diff options
Diffstat (limited to 'src/mesa/drivers/dri/i965')
40 files changed, 746 insertions, 1396 deletions
diff --git a/src/mesa/drivers/dri/i965/Makefile b/src/mesa/drivers/dri/i965/Makefile index 37a470f2e2..2934414d99 100644 --- a/src/mesa/drivers/dri/i965/Makefile +++ b/src/mesa/drivers/dri/i965/Makefile @@ -9,9 +9,10 @@ DRIVER_SOURCES = \ intel_blit.c \ intel_buffer_objects.c \ intel_buffers.c \ + intel_clear.c \ intel_context.c \ intel_decode.c \ - intel_depthstencil.c \ + intel_extensions.c \ intel_fbo.c \ intel_mipmap_tree.c \ intel_regions.c \ @@ -22,6 +23,7 @@ DRIVER_SOURCES = \ intel_pixel_copy.c \ intel_pixel_draw.c \ intel_state.c \ + intel_swapbuffers.c \ intel_tex.c \ intel_tex_copy.c \ intel_tex_format.c \ @@ -49,7 +51,6 @@ DRIVER_SOURCES = \ brw_gs.c \ brw_gs_emit.c \ brw_gs_state.c \ - brw_metaops.c \ brw_misc_state.c \ brw_program.c \ brw_queryobj.c \ diff --git a/src/mesa/drivers/dri/i965/brw_cc.c b/src/mesa/drivers/dri/i965/brw_cc.c index fa8121e02d..82370162f5 100644 --- a/src/mesa/drivers/dri/i965/brw_cc.c +++ b/src/mesa/drivers/dri/i965/brw_cc.c @@ -83,59 +83,60 @@ struct brw_cc_unit_key { static void cc_unit_populate_key(struct brw_context *brw, struct brw_cc_unit_key *key) { - struct gl_stencil_attrib *stencil = brw->attribs.Stencil; + GLcontext *ctx = &brw->intel.ctx; + const unsigned back = ctx->Stencil._BackFace; memset(key, 0, sizeof(*key)); - key->stencil = stencil->Enabled; - key->stencil_two_side = stencil->_TestTwoSide; + key->stencil = ctx->Stencil.Enabled; + key->stencil_two_side = ctx->Stencil._TestTwoSide; if (key->stencil) { - key->stencil_func[0] = stencil->Function[0]; - key->stencil_fail_op[0] = stencil->FailFunc[0]; - key->stencil_pass_depth_fail_op[0] = stencil->ZFailFunc[0]; - key->stencil_pass_depth_pass_op[0] = stencil->ZPassFunc[0]; - key->stencil_ref[0] = stencil->Ref[0]; - key->stencil_write_mask[0] = stencil->WriteMask[0]; - key->stencil_test_mask[0] = stencil->ValueMask[0]; + key->stencil_func[0] = ctx->Stencil.Function[0]; + key->stencil_fail_op[0] = ctx->Stencil.FailFunc[0]; + key->stencil_pass_depth_fail_op[0] = ctx->Stencil.ZFailFunc[0]; + key->stencil_pass_depth_pass_op[0] = ctx->Stencil.ZPassFunc[0]; + key->stencil_ref[0] = ctx->Stencil.Ref[0]; + key->stencil_write_mask[0] = ctx->Stencil.WriteMask[0]; + key->stencil_test_mask[0] = ctx->Stencil.ValueMask[0]; } if (key->stencil_two_side) { - key->stencil_func[1] = stencil->Function[1]; - key->stencil_fail_op[1] = stencil->FailFunc[1]; - key->stencil_pass_depth_fail_op[1] = stencil->ZFailFunc[1]; - key->stencil_pass_depth_pass_op[1] = stencil->ZPassFunc[1]; - key->stencil_ref[1] = stencil->Ref[1]; - key->stencil_write_mask[1] = stencil->WriteMask[1]; - key->stencil_test_mask[1] = stencil->ValueMask[1]; + key->stencil_func[1] = ctx->Stencil.Function[back]; + key->stencil_fail_op[1] = ctx->Stencil.FailFunc[back]; + key->stencil_pass_depth_fail_op[1] = ctx->Stencil.ZFailFunc[back]; + key->stencil_pass_depth_pass_op[1] = ctx->Stencil.ZPassFunc[back]; + key->stencil_ref[1] = ctx->Stencil.Ref[back]; + key->stencil_write_mask[1] = ctx->Stencil.WriteMask[back]; + key->stencil_test_mask[1] = ctx->Stencil.ValueMask[back]; } - if (brw->attribs.Color->_LogicOpEnabled) - key->logic_op = brw->attribs.Color->LogicOp; + if (ctx->Color._LogicOpEnabled) + key->logic_op = ctx->Color.LogicOp; else key->logic_op = GL_COPY; - key->color_blend = brw->attribs.Color->BlendEnabled; + key->color_blend = ctx->Color.BlendEnabled; if (key->color_blend) { - key->blend_eq_rgb = brw->attribs.Color->BlendEquationRGB; - key->blend_eq_a = brw->attribs.Color->BlendEquationA; - key->blend_src_rgb = brw->attribs.Color->BlendSrcRGB; - key->blend_dst_rgb = brw->attribs.Color->BlendDstRGB; - key->blend_src_a = brw->attribs.Color->BlendSrcA; - key->blend_dst_a = brw->attribs.Color->BlendDstA; + key->blend_eq_rgb = ctx->Color.BlendEquationRGB; + key->blend_eq_a = ctx->Color.BlendEquationA; + key->blend_src_rgb = ctx->Color.BlendSrcRGB; + key->blend_dst_rgb = ctx->Color.BlendDstRGB; + key->blend_src_a = ctx->Color.BlendSrcA; + key->blend_dst_a = ctx->Color.BlendDstA; } - key->alpha_enabled = brw->attribs.Color->AlphaEnabled; + key->alpha_enabled = ctx->Color.AlphaEnabled; if (key->alpha_enabled) { - key->alpha_func = brw->attribs.Color->AlphaFunc; - key->alpha_ref = brw->attribs.Color->AlphaRef; + key->alpha_func = ctx->Color.AlphaFunc; + key->alpha_ref = ctx->Color.AlphaRef; } - key->dither = brw->attribs.Color->DitherFlag; + key->dither = ctx->Color.DitherFlag; - key->depth_test = brw->attribs.Depth->Test; + key->depth_test = ctx->Depth.Test; if (key->depth_test) { - key->depth_func = brw->attribs.Depth->Func; - key->depth_write = brw->attribs.Depth->Mask; + key->depth_func = ctx->Depth.Func; + key->depth_write = ctx->Depth.Mask; } } diff --git a/src/mesa/drivers/dri/i965/brw_clip.c b/src/mesa/drivers/dri/i965/brw_clip.c index 38d8b704d7..5cffcebde4 100644 --- a/src/mesa/drivers/dri/i965/brw_clip.c +++ b/src/mesa/drivers/dri/i965/brw_clip.c @@ -145,14 +145,14 @@ static void upload_clip_prog(struct brw_context *brw) /* CACHE_NEW_VS_PROG */ key.attrs = brw->vs.prog_data->outputs_written; /* _NEW_LIGHT */ - key.do_flat_shading = (brw->attribs.Light->ShadeModel == GL_FLAT); + key.do_flat_shading = (ctx->Light.ShadeModel == GL_FLAT); /* _NEW_TRANSFORM */ - key.nr_userclip = brw_count_bits(brw->attribs.Transform->ClipPlanesEnabled); + key.nr_userclip = brw_count_bits(ctx->Transform.ClipPlanesEnabled); key.clip_mode = BRW_CLIPMODE_NORMAL; /* _NEW_POLYGON */ if (key.primitive == GL_TRIANGLES) { - if (brw->attribs.Polygon->CullFaceMode == GL_FRONT_AND_BACK) + if (ctx->Polygon.CullFaceMode == GL_FRONT_AND_BACK) key.clip_mode = BRW_CLIPMODE_REJECT_ALL; else { GLuint fill_front = CLIP_CULL; @@ -160,44 +160,44 @@ static void upload_clip_prog(struct brw_context *brw) GLuint offset_front = 0; GLuint offset_back = 0; - if (!brw->attribs.Polygon->CullFlag || - brw->attribs.Polygon->CullFaceMode != GL_FRONT) { - switch (brw->attribs.Polygon->FrontMode) { + if (!ctx->Polygon.CullFlag || + ctx->Polygon.CullFaceMode != GL_FRONT) { + switch (ctx->Polygon.FrontMode) { case GL_FILL: fill_front = CLIP_FILL; offset_front = 0; break; case GL_LINE: fill_front = CLIP_LINE; - offset_front = brw->attribs.Polygon->OffsetLine; + offset_front = ctx->Polygon.OffsetLine; break; case GL_POINT: fill_front = CLIP_POINT; - offset_front = brw->attribs.Polygon->OffsetPoint; + offset_front = ctx->Polygon.OffsetPoint; break; } } - if (!brw->attribs.Polygon->CullFlag || - brw->attribs.Polygon->CullFaceMode != GL_BACK) { - switch (brw->attribs.Polygon->BackMode) { + if (!ctx->Polygon.CullFlag || + ctx->Polygon.CullFaceMode != GL_BACK) { + switch (ctx->Polygon.BackMode) { case GL_FILL: fill_back = CLIP_FILL; offset_back = 0; break; case GL_LINE: fill_back = CLIP_LINE; - offset_back = brw->attribs.Polygon->OffsetLine; + offset_back = ctx->Polygon.OffsetLine; break; case GL_POINT: fill_back = CLIP_POINT; - offset_back = brw->attribs.Polygon->OffsetPoint; + offset_back = ctx->Polygon.OffsetPoint; break; } } - if (brw->attribs.Polygon->BackMode != GL_FILL || - brw->attribs.Polygon->FrontMode != GL_FILL) { + if (ctx->Polygon.BackMode != GL_FILL || + ctx->Polygon.FrontMode != GL_FILL) { key.do_unfilled = 1; /* Most cases the fixed function units will handle. Cases where @@ -207,17 +207,17 @@ static void upload_clip_prog(struct brw_context *brw) if (offset_back || offset_front) { /* _NEW_POLYGON, _NEW_BUFFERS */ - key.offset_units = brw->attribs.Polygon->OffsetUnits * brw->intel.polygon_offset_scale; - key.offset_factor = brw->attribs.Polygon->OffsetFactor * ctx->DrawBuffer->_MRD; + key.offset_units = ctx->Polygon.OffsetUnits * brw->intel.polygon_offset_scale; + key.offset_factor = ctx->Polygon.OffsetFactor * ctx->DrawBuffer->_MRD; } - switch (brw->attribs.Polygon->FrontFace) { + switch (ctx->Polygon.FrontFace) { case GL_CCW: key.fill_ccw = fill_front; key.fill_cw = fill_back; key.offset_ccw = offset_front; key.offset_cw = offset_back; - if (brw->attribs.Light->Model.TwoSide && + if (ctx->Light.Model.TwoSide && key.fill_cw != CLIP_CULL) key.copy_bfc_cw = 1; break; @@ -226,7 +226,7 @@ static void upload_clip_prog(struct brw_context *brw) key.fill_ccw = fill_back; key.offset_cw = offset_front; key.offset_ccw = offset_back; - if (brw->attribs.Light->Model.TwoSide && + if (ctx->Light.Model.TwoSide && key.fill_ccw != CLIP_CULL) key.copy_bfc_ccw = 1; break; diff --git a/src/mesa/drivers/dri/i965/brw_context.c b/src/mesa/drivers/dri/i965/brw_context.c index d7a2bd95ee..d96ff29310 100644 --- a/src/mesa/drivers/dri/i965/brw_context.c +++ b/src/mesa/drivers/dri/i965/brw_context.c @@ -76,30 +76,6 @@ static void brwInitDriverFunctions( struct dd_function_table *functions ) functions->Viewport = intel_viewport; } - -static void brw_init_attribs( struct brw_context *brw ) -{ - GLcontext *ctx = &brw->intel.ctx; - - brw->attribs.Color = &ctx->Color; - brw->attribs.Depth = &ctx->Depth; - brw->attribs.Fog = &ctx->Fog; - brw->attribs.Hint = &ctx->Hint; - brw->attribs.Light = &ctx->Light; - brw->attribs.Line = &ctx->Line; - brw->attribs.Point = &ctx->Point; - brw->attribs.Polygon = &ctx->Polygon; - brw->attribs.Scissor = &ctx->Scissor; - brw->attribs.Stencil = &ctx->Stencil; - brw->attribs.Texture = &ctx->Texture; - brw->attribs.Transform = &ctx->Transform; - brw->attribs.Viewport = &ctx->Viewport; - brw->attribs.VertexProgram = &ctx->VertexProgram; - brw->attribs.FragmentProgram = &ctx->FragmentProgram; - brw->attribs.PolygonStipple = &ctx->PolygonStipple[0]; -} - - GLboolean brwCreateContext( const __GLcontextModes *mesaVis, __DRIcontextPrivate *driContextPriv, void *sharedContextPrivate) @@ -135,21 +111,21 @@ GLboolean brwCreateContext( const __GLcontextModes *mesaVis, ctx->Const.MaxTextureImageUnits); ctx->Const.MaxVertexTextureImageUnits = 0; /* no vertex shader textures */ - /* Advertise the full hardware capabilities. The new memory - * manager should cope much better with overload situations: + /* Mesa limits textures to 4kx4k; it would be nice to fix that someday */ - ctx->Const.MaxTextureLevels = 12; + ctx->Const.MaxTextureLevels = 13; ctx->Const.Max3DTextureLevels = 9; ctx->Const.MaxCubeTextureLevels = 12; - ctx->Const.MaxTextureRectSize = (1<<11); + ctx->Const.MaxTextureRectSize = (1<<12); /* if conformance mode is set, swrast can handle any size AA point */ ctx->Const.MaxPointSizeAA = 255.0; + /* We want the GLSL compiler to emit code that uses condition codes */ + ctx->Shader.EmitCondCodes = GL_TRUE; + /* ctx->Const.MaxNativeVertexProgramTemps = 32; */ - brw_init_attribs( brw ); - brw_init_metaops( brw ); brw_init_state( brw ); brw->state.dirty.mesa = ~0; diff --git a/src/mesa/drivers/dri/i965/brw_context.h b/src/mesa/drivers/dri/i965/brw_context.h index 5d3f99e025..df90c2027f 100644 --- a/src/mesa/drivers/dri/i965/brw_context.h +++ b/src/mesa/drivers/dri/i965/brw_context.h @@ -131,7 +131,6 @@ struct brw_context; #define BRW_NEW_WM_INPUT_DIMENSIONS 0x100 #define BRW_NEW_INPUT_VARYING 0x200 #define BRW_NEW_PSP 0x800 -#define BRW_NEW_METAOPS 0x1000 #define BRW_NEW_FENCE 0x2000 #define BRW_NEW_INDICES 0x4000 #define BRW_NEW_VERTICES 0x8000 @@ -303,26 +302,6 @@ struct brw_cache { }; - -struct brw_state_pointers { - struct gl_colorbuffer_attrib *Color; - struct gl_depthbuffer_attrib *Depth; - struct gl_fog_attrib *Fog; - struct gl_hint_attrib *Hint; - struct gl_light_attrib *Light; - struct gl_line_attrib *Line; - struct gl_point_attrib *Point; - struct gl_polygon_attrib *Polygon; - GLuint *PolygonStipple; - struct gl_scissor_attrib *Scissor; - struct gl_stencil_attrib *Stencil; - struct gl_texture_attrib *Texture; - struct gl_transform_attrib *Transform; - struct gl_viewport_attrib *Viewport; - struct gl_vertex_program_state *VertexProgram; - struct gl_fragment_program_state *FragmentProgram; -}; - /* Considered adding a member to this struct to document which flags * an update might raise so that ordering of the state atoms can be * checked or derived at runtime. Dropped the idea in favor of having @@ -457,7 +436,6 @@ struct brw_context int validated_bo_count; } state; - struct brw_state_pointers attribs; struct brw_cache cache; struct brw_cached_batch_item *cached_batch_items; @@ -491,28 +469,6 @@ struct brw_context unsigned int offset; } ib; - struct { - /* Will be allocated on demand if needed. - */ - struct brw_state_pointers attribs; - struct gl_vertex_program *vp; - struct gl_fragment_program *fp, *fp_tex; - - struct gl_buffer_object *vbo; - - struct intel_region *saved_draw_region; - GLuint saved_nr_draw_regions; - struct intel_region *saved_depth_region; - - GLuint restore_draw_buffers[MAX_DRAW_BUFFERS]; - GLuint restore_num_draw_buffers; - - struct gl_fragment_program *restore_fp; - - GLboolean active; - } metaops; - - /* Active vertex program: */ const struct gl_vertex_program *vertex_program; @@ -703,13 +659,6 @@ void brw_FrameBufferTexInit( struct brw_context *brw, void brw_FrameBufferTexDestroy( struct brw_context *brw ); void brw_validate_textures( struct brw_context *brw ); -/*====================================================================== - * brw_metaops.c - */ - -void brw_init_metaops( struct brw_context *brw ); -void brw_destroy_metaops( struct brw_context *brw ); - /*====================================================================== * brw_program.c diff --git a/src/mesa/drivers/dri/i965/brw_curbe.c b/src/mesa/drivers/dri/i965/brw_curbe.c index fbf473abf6..4eaaa5f871 100644 --- a/src/mesa/drivers/dri/i965/brw_curbe.c +++ b/src/mesa/drivers/dri/i965/brw_curbe.c @@ -48,6 +48,7 @@ */ static void calculate_curbe_offsets( struct brw_context *brw ) { + GLcontext *ctx = &brw->intel.ctx; /* CACHE_NEW_WM_PROG */ GLuint nr_fp_regs = (brw->wm.prog_data->nr_params + 15) / 16; @@ -58,8 +59,8 @@ static void calculate_curbe_offsets( struct brw_context *brw ) GLuint total_regs; /* _NEW_TRANSFORM */ - if (brw->attribs.Transform->ClipPlanesEnabled) { - GLuint nr_planes = 6 + brw_count_bits(brw->attribs.Transform->ClipPlanesEnabled); + if (ctx->Transform.ClipPlanesEnabled) { + GLuint nr_planes = 6 + brw_count_bits(ctx->Transform.ClipPlanesEnabled); nr_clip_regs = (nr_planes * 4 + 15) / 16; } @@ -233,11 +234,11 @@ static void prepare_constant_buffer(struct brw_context *brw) */ assert(MAX_CLIP_PLANES == 6); for (j = 0; j < MAX_CLIP_PLANES; j++) { - if (brw->attribs.Transform->ClipPlanesEnabled & (1<<j)) { - buf[offset + i * 4 + 0] = brw->attribs.Transform->_ClipUserPlane[j][0]; - buf[offset + i * 4 + 1] = brw->attribs.Transform->_ClipUserPlane[j][1]; - buf[offset + i * 4 + 2] = brw->attribs.Transform->_ClipUserPlane[j][2]; - buf[offset + i * 4 + 