen->chip_family = CHIP_FAMILY_RV100; break; @@ -1482,11 +1484,11 @@ radeonCreateBuffer( __DRIscreenPrivate *driScrnPriv, _mesa_initialize_framebuffer(&rfb->base, mesaVis); if (mesaVis->redBits == 5) - rgbFormat = MESA_FORMAT_RGB565; + rgbFormat = _mesa_little_endian() ? MESA_FORMAT_RGB565 : MESA_FORMAT_RGB565_REV; else if (mesaVis->alphaBits == 0) - rgbFormat = MESA_FORMAT_XRGB8888; + rgbFormat = _mesa_little_endian() ? MESA_FORMAT_XRGB8888 : MESA_FORMAT_XRGB8888_REV; else - rgbFormat = MESA_FORMAT_ARGB8888; + rgbFormat = _mesa_little_endian() ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB8888_REV; /* front color renderbuffer */ rfb->color_rb[0] = radeon_create_renderbuffer(rgbFormat, driDrawPriv); diff --git a/src/mesa/drivers/dri/radeon/radeon_span.c b/src/mesa/drivers/dri/radeon/radeon_span.c index 2bc7d31254..37904dc8dc 100644 --- a/src/mesa/drivers/dri/radeon/radeon_span.c +++ b/src/mesa/drivers/dri/radeon/radeon_span.c @@ -41,6 +41,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ #include "main/glheader.h" +#include "main/texformat.h" #include "swrast/swrast.h" #include "radeon_common.h" @@ -400,6 +401,18 @@ static GLubyte *radeon_ptr_2byte_8x2(const struct radeon_renderbuffer * rrb, #endif #include "spantmp2.h" +#define SPANTMP_PIXEL_FMT GL_RGB +#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_SHORT_5_6_5_REV + +#define TAG(x) radeon##x##_RGB565_REV +#define TAG2(x,y) radeon##x##_RGB565_REV##y +#if defined(RADEON_R600) +#define GET_PTR(X,Y) r600_ptr_color(rrb, (X) + x_off, (Y) + y_off) +#else +#define GET_PTR(X,Y) radeon_ptr_2byte_8x2(rrb, (X) + x_off, (Y) + y_off) +#endif +#include "spantmp2.h" + /* 16 bit, ARGB1555 color spanline and pixel functions */ #define SPANTMP_PIXEL_FMT GL_BGRA @@ -414,6 +427,18 @@ static GLubyte *radeon_ptr_2byte_8x2(const struct radeon_renderbuffer * rrb, #endif #include "spantmp2.h" +#define SPANTMP_PIXEL_FMT GL_BGRA +#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_SHORT_1_5_5_5 + +#define TAG(x) radeon##x##_ARGB1555_REV +#define TAG2(x,y) radeon##x##_ARGB1555_REV##y +#if defined(RADEON_R600) +#define GET_PTR(X,Y) r600_ptr_color(rrb, (X) + x_off, (Y) + y_off) +#else +#define GET_PTR(X,Y) radeon_ptr_2byte_8x2(rrb, (X) + x_off, (Y) + y_off) +#endif +#include "spantmp2.h" + /* 16 bit, RGBA4 color spanline and pixel functions */ #define SPANTMP_PIXEL_FMT GL_BGRA @@ -428,6 +453,18 @@ static GLubyte *radeon_ptr_2byte_8x2(const struct radeon_renderbuffer * rrb, #endif #include "spantmp2.h" +#define SPANTMP_PIXEL_FMT GL_BGRA +#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_SHORT_4_4_4_4 + +#define TAG(x) radeon##x##_ARGB4444_REV +#define TAG2(x,y) radeon##x##_ARGB4444_REV##y +#if defined(RADEON_R600) +#define GET_PTR(X,Y) r600_ptr_color(rrb, (X) + x_off, (Y) + y_off) +#else +#define GET_PTR(X,Y) radeon_ptr_2byte_8x2(rrb, (X) + x_off, (Y) + y_off) +#endif +#include "spantmp2.h" + /* 32 bit, xRGB8888 color spanline and pixel functions */ #define SPANTMP_PIXEL_FMT GL_BGRA @@ -472,6 +509,42 @@ static GLubyte *radeon_ptr_2byte_8x2(const struct radeon_renderbuffer * rrb, #endif #include "spantmp2.h" +/* 32 bit, BGRx8888 color spanline and pixel functions + */ +#define SPANTMP_PIXEL_FMT GL_BGRA +#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8 + +#define TAG(x) radeon##x##_BGRx8888 +#define TAG2(x,y) radeon##x##_BGRx8888##y +#if defined(RADEON_R600) +#define GET_VALUE(_x, _y) ((*(GLuint*)(r600_ptr_color(rrb, _x + x_off, _y + y_off)) | 0x000000ff)) +#define PUT_VALUE(_x, _y, d) { \ + GLuint *_ptr = (GLuint*)r600_ptr_color( rrb, _x + x_off, _y + y_off ); \ + *_ptr = d; \ +} while (0) +#else +#define GET_VALUE(_x, _y) ((*(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off, _y + y_off)) | 0x000000ff)) +#define PUT_VALUE(_x, _y, d) { \ + GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off ); \ + *_ptr = d; \ +} while (0) +#endif +#include "spantmp2.h" + +/* 32 bit, BGRA8888 color spanline and pixel functions + */ +#define SPANTMP_PIXEL_FMT GL_BGRA +#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8 + +#define TAG(x) radeon##x##_BGRA8888 +#define TAG2(x,y) radeon##x##_BGRA8888##y +#if defined(RADEON_R600) +#define GET_PTR(X,Y) r600_ptr_color(rrb, (X) + x_off, (Y) + y_off) +#else +#define GET_PTR(X,Y) radeon_ptr_4byte(rrb, (X) + x_off, (Y) + y_off) +#endif +#include "spantmp2.h" + /* ================================================================ * Depth buffer */ @@ -526,10 +599,10 @@ static GLubyte *radeon_ptr_2byte_8x2(const struct radeon_renderbuffer * rrb, #define WRITE_DEPTH( _x, _y, d ) \ do { \ GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off ); \ - GLuint tmp = *_ptr; \ + GLuint tmp = LE32_TO_CPU(*_ptr); \ tmp &= 0x000000ff; \ tmp |= ((d << 8) & 0xffffff00); \ - *_ptr = tmp; \ + *_ptr = CPU_TO_LE32(tmp); \ } while (0) #elif defined(RADEON_R600) #define WRITE_DEPTH( _x, _y, d ) \ @@ -544,26 +617,26 @@ do { \ #define WRITE_DEPTH( _x, _y, d ) \ do { \ GLuint *_ptr = (GLuint*)r200_depth_4byte( rrb, _x + x_off, _y + y_off ); \ - GLuint tmp = *_ptr; \ + GLuint tmp = LE32_TO_CPU(*_ptr); \ tmp &= 0xff000000; \ tmp |= ((d) & 0x00ffffff); \ - *_ptr = tmp; \ + *_ptr = CPU_TO_LE32(tmp); \ } while (0) #else #define WRITE_DEPTH( _x, _y, d ) \ do { \ GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off ); \ - GLuint tmp = *_ptr; \ + GLuint tmp = LE32_TO_CPU(*_ptr); \ tmp &= 0xff000000; \ tmp |= ((d) & 0x00ffffff); \ - *_ptr = tmp; \ + *_ptr = CPU_TO_LE32(tmp); \ } while (0) #endif #if defined(RADEON_R300) #define READ_DEPTH( d, _x, _y ) \ do { \ - d = (*(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off, _y + y_off)) & 0xffffff00) >> 8; \ + d = (LE32_TO_CPU(*(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off, _y + y_off))) & 0xffffff00) >> 8; \ }while(0) #elif defined(RADEON_R600) #define READ_DEPTH( d, _x, _y ) \ @@ -573,11 +646,11 @@ do { \ #elif defined(RADEON_R200) #define READ_DEPTH( d, _x, _y ) \ do { \ - d = *(GLuint*)(r200_depth_4byte(rrb, _x + x_off, _y + y_off)) & 0x00ffffff; \ + d = LE32_TO_CPU(*(GLuint*)(r200_depth_4byte(rrb, _x + x_off, _y + y_off))) & 0x00ffffff; \ }while(0) #else #define READ_DEPTH( d, _x, _y ) \ - d = *(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off, _y + y_off)) & 0x00ffffff; + d = LE32_TO_CPU(*(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off, _y + y_off))) & 0x00ffffff; #endif #define TAG(x) radeon##x##_z24 @@ -595,7 +668,7 @@ do { \ #define WRITE_DEPTH( _x, _y, d ) \ do { \ GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off ); \ - *_ptr = d; \ + *_ptr = CPU_TO_LE32((((d) & 0xff000000) >> 24) | (((d) & 0x00ffffff) << 8)); \ } while (0) #elif defined(RADEON_R600) #define WRITE_DEPTH( _x, _y, d ) \ @@ -615,20 +688,21 @@ do { \ #define WRITE_DEPTH( _x, _y, d ) \ do { \ GLuint *_ptr = (GLuint*)r200_depth_4byte( rrb, _x + x_off, _y + y_off ); \ - *_ptr = d; \ + *_ptr = CPU_TO_LE32(d); \ } while (0) #else #define WRITE_DEPTH( _x, _y, d ) \ do { \ GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off ); \ - *_ptr = d; \ + *_ptr = CPU_TO_LE32(d); \ } while (0) #endif #if defined(RADEON_R300) #define READ_DEPTH( d, _x, _y ) \ do { \ - d = (*(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off, _y + y_off))); \ + GLuint tmp = (*(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off, _y + y_off))); \ + d = LE32_TO_CPU(((tmp & 0x000000ff) << 24) | ((tmp & 0xffffff00) >> 8)); \ }while(0) #elif defined(RADEON_R600) #define READ_DEPTH( d, _x, _y ) \ @@ -639,11 +713,11 @@ do { \ #elif defined(RADEON_R200) #define READ_DEPTH( d, _x, _y ) \ do { \ - d = *(GLuint*)(r200_depth_4byte(rrb, _x + x_off, _y + y_off)); \ + d = LE32_TO_CPU(*(GLuint*)(r200_depth_4byte(rrb, _x + x_off, _y + y_off))); \ }while(0) #else #define READ_DEPTH( d, _x, _y ) do { \ - d = *(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off, _y + y_off )); \ + d = LE32_TO_CPU(*(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off, _y + y_off))); \ } while (0) #endif @@ -660,10 +734,10 @@ do { \ #define WRITE_STENCIL( _x, _y, d ) \ do { \ GLuint *_ptr = (GLuint*)radeon_ptr_4byte(rrb, _x + x_off, _y + y_off); \ - GLuint tmp = *_ptr; \ + GLuint tmp = LE32_TO_CPU(*_ptr); \ tmp &= 0xffffff00; \ tmp |= (d) & 0xff; \ - *_ptr = tmp; \ + *_ptr = CPU_TO_LE32(tmp); \ } while (0) #elif defined(RADEON_R600) #define WRITE_STENCIL( _x, _y, d ) \ @@ -678,19 +752,19 @@ do { \ #define WRITE_STENCIL( _x, _y, d ) \ do { \ GLuint *_ptr = (GLuint*)r200_depth_4byte(rrb, _x + x_off, _y + y_off); \ - GLuint tmp = *_ptr; \ + GLuint tmp = LE32_TO_CPU(*_ptr); \ tmp &= 0x00ffffff; \ tmp |= (((d) & 0xff) << 24); \ - *_ptr = tmp; \ + *_ptr = CPU_TO_LE32(tmp); \ } while (0) #else #define WRITE_STENCIL( _x, _y, d ) \ do { \ GLuint *_ptr = (GLuint*)radeon_ptr_4byte(rrb, _x + x_off, _y + y_off); \ - GLuint tmp = *_ptr; \ + GLuint tmp = LE32_TO_CPU(*_ptr); \ tmp &= 0x00ffffff; \ tmp |= (((d) & 0xff) << 24); \ - *_ptr = tmp; \ + *_ptr = CPU_TO_LE32(tmp); \ } while (0) #endif @@ -698,7 +772,7 @@ do { \ #define READ_STENCIL( d, _x, _y ) \ do { \ GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off ); \ - GLuint tmp = *_ptr; \ + GLuint tmp = LE32_TO_CPU(*_ptr); \ d = tmp & 0x000000ff; \ } while (0) #elif defined(RADEON_R600) @@ -712,14 +786,14 @@ do { \ #define READ_STENCIL( d, _x, _y ) \ do { \ GLuint *_ptr = (GLuint*)r200_depth_4byte( rrb, _x + x_off, _y + y_off ); \ - GLuint tmp = *_ptr; \ + GLuint tmp = LE32_TO_CPU(*_ptr); \ d = (tmp & 0xff000000) >> 24; \ } while (0) #else #define READ_STENCIL( d, _x, _y ) \ do { \ GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off ); \ - GLuint tmp = *_ptr; \ + GLuint tmp = LE32_TO_CPU(*_ptr); \ d = (tmp & 0xff000000) >> 24; \ } while (0) #endif @@ -848,14 +922,24 @@ static void radeonSetSpanFunctions(struct radeon_renderbuffer *rrb) { if (rrb->base.Format == MESA_FORMAT_RGB565) { radeonInitPointers_RGB565(&rrb->base); + } else if (rrb->base.Format == MESA_FORMAT_RGB565_REV) { + radeonInitPointers_RGB565_REV(&rrb->base); } else if (rrb->base.Format == MESA_FORMAT_XRGB8888) { radeonInitPointers_xRGB8888(&rrb->base); + } else if (rrb->base.Format == MESA_FORMAT_XRGB8888_REV) { + radeonInitPointers_BGRx8888(&rrb->base); } else if (rrb->base.Format == MESA_FORMAT_ARGB8888) { radeonInitPointers_ARGB8888(&rrb->base); + } else if (rrb->base.Format == MESA_FORMAT_ARGB8888_REV) { + radeonInitPointers_BGRA8888(&rrb->base); } else if (rrb->base.Format == MESA_FORMAT_ARGB4444) { radeonInitPointers_ARGB4444(&rrb->base); + } else if (rrb->base.Format == MESA_FORMAT_ARGB4444_REV) { + radeonInitPointers_ARGB4444_REV(&rrb->base); } else if (rrb->base.Format == MESA_FORMAT_ARGB1555) { radeonInitPointers_ARGB1555(&rrb->base); + } else if (rrb->base.Format == MESA_FORMAT_ARGB1555_REV) { + radeonInitPointers_ARGB1555_REV(&rrb->base); } else if (rrb->base.Format == MESA_FORMAT_Z16) { radeonInitDepthPointers_z16(&rrb->base); } else if (rrb->base.Format == MESA_FORMAT_X8_Z24) { diff --git a/src/mesa/drivers/dri/radeon/radeon_state_init.c b/src/mesa/drivers/dri/radeon/radeon_state_init.c index 2d19220d8a..dd82888254 100644 --- a/src/mesa/drivers/dri/radeon/radeon_state_init.c +++ b/src/mesa/drivers/dri/radeon/radeon_state_init.c @@ -645,11 +645,11 @@ static void tex_emit_cs(GLcontext *ctx, struct radeon_state_atom *atom) OUT_BATCH(CP_PACKET0(RADEON_PP_TXOFFSET_0 + (24 * i), 0)); if (t->mt && !t->image_override) { if ((ctx->Texture.Unit[i]._ReallyEnabled & TEXTURE_CUBE_BIT)) { - lvl = &t->mt->levels[0]; + lvl = &t->mt->levels[t->minLod]; OUT_BATCH_RELOC(lvl->faces[5].offset, t->mt->bo, lvl->faces[5].offset, RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0); } else { - OUT_BATCH_RELOC(t->tile_bits, t->mt->bo, 0, + OUT_BATCH_RELOC(t->tile_bits, t->mt->bo, get_base_teximage_offset(t), RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0); } } else { diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.c b/src/mesa/drivers/dri/radeon/radeon_swtcl.c index e61f59eaea..8bf1bfbc57 100644 --- a/src/mesa/drivers/dri/radeon/radeon_swtcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.c @@ -179,7 +179,7 @@ static void radeonSetVertexFormat( GLcontext *ctx ) for (i = 0; i < ctx->Const.MaxTextureUnits; i++) { if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX(i) )) { - GLuint sz = VB->TexCoordPtr[i]->size; + GLuint sz = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->size; switch (sz) { case 1: @@ -309,7 +309,7 @@ void r100_swtcl_flush(GLcontext *ctx, uint32_t current_offset) radeonEmitState(&rmesa->radeon); radeonEmitVertexAOS( rmesa, rmesa->radeon.swtcl.vertex_size, - first_elem(&rmesa->radeon.dma.reserved)->bo, + rmesa->radeon.swtcl.bo, current_offset); diff --git a/src/mesa/drivers/dri/radeon/radeon_tex.c b/src/mesa/drivers/dri/radeon/radeon_tex.c index 60981aada2..749ab75f20 100644 --- a/src/mesa/drivers/dri/radeon/radeon_tex.c +++ b/src/mesa/drivers/dri/radeon/radeon_tex.c @@ -348,17 +348,7 @@ static void radeonTexParameter( GLcontext *ctx, GLenum target, case GL_TEXTURE_MAX_LEVEL: case GL_TEXTURE_MIN_LOD: case GL_TEXTURE_MAX_LOD: - - /* This isn't the most efficient solution but there doesn't appear to - * be a nice alternative. Since there's no LOD clamping, - * we just have to rely on loading the right subset of mipmap levels - * to simulate a clamped LOD. - */ - if (t->mt) { - radeon_miptree_unreference(t->mt); - t->mt = 0; - t->validated = GL_FALSE; - } + t->validated = GL_FALSE; break; default: @@ -388,10 +378,8 @@ static void radeonDeleteTexture( GLcontext *ctx, } } - if (t->mt) { - radeon_miptree_unreference(t->mt); - t->mt = 0; - } + radeon_miptree_unreference(&t->mt); + /* Free mipmap images and the texture object itself */ _mesa_delete_texture_object(ctx, texObj); } diff --git a/src/mesa/drivers/dri/radeon/radeon_texstate.c b/src/mesa/drivers/dri/radeon/radeon_texstate.c index 429977a8bc..3cbe3b4725 100644 --- a/src/mesa/drivers/dri/radeon/radeon_texstate.c +++ b/src/mesa/drivers/dri/radeon/radeon_texstate.c @@ -699,14 +699,10 @@ void radeonSetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_ radeon_bo_unref(rImage->bo); rImage->bo = NULL; } - if (t->mt) { - radeon_miptree_unreference(t->mt); - t->mt = NULL; - } - if (rImage->mt) { - radeon_miptree_unreference(rImage->mt); - rImage->mt = NULL; - } + + radeon_miptree_unreference(&t->mt); + radeon_miptree_unreference(&rImage->mt); + _mesa_init_teximage_fields(radeon->glCtx, target, texImage, rb->base.Width, rb->base.Height, 1, 