diff options
Diffstat (limited to 'src/mesa')
106 files changed, 2368 insertions, 1799 deletions
diff --git a/src/mesa/drivers/dri/Makefile.template b/src/mesa/drivers/dri/Makefile.template index 2fa36bab3f9..5c01d233c13 100644 --- a/src/mesa/drivers/dri/Makefile.template +++ b/src/mesa/drivers/dri/Makefile.template @@ -72,10 +72,11 @@ $(TOP)/$(LIB_DIR)/$(LIBNAME): $(LIBNAME) depend: $(C_SOURCES) $(ASM_SOURCES) $(SYMLINKS) - rm -f depend - touch depend - $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) \ - $(ASM_SOURCES) + @ echo "running $(MKDEP)" + @ rm -f depend + @ touch depend + @ $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) \ + $(ASM_SOURCES) > /dev/null 2>/dev/null # Emacs tags diff --git a/src/mesa/drivers/dri/common/dri_util.c b/src/mesa/drivers/dri/common/dri_util.c index 0ec4adc2325..ae0e61e515b 100644 --- a/src/mesa/drivers/dri/common/dri_util.c +++ b/src/mesa/drivers/dri/common/dri_util.c @@ -62,7 +62,7 @@ __driUtilMessage(const char *f, ...) va_list args; if (getenv("LIBGL_DEBUG")) { - fprintf(stderr, "libGL error: \n"); + fprintf(stderr, "libGL: "); va_start(args, f); vfprintf(stderr, f, args); va_end(args); @@ -314,12 +314,12 @@ static void driSwapBuffers(__DRIdrawable *dPriv) __DRIscreen *psp = dPriv->driScreenPriv; drm_clip_rect_t *rects; int i; - - if (!dPriv->numClipRects) - return; psp->DriverAPI.SwapBuffers(dPriv); + if (!dPriv->numClipRects) + return; + rects = _mesa_malloc(sizeof(*rects) * dPriv->numClipRects); if (!rects) diff --git a/src/mesa/drivers/dri/common/spantmp2.h b/src/mesa/drivers/dri/common/spantmp2.h index f2868cb58a6..89c815722f6 100644 --- a/src/mesa/drivers/dri/common/spantmp2.h +++ b/src/mesa/drivers/dri/common/spantmp2.h @@ -82,6 +82,71 @@ rgba[3] = 0xff; \ } while (0) +#elif (SPANTMP_PIXEL_FMT == GL_BGRA) && (SPANTMP_PIXEL_TYPE == GL_UNSIGNED_SHORT_4_4_4_4_REV) + +/** + ** GL_BGRA, GL_UNSIGNED_SHORT_4_4_4_4_REV + **/ + +#ifndef GET_VALUE +#ifndef GET_PTR +#define GET_PTR(_x, _y) (buf + (_x) * 2 + (_y) * pitch) +#endif + +#define GET_VALUE(_x, _y) *(volatile GLushort *)(GET_PTR(_x, _y)) +#define PUT_VALUE(_x, _y, _v) *(volatile GLushort *)(GET_PTR(_x, _y)) = (_v) +#endif /* GET_VALUE */ + +#define INIT_MONO_PIXEL(p, color) \ + p = PACK_COLOR_4444(color[3], color[0], color[1], color[2]) + +#define WRITE_RGBA( _x, _y, r, g, b, a ) \ + PUT_VALUE(_x, _y, PACK_COLOR_4444(a, r, g, b)) \ + +#define WRITE_PIXEL( _x, _y, p ) PUT_VALUE(_x, _y, p) + +#define READ_RGBA( rgba, _x, _y ) \ + do { \ + GLushort p = GET_VALUE(_x, _y); \ + rgba[0] = ((p >> 8) & 0xf) * 0x11; \ + rgba[1] = ((p >> 4) & 0xf) * 0x11; \ + rgba[2] = ((p >> 0) & 0xf) * 0x11; \ + rgba[3] = ((p >> 12) & 0xf) * 0x11; \ + } while (0) + + +#elif (SPANTMP_PIXEL_FMT == GL_BGRA) && (SPANTMP_PIXEL_TYPE == GL_UNSIGNED_SHORT_1_5_5_5_REV) + +/** + ** GL_BGRA, GL_UNSIGNED_SHORT_1_5_5_5_REV + **/ + +#ifndef GET_VALUE +#ifndef GET_PTR +#define GET_PTR(_x, _y) (buf + (_x) * 2 + (_y) * pitch) +#endif + +#define GET_VALUE(_x, _y) *(volatile GLushort *)(GET_PTR(_x, _y)) +#define PUT_VALUE(_x, _y, _v) *(volatile GLushort *)(GET_PTR(_x, _y)) = (_v) +#endif /* GET_VALUE */ + +#define INIT_MONO_PIXEL(p, color) \ + p = PACK_COLOR_1555(color[3], color[0], color[1], color[2]) + +#define WRITE_RGBA( _x, _y, r, g, b, a ) \ + PUT_VALUE(_x, _y, PACK_COLOR_1555(a, r, g, b)) \ + +#define WRITE_PIXEL( _x, _y, p ) PUT_VALUE(_x, _y, p) + +#define READ_RGBA( rgba, _x, _y ) \ + do { \ + GLushort p = GET_VALUE(_x, _y); \ + rgba[0] = ((p >> 7) & 0xf8) * 255 / 0xf8; \ + rgba[1] = ((p >> 2) & 0xf8) * 255 / 0xf8; \ + rgba[2] = ((p << 3) & 0xf8) * 255 / 0xf8; \ + rgba[3] = ((p >> 15) & 0x1) * 0xff; \ + } while (0) + #elif (SPANTMP_PIXEL_FMT == GL_BGRA) && (SPANTMP_PIXEL_TYPE == GL_UNSIGNED_INT_8_8_8_8_REV) /** diff --git a/src/mesa/drivers/dri/i915/Makefile b/src/mesa/drivers/dri/i915/Makefile index 954a7e2af1f..9f4bd1699f9 100644 --- a/src/mesa/drivers/dri/i915/Makefile +++ b/src/mesa/drivers/dri/i915/Makefile @@ -11,7 +11,6 @@ DRIVER_SOURCES = \ i830_metaops.c \ i830_state.c \ i830_texblend.c \ - i830_tex.c \ i830_texstate.c \ i830_vtbl.c \ intel_render.c \ @@ -36,7 +35,6 @@ DRIVER_SOURCES = \ intel_buffers.c \ intel_blit.c \ intel_swapbuffers.c \ - i915_tex.c \ i915_tex_layout.c \ i915_texstate.c \ i915_context.c \ diff --git a/src/mesa/drivers/dri/i915/i830_context.c b/src/mesa/drivers/dri/i915/i830_context.c index 9c540cb2bb7..10b9bf371c3 100644 --- a/src/mesa/drivers/dri/i915/i830_context.c +++ b/src/mesa/drivers/dri/i915/i830_context.c @@ -47,7 +47,6 @@ i830InitDriverFunctions(struct dd_function_table *functions) { intelInitDriverFunctions(functions); i830InitStateFuncs(functions); - i830InitTextureFuncs(functions); } extern const struct tnl_pipeline_stage *intel_pipeline[]; diff --git a/src/mesa/drivers/dri/i915/i830_tex.c b/src/mesa/drivers/dri/i915/i830_tex.c deleted file mode 100644 index 34ac42a78e0..00000000000 --- a/src/mesa/drivers/dri/i915/i830_tex.c +++ /dev/null @@ -1,100 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "main/glheader.h" -#include "main/mtypes.h" -#include "main/imports.h" -#include "main/simple_list.h" -#include "main/enums.h" -#include "main/image.h" -#include "main/mm.h" -#include "main/texstore.h" -#include "main/texformat.h" -#include "swrast/swrast.h" - -#include "texmem.h" - -#include "i830_context.h" -#include "i830_reg.h" - - - -static void -i830TexEnv(GLcontext * ctx, GLenum target, - GLenum pname, const GLfloat * param) -{ - - switch (pname) { - case GL_TEXTURE_ENV_COLOR: - case GL_TEXTURE_ENV_MODE: - case GL_COMBINE_RGB: - case GL_COMBINE_ALPHA: - case GL_SOURCE0_RGB: - case GL_SOURCE1_RGB: - case GL_SOURCE2_RGB: - case GL_SOURCE0_ALPHA: - case GL_SOURCE1_ALPHA: - case GL_SOURCE2_ALPHA: - case GL_OPERAND0_RGB: - case GL_OPERAND1_RGB: - case GL_OPERAND2_RGB: - case GL_OPERAND0_ALPHA: - case GL_OPERAND1_ALPHA: - case GL_OPERAND2_ALPHA: - case GL_RGB_SCALE: - case GL_ALPHA_SCALE: - break; - - case GL_TEXTURE_LOD_BIAS:{ - struct i830_context *i830 = i830_context(ctx); - GLuint unit = ctx->Texture.CurrentUnit; - int b = (int) ((*param) * 16.0); - if (b > 63) - b = 63; - if (b < -64) - b = -64; - I830_STATECHANGE(i830, I830_UPLOAD_TEX(unit)); - i830->lodbias_tm0s3[unit] = - ((b << TM0S3_LOD_BIAS_SHIFT) & TM0S3_LOD_BIAS_MASK); - break; - } - - default: - break; - } -} - - - - -void -i830InitTextureFuncs(struct dd_function_table *functions) -{ -/* - functions->TexEnv = i830TexEnv; -*/ -} diff --git a/src/mesa/drivers/dri/i915/i830_vtbl.c b/src/mesa/drivers/dri/i915/i830_vtbl.c index 1a949210789..3bf02de61f8 100644 --- a/src/mesa/drivers/dri/i915/i830_vtbl.c +++ b/src/mesa/drivers/dri/i915/i830_vtbl.c @@ -26,12 +26,14 @@ **************************************************************************/ #include "glapi/glapi.h" +#include "main/texformat.h" #include "i830_context.h" #include "i830_reg.h" #include "intel_batchbuffer.h" #include "intel_regions.h" #include "intel_tris.h" +#include "intel_fbo.h" #include "tnl/t_context.h" #include "tnl/t_vertex.h" @@ -614,6 +616,8 @@ i830_state_draw_region(struct intel_context *intel, { struct i830_context *i830 = i830_context(&intel->ctx); GLcontext *ctx = &intel->ctx; + struct gl_renderbuffer *rb = ctx->DrawBuffer->_ColorDrawBuffers[0]; + struct intel_renderbuffer *irb = intel_renderbuffer(rb); GLuint value; ASSERT(state == &i830->state || state == &i830->meta); @@ -651,13 +655,27 @@ i830_state_draw_region(struct intel_context *intel, */ value = (DSTORG_HORT_BIAS(0x8) | /* .5 */ DSTORG_VERT_BIAS(0x8) | DEPTH_IS_Z); /* .5 */ - - if (color_region && color_region->cpp == 4) { - value |= DV_PF_8888; - } - else { - value |= DV_PF_565; + + if (irb != NULL) { + switch (irb->texformat->MesaFormat) { + case MESA_FORMAT_ARGB8888: + value |= DV_PF_8888; + break; + case MESA_FORMAT_RGB565: + value |= DV_PF_565; + break; + case MESA_FORMAT_ARGB1555: + value |= DV_PF_1555; + break; + case MESA_FORMAT_ARGB4444: + value |= DV_PF_4444; + break; + default: + _mesa_problem(ctx, "Bad renderbuffer format: %d\n", + irb->texformat->MesaFormat); + } } + if (depth_region && depth_region->cpp == 4) { value |= DEPTH_FRMT_24_FIXED_8_OTHER; } diff --git a/src/mesa/drivers/dri/i915/i915_context.c b/src/mesa/drivers/dri/i915/i915_context.c index 7549029a1be..fdd2cf61096 100644 --- a/src/mesa/drivers/dri/i915/i915_context.c +++ b/src/mesa/drivers/dri/i915/i915_context.c @@ -83,7 +83,6 @@ i915InitDriverFunctions(struct dd_function_table *functions) { intelInitDriverFunctions(functions); i915InitStateFunctions(functions); - i915InitTextureFuncs(functions); i915InitFragProgFuncs(functions); functions->UpdateState = i915InvalidateState; } diff --git a/src/mesa/drivers/dri/i915/i915_fragprog.c b/src/mesa/drivers/dri/i915/i915_fragprog.c index 52f09a4b1b2..2db10c60e99 100644 --- a/src/mesa/drivers/dri/i915/i915_fragprog.c +++ b/src/mesa/drivers/dri/i915/i915_fragprog.c @@ -162,12 +162,12 @@ src_vector(struct i915_fragment_program *p, GET_SWZ(source->Swizzle, 1), GET_SWZ(source->Swizzle, 2), GET_SWZ(source->Swizzle, 3)); - if (source->NegateBase) + if (source->Negate) src = negate(src, - GET_BIT(source->NegateBase, 0), - GET_BIT(source->NegateBase, 1), - GET_BIT(source->NegateBase, 2), - GET_BIT(source->NegateBase, 3)); + GET_BIT(source->Negate, 0), + GET_BIT(source->Negate, 1), + GET_BIT(source->Negate, 2), + GET_BIT(source->Negate, 3)); return src; } @@ -323,7 +323,8 @@ upload_program(struct i915_fragment_program *p) p->ctx->FragmentProgram._Current; const struct prog_instruction *inst = program->Base.Instructions; -/* _mesa_debug_fp_inst(program->Base.NumInstructions, inst); */ + if (INTEL_DEBUG & DEBUG_WM) + _mesa_print_program(&program->Base); /* Is this a parse-failed program? Ensure a valid program is * loaded, as the flagging of an error isn't sufficient to stop @@ -1049,9 +1050,6 @@ i915ProgramStringNotify(GLcontext * ctx, _mesa_append_fog_code(ctx, &p->FragProg); p->FragProg.FogOption = GL_NONE; } - - if (INTEL_DEBUG & DEBUG_STATE) - _mesa_print_program(prog); } _tnl_program_string(ctx, target, prog); diff --git a/src/mesa/drivers/dri/i915/i915_tex.c b/src/mesa/drivers/dri/i915/i915_tex.c deleted file mode 100644 index e38d8fe79d1..00000000000 --- a/src/mesa/drivers/dri/i915/i915_tex.c +++ /dev/null @@ -1,78 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "main/glheader.h" -#include "main/mtypes.h" -#include "main/imports.h" -#include "main/simple_list.h" -#include "main/enums.h" -#include "main/image.h" -#include "main/mm.h" -#include "main/texstore.h" -#include "main/texformat.h" -#include "swrast/swrast.h" - -#include "texmem.h" - -#include "i915_context.h" -#include "i915_reg.h" - - - -static void -i915TexEnv(GLcontext * ctx, GLenum target, - GLenum pname, const GLfloat * param) -{ - struct i915_context *i915 = I915_CONTEXT(ctx); - - switch (pname) { - case GL_TEXTURE_LOD_BIAS:{ - GLuint unit = ctx->Texture.CurrentUnit; - GLint b = (int) ((*param) * 16.0); - if (b > 255) - b = 255; - if (b < -256) - b = -256; - I915_STATECHANGE(i915, I915_UPLOAD_TEX(unit)); - i915->lodbias_ss2[unit] = - ((b << SS2_LOD_BIAS_SHIFT) & SS2_LOD_BIAS_MASK); - break; - } - - default: - break; - } -} - - -void -i915InitTextureFuncs(struct dd_function_table *functions) -{ -/* - functions->TexEnv = i915TexEnv; -*/ -} diff --git a/src/mesa/drivers/dri/i915/i915_vtbl.c b/src/mesa/drivers/dri/i915/i915_vtbl.c index 3f6d282d342..115004616ff 100644 --- a/src/mesa/drivers/dri/i915/i915_vtbl.c +++ b/src/mesa/drivers/dri/i915/i915_vtbl.c @@ -32,6 +32,7 @@ #include "main/imports.h" #include "main/macros.h" #include "main/colormac.h" +#include "main/texformat.h" #include "tnl/t_context.h" #include "tnl/t_vertex.h" @@ -40,6 +41,7 @@ #include "intel_tex.h" #include "intel_regions.h" #include "intel_tris.h" +#include "intel_fbo.h" #include "i915_reg.h" #include "i915_context.h" @@ -542,6 +544,8 @@ i915_state_draw_region(struct intel_context *intel, { struct i915_context *i915 = i915_context(&intel->ctx); GLcontext *ctx = &intel->ctx; + struct gl_renderbuffer *rb = ctx->DrawBuffer->_ColorDrawBuffers[0]; + struct intel_renderbuffer *irb = intel_renderbuffer(rb); GLuint value; ASSERT(state == &i915->state || state == &i915->meta); @@ -580,12 +584,26 @@ i915_state_draw_region(struct intel_context *intel, value = (DSTORG_HORT_BIAS(0x8) | /* .5 */ DSTORG_VERT_BIAS(0x8) | /* .5 */ LOD_PRECLAMP_OGL | TEX_DEFAULT_COLOR_OGL); - if (color_region && color_region->cpp == 4) { - value |= DV_PF_8888; - } - else { - value |= (DITHER_FULL_ALWAYS | DV_PF_565); + if (irb != NULL) { + switch (irb->texformat->MesaFormat) { + case MESA_FORMAT_ARGB8888: + value |= DV_PF_8888; + break; + case MESA_FORMAT_RGB565: + value |= DV_PF_565 | DITHER_FULL_ALWAYS; + break; + case MESA_FORMAT_ARGB1555: + value |= DV_PF_1555 | DITHER_FULL_ALWAYS; + break; + case MESA_FORMAT_ARGB4444: + value |= DV_PF_4444 | DITHER_FULL_ALWAYS; + break; + default: + _mesa_problem(ctx, "Bad renderbuffer format: %d\n", + irb->texformat->MesaFormat); + } } + if (depth_region && depth_region->cpp == 4) { value |= DEPTH_FRMT_24_FIXED_8_OTHER; } diff --git a/src/mesa/drivers/dri/i965/brw_context.h b/src/mesa/drivers/dri/i965/brw_context.h index 01e07c967fa..a0b3b06309f 100644 --- a/src/mesa/drivers/dri/i965/brw_context.h +++ b/src/mesa/drivers/dri/i965/brw_context.h @@ -141,7 +141,8 @@ struct brw_context; #define BRW_NEW_BATCH 0x10000 /** brw->depth_region updated */ #define BRW_NEW_DEPTH_BUFFER 0x20000 -#define BRW_NEW_NR_SURFACES 0x40000 +#define BRW_NEW_NR_WM_SURFACES 0x40000 +#define BRW_NEW_NR_VS_SURFACES 0x80000 struct brw_state_flags { /** State update flags signalled by mesa internals */ @@ -159,6 +160,7 @@ struct brw_state_flags { struct brw_vertex_program { struct gl_vertex_program program; GLuint id; + dri_bo *const_buffer; /** Program constant buffer/surface */ }; @@ -168,8 +170,7 @@ struct brw_fragment_program { GLuint id; /**< serial no. to identify frag progs, never re-used */ GLboolean isGLSL; /**< really, any IF/LOOP/CONT/BREAK instructions */ - /** Program constant buffer/surface */ - dri_bo *const_buffer; + dri_bo *const_buffer; /** Program constant buffer/surface */ }; @@ -186,7 +187,7 @@ struct brw_wm_prog_data { GLuint total_grf; GLuint total_scratch; - GLuint nr_params; + GLuint nr_params; /**< number of float params/constants */ GLboolean error; /* Pointer to tracked values (only valid once @@ -225,6 +226,7 @@ struct brw_vs_prog_data { GLuint urb_read_length; GLuint total_grf; GLuint outputs_written; + GLuint nr_params; /**< number of float params/constants */ GLuint inputs_read; @@ -245,12 +247,31 @@ struct brw_vs_ouput_sizes { #define BRW_MAX_TEX_UNIT 16 /** - * Size of our surface binding table. + * Size of our surface binding table for the WM. * This contains pointers to the drawing surfaces and current texture - * objects and shader constant buffer (+1). + * objects and shader constant buffers (+2). */ #define BRW_WM_MAX_SURF (MAX_DRAW_BUFFERS + BRW_MAX_TEX_UNIT + 1) +/** + * Helpers to convert drawing buffers, textures and constant buffers + * to surface binding table indexes, for WM. + */ +#define SURF_INDEX_DRAW(d) (d) +#define SURF_INDEX_FRAG_CONST_BUFFER (MAX_DRAW_BUFFERS) +#define SURF_INDEX_TEXTURE(t) (MAX_DRAW_BUFFERS + 1 + (t)) + +/** + * Size of surface binding table for the VS. + * Only one constant buffer for now. + */ +#define BRW_VS_MAX_SURF 1 + +/** + * Only a VS constant buffer + */ +#define SURF_INDEX_VERT_CONST_BUFFER 0 + enum brw_cache_id { BRW_CC_VP, @@ -557,6 +578,11 @@ struct brw_context dri_bo *prog_bo; dri_bo *state_bo; + + /** Binding table of pointers to surf_bo entries */ + dri_bo *bind_bo; + dri_bo *surf_bo[BRW_VS_MAX_SURF]; + GLuint nr_surfaces; } vs; struct { diff --git a/src/mesa/drivers/dri/i965/brw_curbe.c b/src/mesa/drivers/dri/i965/brw_curbe.c index a6bfb7507e7..18b187ed1dc 100644 --- a/src/mesa/drivers/dri/i965/brw_curbe.c +++ b/src/mesa/drivers/dri/i965/brw_curbe.c @@ -45,17 +45,21 @@ #include "brw_util.h" -/* Partition the CURBE between the various users of constant values: +/** + * Partition the CURBE between the various users of constant values: + * Note that vertex and fragment shaders can now fetch constants out + * of constant buffers. We no longer allocatea block of the GRF for + * constants. That greatly reduces the demand for space in the CURBE. + * Some of the comments within are dated... */ 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; + const GLuint nr_fp_regs = (brw->wm.prog_data->nr_params + 15) / 16; /* BRW_NEW_VERTEX_PROGRAM */ - const struct brw_vertex_program *vp = brw_vertex_program_const(brw->vertex_program); - GLuint nr_vp_regs = (vp->program.Base.Parameters->NumParameters * 4 + 15) / 16; + const GLuint nr_vp_regs = (brw->vs.prog_data->nr_params + 15) / 16; GLuint nr_clip_regs = 0; GLuint total_regs; @@ -248,7 +252,7 @@ static void prepare_constant_buffer(struct brw_context *brw) /* vertex shader constants */ if (brw->curbe.vs_size) { GLuint offset = brw->curbe.vs_start * 16; - GLuint nr = vp->program.Base.Parameters->NumParameters; + GLuint nr = brw->vs.prog_data->nr_params / 4; _mesa_load_state_parameters(ctx, vp->program.Base.Parameters); @@ -333,37 +337,69 @@ static void prepare_constant_buffer(struct brw_context *brw) /** - * Vertex/fragment shader constants are stored in a pseudo 1D texture. - * This function updates the constants in that buffer. + * Copy Mesa program parameters into given constant buffer. */ static void -update_texture_constant_buffer(struct brw_context *brw) +update_constant_buffer(struct brw_context *brw, + const struct gl_program_parameter_list *params, + dri_bo *const_buffer) { - struct brw_fragment_program *fp = - (struct brw_fragment_program *) brw->fragment_program; - const struct gl_program_parameter_list *params = fp->program.Base.Parameters; const int size = params->NumParameters * 4 * sizeof(GLfloat); - assert(fp->const_buffer); - assert(fp->const_buffer->size >= size); - - /* copy constants into the buffer */ - if (size > 0) { + /* copy Mesa program constants into the buffer */ + if (const_buffer && size > 0) { GLubyte *map; - dri_bo_map(fp->const_buffer, GL_TRUE); - map = fp->const_buffer->virtual; + + assert(const_buffer); + assert(const_buffer->size >= size); + + dri_bo_map(const_buffer, GL_TRUE); + map = const_buffer->virtual; memcpy(map, params->ParameterValues, size); - dri_bo_unmap(fp->const_buffer); + dri_bo_unmap(const_buffer); + + if (0) { + int i; + for (i = 0; i < params->NumParameters; i++) { + float *p = params->ParameterValues[i]; + printf("%d: %f %f %f %f\n", i, p[0], p[1], p[2], p[3]); + } + } } } +/** Copy current vertex program's parameters into the constant buffer */ +static void +update_vertex_constant_buffer(struct brw_context *brw) +{ + struct brw_vertex_program *vp = + (struct