3] = brw->attribs.Transform->_ClipUserPlane[j][3]; + if (ctx->Transform.ClipPlanesEnabled & (1<<j)) { + buf[offset + i * 4 + 0] = ctx->Transform._ClipUserPlane[j][0]; + buf[offset + i * 4 + 1] = ctx->Transform._ClipUserPlane[j][1]; + buf[offset + i * 4 + 2] = ctx->Transform._ClipUserPlane[j][2]; + buf[offset + i * 4 + 3] = ctx->Transform._ClipUserPlane[j][3]; i++; } } diff --git a/src/mesa/drivers/dri/i965/brw_draw.c b/src/mesa/drivers/dri/i965/brw_draw.c index 785fb784ca..99fd587e9f 100644 --- a/src/mesa/drivers/dri/i965/brw_draw.c +++ b/src/mesa/drivers/dri/i965/brw_draw.c @@ -84,15 +84,17 @@ static const GLenum reduced_prim[GL_POLYGON+1] = { */ static GLuint brw_set_prim(struct brw_context *brw, GLenum prim) { + GLcontext *ctx = &brw->intel.ctx; + if (INTEL_DEBUG & DEBUG_PRIMS) _mesa_printf("PRIM: %s\n", _mesa_lookup_enum_by_nr(prim)); /* Slight optimization to avoid the GS program when not needed: */ if (prim == GL_QUAD_STRIP && - brw->attribs.Light->ShadeModel != GL_FLAT && - brw->attribs.Polygon->FrontMode == GL_FILL && - brw->attribs.Polygon->BackMode == GL_FILL) + ctx->Light.ShadeModel != GL_FLAT && + ctx->Polygon.FrontMode == GL_FILL && + ctx->Polygon.BackMode == GL_FILL) prim = GL_TRIANGLE_STRIP; if (prim != brw->primitive) { @@ -166,14 +168,11 @@ static void brw_merge_inputs( struct brw_context *brw, for (i = 0; i < VERT_ATTRIB_MAX; i++) { brw->vb.inputs[i].glarray = arrays[i]; - /* XXX: metaops passes null arrays */ - if (arrays[i]) { - if (arrays[i]->StrideB != 0) - brw->vb.info.varying |= 1 << i; + if (arrays[i]->StrideB != 0) + brw->vb.info.varying |= 1 << i; brw->vb.info.sizes[i/16] |= (brw->vb.inputs[i].glarray->Size - 1) << ((i%16) * 2); - } } /* Raise statechanges if input sizes and varying have changed: @@ -192,12 +191,13 @@ static GLboolean check_fallbacks( struct brw_context *brw, const struct _mesa_prim *prim, GLuint nr_prims ) { + GLcontext *ctx = &brw->intel.ctx; GLuint i; if (!brw->intel.strict_conformance) return GL_FALSE; - if (brw->attribs.Polygon->SmoothFlag) { + if (ctx->Polygon.SmoothFlag) { for (i = 0; i < nr_prims; i++) if (reduced_prim[prim[i].mode] == GL_TRIANGLES) return GL_TRUE; @@ -206,7 +206,7 @@ static GLboolean check_fallbacks( struct brw_context *brw, /* BRW hardware will do AA lines, but they are non-conformant it * seems. TBD whether we keep this fallback: */ - if (brw->attribs.Line->SmoothFlag) { + if (ctx->Line.SmoothFlag) { for (i = 0; i < nr_prims; i++) if (reduced_prim[prim[i].mode] == GL_LINES) return GL_TRUE; @@ -215,7 +215,7 @@ static GLboolean check_fallbacks( struct brw_context *brw, /* Stipple -- these fallbacks could be resolved with a little * bit of work? */ - if (brw->attribs.Line->StippleFlag) { + if (ctx->Line.StippleFlag) { for (i = 0; i < nr_prims; i++) { /* GS doesn't get enough information to know when to reset * the stipple counter?!? @@ -224,14 +224,14 @@ static GLboolean check_fallbacks( struct brw_context *brw, return GL_TRUE; if (prim[i].mode == GL_POLYGON && - (brw->attribs.Polygon->FrontMode == GL_LINE || - brw->attribs.Polygon->BackMode == GL_LINE)) + (ctx->Polygon.FrontMode == GL_LINE || + ctx->Polygon.BackMode == GL_LINE)) return GL_TRUE; } } - if (brw->attribs.Point->SmoothFlag) { + if (ctx->Point.SmoothFlag) { for (i = 0; i < nr_prims; i++) if (prim[i].mode == GL_POINTS) return GL_TRUE; diff --git a/src/mesa/drivers/dri/i965/brw_draw_upload.c b/src/mesa/drivers/dri/i965/brw_draw_upload.c index 73d6dea01e..02998d5957 100644 --- a/src/mesa/drivers/dri/i965/brw_draw_upload.c +++ b/src/mesa/drivers/dri/i965/brw_draw_upload.c @@ -156,7 +156,13 @@ static GLuint byte_types_scale[5] = { }; -static GLuint get_surface_type( GLenum type, GLuint size, GLboolean normalized ) +/** + * Given vertex array type/size/format/normalized info, return + * the appopriate hardware surface type. + * Format will be GL_RGBA or possibly GL_BGRA for GLubyte[4] color arrays. + */ +static GLuint get_surface_type( GLenum type, GLuint size, + GLenum format, GLboolean normalized ) { if (INTEL_DEBUG & DEBUG_VERTS) _mesa_printf("type %s size %d normalized %d\n", @@ -171,11 +177,20 @@ static GLuint get_surface_type( GLenum type, GLuint size, GLboolean normalized ) case GL_BYTE: return byte_types_norm[size]; case GL_UNSIGNED_INT: return uint_types_norm[size]; case GL_UNSIGNED_SHORT: return ushort_types_norm[size]; - case GL_UNSIGNED_BYTE: return ubyte_types_norm[size]; + case GL_UNSIGNED_BYTE: + if (format == GL_BGRA) { + /* See GL_EXT_vertex_array_bgra */ + assert(size == 4); + return BRW_SURFACEFORMAT_B8G8R8A8_UNORM; + } + else { + return ubyte_types_norm[size]; + } default: assert(0); return 0; } } else { + assert(format == GL_RGBA); /* sanity check */ switch (type) { case GL_DOUBLE: return double_types[size]; case GL_FLOAT: return float_types[size]; @@ -281,7 +296,7 @@ copy_array_to_vbo_array( struct brw_context *brw, } else { void *data; char *dest; - const char *src = element->glarray->Ptr; + const unsigned char *src = element->glarray->Ptr; int i; data = _mesa_malloc(dst_stride * element->count); @@ -484,6 +499,7 @@ static void brw_emit_vertices(struct brw_context *brw) struct brw_vertex_element *input = enabled[i]; uint32_t format = get_surface_type(input->glarray->Type, input->glarray->Size, + input->glarray->Format, input->glarray->Normalized); uint32_t comp0 = BRW_VE1_COMPONENT_STORE_SRC; uint32_t comp1 = BRW_VE1_COMPONENT_STORE_SRC; @@ -549,7 +565,7 @@ static void brw_prepare_indices(struct brw_context *brw) */ dri_bo_subdata(bo, offset, ib_size, index_buffer->ptr); } else { - offset = (GLuint)index_buffer->ptr; + offset = (GLuint) (unsigned long) index_buffer->ptr; /* If the index buffer isn't aligned to its element size, we have to * rebase it into a temporary. diff --git a/src/mesa/drivers/dri/i965/brw_eu.c b/src/mesa/drivers/dri/i965/brw_eu.c index b3ae4eef33..c53efba599 100644 --- a/src/mesa/drivers/dri/i965/brw_eu.c +++ b/src/mesa/drivers/dri/i965/brw_eu.c @@ -129,3 +129,126 @@ const GLuint *brw_get_program( struct brw_compile *p, return (const GLuint *)p->store; } + + +/** + * Subroutine calls require special attention. + * Mesa instructions may be expanded into multiple hardware instructions + * so the prog_instruction::BranchTarget field can't be used as an index + * into the hardware instructions. + * + * The BranchTarget field isn't needed, however. Mesa's GLSL compiler + * emits CAL and BGNSUB instructions with labels that can be used to map + * subroutine calls to actual subroutine code blocks. + * + * The structures and function here implement patching of CAL instructions + * so they jump to the right subroutine code... + */ + + +/** + * For each OPCODE_BGNSUB we create one of these. + */ +struct brw_glsl_label +{ + const char *name; /**< the label string */ + GLuint position; /**< the position of the brw instruction for this label */ + struct brw_glsl_label *next; /**< next in linked list */ +}; + + +/** + * For each OPCODE_CAL we create one of these. + */ +struct brw_glsl_call +{ + GLuint call_inst_pos; /**< location of the CAL instruction */ + const char *sub_name; /**< name of subroutine to call */ + struct brw_glsl_call *next; /**< next in linked list */ +}; + + +/** + * Called for each OPCODE_BGNSUB. + */ +void +brw_save_label(struct brw_compile *c, const char *name, GLuint position) +{ + struct brw_glsl_label *label = CALLOC_STRUCT(brw_glsl_label); + label->name = name; + label->position = position; + label->next = c->first_label; + c->first_label = label; +} + + +/** + * Called for each OPCODE_CAL. + */ +void +brw_save_call(struct brw_compile *c, const char *name, GLuint call_pos) +{ + struct brw_glsl_call *call = CALLOC_STRUCT(brw_glsl_call); + call->call_inst_pos = call_pos; + call->sub_name = name; + call->next = c->first_call; + c->first_call = call; +} + + +/** + * Lookup a label, return label's position/offset. + */ +static GLuint +brw_lookup_label(struct brw_compile *c, const char *name) +{ + const struct brw_glsl_label *label; + for (label = c->first_label; label; label = label->next) { + if (strcmp(name, label->name) == 0) { + return label->position; + } + } + abort(); /* should never happen */ + return ~0; +} + + +/** + * When we're done generating code, this function is called to resolve + * subroutine calls. + */ +void +brw_resolve_cals(struct brw_compile *c) +{ + const struct brw_glsl_call *call; + + for (call = c->first_call; call; call = call->next) { + const GLuint sub_loc = brw_lookup_label(c, call->sub_name); + struct brw_instruction *brw_call_inst = &c->store[call->call_inst_pos]; + struct brw_instruction *brw_sub_inst = &c->store[sub_loc]; + GLint offset = brw_sub_inst - brw_call_inst; + + /* patch brw_inst1 to point to brw_inst2 */ + brw_set_src1(brw_call_inst, brw_imm_d(offset * 16)); + } + + /* free linked list of calls */ + { + struct brw_glsl_call *call, *next; + for (call = c->first_call; call; call = next) { + next = call->next; + _mesa_free(call); + } + c->first_call = NULL; + } + + /* free linked list of labels */ + { + struct brw_glsl_label *label, *next; + for (label = c->first_label; label; label = next) { + next = label->next; + _mesa_free(label); + } + c->first_label = NULL; + } +} diff --git a/src/mesa/drivers/dri/i965/brw_eu.h b/src/mesa/drivers/dri/i965/brw_eu.h index 9e2b39af9b..b36a1976ef 100644 --- a/src/mesa/drivers/dri/i965/brw_eu.h +++ b/src/mesa/drivers/dri/i965/brw_eu.h @@ -91,6 +91,11 @@ struct brw_indirect { }; +struct brw_glsl_label; +struct brw_glsl_call; + + + #define BRW_EU_MAX_INSN_STACK 5 #define BRW_EU_MAX_INSN 1200 @@ -106,9 +111,22 @@ struct brw_compile { GLuint flag_value; GLboolean single_program_flow; struct brw_context *brw; + + struct brw_glsl_label *first_label; /**< linked list of labels */ + struct brw_glsl_call *first_call; /**< linked list of CALs */ }; +void +brw_save_label(struct brw_compile *c, const char *name, GLuint position); + +void +brw_save_call(struct brw_compile *c, const char *name, GLuint call_pos); + +void +brw_resolve_cals(struct brw_compile *c); + + static INLINE int type_sz( GLuint type ) { diff --git a/src/mesa/drivers/dri/i965/brw_eu_emit.c b/src/mesa/drivers/dri/i965/brw_eu_emit.c index 4e099b5945..bac7135bab 100644 --- a/src/mesa/drivers/dri/i965/brw_eu_emit.c +++ b/src/mesa/drivers/dri/i965/brw_eu_emit.c @@ -96,7 +96,7 @@ static void brw_set_dest( struct brw_instruction *insn, } static void brw_set_src0( struct brw_instruction *insn, - struct brw_reg reg ) + struct brw_reg reg ) { assert(reg.file != BRW_MESSAGE_REGISTER_FILE); @@ -169,7 +169,7 @@ static void brw_set_src0( struct brw_instruction *insn, void brw_set_src1( struct brw_instruction *insn, - struct brw_reg reg ) + struct brw_reg reg ) { assert(reg.file != BRW_MESSAGE_REGISTER_FILE); @@ -323,13 +323,13 @@ static void brw_set_dp_read_message( struct brw_instruction *insn, } static void brw_set_sampler_message(struct brw_context *brw, - struct brw_instruction *insn, - GLuint binding_table_index, - GLuint sampler, - GLuint msg_type, - GLuint response_length, - GLuint msg_length, - GLboolean eot) + struct brw_instruction *insn, + GLuint binding_table_index, + GLuint sampler, + GLuint msg_type, + GLuint response_length, + GLuint msg_length, + GLboolean eot) { brw_set_src1(insn, brw_imm_d(0)); @@ -407,7 +407,7 @@ static struct brw_instruction *brw_alu2(struct brw_compile *p, * Convenience routines. */ #define ALU1(OP) \ -struct brw_instruction *brw_##OP(struct brw_compile *p, \ +struct brw_instruction *brw_##OP(struct brw_compile *p, \ struct brw_reg dest, \ struct brw_reg src0) \ { \ @@ -415,7 +415,7 @@ struct brw_instruction *brw_##OP(struct brw_compile *p, \ } #define ALU2(OP) \ -struct brw_instruction *brw_##OP(struct brw_compile *p, \ +struct brw_instruction *brw_##OP(struct brw_compile *p, \ struct brw_reg dest, \ struct brw_reg src0, \ struct brw_reg src1) \ @@ -469,9 +469,9 @@ void brw_NOP(struct brw_compile *p) */ struct brw_instruction *brw_JMPI(struct brw_compile *p, - struct brw_reg dest, - struct brw_reg src0, - struct brw_reg src1) + struct brw_reg dest, + struct brw_reg src0, + struct brw_reg src1) { struct brw_instruction *insn = brw_alu2(p, BRW_OPCODE_JMPI, dest, src0, src1); @@ -674,7 +674,7 @@ struct brw_instruction *brw_DO(struct brw_compile *p, GLuint execute_size) struct brw_instruction *brw_WHILE(struct brw_compile *p, - struct brw_instruction *do_insn) + struct brw_instruction *do_insn) { struct brw_instruction *insn; @@ -931,13 +931,13 @@ void brw_dp_READ_16( struct brw_compile *p, void brw_fb_WRITE(struct brw_compile *p, - struct brw_reg dest, - GLuint msg_reg_nr, - struct brw_reg src0, - GLuint binding_table_index, - GLuint msg_length, - GLuint response_length, - GLboolean eot) + struct brw_reg dest, + GLuint msg_reg_nr, + struct brw_reg src0, + GLuint binding_table_index, + GLuint msg_length, + GLuint response_length, + GLboolean eot) { struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND); @@ -973,8 +973,8 @@ void brw_SAMPLE(struct brw_compile *p, { GLboolean need_stall = 0; - if(writemask == 0) { -/* _mesa_printf("%s: zero writemask??\n", __FUNCTION__); */ + if (writemask == 0) { + /*_mesa_printf("%s: zero writemask??