0, rb->cpp); texImage->RowStride = rb->pitch / rb->cpp; @@ -718,8 +714,6 @@ void radeonSetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_ t->tile_bits = 0; t->image_override = GL_TRUE; t->override_offset = 0; - t->pp_txpitch &= (1 << 13) -1; - pitch_val = rb->pitch; switch (rb->cpp) { case 4: if (glx_texture_format == GLX_TEXTURE_FORMAT_RGB_EXT) @@ -738,12 +732,17 @@ void radeonSetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_ t->pp_txfilter |= tx_table[MESA_FORMAT_RGB565].filter; break; } - t->pp_txsize = ((rb->base.Width - 1) << RADEON_TEX_USIZE_SHIFT) - | ((rb->base.Height - 1) << RADEON_TEX_VSIZE_SHIFT); - t->pp_txformat |= RADEON_TXFORMAT_NON_POWER2; - t->pp_txpitch = pitch_val; - t->pp_txpitch -= 32; + t->pp_txpitch &= (1 << 13) -1; + pitch_val = rb->pitch; + + t->pp_txsize = ((rb->base.Width - 1) << RADEON_TEX_USIZE_SHIFT) + | ((rb->base.Height - 1) << RADEON_TEX_VSIZE_SHIFT); + if (target == GL_TEXTURE_RECTANGLE_NV) { + t->pp_txformat |= RADEON_TXFORMAT_NON_POWER2; + t->pp_txpitch = pitch_val; + t->pp_txpitch -= 32; + } t->validated = GL_TRUE; _mesa_unlock_texture(radeon->glCtx, texObj); return; @@ -1021,7 +1020,7 @@ static GLboolean setup_hardware_state(r100ContextPtr rmesa, radeonTexObj *t, int return GL_TRUE; } - firstImage = t->base.Image[0][t->mt->firstLevel]; + firstImage = t->base.Image[0][t->minLod]; if (firstImage->Border > 0) { fprintf(stderr, "%s: border\n", __FUNCTION__); @@ -1049,9 +1048,9 @@ static GLboolean setup_hardware_state(r100ContextPtr rmesa, radeonTexObj *t, int return GL_FALSE; } } - + t->pp_txfilter &= ~RADEON_MAX_MIP_LEVEL_MASK; - t->pp_txfilter |= (t->mt->lastLevel - t->mt->firstLevel) << RADEON_MAX_MIP_LEVEL_SHIFT; + t->pp_txfilter |= (t->maxLod - t->minLod) << RADEON_MAX_MIP_LEVEL_SHIFT; t->pp_txformat &= ~(RADEON_TXFORMAT_WIDTH_MASK | RADEON_TXFORMAT_HEIGHT_MASK | @@ -1060,9 +1059,9 @@ static GLboolean setup_hardware_state(r100ContextPtr rmesa, radeonTexObj *t, int RADEON_TXFORMAT_F5_HEIGHT_MASK); t->pp_txformat |= ((log2Width << RADEON_TXFORMAT_WIDTH_SHIFT) | (log2Height << RADEON_TXFORMAT_HEIGHT_SHIFT)); - + t->tile_bits = 0; - + if (t->base.Target == GL_TEXTURE_CUBE_MAP) { ASSERT(log2Width == log2Height); t->pp_txformat |= ((log2Width << RADEON_TXFORMAT_F5_WIDTH_SHIFT) | diff --git a/src/mesa/drivers/dri/radeon/radeon_texture.c b/src/mesa/drivers/dri/radeon/radeon_texture.c index baa99b752b..1ee9e2792a 100644 --- a/src/mesa/drivers/dri/radeon/radeon_texture.c +++ b/src/mesa/drivers/dri/radeon/radeon_texture.c @@ -1,4 +1,5 @@ /* + * Copyright (C) 2009 Maciej Cencora. * Copyright (C) 2008 Nicolai Haehnle. * Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. * @@ -46,7 +47,7 @@ #include "radeon_mipmap_tree.h" -static void copy_rows(void* dst, GLuint dststride, const void* src, GLuint srcstride, +void copy_rows(void* dst, GLuint dststride, const void* src, GLuint srcstride, GLuint numrows, GLuint rowsize) { assert(rowsize <= dststride); @@ -81,8 +82,7 @@ void radeonFreeTexImageData(GLcontext *ctx, struct gl_texture_image *timage) radeon_texture_image* image = get_radeon_texture_image(timage); if (image->mt) { - radeon_miptree_unreference(image->mt); - image->mt = 0; + radeon_miptree_unreference(&image->mt); assert(!image->base.Data); } else { _mesa_free_texture_image_data(ctx, timage); @@ -108,7 +108,7 @@ static void teximage_set_map_data(radeon_texture_image *image) lvl = &image->mt->levels[image->mtlevel]; image->base.Data = image->mt->bo->ptr + lvl->faces[image->mtface].offset; - image->base.RowStride = lvl->rowstride / image->mt->bpp; + image->base.RowStride = lvl->rowstride / _mesa_get_format_bytes(image->base.TexFormat); } @@ -174,7 +174,7 @@ void radeonMapTexture(GLcontext *ctx, struct gl_texture_object *texObj) radeon_bo_map(t->mt->bo, GL_FALSE); for(face = 0; face < t->mt->faces; ++face) { - for(level = t->mt->firstLevel; level <= t->mt->lastLevel; ++level) + for(level = t->minLod; level <= t->maxLod; ++level) teximage_set_map_data(get_radeon_texture_image(texObj->Image[face][level])); } } @@ -191,7 +191,7 @@ void radeonUnmapTexture(GLcontext *ctx, struct gl_texture_object *texObj) return; for(face = 0; face < t->mt->faces; ++face) { - for(level = t->mt->firstLevel; level <= t->mt->lastLevel; ++level) + for(level = t->minLod; level <= t->maxLod; ++level) texObj->Image[face][level]->Data = 0; } radeon_bo_unmap(t->mt->bo); @@ -240,8 +240,7 @@ static void radeon_generate_mipmap(GLcontext *ctx, GLenum target, image->mtlevel = i; image->mtface = face; - radeon_miptree_unreference(image->mt); - image->mt = NULL; + radeon_miptree_unreference(&image->mt); } } @@ -510,6 +509,137 @@ gl_format radeonChooseTextureFormat(GLcontext * ctx, return MESA_FORMAT_NONE; /* never get here */ } +static void teximage_assign_miptree(radeonContextPtr rmesa, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage, + unsigned face, + unsigned level) +{ + radeonTexObj *t = radeon_tex_obj(texObj); + radeon_texture_image* image = get_radeon_texture_image(texImage); + + /* Try using current miptree, or create new if there isn't any */ + if (!t->mt || !radeon_miptree_matches_image(t->mt, texImage, face, + radeon_gl_level_to_miptree_level(texObj, level))) { + radeon_miptree_unreference(&t->mt); + radeon_try_alloc_miptree(rmesa, t); + } + + /* Miptree alocation may have failed, + * when there was no image for baselevel specified */ + if (t->mt) { + image->mtface = face; + image->mtlevel = radeon_gl_level_to_miptree_level(texObj, level); + radeon_miptree_reference(t->mt, &image->mt); + } +} + +static GLuint * allocate_image_offsets(GLcontext *ctx, + unsigned alignedWidth, + unsigned height, + unsigned depth) +{ + int i; + GLuint *offsets; + + offsets = _mesa_malloc(depth * sizeof(GLuint)) ; + if (!offsets) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTex[Sub]Image"); + return NULL; + } + + for (i = 0; i < depth; ++i) { + offsets[i] = alignedWidth * height * i; + } + + return offsets; +} + +/** + * Update a subregion of the given texture image. + */ +static void radeon_store_teximage(GLcontext* ctx, int dims, + GLint xoffset, GLint yoffset, GLint zoffset, + GLsizei width, GLsizei height, GLsizei depth, + GLsizei imageSize, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage, + int compressed) +{ + radeonTexObj *t = radeon_tex_obj(texObj); + radeon_texture_image* image = get_radeon_texture_image(texImage); + + GLuint dstRowStride; + GLuint *dstImageOffsets; + + if (image->mt) { + dstRowStride = image->mt->levels[image->mtlevel].rowstride; + } else if (t->bo) { + /* TFP case */ + /* TODO */ + assert(0); + } else { + dstRowStride = _mesa_format_row_stride(texImage->TexFormat, texImage->Width); + } + + if (dims == 3) { + unsigned alignedWidth = dstRowStride/_mesa_get_format_bytes(texImage->TexFormat); + dstImageOffsets = allocate_image_offsets(ctx, alignedWidth, texImage->Height, texImage->Depth); + if (!dstImageOffsets) { + return; + } + } else { + dstImageOffsets = texImage->ImageOffsets; + } + + radeon_teximage_map(image, GL_TRUE); + + if (compressed) { + uint32_t srcRowStride, bytesPerRow, rows, block_width, block_height; + GLubyte *img_start; + + _mesa_get_format_block_size(texImage->TexFormat, &block_width, &block_height); + + if (!image->mt) { + dstRowStride = _mesa_format_row_stride(texImage->TexFormat, texImage->Width); + img_start = _mesa_compressed_image_address(xoffset, yoffset, 0, + texImage->TexFormat, + texImage->Width, texImage->Data); + } + else { + uint32_t offset; + offset = dstRowStride / _mesa_get_format_bytes(texImage->TexFormat) * yoffset / block_height + xoffset / block_width; + offset *= _mesa_get_format_bytes(texImage->TexFormat); + img_start = texImage->Data + offset; + } + srcRowStride = _mesa_format_row_stride(texImage->TexFormat, width); + bytesPerRow = srcRowStride; + rows = (height + block_height - 1) / block_height; + + copy_rows(img_start, dstRowStride, pixels, srcRowStride, rows, bytesPerRow); + } + else { + if (!