brw_vertex_program *) brw->vertex_program; + if (0) { + printf("update VS constants in buffer %p\n", vp->const_buffer); + printf("program %u\n", vp->program.Base.Id); + } + update_constant_buffer(brw, vp->program.Base.Parameters, vp->const_buffer); +} + + +/** Copy current fragment program's parameters into the constant buffer */ +static void +update_fragment_constant_buffer(struct brw_context *brw) +{ + struct brw_fragment_program *fp = + (struct brw_fragment_program *) brw->fragment_program; + update_constant_buffer(brw, fp->program.Base.Parameters, fp->const_buffer); +} + + static void emit_constant_buffer(struct brw_context *brw) { struct intel_context *intel = &brw->intel; GLuint sz = brw->curbe.total_size; - update_texture_constant_buffer(brw); + update_vertex_constant_buffer(brw); + update_fragment_constant_buffer(brw); BEGIN_BATCH(2, IGNORE_CLIPRECTS); if (sz == 0) { diff --git a/src/mesa/drivers/dri/i965/brw_eu.h b/src/mesa/drivers/dri/i965/brw_eu.h index d05f2e6c410..62c98bd8bb3 100644 --- a/src/mesa/drivers/dri/i965/brw_eu.h +++ b/src/mesa/drivers/dri/i965/brw_eu.h @@ -862,9 +862,17 @@ void brw_dp_READ_4( struct brw_compile *p, struct brw_reg dest, GLuint msg_reg_nr, GLboolean relAddr, - GLuint scratch_offset, + GLuint location, GLuint bind_table_index ); +void brw_dp_READ_4_vs( struct brw_compile *p, + struct brw_reg dest, + GLuint oword, + GLboolean relAddr, + struct brw_reg addrReg, + GLuint location, + GLuint bind_table_index ); + void brw_dp_WRITE_16( struct brw_compile *p, struct brw_reg src, GLuint msg_reg_nr, diff --git a/src/mesa/drivers/dri/i965/brw_eu_emit.c b/src/mesa/drivers/dri/i965/brw_eu_emit.c index 21ce8369db4..60ea44f7a96 100644 --- a/src/mesa/drivers/dri/i965/brw_eu_emit.c +++ b/src/mesa/drivers/dri/i965/brw_eu_emit.c @@ -952,7 +952,7 @@ void brw_dp_READ_16( struct brw_compile *p, /** * Read a float[4] vector from the data port Data Cache (const buffer). - * Scratch offset should be a multiple of 16. + * Location (in buffer) should be a multiple of 16. * Used for fetching shader constants. * If relAddr is true, we'll do an indirect fetch using the address register. */ @@ -960,7 +960,7 @@ void brw_dp_READ_4( struct brw_compile *p, struct brw_reg dest, GLuint msg_reg_nr, GLboolean relAddr, - GLuint scratch_offset, + GLuint location, GLuint bind_table_index ) { { @@ -969,18 +969,20 @@ void brw_dp_READ_4( struct brw_compile *p, brw_set_mask_control(p, BRW_MASK_DISABLE); /* set message header global offset field (reg 0, element 2) */ + /* Note that grf[0] will be copied to mrf[1] implicitly by the SEND instr */ brw_MOV(p, retype(brw_vec1_grf(0, 2), BRW_REGISTER_TYPE_UD), - brw_imm_d(scratch_offset)); + brw_imm_d(location)); brw_pop_insn_state(p); } { struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND); - insn->header.predicate_control = 0; /* XXX */ + insn->header.predicate_control = BRW_PREDICATE_NONE; insn->header.compression_control = BRW_COMPRESSION_NONE; insn->header.destreg__conditonalmod = msg_reg_nr; + insn->header.mask_control = BRW_MASK_DISABLE; /* cast dest to a uword[8] vector */ dest = retype(vec8(dest), BRW_REGISTER_TYPE_UW); @@ -989,7 +991,7 @@ void brw_dp_READ_4( struct brw_compile *p, brw_set_src0(insn, retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_UW)); brw_set_dp_read_message(insn, - bind_table_index, /* binding table index (255=stateless) */ + bind_table_index, 0, /* msg_control (0 means 1 Oword) */ BRW_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ, /* msg_type */ 0, /* source cache = data cache */ @@ -1000,6 +1002,78 @@ void brw_dp_READ_4( struct brw_compile *p, } +/** + * Read float[4] constant(s) from VS constant buffer. + * For relative addressing, two float[4] constants will be read into 'dest'. + * Otherwise, one float[4] constant will be read into the lower half of 'dest'. + */ +void brw_dp_READ_4_vs(struct brw_compile *p, + struct brw_reg dest, + GLuint oword, + GLboolean relAddr, + struct brw_reg addrReg, + GLuint location, + GLuint bind_table_index) +{ + GLuint msg_reg_nr = 1; + + assert(oword < 2); + /* + printf("vs const read msg, location %u, msg_reg_nr %d\n", + location, msg_reg_nr); + */ + + /* Setup MRF[1] with location/offset into const buffer */ + { + struct brw_reg b; + + brw_push_insn_state(p); + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_set_mask_control(p, BRW_MASK_DISABLE); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + /*brw_set_access_mode(p, BRW_ALIGN_16);*/ + + /* XXX I think we're setting all the dwords of MRF[1] to 'location'. + * when the docs say only dword[2] should be set. Hmmm. But it works. + */ + b = brw_message_reg(msg_reg_nr); + b = retype(b, BRW_REGISTER_TYPE_UD); + /*b = get_element_ud(b, 2);*/ + if (relAddr) { + brw_ADD(p, b, addrReg, brw_imm_ud(location)); + } + else { + brw_MOV(p, b, brw_imm_ud(location)); + } + + brw_pop_insn_state(p); + } + + { + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND); + + insn->header.predicate_control = BRW_PREDICATE_NONE; + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.destreg__conditonalmod = msg_reg_nr; + insn->header.mask_control = BRW_MASK_DISABLE; + /*insn->header.access_mode = BRW_ALIGN_16;*/ + + brw_set_dest(insn, dest); + brw_set_src0(insn, brw_null_reg()); + + brw_set_dp_read_message(insn, + bind_table_index, + oword, /* 0 = lower Oword, 1 = upper Oword */ + BRW_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ, /* msg_type */ + 0, /* source cache = data cache */ + 1, /* msg_length */ + 1, /* response_length (1 Oword) */ + 0); /* eot */ + } +} + + + void brw_fb_WRITE(struct brw_compile *p, struct brw_reg dest, GLuint msg_reg_nr, diff --git a/src/mesa/drivers/dri/i965/brw_misc_state.c b/src/mesa/drivers/dri/i965/brw_misc_state.c index 5c94a49f60a..9bc5c35139c 100644 --- a/src/mesa/drivers/dri/i965/brw_misc_state.c +++ b/src/mesa/drivers/dri/i965/brw_misc_state.c @@ -101,6 +101,7 @@ const struct brw_tracked_state brw_drawing_rect = { static void prepare_binding_table_pointers(struct brw_context *brw) { + brw_add_validated_bo(brw, brw->vs.bind_bo); brw_add_validated_bo(brw, brw->wm.bind_bo); } @@ -117,13 +118,11 @@ static void upload_binding_table_pointers(struct brw_context *brw) BEGIN_BATCH(6, IGNORE_CLIPRECTS); OUT_BATCH(CMD_BINDING_TABLE_PTRS << 16 | (6 - 2)); - OUT_BATCH(0); /* vs */ + OUT_RELOC(brw->vs.bind_bo, I915_GEM_DOMAIN_SAMPLER, 0, 0); /* vs */ OUT_BATCH(0); /* gs */ OUT_BATCH(0); /* clip */ OUT_BATCH(0); /* sf */ - OUT_RELOC(brw->wm.bind_bo, - I915_GEM_DOMAIN_SAMPLER, 0, - 0); + OUT_RELOC(brw->wm.bind_bo, I915_GEM_DOMAIN_SAMPLER, 0, 0); /* wm/ps */ ADVANCE_BATCH(); } diff --git a/src/mesa/drivers/dri/i965/brw_program.c b/src/mesa/drivers/dri/i965/brw_program.c index 457bc2fc7f4..bac69187c19 100644 --- a/src/mesa/drivers/dri/i965/brw_program.c +++ b/src/mesa/drivers/dri/i965/brw_program.c @@ -95,6 +95,12 @@ static struct gl_program *brwNewProgram( GLcontext *ctx, static void brwDeleteProgram( GLcontext *ctx, struct gl_program *prog ) { + if (prog->Target == GL_FRAGMENT_PROGRAM_ARB) { + struct gl_fragment_program *fprog = (struct gl_fragment_program *) prog; + struct brw_fragment_program *brw_fprog = brw_fragment_program(fprog); + dri_bo_unreference(brw_fprog->const_buffer); + } + _mesa_delete_program( ctx, prog ); } @@ -111,7 +117,6 @@ static void brwProgramStringNotify( GLcontext *ctx, struct gl_program *prog ) { struct brw_context *brw = brw_context(ctx); - struct intel_context *intel = &brw->intel; if (target == GL_FRAGMENT_PROGRAM_ARB) { struct gl_fragment_program *fprog = (struct gl_fragment_program *) prog; @@ -128,24 +133,6 @@ static void brwProgramStringNotify( GLcontext *ctx, brw->state.dirty.brw |= BRW_NEW_FRAGMENT_PROGRAM; newFP->id = brw->program_id++; newFP->isGLSL = brw_wm_is_glsl(fprog); - - /* alloc constant buffer/surface */ - { - const struct gl_program_parameter_list *params = prog->Parameters; - const int size = params->NumParameters * 4 * sizeof(GLfloat); - - /* free old const buffer if too small */ - if (newFP->const_buffer && newFP->const_buffer->size < size) { - dri_bo_unreference(newFP->const_buffer); - newFP->const_buffer = NULL; - } - - if (!newFP->const_buffer) { - newFP->const_buffer = drm_intel_bo_alloc(intel->bufmgr, - "fp_const_buffer", - size, 64); - } - } } else if (target == GL_VERTEX_PROGRAM_ARB) { struct gl_vertex_program *vprog = (struct gl_vertex_program *) prog; diff --git a/src/mesa/drivers/dri/i965/brw_state_dump.c b/src/mesa/drivers/dri/i965/brw_state_dump.c index 5d332d010c2..a7132622696 100644 --- a/src/mesa/drivers/dri/i965/brw_state_dump.c +++ b/src/mesa/drivers/dri/i965/brw_state_dump.c @@ -84,6 +84,19 @@ get_965_surfacetype(unsigned int surfacetype) } } +static const char * +get_965_surface_format(unsigned int surface_format) +{ + switch (surface_format) { + case 0x000: return "r32g32b32a32_float"; + case 0x0c1: return "b8g8r8a8_unorm"; + case 0x100: return "b5g6r5_unorm"; + case 0x102: return "b5g5r5a1_unorm"; + case 0x104: return "b4g4r4a4_unorm"; + default: return "unknown"; + } +} + static void dump_wm_surface_state(struct brw_context *brw) { int i; @@ -95,7 +108,7 @@ static void dump_wm_surface_state(struct brw_context *brw) char name[20]; if (surf_bo == NULL) { - fprintf(stderr, "WM SS%d: NULL\n", i); + fprintf(stderr, " WM SS%d: NULL\n", i); continue; } dri_bo_map(surf_bo, GL_FALSE); @@ -103,8 +116,9 @@ static void dump_wm_surface_state(struct brw_context *brw) surf = (struct brw_surface_state *)(surf_bo->virtual); sprintf(name, "WM SS%d", i); - state_out(name, surf, surfoff, 0, "%s\n", - get_965_surfacetype(surf->ss0.surface_type)); + state_out(name, surf, surfoff, 0, "%s %s\n", + get_965_surfacetype(surf->ss0.surface_type), + get_965_surface_format(surf->ss0.surface_format)); state_out(name, surf, surfoff, 1, "offset\n"); state_out(name, surf, surfoff, 2, "%dx%d size, %d mips\n", surf->ss2.width + 1, surf->ss2.height + 1, surf->ss2.mip_count); diff --git a/src/mesa/drivers/dri/i965/brw_vs.h b/src/mesa/drivers/dri/i965/brw_vs.h index 99d0e937226..d20cf78b8af 100644 --- a/src/mesa/drivers/dri/i965/brw_vs.h +++ b/src/mesa/drivers/dri/i965/brw_vs.h @@ -75,6 +75,13 @@ struct brw_vs_compile { struct brw_reg userplane[6]; + /** using a real constant buffer? */ + GLboolean use_const_buffer; + /** we may need up to 3 constants per instruction (if use_const_buffer) */ + struct { + GLint index; + struct brw_reg reg; + } current_const[3]; }; void brw_vs_emit( struct brw_vs_compile *c ); diff --git a/src/mesa/drivers/dri/i965/brw_vs_constval.c b/src/mesa/drivers/dri/i965/brw_vs_constval.c index d29eb17f8cf..2637344b482 100644 --- a/src/mesa/drivers/dri/i965/brw_vs_constval.c +++ b/src/mesa/drivers/dri/i965/brw_vs_constval.c @@ -96,7 +96,7 @@ static GLubyte get_active( struct tracker *t, struct prog_src_register src ) { GLuint i; - GLubyte active = src.NegateBase; /* NOTE! */ + GLubyte active = src.Negate; /* NOTE! */ if (src.RelAddr) return 0xf; diff --git a/src/mesa/drivers/dri/i965/brw_vs_emit.c b/src/mesa/drivers/dri/i965/brw_vs_emit.c index 0d6c6ab9a8a..524f1211cee 100644 --- a/src/mesa/drivers/dri/i965/brw_vs_emit.c +++ b/src/mesa/drivers/dri/i965/brw_vs_emit.c @@ -38,14 +38,44 @@ #include "brw_vs.h" +static struct brw_reg get_tmp( struct brw_vs_compile *c ) +{ + struct brw_reg tmp = brw_vec8_grf(c->last_tmp, 0); + + if (++c->last_tmp > c->prog_data.total_grf) + c->prog_data.total_grf = c->last_tmp; + + return tmp; +} -/* Do things as simply as possible. Allocate and populate all regs +static void release_tmp( struct brw_vs_compile *c, struct brw_reg tmp ) +{ + if (tmp.nr == c->last_tmp-1) + c->last_tmp--; +} + +static void release_tmps( struct brw_vs_compile *c ) +{ + c->last_tmp = c->first_tmp; +} + + +/** + * Preallocate GRF register before code emit. + * Do things as simply as possible. Allocate and populate all regs * ahead of time. */ static void brw_vs_alloc_regs( struct brw_vs_compile *c ) { GLuint i, reg = 0, mrf; - GLuint nr_params; + +#if 0 + if (c->vp->program.Base.Parameters->NumParameters >= 6) + c->use_const_buffer = 1; + else +#endif + c->use_const_buffer = GL_FALSE; + /*printf("use_const_buffer = %d\n", c->use_const_buffer);*/ /* r0 -- reserved as usual */ @@ -66,13 +96,22 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c ) /* Vertex program parameters from curbe: */ - nr_params = c->vp->program.Base.Parameters->NumParameters; - for (i = 0; i < nr_params; i++) { - c->regs[PROGRAM_STATE_VAR][i] = stride( brw_vec4_grf(reg+i/2, (i%2) * 4), 0, 4, 1); - } - reg += (nr_params + 1) / 2; + if (c->use_const_buffer) { + /* get constants from a real constant buffer */ + c->prog_data.curb_read_length = 0; + c->prog_data.nr_params = 4; /* XXX 0 causes a bug elsewhere... */ + } + else { + /* use a section of the GRF for constants */ + GLuint nr_params = c->vp->program.Base.Parameters->NumParameters; + for (i = 0; i < nr_params; i++) { + c->regs[PROGRAM_STATE_VAR][i] = stride( brw_vec4_grf(reg+i/2, (i%2) * 4), 0, 4, 1); + } + reg += (nr_params + 1) / 2; + c->prog_data.curb_read_length = reg - 1; - c->prog_data.curb_read_length = reg - 1; + c->prog_data.nr_params = nr_params * 4; + } /* Allocate input regs: */ @@ -133,6 +172,14 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c ) reg++; } + if (c->use_const_buffer) { + for (i = 0; i < 3; i++) { + c->current_const[i].index = -1; + c->current_const[i].reg = brw_vec8_grf(reg, 0); + reg++; + } + } + for (i = 0; i < 128; i++) { if (c->output_regs[i].used_in_src) { c->output_regs[i].reg = brw_vec8_grf(reg, 0); @@ -165,28 +212,6 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c ) } -static struct brw_reg get_tmp( struct brw_vs_compile *c ) -{ - struct brw_reg tmp = brw_vec8_grf(c->last_tmp, 0); - - if (++c->last_tmp > c->prog_data.total_grf) - c->prog_data.total_grf = c->last_tmp; - - return tmp; -} - -static void release_tmp( struct brw_vs_compile *c, struct brw_reg tmp ) -{ - if (tmp.nr == c->last_tmp-1) - c->last_tmp--; -} - -static void release_tmps( struct brw_vs_compile *c ) -{ - c->last_tmp = c->first_tmp; -} - - /** * If an instruction uses a temp reg both as a src and the dest, we * sometimes need to allocate an intermediate temporary. @@ -633,6 +658,8 @@ static void emit_lit_noalias( struct brw_vs_compile *c, } brw_ENDIF(p, if_insn); + + release_tmp(c, tmp); } static void emit_lrp_noalias(struct brw_vs_compile *c, @@ -673,13 +700,83 @@ static void emit_nrm( struct brw_vs_compile *c, } +static struct brw_reg +get_constant(struct brw_vs_compile *c, + const struct prog_instruction *inst, + GLuint argIndex) +{ + const struct prog_src_register *src = &inst->SrcReg[argIndex]; + struct brw_compile *p = &c->func; + struct brw_reg const_reg; + struct brw_reg const2_reg; + + assert(argIndex < 3); + + if (c->current_const[argIndex].index != src->Index || src->RelAddr) { + struct brw_reg addrReg = c->regs[PROGRAM_ADDRESS][0]; + + c->current_const[argIndex].index = src->Index; + +#if 0 + printf(" fetch const[%d] for arg %d into reg %d\n", + src->Index, argIndex, c->current_const[argIndex].reg.nr); +#endif + /* need to fetch the constant now */ + brw_dp_READ_4_vs(p, + c->current_const[argIndex].reg,/* writeback dest */ + 0, /* oword */ + src->RelAddr, /* relative indexing? */ + addrReg, /* address register */ + 16 * src->Index, /* byte offset */ + SURF_INDEX_VERT_CONST_BUFFER /* binding table index */ + ); + + if (src->RelAddr) { + /* second read */ + const2_reg = get_tmp(c); + + /* use upper half of address reg for second read */ + addrReg = stride(addrReg, 0, 4, 0); + addrReg.subnr = 16; + + brw_dp_READ_4_vs(p, + const2_reg, /* writeback dest */ + 1, /* oword */ + src->RelAddr, /* relative indexing? */ + addrReg, /* address register */ + 16 * src->Index, /* byte offset */ + SURF_INDEX_VERT_CONST_BUFFER + ); + } + } + + const_reg = c->current_const[argIndex].reg; + + if (src->RelAddr) { + /* merge the two Owords into the constant register */ + /* const_reg[7..4] = const2_reg[7..4] */ + brw_MOV(p, + suboffset(stride(const_reg, 0, 4, 1), 4), + suboffset(stride(const2_reg, 0, 4, 1), 4)); + release_tmp(c, const2_reg); + } + else { + /* replicate lower four floats into upper half (to get XYZWXYZW) */ + const_reg = stride(const_reg, 0, 4, 0); + const_reg.subnr = 0; + } + + return const_reg; +} + + + /* TODO: relative addressing! */ static struct brw_reg get_reg( struct brw_vs_compile *c, gl_register_file file, GLuint index ) { - switch (file) { case PROGRAM_TEMPORARY: case PROGRAM_INPUT: @@ -708,13 +805,17 @@ static struct brw_reg get_reg( struct brw_vs_compile *c, } +/** + * Indirect addressing: get reg[[arg] + offset]. + */ static struct brw_reg deref( struct brw_vs_compile *c, struct brw_reg arg, GLint offset) { struct brw_compile *p = &c->func; struct brw_reg tmp = vec4(get_tmp(c)); - struct brw_reg vp_address = retype(vec1(get_reg(c, PROGRAM_ADDRESS, 0)), BRW_REGISTER_TYPE_UW); + struct brw_reg addr_reg = c->regs[PROGRAM_ADDRESS][0]; + struct brw_reg vp_address = retype(vec1(addr_reg), BRW_REGISTER_TYPE_UW); GLuint byte_offset = arg.nr * 32 + arg.subnr + offset * 16; struct brw_reg indirect = brw_vec4_indirect(0,0); @@ -735,10 +836,67 @@ static struct brw_reg deref( struct brw_vs_compile *c, brw_pop_insn_state(p); } + /* NOTE: tmp not released */ return vec8(tmp); } +/** + * Get brw reg corresponding to the instruction's [argIndex] src reg. + * TODO: relative addressing! + */ +static struct brw_reg +get_src_reg( struct brw_vs_compile *c, + const struct prog_instruction *inst, + GLuint argIndex ) +{ + const GLuint file = inst->SrcReg[argIndex].File; + const GLint index = inst->SrcReg[argIndex].Index; + const GLboolean relAddr = inst->SrcReg[argIndex].RelAddr; + + switch (file) { + case PROGRAM_TEMPORARY: + case PROGRAM_INPUT: + case PROGRAM_OUTPUT: + if (relAddr) { + return deref(c, c->regs[file][0], index); + } + else { + assert(c->regs[file][index].nr != 0); + return c->regs[file][index]; + } + + case PROGRAM_STATE_VAR: + case PROGRAM_CONSTANT: + case PROGRAM_UNIFORM: + if (c->use_const_buffer) { + return get_constant(c, inst, argIndex); + } + else if (relAddr) { + return deref(c, c->regs[PROGRAM_STATE_VAR][0], index); + } + else { + assert(c->regs[PROGRAM_STATE_VAR][index].nr != 0); + return c->regs[PROGRAM_STATE_VAR][index]; + } + case PROGRAM_ADDRESS: + assert(index == 0); + return c->regs[file][index]; + + case PROGRAM_UNDEFINED: + /* this is a normal case since we loop over all three src args */ + return brw_null_reg(); + + case PROGRAM_LOCAL_PARAM: + case PROGRAM_ENV_PARAM: + case PROGRAM_WRITE_ONLY: + default: + assert(0); + return brw_null_reg(); + } +} + + static void emit_arl( struct brw_vs_compile *c, struct brw_reg dst, struct brw_reg arg0 ) @@ -750,30 +908,31 @@ static void emit_arl( struct brw_vs_compile *c, if (need_tmp) tmp = get_tmp(c); - brw_RNDD(p, tmp, arg0); - brw_MUL(p, dst, tmp, brw_imm_d(16)); + brw_RNDD(p, tmp, arg0); /* tmp = round(arg0) */ + brw_MUL(p, dst, tmp, brw_imm_d(16)); /* dst = tmp * 16 */ if (need_tmp) release_tmp(c, tmp); } -/* Will return mangled results for SWZ op. The emit_swz() function +/** + * Return the brw reg for the given instruction's src argument. + * Will return mangled results for SWZ op. The emit_swz() function * ignores this result and recalculates taking extended swizzles into * account. */ static struct brw_reg get_arg( struct brw_vs_compile *c, - struct prog_src_register *src ) + const struct prog_instruction *inst, + GLuint argIndex ) { + const struct prog_src_register *src = &inst->SrcReg[argIndex]; struct brw_reg reg; if (src->File == PROGRAM_UNDEFINED) return brw_null_reg(); - if (src->RelAddr) - reg = deref(c, c->regs[PROGRAM_STATE_VAR][0], src->Index); - else - reg = get_reg(c, src->File, src->Index); + reg = get_src_reg(c, inst, argIndex); /* Convert 3-bit swizzle to 2-bit. */ @@ -784,16 +943,38 @@ static struct brw_reg get_arg( struct brw_vs_compile *c, /* Note this is ok for non-swizzle instructions: */ - reg.negate = src->NegateBase ? 