\n", __FUNCTION__); */ return; } @@ -1006,7 +1006,7 @@ void brw_SAMPLE(struct brw_compile *p, if (newmask != writemask) { need_stall = 1; -/* _mesa_printf("need stall %x %x\n", newmask , writemask); */ + /* _mesa_printf("need stall %x %x\n", newmask , writemask); */ } else { struct brw_reg m1 = brw_message_reg(msg_reg_nr); @@ -1047,8 +1047,7 @@ void brw_SAMPLE(struct brw_compile *p, eot); } - if (need_stall) - { + if (need_stall) { struct brw_reg reg = vec8(offset(dest, response_length-1)); /* mov (8) r9.0<1>:f r9.0<8;8,1>:f { Align1 } diff --git a/src/mesa/drivers/dri/i965/brw_fallback.c b/src/mesa/drivers/dri/i965/brw_fallback.c index 4ea660a51a..e63098fdd4 100644 --- a/src/mesa/drivers/dri/i965/brw_fallback.c +++ b/src/mesa/drivers/dri/i965/brw_fallback.c @@ -47,20 +47,12 @@ static GLboolean do_check_fallback(struct brw_context *brw) GLcontext *ctx = &brw->intel.ctx; GLuint i; - /* BRW_NEW_METAOPS - */ - if (brw->metaops.active) - return GL_FALSE; - if (brw->intel.no_rast) { DBG("FALLBACK: rasterization disabled\n"); return GL_TRUE; } /* _NEW_RENDERMODE - * - * XXX: need to save/restore RenderMode in metaops state, or - * somehow move to a new attribs pointer: */ if (ctx->RenderMode != GL_RENDER) { DBG("FALLBACK: render mode\n"); @@ -70,7 +62,7 @@ static GLboolean do_check_fallback(struct brw_context *brw) /* _NEW_TEXTURE: */ for (i = 0; i < BRW_MAX_TEX_UNIT; i++) { - struct gl_texture_unit *texUnit = &brw->attribs.Texture->Unit[i]; + struct gl_texture_unit *texUnit = &ctx->Texture.Unit[i]; if (texUnit->_ReallyEnabled) { struct intel_texture_object *intelObj = intel_texture_object(texUnit->_Current); struct gl_texture_image *texImage = intelObj->base.Image[0][intelObj->firstLevel]; @@ -83,7 +75,7 @@ static GLboolean do_check_fallback(struct brw_context *brw) /* _NEW_STENCIL */ - if (brw->attribs.Stencil->Enabled && + if (ctx->Stencil.Enabled && !brw->intel.hw_stencil) { DBG("FALLBACK: stencil\n"); return GL_TRUE; @@ -101,7 +93,7 @@ static void check_fallback(struct brw_context *brw) const struct brw_tracked_state brw_check_fallback = { .dirty = { .mesa = _NEW_BUFFERS | _NEW_RENDERMODE | _NEW_TEXTURE | _NEW_STENCIL, - .brw = BRW_NEW_METAOPS, + .brw = 0, .cache = 0 }, .prepare = check_fallback diff --git a/src/mesa/drivers/dri/i965/brw_metaops.c b/src/mesa/drivers/dri/i965/brw_metaops.c deleted file mode 100644 index 41bfa2e256..0000000000 --- a/src/mesa/drivers/dri/i965/brw_metaops.c +++ /dev/null @@ -1,583 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell <keith@tungstengraphics.com> - * frame buffer texture by Gary Wong <gtw@gnu.org> - */ - - - -#include "main/glheader.h" -#include "main/context.h" -#include "main/macros.h" - -#include "shader/arbprogparse.h" - -#include "intel_screen.h" -#include "intel_batchbuffer.h" -#include "intel_regions.h" -#include "intel_buffers.h" - -#include "brw_context.h" -#include "brw_defines.h" -#include "brw_draw.h" -#include "brw_fallback.h" - -#define INIT(brw, STRUCT, ATTRIB) \ -do { \ - brw->attribs.ATTRIB = &ctx->ATTRIB; \ -} while (0) - -#define DUP(brw, STRUCT, ATTRIB) \ -do { \ - brw->metaops.attribs.ATTRIB = MALLOC_STRUCT(STRUCT); \ - memcpy(brw->metaops.attribs.ATTRIB, \ - brw->attribs.ATTRIB, \ - sizeof(struct STRUCT)); \ -} while (0) - - -#define INSTALL(brw, ATTRIB, STATE) \ -do { \ - brw->attribs.ATTRIB = brw->metaops.attribs.ATTRIB; \ - brw->state.dirty.mesa |= STATE; \ -} while (0) - -#define RESTORE(brw, ATTRIB, STATE) \ -do { \ - brw->attribs.ATTRIB = &brw->intel.ctx.ATTRIB; \ - brw->state.dirty.mesa |= STATE; \ -} while (0) - -static void init_attribs( struct brw_context *brw ) -{ - DUP(brw, gl_colorbuffer_attrib, Color); - DUP(brw, gl_depthbuffer_attrib, Depth); - DUP(brw, gl_fog_attrib, Fog); - DUP(brw, gl_hint_attrib, Hint); - DUP(brw, gl_light_attrib, Light); - DUP(brw, gl_line_attrib, Line); - DUP(brw, gl_point_attrib, Point); - DUP(brw, gl_polygon_attrib, Polygon); - DUP(brw, gl_scissor_attrib, Scissor); - DUP(brw, gl_stencil_attrib, Stencil); - DUP(brw, gl_texture_attrib, Texture); - DUP(brw, gl_transform_attrib, Transform); - DUP(brw, gl_viewport_attrib, Viewport); - DUP(brw, gl_vertex_program_state, VertexProgram); - DUP(brw, gl_fragment_program_state, FragmentProgram); -} - -static void install_attribs( struct brw_context *brw ) -{ - INSTALL(brw, Color, _NEW_COLOR); - INSTALL(brw, Depth, _NEW_DEPTH); - INSTALL(brw, Fog, _NEW_FOG); - INSTALL(brw, Hint, _NEW_HINT); - INSTALL(brw, Light, _NEW_LIGHT); - INSTALL(brw, Line, _NEW_LINE); - INSTALL(brw, Point, _NEW_POINT); - INSTALL(brw, Polygon, _NEW_POLYGON); - INSTALL(brw, Scissor, _NEW_SCISSOR); - INSTALL(brw, Stencil, _NEW_STENCIL); - INSTALL(brw, Texture, _NEW_TEXTURE); - INSTALL(brw, Transform, _NEW_TRANSFORM); - INSTALL(brw, Viewport, _NEW_VIEWPORT); - INSTALL(brw, VertexProgram, _NEW_PROGRAM); - INSTALL(brw, FragmentProgram, _NEW_PROGRAM); -} - -static void restore_attribs( struct brw_context *brw ) -{ - RESTORE(brw, Color, _NEW_COLOR); - RESTORE(brw, Depth, _NEW_DEPTH); - RESTORE(brw, Fog, _NEW_FOG); - RESTORE(brw, Hint, _NEW_HINT); - RESTORE(brw, Light, _NEW_LIGHT); - RESTORE(brw, Line, _NEW_LINE); - RESTORE(brw, Point, _NEW_POINT); - RESTORE(brw, Polygon, _NEW_POLYGON); - RESTORE(brw, Scissor, _NEW_SCISSOR); - RESTORE(brw, Stencil, _NEW_STENCIL); - RESTORE(brw, Texture, _NEW_TEXTURE); - RESTORE(brw, Transform, _NEW_TRANSFORM); - RESTORE(brw, Viewport, _NEW_VIEWPORT); - RESTORE(brw, VertexProgram, _NEW_PROGRAM); - RESTORE(brw, FragmentProgram, _NEW_PROGRAM); -} - - -static const char *vp_prog = - "!!ARBvp1.0\n" - "MOV result.color, vertex.color;\n" - "MOV result.position, vertex.position;\n" - "END\n"; - -static const char *fp_prog = - "!!ARBfp1.0\n" - "MOV result.color, fragment.color;\n" - "END\n"; - -static const char *fp_tex_prog = - "!!ARBfp1.0\n" - "TEMP a;\n" - "ADD a, fragment.position, program.local[0];\n" - "MUL a, a, program.local[1];\n" - "TEX result.color, a, texture[0], 2D;\n" - "MOV result.depth.z, fragment.position;\n" - "END\n"; - -/* Derived values of importance: - * - * FragmentProgram->_Current - * VertexProgram->_Enabled - * brw->vertex_program - * DrawBuffer->_ColorDrawBufferIndexes[0] - * - * - * More if drawpixels-through-texture is added. - */ -static void init_metaops_state( struct brw_context *brw ) -{ - GLcontext *ctx = &brw->intel.ctx; - - brw->metaops.vbo = ctx->Driver.NewBufferObject(ctx, 1, GL_ARRAY_BUFFER_ARB); - - ctx->Driver.BufferData(ctx, - GL_ARRAY_BUFFER_ARB, - 4096, - NULL, - GL_DYNAMIC_DRAW_ARB, - brw->metaops.vbo); - - brw->metaops.fp = (struct gl_fragment_program *) - ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 1 ); - - brw->metaops.fp_tex = (struct gl_fragment_program *) - ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 1 ); - - brw->metaops.vp = (struct gl_vertex_program *) - ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 1 ); - - _mesa_parse_arb_fragment_program(ctx, GL_FRAGMENT_PROGRAM_ARB, - fp_prog, strlen(fp_prog), - brw->metaops.fp); - - _mesa_parse_arb_fragment_program(ctx, GL_FRAGMENT_PROGRAM_ARB, - fp_tex_prog, strlen(fp_tex_prog), - brw->metaops.fp_tex); - - _mesa_parse_arb_vertex_program(ctx, GL_VERTEX_PROGRAM_ARB, - vp_prog, strlen(vp_prog), - brw->metaops.vp); - - brw->metaops.attribs.VertexProgram->_Current = brw->metaops.vp; - brw->metaops.attribs.VertexProgram->_Enabled = GL_TRUE; - - brw->metaops.attribs.FragmentProgram->_Current = brw->metaops.fp; -} - -static void meta_flat_shade( struct intel_context *intel ) -{ - struct brw_context *brw = brw_context(&intel->ctx); - - brw->metaops.attribs.Light->ShadeModel = GL_FLAT; - brw->state.dirty.mesa |= _NEW_LIGHT; -} - - -static void meta_no_stencil_write( struct intel_context *intel ) -{ - struct brw_context *brw = brw_context(&intel->ctx); - - brw->metaops.attribs.Stencil->Enabled = GL_FALSE; - brw->metaops.attribs.Stencil->WriteMask[0] = GL_FALSE; - brw->state.dirty.mesa |= _NEW_STENCIL; -} - -static void meta_no_depth_write( struct intel_context *intel ) -{ - struct brw_context *brw = brw_context(&intel->ctx); - - brw->metaops.attribs.Depth->Test = GL_FALSE; - brw->metaops.attribs.Depth->Mask = GL_FALSE; - brw->state.dirty.mesa |= _NEW_DEPTH; -} - - -static void meta_depth_replace( struct intel_context *intel ) -{ - struct brw_context *brw = brw_context(&intel->ctx); - - /* ctx->Driver.Enable( ctx, GL_DEPTH_TEST, GL_TRUE ) - * ctx->Driver.DepthMask( ctx, GL_TRUE ) - */ - brw->metaops.attribs.Depth->Test = GL_TRUE; - brw->metaops.attribs.Depth->Mask = GL_TRUE; - brw->state.dirty.mesa |= _NEW_DEPTH; - - /* ctx->Driver.DepthFunc( ctx, GL_ALWAYS ) - */ - brw->metaops.attribs.Depth->Func = GL_ALWAYS; - - brw->state.dirty.mesa |= _NEW_DEPTH; -} - - -static void meta_stencil_replace( struct intel_context *intel, - GLuint s_mask, - GLuint s_clear) -{ - struct brw_context *brw = brw_context(&intel->ctx); - - brw->metaops.attribs.Stencil->Enabled = GL_TRUE; - brw->metaops.attribs.Stencil->WriteMask[0] = s_mask; - brw->metaops.attribs.Stencil->ValueMask[0] = 0xff; - brw->metaops.attribs.Stencil->Ref[0] = s_clear; - brw->metaops.attribs.Stencil->Function[0] = GL_ALWAYS; - brw->metaops.attribs.Stencil->FailFunc[0] = GL_REPLACE; - brw->metaops.attribs.Stencil->ZPassFunc[0] = GL_REPLACE; - brw->metaops.attribs.Stencil->ZFailFunc[0] = GL_REPLACE; - brw->state.dirty.mesa |= _NEW_STENCIL; -} - - -static void meta_color_mask( struct intel_context *intel, GLboolean state ) -{ - struct brw_context *brw = brw_context(&intel->ctx); - - if (state) - COPY_4V(brw->metaops.attribs.Color->ColorMask, - brw->intel.ctx.Color.ColorMask); - else - ASSIGN_4V(brw->metaops.attribs.Color->ColorMask, 0, 0, 0, 0); - - brw->state.dirty.mesa |= _NEW_COLOR; -} - -static void meta_no_texture( struct intel_context *intel ) -{ - struct brw_context *brw = brw_context(&intel->ctx); - - brw->metaops.attribs.FragmentProgram->_Current = brw->metaops.fp; - - brw->metaops.attribs.Texture->CurrentUnit = 0; - brw->metaops.attribs.Texture->_EnabledUnits = 0; - brw->metaops.attribs.Texture->_EnabledCoordUnits = 0; - brw->metaops.attribs.Texture->Unit[ 0 ].Enabled = 0; - brw->metaops.attribs.Texture->Unit[ 0 ]._ReallyEnabled = 0; - - brw->state.dirty.mesa |= _NEW_TEXTURE | _NEW_PROGRAM; -} - -static void meta_texture_blend_replace(struct intel_context *intel) -{ - struct brw_context *brw = brw_context(&intel->ctx); - - brw->metaops.attribs.Texture->CurrentUnit = 0; - brw->metaops.attribs.Texture->_EnabledUnits = 1; - brw->metaops.attribs.Texture->_EnabledCoordUnits = 1; - brw->metaops.attribs.Texture->Unit[ 0 ].Enabled = TEXTURE_2D_BIT; - brw->metaops.attribs.Texture->Unit[ 0 ]._ReallyEnabled = TEXTURE_2D_BIT; - brw->metaops.attribs.Texture->Unit[ 0 ].Current2D = - intel->frame_buffer_texobj; - brw->metaops.attribs.Texture->Unit[ 0 ]._Current = - intel->frame_buffer_texobj; - - brw->state.dirty.mesa |= _NEW_TEXTURE | _NEW_PROGRAM; -} - -static void meta_import_pixel_state(struct intel_context *intel) -{ - struct brw_context *brw = brw_context(&intel->ctx); - - RESTORE(brw, Color, _NEW_COLOR); - RESTORE(brw, Depth, _NEW_DEPTH); - RESTORE(brw, Fog, _NEW_FOG); - RESTORE(brw, Scissor, _NEW_SCISSOR); - RESTORE(brw, Stencil, _NEW_STENCIL); - RESTORE(brw, Texture, _NEW_TEXTURE); - RESTORE(brw, FragmentProgram, _NEW_PROGRAM); -} - -static void meta_frame_buffer_texture( struct intel_context *intel, - GLint xoff, GLint yoff ) -{ - struct brw_context *brw = brw_context(&intel->ctx); - struct intel_region *region = intel_drawbuf_region( intel ); - - INSTALL(brw, FragmentProgram, _NEW_PROGRAM); - - brw->metaops.attribs.FragmentProgram->_Current = brw->metaops.fp_tex; - /* This is unfortunate, but seems to be necessary, since later on we - will end up calling _mesa_load_state_parameters to lookup the - local params (below), and that will want to look in ctx.FragmentProgram - instead of brw->attribs.FragmentProgram. */ - intel->ctx.FragmentProgram.Current = brw->metaops.fp_tex; - - brw->metaops.fp_tex->Base.LocalParams[ 0 ][ 0 ] = xoff; - brw->metaops.fp_tex->Base.LocalParams[ 0 ][ 1 ] = yoff; - brw->metaops.fp_tex->Base.LocalParams[ 0 ][ 2 ] = 0.0; - brw->metaops.fp_tex->Base.LocalParams[ 0 ][ 3 ] = 0.0; - brw->metaops.fp_tex->Base.LocalParams[ 1 ][ 0 ] = - 1.0 / region->pitch; - brw->metaops.fp_tex->Base.LocalParams[ 1 ][ 1 ] = - -1.0 / region->height; - brw->metaops.fp_tex->Base.LocalParams[ 1 ][ 2 ] = 0.0; - brw->metaops.fp_tex->Base.LocalParams[ 1 ][ 3 ] = 1.0; - - brw->state.dirty.mesa |= _NEW_PROGRAM; -} - - -static void meta_draw_region( struct intel_context *intel, - struct intel_region *draw_region, - struct intel_region *depth_region ) -{ - struct brw_context *brw = brw_context(&intel->ctx); - - if (!brw->metaops.saved_draw_region) { - brw->metaops.saved_draw_region = brw->state.draw_regions[0]; - brw->metaops.saved_nr_draw_regions = brw->state.nr_draw_regions; - brw->metaops.saved_depth_region = brw->state.depth_region; - } - - brw->state.draw_regions[0] = draw_region; - brw->state.nr_draw_regions = 1; - brw->state.depth_region = depth_region; - - if (intel->frame_buffer_texobj != NULL) - brw_FrameBufferTexDestroy(brw); - - if (draw_region) - brw_FrameBufferTexInit(brw, draw_region); - - brw->state.dirty.mesa |= _NEW_BUFFERS; -} - - -static void 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) -{ - GLcontext *ctx = &intel->ctx; - struct brw_context *brw = brw_context(&intel->ctx); - struct gl_client_array pos_array; - struct gl_client_array color_array; - struct gl_client_array *attribs[VERT_ATTRIB_MAX]; - struct _mesa_prim prim[1]; - GLfloat pos[4][3]; - - ctx->Driver.BufferData(ctx, - GL_ARRAY_BUFFER_ARB, - sizeof(pos) + sizeof(color), - NULL, - GL_DYNAMIC_DRAW_ARB, - brw->metaops.vbo); - - pos[0][0] = x0; - pos[0][1] = y0; - pos[0][2] = z; - - pos[1][0] = x1; - pos[1][1] = y0; - pos[1][2] = z; - - pos[2][0] = x1; - pos[2][1] = y1; - pos[2][2] = z; - - pos[3][0] = x0; - pos[3][1] = y1; - pos[3][2] = z; - - ctx->Driver.BufferSubData(ctx, - GL_ARRAY_BUFFER_ARB, - 0, - sizeof(pos), - pos, - brw->metaops.vbo); - - /* Convert incoming ARGB to required RGBA */ - /* Note this color is stored as GL_UNSIGNED_BYTE */ - color = (color & 0xff00ff00) | (((color >> 16) | (color << 16)) & 0xff00ff); - - ctx->Driver.BufferSubData(ctx, - GL_ARRAY_BUFFER_ARB, - sizeof(pos), - sizeof(color), - &color, - brw->metaops.vbo); - - /* Ignoring texture coords. - */ - - memset(attribs, 0, VERT_ATTRIB_MAX * sizeof(*attribs)); - - attribs[VERT_ATTRIB_POS] = &pos_array; - attribs[VERT_ATTRIB_POS]->Ptr = 0; - attribs[VERT_ATTRIB_POS]->Type = GL_FLOAT; - attribs[VERT_ATTRIB_POS]->Enabled = 1; - attribs[VERT_ATTRIB_POS]->Size = 3; - attribs[VERT_ATTRIB_POS]->StrideB = 3 * sizeof(GLfloat); - attribs[VERT_ATTRIB_POS]->Stride = 3 * sizeof(GLfloat); - attribs[VERT_ATTRIB_POS]->_MaxElement = 4; - attribs[VERT_ATTRIB_POS]->Normalized = 0; - attribs[VERT_ATTRIB_POS]->BufferObj = brw->metaops.vbo; - - attribs[VERT_ATTRIB_COLOR0] = &color_array; - attribs[VERT_ATTRIB_COLOR0]->Ptr = (const GLubyte *)sizeof(pos); - attribs[VERT_ATTRIB_COLOR0]->Type = GL_UNSIGNED_BYTE; - attribs[VERT_ATTRIB_COLOR0]->Enabled = 1; - attribs[VERT_ATTRIB_COLOR0]->Size = 4; - attribs[VERT_ATTRIB_COLOR0]->StrideB = 0; - attribs[VERT_ATTRIB_COLOR0]->Stride = 0; - attribs[VERT_ATTRIB_COLOR0]->_MaxElement = 1; - attribs[VERT_ATTRIB_COLOR0]->Normalized = 1; - attribs[VERT_ATTRIB_COLOR0]->BufferObj = brw->metaops.vbo; - - /* Just ignoring texture coordinates for now. - */ - - memset(prim, 0, sizeof(*prim)); - - prim[0].mode = GL_TRIANGLE_FAN; - prim[0].begin = 1; - prim[0].end = 1; - prim[0].weak = 0; - prim[0].pad = 0; - prim[0].start = 0; - prim[0].count = 4; - - brw_draw_prims(&brw->intel.ctx, - (const struct gl_client_array **)attribs, - prim, 1, - NULL, - 0, - 3 ); -} - - -static void install_meta_state( struct intel_context *intel ) -{ - GLcontext *ctx = &intel->ctx; - struct brw_context *brw = brw_context(ctx); - GLuint i; - - if (!brw->metaops.vbo) { - init_metaops_state(brw); - } - - install_attribs(brw); - - meta_no_texture(&brw->intel); - meta_flat_shade(&brw->intel); - for (i = 0; i < ctx->Const.MaxDrawBuffers; i++) { - brw->metaops.restore_draw_buffers[i] - = ctx->DrawBuffer->_ColorDrawBufferIndexes[i]; - } - brw->metaops.restore_num_draw_buffers = ctx->DrawBuffer->_NumColorDrawBuffers; - - brw->metaops.restore_fp = ctx->FragmentProgram.Current; - - /* This works without adjusting refcounts. Fix later? - */ - brw->metaops.saved_draw_region = brw->state.draw_regions[0]; - brw->metaops.saved_nr_draw_regions = brw->state.nr_draw_regions; - brw->metaops.saved_depth_region = brw->state.depth_region; - brw->metaops.active = 1; - - brw->state.dirty.brw |= BRW_NEW_METAOPS; -} - -static void leave_meta_state( struct intel_context *intel ) -{ - GLcontext *ctx = &intel->ctx; - struct brw_context *brw = brw_context(ctx); - GLuint i; - - restore_attribs(brw); - - for (i = 0; i < ctx->Const.MaxDrawBuffers; i++) { - ctx->DrawBuffer->_ColorDrawBufferIndexes[i] - = brw->metaops.restore_draw_buffers[i]; - } - ctx->DrawBuffer->_NumColorDrawBuffers = brw->metaops.restore_num_draw_buffers; - - ctx->FragmentProgram.Current = brw->metaops.restore_fp; - - brw->state.draw_regions[0] = brw->metaops.saved_draw_region; - brw->state.nr_draw_regions = brw->metaops.saved_nr_draw_regions; - brw->state.depth_region = brw->metaops.saved_depth_region; - brw->metaops.saved_draw_region = NULL; - brw->metaops.saved_depth_region = NULL; - brw->metaops.active = 0; - - brw->state.dirty.mesa |= _NEW_BUFFERS; - brw->state.dirty.brw |= BRW_NEW_METAOPS; -} - - - -void brw_init_metaops( struct brw_context *brw ) -{ - init_attribs(brw); - - - brw->intel.vtbl.install_meta_state = install_meta_state; - brw->intel.vtbl.leave_meta_state = leave_meta_state; - brw->intel.vtbl.meta_no_depth_write = meta_no_depth_write; - brw->intel.vtbl.meta_no_stencil_write = meta_no_stencil_write; - brw->intel.vtbl.meta_stencil_replace = meta_stencil_replace; - brw->intel.vtbl.meta_depth_replace = meta_depth_replace; - brw->intel.vtbl.meta_color_mask = meta_color_mask; - brw->intel.vtbl.meta_no_texture = meta_no_texture; - brw->intel.vtbl.meta_import_pixel_state = meta_import_pixel_state; - brw->intel.vtbl.meta_frame_buffer_texture = meta_frame_buffer_texture; - brw->intel.vtbl.meta_draw_region = meta_draw_region; - brw->intel.vtbl.meta_draw_quad = meta_draw_quad; - brw->intel.vtbl.meta_texture_blend_replace = meta_texture_blend_replace; -/* brw->intel.vtbl.meta_tex_rect_source = meta_tex_rect_source; */ -/* brw->intel.vtbl.meta_draw_format = set_draw_format; */ -} - -void brw_destroy_metaops( struct brw_context *brw ) -{ - GLcontext *ctx = &brw->intel.ctx; - - if (brw->metaops.vbo) - ctx->Driver.DeleteBuffer( ctx, brw->metaops.vbo ); - -/* ctx->Driver.DeleteProgram( ctx, brw->metaops.fp ); */ -/* ctx->Driver.DeleteProgram( ctx, brw->metaops.fp_tex ); */ -/* ctx->Driver.DeleteProgram( ctx, brw->metaops.vp ); */ -} diff --git a/src/mesa/drivers/dri/i965/brw_misc_state.c b/src/mesa/drivers/dri/i965/brw_misc_state.c index 627705fa9b..9dcdad7b4e 100644 --- a/src/mesa/drivers/dri/i965/brw_misc_state.c +++ b/src/mesa/drivers/dri/i965/brw_misc_state.c @@ -48,15 +48,16 @@ static void upload_blend_constant_color(struct brw_context *brw) { + GLcontext *ctx = &brw->intel.ctx; struct brw_blend_constant_color bcc; memset(&bcc, 0, sizeof(bcc)); bcc.header.opcode = CMD_BLEND_CONSTANT_COLOR; bcc.header.length = sizeof(bcc)/4-2; - bcc.blend_constant_color[0] = brw->attribs.Color->BlendColor[0]; - bcc.blend_constant_color[1] = brw->attribs.Color->BlendColor[1]; - bcc.blend_constant_color[2] = brw->attribs.Color->BlendColor[2]; - bcc.blend_constant_color[3] = brw->attribs.Color->BlendColor[3]; + bcc.blend_constant_color[0] = ctx->Color.BlendColor[0]; + bcc.blend_constant_color[1] = ctx->Color.BlendColor[1]; + bcc.blend_constant_color[2] = ctx->Color.BlendColor[2]; + bcc.blend_constant_color[3] = ctx->Color.BlendColor[3]; BRW_CACHED_BATCH_STRUCT(brw, &bcc); } @@ -154,10 +155,7 @@ static void upload_pipelined_state_pointers(struct brw_context *brw ) OUT_RELOC(brw->gs.state_bo, I915_GEM_DOMAIN_INSTRUCTION, 0, 1); else OUT_BATCH(0); - if (!brw->metaops.active) - OUT_RELOC(brw->clip.state_bo, I915_GEM_DOMAIN_INSTRUCTION, 0, 1); - else - OUT_BATCH(0); + OUT_RELOC(brw->clip.state_bo, I915_GEM_DOMAIN_INSTRUCTION, 0, 1); OUT_RELOC(brw->sf.state_bo, I915_GEM_DOMAIN_INSTRUCTION, 0, 0); OUT_RELOC(brw->wm.state_bo, I915_GEM_DOMAIN_INSTRUCTION, 0, 0); OUT_RELOC(brw->cc.state_bo, I915_GEM_DOMAIN_INSTRUCTION, 0, 0); @@ -186,7 +184,7 @@ static void upload_psp_urb_cbs(struct brw_context *brw ) const struct brw_tracked_state brw_psp_urb_cbs = { .dirty = { .mesa = 0, - .brw = BRW_NEW_URB_FENCE | BRW_NEW_METAOPS | BRW_NEW_BATCH, + .brw = BRW_NEW_URB_FENCE | BRW_NEW_BATCH, .cache = (CACHE_NEW_VS_UNIT | CACHE_NEW_GS_UNIT | CACHE_NEW_GS_PROG | @@ -284,6 +282,7 @@ const struct brw_tracked_state brw_depthbuffer = { static void upload_polygon_stipple(struct brw_context *brw) { + GLcontext *ctx = &brw->intel.ctx; struct brw_polygon_stipple bps; GLuint i; @@ -292,7 +291,7 @@ static void upload_polygon_stipple(struct brw_context *brw) bps.header.length = sizeof(bps)/4-2; for (i = 0; i < 32; i++) - bps.stipple[i] = brw->attribs.PolygonStipple[31 - i]; /* invert */ + bps.stipple[i] = ctx->PolygonStipple[31 - i]; /* invert */ BRW_CACHED_BATCH_STRUCT(brw, &bps); } @@ -370,6 +369,7 @@ const struct brw_tracked_state brw_aa_line_parameters = { static void upload_line_stipple(struct brw_context *brw) { + GLcontext *ctx = &brw->intel.ctx; struct brw_line_stipple bls; GLfloat tmp; GLint tmpi; @@ -378,10 +378,10 @@ static void upload_line_stipple(struct brw_context *brw) bls.header.opcode = CMD_LINE_STIPPLE_PATTERN; bls.header.length = sizeof(bls)/4 - 2; - bls.bits0.pattern = brw->attribs.Line->StipplePattern; - bls.bits1.repeat_count = brw->attribs.Line->StippleFactor; + bls.bits0.pattern = ctx->Line.StipplePattern; + bls.bits1.repeat_count = ctx->Line.StippleFactor; - tmp = 1.0 / (GLfloat) brw->attribs.Line->StippleFactor; + tmp = 1.0 / (GLfloat) ctx->Line.StippleFactor; tmpi = tmp * (1<<13); diff --git a/src/mesa/drivers/dri/i965/brw_sf.c b/src/mesa/drivers/dri/i965/brw_sf.c index 1a11d54621..8c1711538a 100644 --- a/src/mesa/drivers/dri/i965/brw_sf.c +++ b/src/mesa/drivers/dri/i965/brw_sf.c @@ -46,6 +46,7 @@ static void compile_sf_prog( struct brw_context *brw, struct brw_sf_prog_key *key ) { + GLcontext *ctx = &brw->intel.ctx; struct brw_sf_compile c; const GLuint *program; GLuint program_size; @@ -74,7 +75,7 @@ static void compile_sf_prog( struct brw_context *brw, c.idx_to_attr[idx] = i; if (i >= VERT_RESULT_TEX0 && i <= VERT_RESULT_TEX7) { c.point_attrs[i].CoordReplace = - brw->attribs.Point->CoordReplace[i - VERT_RESULT_TEX0]; + ctx->Point.CoordReplace[i - VERT_RESULT_TEX0]; } else { c.point_attrs[i].CoordReplace = GL_FALSE; @@ -128,6 +129,7 @@ static void compile_sf_prog( struct brw_context *brw, */ static void upload_sf_prog(struct brw_context *brw) { + GLcontext *ctx = &brw->intel.ctx; struct brw_sf_prog_key key; memset(&key, 0, sizeof(key)); @@ -158,15 +160,15 @@ static void upload_sf_prog(struct brw_context *brw) break; } - key.do_point_sprite = brw->attribs.Point->PointSprite; - key.SpriteOrigin = brw->attribs.Point->SpriteOrigin; + key.do_point_sprite = ctx->Point.PointSprite; + key.SpriteOrigin = ctx->Point.SpriteOrigin; /* _NEW_LIGHT */ - key.do_flat_shading = (brw->attribs.Light->ShadeModel == GL_FLAT); - key.do_twoside_color = (brw->attribs.Light->Enabled && brw->attribs.Light->Model.TwoSide); + key.do_flat_shading = (ctx->Light.ShadeModel == GL_FLAT); + key.do_twoside_color = (ctx->Light.Enabled && ctx->Light.Model.TwoSide); /* _NEW_POLYGON */ if (key.do_twoside_color) - key.frontface_ccw = (brw->attribs.Polygon->FrontFace == GL_CCW); + key.frontface_ccw = (ctx->Polygon.FrontFace == GL_CCW); dri_bo_unreference(brw->sf.prog_bo); brw->sf.prog_bo = brw_search_cache(&brw->cache, BRW_SF_PROG, diff --git a/src/mesa/drivers/dri/i965/brw_sf_state.c b/src/mesa/drivers/dri/i965/brw_sf_state.c index 242b7047a1..e96d5354b3 100644 --- a/src/mesa/drivers/dri/i965/brw_sf_state.c +++ b/src/mesa/drivers/dri/i965/brw_sf_state.c @@ -43,10 +43,11 @@ static void upload_sf_vp(struct brw_context *brw) const GLfloat depth_scale = 1.0F / ctx->DrawBuffer->_DepthMaxF; struct brw_sf_viewport sfv; GLfloat y_scale, y_bias; + const GLboolean render_to_fbo = (ctx->DrawBuffer->Name != 0); memset(&sfv, 0, sizeof(sfv)); - if (intel_rendering_to_texture(ctx)) { + if (render_to_fbo) { y_scale = 1.0; y_bias = 0; } @@ -55,25 +56,16 @@ static void upload_sf_vp(struct brw_context *brw) y_bias = ctx->DrawBuffer->Height; } - /* _NEW_VIEWPORT, BRW_NEW_METAOPS */ - - if (!brw->metaops.active) { - const GLfloat *v = ctx->Viewport._WindowMap.m; - - sfv.viewport.m00 = v[MAT_SX]; - sfv.viewport.m11 = v[MAT_SY] * y_scale; - sfv.viewport.m22 = v[MAT_SZ] * depth_scale; - sfv.viewport.m30 = v[MAT_TX]; - sfv.viewport.m31 = v[MAT_TY] * y_scale + y_bias; - sfv.viewport.m32 = v[MAT_TZ] * depth_scale; - } else { - sfv.viewport.m00 = 1; - sfv.viewport.m11 = - 1; - sfv.viewport.m22 = 1; - sfv.viewport.m30 = 0; - sfv.viewport.m31 = ctx->DrawBuffer->Height; - sfv.viewport.m32 = 0; - } + /* _NEW_VIEWPORT */ + + const GLfloat *v = ctx->Viewport._WindowMap.m; + + sfv.viewport.m00 = v[MAT_SX]; + sfv.viewport.m11 = v[MAT_SY] * y_scale; + sfv.viewport.m22 = v[MAT_SZ] * depth_scale; + sfv.viewport.m30 = v[MAT_TX]; + sfv.viewport.m31 = v[MAT_TY] * y_scale + y_bias; + sfv.viewport.m32 = v[MAT_TZ] * depth_scale; /* _NEW_SCISSOR */ @@ -84,10 +76,20 @@ static void upload_sf_vp(struct brw_context *brw) * Note that the hardware's coordinates are inclusive, while Mesa's min is * inclusive but max is exclusive. */ - sfv.scissor.xmin = ctx->DrawBuffer->_Xmin; - sfv.scissor.xmax = ctx->DrawBuffer->_Xmax - 1; - sfv.scissor.ymin = ctx->DrawBuffer->Height - ctx->DrawBuffer->_Ymax; - sfv.scissor.ymax = ctx->DrawBuffer->Height - ctx->DrawBuffer->_Ymin - 1; + if (render_to_fbo) { + /* texmemory: Y=0=bottom */ + sfv.scissor.xmin = ctx->DrawBuffer->_Xmin; + sfv.scissor.xmax = ctx->DrawBuffer->_Xmax - 1; + sfv.scissor.ymin = ctx->DrawBuffer->_Ymin; + sfv.scissor.ymax = ctx->DrawBuffer->_Ymax - 1; + } + else { + /* memory: Y=0=top */ + sfv.scissor.xmin = ctx->DrawBuffer->_Xmin; + sfv.scissor.xmax = ctx->DrawBuffer->_Xmax - 1; + sfv.scissor.ymin = ctx->DrawBuffer->Height - ctx->DrawBuffer->_Ymax; + sfv.scissor.ymax = ctx->DrawBuffer->Height - ctx->DrawBuffer->_Ymin - 1; + } dri_bo_unreference(brw->sf.vp_bo); brw->sf.vp_bo = brw_cache_data( &brw->cache, BRW_SF_VP, &sfv, NULL, 0 ); @@ -97,7 +99,7 @@ const struct brw_tracked_state brw_sf_vp = { .dirty = { .mesa = (_NEW_VIEWPORT | _NEW_SCISSOR), - .brw = BRW_NEW_METAOPS, + .brw = 0, .cache = 0 }, .prepare = upload_sf_vp @@ -110,15 +112,19 @@ struct brw_sf_unit_key { unsigned int nr_urb_entries, urb_size, sfsize; GLenum front_face, cull_face; - GLboolean scissor, line_smooth, point_sprite, point_attenuated; + unsigned scissor:1; + unsigned line_smooth:1; + unsigned point_sprite:1; + unsigned point_attenuated:1; + unsigned render_to_fbo:1; float line_width; float point_size; - GLboolean render_to_texture; }; static void sf_unit_populate_key(struct brw_context *brw, struct brw_sf_unit_key *key) { + GLcontext *ctx = &brw->intel.ctx; memset(key, 0, sizeof(*key)); /* CACHE_NEW_SF_PROG */ @@ -130,22 +136,22 @@ sf_unit_populate_key(struct brw_context *brw, struct brw_sf_unit_key *key) key->urb_size = brw->urb.vsize; key->sfsize = brw->urb.sfsize; - key->scissor = brw->attribs.Scissor->Enabled; - key->front_face = brw->attribs.Polygon->FrontFace; + key->scissor = ctx->Scissor.Enabled; + key->front_face = ctx->Polygon.FrontFace; - if (brw->attribs.Polygon->CullFlag) - key->cull_face = brw->attribs.Polygon->CullFaceMode; + if (ctx->Polygon.CullFlag) + key->cull_face = ctx->Polygon.CullFaceMode; else key->cull_face = GL_NONE; - key->line_width = brw->attribs.Line->Width; - key->line_smooth = brw->attribs.Line->SmoothFlag; + key->line_width = ctx->Line.Width; + key->line_smooth = ctx->Line.SmoothFlag; - key->point_sprite = brw->attribs.Point->PointSprite; - key->point_size = brw->attribs.Point->Size; - key->point_attenuated = brw->attribs.Point->_Attenuated; + key->point_sprite = ctx->Point.PointSprite; + key->point_size = ctx->Point.Size; + key->point_attenuated = ctx->Point._Attenuated; - key->render_to_texture = intel_rendering_to_texture(&brw->intel.ctx); + key->render_to_fbo = brw->intel.ctx.DrawBuffer->Name != 0; } static dri_bo * @@ -192,10 +198,10 @@ sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key, else sf.sf5.front_winding = BRW_FRONTWINDING_CW; - /* The viewport is inverted for rendering to texture, and that inverts + /* The viewport is inverted for rendering to a FBO, and that inverts * polygon front/back orientation. */ - sf.sf5.front_winding ^= key->render_to_texture; + sf.sf5.front_winding ^= key->render_to_fbo; switch (key->cull_face) { case GL_FRONT: @@ -295,8 +301,7 @@ const struct brw_tracked_state brw_sf_unit = { _NEW_LINE | _NEW_POINT | _NEW_SCISSOR), - .brw = (BRW_NEW_URB_FENCE | - BRW_NEW_METAOPS), + .brw = BRW_NEW_URB_FENCE, .cache = (CACHE_NEW_SF_VP | CACHE_NEW_SF_PROG) }, diff --git a/src/mesa/drivers/dri/i965/brw_state_upload.c b/src/mesa/drivers/dri/i965/brw_state_upload.c index 4845859b3e..5de1450e61 100644 --- a/src/mesa/drivers/dri/i965/brw_state_upload.c +++ b/src/mesa/drivers/dri/i965/brw_state_upload.c @@ -233,7 +233,6 @@ static struct dirty_bit_map brw_bits[] = { DEFINE_BIT(BRW_NEW_WM_INPUT_DIMENSIONS), DEFINE_BIT(BRW_NEW_INPUT_VARYING), DEFINE_BIT(BRW_NEW_PSP), - DEFINE_BIT(BRW_NEW_METAOPS), DEFINE_BIT(BRW_NEW_FENCE), DEFINE_BIT(BRW_NEW_INDICES), DEFINE_BIT(BRW_NEW_VERTICES), @@ -298,6 +297,7 @@ brw_print_dirty_count(struct dirty_bit_map *bit_map, int32_t bits) */ void brw_validate_state( struct brw_context *brw ) { + GLcontext *ctx = &brw->intel.ctx; struct intel_context *intel = &brw->intel; struct brw_state_flags *state = &brw->state.dirty; GLuint i; @@ -314,13 +314,13 @@ void brw_validate_state( struct brw_context *brw ) state->brw |= ~0; } - if (brw->fragment_program != brw->attribs.FragmentProgram->_Current) { - brw->fragment_program = brw->attribs.FragmentProgram->_Current; + if (brw->fragment_program != ctx->FragmentProgram._Current) { + brw->fragment_program = ctx->FragmentProgram._Current; brw->state.dirty.brw |= BRW_NEW_FRAGMENT_PROGRAM; } - if (brw->vertex_program != brw->attribs.VertexProgram->_Current) { - brw->vertex_program = brw->attribs.VertexProgram->_Current; + if (brw->vertex_program != ctx->VertexProgram._Current) { + brw->vertex_program = ctx->VertexProgram._Current; brw->state.dirty.brw |= BRW_NEW_VERTEX_PROGRAM; } diff --git a/src/mesa/drivers/dri/i965/brw_structs.h b/src/mesa/drivers/dri/i965/brw_structs.h index 4e577d0f6a..d97ff27f0a 100644 --- a/src/mesa/drivers/dri/i965/brw_structs.h +++ b/src/mesa/drivers/dri/i965/brw_structs.h @@ -1031,10 +1031,10 @@ struct brw_surface_state GLuint writedisable_green:1; GLuint writedisable_red:1; GLuint writedisable_alpha:1; - GLuint surface_format:9; + GLuint surface_format:9; /**< BRW_SURFACEFORMAT_x */ GLuint data_return_format:1; GLuint pad0:1; - GLuint surface_type:3; + GLuint surface_type:3; /**< BRW_SURFACE_1D/2D/3D/CUBE */ } ss0; struct { diff --git a/src/mesa/drivers/dri/i965/brw_tex.c b/src/mesa/drivers/dri/i965/brw_tex.c index 0bb6f176a0..ef99e9c1ae 100644 --- a/src/mesa/drivers/dri/i965/brw_tex.c +++ b/src/mesa/drivers/dri/i965/brw_tex.c @@ -86,11 +86,12 @@ void brw_FrameBufferTexDestroy( struct brw_context *brw ) */ void brw_validate_textures( struct brw_context *brw ) { + GLcontext *ctx = &brw->intel.ctx; struct intel_context *intel = &brw->intel; int i; for (i = 0; i < BRW_MAX_TEX_UNIT; i++) { - struct gl_texture_unit *texUnit = &brw->attribs.Texture->Unit[i]; + struct gl_texture_unit *texUnit = &ctx->Texture.Unit[i]; if (texUnit->_ReallyEnabled) { intel_finalize_mipmap_tree(intel, i); diff --git a/src/mesa/drivers/dri/i965/brw_vs.c b/src/mesa/drivers/dri/i965/brw_vs.c index 1db7ceebcf..e3111c6680 100644 --- a/src/mesa/drivers/dri/i965/brw_vs.c +++ b/src/mesa/drivers/dri/i965/brw_vs.c @@ -85,6 +85,7 @@ static void do_vs_prog( struct brw_context *brw, static void brw_upload_vs_prog(struct brw_context *brw) { + GLcontext *ctx = &brw->intel.ctx; struct brw_vs_prog_key key; struct brw_vertex_program *vp = (struct brw_vertex_program *)brw->vertex_program; @@ -97,14 +98,9 @@ static void brw_upload_vs_prog(struct brw_context *brw) * the inputs it asks for, whether they are varying or not. */ key.program_string_id = vp->id; - key.nr_userclip = brw_count_bits(brw->attribs.Transform->ClipPlanesEnabled); - key.copy_edgeflag = (brw->attribs.Polygon->FrontMode != GL_FILL || - brw->attribs.Polygon->BackMode != GL_FILL); - - /* BRW_NEW_METAOPS - */ - if (brw->metaops.active) - key.know_w_is_one = 1; + key.nr_userclip = brw_count_bits(ctx->Transform.ClipPlanesEnabled); + key.copy_edgeflag = (ctx->Polygon.FrontMode != GL_FILL || + ctx->Polygon.BackMode != GL_FILL); /* Make an early check for the key. */ @@ -123,7 +119,7 @@ static void brw_upload_vs_prog(struct brw_context *brw) const struct brw_tracked_state brw_vs_prog = { .dirty = { .mesa = _NEW_TRANSFORM | _NEW_POLYGON, - .brw = BRW_NEW_VERTEX_PROGRAM | BRW_NEW_METAOPS, + .brw = BRW_NEW_VERTEX_PROGRAM, .cache = 0 }, .prepare = brw_upload_vs_prog diff --git a/src/mesa/drivers/dri/i965/brw_vs.h b/src/mesa/drivers/dri/i965/brw_vs.h index 22388ec99d..99d0e93722 100644 --- a/src/mesa/drivers/dri/i965/brw_vs.h +++ b/src/mesa/drivers/dri/i965/brw_vs.h @@ -43,7 +43,6 @@ struct brw_vs_prog_key { GLuint program_string_id; GLuint nr_userclip:4; GLuint copy_edgeflag:1; - GLuint know_w_is_one:1; GLuint pad:26; }; diff --git a/src/mesa/drivers/dri/i965/brw_vs_constval.c b/src/mesa/drivers/dri/i965/brw_vs_constval.c index 6fbac02de6..9977677fd7 100644 --- a/src/mesa/drivers/dri/i965/brw_vs_constval.c +++ b/src/mesa/drivers/dri/i965/brw_vs_constval.c @@ -168,6 +168,7 @@ static GLuint get_input_size(struct brw_context *brw, */ static void calc_wm_input_sizes( struct brw_context *brw ) { + GLcontext *ctx = &brw->intel.ctx; /* BRW_NEW_VERTEX_PROGRAM */ struct brw_vertex_program *vp = (struct brw_vertex_program *)brw->vertex_program; @@ -179,7 +180,7 @@ static void calc_wm_input_sizes( struct brw_context *brw ) memset(&t, 0, sizeof(t)); /* _NEW_LIGHT */ - if (brw->attribs.Light->Model.TwoSide) + if (ctx->Light.Model.TwoSide) t.twoside = 1; for (i = 0; i < VERT_ATTRIB_MAX; i++) diff --git a/src/mesa/drivers/dri/i965/brw_vs_emit.c b/src/mesa/drivers/dri/i965/brw_vs_emit.c index 174331a765..235f826812 100644 --- a/src/mesa/drivers/dri/i965/brw_vs_emit.c +++ b/src/mesa/drivers/dri/i965/brw_vs_emit.c @@ -871,21 +871,15 @@ static void emit_vertex_write( struct brw_vs_compile *c) } /* Build ndc coords */ - if (!c->key.know_w_is_one) { - ndc = get_tmp(c); - emit_math1(c, BRW_MATH_FUNCTION_INV, ndc, brw_swizzle1(pos, 3), BRW_MATH_PRECISION_FULL); - brw_MUL(p, brw_writemask(ndc, WRITEMASK_XYZ), pos, ndc); - } - else { - ndc = pos; - } + ndc = get_tmp(c); + emit_math1(c, BRW_MATH_FUNCTION_INV, ndc, brw_swizzle1(pos, 3), BRW_MATH_PRECISION_FULL); + brw_MUL(p, brw_writemask(ndc, WRITEMASK_XYZ), pos, ndc); /* Update the header for point size, user clipping flags, and -ve rhw * workaround. */ if ((c->prog_data.outputs_written & (1<<VERT_RESULT_PSIZ)) || - c->key.nr_userclip || - (!BRW_IS_G4X(p->brw) && !c->key.know_w_is_one)) + c->key.nr_userclip || !BRW_IS_G4X(p->brw)) { struct brw_reg header1 = retype(get_tmp(c), BRW_REGISTER_TYPE_UD); GLuint i; @@ -916,7 +910,7 @@ static void emit_vertex_write( struct brw_vs_compile *c) * Later, clipping will detect ucp[6] and ensure the primitive is * clipped against all fixed planes. */ - if (!BRW_IS_G4X(p->brw) && !c->key.know_w_is_one) { + if (!BRW_IS_G4X(p->brw)) { brw_CMP(p, vec8(brw_null_reg()), BRW_CONDITIONAL_L, @@ -960,36 +954,27 @@ static void emit_vertex_write( struct brw_vs_compile *c) } +/** + * Called after code generation to resolve subroutine calls and the + * END instruction. + * \param end_inst points to brw code for END instruction + * \param last_inst points to last instruction emitted before vertex write + */ static void -post_vs_emit( struct brw_vs_compile *c, struct brw_instruction *end_inst ) +post_vs_emit( struct brw_vs_compile *c, + struct brw_instruction *end_inst, + struct brw_instruction *last_inst ) { - GLuint nr_insns = c->vp->program.Base.NumInstructions; - GLuint insn, target_insn; - struct prog_instruction *inst1, *inst2; - struct brw_instruction *brw_inst1, *brw_inst2; - int offset; - for (insn = 0; insn < nr_insns; insn++) { - inst1 = &c->vp->program.Base.Instructions[insn]; - brw_inst1 = inst1->Data; - switch (inst1->Opcode) { - case OPCODE_CAL: - case OPCODE_BRA: - target_insn = inst1->BranchTarget; - inst2 = &c->vp->program.Base.Instructions[target_insn]; - brw_inst2 = inst2->Data; - offset = brw_inst2 - brw_inst1; - brw_set_src1(brw_inst1, brw_imm_d(offset*16)); - break; - case OPCODE_END: - offset = end_inst - brw_inst1; - brw_set_src1(brw_inst1, brw_imm_d(offset*16)); - break; - default: - break; - } - } + GLint offset; + + brw_resolve_cals(&c->func); + + /* patch up the END code to jump past subroutines, etc */ + offset = last_inst - end_inst; + brw_set_src1(end_inst, brw_imm_d(offset * 16)); } + /* Emit the fragment program instructions here. */ void brw_vs_emit(struct brw_vs_compile *c ) @@ -998,7 +983,8 @@ void brw_vs_emit(struct brw_vs_compile *c ) struct brw_compile *p = &c->func; GLuint nr_insns = c->vp->program.Base.NumInstructions; GLuint insn, if_insn = 0; - struct brw_instruction *end_inst; + GLuint end_offset = 0; + struct brw_instruction *end_inst, *last_inst; struct brw_instruction *if_inst[MAX_IFSN]; struct brw_indirect stack_index = brw_indirect(0, 0); @@ -1041,7 +1027,6 @@ void brw_vs_emit(struct brw_vs_compile *c ) /* Get argument regs. SWZ is special and does this itself. */ - inst->Data = &p->store[p->nr_insn]; if (inst->Opcode != OPCODE_SWZ) for (i = 0; i < 3; i++) { struct prog_src_register *src = &inst->SrcReg[i]; @@ -1209,7 +1194,7 @@ void brw_vs_emit(struct brw_vs_compile *c ) brw_set_access_mode(p, BRW_ALIGN_16); brw_ADD(p, get_addr_reg(stack_index), get_addr_reg(stack_index), brw_imm_d(4)); - inst->Data = &p->store[p->nr_insn]; + brw_save_call(p, inst->Comment, p->nr_insn); brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16)); break; case OPCODE_RET: @@ -1218,14 +1203,23 @@ void brw_vs_emit(struct brw_vs_compile *c ) brw_set_access_mode(p, BRW_ALIGN_1); brw_MOV(p, brw_ip_reg(), deref_1d(stack_index, 0)); brw_set_access_mode(p, BRW_ALIGN_16); + break; case OPCODE_END: + end_offset = p->nr_insn; + /* this instruction will get patched later to jump past subroutine + * code, etc. + */ brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16)); break; case OPCODE_PRINT: + /* no-op */ + break; case OPCODE_BGNSUB: + brw_save_label(p, inst->Comment, p->nr_insn); + break; case OPCODE_ENDSUB: - /* no-op instructions */ - break; + /* no-op */ + break; default: _mesa_problem(NULL, "Unsupported opcode %i (%s) in vertex shader", inst->Opcode, inst->Opcode < MAX_OPCODE ? @@ -1263,9 +1257,11 @@ void brw_vs_emit(struct brw_vs_compile *c ) release_tmps(c); } - end_inst = &p->store[p->nr_insn]; + end_inst = &p->store[end_offset]; + last_inst = &p->store[p->nr_insn]; + + /* The END instruction will be patched to jump to this code */ emit_vertex_write(c); - post_vs_emit(c, end_inst); - for (insn = 0; insn < nr_insns; insn++) - c->vp->program.Base.Instructions[insn].Data = NULL; + + post_vs_emit(c, end_inst, last_inst); } diff --git a/src/mesa/drivers/dri/i965/brw_vs_state.c b/src/mesa/drivers/dri/i965/brw_vs_state.c index 942581696d..1a63766ea1 100644 --- a/src/mesa/drivers/dri/i965/brw_vs_state.c +++ b/src/mesa/drivers/dri/i965/brw_vs_state.c @@ -49,6 +49,8 @@ struct brw_vs_unit_key { static void vs_unit_populate_key(struct brw_context *brw, struct brw_vs_unit_key *key) { + GLcontext *ctx = &brw->intel.ctx; + memset(key, 0, sizeof(*key)); /* CACHE_NEW_VS_PROG */ @@ -61,7 +63,7 @@ vs_unit_populate_key(struct brw_context *brw, struct brw_vs_unit_key *key) key->urb_size = brw->urb.vsize; /* BRW_NEW_CURBE_OFFSETS, _NEW_TRANSFORM */ - if (brw->attribs.Transform->ClipPlanesEnabled) { + if (ctx->Transform.ClipPlanesEnabled) { /* Note that we read in the userclip planes as well, hence * clip_start: */ diff --git a/src/mesa/drivers/dri/i965/brw_vtbl.c b/src/mesa/drivers/dri/i965/brw_vtbl.c index f7293ef423..b501a59ccd 100644 --- a/src/mesa/drivers/dri/i965/brw_vtbl.c +++ b/src/mesa/drivers/dri/i965/brw_vtbl.c @@ -23,14 +23,12 @@ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - **********************************************************************/ - /* - * Authors: - * Keith Whitwell <keith@tungstengraphics.com> - */ - - +**********************************************************************/ +/* + * Authors: + * Keith Whitwell <keith@tungstengraphics.com> + */ #include "main/glheader.h" #include "main/mtypes.h" @@ -44,12 +42,11 @@ #include "brw_context.h" #include "brw_defines.h" #include "brw_state.h" - #include "brw_draw.h" #include "brw_state.h" #include "brw_fallback.h" #include "brw_vs.h" -#include <stdarg.h> + static void dri_bo_release(dri_bo **bo) @@ -58,21 +55,22 @@ dri_bo_release(dri_bo **bo) *bo = NULL; } -/* called from intelDestroyContext() + +/** + * called from intelDestroyContext() */ static void brw_destroy_context( struct intel_context *intel ) { struct brw_context *brw = brw_context(&intel->ctx); int i; - brw_destroy_metaops(brw); brw_destroy_state(brw); brw_draw_destroy( brw ); brw_FrameBufferTexDestroy( brw ); for (i = 0; i < brw->state.nr_draw_regions; i++) - intel_region_release(&brw->state.draw_regions[i]); + intel_region_release(&brw->state.draw_regions[i]); brw->state.nr_draw_regions = 0; intel_region_release(&brw->state.depth_region); @@ -99,37 +97,46 @@ static void brw_destroy_context( struct intel_context *intel ) dri_bo_release(&brw->cc.vp_bo); } -/* called from intelDrawBuffer() + +/** + * called from intelDrawBuffer() */ static void brw_set_draw_region( struct intel_context *intel, - struct intel_region *draw_regions[], - struct intel_region *depth_region, - GLuint num_regions) + struct intel_region *draw_regions[], + struct intel_region *depth_region, + GLuint num_regions) { struct brw_context *brw = brw_context(&intel->ctx); int i; + + /* release old color/depth regions */ if (brw->state.depth_region != depth_region) brw->state.dirty.brw |= BRW_NEW_DEPTH_BUFFER; for (i = 0; i < brw->state.nr_draw_regions; i++) intel_region_release(&brw->state.draw_regions[i]); intel_region_release(&brw->state.depth_region); + + /* reference new color/depth regions */ for (i = 0; i < num_regions; i++) intel_region_reference(&brw->state.draw_regions[i], draw_regions[i]); intel_region_reference(&brw->state.depth_region, depth_region); brw->state.nr_draw_regions = num_regions; } -/* called from intel_batchbuffer_flush and children before sending a + +/** + * called from intel_batchbuffer_flush and children before sending a * batchbuffer off. */ static void brw_finish_batch(struct intel_context *intel) { struct brw_context *brw = brw_context(&intel->ctx); - brw_emit_query_end(brw); } -/* called from intelFlushBatchLocked + +/** + * called from intelFlushBatchLocked */ static void brw_new_batch( struct intel_context *intel ) { @@ -160,22 +167,21 @@ static void brw_new_batch( struct intel_context *intel ) } } -static void brw_note_fence( struct intel_context *intel, - GLuint fence ) + +static void brw_note_fence( struct intel_context *intel, GLuint fence ) { brw_context(&intel->ctx)->state.dirty.brw |= BRW_NEW_FENCE; } - + + static void brw_note_unlock( struct intel_context *intel ) { struct brw_context *brw = brw_context(&intel->ctx); - brw_state_cache_check_size(brw); } -void brw_do_flush( struct brw_context *brw, - GLuint flags ) +void brw_do_flush( struct brw_context *brw, GLuint flags ) { struct brw_mi_flush flush; memset(&flush, 0, sizeof(flush)); @@ -185,8 +191,7 @@ void brw_do_flush( struct brw_context *brw, } -static void brw_emit_flush( struct intel_context *intel, - GLuint unused ) +static void brw_emit_flush( struct intel_context *intel, GLuint unused ) { brw_do_flush(brw_context(&intel->ctx), BRW_FLUSH_STATE_CACHE|BRW_FLUSH_READ_CACHE); @@ -206,6 +211,7 @@ static GLuint brw_flush_cmd( void ) return *(GLuint *)&flush; } + static void brw_invalidate_state( struct intel_context *intel, GLuint new_state ) { /* nothing */ @@ -215,14 +221,14 @@ static void brw_invalidate_state( struct intel_context *intel, GLuint new_state void brwInitVtbl( struct brw_context *brw ) { brw->intel.vtbl.check_vertex_size = 0; - brw->intel.vtbl.emit_state = 0; - brw->intel.vtbl.reduced_primitive_state = 0; + brw->intel.vtbl.emit_state = 0; + brw->intel.vtbl.reduced_primitive_state = 0; brw->intel.vtbl.render_start = 0; - brw->intel.vtbl.update_texture_state = 0; + brw->intel.vtbl.update_texture_state = 0; - brw->intel.vtbl.invalidate_state = brw_invalidate_state; - brw->intel.vtbl.note_fence = brw_note_fence; - brw->intel.vtbl.note_unlock = brw_note_unlock; + brw->intel.vtbl.invalidate_state = brw_invalidate_state; + brw->intel.vtbl.note_fence = brw_note_fence; + brw->intel.vtbl.note_unlock = brw_note_unlock; brw->intel.vtbl.new_batch = brw_new_batch; brw->intel.vtbl.finish_batch = brw_finish_batch; brw->intel.vtbl.destroy = brw_destroy_context; @@ -231,4 +237,3 @@ void brwInitVtbl( struct brw_context *brw ) brw->intel.vtbl.emit_flush = brw_emit_flush; brw->intel.vtbl.debug_batch = brw_debug_batch; } - diff --git a/src/mesa/drivers/dri/i965/brw_wm.c b/src/mesa/drivers/dri/i965/brw_wm.c index c50b0d2dd9..4501e0472f 100644 --- a/src/mesa/drivers/dri/i965/brw_wm.c +++ b/src/mesa/drivers/dri/i965/brw_wm.c @@ -80,6 +80,53 @@ GLuint brw_wm_is_scalar_result( GLuint opcode ) } +/** + * Do GPU code generation for non-GLSL shader. non-GLSL shaders have + * no flow control instructions so we can more readily do SSA-style + * optimizations. + */ +static void +brw_wm_non_glsl_emit(struct brw_context *brw, struct brw_wm_compile *c) +{ + /* Augment fragment program. Add instructions for pre- and + * post-fragment-program tasks such as interpolation and fogging. + */ + brw_wm_pass_fp(c); + + /* Translate to intermediate representation. Build register usage + * chains. + */ + brw_wm_pass0(c); + + /* Dead code removal. + */ + brw_wm_pass1(c); + + /* Register allocation. + */ + c->grf_limit = BRW_WM_MAX_GRF / 2; + + brw_wm_pass2(c); + + c->prog_data.total_grf = c->max_wm_grf; + if (c->last_scratch) { + c->prog_data.total_scratch = c->last_scratch + 0x40; + } + else { + c->prog_data.total_scratch = 0; + } + + /* Emit GEN4 code. + */ + brw_wm_emit(c); +} + + +/** + * All Mesa program -> GPU code generation goes through this function. + * Depending on the instructions used (i.e. flow control instructions) + * we'll use one of two code generators. + */ static void do_wm_prog( struct brw_context *brw, struct brw_fragment_program *fp, struct brw_wm_prog_key *key) @@ -90,52 +137,29 @@ static void do_wm_prog( struct brw_context *brw, c = brw->wm.compile_data; if (c == NULL) { - brw->wm.compile_data = calloc(1, sizeof(*brw->wm.compile_data)); - c = brw->wm.compile_data; + brw->wm.compile_data = calloc(1, sizeof(*brw->wm.compile_data)); + c = brw->wm.compile_data; } else { - memset(c, 0, sizeof(*brw->wm.compile_data)); + memset(c, 0, sizeof(*brw->wm.compile_data)); } memcpy(&c->key, key, sizeof(*key)); c->fp = fp; c->env_param = brw->intel.ctx.FragmentProgram.Parameters; - brw_init_compile(brw, &c->func); + brw_init_compile(brw, &c->func); + + /* + * Shader which use GLSL features such as flow control are handled + * differently from "simple" shaders. + */ if (brw_wm_is_glsl(&c->fp->program)) { - brw_wm_glsl_emit(brw, c); - } else { - /* Augment fragment program. Add instructions for pre- and - * post-fragment-program tasks such as interpolation and fogging. - */ - brw_wm_pass_fp(c); - - /* Translate to intermediate representation. Build register usage - * chains. - */ - brw_wm_pass0(c); - - /* Dead code removal. - */ - brw_wm_pass1(c); - - /* Register allocation. - */ - c->grf_limit = BRW_WM_MAX_GRF/2; - - brw_wm_pass2(c); - - c->prog_data.total_grf = c->max_wm_grf; - if (c->last_scratch) { - c->prog_data.total_scratch = - c->last_scratch + 0x40; - } else { - c->prog_data.total_scratch = 0; - } - - /* Emit GEN4 code. - */ - brw_wm_emit(c); + brw_wm_glsl_emit(brw, c); } + else { + brw_wm_non_glsl_emit(brw, c); + } + if (INTEL_DEBUG & DEBUG_WM) fprintf(stderr, "\n"); @@ -157,6 +181,7 @@ static void do_wm_prog( struct brw_context *brw, static void brw_wm_populate_key( struct brw_context *brw, struct brw_wm_prog_key *key ) { + GLcontext *ctx = &brw->intel.ctx; /* BRW_NEW_FRAGMENT_PROGRAM */ struct brw_fragment_program *fp = (struct brw_fragment_program *)brw->fragment_program; @@ -170,51 +195,50 @@ static void brw_wm_populate_key( struct brw_context *brw, */ /* _NEW_COLOR */ if (fp->program.UsesKill || - brw->attribs.Color->AlphaEnabled) + ctx->Color.AlphaEnabled) lookup |= IZ_PS_KILL_ALPHATEST_BIT; if (fp->program.Base.OutputsWritten & (1<<FRAG_RESULT_DEPR)) lookup |= IZ_PS_COMPUTES_DEPTH_BIT; /* _NEW_DEPTH */ - if (brw->attribs.Depth->Test) + if (ctx->Depth.Test) lookup |= IZ_DEPTH_TEST_ENABLE_BIT; - if (brw->attribs.Depth->Test && - brw->attribs.Depth->Mask) /* ?? */ + if (ctx->Depth.Test && + ctx->Depth.Mask) /* ?? */ lookup |= IZ_DEPTH_WRITE_ENABLE_BIT; /* _NEW_STENCIL */ - if (brw->attribs.Stencil->Enabled) { + if (ctx->Stencil.Enabled) { lookup |= IZ_STENCIL_TEST_ENABLE_BIT; - if (brw->attribs.Stencil->WriteMask[0] || - (brw->attribs.Stencil->_TestTwoSide && - brw->attribs.Stencil->WriteMask[1])) + if (ctx->Stencil.WriteMask[0] || + ctx->Stencil.WriteMask[ctx->Stencil._BackFace]) lookup |= IZ_STENCIL_WRITE_ENABLE_BIT; } line_aa = AA_NEVER; /* _NEW_LINE, _NEW_POLYGON, BRW_NEW_REDUCED_PRIMITIVE */ - if (brw->attribs.Line->SmoothFlag) { + if (ctx->Line.SmoothFlag) { if (brw->intel.reduced_primitive == GL_LINES) { line_aa = AA_ALWAYS; } else if (brw->intel.reduced_primitive == GL_TRIANGLES) { - if (brw->attribs.Polygon->FrontMode == GL_LINE) { + if (ctx->Polygon.FrontMode == GL_LINE) { line_aa = AA_SOMETIMES; - if (brw->attribs.Polygon->BackMode == GL_LINE || - (brw->attribs.Polygon->CullFlag && - brw->attribs.Polygon->CullFaceMode == GL_BACK)) + if (ctx->Polygon.BackMode == GL_LINE || + (ctx->Polygon.CullFlag && + ctx->Polygon.CullFaceMode == GL_BACK)) line_aa = AA_ALWAYS; } - else if (brw->attribs.Polygon->BackMode == GL_LINE) { + else if (ctx->Polygon.BackMode == GL_LINE) { line_aa = AA_SOMETIMES; - if ((brw->attribs.Polygon->CullFlag && - brw->attribs.Polygon->CullFaceMode == GL_FRONT)) + if ((ctx->Polygon.CullFlag && + ctx->Polygon.CullFaceMode == GL_FRONT)) line_aa = AA_ALWAYS; } } @@ -229,20 +253,25 @@ static void brw_wm_populate_key( struct brw_context *brw, key->projtex_mask = brw->wm.input_size_masks[4-1] >> (FRAG_ATTRIB_TEX0 - FRAG_ATTRIB_WPOS); /* _NEW_LIGHT */ - key->flat_shade = (brw->attribs.Light->ShadeModel == GL_FLAT); + key->flat_shade = (ctx->Light.ShadeModel == GL_FLAT); /* _NEW_TEXTURE */ for (i = 0; i < BRW_MAX_TEX_UNIT; i++) { - const struct gl_texture_unit *unit = &brw->attribs.Texture->Unit[i]; - const struct gl_texture_object *t = unit->_Current; + const struct gl_texture_unit *unit = &ctx->Texture.Unit[i]; if (unit->_ReallyEnabled) { - if (t->Image[0][t->BaseLevel]->InternalFormat == GL_YCBCR_MESA) { - key->yuvtex_mask |= 1<<i; - if (t->Image[0][t->BaseLevel]->TexFormat->MesaFormat == - MESA_FORMAT_YCBCR) - key->yuvtex_swap_mask |= 1<< i; + const struct gl_texture_object *t = unit->_Current; + const struct gl_texture_image *img = t->Image[0][t->BaseLevel]; + if (img->InternalFormat == GL_YCBCR_MESA) { + key->yuvtex_mask |= 1 << i; + if (img->TexFormat->MesaFormat == MESA_FORMAT_YCBCR) + key->yuvtex_swap_mask |= 1 << i; } + + key->tex_swizzles[i] = t->_Swizzle; + } + else { + key->tex_swizzles[i] = SWIZZLE_NOOP; } } @@ -300,8 +329,6 @@ static void brw_prepare_wm_prog(struct brw_context *brw) } -/* See brw_wm.c: - */ const struct brw_tracked_state brw_wm_prog = { .dirty = { .mesa = (_NEW_COLOR | diff --git a/src/mesa/drivers/dri/i965/brw_wm.h b/src/mesa/drivers/dri/i965/brw_wm.h index f990bde839..582031f5f6 100644 --- a/src/mesa/drivers/dri/i965/brw_wm.h +++ b/src/mesa/drivers/dri/i965/brw_wm.h @@ -69,7 +69,8 @@ struct brw_wm_prog_key { GLuint shadowtex_mask:16; GLuint yuvtex_mask:16; GLuint yuvtex_swap_mask:16; /* UV swaped */ - // GLuint pad1:16; + + GLuint tex_swizzles[BRW_MAX_TEX_UNIT]; GLuint program_string_id:32; GLuint origin_x, origin_y; @@ -280,4 +281,6 @@ void brw_wm_lookup_iz( GLuint line_aa, GLboolean brw_wm_is_glsl(const struct gl_fragment_program *fp); void brw_wm_glsl_emit(struct brw_context *brw, struct brw_wm_compile *c); + + #endif diff --git a/src/mesa/drivers/dri/i965/brw_wm_fp.c b/src/mesa/drivers/dri/i965/brw_wm_fp.c index 6df2c95d80..8b5ccdba93 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_fp.c +++ b/src/mesa/drivers/dri/i965/brw_wm_fp.c @@ -111,6 +111,12 @@ static struct prog_src_register src_swizzle1( struct prog_src_register reg, int return src_swizzle(reg, x, x, x, x); } +static struct prog_src_register src_swizzle4( struct prog_src_register reg, uint swizzle ) +{ + reg.Swizzle = swizzle; + return reg; +} + /*********************************************************************** * Dest regs @@ -177,7 +183,6 @@ static struct prog_instruction *emit_insn(struct brw_wm_compile *c, { struct prog_instruction *inst = get_fp_inst(c); *inst = *inst0; - inst->Data = (void *)inst0; return inst; } @@ -554,12 +559,19 @@ static void precalc_lit( struct brw_wm_compile *c, } } + +/** + * Some TEX instructions require extra code, cube map coordinate + * normalization, or coordinate scaling for RECT textures, etc. + * This function emits those extra instructions and the TEX + * instruction itself. + */ static void precalc_tex( struct brw_wm_compile *c, const struct prog_instruction *inst ) { struct prog_src_register coord; struct prog_dst_register tmpcoord; - GLuint unit = c->fp->program.Base.SamplerUnits[inst->TexSrcUnit]; + const GLuint unit = c->fp->program.Base.SamplerUnits[inst->TexSrcUnit]; if (inst->TexSrcTarget == TEXTURE_CUBE_INDEX) { struct prog_instruction *out; @@ -569,9 +581,11 @@ static void precalc_tex( struct brw_wm_compile *c, struct prog_src_register tmp1src = src_reg_from_dst(tmp1); struct prog_src_register src0 = inst->SrcReg[0]; + /* find longest component of coord vector and normalize it */ tmpcoord = get_temp(c); coord = src_reg_from_dst(tmpcoord); + /* tmpcoord = src0 (i.e.: coord = src0) */ out = emit_op(c, OPCODE_MOV, tmpcoord, 0, 0, 0, @@ -581,6 +595,7 @@ static void precalc_tex( struct brw_wm_compile *c, out->SrcReg[0].NegateBase = 0; out->SrcReg[0].Abs = 1; + /* tmp0 = MAX(coord.X, coord.Y) */ emit_op(c, OPCODE_MAX, tmp0, 0, 0, 0, @@ -588,6 +603,7 @@ static void precalc_tex( struct brw_wm_compile *c, src_swizzle1(coord, Y), src_undef()); + /* tmp1 = MAX(tmp0, coord.Z) */ emit_op(c, OPCODE_MAX, tmp1, 0, 0, 0, @@ -595,6 +611,7 @@ static void precalc_tex( struct brw_wm_compile *c, src_swizzle1(coord, Z), src_undef()); + /* tmp0 = 1 / tmp1 */ emit_op(c, OPCODE_RCP, tmp0, 0, 0, 0, @@ -602,6 +619,7 @@ static void precalc_tex( struct brw_wm_compile *c, src_undef(), src_undef()); + /* tmpCoord = src0 * tmp0 */ emit_op(c, OPCODE_MUL, tmpcoord, 0, 0, 0, @@ -611,7 +629,8 @@ static void precalc_tex( struct brw_wm_compile *c, release_temp(c, tmp0); release_temp(c, tmp1); - } else if (inst->TexSrcTarget == TEXTURE_RECT_INDEX) { + } + else if (inst->TexSrcTarget == TEXTURE_RECT_INDEX) { struct prog_src_register scale = search_or_add_param5( c, STATE_INTERNAL, @@ -642,19 +661,9 @@ static void precalc_tex( struct brw_wm_compile *c, * conversion requires allocating a temporary variable which we * don't have the facility to do that late in the compilation. */ - if (!(c->key.yuvtex_mask & (1<<unit))) { - emit_op(c, - OPCODE_TEX, - inst->DstReg, - inst->SaturateMode, - unit, - inst->TexSrcTarget, - coord, - src_undef(), - src_undef()); - } - else { - GLboolean swap_uv = c->key.yuvtex_swap_mask & (1<<unit); + if (c->key.yuvtex_mask & (1 << unit)) { + /* convert ycbcr to RGBA */ + GLboolean swap_uv = c->key.yuvtex_swap_mask & (1<<unit); /* CONST C0 = { -.5, -.0625, -.5, 1.164 } @@ -734,6 +743,31 @@ static void precalc_tex( struct brw_wm_compile *c, release_temp(c, tmp); } + else { + /* ordinary RGBA tex instruction */ + emit_op(c, + OPCODE_TEX, + inst->DstReg, + inst->SaturateMode, + unit, + inst->TexSrcTarget, + coord, + src_undef(), + src_undef()); + } + + /* For GL_EXT_texture_swizzle: */ + if (c->key.tex_swizzles[unit] != SWIZZLE_NOOP) { + /* swizzle the result of the TEX instruction */ + struct prog_src_register tmpsrc = src_reg_from_dst(inst->DstReg); + emit_op(c, OPCODE_SWZ, + inst->DstReg, + SATURATE_OFF, /* saturate already done above */ + 0, 0, /* tex unit, target N/A */ + src_swizzle4(tmpsrc, c->key.tex_swizzles[unit]), + src_undef(), + src_undef()); + } if ((inst->TexSrcTarget == TEXTURE_RECT_INDEX) || (inst->TexSrcTarget == TEXTURE_CUBE_INDEX)) @@ -907,6 +941,11 @@ static void print_insns( const struct prog_instruction *insn, } } + +/** + * Initial pass for fragment program code generation. + * This function is used by both the GLSL and non-GLSL paths. + */ void brw_wm_pass_fp( struct brw_wm_compile *c ) { struct brw_fragment_program *fp = c->fp; @@ -923,15 +962,19 @@ void brw_wm_pass_fp( struct brw_wm_compile *c ) c->pixel_w = src_undef(); c->nr_fp_insns = 0; - /* Emit preamble instructions: + /* Emit preamble instructions. This is where special instructions such as + * WM_CINTERP, WM_LINTERP, WM_PINTERP and WM_WPOSXY are emitted to + * compute shader inputs from varying vars. */ - - for (insn = 0; insn < fp->program.Base.NumInstructions; insn++) { const struct prog_instruction *inst = &fp->program.Base.Instructions[insn]; validate_src_regs(c, inst); validate_dst_regs(c, inst); } + + /* Loop over all instructions doing assorted simplifications and + * transformations. + */ for (insn = 0; insn < fp->program.Base.NumInstructions; insn++) { const struct prog_instruction *inst = &fp->program.Base.Instructions[insn]; struct prog_instruction *out; @@ -940,7 +983,6 @@ void brw_wm_pass_fp( struct brw_wm_compile *c ) * necessary: */ - switch (inst->Opcode) { case OPCODE_SWZ: out = emit_insn(c, inst); diff --git a/src/mesa/drivers/dri/i965/brw_wm_glsl.c b/src/mesa/drivers/dri/i965/brw_wm_glsl.c index 8fd776ac39..d99aa372a9 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_glsl.c +++ b/src/mesa/drivers/dri/i965/brw_wm_glsl.c @@ -8,12 +8,17 @@ enum _subroutine { SUB_NOISE1, SUB_NOISE2, SUB_NOISE3, SUB_NOISE4 }; -/* Only guess, need a flag in gl_fragment_program later */ + +/** + * Determine if the given fragment program uses GLSL features such + * as flow conditionals, loops, subroutines. + * Some GLSL shaders may use these features, others might not. + */ GLboolean brw_wm_is_glsl(const struct gl_fragment_program *fp) { int i; for (i = 0; i < fp->Base.NumInstructions; i++) { - struct prog_instruction *inst = &fp->Base.Instructions[i]; + const struct prog_instruction *inst = &fp->Base.Instructions[i]; switch (inst->Opcode) { case OPCODE_IF: case OPCODE_TRUNC: @@ -176,13 +181,15 @@ static struct brw_reg get_src_reg(struct brw_wm_compile *c, src->NegateBase, src->Abs); } -/* Subroutines are minimal support for resusable instruction sequences. - They are implemented as simply as possible to minimise overhead: there - is no explicit support for communication between the caller and callee - other than saving the return address in a temporary register, nor is - there any automatic local storage. This implies that great care is - required before attempting reentrancy or any kind of nested - subroutine invocations. */ +/** + * Subroutines are minimal support for resusable instruction sequences. + * They are implemented as simply as possible to minimise overhead: there + * is no explicit support for communication between the caller and callee + * other than saving the return address in a temporary register, nor is + * there any automatic local