_mesa_texstore(ctx, dims, texImage->_BaseFormat, + texImage->TexFormat, texImage->Data, + xoffset, yoffset, zoffset, + dstRowStride, + dstImageOffsets, + width, height, depth, + format, type, pixels, packing)) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage"); + } + } + + if (dims == 3) { + _mesa_free(dstImageOffsets); + } + + radeon_teximage_unmap(image); +} + /** * All glTexImage calls go through this function. */ @@ -528,13 +658,17 @@ static void radeon_teximage( radeonContextPtr rmesa = RADEON_CONTEXT(ctx); radeonTexObj* t = radeon_tex_obj(texObj); radeon_texture_image* image = get_radeon_texture_image(texImage); - GLuint dstRowStride; GLint postConvWidth = width; GLint postConvHeight = height; - GLuint texelBytes; GLuint face = radeon_face_for_target(target); - radeon_firevertices(rmesa); + { + struct radeon_bo *bo; + bo = !image->mt ? image->bo : image->mt->bo; + if (bo && radeon_bo_is_referenced_by_cs(bo, rmesa->cmdbuf.cs)) { + radeon_firevertices(rmesa); + } + } t->validated = GL_FALSE; @@ -543,53 +677,30 @@ static void radeon_teximage( &postConvHeight); } - if (_mesa_is_format_compressed(texImage->TexFormat)) { - texelBytes = 0; - } else { - texelBytes = _mesa_get_format_bytes(texImage->TexFormat); + if (!_mesa_is_format_compressed(texImage->TexFormat)) { + GLuint texelBytes = _mesa_get_format_bytes(texImage->TexFormat); /* Minimum pitch of 32 bytes */ if (postConvWidth * texelBytes < 32) { - postConvWidth = 32 / texelBytes; - texImage->RowStride = postConvWidth; + postConvWidth = 32 / texelBytes; + texImage->RowStride = postConvWidth; } - if (!image->mt) { + if (!image->mt) { assert(texImage->RowStride == postConvWidth); } } - /* Allocate memory for image */ - radeonFreeTexImageData(ctx, texImage); /* Mesa core only clears texImage->Data but not image->mt */ + /* Mesa core only clears texImage->Data but not image->mt */ + radeonFreeTexImageData(ctx, texImage); - if (t->mt && - t->mt->firstLevel == level && - t->mt->lastLevel == level && - t->mt->target != GL_TEXTURE_CUBE_MAP_ARB && - !radeon_miptree_matches_image(t->mt, texImage, face, level)) { - radeon_miptree_unreference(t->mt); - t->mt = NULL; - } - - if (!t->mt) - radeon_try_alloc_miptree(rmesa, t, image, face, level); - if (t->mt && radeon_miptree_matches_image(t->mt, texImage, face, level)) { - radeon_mipmap_level *lvl; - image->mt = t->mt; - image->mtlevel = level - t->mt->firstLevel; - image->mtface = face; - radeon_miptree_reference(t->mt); - lvl = &image->mt->levels[image->mtlevel]; - dstRowStride = lvl->rowstride; - } else { - int size; - if (_mesa_is_format_compressed(texImage->TexFormat)) { - size = _mesa_format_image_size(texImage->TexFormat, - texImage->Width, - texImage->Height, - texImage->Depth); - } else { - size = texImage->Width * texImage->Height * texImage->Depth * _mesa_get_format_bytes(texImage->TexFormat); + if (!t->bo) { + teximage_assign_miptree(rmesa, texObj, texImage, face, level); + if (!t->mt) { + int size = _mesa_format_image_size(texImage->TexFormat, + texImage->Width, + texImage->Height, + texImage->Depth); + texImage->Data = _mesa_alloc_texmemory(size); } - texImage->Data = _mesa_alloc_texmemory(size); } /* Upload texture image; note that the spec allows pixels to be NULL */ @@ -603,65 +714,16 @@ static void radeon_teximage( } if (pixels) { - radeon_teximage_map(image, GL_TRUE); - if (compressed) { - if (image->mt) { - uint32_t srcRowStride, bytesPerRow, rows; - srcRowStride = _mesa_format_row_stride(texImage->TexFormat, width); - bytesPerRow = srcRowStride; - rows = (height + 3) / 4; - copy_rows(texImage->Data, image->mt->levels[level].rowstride, - pixels, srcRowStride, rows, bytesPerRow); - } else { - memcpy(texImage->Data, pixels, imageSize); - } - } else { - GLuint dstRowStride; - GLuint *dstImageOffsets; - - if (image->mt) { - radeon_mipmap_level *lvl = &image->mt->levels[image->mtlevel]; - dstRowStride = lvl->rowstride; - } else { - dstRowStride = texImage->Width * _mesa_get_format_bytes(texImage->TexFormat); - } - - if (dims == 3) { - int i; - - dstImageOffsets = _mesa_malloc(depth * sizeof(GLuint)) ; - if (!dstImageOffsets) - _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); - - for (i = 0; i < depth; ++i) { - dstImageOffsets[i] = dstRowStride/_mesa_get_format_bytes(texImage->TexFormat) * height * i; - } - } else { - dstImageOffsets = texImage->ImageOffsets; - } - - if (!_mesa_texstore(ctx, dims, - texImage->_BaseFormat, - texImage->TexFormat, - texImage->Data, 0, 0, 0, /* dstX/Y/Zoffset */ - dstRowStride, - dstImageOffsets, - width, height, depth, - format, type, pixels, packing)) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); - } - - if (dims == 3) - _mesa_free(dstImageOffsets); - } + radeon_store_teximage(ctx, dims, + 0, 0, 0, + width, height, depth, + imageSize, format, type, + pixels, packing, + texObj, texImage, + compressed); } _mesa_unmap_teximage_pbo(ctx, packing); - - if (pixels) - radeon_teximage_unmap(image); - - } void radeonTexImage1D(GLcontext * ctx, GLenum target, GLint level, @@ -714,7 +776,7 @@ void radeonTexImage3D(GLcontext * ctx, GLenum target, GLint level, } /** - * Update a subregion of the given texture image. + * All glTexSubImage calls go through this function. */ static void radeon_texsubimage(GLcontext* ctx, int dims, GLenum target, int level, GLint xoffset, GLint yoffset, GLint zoffset, @@ -731,66 +793,34 @@ static void radeon_texsubimage(GLcontext* ctx, int dims, GLenum target, int leve radeonTexObj* t = radeon_tex_obj(texObj); radeon_texture_image* image = get_radeon_texture_image(texImage); - radeon_firevertices(rmesa); + { + struct radeon_bo *bo; + bo = !image->mt ? image->bo : image->mt->bo; + if (bo && radeon_bo_is_referenced_by_cs(bo, rmesa->cmdbuf.cs)) { + radeon_firevertices(rmesa); + } + } t->validated = GL_FALSE; if (compressed) { pixels = _mesa_validate_pbo_compressed_teximage( - ctx, imageSize, pixels, packing, "glCompressedTexImage"); + ctx, imageSize, pixels, packing, "glCompressedTexSubImage"); } else { pixels = _mesa_validate_pbo_teximage(ctx, dims, - width, height, depth, format, type, pixels, packing, "glTexSubImage1D"); + width, height, depth, format, type, pixels, packing, "glTexSubImage"); } if (pixels) { - GLint dstRowStride; - radeon_teximage_map(image, GL_TRUE); - - if (image->mt) { - radeon_mipmap_level *lvl = &image->mt->levels[image->mtlevel]; - dstRowStride = lvl->rowstride; - } else { - dstRowStride = texImage->RowStride * _mesa_get_format_bytes(texImage->TexFormat); - } - - if (compressed) { - uint32_t srcRowStride, bytesPerRow, rows; - GLubyte *img_start; - if (!image->mt) { - dstRowStride = _mesa_format_row_stride(texImage->TexFormat, texImage->Width); - img_start = _mesa_compressed_image_address(xoffset, yoffset, 0, - texImage->TexFormat, - texImage->Width, texImage->Data); - } - else { - uint32_t blocks_x = dstRowStride / (image->mt->bpp * 4); - img_start = texImage->Data + image->mt->bpp * 4 * (blocks_x * (yoffset / 4) + xoffset / 4); - } - srcRowStride = _mesa_format_row_stride(texImage->TexFormat, width); - bytesPerRow = srcRowStride; - rows = (height + 3) / 4; - - copy_rows(img_start, dstRowStride, pixels, srcRowStride, rows, bytesPerRow); - - } - else { - if (!