1 : 0; + reg.negate = src->Negate ? 1 : 0; return reg; } +/** + * Get brw register for the given program dest register. + */ static struct brw_reg get_dst( struct brw_vs_compile *c, struct prog_dst_register dst ) { - struct brw_reg reg = get_reg(c, dst.File, dst.Index); + struct brw_reg reg; + + switch (dst.File) { + case PROGRAM_TEMPORARY: + case PROGRAM_OUTPUT: + assert(c->regs[dst.File][dst.Index].nr != 0); + reg = c->regs[dst.File][dst.Index]; + break; + case PROGRAM_ADDRESS: + assert(dst.Index == 0); + reg = c->regs[dst.File][dst.Index]; + break; + case PROGRAM_UNDEFINED: + /* we may hit this for OPCODE_END, OPCODE_KIL, etc */ + reg = brw_null_reg(); + break; + default: + assert(0); + reg = brw_null_reg(); + } reg.dw1.bits.writemask = dst.WriteMask; @@ -803,14 +984,16 @@ static struct brw_reg get_dst( struct brw_vs_compile *c, static void emit_swz( struct brw_vs_compile *c, struct brw_reg dst, - struct prog_src_register src ) + const struct prog_instruction *inst) { + const GLuint argIndex = 0; + const struct prog_src_register src = inst->SrcReg[argIndex]; struct brw_compile *p = &c->func; GLuint zeros_mask = 0; GLuint ones_mask = 0; GLuint src_mask = 0; GLubyte src_swz[4]; - GLboolean need_tmp = (src.NegateBase && + GLboolean need_tmp = (src.Negate && dst.file != BRW_GENERAL_REGISTER_FILE); struct brw_reg tmp = dst; GLuint i; @@ -844,10 +1027,7 @@ static void emit_swz( struct brw_vs_compile *c, if (src_mask) { struct brw_reg arg0; - if (src.RelAddr) - arg0 = deref(c, c->regs[PROGRAM_STATE_VAR][0], src.Index); - else - arg0 = get_reg(c, src.File, src.Index); + arg0 = get_src_reg(c, inst, argIndex); arg0 = brw_swizzle(arg0, src_swz[0], src_swz[1], @@ -862,8 +1042,8 @@ static void emit_swz( struct brw_vs_compile *c, if (ones_mask) brw_MOV(p, brw_writemask(tmp, ones_mask), brw_imm_f(1)); - if (src.NegateBase) - brw_MOV(p, brw_writemask(tmp, src.NegateBase), negate(tmp)); + if (src.Negate) + brw_MOV(p, brw_writemask(tmp, src.Negate), negate(tmp)); if (need_tmp) { brw_MOV(p, dst, tmp); @@ -1043,6 +1223,11 @@ void brw_vs_emit(struct brw_vs_compile *c ) struct brw_reg args[3], dst; GLuint i; +#if 0 + printf("%d: ", insn); + _mesa_print_instruction(inst); +#endif + /* Get argument regs. SWZ is special and does this itself. */ if (inst->Opcode != OPCODE_SWZ) @@ -1053,7 +1238,7 @@ void brw_vs_emit(struct brw_vs_compile *c ) if (file == PROGRAM_OUTPUT && c->output_regs[index].used_in_src) args[i] = c->output_regs[index].reg; else - args[i] = get_arg(c, src); + args[i] = get_arg(c, inst, i); } /* Get dest regs. Note that it is possible for a reg to be both @@ -1181,7 +1366,7 @@ void brw_vs_emit(struct brw_vs_compile *c ) /* The args[0] value can't be used here as it won't have * correctly encoded the full swizzle: */ - emit_swz(c, dst, inst->SrcReg[0] ); + emit_swz(c, dst, inst); break; case OPCODE_TRUNC: /* round toward zero */ diff --git a/src/mesa/drivers/dri/i965/brw_vs_state.c b/src/mesa/drivers/dri/i965/brw_vs_state.c index 1a63766ea1f..3d295388437 100644 --- a/src/mesa/drivers/dri/i965/brw_vs_state.c +++ b/src/mesa/drivers/dri/i965/brw_vs_state.c @@ -44,6 +44,8 @@ struct brw_vs_unit_key { unsigned int curbe_offset; unsigned int nr_urb_entries, urb_size; + + unsigned int nr_surfaces; }; static void @@ -62,6 +64,9 @@ vs_unit_populate_key(struct brw_context *brw, struct brw_vs_unit_key *key) key->nr_urb_entries = brw->urb.nr_vs_entries; key->urb_size = brw->urb.vsize; + /* BRW_NEW_NR_VS_SURFACES */ + key->nr_surfaces = brw->vs.nr_surfaces; + /* BRW_NEW_CURBE_OFFSETS, _NEW_TRANSFORM */ if (ctx->Transform.ClipPlanesEnabled) { /* Note that we read in the userclip planes as well, hence @@ -92,6 +97,8 @@ vs_unit_create_from_key(struct brw_context *brw, struct brw_vs_unit_key *key) * brw_urb_WRITE() results. */ vs.thread1.single_program_flow = 0; + vs.thread1.binding_table_entry_count = key->nr_surfaces; + vs.thread3.urb_entry_read_length = key->urb_entry_read_length; vs.thread3.const_urb_entry_read_length = key->curb_entry_read_length; vs.thread3.dispatch_grf_start_reg = 1; @@ -158,6 +165,7 @@ const struct brw_tracked_state brw_vs_unit = { .dirty = { .mesa = _NEW_TRANSFORM, .brw = (BRW_NEW_CURBE_OFFSETS | + BRW_NEW_NR_VS_SURFACES | BRW_NEW_URB_FENCE), .cache = CACHE_NEW_VS_PROG }, diff --git a/src/mesa/drivers/dri/i965/brw_vtbl.c b/src/mesa/drivers/dri/i965/brw_vtbl.c index 960bbb311e3..ba03afd6c13 100644 --- a/src/mesa/drivers/dri/i965/brw_vtbl.c +++ b/src/mesa/drivers/dri/i965/brw_vtbl.c @@ -79,6 +79,7 @@ static void brw_destroy_context( struct intel_context *intel ) dri_bo_release(&brw->curbe.curbe_bo); dri_bo_release(&brw->vs.prog_bo); dri_bo_release(&brw->vs.state_bo); + dri_bo_release(&brw->vs.bind_bo); dri_bo_release(&brw->gs.prog_bo); dri_bo_release(&brw->gs.state_bo); dri_bo_release(&brw->clip.prog_bo); diff --git a/src/mesa/drivers/dri/i965/brw_wm_emit.c b/src/mesa/drivers/dri/i965/brw_wm_emit.c index d65b1332c61..72fc21d2eba 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_emit.c +++ b/src/mesa/drivers/dri/i965/brw_wm_emit.c @@ -742,7 +742,7 @@ static void emit_tex( struct brw_wm_compile *c, retype(vec16(dst[0]), BRW_REGISTER_TYPE_UW), 1, retype(c->payload.depth[0].hw_reg, BRW_REGISTER_TYPE_UW), - inst->tex_unit + MAX_DRAW_BUFFERS, /* surface */ + SURF_INDEX_TEXTURE(inst->tex_unit), inst->tex_unit, /* sampler */ inst->writemask, (inst->tex_shadow ? @@ -791,7 +791,7 @@ static void emit_txb( struct brw_wm_compile *c, retype(vec16(dst[0]), BRW_REGISTER_TYPE_UW), 1, retype(c->payload.depth[0].hw_reg, BRW_REGISTER_TYPE_UW), - inst->tex_unit + MAX_DRAW_BUFFERS, /* surface */ + SURF_INDEX_TEXTURE(inst->tex_unit), inst->tex_unit, /* sampler */ inst->writemask, BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS, diff --git a/src/mesa/drivers/dri/i965/brw_wm_fp.c b/src/mesa/drivers/dri/i965/brw_wm_fp.c index a7f5f1b9a28..1798d842c79 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_fp.c +++ b/src/mesa/drivers/dri/i965/brw_wm_fp.c @@ -80,9 +80,8 @@ static struct prog_src_register src_reg(GLuint file, GLuint idx) reg.Index = idx; reg.Swizzle = SWIZZLE_NOOP; reg.RelAddr = 0; - reg.NegateBase = 0; + reg.Negate = NEGATE_NONE; reg.Abs = 0; - reg.NegateAbs = 0; return reg; } @@ -569,7 +568,7 @@ static void precalc_dst( struct brw_wm_compile *c, src_undef(), src_undef()); /* Avoid letting negation flag of src0 affect our 1 constant. */ - swz->SrcReg[0].NegateBase &= ~NEGATE_X; + swz->SrcReg[0].Negate &= ~NEGATE_X; } if (dst.WriteMask & WRITEMASK_W) { /* dst.w = mov src1.w @@ -604,7 +603,7 @@ static void precalc_lit( struct brw_wm_compile *c, src_undef(), src_undef()); /* Avoid letting the negation flag of src0 affect our 1 constant. */ - swz->SrcReg[0].NegateBase = 0; + swz->SrcReg[0].Negate = NEGATE_NONE; } if (dst.WriteMask & WRITEMASK_YZ) { @@ -651,7 +650,7 @@ static void precalc_tex( struct brw_wm_compile *c, src0, src_undef(), src_undef()); - out->SrcReg[0].NegateBase = 0; + out->SrcReg[0].Negate = NEGATE_NONE; out->SrcReg[0].Abs = 1; /* tmp0 = MAX(coord.X, coord.Y) */ @@ -1050,14 +1049,14 @@ void brw_wm_pass_fp( struct brw_wm_compile *c ) case OPCODE_ABS: out = emit_insn(c, inst); out->Opcode = OPCODE_MOV; - out->SrcReg[0].NegateBase = 0; + out->SrcReg[0].Negate = NEGATE_NONE; out->SrcReg[0].Abs = 1; break; case OPCODE_SUB: out = emit_insn(c, inst); out->Opcode = OPCODE_ADD; - out->SrcReg[1].NegateBase ^= 0xf; + out->SrcReg[1].Negate ^= NEGATE_XYZW; break; case OPCODE_SCS: diff --git a/src/mesa/drivers/dri/i965/brw_wm_glsl.c b/src/mesa/drivers/dri/i965/brw_wm_glsl.c index 575cd45d572..22e17622c6d 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_glsl.c +++ b/src/mesa/drivers/dri/i965/brw_wm_glsl.c @@ -56,7 +56,7 @@ static void set_reg(struct brw_wm_compile *c, int file, int index, * Examine instruction's write mask to find index of first component * enabled for writing. */ -static int get_scalar_dst_index(struct prog_instruction *inst) +static int get_scalar_dst_index(const struct prog_instruction *inst) { int i; for (i = 0; i < 4; i++) @@ -254,14 +254,15 @@ static void prealloc_reg(struct brw_wm_compile *c) * XXX alloc these on demand! */ if (c->use_const_buffer) { - c->current_const[0].reg = alloc_tmp(c); - c->current_const[1].reg = alloc_tmp(c); - c->current_const[2].reg = alloc_tmp(c); + for (i = 0; i < 3; i++) { + c->current_const[i].index = -1; + c->current_const[i].reg = alloc_tmp(c); + } } - /* +#if 0 printf("USE CONST BUFFER? %d\n", c->use_const_buffer); printf("AFTER PRE_ALLOC, reg_index = %d\n", c->reg_index); - */ +#endif } @@ -283,14 +284,12 @@ static void fetch_constants(struct brw_wm_compile *c, src->File == PROGRAM_CONSTANT || src->File == PROGRAM_UNIFORM) { if (c->current_const[i].index != src->Index) { - c->current_const[i].index = src->Index; - /*c->current_const[i].reg = alloc_tmp(c);*/ - /* +#if 0 printf(" fetch const[%d] for arg %d into reg %d\n", src->Index, i, c->current_const[i].reg.nr); - */ +#endif /* need to fetch the constant now */ brw_dp_READ_4(p, @@ -298,28 +297,8 @@ static void fetch_constants(struct brw_wm_compile *c, 1, /* msg_reg */ src->RelAddr, /* relative indexing? */ 16 * src->Index, /* byte offset */ - BRW_WM_MAX_SURF - 1 /* binding table index */ + SURF_INDEX_FRAG_CONST_BUFFER/* binding table index */ ); - -#if 0 - /* dependency stall */ - { - int response_length = 1; - int mark = mark_tmps( c ); - struct brw_reg src = c->current_const[i].reg; - struct brw_reg tmp = alloc_tmp(c); - - /* mov (8) r9.0<1>:f r9.0<8;8,1>:f { Align1 } - */ - brw_push_insn_state(p); - brw_set_compression_control(p, BRW_COMPRESSION_NONE); - brw_MOV(p, tmp, src); - brw_MOV(p, src, tmp); - brw_pop_insn_state(p); - - release_tmps( c, mark ); - } -#endif } } } @@ -361,18 +340,18 @@ get_src_reg_const(struct brw_wm_compile *c, const_reg = stride(const_reg, 0, 1, 0); const_reg.subnr = component * 4; - if (src->NegateBase) + if (src->Negate & (1 << component)) const_reg = negate(const_reg); if (src->Abs) const_reg = brw_abs(const_reg); - /* - printf(" form const[%d] for arg %d, comp %d, reg %d\n", +#if 0 + printf(" form const[%d].%d for arg %d, reg %d\n", c->current_const[srcRegIndex].index, - srcRegIndex, component, + srcRegIndex, const_reg.nr); - */ +#endif return const_reg; } @@ -398,7 +377,7 @@ static struct brw_reg get_src_reg(struct brw_wm_compile *c, else { /* other type of source register */ return get_reg(c, src->File, src->Index, component, nr, - src->NegateBase, src->Abs); + src->Negate, src->Abs); } } @@ -423,11 +402,13 @@ static struct brw_reg get_src_reg_imm(struct brw_wm_compile *c, const GLfloat *param = c->fp->program.Base.Parameters->ParameterValues[src->Index]; GLfloat value = param[component]; - if (src->NegateBase) + if (src->Negate & (1 << channel)) value = -value; if (src->Abs) value = FABSF(value); - /*printf(" form imm reg %f\n", value);*/ +#if 0 + printf(" form immed value %f for chan %d\n", value, channel); +#endif return brw_imm_f(value); } else { @@ -501,7 +482,7 @@ static void invoke_subroutine( struct brw_wm_compile *c, } static void emit_abs( struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { int i; struct brw_compile *p = &c->func; @@ -518,7 +499,7 @@ static void emit_abs( struct brw_wm_compile *c, } static void emit_trunc( struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { int i; struct brw_compile *p = &c->func; @@ -536,7 +517,7 @@ static void emit_trunc( struct brw_wm_compile *c, } static void emit_mov( struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { int i; struct brw_compile *p = &c->func; @@ -546,7 +527,9 @@ static void emit_mov( struct brw_wm_compile *c, if (mask & (1<<i)) { struct brw_reg src, dst; dst = get_dst_reg(c, inst, i); - src = get_src_reg_imm(c, inst, 0, i); + /* XXX some moves from immediate value don't work reliably!!! */ + /*src = get_src_reg_imm(c, inst, 0, i);*/ + src = get_src_reg(c, inst, 0, i); brw_MOV(p, dst, src); } } @@ -554,7 +537,7 @@ static void emit_mov( struct brw_wm_compile *c, } static void emit_pixel_xy(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { struct brw_reg r1 = brw_vec1_grf(1, 0); struct brw_reg r1_uw = retype(r1, BRW_REGISTER_TYPE_UW); @@ -584,7 +567,7 @@ static void emit_pixel_xy(struct brw_wm_compile *c, } static void emit_delta_xy(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { struct brw_reg r1 = brw_vec1_grf(1, 0); struct brw_reg dst0, dst1, src0, src1; @@ -644,7 +627,7 @@ static void fire_fb_write( struct brw_wm_compile *c, } static void emit_fb_write(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { struct brw_compile *p = &c->func; int nr = 2; @@ -713,7 +696,7 @@ static void emit_fb_write(struct brw_wm_compile *c, } static void emit_pixel_w( struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { struct brw_compile *p = &c->func; GLuint mask = inst->DstReg.WriteMask; @@ -743,7 +726,7 @@ static void emit_pixel_w( struct brw_wm_compile *c, } static void emit_linterp(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { struct brw_compile *p = &c->func; GLuint mask = inst->DstReg.WriteMask; @@ -772,7 +755,7 @@ static void emit_linterp(struct brw_wm_compile *c, } static void emit_cinterp(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { struct brw_compile *p = &c->func; GLuint mask = inst->DstReg.WriteMask; @@ -798,7 +781,7 @@ static void emit_cinterp(struct brw_wm_compile *c, } static void emit_pinterp(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { struct brw_compile *p = &c->func; GLuint mask = inst->DstReg.WriteMask; @@ -832,7 +815,7 @@ static void emit_pinterp(struct brw_wm_compile *c, /* Sets the destination channels to 1.0 or 0.0 according to glFrontFacing. */ static void emit_frontfacing(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { struct brw_compile *p = &c->func; struct brw_reg r1_6ud = retype(brw_vec1_grf(1, 6), BRW_REGISTER_TYPE_UD); @@ -861,7 +844,7 @@ static void emit_frontfacing(struct brw_wm_compile *c, } static void emit_xpd(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { int i; struct brw_compile *p = &c->func; @@ -886,7 +869,7 @@ static void emit_xpd(struct brw_wm_compile *c, } static void emit_dp3(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { struct brw_reg src0[3], src1[3], dst; int i; @@ -905,7 +888,7 @@ static void emit_dp3(struct brw_wm_compile *c, } static void emit_dp4(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { struct brw_reg src0[4], src1[4], dst; int i; @@ -924,7 +907,7 @@ static void emit_dp4(struct brw_wm_compile *c, } static void emit_dph(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { struct brw_reg src0[4], src1[4], dst; int i; @@ -948,7 +931,7 @@ static void emit_dph(struct brw_wm_compile *c, * register's X, Y, Z and W channels (subject to writemasking of course). */ static void emit_math1(struct brw_wm_compile *c, - struct prog_instruction *inst, GLuint func) + const struct prog_instruction *inst, GLuint func) { struct brw_compile *p = &c->func; struct brw_reg src0, dst, tmp; @@ -985,43 +968,43 @@ static void emit_math1(struct brw_wm_compile *c, } static void emit_rcp(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { emit_math1(c, inst, BRW_MATH_FUNCTION_INV); } static void emit_rsq(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { emit_math1(c, inst, BRW_MATH_FUNCTION_RSQ); } static void emit_sin(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { emit_math1(c, inst, BRW_MATH_FUNCTION_SIN); } static void emit_cos(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { emit_math1(c, inst, BRW_MATH_FUNCTION_COS); } static void emit_ex2(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { emit_math1(c, inst, BRW_MATH_FUNCTION_EXP); } static void emit_lg2(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { emit_math1(c, inst, BRW_MATH_FUNCTION_LOG); } static void emit_add(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { struct brw_compile *p = &c->func; struct brw_reg src0, src1, dst; @@ -1040,7 +1023,7 @@ static void emit_add(struct brw_wm_compile *c, } static void emit_arl(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { struct brw_compile *p = &c->func; struct brw_reg src0, addr_reg; @@ -1053,7 +1036,7 @@ static void emit_arl(struct brw_wm_compile *c, } static void emit_sub(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { struct brw_compile *p = &c->func; struct brw_reg src0, src1, dst; @@ -1072,7 +1055,7 @@ static void emit_sub(struct brw_wm_compile *c, } static void emit_mul(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { struct brw_compile *p = &c->func; struct brw_reg src0, src1, dst; @@ -1091,7 +1074,7 @@ static void emit_mul(struct brw_wm_compile *c, } static void emit_frc(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { struct brw_compile *p = &c->func; struct brw_reg src0, dst; @@ -1110,7 +1093,7 @@ static void emit_frc(struct brw_wm_compile *c, } static void emit_flr(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { struct brw_compile *p = &c->func; struct brw_reg src0, dst; @@ -1177,7 +1160,7 @@ static void emit_min_max(struct brw_wm_compile *c, } static void emit_pow(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { struct brw_compile *p = &c->func; struct brw_reg dst, src0, src1; @@ -1199,7 +1182,7 @@ static void emit_pow(struct brw_wm_compile *c, } static void emit_lrp(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { struct brw_compile *p = &c->func; GLuint mask = inst->DstReg.WriteMask; @@ -1252,7 +1235,7 @@ static void emit_kil(struct brw_wm_compile *c) } static void emit_mad(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { struct brw_compile *p = &c->func; GLuint mask = inst->DstReg.WriteMask; @@ -1275,7 +1258,7 @@ static void