storage. This implies that great care is + * required before attempting reentrancy or any kind of nested + * subroutine invocations. + */ static void invoke_subroutine( struct brw_wm_compile *c, enum _subroutine subroutine, void (*emit)( struct brw_wm_compile * ) ) @@ -319,11 +326,10 @@ static void emit_pixel_xy(struct brw_wm_compile *c, stride(suboffset(r1_uw, 5), 2, 4, 0), brw_imm_v(0x11001100)); } - } static void emit_delta_xy(struct brw_wm_compile *c, - struct prog_instruction *inst) + struct prog_instruction *inst) { struct brw_reg r1 = brw_vec1_grf(1, 0); struct brw_reg dst0, dst1, src0, src1; @@ -351,10 +357,8 @@ static void emit_delta_xy(struct brw_wm_compile *c, negate(suboffset(r1,1))); } - } - static void fire_fb_write( struct brw_wm_compile *c, GLuint base_reg, GLuint nr, @@ -397,31 +401,31 @@ static void emit_fb_write(struct brw_wm_compile *c, */ if (c->key.aa_dest_stencil_reg) nr += 1; - { - brw_push_insn_state(p); - for (channel = 0; channel < 4; channel++) { - src0 = get_src_reg(c, &inst->SrcReg[0], channel, 1); - /* mov (8) m2.0<1>:ud r28.0<8;8,1>:ud { Align1 } */ - /* mov (8) m6.0<1>:ud r29.0<8;8,1>:ud { Align1 SecHalf } */ - brw_MOV(p, brw_message_reg(nr + channel), src0); - } - /* skip over the regs populated above: */ - nr += 8; - brw_pop_insn_state(p); + + brw_push_insn_state(p); + for (channel = 0; channel < 4; channel++) { + src0 = get_src_reg(c, &inst->SrcReg[0], channel, 1); + /* mov (8) m2.0<1>:ud r28.0<8;8,1>:ud { Align1 } */ + /* mov (8) m6.0<1>:ud r29.0<8;8,1>:ud { Align1 SecHalf } */ + brw_MOV(p, brw_message_reg(nr + channel), src0); + } + /* skip over the regs populated above: */ + nr += 8; + brw_pop_insn_state(p); + + if (c->key.source_depth_to_render_target) { + if (c->key.computes_depth) { + src0 = get_src_reg(c, &inst->SrcReg[2], 2, 1); + brw_MOV(p, brw_message_reg(nr), src0); + } + else { + src0 = get_src_reg(c, &inst->SrcReg[1], 1, 1); + brw_MOV(p, brw_message_reg(nr), src0); + } + + nr += 2; } - if (c->key.source_depth_to_render_target) - { - if (c->key.computes_depth) { - src0 = get_src_reg(c, &inst->SrcReg[2], 2, 1); - brw_MOV(p, brw_message_reg(nr), src0); - } else { - src0 = get_src_reg(c, &inst->SrcReg[1], 1, 1); - brw_MOV(p, brw_message_reg(nr), src0); - } - - nr += 2; - } target = inst->Sampler >> 1; eot = inst->Sampler & 1; fire_fb_write(c, 0, nr, target, eot); @@ -1045,23 +1049,23 @@ static void emit_ddy(struct brw_wm_compile *c, brw_set_saturate(p, 0); } -static __inline struct brw_reg high_words( struct brw_reg reg ) +static INLINE struct brw_reg high_words( struct brw_reg reg ) { return stride( suboffset( retype( reg, BRW_REGISTER_TYPE_W ), 1 ), 0, 8, 2 ); } -static __inline struct brw_reg low_words( struct brw_reg reg ) +static INLINE struct brw_reg low_words( struct brw_reg reg ) { return stride( retype( reg, BRW_REGISTER_TYPE_W ), 0, 8, 2 ); } -static __inline struct brw_reg even_bytes( struct brw_reg reg ) +static INLINE struct brw_reg even_bytes( struct brw_reg reg ) { return stride( retype( reg, BRW_REGISTER_TYPE_B ), 0, 16, 2 ); } -static __inline struct brw_reg odd_bytes( struct brw_reg reg ) +static INLINE struct brw_reg odd_bytes( struct brw_reg reg ) { return stride( suboffset( retype( reg, BRW_REGISTER_TYPE_B ), 1 ), 0, 16, 2 ); @@ -1366,9 +1370,11 @@ static void emit_noise2( struct brw_wm_compile *c, release_tmps( c, mark ); } -/* The three-dimensional case is much like the one- and two- versions above, - but since the number of corners is rapidly growing we now pack 16 16-bit - hashes into each register to extract more parallelism from the EUs. */ +/** + * The three-dimensional case is much like the one- and two- versions above, + * but since the number of corners is rapidly growing we now pack 16 16-bit + * hashes into each register to extract more parallelism from the EUs. + */ static void noise3_sub( struct brw_wm_compile *c ) { struct brw_compile *p = &c->func; @@ -1670,13 +1676,15 @@ static void emit_noise3( struct brw_wm_compile *c, release_tmps( c, mark ); } -/* For the four-dimensional case, the little micro-optimisation benefits - we obtain by unrolling all the loops aren't worth the massive bloat it - now causes. Instead, we loop twice around performing a similar operation - to noise3, once for the w=0 cube and once for the w=1, with a bit more - code to glue it all together. */ -static void noise4_sub( struct brw_wm_compile *c ) { - +/** + * For the four-dimensional case, the little micro-optimisation benefits + * we obtain by unrolling all the loops aren't worth the massive bloat it + * now causes. Instead, we loop twice around performing a similar operation + * to noise3, once for the w=0 cube and once for the w=1, with a bit more + * code to glue it all together. + */ +static void noise4_sub( struct brw_wm_compile *c ) +{ struct brw_compile *p = &c->func; struct brw_reg param[ 4 ], x0y0, x0y1, x1y0, x1y1, /* gradients at four of the corners */ @@ -2244,28 +2252,12 @@ static void emit_tex(struct brw_wm_compile *c, brw_MOV(p, dst[3], brw_imm_f(1.0)); } +/** + * Resolve subroutine calls after code emit is done. + */ static void post_wm_emit( struct brw_wm_compile *c ) { - GLuint nr_insns = c->fp->program.Base.NumInstructions; - GLuint insn, target_insn; - struct prog_instruction *inst1, *inst2; - struct brw_instruction *brw_inst1, *brw_inst2; - int offset; - for (insn = 0; insn < nr_insns; insn++) { - inst1 = &c->fp->program.Base.Instructions[insn]; - brw_inst1 = inst1->Data; - switch (inst1->Opcode) { - case OPCODE_CAL: - target_insn = inst1->BranchTarget; - inst2 = &c->fp->program.Base.Instructions[target_insn]; - brw_inst2 = inst2->Data; - offset = brw_inst2 - brw_inst1; - brw_set_src1(brw_inst1, brw_imm_d(offset*16)); - break; - default: - break; - } - } + brw_resolve_cals(&c->func); } static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c) @@ -2285,10 +2277,6 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c) for (i = 0; i < c->nr_fp_insns; i++) { struct prog_instruction *inst = &c->prog_instructions[i]; - struct prog_instruction *orig_inst; - - if ((orig_inst = inst->Data) != 0) - orig_inst->Data = current_insn(p); if (inst->CondUpdate) brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); @@ -2446,7 +2434,10 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c) brw_ENDIF(p, if_inst[--if_insn]); break; case OPCODE_BGNSUB: + brw_save_label(p, inst->Comment, p->nr_insn); + break; case OPCODE_ENDSUB: + /* no-op */ break; case OPCODE_CAL: brw_push_insn_state(p); @@ -2456,8 +2447,7 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c) brw_set_access_mode(p, BRW_ALIGN_16); brw_ADD(p, get_addr_reg(stack_index), get_addr_reg(stack_index), brw_imm_d(4)); - orig_inst = inst->Data; - orig_inst->Data = &p->store[p->nr_insn]; + brw_save_call(&c->func, inst->Comment, p->nr_insn); brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16)); brw_pop_insn_state(p); break; @@ -2510,14 +2500,21 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c) brw_set_predicate_control(p, BRW_PREDICATE_NONE); } post_wm_emit(c); - for (i = 0; i < c->fp->program.Base.NumInstructions; i++) - c->fp->program.Base.Instructions[i].Data = NULL; } + +/** + * Do GPU code generation for shaders that use GLSL features such as + * flow control. Other shaders will be compiled with the + */ void brw_wm_glsl_emit(struct brw_context *brw, struct brw_wm_compile *c) { + /* initial instruction translation/simplification */ brw_wm_pass_fp(c); + + /* actual code generation */ brw_wm_emit_glsl(brw, c); + c->prog_data.total_grf = c->reg_index; c->prog_data.total_scratch = 0; } diff --git a/src/mesa/drivers/dri/i965/brw_wm_pass0.c b/src/mesa/drivers/dri/i965/brw_wm_pass0.c index fca7b7a50c..3ade6aa6c3 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_pass0.c +++ b/src/mesa/drivers/dri/i965/brw_wm_pass0.c @@ -51,6 +51,7 @@ static struct brw_wm_value *get_value( struct brw_wm_compile *c) return &c->vreg[c->nr_vreg++]; } +/** return pointer to a newly allocated instruction */ static struct brw_wm_instruction *get_instruction( struct brw_wm_compile *c ) { assert(c->nr_insns < BRW_WM_MAX_INSN); @@ -60,6 +61,7 @@ static struct brw_wm_instruction *get_instruction( struct brw_wm_compile *c ) /*********************************************************************** */ +/** Init the "undef" register */ static void pass0_init_undef( struct brw_wm_compile *c) { struct brw_wm_ref *ref = &c->undef_ref; @@ -69,6 +71,7 @@ static void pass0_init_undef( struct brw_wm_compile *c) ref->prevuse = NULL; } +/** Set a FP register to a value */ static void pass0_set_fpreg_value( struct brw_wm_compile *c, GLuint file, GLuint idx, @@ -83,6 +86,7 @@ static void pass0_set_fpreg_value( struct brw_wm_compile *c, c->pass0_fp_reg[file][idx][component] = ref; } +/** Set a FP register to a ref */ static void pass0_set_fpreg_ref( struct brw_wm_compile *c, GLuint file, GLuint idx, @@ -115,12 +119,13 @@ static const struct brw_wm_ref *get_param_ref( struct brw_wm_compile *c, ref->value = &c->creg[i/16]; ref->insn = 0; ref->prevuse = NULL; - + return ref; } } +/** Return a ref to a constant/literal value */ static const struct brw_wm_ref *get_const_ref( struct brw_wm_compile *c, const GLfloat *constval ) { @@ -142,7 +147,7 @@ static const struct brw_wm_ref *get_const_ref( struct brw_wm_compile *c, */ c->constref[i].constval = *constval; c->constref[i].ref = get_param_ref(c, constval); - + return c->constref[i].ref; } else { @@ -187,7 +192,7 @@ static const struct brw_wm_ref *pass0_get_reg( struct brw_wm_compile *c, /* There's something really hokey about parameters parsed in * arb programs - they all end up in here, whether they be - * state values, paramters or constants. This duplicates the + * state values, parameters or constants. This duplicates the * structure above & also seems to subvert the limits set for * each type of constant/param. */ @@ -198,7 +203,7 @@ static const struct brw_wm_ref *pass0_get_reg( struct brw_wm_compile *c, */ ref = get_const_ref(c, &plist->ParameterValues[idx][component]); break; - + case PROGRAM_STATE_VAR: case PROGRAM_UNIFORM: /* These may change from run to run: @@ -229,14 +234,13 @@ static const struct brw_wm_ref *pass0_get_reg( struct brw_wm_compile *c, - /*********************************************************************** * Straight translation to internal instruction format */ static void pass0_set_dst( struct brw_wm_compile *c, - struct brw_wm_instruction *out, - const struct prog_instruction *inst, + struct brw_wm_instruction *out, + const struct prog_instruction *inst, GLuint writemask ) { const struct prog_dst_register *dst = &inst->DstReg; @@ -245,18 +249,17 @@ static void pass0_set_dst( struct brw_wm_compile *c, for (i = 0; i < 4; i++) { if (writemask & (1<<i)) { out->dst[i] = get_value(c); - pass0_set_fpreg_value(c, dst->File, dst->Index, i, out->dst[i]); } } - + out->writemask = writemask; } static void pass0_set_dst_scalar( struct brw_wm_compile *c, - struct brw_wm_instruction *out, - const struct prog_instruction *inst, + struct brw_wm_instruction *out, + const struct prog_instruction *inst, GLuint writemask ) { if (writemask) { @@ -282,7 +285,6 @@ static void pass0_set_dst_scalar( struct brw_wm_compile *c, } - static const struct brw_wm_ref *get_fp_src_reg_ref( struct brw_wm_compile *c, struct prog_src_register src, GLuint i ) @@ -292,14 +294,13 @@ static const struct brw_wm_ref *get_fp_src_reg_ref( struct brw_wm_compile *c, static const GLfloat const_zero = 0.0; static const GLfloat const_one = 1.0; - if (component == SWIZZLE_ZERO) src_ref = get_const_ref(c, &const_zero); else if (component == SWIZZLE_ONE) src_ref = get_const_ref(c, &const_one); else src_ref = pass0_get_reg(c, src.File, src.Index, component); - + return src_ref; } @@ -311,19 +312,19 @@ static struct brw_wm_ref *get_new_ref( struct brw_wm_compile *c, { const struct brw_wm_ref *ref = get_fp_src_reg_ref(c, src, i); struct brw_wm_ref *newref = get_ref(c); - + newref->value = ref->value; newref->hw_reg = ref->hw_reg; - if (insn) { + if (insn) { newref->insn = insn - c->instruction; newref->prevuse = newref->value->lastuse; newref->value->lastuse = newref; } - if (src.NegateBase & (1<<i)) + if (src.NegateBase & (1<<i)) newref->hw_reg.negate ^= 1; - + if (src.Abs) { newref->hw_reg.negate = 0; newref->hw_reg.abs = 1; @@ -333,9 +334,9 @@ static struct brw_wm_ref *get_new_ref( struct brw_wm_compile *c, } - -static struct brw_wm_instruction *translate_insn( struct brw_wm_compile *c, - const struct prog_instruction *inst ) +static void +translate_insn(struct brw_wm_compile *c, + const struct prog_instruction *inst) { struct brw_wm_instruction *out = get_instruction(c); GLuint writemask = inst->DstReg.WriteMask; @@ -365,8 +366,6 @@ static struct brw_wm_instruction *translate_insn( struct brw_wm_compile *c, pass0_set_dst_scalar(c, out, inst, writemask); else pass0_set_dst(c, out, inst, writemask); - - return out; } @@ -426,6 +425,7 @@ static void pass0_init_payload( struct brw_wm_compile *c ) &c->payload.input_interp[i] ); } + /*********************************************************************** * PASS 0 * @@ -448,7 +448,6 @@ void brw_wm_pass0( struct brw_wm_compile *c ) for (insn = 0; insn < c->nr_fp_insns; insn++) { const struct prog_instruction *inst = &c->prog_instructions[insn]; - /* Optimize away moves, otherwise emit translated instruction: */ switch (inst->Opcode) { @@ -461,8 +460,6 @@ void brw_wm_pass0( struct brw_wm_compile *c ) translate_insn(c, inst); } break; - - default: translate_insn(c, inst); break; @@ -473,4 +470,3 @@ void brw_wm_pass0( struct brw_wm_compile *c ) brw_wm_print_program(c, "pass0"); } } - diff --git a/src/mesa/drivers/dri/i965/brw_wm_pass1.c b/src/mesa/drivers/dri/i965/brw_wm_pass1.c index f6f3a38e9e..6eaed8a665 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_pass1.c +++ b/src/mesa/drivers/dri/i965/brw_wm_pass1.c @@ -58,7 +58,8 @@ static void unlink_ref(struct brw_wm_ref *ref) if (ref == value->lastuse) { value->lastuse = ref->prevuse; - } else { + } + else { struct brw_wm_ref *i = value->lastuse; while (i->prevuse != ref) i = i->prevuse; i->prevuse = ref->prevuse; @@ -75,8 +76,9 @@ static void track_arg(struct brw_wm_compile *c, for (i = 0; i < 4; i++) { struct brw_wm_ref *ref = inst->src[arg][i]; if (ref) { - if (readmask & (1<<i)) + if (readmask & (1<<i)) { ref->value->contributes_to_output = 1; + } else { unlink_ref(ref); inst->src[arg][i] = NULL; @@ -88,15 +90,21 @@ static void track_arg(struct brw_wm_compile *c, static GLuint get_texcoord_mask( GLuint tex_idx ) { switch (tex_idx) { - case TEXTURE_1D_INDEX: return WRITEMASK_X; - case TEXTURE_2D_INDEX: return WRITEMASK_XY; - case TEXTURE_3D_INDEX: return WRITEMASK_XYZ; - case