_mesa_texstore(ctx, dims, texImage->_BaseFormat, - texImage->TexFormat, texImage->Data, - xoffset, yoffset, zoffset, - dstRowStride, - texImage->ImageOffsets, - width, height, depth, - format, type, pixels, packing)) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage"); - } - } + radeon_store_teximage(ctx, dims, + xoffset, yoffset, zoffset, + width, height, depth, + imageSize, format, type, + pixels, packing, + texObj, texImage, + compressed); } - radeon_teximage_unmap(image); - _mesa_unmap_teximage_pbo(ctx, packing); - - } void radeonTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, @@ -846,143 +876,6 @@ void radeonTexSubImage3D(GLcontext * ctx, GLenum target, GLint level, format, type, pixels, packing, texObj, texImage, 0); } - - -/** - * Ensure that the given image is stored in the given miptree from now on. - */ -static void migrate_image_to_miptree(radeon_mipmap_tree *mt, radeon_texture_image *image, int face, int level) -{ - radeon_mipmap_level *dstlvl = &mt->levels[level - mt->firstLevel]; - unsigned char *dest; - - assert(image->mt != mt); - assert(dstlvl->width == image->base.Width); - assert(dstlvl->height == image->base.Height); - assert(dstlvl->depth == image->base.Depth); - - - radeon_bo_map(mt->bo, GL_TRUE); - dest = mt->bo->ptr + dstlvl->faces[face].offset; - - if (image->mt) { - /* Format etc. should match, so we really just need a memcpy(). - * In fact, that memcpy() could be done by the hardware in many - * cases, provided that we have a proper memory manager. - */ - radeon_mipmap_level *srclvl = &image->mt->levels[image->mtlevel-image->mt->firstLevel]; - - assert(srclvl->size == dstlvl->size); - assert(srclvl->rowstride == dstlvl->rowstride); - - radeon_bo_map(image->mt->bo, GL_FALSE); - - memcpy(dest, - image->mt->bo->ptr + srclvl->faces[face].offset, - dstlvl->size); - radeon_bo_unmap(image->mt->bo); - - radeon_miptree_unreference(image->mt); - } else { - uint32_t srcrowstride; - uint32_t height; - /* need to confirm this value is correct */ - if (mt->compressed) { - height = (image->base.Height + 3) / 4; - srcrowstride = _mesa_format_row_stride(image->base.TexFormat, image->base.Width); - } else { - height = image->base.Height * image->base.Depth; - srcrowstride = image->base.Width * _mesa_get_format_bytes(image->base.TexFormat); - } - -// if (mt->tilebits) -// WARN_ONCE("%s: tiling not supported yet", __FUNCTION__); - - copy_rows(dest, dstlvl->rowstride, image->base.Data, srcrowstride, - height, srcrowstride); - - _mesa_free_texmemory(image->base.Data); - image->base.Data = 0; - } - - radeon_bo_unmap(mt->bo); - - image->mt = mt; - image->mtface = face; - image->mtlevel = level; - radeon_miptree_reference(image->mt); -} - -int radeon_validate_texture_miptree(GLcontext * ctx, struct gl_texture_object *texObj) -{ - radeonContextPtr rmesa = RADEON_CONTEXT(ctx); - radeonTexObj *t = radeon_tex_obj(texObj); - radeon_texture_image *baseimage = get_radeon_texture_image(texObj->Image[0][texObj->BaseLevel]); - int face, level; - - if (t->validated || t->image_override) - return GL_TRUE; - - if (RADEON_DEBUG & RADEON_TEXTURE) - fprintf(stderr, "%s: Validating texture %p now\n", __FUNCTION__, texObj); - - if (baseimage->base.Border > 0) - return GL_FALSE; - - /* Ensure a matching miptree exists. - * - * Differing mipmap trees can result when the app uses TexImage to - * change texture dimensions. - * - * Prefer to use base image's miptree if it - * exists, since that most likely contains more valid data (remember - * that the base level is usually significantly larger than the rest - * of the miptree, so cubemaps are the only possible exception). - */ - if (baseimage->mt && - baseimage->mt != t->mt && - radeon_miptree_matches_texture(baseimage->mt, &t->base)) { - radeon_miptree_unreference(t->mt); - t->mt = baseimage->mt; - radeon_miptree_reference(t->mt); - } else if (t->mt && !radeon_miptree_matches_texture(t->mt, &t->base)) { - radeon_miptree_unreference(t->mt); - t->mt = 0; - } - - if (!t->mt) { - if (RADEON_DEBUG & RADEON_TEXTURE) - fprintf(stderr, " Allocate new miptree\n"); - radeon_try_alloc_miptree(rmesa, t, baseimage, 0, texObj->BaseLevel); - if (!t->mt) { - _mesa_problem(ctx, "radeon_validate_texture failed to alloc miptree"); - return GL_FALSE; - } - } - - /* Ensure all images are stored in the single main miptree */ - for(face = 0; face < t->mt->faces; ++face) { - for(level = t->mt->firstLevel; level <= t->mt->lastLevel; ++level) { - radeon_texture_image *image = get_radeon_texture_image(texObj->Image[face][level]); - if (RADEON_DEBUG & RADEON_TEXTURE) - fprintf(stderr, " face %i, level %i... %p vs %p ", face, level, t->mt, image->mt); - if (t->mt == image->mt || (!image->mt && !image->base.Data)) { - if (RADEON_DEBUG & RADEON_TEXTURE) - fprintf(stderr, "OK\n"); - - continue; - } - - if (RADEON_DEBUG & RADEON_TEXTURE) - fprintf(stderr, "migrating\n"); - migrate_image_to_miptree(t->mt, image, face, level); - } - } - - return GL_TRUE; -} - - /** * Need to map texture image into memory before copying image data, * then unmap it. diff --git a/src/mesa/drivers/dri/radeon/radeon_texture.h b/src/mesa/drivers/dri/radeon/radeon_texture.h index 8995546d77..906daf12d0 100644 --- a/src/mesa/drivers/dri/radeon/radeon_texture.h +++ b/src/mesa/drivers/dri/radeon/radeon_texture.h @@ -33,7 +33,8 @@ #include "main/formats.h" - +void copy_rows(void* dst, GLuint dststride, const void* src, GLuint srcstride, + GLuint numrows, GLuint rowsize); struct gl_texture_image *radeonNewTextureImage(GLcontext *ctx); void radeonFreeTexImageData(GLcontext *ctx, struct gl_texture_image *timage); diff --git a/src/mesa/drivers/dri/savage/savagerender.c b/src/mesa/drivers/dri/savage/savagerender.c index 32c74f9467..8221edf387 100644 --- a/src/mesa/drivers/dri/savage/savagerender.c +++ b/src/mesa/drivers/dri/savage/savagerender.c @@ -252,13 +252,13 @@ static GLboolean run_texnorm_stage( GLcontext *ctx, const GLboolean normalizeS = (texObj->WrapS == GL_REPEAT); const GLboolean normalizeT = (reallyEnabled & TEXTURE_2D_BIT) && (texObj->WrapT == GL_REPEAT); - const GLfloat *in = (GLfloat *)VB->TexCoordPtr[i]->data; - const GLint instride = VB->TexCoordPtr[i]->stride; + const GLfloat *in = (GLfloat *)VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->data; + const GLint instride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->stride; GLfloat (*out)[4] = store->texcoord[i].data; GLint j; if (!ctx->Texture.Unit[i]._ReallyEnabled || - VB->TexCoordPtr[i]->size == 4) + VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->size == 4) /* Never try to normalize homogenous tex coords! */ continue; @@ -297,7 +297,7 @@ static GLboolean run_texnorm_stage( GLcontext *ctx, } if (normalizeS || normalizeT) - VB->AttribPtr[VERT_ATTRIB_TEX0+i] = VB->TexCoordPtr[i] = &store->texcoord[i]; + VB->AttribPtr[_TNL_ATTRIB_TEX0 + i] = &store->texcoord[i]; } } diff --git a/src/mesa/drivers/dri/savage/savagetris.c b/src/mesa/drivers/dri/savage/savagetris.c index c04763b40e..e9529d1939 100644 --- a/src/mesa/drivers/dri/savage/savagetris.c +++ b/src/mesa/drivers/dri/savage/savagetris.c @@ -879,13 +879,13 @@ static GLboolean savageCheckPTexHack( GLcontext *ctx ) RENDERINPUTS_COPY( index_bitset, tnl->render_inputs_bitset ); - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 ) && VB->TexCoordPtr[0]->size == 4) { + if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 ) && VB->AttribPtr[_TNL_ATTRIB_TEX0]->size == 4) { if (!RENDERINPUTS_TEST_RANGE( index_bitset, _TNL_ATTRIB_TEX1, _TNL_LAST_TEX )) return GL_TRUE; /* apply ptex hack */ else FALLBACK(ctx, SAVAGE_FALLBACK_PROJ_TEXTURE, GL_TRUE); } - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX1 ) && VB->TexCoordPtr[1]->size == 4) + if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX1 ) && VB->AttribPtr[_TNL_ATTRIB_TEX1]->size == 4) FALLBACK(ctx, SAVAGE_FALLBACK_PROJ_TEXTURE, GL_TRUE); return GL_FALSE; /* don't apply ptex hack */ @@ -976,13 +976,13 @@ static INLINE GLuint savageChooseVertexFormat_s3d( GLcontext *ctx ) if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 )) { if (imesa->ptexHack) EMIT_ATTR( _TNL_ATTRIB_TEX0, EMIT_3F_XYW, SAVAGE_EMIT_STQ0, SAVAGE_SKIP_ST0); - else if (VB->TexCoordPtr[0]->size == 4) + else if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size == 4) assert (0); /* should be caught by savageCheckPTexHack */ - else if (VB->TexCoordPtr[0]->size >= 2) + else if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size >= 2) /* The chromium menu emits some 3D tex coords even though no * 3D texture is enabled. Ignore the 3rd coordinate. */ EMIT_ATTR( _TNL_ATTRIB_TEX0, EMIT_2F, SAVAGE_EMIT_ST0, SAVAGE_SKIP_ST0 ); - else if (VB->TexCoordPtr[0]->size == 1) { + else if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size == 1) { EMIT_ATTR( _TNL_ATTRIB_TEX0, EMIT_1F, SAVAGE_EMIT_S0, SAVAGE_SKIP_S0 ); EMIT_PAD( 4 ); } else @@ -1025,9 +1025,9 @@ static INLINE GLuint savageChooseVertexFormat_s4( GLcontext *ctx ) if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 )) { if (imesa->ptexHack) NEED_ATTR( SAVAGE_EMIT_STQ0, SAVAGE_SKIP_ST0); - else if (VB->TexCoordPtr[0]->size == 4) + else if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size == 4) assert (0); /* should be caught by savageCheckPTexHack */ - else if (VB->TexCoordPtr[0]->size >= 2) + else if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size >= 2) /* The chromium menu emits some 3D tex coords even though no * 3D texture is enabled. Ignore the 3rd coordinate. */ NEED_ATTR( SAVAGE_EMIT_ST0, SAVAGE_SKIP_ST0 ); @@ -1035,10 +1035,10 @@ static INLINE GLuint savageChooseVertexFormat_s4( GLcontext *ctx ) NEED_ATTR( SAVAGE_EMIT_S0, SAVAGE_SKIP_S0 ); } if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX1 )) { - if (VB->TexCoordPtr[1]->size == 4) + if (VB->AttribPtr[_TNL_ATTRIB_TEX1]->size == 4) /* projective textures are not supported by the hardware */ assert (0); /* should be caught by savageCheckPTexHack */ - else if (VB->TexCoordPtr[1]->size >= 2) + else if (VB->AttribPtr[_TNL_ATTRIB_TEX1]->size >= 2) NEED_ATTR( SAVAGE_EMIT_ST1, SAVAGE_SKIP_ST1 ); else NEED_ATTR( SAVAGE_EMIT_S1, SAVAGE_SKIP_S1 ); diff --git a/src/mesa/drivers/dri/sis/sis_tris.c b/src/mesa/drivers/dri/sis/sis_tris.c index 76d12d07b3..3cf10007b5 100644 --- a/src/mesa/drivers/dri/sis/sis_tris.c +++ b/src/mesa/drivers/dri/sis/sis_tris.c @@ -903,14 +903,14 @@ static void sisRenderStart( GLcontext *ctx ) /* projective textures are not supported by the hardware */ if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 )) { - if (VB->TexCoordPtr[0]->size > 2) + if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size > 2) tex_fallback = GL_TRUE; EMIT_ATTR(_TNL_ATTRIB_TEX0, EMIT_2F); AGPParseSet |= SiS_PS_HAS_UV0; } /* Will only hit tex1 on SiS300 */ if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX1 )) { - if (VB->TexCoordPtr[1]->size > 2) + if (VB->AttribPtr[_TNL_ATTRIB_TEX1]->size > 2) tex_fallback = GL_TRUE; EMIT_ATTR(_TNL_ATTRIB_TEX1, EMIT_2F); AGPParseSet |= SiS_PS_HAS_UV1; diff --git a/src/mesa/drivers/dri/tdfx/tdfx_vb.c b/src/mesa/drivers/dri/tdfx/tdfx_vb.c index 4928802232..c200ba3255 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_vb.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_vb.c @@ -69,11 +69,11 @@ static void interp_extras( GLcontext *ctx, /*fprintf(stderr, "%s\n", __FUNCTION__);*/ - if (VB->ColorPtr[1]) { + if (VB->BackfaceColorPtr) { INTERP_4F( t, - GET_COLOR(VB->ColorPtr[1], dst), - GET_COLOR(VB->ColorPtr[1], out), - GET_COLOR(VB->ColorPtr[1], in) ); + GET_COLOR(VB->BackfaceColorPtr, dst), + GET_COLOR(VB->BackfaceColorPtr, out), + GET_COLOR(VB->BackfaceColorPtr, in) ); } if (VB->EdgeFlag) { @@ -88,9 +88,9 @@ static void copy_pv_extras( GLcontext *ctx, GLuint dst, GLuint src ) { struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; - if (VB->ColorPtr[1]) { - COPY_4FV( GET_COLOR(VB->ColorPtr[1], dst), - GET_COLOR(VB->ColorPtr[1], src) ); + if (VB->BackfaceColorPtr) { + COPY_4FV( GET_COLOR(VB->BackfaceColorPtr, dst), + GET_COLOR(VB->BackfaceColorPtr, src) ); } setup_tab[TDFX_CONTEXT(ctx)->SetupIndex].copy_pv(ctx, dst, src); diff --git a/src/mesa/drivers/dri/tdfx/tdfx_vbtmp.h b/src/mesa/drivers/dri/tdfx/tdfx_vbtmp.h index 9b780761f4..19baf7d0d2 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_vbtmp.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_vbtmp.h @@ -58,32 +58,32 @@ static void TAG(emit)( GLcontext *ctx, /* fprintf(stderr, "%s\n", __FUNCTION__); */ if (IND & TDFX_TEX0_BIT) { - tc0_stride = VB->TexCoordPtr[tmu0_source]->stride; - tc0 = VB->TexCoordPtr[tmu0_source]->data; + tc0_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu0_source]->stride; + tc0 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu0_source]->data; u0scale = fxMesa->sScale0; v0scale = fxMesa->tScale0; if (IND & TDFX_PTEX_BIT) - tc0_size = VB->TexCoordPtr[tmu0_source]->size; + tc0_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu0_source]->size; } if (IND & TDFX_TEX1_BIT) { - tc1 = VB->TexCoordPtr[tmu1_source]->data; - tc1_stride = VB->TexCoordPtr[tmu1_source]->stride; + tc1 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu1_source]->data; + tc1_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu1_source]->stride; u1scale = fxMesa->sScale1; v1scale = fxMesa->tScale1; if (IND & TDFX_PTEX_BIT) - tc1_size = VB->TexCoordPtr[tmu1_source]->size; + tc1_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu1_source]->size; } if (IND & TDFX_RGBA_BIT) { - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; - col_size = VB->ColorPtr[0]->size; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; + col_size = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->size; } if (IND & TDFX_FOGC_BIT) { - fog = VB->FogCoordPtr->data; - fog_stride = VB->FogCoordPtr->stride; + fog = VB->AttribPtr[_TNL_ATTRIB_FOG]->data; + fog_stride = VB->AttribPtr[_TNL_ATTRIB_FOG]->stride; } { @@ -168,14 +168,14 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; if (IND & TDFX_TEX1_BIT) { - if (VB->TexCoordPtr[0] == 0) - VB->TexCoordPtr[0] = VB->TexCoordPtr[1]; + if (VB->AttribPtr[_TNL_ATTRIB_TEX0] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0] = VB->AttribPtr[_TNL_ATTRIB_TEX1]; - if (VB->TexCoordPtr[1]->size == 4) + if (VB->AttribPtr[_TNL_ATTRIB_TEX1]->size == 4) return GL_FALSE; } - if (VB->TexCoordPtr[0]->size == 4) + if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size == 4) return GL_FALSE; } diff --git a/src/mesa/drivers/dri/unichrome/via_tris.c b/src/mesa/drivers/dri/unichrome/via_tris.c index 79e67620c9..ab457d41dc 100644 --- a/src/mesa/drivers/dri/unichrome/via_tris.c +++ b/src/mesa/drivers/dri/unichrome/via_tris.c @@ -832,13 +832,13 @@ static GLboolean viaCheckPTexHack( GLcontext *ctx ) RENDERINPUTS_COPY( index_bitset, tnl->render_inputs_bitset ); - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 ) && VB->TexCoordPtr[0]->size == 4) { + if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 ) && VB->AttribPtr[_TNL_ATTRIB_TEX0]->size == 4) { if (!RENDERINPUTS_TEST_RANGE( index_bitset, _TNL_ATTRIB_TEX1, _TNL_LAST_TEX )) ptexHack = GL_TRUE; else fallback = GL_TRUE; } - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX1 ) && VB->TexCoordPtr[1]->size == 4) + if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX1 ) && VB->AttribPtr[_TNL_ATTRIB_TEX1]->size == 4) fallback = GL_TRUE; FALLBACK(VIA_CONTEXT(ctx), VIA_FALLBACK_PROJ_TEXTURE, fallback); diff --git a/src/mesa/drivers/glide/fxvb.c b/src/mesa/drivers/glide/fxvb.c index 1dc5f9891a..cc9ad0e8b8 100644 --- a/src/mesa/drivers/glide/fxvb.c +++ b/src/mesa/drivers/glide/fxvb.c @@ -104,24 +104,24 @@ static void interp_extras( GLcontext *ctx, { struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; - if (VB->ColorPtr[1]) { - /* If stride is zero, ColorPtr[1] is constant across the VB, so + if (VB->BackfaceColorPtr) { + /* If stride is zero, BackfaceColorPtr is constant across the VB, so * there is no point interpolating between two values as they will * be identical. This case is handled in t_dd_tritmp.h */ - if (VB->ColorPtr[1]->stride) { - assert(VB->ColorPtr[1]->stride == 4 * sizeof(GLfloat)); + if (VB->BackfaceColorPtr->stride) { + assert(VB->BackfaceColorPtr->stride == 4 * sizeof(GLfloat)); INTERP_4F( t, - GET_COLOR(VB->ColorPtr[1], dst), - GET_COLOR(VB->ColorPtr[1], out), - GET_COLOR(VB->ColorPtr[1], in) ); + GET_COLOR(VB->BackfaceColorPtr, dst), + GET_COLOR(VB->BackfaceColorPtr, out), + GET_COLOR(VB->BackfaceColorPtr, in) ); } - if (VB->SecondaryColorPtr[1]) { + if (VB->BackfaceSecondaryColorPtr) { INTERP_3F( t, - GET_COLOR(VB->SecondaryColorPtr[1], dst), - GET_COLOR(VB->SecondaryColorPtr[1], out), - GET_COLOR(VB->SecondaryColorPtr[1], in) ); + GET_COLOR(VB->BackfaceSecondaryColorPtr, dst), + GET_COLOR(VB->BackfaceSecondaryColorPtr, out), + GET_COLOR(VB->BackfaceSecondaryColorPtr, in) ); } } @@ -137,13 +137,13 @@ static void copy_pv_extras( GLcontext *ctx, GLuint dst, GLuint src ) { struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; - if (VB->ColorPtr[1]) { - COPY_4FV( GET_COLOR(VB->ColorPtr[1], dst), - GET_COLOR(VB->ColorPtr[1], src) ); + if (VB->BackfaceColorPtr) { + COPY_4FV( GET_COLOR(VB->BackfaceColorPtr, dst), + GET_COLOR(VB->BackfaceColorPtr, src) ); - if (VB->SecondaryColorPtr[1]) { - COPY_3FV( GET_COLOR(VB->SecondaryColorPtr[1], dst), - GET_COLOR(VB->SecondaryColorPtr[1], src) ); + if (VB->BackfaceSecondaryColorPtr) { + COPY_3FV( GET_COLOR(VB->BackfaceSecondaryColorPtr, dst), + GET_COLOR(VB->BackfaceSecondaryColorPtr, src) ); } } diff --git a/src/mesa/drivers/glide/fxvbtmp.h b/src/mesa/drivers/glide/fxvbtmp.h index f7970c78e2..f7893c1573 100644 --- a/src/mesa/drivers/glide/fxvbtmp.h +++ b/src/mesa/drivers/glide/fxvbtmp.h @@ -62,37 +62,37 @@ static void TAG(emit)( GLcontext *ctx, } if (IND & SETUP_TMU0) { - tc0 = VB->TexCoordPtr[tmu0_source]->data; - tc0_stride = VB->TexCoordPtr[tmu0_source]->stride; + tc0 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu0_source]->data; + tc0_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu0_source]->stride; u0scale = fxMesa->s0scale; v0scale = fxMesa->t0scale; if (IND & SETUP_PTEX) - tc0_size = VB->TexCoordPtr[tmu0_source]->size; + tc0_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu0_source]->size; } if (IND & SETUP_TMU1) { - tc1 = VB->TexCoordPtr[tmu1_source]->data; - tc1_stride = VB->TexCoordPtr[tmu1_source]->stride; + tc1 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu1_source]->data; + tc1_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu1_source]->stride; u1scale = fxMesa->s1scale; /* wrong if tmu1_source == 0, possible? */ v1scale = fxMesa->t1scale; if (IND & SETUP_PTEX) - tc1_size = VB->TexCoordPtr[tmu1_source]->size; + tc1_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu1_source]->size; } if (IND & SETUP_RGBA) { - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; - col_size = VB->ColorPtr[0]->size; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; + col_size = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->size; } if (IND & SETUP_SPEC) { - spec = VB->SecondaryColorPtr[0]->data; - spec_stride = VB->SecondaryColorPtr[0]->stride; + spec = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->data; + spec_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->stride; } if (IND & SETUP_FOGC) { - fog = VB->FogCoordPtr->data; - fog_stride = VB->FogCoordPtr->stride; + fog = VB->AttribPtr[_TNL_ATTRIB_FOG]->data; + fog_stride = VB->AttribPtr[_TNL_ATTRIB_FOG]->stride; } if (start) { @@ -220,14 +220,15 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; if (IND & SETUP_TMU1) { - if (VB->TexCoordPtr[0] == 0) - VB->TexCoordPtr[0] = VB->TexCoordPtr[1]; + if (VB->AttribPtr[_TNL_ATTRIB_TEX0] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0] = VB->AttribPtr[_TNL_ATTRIB_TEX1]; - if (VB->TexCoordPtr[1]->size == 4) + if (VB->AttribPtr[_TNL_ATTRIB_TEX1]->size == 4) return GL_FALSE; } - if (VB->TexCoordPtr[0] && VB->TexCoordPtr[0]->size == 4) + if (VB->AttribPtr[_TNL_ATTRIB_TEX0] && + VB->AttribPtr[_TNL_ATTRIB_TEX0]->size == 4) return GL_FALSE; } diff --git a/src/mesa/drivers/windows/gldirect/dx7/gld_primitive_dx7.c b/src/mesa/drivers/windows/gldirect/dx7/gld_primitive_dx7.c index c99ba0bba5..0b373814fe 100644 --- a/src/mesa/drivers/windows/gldirect/dx7/gld_primitive_dx7.c +++ b/src/mesa/drivers/windows/gldirect/dx7/gld_primitive_dx7.c @@ -189,9 +189,9 @@ GLfloat ex,ey,fx,fy,cc; \ /* Get vars for later */ \ VB = &TNL_CONTEXT(ctx)->vb; \ - vbcolor = (GLchan (*)[4])VB->ColorPtr[1]->data; \ - if (VB->SecondaryColorPtr[1]) { \ - vbspec = (GLchan (*)[4])VB->SecondaryColorPtr[1]->data; \ + vbcolor = (GLchan (*)[4])VB->BackfaceColorPtr->data; \ + if (VB->BackfaceSecondaryColorPtr) { \ + vbspec = (GLchan (*)[4])VB->BackfaceSecondaryColorPtr->data; \ } else { \ vbspec = NULL; \ } \ @@ -241,33 +241,33 @@ DWORD dwColor; #define GLD_SETUP_3D_VERTEX(v) \ - p4f = VB->ObjPtr->data; \ + p4f = VB->AttribPtr[_TNL_ATTRIB_POS]->data; \ pV->Position.x = p4f[##v][0]; \ pV->Position.y = p4f[##v][1]; \ pV->Position.z = p4f[##v][2]; #define GLD_SETUP_SMOOTH_COLOUR_3D(v) \ - p4f = (GLfloat (*)[4])VB->ColorPtr[0]->data; \ + p4f = (GLfloat (*)[4])VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; \ pV->Diffuse = D3DCOLOR_COLORVALUE(p4f[##v][0], p4f[##v][1], p4f[##v][2], p4f[##v][3]); #define GLD_SETUP_GET_FLAT_COLOUR_3D(v) \ - p4f = (GLfloat (*)[4])VB->ColorPtr[0]->data; \ + p4f = (GLfloat (*)[4])VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; \ dwColor = D3DCOLOR_COLORVALUE(p4f[##v][0], p4f[##v][1], p4f[##v][2], p4f[##v][3]); #define GLD_SETUP_USE_FLAT_COLOUR_3D \ pV->Diffuse = dwColor; #define GLD_SETUP_TEX0_3D(v) \ - if (VB->TexCoordPtr[0]) { \ - tc = VB->TexCoordPtr[0]->data; \ + if (VB->AttribPtr[_TNL_ATTRIB_TEX0]) { \ + tc = VB->AttribPtr[_TNL_ATTRIB_TEX0]->data; \ pV->TexUnit0.x = tc[##v][0]; \ pV->TexUnit0.y = tc[##v][1]; \ } #define GLD_SETUP_TEX1_3D(v) \ - if (VB->TexCoordPtr[1]) { \ - tc = VB->TexCoordPtr[1]->data; \ + if (VB->AttribPtr[_TNL_ATTRIB_TEX1]) { \ + tc = VB->AttribPtr[_TNL_ATTRIB_TEX1]->data; \ pV->TexUnit1.x = tc[##v][0]; \ pV->TexUnit1.y = tc[##v][1]; \ } diff --git a/src/mesa/drivers/windows/gldirect/dx7/gld_vb_d3d_render_dx7.c b/src/mesa/drivers/windows/gldirect/dx7/gld_vb_d3d_render_dx7.c index a85620dde8..c39775cad3 100644 --- a/src/mesa/drivers/windows/gldirect/dx7/gld_vb_d3d_render_dx7.c +++ b/src/mesa/drivers/windows/gldirect/dx7/gld_vb_d3d_render_dx7.c @@ -151,7 +151,7 @@ static GLboolean gld_d3d_render_stage_run( #if 0 // For debugging: Useful to see if an app passes colour data in // an unusual format. - switch (VB->ColorPtr[0]->Type) { + switch (VB->AttribPtr[_TNL_ATTRIB_COLOR0]->Type) { case GL_FLOAT: ddlogMessage(GLDLOG_SYSTEM, "ColorPtr: GL_FLOAT\n"); break; diff --git a/src/mesa/drivers/windows/gldirect/dx8/gld_primitive_dx8.c b/src/mesa/drivers/windows/gldirect/dx8/gld_primitive_dx8.c index a5b5462f03..990922580a 100644 --- a/src/mesa/drivers/windows/gldirect/dx8/gld_primitive_dx8.c +++ b/src/mesa/drivers/windows/gldirect/dx8/gld_primitive_dx8.c @@ -189,9 +189,9 @@ GLfloat ex,ey,fx,fy,cc; \ /* Get vars for later */ \ VB = &TNL_CONTEXT(ctx)->vb; \ - vbcolor = (GLchan (*)[4])VB->ColorPtr[1]->data; \ - if (VB->SecondaryColorPtr[1]) { \ - vbspec = (GLchan (*)[4])VB->SecondaryColorPtr[1]->data; \ + vbcolor = (GLchan (*)[4])VB->BackfaceColorPtr->data; \ + if (VB->BackfaceSecondaryColorPtr) { \ + vbspec = (GLchan (*)[4])VB->BackfaceSecondaryColorPtr->data; \ } else { \ vbspec = NULL; \ } \ @@ -241,33 +241,33 @@ DWORD dwColor; #define GLD_SETUP_3D_VERTEX(v) \ - p4f = VB->ObjPtr->data; \ + p4f = VB->AttribPtr[_TNL_ATTRIB_POS]->data; \ pV->Position.x = p4f[##v][0]; \ pV->Position.y = p4f[##v][1]; \ pV->Position.z = p4f[##v][2]; #define GLD_SETUP_SMOOTH_COLOUR_3D(v) \ - p4f = (GLfloat (*)[4])VB->ColorPtr[0]->data; \ + p4f = (GLfloat (*)[4])VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; \ pV->Diffuse = D3DCOLOR_COLORVALUE(p4f[##v][0], p4f[##v][1], p4f[##v][2], p4f[##v][3]); #define GLD_SETUP_GET_FLAT_COLOUR_3D(v) \ - p4f = (GLfloat (*)[4])VB->ColorPtr[0]->data; \ + p4f = (GLfloat (*)[4])VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; \ dwColor = D3DCOLOR_COLORVALUE(p4f[##v][0], p4f[##v][1], p4f[##v][2], p4f[##v][3]); #define GLD_SETUP_USE_FLAT_COLOUR_3D \ pV->Diffuse = dwColor; #define GLD_SETUP_TEX0_3D(v) \ - if (VB->TexCoordPtr[0]) { \ - tc = VB->TexCoordPtr[0]->data; \ + if (VB->AttribPtr[_TNL_ATTRIB_TEX0]) { \ + tc = VB->TnlAttribPtr[_TNL_ATTRIB_TEX0]->data; \ pV->TexUnit0.x = tc[##v][0]; \ pV->TexUnit0.y = tc[##v][1]; \ } #define GLD_SETUP_TEX1_3D(v) \ - if (VB->TexCoordPtr[1]) { \ - tc = VB->TexCoordPtr[1]->data; \ + if (VB->TnlAttribPtr[_TNL_ATTRIB_TEX1]) { \ + tc = VB->TnlAttribPtr[_TNL_ATTRIB_TEX1]->data; \ pV->TexUnit1.x = tc[##v][0]; \ pV->TexUnit1.y = tc[##v][1]; \ } diff --git a/src/mesa/drivers/windows/gldirect/dx8/gld_vb_d3d_render_dx8.c b/src/mesa/drivers/windows/gldirect/dx8/gld_vb_d3d_render_dx8.c index cafbf4f5c5..265c81fb4a 100644 --- a/src/mesa/drivers/windows/gldirect/dx8/gld_vb_d3d_render_dx8.c +++ b/src/mesa/drivers/windows/gldirect/dx8/gld_vb_d3d_render_dx8.c @@ -149,7 +149,7 @@ static GLboolean gld_d3d_render_stage_run( #if 0 // For debugging: Useful to see if an app passes colour data in // an unusual format. - switch (VB->ColorPtr[0]->Type) { + switch (VB->AttribPtr[_TNL_ATTRIB_COLOR0]->Type) { case GL_FLOAT: ddlogMessage(GLDLOG_SYSTEM, "ColorPtr: GL_FLOAT\n"); break; diff --git a/src/mesa/drivers/windows/gldirect/dx9/gld_primitive_dx9.c b/src/mesa/drivers/windows/gldirect/dx9/gld_primitive_dx9.c index 403a9d5f86..fd4dd4ed75 100644 --- a/src/mesa/drivers/windows/gldirect/dx9/gld_primitive_dx9.c +++ b/src/mesa/drivers/windows/gldirect/dx9/gld_primitive_dx9.c @@ -189,9 +189,9 @@ GLfloat ex,ey,fx,fy,cc; \ /* Get vars for later */ \ VB = &TNL_CONTEXT(ctx)->vb; \ - vbcolor = (GLchan (*)[4])VB->ColorPtr[1]->data; \ - if (VB->SecondaryColorPtr[1]) { \ - vbspec = (GLchan (*)[4])VB->SecondaryColorPtr[1]->data; \ + vbcolor = (GLchan (*)[4])VB->BackfaceColorPtr->data; \ + if (VB->BackfaceSecondaryColorPtr) { \ + vbspec = (GLchan (*)[4])VB->BackfaceSecondaryColorPtr->data; \ } else { \ vbspec = NULL; \ } \ @@ -241,33 +241,33 @@ DWORD dwColor; #define GLD_SETUP_3D_VERTEX(v) \ - p4f = VB->ObjPtr->data; \ + p4f = VB->AttribPtr[_TNL_ATTRIB_POS]->data; \ pV->Position.x = p4f[##v][0]; \ pV->Position.y = p4f[##v][1]; \ pV->Position.z = p4f[##v][2]; #define GLD_SETUP_SMOOTH_COLOUR_3D(v) \ - p4f = (GLfloat (*)[4])VB->ColorPtr[0]->data; \ + p4f = (GLfloat (*)[4])VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; \ pV->Diffuse = D3DCOLOR_COLORVALUE(p4f[##v][0], p4f[##v][1], p4f[##v][2], p4f[##v][3]); #define GLD_SETUP_GET_FLAT_COLOUR_3D(v) \ - p4f = (GLfloat (*)[4])VB->ColorPtr[0]->data; \ + p4f = (GLfloat (*)[4])VB->AttribPtr[_TNL_ATTRIB_COLOR00]->data; \ dwColor = D3DCOLOR_COLORVALUE(p4f[##v][0], p4f[##v][1], p4f[##v][2], p4f[##v][3]); #define GLD_SETUP_USE_FLAT_COLOUR_3D \ pV->Diffuse = dwColor; #define GLD_SETUP_TEX0_3D(v) \ - if (VB->TexCoordPtr[0]) { \ - tc = VB->TexCoordPtr[0]->data; \ + if (VB->AttribPtr[_TNL_ATTRIB_TEX0]) { \ + tc = VB->AttribPtr[_TNL_ATTRIB_TEX0]->data; \ pV->TexUnit0.x = tc[##v][0]; \ pV->TexUnit0.y = tc[##v][1]; \ } #define GLD_SETUP_TEX1_3D(v) \ - if (VB->TexCoordPtr[1]) { \ - tc = VB->TexCoordPtr[1]->data; \ + if (VB->AttribPtr[_TNL_ATTRIB_TEX1]) { \ + tc = VB->AttribPtr[_TNL_ATTRIB_TEX1]->data; \ pV->TexUnit1.x = tc[##v][0]; \ pV->TexUnit1.y = tc[##v][1]; \ } diff --git a/src/mesa/drivers/windows/gldirect/dx9/gld_vb_d3d_render_dx9.c b/src/mesa/drivers/windows/gldirect/dx9/gld_vb_d3d_render_dx9.c index 4fa6bcaf1a..91a68b3f2d 100644 --- a/src/mesa/drivers/windows/gldirect/dx9/gld_vb_d3d_render_dx9.c +++ b/src/mesa/drivers/windows/gldirect/dx9/gld_vb_d3d_render_dx9.c @@ -149,7 +149,7 @@ static GLboolean gld_d3d_render_stage_run( #if 0 // For debugging: Useful to see if an app passes colour data in // an unusual format. - switch (VB->ColorPtr[0]->Type) { + switch (VB->AttribPtr[_TNL_ATTRIB_COLOR0]->Type) { case GL_FLOAT: ddlogMessage(GLDLOG_SYSTEM, "ColorPtr: GL_FLOAT\n"); break; |