emit_mad(struct brw_wm_compile *c, } static void emit_sop(struct brw_wm_compile *c, - struct prog_instruction *inst, GLuint cond) + const struct prog_instruction *inst, GLuint cond) { struct brw_compile *p = &c->func; GLuint mask = inst->DstReg.WriteMask; @@ -1299,43 +1282,43 @@ static void emit_sop(struct brw_wm_compile *c, } static void emit_slt(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { emit_sop(c, inst, BRW_CONDITIONAL_L); } static void emit_sle(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { emit_sop(c, inst, BRW_CONDITIONAL_LE); } static void emit_sgt(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { emit_sop(c, inst, BRW_CONDITIONAL_G); } static void emit_sge(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { emit_sop(c, inst, BRW_CONDITIONAL_GE); } static void emit_seq(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { emit_sop(c, inst, BRW_CONDITIONAL_EQ); } static void emit_sne(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { emit_sop(c, inst, BRW_CONDITIONAL_NEQ); } static void emit_ddx(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { struct brw_compile *p = &c->func; GLuint mask = inst->DstReg.WriteMask; @@ -1362,7 +1345,7 @@ static void emit_ddx(struct brw_wm_compile *c, } static void emit_ddy(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { struct brw_compile *p = &c->func; GLuint mask = inst->DstReg.WriteMask; @@ -1505,7 +1488,7 @@ static void noise1_sub( struct brw_wm_compile *c ) { } static void emit_noise1( struct brw_wm_compile *c, - struct prog_instruction *inst ) + const struct prog_instruction *inst ) { struct brw_compile *p = &c->func; struct brw_reg src, param, dst; @@ -1675,7 +1658,7 @@ static void noise2_sub( struct brw_wm_compile *c ) { } static void emit_noise2( struct brw_wm_compile *c, - struct prog_instruction *inst ) + const struct prog_instruction *inst ) { struct brw_compile *p = &c->func; struct brw_reg src0, src1, param0, param1, dst; @@ -1978,7 +1961,7 @@ static void noise3_sub( struct brw_wm_compile *c ) { } static void emit_noise3( struct brw_wm_compile *c, - struct prog_instruction *inst ) + const struct prog_instruction *inst ) { struct brw_compile *p = &c->func; struct brw_reg src0, src1, src2, param0, param1, param2, dst; @@ -2401,7 +2384,7 @@ static void noise4_sub( struct brw_wm_compile *c ) } static void emit_noise4( struct brw_wm_compile *c, - struct prog_instruction *inst ) + const struct prog_instruction *inst ) { struct brw_compile *p = &c->func; struct brw_reg src0, src1, src2, src3, param0, param1, param2, param3, dst; @@ -2443,7 +2426,7 @@ static void emit_noise4( struct brw_wm_compile *c, } static void emit_wpos_xy(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { struct brw_compile *p = &c->func; GLuint mask = inst->DstReg.WriteMask; @@ -2479,7 +2462,7 @@ static void emit_wpos_xy(struct brw_wm_compile *c, BIAS on SIMD8 not working yet... */ static void emit_txb(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { struct brw_compile *p = &c->func; struct brw_reg dst[4], src[4], payload_reg; @@ -2517,7 +2500,7 @@ static void emit_txb(struct brw_wm_compile *c, retype(vec8(dst[0]), BRW_REGISTER_TYPE_UW), /* dest */ 1, /* msg_reg_nr */ retype(payload_reg, BRW_REGISTER_TYPE_UW), /* src0 */ - unit + MAX_DRAW_BUFFERS, /* surface */ + SURF_INDEX_TEXTURE(unit), unit, /* sampler */ inst->DstReg.WriteMask, /* writemask */ BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS, /* msg_type */ @@ -2528,7 +2511,7 @@ static void emit_txb(struct brw_wm_compile *c, static void emit_tex(struct brw_wm_compile *c, - struct prog_instruction *inst) + const struct prog_instruction *inst) { struct brw_compile *p = &c->func; struct brw_reg dst[4], src[4], payload_reg; @@ -2581,7 +2564,7 @@ static void emit_tex(struct brw_wm_compile *c, retype(vec8(dst[0]), BRW_REGISTER_TYPE_UW), /* dest */ 1, /* msg_reg_nr */ retype(payload_reg, BRW_REGISTER_TYPE_UW), /* src0 */ - unit + MAX_DRAW_BUFFERS, /* surface */ + SURF_INDEX_TEXTURE(unit), unit, /* sampler */ inst->DstReg.WriteMask, /* writemask */ BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE, /* msg_type */ @@ -2618,22 +2601,22 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c) brw_MOV(p, get_addr_reg(stack_index), brw_address(c->stack)); for (i = 0; i < c->nr_fp_insns; i++) { - struct prog_instruction *inst = &c->prog_instructions[i]; + const struct prog_instruction *inst = &c->prog_instructions[i]; - if (inst->CondUpdate) - brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); - else - brw_set_conditionalmod(p, BRW_CONDITIONAL_NONE); - - /* +#if 0 _mesa_printf("Inst %d: ", i); _mesa_print_instruction(inst); - */ +#endif /* fetch any constants that this instruction needs */ if (c->use_const_buffer) fetch_constants(c, inst); + if (inst->CondUpdate) + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + else + brw_set_conditionalmod(p, BRW_CONDITIONAL_NONE); + switch (inst->Opcode) { case WM_PIXELXY: emit_pixel_xy(c, inst); @@ -2819,6 +2802,7 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c) break; case OPCODE_BGNLOOP: + /* XXX may need to invalidate the current_constant regs */ loop_inst[loop_insn++] = brw_DO(p, BRW_EXECUTE_8); break; case OPCODE_BRK: diff --git a/src/mesa/drivers/dri/i965/brw_wm_pass0.c b/src/mesa/drivers/dri/i965/brw_wm_pass0.c index 2debd0678a5..92142764f5d 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_pass0.c +++ b/src/mesa/drivers/dri/i965/brw_wm_pass0.c @@ -322,7 +322,7 @@ static struct brw_wm_ref *get_new_ref( struct brw_wm_compile *c, newref->value->lastuse = newref; } - if (src.NegateBase & (1<<i)) + if (src.Negate & (1 << i)) newref->hw_reg.negate ^= 1; if (src.Abs) { diff --git a/src/mesa/drivers/dri/i965/brw_wm_state.c b/src/mesa/drivers/dri/i965/brw_wm_state.c index 58fa6aaf8f9..67b41173fb2 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_state.c +++ b/src/mesa/drivers/dri/i965/brw_wm_state.c @@ -290,7 +290,7 @@ const struct brw_tracked_state brw_wm_unit = { .brw = (BRW_NEW_FRAGMENT_PROGRAM | BRW_NEW_CURBE_OFFSETS | - BRW_NEW_NR_SURFACES), + BRW_NEW_NR_WM_SURFACES), .cache = (CACHE_NEW_WM_PROG | CACHE_NEW_SAMPLER) 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 e7d55d5dbd1..71840d1e4e8 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c +++ b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c @@ -38,7 +38,7 @@ #include "intel_mipmap_tree.h" #include "intel_batchbuffer.h" #include "intel_tex.h" - +#include "intel_fbo.h" #include "brw_context.h" #include "brw_state.h" @@ -176,7 +176,11 @@ static GLuint translate_tex_format( GLuint mesa_format, GLenum internal_format, } } -struct brw_wm_surface_key { + +/** + * Use same key for WM and VS surfaces. + */ +struct brw_surface_key { GLenum target, depthmode; dri_bo *bo; GLint format, internal_format; @@ -187,6 +191,7 @@ struct brw_wm_surface_key { GLuint offset; }; + static void brw_set_surface_tiling(struct brw_surface_state *surf, uint32_t tiling) { @@ -208,7 +213,7 @@ brw_set_surface_tiling(struct brw_surface_state *surf, uint32_t tiling) static dri_bo * brw_create_texture_surface( struct brw_context *brw, - struct brw_wm_surface_key *key ) + struct brw_surface_key *key ) { struct brw_surface_state surf; dri_bo *bo; @@ -287,8 +292,8 @@ brw_update_texture_surface( GLcontext *ctx, GLuint unit ) 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; - const GLuint j = MAX_DRAW_BUFFERS + unit; + struct brw_surface_key key; + const GLuint surf = SURF_INDEX_TEXTURE(unit); memset(&key, 0, sizeof(key)); @@ -315,25 +320,25 @@ brw_update_texture_surface( GLcontext *ctx, GLuint unit ) key.cpp = intelObj->mt->cpp; key.tiling = intelObj->mt->region->tiling; - dri_bo_unreference(brw->wm.surf_bo[j]); - brw->wm.surf_bo[j] = brw_search_cache(&brw->cache, BRW_SS_SURFACE, + dri_bo_unreference(brw->wm.surf_bo[surf]); + brw->wm.surf_bo[surf] = brw_search_cache(&brw->cache, BRW_SS_SURFACE, &key, sizeof(key), &key.bo, key.bo ? 1 : 0, NULL); - if (brw->wm.surf_bo[j] == NULL) { - brw->wm.surf_bo[j] = brw_create_texture_surface(brw, &key); + if (brw->wm.surf_bo[surf] == NULL) { + brw->wm.surf_bo[surf] = brw_create_texture_surface(brw, &key); } } /** - * Create the constant buffer surface. Fragment shader constanst will be + * Create the constant buffer surface. Vertex/fragment shader constants will be * read from this buffer with Data Port Read instructions/messages. */ static dri_bo * brw_create_constant_surface( struct brw_context *brw, - struct brw_wm_surface_key *key ) + struct brw_surface_key *key ) { const GLint w = key->width - 1; struct brw_surface_state surf; @@ -345,8 +350,6 @@ brw_create_constant_surface( struct brw_context *brw, surf.ss0.surface_type = BRW_SURFACE_BUFFER; surf.ss0.surface_format = BRW_SURFACEFORMAT_R32G32B32A32_FLOAT; - /* This is ok for all textures with channel width 8bit or less: - */ assert(key->bo); if (key->bo) surf.ss1.base_addr = key->bo->offset; /* reloc */ @@ -356,8 +359,8 @@ brw_create_constant_surface( struct brw_context *brw, surf.ss2.width = w & 0x7f; /* bits 6:0 of size or width */ surf.ss2.height = (w >> 7) & 0x1fff; /* bits 19:7 of size or width */ surf.ss3.depth = (w >> 20) & 0x7f; /* bits 26:20 of size or width */ - surf.ss3.pitch = (key->pitch * key->cpp) - 1; - brw_set_surface_tiling(&surf, key->tiling); + surf.ss3.pitch = (key->pitch * key->cpp) - 1; /* ignored?? */ + brw_set_surface_tiling(&surf, key->tiling); /* tiling now allowed */ bo = brw_upload_cache(&brw->cache, BRW_SS_SURFACE, key, sizeof(*key), @@ -379,26 +382,100 @@ brw_create_constant_surface( struct brw_context *brw, /** - * Update the constant buffer surface. + * Update the surface state for a WM constant buffer. + * The constant buffer will be (re)allocated here if needed. */ -static void -brw_update_constant_surface( GLcontext *ctx, - const struct brw_fragment_program *fp ) +static dri_bo * +brw_update_wm_constant_surface( GLcontext *ctx, + GLuint surf, + dri_bo *const_buffer, + const struct gl_program_parameter_list *params) { struct brw_context *brw = brw_context(ctx); - struct brw_wm_surface_key key; - const GLuint j = BRW_WM_MAX_SURF - 1; - const GLuint numParams = fp->program.Base.Parameters->NumParameters; + struct brw_surface_key key; + struct intel_context *intel = &brw->intel; + const int size = params->NumParameters * 4 * sizeof(GLfloat); + + /* free old const buffer if too small */ + if (const_buffer && const_buffer->size < size) { + dri_bo_unreference(const_buffer); + const_buffer = NULL; + } + + /* alloc new buffer if needed */ + if (!const_buffer) { + const_buffer = + drm_intel_bo_alloc(intel->bufmgr, "fp_const_buffer", size, 64); + } memset(&key, 0, sizeof(key)); key.format = MESA_FORMAT_RGBA_FLOAT32; key.internal_format = GL_RGBA; - key.bo = fp->const_buffer; + key.bo = const_buffer; + key.depthmode = GL_NONE; + key.pitch = params->NumParameters; + key.width = params->NumParameters; + key.height = 1; + key.depth = 1; + key.cpp = 16; + + /* + printf("%s:\n", __FUNCTION__); + printf(" width %d height %d depth %d cpp %d pitch %d\n", + key.width, key.height, key.depth, key.cpp, key.pitch); + */ + + dri_bo_unreference(brw->wm.surf_bo[surf]); + brw->wm.surf_bo[surf] = brw_search_cache(&brw->cache, BRW_SS_SURFACE, + &key, sizeof(key), + &key.bo, key.bo ? 1 : 0, + NULL); + if (brw->wm.surf_bo[surf] == NULL) { + brw->wm.surf_bo[surf] = brw_create_constant_surface(brw, &key); + } + + return const_buffer; +} + +/** + * Update the surface state for a VS constant buffer. + * The constant buffer will be (re)allocated here if needed. + */ +static dri_bo * +brw_update_vs_constant_surface( GLcontext *ctx, + GLuint surf, + dri_bo *const_buffer, + const struct gl_program_parameter_list *params) +{ + struct brw_context *brw = brw_context(ctx); + struct brw_surface_key key; + struct intel_context *intel = &brw->intel; + const int size = params->NumParameters * 4 * sizeof(GLfloat); + + assert(surf == 0); + + /* free old const buffer if too small */ + if (const_buffer && const_buffer->size < size) { + dri_bo_unreference(const_buffer); + const_buffer = NULL; + } + + /* alloc new buffer if needed */ + if (!const_buffer) { + const_buffer = + drm_intel_bo_alloc(intel->bufmgr, "vp_const_buffer", size, 64); + } + + memset(&key, 0, sizeof(key)); + + key.format = MESA_FORMAT_RGBA_FLOAT32; + key.internal_format = GL_RGBA; + key.bo = const_buffer; key.depthmode = GL_NONE; - key.pitch = numParams; - key.width = numParams; + key.pitch = params->NumParameters; + key.width = params->NumParameters; key.height = 1; key.depth = 1; key.cpp = 16; @@ -409,14 +486,16 @@ brw_update_constant_surface( GLcontext *ctx, key.width, key.height, key.depth, key.cpp, key.pitch); */ - dri_bo_unreference(brw->wm.surf_bo[j]); - brw->wm.surf_bo[j] = brw_search_cache(&brw->cache, BRW_SS_SURFACE, - &key, sizeof(key), - &key.bo, key.bo ? 1 : 0, - NULL); - if (brw->wm.surf_bo[j] == NULL) { - brw->wm.surf_bo[j] = brw_create_constant_surface(brw, &key); + dri_bo_unreference(brw->vs.surf_bo[surf]); + brw->vs.surf_bo[surf] = brw_search_cache(&brw->cache, BRW_SS_SURFACE, + &key, sizeof(key), + &key.bo, key.bo ? 1 : 0, + NULL); + if (brw->vs.surf_bo[surf] == NULL) { + brw->vs.surf_bo[surf] = brw_create_constant_surface(brw, &key); } + + return const_buffer; } @@ -426,15 +505,18 @@ brw_update_constant_surface( GLcontext *ctx, * usable for further buffers when doing ARB_draw_buffer support. */ static void -brw_update_region_surface(struct brw_context *brw, struct intel_region *region, - unsigned int unit, GLboolean cached) +brw_update_renderbuffer_surface(struct brw_context *brw, + struct gl_renderbuffer *rb, + unsigned int unit, GLboolean cached) { GLcontext *ctx = &brw->intel.ctx; dri_bo *region_bo = NULL; + struct intel_renderbuffer *irb = intel_renderbuffer(rb); + struct intel_region *region = irb ? irb->region : NULL; struct { unsigned int surface_type; unsigned int surface_format; - unsigned int width, height, cpp; + unsigned int width, height, pitch, cpp; GLubyte color_mask[4]; GLboolean color_blend; uint32_t tiling; @@ -446,13 +528,27 @@ brw_update_region_surface(struct brw_context *brw, struct intel_region *region, region_bo = region->buffer; key.surface_type = BRW_SURFACE_2D; - if (region->cpp == 4) + switch (irb->texformat->MesaFormat) { + case MESA_FORMAT_ARGB8888: key.surface_format = BRW_SURFACEFORMAT_B8G8R8A8_UNORM; - else + break; + case MESA_FORMAT_RGB565: key.surface_format = BRW_SURFACEFORMAT_B5G6R5_UNORM; + break; + case MESA_FORMAT_ARGB1555: + key.surface_format = BRW_SURFACEFORMAT_B5G5R5A1_UNORM; + break; + case MESA_FORMAT_ARGB4444: + key.surface_format = BRW_SURFACEFORMAT_B4G4R4A4_UNORM; + break; + default: + _mesa_problem(ctx, "Bad renderbuffer format: %d\n", + irb->texformat->MesaFormat); + } key.tiling = region->tiling; - key.width = region->pitch; /* XXX: not really! */ + key.width = region->width; key.height = region->height; + key.pitch = region->pitch; key.cpp = region->cpp; } else { key.surface_type = BRW_SURFACE_NULL; @@ -488,7 +584,7 @@ brw_update_region_surface(struct brw_context *brw, struct intel_region *region, surf.ss2.width = key.width - 1; surf.ss2.height = key.height - 1; brw_set_surface_tiling(&surf, key.tiling); - surf.ss3.pitch = (key.width * key.cpp) - 1; + surf.ss3.pitch = (key.pitch * key.cpp) - 1; /* _NEW_COLOR */ surf.ss0.color_blend = key.color_blend; @@ -499,7 +595,7 @@ brw_update_region_surface(struct brw_context *brw, struct intel_region *region, /* Key size will never match key size for textures, so we're safe. */ brw->wm.surf_bo[unit] = brw_upload_cache(&brw->cache, BRW_SS_SURFACE, - &key, sizeof(key), + &key, sizeof(key), ®ion_bo, 1, &surf, sizeof(surf), NULL, NULL); @@ -528,6 +624,8 @@ brw_wm_get_binding_table(struct brw_context *brw) { dri_bo *bind_bo; + assert(brw->wm.nr_surfaces <= BRW_WM_MAX_SURF); + bind_bo = brw_search_cache(&brw->cache, BRW_SS_SURF_BIND, NULL, 0, brw->wm.surf_bo, brw->wm.nr_surfaces, @@ -574,69 +672,159 @@ static void prepare_wm_surfaces(struct brw_context *brw ) GLuint i; int old_nr_surfaces; + /* _NEW_BUFFERS */ /* Update surfaces for drawing buffers */ - if (brw->state.nr_color_regions > 1) { - for (i = 0; i < brw->state.nr_color_regions; i++) { - brw_update_region_surface(brw, brw->state.color_regions[i], i, - GL_FALSE); + if (ctx->DrawBuffer->_NumColorDrawBuffers >= 1) { + for (i = 0; i < ctx->DrawBuffer->_NumColorDrawBuffers; i++) { + brw_update_renderbuffer_surface(brw, + ctx->DrawBuffer->_ColorDrawBuffers[i], + i, + GL_FALSE); } } else { - brw_update_region_surface(brw, brw->state.color_regions[0], 0, GL_TRUE); + brw_update_renderbuffer_surface(brw, NULL, 0, GL_TRUE); } old_nr_surfaces = brw->wm.nr_surfaces; brw->wm.nr_surfaces = MAX_DRAW_BUFFERS; + /* Update surface / buffer for fragment shader constant buffer */ + { + const GLuint surf = SURF_INDEX_FRAG_CONST_BUFFER; + struct brw_fragment_program *fp = + (struct brw_fragment_program *) brw->fragment_program; + fp->const_buffer = + brw_update_wm_constant_surface(ctx, surf, fp->const_buffer, + fp->program.Base.Parameters); + + brw->wm.nr_surfaces = surf + 1; + } + /* Update surfaces for textures */ for (i = 0; i < BRW_MAX_TEX_UNIT; i++) { const struct gl_texture_unit *texUnit = &ctx->Texture.Unit[i]; - const GLuint j = MAX_DRAW_BUFFERS + i; + const GLuint surf = SURF_INDEX_TEXTURE(i); /* _NEW_TEXTURE, BRW_NEW_TEXDATA */ if (texUnit->_ReallyEnabled) { if (texUnit->_Current == intel->frame_buffer_texobj) { /* render to texture */ - dri_bo_unreference(brw->wm.surf_bo[j]); - brw->wm.surf_bo[j] = brw->wm.surf_bo[0]; - dri_bo_reference(brw->wm.surf_bo[j]); - brw->wm.nr_surfaces = j + 1; + dri_bo_unreference(brw->wm.surf_bo[surf]); + brw->wm.surf_bo[surf] = brw->wm.surf_bo[0]; + dri_bo_reference(brw->wm.surf_bo[surf]); + brw->wm.nr_surfaces = surf + 1; } else { /* regular texture */ brw_update_texture_surface(ctx, i); - brw->wm.nr_surfaces = j + 1; + brw->wm.nr_surfaces = surf + 1; } } else { - dri_bo_unreference(brw->wm.surf_bo[j]); - brw->wm.surf_bo[j] = NULL; + dri_bo_unreference(brw->wm.surf_bo[surf]); + brw->wm.surf_bo[surf] = NULL; } } - /* Update surface for fragment shader constant buffer */ - { - const GLuint j = BRW_WM_MAX_SURF - 1; - const struct brw_fragment_program *fp = - brw_fragment_program_const(brw->fragment_program); + dri_bo_unreference(brw->wm.bind_bo); + brw->wm.bind_bo = brw_wm_get_binding_table(brw); + + if (brw->wm.nr_surfaces != old_nr_surfaces) + brw->state.dirty.brw |= BRW_NEW_NR_WM_SURFACES; +} + + +/** + * Constructs the binding table for the VS surface state. + */ +static dri_bo * +brw_vs_get_binding_table(struct brw_context *brw) +{ + dri_bo *bind_bo; + + assert(brw->vs.nr_surfaces <= BRW_VS_MAX_SURF); + + bind_bo = brw_search_cache(&brw->cache, BRW_SS_SURF_BIND, + NULL, 0, + brw->vs.surf_bo, brw->vs.nr_surfaces, + NULL); + + if (bind_bo == NULL) { + GLuint data_size = brw->vs.nr_surfaces * sizeof(GLuint); + uint32_t *data = malloc(data_size); + int i; + + for (i = 0; i < brw->vs.nr_surfaces; i++) + if (brw->vs.surf_bo[i]) + data[i] = brw->vs.surf_bo[i]->offset; + else + data[i] = 0; + + bind_bo = brw_upload_cache( &brw->cache, BRW_SS_SURF_BIND, + NULL, 0, + brw->vs.surf_bo, brw->vs.nr_surfaces, + data, data_size, + NULL, NULL); - brw_update_constant_surface(ctx, fp); - brw->wm.nr_surfaces = j + 1; + /* Emit binding table relocations to surface state */ + for (i = 0; i < BRW_VS_MAX_SURF; i++) { + if (brw->vs.surf_bo[i] != NULL) { + dri_bo_emit_reloc(bind_bo, + I915_GEM_DOMAIN_INSTRUCTION, 0, + 0, + i * sizeof(GLuint), + brw->vs.surf_bo[i]); + } + } + + free(data); } + return bind_bo; +} - dri_bo_unreference(brw->wm.bind_bo); - brw->wm.bind_bo = brw_wm_get_binding_table(brw); - if (brw->wm.nr_surfaces != old_nr_surfaces) - brw->state.dirty.brw |= BRW_NEW_NR_SURFACES; +/** + * Vertex shader surfaces. Just constant buffer for now. Could add vertex + * shader textures in the future. + */ +static void prepare_vs_surfaces(struct brw_context *brw ) +{ + GLcontext *ctx = &brw->intel.ctx; + + /* Update surface / buffer for vertex shader constant buffer */ + { + const GLuint surf = SURF_INDEX_VERT_CONST_BUFFER; + struct brw_vertex_program *vp = + (struct brw_vertex_program *) brw->vertex_program; + vp->const_buffer = + brw_update_vs_constant_surface(ctx, surf, vp->const_buffer, + vp->program.Base.Parameters); + + brw->vs.nr_surfaces = 1; + } + + dri_bo_unreference(brw->vs.bind_bo); + brw->vs.bind_bo = brw_vs_get_binding_table(brw); + + if (1) + brw->state.dirty.brw |= BRW_NEW_NR_VS_SURFACES; +} + + +static void +prepare_surfaces(struct brw_context *brw) +{ + prepare_wm_surfaces(brw); + prepare_vs_surfaces(brw); } const struct brw_tracked_state brw_wm_surfaces = { .dirty = { - .mesa = _NEW_COLOR | _NEW_TEXTURE | _NEW_BUFFERS, + .mesa = _NEW_COLOR | _NEW_TEXTURE | _NEW_BUFFERS | _NEW_PROGRAM, .brw = BRW_NEW_CONTEXT, .cache = 0 }, - .prepare = prepare_wm_surfaces, + .prepare = prepare_surfaces, }; diff --git a/src/mesa/drivers/dri/intel/intel_blit.c b/src/mesa/drivers/dri/intel/intel_blit.c index 4ae9b118a3d..49198281316 100644 --- a/src/mesa/drivers/dri/intel/intel_blit.c +++ b/src/mesa/drivers/dri/intel/intel_blit.c @@ -32,6 +32,8 @@ #include "main/mtypes.h" #include "main/context.h" #include "main/enums.h" +#include "main/texformat.h" +#include "main/colormac.h" #include "intel_blit.h" #include "intel_buffers.h" @@ -484,10 +486,9 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask) const GLbitfield bufBit = 1 << buf; if ((clearMask & bufBit) && !