TEXTURE_CUBE_INDEX: return WRITEMASK_XYZ; - case TEXTURE_RECT_INDEX: return WRITEMASK_XY; + case TEXTURE_1D_INDEX: + return WRITEMASK_X; + case TEXTURE_2D_INDEX: + return WRITEMASK_XY; + case TEXTURE_3D_INDEX: + return WRITEMASK_XYZ; + case TEXTURE_CUBE_INDEX: + return WRITEMASK_XYZ; + case TEXTURE_RECT_INDEX: + return WRITEMASK_XY; default: return 0; } } + /* Step two: Basically this is dead code elimination. * * Iterate backwards over instructions, noting which values @@ -273,6 +281,3 @@ void brw_wm_pass1( struct brw_wm_compile *c ) brw_wm_print_program(c, "pass1"); } } - - - diff --git a/src/mesa/drivers/dri/i965/brw_wm_pass2.c b/src/mesa/drivers/dri/i965/brw_wm_pass2.c index 6fca9ad220..780edbc42e 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_pass2.c +++ b/src/mesa/drivers/dri/i965/brw_wm_pass2.c @@ -84,7 +84,7 @@ 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++) + for (j = 0; j < FRAG_ATTRIB_MAX; j++) { if (inputs & (1<<j)) { /* index for vs output and ps input are not the same in shader varying */ @@ -96,6 +96,7 @@ static void init_registers( struct brw_wm_compile *c ) nr_interp_regs++; prealloc_reg(c, &c->payload.input_interp[index], i++); } + } assert(nr_interp_regs >= 1); @@ -120,7 +121,7 @@ static void update_register_usage(struct brw_wm_compile *c, /* Only search those which can change: */ if (grf->nextuse < thisinsn) { - struct brw_wm_ref *ref = grf->value->lastuse; + const struct brw_wm_ref *ref = grf->value->lastuse; /* Has last use of value been passed? */ @@ -148,7 +149,7 @@ static void spill_value(struct brw_wm_compile *c, /* Allocate a spill slot. Note that allocations start from 0x40 - * the first slot is reserved to mean "undef" in brw_wm_emit.c */ - if (!value->spill_slot) { + if (!value->spill_slot) { c->last_scratch += 0x40; value->spill_slot = c->last_scratch; } @@ -189,7 +190,7 @@ static GLuint search_contiguous_regs(struct brw_wm_compile *c, if (grf[i+j].nextuse < group_nextuse) group_nextuse = grf[i+j].nextuse; } - + if (group_nextuse > furthest) { furthest = group_nextuse; reg = i; @@ -197,7 +198,7 @@ static GLuint search_contiguous_regs(struct brw_wm_compile *c, } assert(furthest != thisinsn); - + /* Any non-empty regs will need to be spilled: */ for (j = 0; j < nr; j++) @@ -243,7 +244,7 @@ static void alloc_contiguous_dest(struct brw_wm_compile *c, static void load_args(struct brw_wm_compile *c, struct brw_wm_instruction *inst) -{ +{ GLuint thisinsn = inst - c->instruction; GLuint i,j; @@ -258,17 +259,17 @@ static void load_args(struct brw_wm_compile *c, * register allocation and mark the ref as requiring a fill. */ GLuint reg = search_contiguous_regs(c, 1, thisinsn); - + c->pass2_grf[reg].value = ref->value; c->pass2_grf[reg].nextuse = thisinsn; - + ref->value->resident = &c->pass2_grf[reg]; /* Note that a fill is required: */ ref->unspill_reg = reg*2; } - + /* Adjust the hw_reg to point at the value's current location: */ assert(ref->value == ref->value->resident->value); @@ -294,7 +295,7 @@ void brw_wm_pass2( struct brw_wm_compile *c ) for (insn = 0; insn < c->nr_insns; insn++) { struct brw_wm_instruction *inst = &c->instruction[insn]; - + /* Update registers' nextuse values: */ update_register_usage(c, insn); @@ -322,11 +323,11 @@ void brw_wm_pass2( struct brw_wm_compile *c ) break; } - if (TEST_DST_SPILLS && inst->opcode != WM_PIXELXY) + if (TEST_DST_SPILLS && inst->opcode != WM_PIXELXY) { for (i = 0; i < 4; i++) if (inst->dst[i]) spill_value(c, inst->dst[i]); - + } } if (INTEL_DEBUG & DEBUG_WM) { @@ -339,6 +340,3 @@ void brw_wm_pass2( struct brw_wm_compile *c ) brw_wm_print_program(c, "pass2/done"); } } - - - diff --git a/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c b/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c index 8c9cb78945..b6dac0d698 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c +++ b/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c @@ -220,14 +220,15 @@ static void brw_wm_sampler_populate_key(struct brw_context *brw, struct wm_sampler_key *key) { + GLcontext *ctx = &brw->intel.ctx; int unit; memset(key, 0, sizeof(*key)); for (unit = 0; unit < BRW_MAX_TEX_UNIT; unit++) { - if (brw->attribs.Texture->Unit[unit]._ReallyEnabled) { + if (ctx->Texture.Unit[unit]._ReallyEnabled) { struct wm_sampler_entry *entry = &key->sampler[unit]; - struct gl_texture_unit *texUnit = &brw->attribs.Texture->Unit[unit]; + struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; struct gl_texture_object *texObj = texUnit->_Current; struct intel_texture_object *intelObj = intel_texture_object(texObj); struct gl_texture_image *firstImage = @@ -274,6 +275,7 @@ brw_wm_sampler_populate_key(struct brw_context *brw, */ static void upload_wm_samplers( struct brw_context *brw ) { + GLcontext *ctx = &brw->intel.ctx; struct wm_sampler_key key; int i; @@ -317,7 +319,7 @@ static void upload_wm_samplers( struct brw_context *brw ) /* Emit SDC relocations */ for (i = 0; i < BRW_MAX_TEX_UNIT; i++) { - if (!brw->attribs.Texture->Unit[i]._ReallyEnabled) + if (!ctx->Texture.Unit[i]._ReallyEnabled) continue; dri_bo_emit_reloc(brw->wm.sampler_bo, diff --git a/src/mesa/drivers/dri/i965/brw_wm_state.c b/src/mesa/drivers/dri/i965/brw_wm_state.c index 5302405eda..3c3b3473d6 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_state.c +++ b/src/mesa/drivers/dri/i965/brw_wm_state.c @@ -60,6 +60,7 @@ struct brw_wm_unit_key { static void wm_unit_populate_key(struct brw_context *brw, struct brw_wm_unit_key *key) { + GLcontext *ctx = &brw->intel.ctx; const struct gl_fragment_program *fp = brw->fragment_program; struct intel_context *intel = &brw->intel; @@ -95,7 +96,7 @@ wm_unit_populate_key(struct brw_context *brw, struct brw_wm_unit_key *key) key->sampler_count = brw->wm.sampler_count; /* _NEW_POLYGONSTIPPLE */ - key->polygon_stipple = brw->attribs.Polygon->StippleFlag; + key->polygon_stipple = ctx->Polygon.StippleFlag; /* BRW_NEW_FRAGMENT_PROGRAM */ key->uses_depth = (fp->Base.InputsRead & (1 << FRAG_ATTRIB_WPOS)) != 0; @@ -105,19 +106,19 @@ wm_unit_populate_key(struct brw_context *brw, struct brw_wm_unit_key *key) (fp->Base.OutputsWritten & (1 << FRAG_RESULT_DEPR)) != 0; /* _NEW_COLOR */ - key->uses_kill = fp->UsesKill || brw->attribs.Color->AlphaEnabled; + key->uses_kill = fp->UsesKill || ctx->Color.AlphaEnabled; key->is_glsl = brw_wm_is_glsl(fp); /* XXX: This needs a flag to indicate when it changes. */ key->stats_wm = intel->stats_wm; /* _NEW_LINE */ - key->line_stipple = brw->attribs.Line->StippleFlag; + key->line_stipple = ctx->Line.StippleFlag; /* _NEW_POLYGON */ - key->offset_enable = brw->attribs.Polygon->OffsetFill; - key->offset_units = brw->attribs.Polygon->OffsetUnits; - key->offset_factor = brw->attribs.Polygon->OffsetFactor; + key->offset_enable = ctx->Polygon.OffsetFill; + key->offset_units = ctx->Polygon.OffsetUnits; + key->offset_factor = ctx->Polygon.OffsetFactor; } static dri_bo * 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 06e71e6d69..d70f9c646c 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c +++ b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c @@ -253,7 +253,7 @@ static void brw_update_texture_surface( GLcontext *ctx, GLuint unit ) { struct brw_context *brw = brw_context(ctx); - struct gl_texture_object *tObj = brw->attribs.Texture->Unit[unit]._Current; + struct gl_texture_object *tObj = ctx->Texture.Unit[unit]._Current; struct intel_texture_object *intelObj = intel_texture_object(tObj); struct gl_texture_image *firstImage = tObj->Image[0][intelObj->firstLevel]; struct brw_wm_surface_key key; @@ -301,6 +301,7 @@ static void brw_update_region_surface(struct brw_context *brw, struct intel_region *region, unsigned int unit, GLboolean cached) { + GLcontext *ctx = &brw->intel.ctx; dri_bo *region_bo = NULL; struct { unsigned int surface_type; @@ -333,10 +334,10 @@ brw_update_region_surface(struct brw_context *brw, struct intel_region *region, key.height = 1; key.cpp = 4; } - memcpy(key.color_mask, brw->attribs.Color->ColorMask, + memcpy(key.color_mask, ctx->Color.ColorMask, sizeof(key.color_mask)); - key.color_blend = (!brw->attribs.Color->_LogicOpEnabled && - brw->attribs.Color->BlendEnabled); + key.color_blend = (!ctx->Color._LogicOpEnabled && + ctx->Color.BlendEnabled); dri_bo_unreference(brw->wm.surf_bo[unit]); brw->wm.surf_bo[unit] = NULL; @@ -459,7 +460,7 @@ static void prepare_wm_surfaces(struct brw_context *brw ) brw->wm.nr_surfaces = MAX_DRAW_BUFFERS; for (i = 0; i < BRW_MAX_TEX_UNIT; i++) { - struct gl_texture_unit *texUnit = &brw->attribs.Texture->Unit[i]; + struct gl_texture_unit *texUnit = &ctx->Texture.Unit[i]; /* _NEW_TEXTURE, BRW_NEW_TEXDATA */ if(texUnit->_ReallyEnabled) { diff --git a/src/mesa/drivers/dri/i965/intel_clear.c b/src/mesa/drivers/dri/i965/intel_clear.c new file mode 120000 index 0000000000..9a2a742a0d --- /dev/null +++ b/src/mesa/drivers/dri/i965/intel_clear.c @@ -0,0 +1 @@ +../intel/intel_clear.c
\ No newline at end of file diff --git a/src/mesa/drivers/dri/i965/intel_depthstencil.c b/src/mesa/drivers/dri/i965/intel_depthstencil.c deleted file mode 120000 index 4ac4ae690a..0000000000 --- a/src/mesa/drivers/dri/i965/intel_depthstencil.c +++ /dev/null @@ -1 +0,0 @@ -../intel/intel_depthstencil.c
\ No newline at end of file diff --git a/src/mesa/drivers/dri/i965/intel_extensions.c b/src/mesa/drivers/dri/i965/intel_extensions.c new file mode 120000 index 0000000000..a2f3e8cd20 --- /dev/null +++ b/src/mesa/drivers/dri/i965/intel_extensions.c @@ -0,0 +1 @@ +../intel/intel_extensions.c
\ No newline at end of file diff --git a/src/mesa/drivers/dri/i965/intel_state.c b/src/mesa/drivers/dri/i965/intel_state.c index 67ef5f78c1..519672fc35 100644..120000 --- a/src/mesa/drivers/dri/i965/intel_state.c +++ b/src/mesa/drivers/dri/i965/intel_state.c @@ -1,225 +1 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "main/glheader.h" -#include "main/context.h" -#include "main/macros.h" -#include "main/enums.h" -#include "main/colormac.h" -#include "main/dd.h" - -#include "intel_screen.h" -#include "intel_context.h" -#include "intel_regions.h" -#include "swrast/swrast.h" - -int intel_translate_shadow_compare_func( GLenum func ) -{ - switch(func) { - case GL_NEVER: - return COMPAREFUNC_ALWAYS; - case GL_LESS: - return COMPAREFUNC_LEQUAL; - case GL_LEQUAL: - return COMPAREFUNC_LESS; - case GL_GREATER: - return COMPAREFUNC_GEQUAL; - case GL_GEQUAL: - return COMPAREFUNC_GREATER; - case GL_NOTEQUAL: - return COMPAREFUNC_EQUAL; - case GL_EQUAL: - return COMPAREFUNC_NOTEQUAL; - case GL_ALWAYS: - return COMPAREFUNC_NEVER; - } - - fprintf(stderr, "Unknown value in %s: %x\n", __FUNCTION__, func); - return COMPAREFUNC_NEVER; -} - -int intel_translate_compare_func( GLenum func ) -{ - switch(func) { - case GL_NEVER: - return COMPAREFUNC_NEVER; - case GL_LESS: - return COMPAREFUNC_LESS; - case GL_LEQUAL: - return COMPAREFUNC_LEQUAL; - case GL_GREATER: - return COMPAREFUNC_GREATER; - case GL_GEQUAL: - return COMPAREFUNC_GEQUAL; - case GL_NOTEQUAL: - return COMPAREFUNC_NOTEQUAL; - case GL_EQUAL: - return COMPAREFUNC_EQUAL; - case GL_ALWAYS: - return COMPAREFUNC_ALWAYS; - } - - fprintf(stderr, "Unknown value in %s: %x\n", __FUNCTION__, func); - return COMPAREFUNC_ALWAYS; -} - -int intel_translate_stencil_op( GLenum op ) -{ - switch(op) { - case GL_KEEP: - return STENCILOP_KEEP; - case GL_ZERO: - return STENCILOP_ZERO; - case GL_REPLACE: - return STENCILOP_REPLACE; - case GL_INCR: - return STENCILOP_INCRSAT; - case GL_DECR: - return STENCILOP_DECRSAT; - case GL_INCR_WRAP: - return STENCILOP_INCR; - case GL_DECR_WRAP: - return STENCILOP_DECR; - case GL_INVERT: - return STENCILOP_INVERT; - default: - return STENCILOP_ZERO; - } -} - -int intel_translate_blend_factor( GLenum factor ) -{ - switch(factor) { - case GL_ZERO: - return BLENDFACT_ZERO; - case GL_SRC_ALPHA: - return BLENDFACT_SRC_ALPHA; - case GL_ONE: - return BLENDFACT_ONE; - case GL_SRC_COLOR: - return BLENDFACT_SRC_COLR; - case GL_ONE_MINUS_SRC_COLOR: - return BLENDFACT_INV_SRC_COLR; - case GL_DST_COLOR: - return BLENDFACT_DST_COLR; - case GL_ONE_MINUS_DST_COLOR: - return BLENDFACT_INV_DST_COLR; - case GL_ONE_MINUS_SRC_ALPHA: - return BLENDFACT_INV_SRC_ALPHA; - case GL_DST_ALPHA: - return BLENDFACT_DST_ALPHA; - case GL_ONE_MINUS_DST_ALPHA: - return BLENDFACT_INV_DST_ALPHA; - case GL_SRC_ALPHA_SATURATE: - return BLENDFACT_SRC_ALPHA_SATURATE; - case GL_CONSTANT_COLOR: - return BLENDFACT_CONST_COLOR; - case GL_ONE_MINUS_CONSTANT_COLOR: - return BLENDFACT_INV_CONST_COLOR; - case GL_CONSTANT_ALPHA: - return BLENDFACT_CONST_ALPHA; - case GL_ONE_MINUS_CONSTANT_ALPHA: - return BLENDFACT_INV_CONST_ALPHA; - } - - fprintf(stderr, "Unknown value in %s: %x\n", __FUNCTION__, factor); - return BLENDFACT_ZERO; -} - -int intel_translate_logic_op( GLenum opcode ) -{ - switch(opcode) { - case GL_CLEAR: - return LOGICOP_CLEAR; - case GL_AND: - return LOGICOP_AND; - case GL_AND_REVERSE: - return LOGICOP_AND_RVRSE; - case GL_COPY: - return LOGICOP_COPY; - case GL_COPY_INVERTED: - return LOGICOP_COPY_INV; - case GL_AND_INVERTED: - return LOGICOP_AND_INV; - case GL_NOOP: - return LOGICOP_NOOP; - case GL_XOR: - return LOGICOP_XOR; - case GL_OR: - return LOGICOP_OR; - case GL_OR_INVERTED: - return LOGICOP_OR_INV; - case GL_NOR: - return LOGICOP_NOR; - case GL_EQUIV: - return LOGICOP_EQUIV; - case GL_INVERT: - return LOGICOP_INV; - case GL_OR_REVERSE: - return LOGICOP_OR_RVRSE; - case GL_NAND: - return LOGICOP_NAND; - case GL_SET: - return LOGICOP_SET; - default: - return LOGICOP_SET; - } -} - - -static void intelClearColor(GLcontext *ctx, const GLfloat color[4]) -{ - struct intel_context *intel = intel_context(ctx); - - UNCLAMPED_FLOAT_TO_RGBA_CHAN(intel->clear_chan, color); - - intel->ClearColor8888 = INTEL_PACKCOLOR8888(intel->clear_chan[0], - intel->clear_chan[1], - intel->clear_chan[2], - intel->clear_chan[3]); - intel->ClearColor565 = INTEL_PACKCOLOR565(intel->clear_chan[0], - intel->clear_chan[1], - intel->clear_chan[2]); -} - - - -/* Fallback to swrast for select and feedback. - */ -static void intelRenderMode( GLcontext *ctx, GLenum mode ) -{ - struct intel_context *intel = intel_context(ctx); - FALLBACK( intel, INTEL_FALLBACK_RENDERMODE, (mode != GL_RENDER) ); -} - - -void intelInitStateFuncs( struct dd_function_table *functions ) -{ - functions->RenderMode = intelRenderMode; - functions->ClearColor = intelClearColor; -} +../intel/intel_state.c
\ No newline at end of file diff --git a/src/mesa/drivers/dri/i965/intel_swapbuffers.c b/src/mesa/drivers/dri/i965/intel_swapbuffers.c new file mode 120000 index 0000000000..148d5215aa --- /dev/null +++ b/src/mesa/drivers/dri/i965/intel_swapbuffers.c @@ -0,0 +1 @@ +../intel/intel_swapbuffers.c
\ No newline at end of file |