(bufBit & skipBuffers)) { /* OK, clear this renderbuffer */ - struct intel_region *irb_region = - intel_get_rb_region(fb, buf); + struct intel_renderbuffer *irb = intel_get_renderbuffer(fb, buf); dri_bo *write_buffer = - intel_region_buffer(intel, irb_region, + intel_region_buffer(intel, irb->region, all ? INTEL_WRITE_FULL : INTEL_WRITE_PART); @@ -495,15 +496,13 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask) GLint pitch, cpp; GLuint BR13, CMD; - ASSERT(irb_region); - - pitch = irb_region->pitch; - cpp = irb_region->cpp; + pitch = irb->region->pitch; + cpp = irb->region->cpp; DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", __FUNCTION__, - irb_region->buffer, (pitch * cpp), - irb_region->draw_offset, + irb->region->buffer, (pitch * cpp), + irb->region->draw_offset, b.x1, b.y1, b.x2 - b.x1, b.y2 - b.y1); BR13 = 0xf0 << 16; @@ -529,7 +528,7 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask) } #ifndef I915 - if (irb_region->tiling != I915_TILING_NONE) { + if (irb->region->tiling != I915_TILING_NONE) { CMD |= XY_DST_TILED; pitch /= 4; } @@ -540,9 +539,36 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask) clearVal = clear_depth; } else { - clearVal = (cpp == 4) - ? intel->ClearColor8888 : intel->ClearColor565; - } + uint8_t clear[4]; + GLclampf *color = ctx->Color.ClearColor; + + CLAMPED_FLOAT_TO_UBYTE(clear[0], color[0]); + CLAMPED_FLOAT_TO_UBYTE(clear[1], color[1]); + CLAMPED_FLOAT_TO_UBYTE(clear[2], color[2]); + CLAMPED_FLOAT_TO_UBYTE(clear[3], color[3]); + + switch (irb->texformat->MesaFormat) { + case MESA_FORMAT_ARGB8888: + clearVal = intel->ClearColor8888; + break; + case MESA_FORMAT_RGB565: + clearVal = intel->ClearColor565; + break; + case MESA_FORMAT_ARGB4444: + clearVal = PACK_COLOR_4444(clear[3], clear[0], + clear[1], clear[2]); + break; + case MESA_FORMAT_ARGB1555: + clearVal = PACK_COLOR_1555(clear[3], clear[0], + clear[1], clear[2]); + break; + default: + _mesa_problem(ctx, "Unexpected renderbuffer format: %d\n", + irb->texformat->MesaFormat); + clearVal = 0; + } + } + /* _mesa_debug(ctx, "hardware blit clear buf %d rb id %d\n", buf, irb->Base.Name); @@ -558,7 +584,7 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask) OUT_BATCH((b.y2 << 16) | b.x2); OUT_RELOC(write_buffer, I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER, - irb_region->draw_offset); + irb->region->draw_offset); OUT_BATCH(clearVal); ADVANCE_BATCH(); clearMask &= ~bufBit; /* turn off bit, for faster loop exit */ diff --git a/src/mesa/drivers/dri/intel/intel_buffers.c b/src/mesa/drivers/dri/intel/intel_buffers.c index 0929a2c223c..90964df3553 100644 --- a/src/mesa/drivers/dri/intel/intel_buffers.c +++ b/src/mesa/drivers/dri/intel/intel_buffers.c @@ -202,6 +202,8 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb) intel_batchbuffer_flush(intel->batch); intel->front_cliprects = GL_TRUE; colorRegions[0] = intel_get_rb_region(fb, BUFFER_FRONT_LEFT); + + intel->front_buffer_dirty = GL_TRUE; } else { if (!intel->constant_cliprect && intel->front_cliprects) @@ -319,6 +321,12 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb) static void intelDrawBuffer(GLcontext * ctx, GLenum mode) { + if ((ctx->DrawBuffer != NULL) && (ctx->DrawBuffer->Name == 0)) { + struct intel_context *const intel = intel_context(ctx); + + intel->is_front_buffer_rendering = (mode == GL_FRONT_LEFT); + } + intel_draw_buffer(ctx, ctx->DrawBuffer); } diff --git a/src/mesa/drivers/dri/intel/intel_clear.c b/src/mesa/drivers/dri/intel/intel_clear.c index 28281b38615..aed95c7c561 100644 --- a/src/mesa/drivers/dri/intel/intel_clear.c +++ b/src/mesa/drivers/dri/intel/intel_clear.c @@ -38,6 +38,7 @@ #include "main/enable.h" #include "main/macros.h" #include "main/matrix.h" +#include "main/polygon.h" #include "main/texstate.h" #include "main/shaders.h" #include "main/stencil.h" @@ -93,6 +94,7 @@ intel_clear_tris(GLcontext *ctx, GLbitfield mask) GL_CURRENT_BIT | GL_DEPTH_BUFFER_BIT | GL_ENABLE_BIT | + GL_POLYGON_BIT | GL_STENCIL_BUFFER_BIT | GL_TRANSFORM_BIT | GL_CURRENT_BIT); @@ -114,6 +116,7 @@ intel_clear_tris(GLcontext *ctx, GLbitfield mask) _mesa_Disable(GL_CLIP_PLANE3); _mesa_Disable(GL_CLIP_PLANE4); _mesa_Disable(GL_CLIP_PLANE5); + _mesa_PolygonMode(GL_FRONT_AND_BACK, GL_FILL); if (ctx->Extensions.ARB_fragment_program && ctx->FragmentProgram.Enabled) { saved_fp_enable = GL_TRUE; _mesa_Disable(GL_FRAGMENT_PROGRAM_ARB); @@ -146,6 +149,11 @@ intel_clear_tris(GLcontext *ctx, GLbitfield mask) } } +#if FEATURE_ARB_vertex_buffer_object + _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, 0); + _mesa_BindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, 0); +#endif + intel_meta_set_passthrough_transform(intel); for (i = 0; i < 4; i++) { diff --git a/src/mesa/drivers/dri/intel/intel_context.c b/src/mesa/drivers/dri/intel/intel_context.c index a664e749360..3436b8ecd30 100644 --- a/src/mesa/drivers/dri/intel/intel_context.c +++ b/src/mesa/drivers/dri/intel/intel_context.c @@ -235,6 +235,11 @@ intel_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable) region_name = "dri2 front buffer"; break; + case __DRI_BUFFER_FAKE_FRONT_LEFT: + rb = intel_fb->color_rb[0]; + region_name = "dri2 fake front buffer"; + break; + case __DRI_BUFFER_BACK_LEFT: rb = intel_fb->color_rb[1]; region_name = "dri2 back buffer"; @@ -391,6 +396,27 @@ intel_flush(GLcontext *ctx, GLboolean needs_mi_flush) if (intel->batch->map != intel->batch->ptr) intel_batchbuffer_flush(intel->batch); + + if ((ctx->DrawBuffer->Name == 0) && intel->front_buffer_dirty) { + __DRIscreen *const screen = intel->intelScreen->driScrnPriv; + + if (screen->dri2.loader && + (screen->dri2.loader->base.version >= 2) + && (screen->dri2.loader->flushFrontBuffer != NULL)) { + (*screen->dri2.loader->flushFrontBuffer)(intel->driDrawable, + intel->driDrawable->loaderPrivate); + + /* Only clear the dirty bit if front-buffer rendering is no longer + * enabled. This is done so that the dirty bit can only be set in + * glDrawBuffer. Otherwise the dirty bit would have to be set at + * each of N places that do rendering. This has worse performances, + * but it is much easier to get correct. + */ + if (intel->is_front_buffer_rendering) { + intel->front_buffer_dirty = GL_FALSE; + } + } + } } void diff --git a/src/mesa/drivers/dri/intel/intel_context.h b/src/mesa/drivers/dri/intel/intel_context.h index d635f3f50dc..d798225ddd9 100644 --- a/src/mesa/drivers/dri/intel/intel_context.h +++ b/src/mesa/drivers/dri/intel/intel_context.h @@ -48,6 +48,8 @@ #define DV_PF_555 (1<<8) #define DV_PF_565 (2<<8) #define DV_PF_8888 (3<<8) +#define DV_PF_4444 (8<<8) +#define DV_PF_1555 (9<<8) struct intel_region; struct intel_context; @@ -262,11 +264,29 @@ struct intel_context * flush time while the lock is held. */ GLboolean constant_cliprect; + /** * In !constant_cliprect mode, set to true if the front cliprects should be * used instead of back. */ GLboolean front_cliprects; + + /** + * Set if rendering has occured to the drawable's front buffer. + * + * This is used in the DRI2 case to detect that glFlush should also copy + * the contents of the fake front buffer to the real front buffer. + */ + GLboolean front_buffer_dirty; + + /** + * Track whether front-buffer rendering is currently enabled + * + * A separate flag is used to track this in order to support MRT more + * easily. + */ + GLboolean is_front_buffer_rendering; + drm_clip_rect_t fboRect; /**< cliprect for FBO rendering */ int perf_boxes; @@ -319,6 +339,7 @@ extern char *__progname; #define ARRAY_SIZE(x) (sizeof(x) / sizeof(x[0])) #define ALIGN(value, alignment) ((value + alignment - 1) & ~(alignment - 1)) +#define IS_POWER_OF_TWO(val) (((val) & (val - 1)) == 0) #define INTEL_FIREVERTICES(intel) \ do { \ diff --git a/src/mesa/drivers/dri/intel/intel_decode.c b/src/mesa/drivers/dri/intel/intel_decode.c index f04638206d5..a9dfe281cbd 100644 --- a/src/mesa/drivers/dri/intel/intel_decode.c +++ b/src/mesa/drivers/dri/intel/intel_decode.c @@ -800,6 +800,7 @@ static int decode_3d_1d(uint32_t *data, int count, uint32_t hw_offset, int *failures, int i830) { unsigned int len, i, c, opcode, word, map, sampler, instr; + char *format; struct { uint32_t opcode; @@ -1001,6 +1002,35 @@ decode_3d_1d(uint32_t *data, int count, uint32_t hw_offset, int *failures, int i (*failures)++; } return len; + case 0x85: + len = (data[0] & 0x0000000f) + 2; + + if (len != 2) + fprintf(out, "Bad count in 3DSTATE_DEST_BUFFER_VARIABLES\n"); + if (count < 2) + BUFFER_FAIL(count, len, "3DSTATE_DEST_BUFFER_VARIABLES"); + + instr_out(data, hw_offset, 0, + "3DSTATE_DEST_BUFFER_VARIABLES\n"); + + switch ((data[1] >> 8) & 0xf) { + case 0x0: format = "g8"; break; + case 0x1: format = "x1r5g5b5"; break; + case 0x2: format = "r5g6b5"; break; + case 0x3: format = "a8r8g8b8"; break; + case 0x4: format = "ycrcb_swapy"; break; + case 0x5: format = "ycrcb_normal"; break; + case 0x6: format = "ycrcb_swapuv"; break; + case 0x7: format = "ycrcb_swapuvy"; break; + case 0x8: format = "a4r4g4b4"; break; + case 0x9: format = "a1r5g5b5"; break; + case 0xa: format = "a2r10g10b10"; break; + default: format = "BAD"; break; + } + instr_out(data, hw_offset, 1, "%s format, early Z %sabled\n", + format, + (data[1] & (1 << 31)) ? "en" : "dis"); + return len; } for (opcode = 0; opcode < sizeof(opcodes_3d_1d) / sizeof(opcodes_3d_1d[0]); diff --git a/src/mesa/drivers/dri/intel/intel_fbo.c b/src/mesa/drivers/dri/intel/intel_fbo.c index a401f730ba2..52647ddf8b2 100644 --- a/src/mesa/drivers/dri/intel/intel_fbo.c +++ b/src/mesa/drivers/dri/intel/intel_fbo.c @@ -119,6 +119,7 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, rb->RedBits = 5; rb->GreenBits = 6; rb->BlueBits = 5; + irb->texformat = &_mesa_texformat_rgb565; cpp = 2; break; case GL_RGB: @@ -132,6 +133,7 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, rb->GreenBits = 8; rb->BlueBits = 8; rb->AlphaBits = 0; + irb->texformat = &_mesa_texformat_argb8888; /* XXX: Need xrgb8888 */ cpp = 4; break; case GL_RGBA: @@ -148,6 +150,7 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, rb->GreenBits = 8; rb->BlueBits = 8; rb->AlphaBits = 8; + irb->texformat = &_mesa_texformat_argb8888; cpp = 4; break; case GL_STENCIL_INDEX: @@ -160,12 +163,14 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, rb->DataType = GL_UNSIGNED_INT_24_8_EXT; rb->StencilBits = 8; cpp = 4; + irb->texformat = &_mesa_texformat_s8_z24; break; case GL_DEPTH_COMPONENT16: rb->_ActualFormat = GL_DEPTH_COMPONENT16; rb->DataType = GL_UNSIGNED_SHORT; rb->DepthBits = 16; cpp = 2; + irb->texformat = &_mesa_texformat_z16; break; case GL_DEPTH_COMPONENT: case GL_DEPTH_COMPONENT24: @@ -174,6 +179,7 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, rb->DataType = GL_UNSIGNED_INT_24_8_EXT; rb->DepthBits = 24; cpp = 4; + irb->texformat = &_mesa_texformat_s8_z24; break; case GL_DEPTH_STENCIL_EXT: case GL_DEPTH24_STENCIL8_EXT: @@ -182,6 +188,7 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, rb->DepthBits = 24; rb->StencilBits = 8; cpp = 4; + irb->texformat = &_mesa_texformat_s8_z24; break; default: _mesa_problem(ctx, @@ -322,6 +329,7 @@ intel_create_renderbuffer(GLenum intFormat) irb->Base.GreenBits = 6; irb->Base.BlueBits = 5; irb->Base.DataType = GL_UNSIGNED_BYTE; + irb->texformat = &_mesa_texformat_rgb565; break; case GL_RGB8: irb->Base._ActualFormat = GL_RGB8; @@ -331,6 +339,7 @@ intel_create_renderbuffer(GLenum intFormat) irb->Base.BlueBits = 8; irb->Base.AlphaBits = 0; irb->Base.DataType = GL_UNSIGNED_BYTE; + irb->texformat = &_mesa_texformat_argb8888; /* XXX: Need xrgb8888 */ break; case GL_RGBA8: irb->Base._ActualFormat = GL_RGBA8; @@ -340,24 +349,28 @@ intel_create_renderbuffer(GLenum intFormat) irb->Base.BlueBits = 8; irb->Base.AlphaBits = 8; irb->Base.DataType = GL_UNSIGNED_BYTE; + irb->texformat = &_mesa_texformat_argb8888; break; case GL_STENCIL_INDEX8_EXT: irb->Base._ActualFormat = GL_STENCIL_INDEX8_EXT; irb->Base._BaseFormat = GL_STENCIL_INDEX; irb->Base.StencilBits = 8; irb->Base.DataType = GL_UNSIGNED_BYTE; + irb->texformat = &_mesa_texformat_s8_z24; break; case GL_DEPTH_COMPONENT16: irb->Base._ActualFormat = GL_DEPTH_COMPONENT16; irb->Base._BaseFormat = GL_DEPTH_COMPONENT; irb->Base.DepthBits = 16; irb->Base.DataType = GL_UNSIGNED_SHORT; + irb->texformat = &_mesa_texformat_z16; break; case GL_DEPTH_COMPONENT24: irb->Base._ActualFormat = GL_DEPTH24_STENCIL8_EXT; irb->Base._BaseFormat = GL_DEPTH_COMPONENT; irb->Base.DepthBits = 24; irb->Base.DataType = GL_UNSIGNED_INT; + irb->texformat = &_mesa_texformat_s8_z24; break; case GL_DEPTH24_STENCIL8_EXT: irb->Base._ActualFormat = GL_DEPTH24_STENCIL8_EXT; @@ -365,6 +378,7 @@ intel_create_renderbuffer(GLenum intFormat) irb->Base.DepthBits = 24; irb->Base.StencilBits = 8; irb->Base.DataType = GL_UNSIGNED_INT_24_8_EXT; + irb->texformat = &_mesa_texformat_s8_z24; break; default: _mesa_problem(NULL, @@ -449,6 +463,8 @@ static GLboolean intel_update_wrapper(GLcontext *ctx, struct intel_renderbuffer *irb, struct gl_texture_image *texImage) { + irb->texformat = texImage->TexFormat; + if (texImage->TexFormat == &_mesa_texformat_argb8888) { irb->Base._ActualFormat = GL_RGBA8; irb->Base._BaseFormat = GL_RGBA; @@ -458,9 +474,21 @@ intel_update_wrapper(GLcontext *ctx, struct intel_renderbuffer *irb, else if (texImage->TexFormat == &_mesa_texformat_rgb565) { irb->Base._ActualFormat = GL_RGB5; irb->Base._BaseFormat = GL_RGB; - irb->Base.DataType = GL_UNSIGNED_SHORT; + irb->Base.DataType = GL_UNSIGNED_BYTE; DBG("Render to RGB5 texture OK\n"); } + else if (texImage->TexFormat == &_mesa_texformat_argb1555) { + irb->Base._ActualFormat = GL_RGB5_A1; + irb->Base._BaseFormat = GL_RGBA; + irb->Base.DataType = GL_UNSIGNED_BYTE; + DBG("Render to ARGB1555 texture OK\n"); + } + else if (texImage->TexFormat == &_mesa_texformat_argb4444) { + irb->Base._ActualFormat = GL_RGBA4; + irb->Base._BaseFormat = GL_RGBA; + irb->Base.DataType = GL_UNSIGNED_BYTE; + DBG("Render to ARGB4444 texture OK\n"); + } else if (texImage->TexFormat == &_mesa_texformat_z16) { irb->Base._ActualFormat = GL_DEPTH_COMPONENT16; irb->Base._BaseFormat = GL_DEPTH_COMPONENT; @@ -631,11 +659,11 @@ intel_finish_render_texture(GLcontext * ctx, static void intel_validate_framebuffer(GLcontext *ctx, struct gl_framebuffer *fb) { - struct intel_context *intel = intel_context(ctx); const struct intel_renderbuffer *depthRb = intel_get_renderbuffer(fb, BUFFER_DEPTH); const struct intel_renderbuffer *stencilRb = intel_get_renderbuffer(fb, BUFFER_STENCIL); + int i; if (stencilRb && stencilRb != depthRb) { /* we only support combined depth/stencil buffers, not separate @@ -644,32 +672,21 @@ intel_validate_framebuffer(GLcontext *ctx, struct gl_framebuffer *fb) fb->_Status = GL_FRAMEBUFFER_UNSUPPORTED_EXT; } - /* check that texture color buffers are a format we can render into */ - { - const struct gl_texture_format *supportedFormat; - GLuint i; + for (i = 0; i < ctx->Const.MaxDrawBuffers; i++) { + struct gl_renderbuffer *rb = ctx->DrawBuffer->_ColorDrawBuffers[i]; + struct intel_renderbuffer *irb = intel_renderbuffer(rb); - /* The texture format we can render into seems to depend on the - * screen depth. There currently seems to be a problem when - * rendering into a rgb565 texture when the screen is abgr8888. - */ + if (rb == NULL) + continue; - if (intel->ctx.Visual.rgbBits >= 24) - supportedFormat = &_mesa_texformat_argb8888; - else - supportedFormat = &_mesa_texformat_rgb565; - - for (i = 0; i < ctx->Const.MaxDrawBuffers; i++) { - const struct gl_texture_object *texObj = - fb->Attachment[BUFFER_COLOR0 + i].Texture; - if (texObj) { - const struct gl_texture_image *texImg = - texObj->Image[0][texObj->BaseLevel]; - if (texImg && texImg->TexFormat != supportedFormat) { - fb->_Status = GL_FRAMEBUFFER_UNSUPPORTED_EXT; - break; - } - } + switch (irb->texformat->MesaFormat) { + case MESA_FORMAT_ARGB8888: + case MESA_FORMAT_RGB565: + case MESA_FORMAT_ARGB1555: + case MESA_FORMAT_ARGB4444: + break; + default: + fb->_Status = GL_FRAMEBUFFER_UNSUPPORTED_EXT; } } } diff --git a/src/mesa/drivers/dri/intel/intel_fbo.h b/src/mesa/drivers/dri/intel/intel_fbo.h index 7226ee026f6..f0665af482e 100644 --- a/src/mesa/drivers/dri/intel/intel_fbo.h +++ b/src/mesa/drivers/dri/intel/intel_fbo.h @@ -61,6 +61,8 @@ struct intel_renderbuffer struct gl_renderbuffer Base; struct intel_region *region; + const struct gl_texture_format *texformat; + GLuint vbl_pending; /**< vblank sequence number of pending flip */ uint8_t *span_cache; diff --git a/src/mesa/drivers/dri/intel/intel_span.c b/src/mesa/drivers/dri/intel/intel_span.c index c3a873f1abd..34b78ebc1ab 100644 --- a/src/mesa/drivers/dri/intel/intel_span.c +++ b/src/mesa/drivers/dri/intel/intel_span.c @@ -29,6 +29,7 @@ #include "main/macros.h" #include "main/mtypes.h" #include "main/colormac.h" +#include "main/texformat.h" #include "intel_buffers.h" #include "intel_fbo.h" @@ -313,6 +314,22 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb, #define INTEL_TAG(x) x##_RGB565 #include "intel_spantmp.h" +/* a4r4g4b4 color span and pixel functions */ +#define INTEL_PIXEL_FMT GL_BGRA +#define INTEL_PIXEL_TYPE GL_UNSIGNED_SHORT_4_4_4_4_REV +#define INTEL_READ_VALUE(offset) pread_16(irb, offset) +#define INTEL_WRITE_VALUE(offset, v) pwrite_16(irb, offset, v) +#define INTEL_TAG(x) x##_ARGB4444 +#include "intel_spantmp.h" + +/* a1r5g5b5 color span and pixel functions */ +#define INTEL_PIXEL_FMT GL_BGRA +#define INTEL_PIXEL_TYPE GL_UNSIGNED_SHORT_1_5_5_5_REV +#define INTEL_READ_VALUE(offset) pread_16(irb, offset) +#define INTEL_WRITE_VALUE(offset, v) pwrite_16(irb, offset, v) +#define INTEL_TAG(x) x##_ARGB1555 +#include "intel_spantmp.h" + /* a8r8g8b8 color span and pixel functions */ #define INTEL_PIXEL_FMT GL_BGRA #define INTEL_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV @@ -561,8 +578,8 @@ intel_set_span_functions(struct intel_context *intel, else tiling = I915_TILING_NONE; - if (rb->_ActualFormat == GL_RGB5) { - /* 565 RGB */ + switch (irb->texformat->MesaFormat) { + case MESA_FORMAT_RGB565: switch (tiling) { case I915_TILING_NONE: default: @@ -575,38 +592,67 @@ intel_set_span_functions(struct intel_context *intel, intel_YTile_InitPointers_RGB565(rb); break; } - } - else if (rb->_ActualFormat == GL_RGB8) { - /* 8888 RGBx */ + break; + case MESA_FORMAT_ARGB4444: switch (tiling) { case I915_TILING_NONE: default: - intelInitPointers_xRGB8888(rb); + intelInitPointers_ARGB4444(rb); break; case I915_TILING_X: - intel_XTile_InitPointers_xRGB8888(rb); + intel_XTile_InitPointers_ARGB4444(rb); break; case I915_TILING_Y: - intel_YTile_InitPointers_xRGB8888(rb); + intel_YTile_InitPointers_ARGB4444(rb); break; } - } - else if (rb->_ActualFormat == GL_RGBA8) { - /* 8888 RGBA */ + break; + case MESA_FORMAT_ARGB1555: switch (tiling) { case I915_TILING_NONE: default: - intelInitPointers_ARGB8888(rb); + intelInitPointers_ARGB1555(rb); break; case I915_TILING_X: - intel_XTile_InitPointers_ARGB8888(rb); + intel_XTile_InitPointers_ARGB1555(rb); break; case I915_TILING_Y: - intel_YTile_InitPointers_ARGB8888(rb); + intel_YTile_InitPointers_ARGB1555(rb); break; } - } - else if (rb->_ActualFormat == GL_DEPTH_COMPONENT16) { + break; + case MESA_FORMAT_ARGB8888: + if (rb->AlphaBits == 0) { /* XXX: Need xRGB8888 Mesa format */ + /* 8888 RGBx */ + switch (tiling) { + case I915_TILING_NONE: + default: + intelInitPointers_xRGB8888(rb); + break; + case I915_TILING_X: + intel_XTile_InitPointers_xRGB8888(rb); + break; + case I915_TILING_Y: + intel_YTile_InitPointers_xRGB8888(rb); + break; + } + } else { + /* 8888 RGBA */ + switch (tiling) { + case I915_TILING_NONE: + default: + intelInitPointers_ARGB8888(rb); + break; + case I915_TILING_X: + intel_XTile_InitPointers_ARGB8888(rb); + break; + case I915_TILING_Y: + intel_YTile_InitPointers_ARGB8888(rb); + break; + } + } + break; + case MESA_FORMAT_Z16: switch (tiling) { case I915_TILING_NONE: default: @@ -619,51 +665,57 @@ intel_set_span_functions(struct intel_context *intel, intel_YTile_InitDepthPointers_z16(rb); break; } - } - else if (rb->_ActualFormat == GL_DEPTH_COMPONENT24) { - switch (tiling) { - case I915_TILING_NONE: - default: - intelInitDepthPointers_z24(rb); - break; - case I915_TILING_X: - intel_XTile_InitDepthPointers_z24(rb); - break; - case I915_TILING_Y: - intel_YTile_InitDepthPointers_z24(rb); - break; - } - } - else if (rb->_ActualFormat == GL_DEPTH24_STENCIL8_EXT) { - switch (tiling) { - case I915_TILING_NONE: - default: - intelInitDepthPointers_z24_s8(rb); - break; - case I915_TILING_X: - intel_XTile_InitDepthPointers_z24_s8(rb); - break; - case I915_TILING_Y: - intel_YTile_InitDepthPointers_z24_s8(rb); - break; - } - } - else if (rb->_ActualFormat == GL_STENCIL_INDEX8_EXT) { - switch (tiling) { - case I915_TILING_NONE: - default: - intelInitStencilPointers_z24_s8(rb); - break; - case I915_TILING_X: - intel_XTile_InitStencilPointers_z24_s8(rb); - break; - case I915_TILING_Y: - intel_YTile_InitStencilPointers_z24_s8(rb); - break; + break; + case MESA_FORMAT_S8_Z24: + /* There are a few different ways SW asks us to access the S8Z24 data: + * Z24 depth-only depth reads + * S8Z24 depth reads + * S8Z24 stencil reads. + */ + if (rb->_ActualFormat == GL_DEPTH_COMPONENT24) { + switch (tiling) { + case I915_TILING_NONE: + default: + intelInitDepthPointers_z24(rb); + break; + case I915_TILING_X: + intel_XTile_InitDepthPointers_z24(rb); + break; + case I915_TILING_Y: + intel_YTile_InitDepthPointers_z24(rb); + break; + } + } else if (rb->_ActualFormat == GL_DEPTH24_STENCIL8_EXT) { + switch (tiling) { + case I915_TILING_NONE: + default: + intelInitDepthPointers_z24_s8(rb); + break; + case I915_TILING_X: + intel_XTile_InitDepthPointers_z24_s8(rb); + break; + case I915_TILING_Y: + intel_YTile_InitDepthPointers_z24_s8(rb); + break; + } + } else if (rb->_ActualFormat == GL_STENCIL_INDEX8_EXT) { + switch (tiling) { + case I915_TILING_NONE: + default: + intelInitStencilPointers_z24_s8(rb); + break; + case I915_TILING_X: + intel_XTile_InitStencilPointers_z24_s8(rb); + break; + case I915_TILING_Y: + intel_YTile_InitStencilPointers_z24_s8(rb); + break; + } } - } - else { + break; + default: _mesa_problem(NULL, - "Unexpected _ActualFormat in intelSetSpanFunctions"); + "Unexpected MesaFormat in intelSetSpanFunctions"); + break; } } diff --git a/src/mesa/drivers/dri/intel/intel_tex_image.c b/src/mesa/drivers/dri/intel/intel_tex_image.c index 71561cf85cd..1f192dafbe1 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_image.c +++ b/src/mesa/drivers/dri/intel/intel_tex_image.c @@ -315,8 +315,8 @@ intelTexImage(GLcontext * ctx, GLint postConvWidth = width; GLint postConvHeight = height; GLint texelBytes, sizeInBytes; - GLuint dstRowStride, srcRowStride = texImage->RowStride; - + GLuint dstRowStride = 0, srcRowStride = texImage->RowStride; + GLboolean needs_map; DBG("%s target %s level %d %dx%dx%d border %d\n", __FUNCTION__, _mesa_lookup_enum_by_nr(target), level, width, height, depth, border); @@ -482,8 +482,15 @@ intelTexImage(GLcontext * ctx, LOCK_HARDWARE(intel); + /* Two cases where we need a mapping of the miptree: when the user supplied + * data is mapped as well (non-PBO, memcpy upload) or when we're going to do + * (software) mipmap generation. + */ + needs_map = (pixels != NULL) || (level == texObj->BaseLevel && + texObj->GenerateMipmap); + if (intelImage->mt) { - if (pixels) + if (needs_map) texImage->Data = intel_miptree_image_map(intel, intelImage->mt, intelImage->face, @@ -509,8 +516,9 @@ intelTexImage(GLcontext * ctx, } DBG("Upload image %dx%dx%d row_len %d " - "pitch %d\n", - width, height, depth, width * texelBytes, dstRowStride); + "pitch %d pixels %d compressed %d\n", + width, height, depth, width * texelBytes, dstRowStride, + pixels ? 1 : 0, compressed); /* Copy data. Would like to know when it's ok for us to eg. use * the blitter to copy. Or, use the hardware to do the format @@ -523,7 +531,7 @@ intelTexImage(GLcontext * ctx, _mesa_copy_rect(texImage->Data, dst->cpp, dst->pitch, 0, 0, intelImage->mt->level[level].width, - intelImage->mt->level[level].height/4, + (intelImage->mt->level[level].height+3)/4, pixels, srcRowStride, 0, 0); @@ -549,7 +557,7 @@ intelTexImage(GLcontext * ctx, _mesa_unmap_teximage_pbo(ctx, unpack); if (intelImage->mt) { - if (pixels) + if (needs_map) intel_miptree_image_unmap(intel, intelImage->mt); texImage->Data = NULL; } diff --git a/src/mesa/drivers/dri/r200/r200_vertprog.c b/src/mesa/drivers/dri/r200/r200_vertprog.c index aadd1443ad9..620f29b5c6e 100644 --- a/src/mesa/drivers/dri/r200/r200_vertprog.c +++ b/src/mesa/drivers/dri/r200/r200_vertprog.c @@ -290,7 +290,7 @@ static unsigned long t_src(struct r200_vertex_program *vp, struct prog_src_regis t_swizzle(GET_SWZ(src->Swizzle, 2)), t_swizzle(GET_SWZ(src->Swizzle, 3)), t_src_class(src->File), - src->NegateBase) | (src->RelAddr << 4); + src->Negate) | (src->RelAddr << 4); } static unsigned long t_src_scalar(struct r200_vertex_program *vp, struct prog_src_register *src) @@ -302,7 +302,7 @@ static unsigned long t_src_scalar(struct r200_vertex_program *vp, struct prog_sr t_swizzle(GET_SWZ(src->Swizzle, 0)), t_swizzle(GET_SWZ(src->Swizzle, 0)), t_src_class(src->File), - src->NegateBase ? VSF_FLAG_ALL : VSF_FLAG_NONE) | (src->RelAddr << 4); + src->Negate ? VSF_FLAG_ALL : VSF_FLAG_NONE) | (src->RelAddr << 4); } static unsigned long t_opcode(enum prog_opcode opcode) @@ -700,7 +700,7 @@ static GLboolean r200_translate_vertex_program(GLcontext *ctx, struct r200_verte t_swizzle(GET_SWZ(src[1].Swizzle, 0)), SWIZZLE_ZERO, t_src_class(src[0].File), - src[0].NegateBase) | (src[0].RelAddr << 4); + src[0].Negate) | (src[0].RelAddr << 4); o_inst->src1 = UNUSED_SRC_0; o_inst->src2 = UNUSED_SRC_0; } @@ -712,12 +712,12 @@ static GLboolean r200_translate_vertex_program(GLcontext *ctx, struct r200_verte t_swizzle(GET_SWZ(src[0].Swizzle, 0)), SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ZERO, t_src_class(src[0].File), - src[0].NegateBase ? VSF_FLAG_ALL : VSF_FLAG_NONE) | (src[0].RelAddr << 4); + src[0].Negate ? VSF_FLAG_ALL : VSF_FLAG_NONE) | (src[0].RelAddr << 4); o_inst->src1 = MAKE_VSF_SOURCE(t_src_index(vp, &src[1]), SWIZZLE_ZERO, SWIZZLE_ZERO, t_swizzle(GET_SWZ(src[1].Swizzle, 0)), SWIZZLE_ZERO, t_src_class(src[1].File), - src[1].NegateBase ? VSF_FLAG_ALL : VSF_FLAG_NONE) | (src[1].RelAddr << 4); + src[1].Negate ? VSF_FLAG_ALL : VSF_FLAG_NONE) | (src[1].RelAddr << 4); o_inst->src2 = UNUSED_SRC_1; o_inst++; @@ -766,11 +766,11 @@ if ((o_inst - vp->instr) == 31) { o_inst->src1 = MAKE_VSF_SOURCE(t_src_index(vp, &src[1]), SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, t_src_class(src[1].File), - src[1].NegateBase) | (src[1].RelAddr << 4); + src[1].Negate) | (src[1].RelAddr << 4); o_inst->src2 = MAKE_VSF_SOURCE(t_src_index(vp, &src[1]), SWIZZLE_Y, SWIZZLE_Y, SWIZZLE_Y, SWIZZLE_Y, t_src_class(src[1].File), - src[1].NegateBase) | (src[1].RelAddr << 4); + src[1].Negate) | (src[1].RelAddr << 4); } else { o_inst->src1 = t_src(vp, &src[1]); @@ -792,7 +792,7 @@ else { t_swizzle(GET_SWZ(src[0].Swizzle, 2)), SWIZZLE_ZERO, t_src_class(src[0].File), - src[0].NegateBase) | (src[0].RelAddr << 4); + src[0].Negate) | (src[0].RelAddr << 4); o_inst->src1 = MAKE_VSF_SOURCE(t_src_index(vp, &src[1]), t_swizzle(GET_SWZ(src[1].Swizzle, 0)), @@ -800,7 +800,7 @@ else { t_swizzle(GET_SWZ(src[1].Swizzle, 2)), SWIZZLE_ZERO, t_src_class(src[1].File), - src[1].NegateBase) | (src[1].RelAddr << 4); + src[1].Negate) | (src[1].RelAddr << 4); o_inst->src2 = UNUSED_SRC_1; goto next; @@ -815,7 +815,7 @@ else { t_swizzle(GET_SWZ(src[0].Swizzle, 2)), VSF_IN_COMPONENT_ONE, t_src_class(src[0].File), - src[0].NegateBase) | (src[0].RelAddr << 4); + src[0].Negate) | (src[0].RelAddr << 4); o_inst->src1 = t_src(vp, &src[1]); o_inst->src2 = UNUSED_SRC_1; goto next; @@ -831,7 +831,7 @@ else { t_swizzle(GET_SWZ(src[1].Swizzle, 2)), t_swizzle(GET_SWZ(src[1].Swizzle, 3)), t_src_class(src[1].File), - (!src[1].NegateBase) ? VSF_FLAG_ALL : VSF_FLAG_NONE) | (src[1].RelAddr << 4); + (!src[1].Negate) ? VSF_FLAG_ALL : VSF_FLAG_NONE) | (src[1].RelAddr << 4); o_inst->src2 = UNUSED_SRC_1; goto next; @@ -846,7 +846,7 @@ else { t_swizzle(GET_SWZ(src[0].Swizzle, 2)), t_swizzle(GET_SWZ(src[0].Swizzle, 3)), t_src_class(src[0].File), - (!src[0].NegateBase) ? VSF_FLAG_ALL : VSF_FLAG_NONE) | (src[0].RelAddr << 4); + (!src[0].Negate) ? VSF_FLAG_ALL : VSF_FLAG_NONE) | (src[0].RelAddr << 4); o_inst->src2 = UNUSED_SRC_1; goto next; @@ -874,7 +874,7 @@ else { VSF_IN_COMPONENT_W, VSF_IN_CLASS_TMP, /* Not 100% sure about this */ - (!src[0].NegateBase) ? VSF_FLAG_ALL : VSF_FLAG_NONE/*VSF_FLAG_ALL*/); + (!src[0].Negate) ? VSF_FLAG_ALL : VSF_FLAG_NONE/*VSF_FLAG_ALL*/); o_inst->src2 = UNUSED_SRC_0; u_temp_i--; @@ -899,7 +899,7 @@ else { t_swizzle(GET_SWZ(src[0].Swizzle, 0)), // x t_swizzle(GET_SWZ(src[0].Swizzle, 3)), // w t_src_class(src[0].File), - src[0].NegateBase) | (src[0].RelAddr << 4); + src[0].Negate) | (src[0].RelAddr << 4); o_inst->src1 = MAKE_VSF_SOURCE(t_src_index(vp, &src[1]), t_swizzle(GET_SWZ(src[1].Swizzle, 2)), // z @@ -907,7 +907,7 @@ else { t_swizzle(GET_SWZ(src[1].Swizzle, 1)), // y t_swizzle(GET_SWZ(src[1].Swizzle, 3)), // w t_src_class(src[1].File), - src[1].NegateBase) | (src[1].RelAddr << 4); + src[1].Negate) | (src[1].RelAddr << 4); o_inst->src2 = UNUSED_SRC_1; o_inst++; @@ -922,7 +922,7 @@ else { t_swizzle(GET_SWZ(src[1].Swizzle, 0)), // x t_swizzle(GET_SWZ(src[1].Swizzle, 3)), // w t_src_class(src[1].File), - (!src[1].NegateBase) ? VSF_FLAG_ALL : VSF_FLAG_NONE) | (src[1].RelAddr << 4); + (!src[1].Negate) ? VSF_FLAG_ALL : VSF_FLAG_NONE) | (src[1].RelAddr << 4); o_inst->src1 = MAKE_VSF_SOURCE(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 2)), // z @@ -930,7 +930,7 @@ else { t_swizzle(GET_SWZ(src[0].Swizzle, 1)), // y t_swizzle(GET_SWZ(src[0].Swizzle, 3)), // w t_src_class(src[0].File), - src[0].NegateBase) | (src[0].RelAddr << 4); + src[0].Negate) | (src[0].RelAddr << 4); o_inst->src2 = MAKE_VSF_SOURCE(u_temp_i+1, VSF_IN_COMPONENT_X, diff --git a/src/mesa/drivers/dri/r300/r300_fragprog.c b/src/mesa/drivers/dri/r300/r300_fragprog.c index 921ca33c750..ea530fd00e7 100644 --- a/src/mesa/drivers/dri/r300/r300_fragprog.c +++ b/src/mesa/drivers/dri/r300/r300_fragprog.c @@ -194,9 +194,9 @@ GLboolean r300_transform_TEX( * r < tex <=> -tex+r < 0 * r >= tex <=> not (-tex+r < 0 */ if (comparefunc == GL_LESS || comparefunc == GL_GEQUAL) - tgt[1].SrcReg[2].NegateBase = tgt[0].SrcReg[2].NegateBase ^ NEGATE_XYZW; + tgt[1].SrcReg[2].Negate = tgt[0].SrcReg[2].Negate ^ NEGATE_XYZW; else - tgt[1].SrcReg[0].NegateBase = tgt[0].SrcReg[0].NegateBase ^ NEGATE_XYZW; + tgt[1].SrcReg[0].Negate = tgt[0].SrcReg[0].Negate ^ NEGATE_XYZW; tgt[2].Opcode = OPCODE_CMP; tgt[2].DstReg = orig_inst->DstReg; diff --git a/src/mesa/drivers/dri/r300/r300_fragprog_swizzle.c b/src/mesa/drivers/dri/r300/r300_fragprog_swizzle.c index a86d2bd4712..fc9d855bce6 100644 --- a/src/mesa/drivers/dri/r300/r300_fragprog_swizzle.c +++ b/src/mesa/drivers/dri/r300/r300_fragprog_swizzle.c @@ -92,7 +92,7 @@ static const struct swizzle_data* lookup_native_swizzle(GLuint swizzle) GLboolean r300FPIsNativeSwizzle(GLuint opcode, struct prog_src_register reg) { if (reg.Abs) - reg.NegateBase = 0; + reg.Negate = NEGATE_NONE; if (opcode == OPCODE_KIL || opcode == OPCODE_TEX || @@ -100,7 +100,7 @@ GLboolean r300FPIsNativeSwizzle(GLuint opcode, struct prog_src_register reg) opcode == OPCODE_TXP) { int j; - if (reg.Abs || reg.NegateBase != (15*reg.NegateAbs)) + if (reg.Abs || reg.Negate) return GL_FALSE; for(j = 0; j < 4; ++j) { @@ -121,7 +121,7 @@ GLboolean r300FPIsNativeSwizzle(GLuint opcode, struct prog_src_register reg) if (GET_SWZ(reg.Swizzle, j) != SWIZZLE_NIL) relevant |= 1 << j; - if ((reg.NegateBase & relevant) && (reg.NegateBase & relevant) != relevant) + if ((reg.Negate & relevant) && ((reg.Negate & relevant) != relevant)) return GL_FALSE; if (!lookup_native_swizzle(reg.Swizzle)) @@ -137,13 +137,12 @@ GLboolean r300FPIsNativeSwizzle(GLuint opcode, struct prog_src_register reg) void r300FPBuildSwizzle(struct nqssadce_state *s, struct prog_dst_register dst, struct prog_src_register src) { if (src.Abs) - src.NegateBase = 0; + src.Negate = NEGATE_NONE; while(dst.WriteMask) { const struct swizzle_data *best_swizzle = 0; GLuint best_matchcount = 0; GLuint best_matchmask = 0; - GLboolean rgbnegate; int i, comp; for(i = 0; i < num_native_swizzles; ++i) { @@ -157,6 +156,11 @@ void r300FPBuildSwizzle(struct nqssadce_state *s, struct prog_dst_register dst, if (swz == SWIZZLE_NIL) continue; if (swz == GET_SWZ(sd->hash, comp)) { + /* check if the negate bit of current component + * is the same for already matched components */ + if (matchmask && (!!(src.Negate & matchmask) != !!(src.Negate & (1 << comp)))) + continue; + matchcount++; matchmask |= 1 << comp; } @@ -170,13 +174,6 @@ void r300FPBuildSwizzle(struct nqssadce_state *s, struct prog_dst_register dst, } } - if ((src.NegateBase & best_matchmask) != 0) { - best_matchmask &= src.NegateBase; - rgbnegate = !src.NegateAbs; - } else { - rgbnegate = src.NegateAbs; - } - struct prog_instruction *inst; _mesa_insert_instructions(s->Program, s->IP, 1); @@ -185,6 +182,7 @@ void r300FPBuildSwizzle(struct nqssadce_state *s, struct prog_dst_register dst, inst->DstReg = dst; inst->DstReg.WriteMask &= (best_matchmask | WRITEMASK_W); inst->SrcReg[0] = src; + inst->SrcReg[0].Negate = (best_matchmask & src.Negate) ? NEGATE_XYZW : NEGATE_NONE; /* Note: We rely on NqSSA/DCE to set unused swizzle components to NIL */ dst.WriteMask &= ~inst->DstReg.WriteMask; diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.c b/src/mesa/drivers/dri/r300/r300_vertprog.c index 50806575ced..146daa367cd 100644 --- a/src/mesa/drivers/dri/r300/r300_vertprog.c +++ b/src/mesa/drivers/dri/r300/r300_vertprog.c @@ -245,7 +245,7 @@ static unsigned long t_src_index(struct r300_vertex_program *vp, static unsigned long t_src(struct r300_vertex_program *vp, struct prog_src_register *src) { - /* src->NegateBase uses the NEGATE_ flags from program_instruction.h, + /* src->Negate uses the NEGATE_ flags from program_instruction.h, * which equal our VSF_FLAGS_ values, so it's safe to just pass it here. */ return PVS_SRC_OPERAND(t_src_index(vp, src), @@ -254,13 +254,13 @@ static unsigned long t_src(struct r300_vertex_program *vp, t_swizzle(GET_SWZ(src->Swizzle, 2)), t_swizzle(GET_SWZ(src->Swizzle, 3)), t_src_class(src->File), - src->NegateBase) | (src->RelAddr << 4); + src->Negate) | (src->RelAddr << 4); } static unsigned long t_src_scalar(struct r300_vertex_program *vp, struct prog_src_register *src) { - /* src->NegateBase uses the NEGATE_ flags from program_instruction.h, + /* src->Negate uses the NEGATE_ flags from program_instruction.h, * which equal our VSF_FLAGS_ values, so it's safe to just pass it here. */ return PVS_SRC_OPERAND(t_src_index(vp, src), @@ -269,8 +269,7 @@ static unsigned long t_src_scalar(struct r300_vertex_program *vp, t_swizzle(GET_SWZ(src->Swizzle, 0)), t_swizzle(GET_SWZ(src->Swizzle, 0)), t_src_class(src->File), - src-> - NegateBase ? VSF_FLAG_ALL : VSF_FLAG_NONE) | + src->Negate ? VSF_FLAG_ALL : VSF_FLAG_NONE) | (src->RelAddr << 4); } @@ -307,7 +306,7 @@ static GLuint *r300TranslateOpcodeABS(struct r300_vertex_program *vp, t_swizzle(GET_SWZ(src[0].Swizzle, 3)), t_src_class(src[0].File), (!src[0]. - NegateBase) ? VSF_FLAG_ALL : VSF_FLAG_NONE) | + Negate) ? VSF_FLAG_ALL : VSF_FLAG_NONE) | (src[0].RelAddr << 4); inst[3] = 0; @@ -369,8 +368,7 @@ static GLuint *r300TranslateOpcodeDP3(struct r300_vertex_program *vp, t_swizzle(GET_SWZ(src[0].Swizzle, 2)), SWIZZLE_ZERO, t_src_class(src[0].File), - src[0]. - NegateBase ? VSF_FLAG_XYZ : VSF_FLAG_NONE) | + src[0].Negate ? VSF_FLAG_XYZ : VSF_FLAG_NONE) | (src[0].RelAddr << 4); inst[2] = PVS_SRC_OPERAND(t_src_index(vp, &src[1]), @@ -378,8 +376,7 @@ static GLuint *r300TranslateOpcodeDP3(struct r300_vertex_program *vp, t_swizzle(GET_SWZ(src[1].Swizzle, 1)), t_swizzle(GET_SWZ(src[1].Swizzle, 2)), SWIZZLE_ZERO, t_src_class(src[1].File), - src[1]. - NegateBase ? VSF_FLAG_XYZ : VSF_FLAG_NONE) | + src[1].Negate ? VSF_FLAG_XYZ : VSF_FLAG_NONE) | (src[1].RelAddr << 4); inst[3] = __CONST(1, SWIZZLE_ZERO); @@ -422,8 +419,7 @@ static GLuint *r300TranslateOpcodeDPH(struct r300_vertex_program *vp, t_swizzle(GET_SWZ(src[0].Swizzle, 2)), PVS_SRC_SELECT_FORCE_1, t_src_class(src[0].File), - src[0]. - NegateBase ? VSF_FLAG_XYZ : VSF_FLAG_NONE) | + src[0].Negate ? VSF_FLAG_XYZ : VSF_FLAG_NONE) | (src[0].RelAddr << 4); inst[2] = t_src(vp, &src[1]); inst[3] = __CONST(1, SWIZZLE_ZERO); @@ -519,7 +515,7 @@ static GLuint *r300TranslateOpcodeFLR(struct r300_vertex_program *vp, PVS_SRC_SELECT_W, PVS_SRC_REG_TEMPORARY, /* Not 100% sure about this */ (!src[0]. - NegateBase) ? VSF_FLAG_ALL : VSF_FLAG_NONE + Negate) ? VSF_FLAG_ALL : VSF_FLAG_NONE /*VSF_FLAG_ALL */ ); inst[3] = __CONST(0, SWIZZLE_ZERO); (*u_temp_i)--; @@ -564,8 +560,7 @@ static GLuint *r300TranslateOpcodeLG2(struct r300_vertex_program *vp, t_swizzle(GET_SWZ(src[0].Swizzle, 0)), t_swizzle(GET_SWZ(src[0].Swizzle, 0)), t_src_class(src[0].File), - src[0]. - NegateBase ? VSF_FLAG_ALL : VSF_FLAG_NONE) | + src[0].Negate ? VSF_FLAG_ALL : VSF_FLAG_NONE) | (src[0].RelAddr << 4); inst[2] = __CONST(0, SWIZZLE_ZERO); inst[3] = __CONST(0, SWIZZLE_ZERO); @@ -592,24 +587,21 @@ static GLuint *r300TranslateOpcodeLIT(struct r300_vertex_program *vp, PVS_SRC_SELECT_FORCE_0, // Z t_swizzle(GET_SWZ(src[0].Swizzle, 1)), // Y t_src_class(src[0].File), - src[0]. - NegateBase ? VSF_FLAG_ALL : VSF_FLAG_NONE) | + src[0].Negate ? VSF_FLAG_ALL : VSF_FLAG_NONE) | (src[0].RelAddr << 4); inst[2] = PVS_SRC_OPERAND(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 1)), // Y t_swizzle(GET_SWZ(src[0].Swizzle, 3)), // W PVS_SRC_SELECT_FORCE_0, // Z t_swizzle(GET_SWZ(src[0].Swizzle, 0)), // X t_src_class(src[0].File), - src[0]. - NegateBase ? VSF_FLAG_ALL : VSF_FLAG_NONE) | + src[0].Negate ? VSF_FLAG_ALL : VSF_FLAG_NONE) | (src[0].RelAddr << 4); inst[3] = PVS_SRC_OPERAND(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 1)), // Y t_swizzle(GET_SWZ(src[0].Swizzle, 0)), // X PVS_SRC_SELECT_FORCE_0, // Z t_swizzle(GET_SWZ(src[0].Swizzle, 3)), // W t_src_class(src[0].File), - src[0]. - NegateBase ? VSF_FLAG_ALL : VSF_FLAG_NONE) | + src[0].Negate ? VSF_FLAG_ALL : VSF_FLAG_NONE) | (src[0].RelAddr << 4); return inst; @@ -837,7 +829,7 @@ static GLuint *r300TranslateOpcodeSUB(struct r300_vertex_program *vp, t_swizzle(GET_SWZ(src[1].Swizzle, 3)), t_src_class(src[1].File), (!src[1]. - NegateBase) ? VSF_FLAG_ALL : VSF_FLAG_NONE) | + Negate) ? VSF_FLAG_ALL : VSF_FLAG_NONE) | (src[1].RelAddr << 4); inst[3] = 0; #else @@ -857,7 +849,7 @@ static GLuint *r300TranslateOpcodeSUB(struct r300_vertex_program *vp, t_swizzle(GET_SWZ(src[1].Swizzle, 3)), t_src_class(src[1].File), (!src[1]. - NegateBase) ? VSF_FLAG_ALL : VSF_FLAG_NONE) | + Negate) ? VSF_FLAG_ALL : VSF_FLAG_NONE) | (src[1].RelAddr << 4); #endif @@ -905,16 +897,14 @@ static GLuint *r300TranslateOpcodeXPD(struct r300_vertex_program *vp, t_swizzle(GET_SWZ(src[0].Swizzle, 0)), // X t_swizzle(GET_SWZ(src[0].Swizzle, 3)), // W t_src_class(src[0].File), - src[0]. - NegateBase ? VSF_FLAG_ALL : VSF_FLAG_NONE) | + src[0].Negate ? VSF_FLAG_ALL : VSF_FLAG_NONE) | (src[0].RelAddr << 4); inst[2] = PVS_SRC_OPERAND(t_src_index(vp, &src[1]), t_swizzle(GET_SWZ(src[1].Swizzle, 2)), // Z t_swizzle(GET_SWZ(src[1].Swizzle, 0)), // X t_swizzle(GET_SWZ(src[1].Swizzle, 1)), // Y t_swizzle(GET_SWZ(src[1].Swizzle, 3)), // W t_src_class(src[1].File), - src[1]. - NegateBase ? VSF_FLAG_ALL : VSF_FLAG_NONE) | + src[1].Negate ? VSF_FLAG_ALL : VSF_FLAG_NONE) | (src[1].RelAddr << 4); inst[3] = __CONST(1, SWIZZLE_ZERO); inst += 4; @@ -931,15 +921,14 @@ static GLuint *r300TranslateOpcodeXPD(struct r300_vertex_program *vp, t_swizzle(GET_SWZ(src[1].Swizzle, 3)), // W t_src_class(src[1].File), (!src[1]. - NegateBase) ? VSF_FLAG_ALL : VSF_FLAG_NONE) | + Negate) ? VSF_FLAG_ALL : VSF_FLAG_NONE) | (src[1].RelAddr << 4); inst[2] = PVS_SRC_OPERAND(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 2)), // Z t_swizzle(GET_SWZ(src[0].Swizzle, 0)), // X t_swizzle(GET_SWZ(src[0].Swizzle, 1)), // Y t_swizzle(GET_SWZ(src[0].Swizzle, 3)), // W t_src_class(src[0].File), - src[0]. - NegateBase ? VSF_FLAG_ALL : VSF_FLAG_NONE) | + src[0].Negate ? VSF_FLAG_ALL : VSF_FLAG_NONE) | (src[0].RelAddr << 4); inst[3] = PVS_SRC_OPERAND(*u_temp_i, PVS_SRC_SELECT_X, PVS_SRC_SELECT_Y, diff --git a/src/mesa/drivers/dri/r300/r500_fragprog.c b/src/mesa/drivers/dri/r300/r500_fragprog.c index e9c0d89dd43..8f0b70ad3a3 100644 --- a/src/mesa/drivers/dri/r300/r500_fragprog.c +++ b/src/mesa/drivers/dri/r300/r500_fragprog.c @@ -152,9 +152,9 @@ GLboolean r500_transform_TEX( * r < tex <=> -tex+r < 0 * r >= tex <=> not (-tex+r < 0 */ if (comparefunc == GL_LESS || comparefunc == GL_GEQUAL) - tgt[1].SrcReg[2].NegateBase = tgt[0].SrcReg[2].NegateBase ^ NEGATE_XYZW; + tgt[1].SrcReg[2].Negate = tgt[0].SrcReg[2].Negate ^ NEGATE_XYZW; else - tgt[1].SrcReg[0].NegateBase = tgt[0].SrcReg[0].NegateBase ^ NEGATE_XYZW; + tgt[1].SrcReg[0].Negate = tgt[0].SrcReg[0].Negate ^ NEGATE_XYZW; tgt[2].Opcode = OPCODE_CMP; tgt[2].DstReg = orig_inst->DstReg; @@ -196,8 +196,8 @@ GLboolean r500FPIsNativeSwizzle(GLuint opcode, struct prog_src_register reg) if (reg.Abs) return GL_FALSE; - if (reg.NegateAbs) - reg.NegateBase ^= 15; + if (reg.Negate) + reg.Negate ^= NEGATE_XYZW; if (opcode == OPCODE_KIL) { if (reg.Swizzle != SWIZZLE_NOOP) @@ -206,7 +206,7 @@ GLboolean r500FPIsNativeSwizzle(GLuint opcode, struct prog_src_register reg) for(i = 0; i < 4; ++i) { GLuint swz = GET_SWZ(reg.Swizzle, i); if (swz == SWIZZLE_NIL) { - reg.NegateBase &= ~(1 << i); + reg.Negate &= ~(1 << i); continue; } if (swz >= 4) @@ -214,15 +214,14 @@ GLboolean r500FPIsNativeSwizzle(GLuint opcode, struct prog_src_register reg) } } - if (reg.NegateBase) + if (reg.Negate) return GL_FALSE; return GL_TRUE; } else if (opcode == OPCODE_DDX || opcode == OPCODE_DDY) { /* DDX/MDH and DDY/MDV explicitly ignore incoming swizzles; * if it doesn't fit perfectly into a .xyzw case... */ - if (reg.Swizzle == SWIZZLE_NOOP && !reg.Abs - && !reg.NegateBase && !reg.NegateAbs) + if (reg.Swizzle == SWIZZLE_NOOP && !reg.Abs && !reg.Negate) return GL_TRUE; return GL_FALSE; @@ -237,7 +236,7 @@ GLboolean r500FPIsNativeSwizzle(GLuint opcode, struct prog_src_register reg) if (swz != SWIZZLE_NIL && swz != SWIZZLE_ZERO) relevant |= 1 << i; } - if ((reg.NegateBase & relevant) && ((reg.NegateBase & relevant) != relevant)) + if ((reg.Negate & relevant) && ((reg.Negate & relevant) != relevant)) return GL_FALSE; return GL_TRUE; @@ -260,7 +259,7 @@ void r500FPBuildSwizzle(struct nqssadce_state *s, struct prog_dst_register dst, GLuint swz = GET_SWZ(src.Swizzle, i); if (swz == SWIZZLE_NIL) continue; - negatebase[GET_BIT(src.NegateBase, i)] |= 1 << i; + negatebase[GET_BIT(src.Negate, i)] |= 1 << i; } _mesa_insert_instructions(s->Program, s->IP, (negatebase[0] ? 1 : 0) + (negatebase[1] ? 1 : 0)); diff --git a/src/mesa/drivers/dri/r300/radeon_nqssadce.c b/src/mesa/drivers/dri/r300/radeon_nqssadce.c index a083c3d2436..4a2e1cba402 100644 --- a/src/mesa/drivers/dri/r300/radeon_nqssadce.c +++ b/src/mesa/drivers/dri/r300/radeon_nqssadce.c @@ -61,12 +61,12 @@ static struct prog_src_register lmul_swizzle(GLuint swizzle, struct prog_src_reg struct prog_src_register tmp = srcreg; int i; tmp.Swizzle = 0; - tmp.NegateBase = 0; + tmp.Negate = NEGATE_NONE; for(i = 0; i < 4; ++i) { GLuint swz = GET_SWZ(swizzle, i); if (swz < 4) { tmp.Swizzle |= GET_SWZ(srcreg.Swizzle, swz) << (i*3); - tmp.NegateBase |= GET_BIT(srcreg.NegateBase, swz) << i; + tmp.Negate |= GET_BIT(srcreg.Negate, swz) << i; } else { tmp.Swizzle |= swz << (i*3); } @@ -103,9 +103,8 @@ static struct prog_instruction* track_used_srcreg(struct nqssadce_state* s, inst->SrcReg[src].File = PROGRAM_TEMPORARY; inst->SrcReg[src].Index = dstreg.Index; inst->SrcReg[src].Swizzle = 0; - inst->SrcReg[src].NegateBase = 0; + inst->SrcReg[src].Negate = NEGATE_NONE; inst->SrcReg[src].Abs = 0; - inst->SrcReg[src].NegateAbs = 0; for(i = 0; i < 4; ++i) { if (GET_BIT(sourced, i)) inst->SrcReg[src].Swizzle |= i << (3*i); diff --git a/src/mesa/drivers/dri/r300/radeon_program_alu.c b/src/mesa/drivers/dri/r300/radeon_program_alu.c index 1ef71e74dcf..8283723bad7 100644 --- a/src/mesa/drivers/dri/r300/radeon_program_alu.c +++ b/src/mesa/drivers/dri/r300/radeon_program_alu.c @@ -81,18 +81,6 @@ static struct prog_instruction *emit3(struct gl_program* p, return fpi; } -static void set_swizzle(struct prog_src_register *SrcReg, int coordinate, int swz) -{ - SrcReg->Swizzle &= ~(7 << (3*coordinate)); - SrcReg->Swizzle |= swz << (3*coordinate); -} - -static void set_negate_base(struct prog_src_register *SrcReg, int coordinate, int negate) -{ - SrcReg->NegateBase &= ~(1 << coordinate); - SrcReg->NegateBase |= (negate << coordinate); -} - static struct prog_dst_register dstreg(int file, int index) { struct prog_dst_register dst; @@ -156,15 +144,14 @@ static struct prog_src_register absolute(struct prog_src_register reg) { struct prog_src_register newreg = reg; newreg.Abs = 1; - newreg.NegateBase = 0; - newreg.NegateAbs = 0; + newreg.Negate = NEGATE_NONE; return newreg; } static struct prog_src_register negate(struct prog_src_register reg) { struct prog_src_register newreg = reg; - newreg.NegateAbs = !newreg.NegateAbs; + newreg.Negate = newreg.Negate ^ NEGATE_XYZW; return newreg; } @@ -189,8 +176,7 @@ static void transform_ABS(struct radeon_transform_context* t, { struct prog_src_register src = inst->SrcReg[0]; src.Abs = 1; - src.NegateBase = 0; - src.NegateAbs = 0; + src.Negate = NEGATE_NONE; emit1(t->Program, OPCODE_MOV, inst->SaturateMode, inst->DstReg, src); } @@ -198,18 +184,9 @@ static void transform_DPH(struct radeon_transform_context* t, struct prog_instruction* inst) { struct prog_src_register src0 = inst->SrcReg[0]; - if (src0.NegateAbs) { - if (src0.Abs) { - int tempreg = radeonFindFreeTemporary(t); - emit1(t->Program, OPCODE_MOV, 0, dstreg(PROGRAM_TEMPORARY, tempreg), src0); - src0 = srcreg(src0.File, src0.Index); - } else { - src0.NegateAbs = 0; - src0.NegateBase ^= NEGATE_XYZW; - } - } - set_swizzle(&src0, 3, SWIZZLE_ONE); - set_negate_base(&src0, 3, 0); + src0.Negate &= ~NEGATE_W; + src0.Swizzle &= ~(7 << (3 * 3)); + src0.Swizzle |= SWIZZLE_ONE << (3 * 3); emit2(t->Program, OPCODE_DP4, inst->SaturateMode, inst->DstReg, src0, inst->SrcReg[1]); } @@ -649,7 +626,7 @@ GLboolean radeonTransformDeriv(struct radeon_transform_context* t, B.Swizzle = MAKE_SWIZZLE4(SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE); - B.NegateBase = NEGATE_XYZW; + B.Negate = NEGATE_XYZW; emit2(t->Program, inst->Opcode, inst->SaturateMode, inst->DstReg, inst->SrcReg[0], B); diff --git a/src/mesa/drivers/dri/r300/radeon_program_pair.c b/src/mesa/drivers/dri/r300/radeon_program_pair.c index 49aa90dd94a..5c6594bc2e1 100644 --- a/src/mesa/drivers/dri/r300/radeon_program_pair.c +++ b/src/mesa/drivers/dri/r300/radeon_program_pair.c @@ -255,8 +255,7 @@ static void final_rewrite(struct pair_state *s, struct prog_instruction *inst) inst->SrcReg[2] = inst->SrcReg[1]; inst->SrcReg[1].File = PROGRAM_BUILTIN; inst->SrcReg[1].Swizzle = SWIZZLE_1111; - inst->SrcReg[1].NegateBase = 0; - inst->SrcReg[1].NegateAbs = 0; + inst->SrcReg[1].Negate = NEGATE_NONE; inst->Opcode = OPCODE_MAD; break; case OPCODE_CMP: @@ -722,7 +721,6 @@ static GLboolean fill_instruction_into_pair(struct pair_state *s, struct radeon_ if (pairinst->NeedRGB && !pairinst->IsTranscendent) { GLboolean srcrgb = GL_FALSE; GLboolean srcalpha = GL_FALSE; - GLuint negatebase = 0; int j; for(j = 0; j < 3; ++j) { GLuint swz = GET_SWZ(inst->SrcReg[i].Swizzle, j); @@ -730,8 +728,6 @@ static GLboolean fill_instruction_into_pair(struct pair_state *s, struct radeon_ srcrgb = GL_TRUE; else if (swz < 4) srcalpha = GL_TRUE; - if (swz != SWIZZLE_NIL && GET_BIT(inst->SrcReg[i].NegateBase, j)) - negatebase = 1; } source = alloc_pair_source(s, pair, inst->SrcReg[i], srcrgb, srcalpha); if (source < 0) @@ -739,12 +735,11 @@ static GLboolean fill_instruction_into_pair(struct pair_state *s, struct radeon_ pair->RGB.Arg[i].Source = source; pair->RGB.Arg[i].Swizzle = inst->SrcReg[i].Swizzle & 0x1ff; pair->RGB.Arg[i].Abs = inst->SrcReg[i].Abs; - pair->RGB.Arg[i].Negate = (negatebase & ~pair->RGB.Arg[i].Abs) ^ inst->SrcReg[i].NegateAbs; + pair->RGB.Arg[i].Negate = !!(inst->SrcReg[i].Negate & (NEGATE_X | NEGATE_Y | NEGATE_Z)); } if (pairinst->NeedAlpha) { GLboolean srcrgb = GL_FALSE; GLboolean srcalpha = GL_FALSE; - GLuint negatebase = GET_BIT(inst->SrcReg[i].NegateBase, pairinst->IsTranscendent ? 0 : 3); GLuint swz = GET_SWZ(inst->SrcReg[i].Swizzle, pairinst->IsTranscendent ? 0 : 3); if (swz < 3) srcrgb = GL_TRUE; @@ -756,7 +751,7 @@ static GLboolean fill_instruction_into_pair(struct pair_state *s, struct radeon_ pair->Alpha.Arg[i].Source = source; pair->Alpha.Arg[i].Swizzle = swz; pair->Alpha.Arg[i].Abs = inst->SrcReg[i].Abs; - pair->Alpha.Arg[i].Negate = (negatebase & ~pair->RGB.Arg[i].Abs) ^ inst->SrcReg[i].NegateAbs; + pair->Alpha.Arg[i].Negate = !!(inst->SrcReg[i].Negate & NEGATE_W); } } diff --git a/src/mesa/drivers/windows/gdi/wmesa.c b/src/mesa/drivers/windows/gdi/wmesa.c index d1103f31670..e1971db6933 100644 --- a/src/mesa/drivers/windows/gdi/wmesa.c +++ b/src/mesa/drivers/windows/gdi/wmesa.c @@ -1679,80 +1679,3 @@ void WMesaShareLists(WMesaContext ctx_to_share, WMesaContext ctx) _mesa_share_state(&ctx->gl_ctx, &ctx_to_share->gl_ctx); } -/* This is hopefully a temporary hack to define some needed dispatch - * table entries. Hopefully, I'll find a better solution. The - * dispatch table generation scripts ought to be making these dummy - * stubs as well. */ -#if !defined(__MINGW32__) || !defined(GL_NO_STDCALL) -void gl_dispatch_stub_543(void){} -void gl_dispatch_stub_544(void){} -void gl_dispatch_stub_545(void){} -void gl_dispatch_stub_546(void){} -void gl_dispatch_stub_547(void){} -void gl_dispatch_stub_548(void){} -void gl_dispatch_stub_549(void){} -void gl_dispatch_stub_550(void){} -void gl_dispatch_stub_551(void){} -void gl_dispatch_stub_552(void){} -void gl_dispatch_stub_553(void){} -void gl_dispatch_stub_554(void){} -void gl_dispatch_stub_555(void){} -void gl_dispatch_stub_556(void){} -void gl_dispatch_stub_557(void){} -void gl_dispatch_stub_558(void){} -void gl_dispatch_stub_559(void){} -void gl_dispatch_stub_560(void){} -void gl_dispatch_stub_561(void){} -void gl_dispatch_stub_565(void){} -void gl_dispatch_stub_566(void){} -void gl_dispatch_stub_577(void){} -void gl_dispatch_stub_578(void){} -void gl_dispatch_stub_603(void){} -void gl_dispatch_stub_645(void){} -void gl_dispatch_stub_646(void){} -void gl_dispatch_stub_647(void){} -void gl_dispatch_stub_648(void){} -void gl_dispatch_stub_649(void){} -void gl_dispatch_stub_650(void){} -void gl_dispatch_stub_651(void){} -void gl_dispatch_stub_652(void){} -void gl_dispatch_stub_653(void){} -void gl_dispatch_stub_733(void){} -void gl_dispatch_stub_734(void){} -void gl_dispatch_stub_735(void){} -void gl_dispatch_stub_736(void){} -void gl_dispatch_stub_737(void){} -void gl_dispatch_stub_738(void){} -void gl_dispatch_stub_744(void){} -void gl_dispatch_stub_745(void){} -void gl_dispatch_stub_746(void){} -void gl_dispatch_stub_760(void){} -void gl_dispatch_stub_761(void){} -void gl_dispatch_stub_763(void){} -void gl_dispatch_stub_765(void){} -void gl_dispatch_stub_766(void){} -void gl_dispatch_stub_767(void){} -void gl_dispatch_stub_768(void){} - -void gl_dispatch_stub_562(void){} -void gl_dispatch_stub_563(void){} -void gl_dispatch_stub_564(void){} -void gl_dispatch_stub_567(void){} -void gl_dispatch_stub_568(void){} -void gl_dispatch_stub_569(void){} -void gl_dispatch_stub_580(void){} -void gl_dispatch_stub_581(void){} -void gl_dispatch_stub_606(void){} -void gl_dispatch_stub_654(void){} -void gl_dispatch_stub_655(void){} -void gl_dispatch_stub_656(void){} -void gl_dispatch_stub_739(void){} -void gl_dispatch_stub_740(void){} -void gl_dispatch_stub_741(void){} -void gl_dispatch_stub_748(void){} -void gl_dispatch_stub_749(void){} -void gl_dispatch_stub_769(void){} -void gl_dispatch_stub_770(void){} -void gl_dispatch_stub_771(void){} - -#endif diff --git a/src/mesa/glapi/gl_procs.py b/src/mesa/glapi/gl_procs.py index 96e59a58ca1..cd1a68cee17 100644 --- a/src/mesa/glapi/gl_procs.py +++ b/src/mesa/glapi/gl_procs.py @@ -134,12 +134,11 @@ typedef struct { print "#endif /* USE_MGL_NAMESPACE */" print '' print '' - print '/* FIXME: Having these (incorrect) prototypes here is ugly. */' print '#if defined(NEED_FUNCTION_POINTER) || defined(GLX_INDIRECT_RENDERING)' for func in api.functionIterateByOffset(): for n in func.entry_points: if (not func.is_static_entry_point(func.name)) or (func.has_different_protocol(n) and not func.is_static_entry_point(n)): - print 'extern void gl_dispatch_stub_%u(void);' % (func.offset) + print '%s GLAPIENTRY gl_dispatch_stub_%u(%s);' % (func.return_type, func.offset, func.get_parameter_string()) break print '#endif /* defined(NEED_FUNCTION_POINTER) || defined(GLX_INDIRECT_RENDERING) */' diff --git a/src/mesa/glapi/glprocs.h b/src/mesa/glapi/glprocs.h index 021a64efcd9..680893cfc4e 100644 --- a/src/mesa/glapi/glprocs.h +++ b/src/mesa/glapi/glprocs.h @@ -1178,54 +1178,53 @@ static const char gl_string_table[] = #endif /* USE_MGL_NAMESPACE */ -/* FIXME: Having these (incorrect) prototypes here is ugly. */ #if defined(NEED_FUNCTION_POINTER) || defined(GLX_INDIRECT_RENDERING) -extern void gl_dispatch_stub_343(void); -extern void gl_dispatch_stub_344(void); -extern void gl_dispatch_stub_345(void); -extern void gl_dispatch_stub_356(void); -extern void gl_dispatch_stub_357(void); -extern void gl_dispatch_stub_358(void); -extern void gl_dispatch_stub_359(void); -extern void gl_dispatch_stub_361(void); -extern void gl_dispatch_stub_362(void); -extern void gl_dispatch_stub_363(void); -extern void gl_dispatch_stub_364(void); -extern void gl_dispatch_stub_365(void); -extern void gl_dispatch_stub_366(void); -extern void gl_dispatch_stub_563(void); -extern void gl_dispatch_stub_564(void); -extern void gl_dispatch_stub_565(void); -extern void gl_dispatch_stub_566(void); -extern void gl_dispatch_stub_567(void); -extern void gl_dispatch_stub_568(void); -extern void gl_dispatch_stub_569(void); -extern void gl_dispatch_stub_570(void); -extern void gl_dispatch_stub_581(void); -extern void gl_dispatch_stub_582(void); -extern void gl_dispatch_stub_607(void); -extern void gl_dispatch_stub_649(void); -extern void gl_dispatch_stub_650(void); -extern void gl_dispatch_stub_651(void); -extern void gl_dispatch_stub_652(void); -extern void gl_dispatch_stub_653(void); -extern void gl_dispatch_stub_654(void); -extern void gl_dispatch_stub_655(void); -extern void gl_dispatch_stub_656(void); -extern void gl_dispatch_stub_657(void); -extern void gl_dispatch_stub_738(void); -extern void gl_dispatch_stub_739(void); -extern void gl_dispatch_stub_740(void); -extern void gl_dispatch_stub_741(void); -extern void gl_dispatch_stub_742(void); -extern void gl_dispatch_stub_749(void); -extern void gl_dispatch_stub_750(void); -extern void gl_dispatch_stub_768(void); -extern void gl_dispatch_stub_770(void); -extern void gl_dispatch_stub_771(void); -extern void gl_dispatch_stub_772(void); -extern void gl_dispatch_stub_773(void); -extern void gl_dispatch_stub_774(void); +void GLAPIENTRY gl_dispatch_stub_343(GLenum target, GLenum format, GLenum type, GLvoid * table); +void GLAPIENTRY gl_dispatch_stub_344(GLenum target, GLenum pname, GLfloat * params); +void GLAPIENTRY gl_dispatch_stub_345(GLenum target, GLenum pname, GLint * params); +void GLAPIENTRY gl_dispatch_stub_356(GLenum target, GLenum format, GLenum type, GLvoid * image); +void GLAPIENTRY gl_dispatch_stub_357(GLenum target, GLenum pname, GLfloat * params); +void GLAPIENTRY gl_dispatch_stub_358(GLenum target, GLenum pname, GLint * params); +void GLAPIENTRY gl_dispatch_stub_359(GLenum target, GLenum format, GLenum type, GLvoid * row, GLvoid * column, GLvoid * span); +void GLAPIENTRY gl_dispatch_stub_361(GLenum target, GLboolean reset, GLenum format, GLenum type, GLvoid * values); +void GLAPIENTRY gl_dispatch_stub_362(GLenum target, GLenum pname, GLfloat * params); +void GLAPIENTRY gl_dispatch_stub_363(GLenum target, GLenum pname, GLint * params); +void GLAPIENTRY gl_dispatch_stub_364(GLenum target, GLboolean reset, GLenum format, GLenum type, GLvoid * values); +void GLAPIENTRY gl_dispatch_stub_365(GLenum target, GLenum pname, GLfloat * params); +void GLAPIENTRY gl_dispatch_stub_366(GLenum target, GLenum pname, GLint * params); +void GLAPIENTRY gl_dispatch_stub_563(GLenum pname, GLfloat * params); +void GLAPIENTRY gl_dispatch_stub_564(GLenum pname, GLint * params); +void GLAPIENTRY gl_dispatch_stub_565(GLenum pname, GLfloat param); +void GLAPIENTRY gl_dispatch_stub_566(GLenum pname, const GLfloat * params); +void GLAPIENTRY gl_dispatch_stub_567(GLenum pname, GLint param); +void GLAPIENTRY gl_dispatch_stub_568(GLenum pname, const GLint * params); +void GLAPIENTRY gl_dispatch_stub_569(GLclampf value, GLboolean invert); +void GLAPIENTRY gl_dispatch_stub_570(GLenum pattern); +void GLAPIENTRY gl_dispatch_stub_581(GLenum pname, GLdouble * params); +void GLAPIENTRY gl_dispatch_stub_582(GLenum pname, GLfloat * params); +void GLAPIENTRY gl_dispatch_stub_607(GLenum mode); +void GLAPIENTRY gl_dispatch_stub_649(const GLenum * mode, const GLint * first, const GLsizei * count, GLsizei primcount, GLint modestride); +void GLAPIENTRY gl_dispatch_stub_650(const GLenum * mode, const GLsizei * count, GLenum type, const GLvoid * const * indices, GLsizei primcount, GLint modestride); +void GLAPIENTRY gl_dispatch_stub_651(GLsizei n, const GLuint * fences); +void GLAPIENTRY gl_dispatch_stub_652(GLuint fence); +void GLAPIENTRY gl_dispatch_stub_653(GLsizei n, GLuint * fences); +void GLAPIENTRY gl_dispatch_stub_654(GLuint fence, GLenum pname, GLint * params); +GLboolean GLAPIENTRY gl_dispatch_stub_655(GLuint fence); +void GLAPIENTRY gl_dispatch_stub_656(GLuint fence, GLenum condition); +GLboolean GLAPIENTRY gl_dispatch_stub_657(GLuint fence); +void GLAPIENTRY gl_dispatch_stub_738(GLenum face); +void GLAPIENTRY gl_dispatch_stub_739(GLuint array); +void GLAPIENTRY gl_dispatch_stub_740(GLsizei n, const GLuint * arrays); +void GLAPIENTRY gl_dispatch_stub_741(GLsizei n, GLuint * arrays); +GLboolean GLAPIENTRY gl_dispatch_stub_742(GLuint array); +void GLAPIENTRY gl_dispatch_stub_749(GLclampd zmin, GLclampd zmax); +void GLAPIENTRY gl_dispatch_stub_750(GLenum modeRGB, GLenum modeA); +void GLAPIENTRY gl_dispatch_stub_768(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter); +void GLAPIENTRY gl_dispatch_stub_770(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask); +void GLAPIENTRY gl_dispatch_stub_771(GLenum target, GLuint index, GLsizei count, const GLfloat * params); +void GLAPIENTRY gl_dispatch_stub_772(GLenum target, GLuint index, GLsizei count, const GLfloat * params); +void GLAPIENTRY gl_dispatch_stub_773(GLuint id, GLenum pname, GLint64EXT * params); +void GLAPIENTRY gl_dispatch_stub_774(GLuint id, GLenum pname, GLuint64EXT * params); #endif /* defined(NEED_FUNCTION_POINTER) || defined(GLX_INDIRECT_RENDERING) */ static const glprocs_table_t static_functions[] = { diff --git a/src/mesa/main/attrib.c b/src/mesa/main/attrib.c index d5d0a552dbf..e43fa96dd31 100644 --- a/src/mesa/main/attrib.c +++ b/src/mesa/main/attrib.c @@ -1265,7 +1265,7 @@ _mesa_PopAttrib(void) /* restore clip planes */ for (i = 0; i < MAX_CLIP_PLANES; i++) { - const GLuint mask = 1 << 1; + const GLuint mask = 1 << i; const GLfloat *eyePlane = xform->EyeUserPlane[i]; COPY_4V(ctx->Transform.EyeUserPlane[i], eyePlane); if (xform->ClipPlanesEnabled & mask) { diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index b24a3b4409b..5726dbd983e 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -187,7 +187,7 @@ GLfloat _mesa_ubyte_to_float_color_tab[256]; void _mesa_notifySwapBuffers(__GLcontext *ctx) { - FLUSH_VERTICES( ctx, 0 ); + FLUSH_CURRENT( ctx, 0 ); if (ctx->Driver.Flush) { ctx->Driver.Flush(ctx); } @@ -434,7 +434,7 @@ one_time_init( GLcontext *ctx ) } #if defined(DEBUG) && defined(__DATE__) && defined(__TIME__) - _mesa_debug(ctx, "Mesa %s DEBUG build %s %s\n", + _mesa_debug(ctx, "Mesa %s DEBUG build %s %s", MESA_VERSION_STRING, __DATE__, __TIME__); #endif @@ -1514,6 +1514,7 @@ _mesa_Flush(void) { GET_CURRENT_CONTEXT(ctx); ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx); + FLUSH_CURRENT( ctx, 0 ); if (ctx->Driver.Flush) { ctx->Driver.Flush(ctx); } diff --git a/src/mesa/main/dd.h b/src/mesa/main/dd.h index d994401e55f..32b1d4e9faf 100644 --- a/src/mesa/main/dd.h +++ b/src/mesa/main/dd.h @@ -911,7 +911,7 @@ struct dd_function_table { void (*Uniform)(GLcontext *ctx, GLint location, GLsizei count, const GLvoid *values, GLenum type); void (*UniformMatrix)(GLcontext *ctx, GLint cols, GLint rows, - GLenum matrixType, GLint location, GLsizei count, + GLint location, GLsizei count, GLboolean transpose, const GLfloat *values); void (*UseProgram)(GLcontext *ctx, GLuint program); void (*ValidateProgram)(GLcontext *ctx, GLuint program); diff --git a/src/mesa/main/extensions.c b/src/mesa/main/extensions.c index 147d923e64f..5c4bea9cf62 100644 --- a/src/mesa/main/extensions.c +++ b/src/mesa/main/extensions.c @@ -31,7 +31,7 @@ #include "mtypes.h" -#define F(x) (int)(uintptr_t)&(((struct gl_extensions *)0)->x) +#define F(x) offsetof(struct gl_extensions, x) #define ON GL_TRUE #define OFF GL_FALSE diff --git a/src/mesa/main/fbobject.c b/src/mesa/main/fbobject.c index 151e29053ab..9c5a5908a22 100644 --- a/src/mesa/main/fbobject.c +++ b/src/mesa/main/fbobject.c @@ -692,7 +692,7 @@ _mesa_BindRenderbufferEXT(GLenum target, GLuint renderbuffer) return; } - FLUSH_VERTICES(ctx, _NEW_BUFFERS); + FLUSH_CURRENT(ctx, _NEW_BUFFERS); /* The above doesn't fully flush the drivers in the way that a * glFlush does, but that is required here: */ @@ -1187,8 +1187,7 @@ _mesa_BindFramebufferEXT(GLenum target, GLuint framebuffer) return; } - FLUSH_VERTICES(ctx, _NEW_BUFFERS); - + FLUSH_CURRENT(ctx, _NEW_BUFFERS); if (ctx->Driver.Flush) { ctx->Driver.Flush(ctx); } @@ -1269,7 +1268,7 @@ _mesa_DeleteFramebuffersEXT(GLsizei n, const GLuint *framebuffers) GET_CURRENT_CONTEXT(ctx); ASSERT_OUTSIDE_BEGIN_END(ctx); - FLUSH_VERTICES(ctx, _NEW_BUFFERS); + FLUSH_CURRENT(ctx, _NEW_BUFFERS); /* The above doesn't fully flush the drivers in the way that a * glFlush does, but that is required here: */ @@ -1507,7 +1506,7 @@ framebuffer_texture(GLcontext *ctx, const char *caller, GLenum target, } } - FLUSH_VERTICES(ctx, _NEW_BUFFERS); + FLUSH_CURRENT(ctx, _NEW_BUFFERS); /* The above doesn't fully flush the drivers in the way that a * glFlush does, but that is required here: */ @@ -1688,7 +1687,7 @@ _mesa_FramebufferRenderbufferEXT(GLenum target, GLenum attachment, } - FLUSH_VERTICES(ctx, _NEW_BUFFERS); + FLUSH_CURRENT(ctx, _NEW_BUFFERS); /* The above doesn't fully flush the drivers in the way that a * glFlush does, but that is required here: */ @@ -1769,7 +1768,7 @@ _mesa_GetFramebufferAttachmentParameterivEXT(GLenum target, GLenum attachment, } } - FLUSH_VERTICES(ctx, _NEW_BUFFERS); + FLUSH_CURRENT(ctx, _NEW_BUFFERS); /* The above doesn't fully flush the drivers in the way that a * glFlush does, but that is required here: */ diff --git a/src/mesa/main/ffvertex_prog.c b/src/mesa/main/ffvertex_prog.c index 03f42704a7d..1ce5685af45 100644 --- a/src/mesa/main/ffvertex_prog.c +++ b/src/mesa/main/ffvertex_prog.c @@ -570,9 +570,8 @@ static void emit_arg( struct prog_src_register *src, src->File = reg.file; src->Index = reg.idx; src->Swizzle = reg.swz; - src->NegateBase = reg.negate ? NEGATE_XYZW : 0; + src->Negate = reg.negate ? NEGATE_XYZW : NEGATE_NONE; src->Abs = 0; - src->NegateAbs = 0; src->RelAddr = 0; /* Check that bitfield sizes aren't exceeded */ ASSERT(src->Index == reg.idx); diff --git a/src/mesa/main/glheader.h b/src/mesa/main/glheader.h index ad095321e39..81d4ccf919f 100644 --- a/src/mesa/main/glheader.h +++ b/src/mesa/main/glheader.h @@ -74,6 +74,11 @@ #endif +#ifndef GL_PROGRAM_BINARY_LENGTH_OES +#define GL_PROGRAM_BINARY_LENGTH_OES 0x8741 +#endif + + /** * Special, internal token */ diff --git a/src/mesa/main/hash.c b/src/mesa/main/hash.c index 976f9d999b7..08c64568c83 100644 --- a/src/mesa/main/hash.c +++ b/src/mesa/main/hash.c @@ -63,6 +63,7 @@ struct _mesa_HashTable { struct HashEntry *Table[TABLE_SIZE]; /**< the lookup table */ GLuint MaxKey; /**< highest key inserted so far */ _glthread_Mutex Mutex; /**< mutual exclusion lock */ + _glthread_Mutex WalkMutex; /**< for _mesa_HashWalk() */ GLboolean InDeleteAll; /**< Debug check */ }; @@ -79,6 +80,7 @@ _mesa_NewHashTable(void) struct _mesa_HashTable *table = CALLOC_STRUCT(_mesa_HashTable); if (table) { _glthread_INIT_MUTEX(table->Mutex); + _glthread_INIT_MUTEX(table->WalkMutex); } return table; } @@ -111,6 +113,7 @@ _mesa_DeleteHashTable(struct _mesa_HashTable *table) } } _glthread_DESTROY_MUTEX(table->Mutex); + _glthread_DESTROY_MUTEX(table->WalkMutex); _mesa_free(table); } @@ -285,6 +288,11 @@ _mesa_HashDeleteAll(struct _mesa_HashTable *table, /** * Walk over all entries in a hash table, calling callback function for each. + * Note: we use a separate mutex in this function to avoid a recursive + * locking deadlock (in case the callback calls _mesa_HashRemove()) and to + * prevent multiple threads/contexts from getting tangled up. + * A lock-less version of this function could be used when the table will + * not be modified. * \param table the hash table to walk * \param callback the callback function * \param userData arbitrary pointer to pass along to the callback @@ -300,14 +308,16 @@ _mesa_HashWalk(const struct _mesa_HashTable *table, GLuint pos; ASSERT(table); ASSERT(callback); - _glthread_LOCK_MUTEX(table2->Mutex); + _glthread_LOCK_MUTEX(table2->WalkMutex); for (pos = 0; pos < TABLE_SIZE; pos++) { - struct HashEntry *entry; - for (entry = table->Table[pos]; entry; entry = entry->Next) { + struct HashEntry *entry, *next; + for (entry = table->Table[pos]; entry; entry = next) { + /* save 'next' pointer now in case the callback deletes the entry */ + next = entry->Next; callback(entry->Key, entry->Data, userData); } } - _glthread_UNLOCK_MUTEX(table2->Mutex); + _glthread_UNLOCK_MUTEX(table2->WalkMutex); } diff --git a/src/mesa/main/matrix.c b/src/mesa/main/matrix.c index 7388276c458..39b4967a58e 100644 --- a/src/mesa/main/matrix.c +++ b/src/mesa/main/matrix.c @@ -161,7 +161,7 @@ _mesa_MatrixMode( GLenum mode ) break; case GL_TEXTURE: if (ctx->Texture.CurrentUnit >= ctx->Const.MaxTextureCoordUnits) { - _mesa_error(ctx, GL_INVALID_OPERATION, "glMatrixMode(invalid unit %d)", + _mesa_error(ctx, GL_INVALID_OPERATION, "glMatrixMode(invalid tex unit %d)", ctx->Texture.CurrentUnit); return; } diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h index 3f315900000..30c7cca3b5a 100644 --- a/src/mesa/main/mtypes.h +++ b/src/mesa/main/mtypes.h @@ -2615,6 +2615,7 @@ struct gl_matrix_stack #define _NEW_TRACK_MATRIX 0x4000000 /**< __GLcontextRec::VertexProgram */ #define _NEW_PROGRAM 0x8000000 /**< __GLcontextRec::VertexProgram */ #define _NEW_CURRENT_ATTRIB 0x10000000 /**< __GLcontextRec::Current */ +#define _NEW_PROGRAM_CONSTANTS 0x20000000 #define _NEW_ALL ~0 /*@}*/ diff --git a/src/mesa/main/shaders.c b/src/mesa/main/shaders.c index be93b45a7da..bc76b912913 100644 --- a/src/mesa/main/shaders.c +++ b/src/mesa/main/shaders.c @@ -638,8 +638,7 @@ _mesa_UniformMatrix2fvARB(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value) { GET_CURRENT_CONTEXT(ctx); - ctx->Driver.UniformMatrix(ctx, 2, 2, GL_FLOAT_MAT2, - location, count, transpose, value); + ctx->Driver.UniformMatrix(ctx, 2, 2, location, count, transpose, value); } void GLAPIENTRY @@ -647,8 +646,7 @@ _mesa_UniformMatrix3fvARB(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value) { GET_CURRENT_CONTEXT(ctx); - ctx->Driver.UniformMatrix(ctx, 3, 3, GL_FLOAT_MAT3, - location, count, transpose, value); + ctx->Driver.UniformMatrix(ctx, 3, 3, location, count, transpose, value); } void GLAPIENTRY @@ -656,8 +654,7 @@ _mesa_UniformMatrix4fvARB(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value) { GET_CURRENT_CONTEXT(ctx); - ctx->Driver.UniformMatrix(ctx, 4, 4, GL_FLOAT_MAT4, - location, count, transpose, value); + ctx->Driver.UniformMatrix(ctx, 4, 4, location, count, transpose, value); } @@ -669,8 +666,7 @@ _mesa_UniformMatrix2x3fv(GLint location, GLsizei count, GLboolean transpose, const GLfloat *value) { GET_CURRENT_CONTEXT(ctx); - ctx->Driver.UniformMatrix(ctx, 2, 3, GL_FLOAT_MAT2x3, - location, count, transpose, value); + ctx->Driver.UniformMatrix(ctx, 2, 3, location, count, transpose, value); } void GLAPIENTRY @@ -678,8 +674,7 @@ _mesa_UniformMatrix3x2fv(GLint location, GLsizei count, GLboolean transpose, const GLfloat *value) { GET_CURRENT_CONTEXT(ctx); - ctx->Driver.UniformMatrix(ctx, 3, 2, GL_FLOAT_MAT3x2, - location, count, transpose, value); + ctx->Driver.UniformMatrix(ctx, 3, 2, location, count, transpose, value); } void GLAPIENTRY @@ -687,8 +682,7 @@ _mesa_UniformMatrix2x4fv(GLint location, GLsizei count, GLboolean transpose, const GLfloat *value) { GET_CURRENT_CONTEXT(ctx); - ctx->Driver.UniformMatrix(ctx, 2, 4, GL_FLOAT_MAT2x4, - location, count, transpose, value); + ctx->Driver.UniformMatrix(ctx, 2, 4, location, count, transpose, value); } void GLAPIENTRY @@ -696,8 +690,7 @@ _mesa_UniformMatrix4x2fv(GLint location, GLsizei count, GLboolean transpose, const GLfloat *value) { GET_CURRENT_CONTEXT(ctx); - ctx->Driver.UniformMatrix(ctx, 4, 2, GL_FLOAT_MAT4x2, - location, count, transpose, value); + ctx->Driver.UniformMatrix(ctx, 4, 2, location, count, transpose, value); } void GLAPIENTRY @@ -705,8 +698,7 @@ _mesa_UniformMatrix3x4fv(GLint location, GLsizei count, GLboolean transpose, const GLfloat *value) { GET_CURRENT_CONTEXT(ctx); - ctx->Driver.UniformMatrix(ctx, 3, 4, GL_FLOAT_MAT3x4, - location, count, transpose, value); + ctx->Driver.UniformMatrix(ctx, 3, 4, location, count, transpose, value); } void GLAPIENTRY @@ -714,8 +706,7 @@ _mesa_UniformMatrix4x3fv(GLint location, GLsizei count, GLboolean transpose, const GLfloat *value) { GET_CURRENT_CONTEXT(ctx); - ctx->Driver.UniformMatrix(ctx, 4, 3, GL_FLOAT_MAT4x3, - location, count, transpose, value); + ctx->Driver.UniformMatrix(ctx, 4, 3, location, count, transpose, value); } diff --git a/src/mesa/main/state.c b/src/mesa/main/state.c index cc37d636369..f18fc8f6837 100644 --- a/src/mesa/main/state.c +++ b/src/mesa/main/state.c @@ -44,6 +44,7 @@ #include "pixel.h" #endif #include "shader/program.h" +#include "shader/prog_parameter.h" #include "state.h" #include "stencil.h" #include "texenvprogram.h" @@ -303,6 +304,36 @@ update_program(GLcontext *ctx) } +/** + * Examine shader constants and return either _NEW_PROGRAM_CONSTANTS or 0. + */ +static GLbitfield +update_program_constants(GLcontext *ctx) +{ + GLbitfield new_state = 0x0; + + if (ctx->FragmentProgram._Current) { + const struct gl_program_parameter_list *params = + ctx->FragmentProgram._Current->Base.Parameters; + if (params && params->StateFlags & ctx->NewState) { + new_state |= _NEW_PROGRAM_CONSTANTS; + } + } + + if (ctx->VertexProgram._Current) { + const struct gl_program_parameter_list *params = + ctx->VertexProgram._Current->Base.Parameters; + if (params && params->StateFlags & ctx->NewState) { + new_state |= _NEW_PROGRAM_CONSTANTS; + } + } + + return new_state; +} + + + + static void update_viewport_matrix(GLcontext *ctx) { @@ -470,7 +501,8 @@ _mesa_update_state_locked( GLcontext *ctx ) /* Determine which state flags effect vertex/fragment program state */ if (ctx->FragmentProgram._MaintainTexEnvProgram) { - prog_flags |= (_NEW_TEXTURE | _NEW_FOG | _DD_NEW_SEPARATE_SPECULAR); + prog_flags |= (_NEW_TEXTURE | _NEW_FOG | _DD_NEW_SEPARATE_SPECULAR | + _NEW_ARRAY); } if (ctx->VertexProgram._MaintainTnlProgram) { prog_flags |= (_NEW_ARRAY | _NEW_TEXTURE | _NEW_TEXTURE_MATRIX | @@ -553,6 +585,10 @@ _mesa_update_state_locked( GLcontext *ctx ) new_prog_state |= update_program( ctx ); } + + out: + new_prog_state |= update_program_constants(ctx); + /* * Give the driver a chance to act upon the new_state flags. * The driver might plug in different span functions, for example. @@ -562,7 +598,6 @@ _mesa_update_state_locked( GLcontext *ctx ) * Set ctx->NewState to zero to avoid recursion if * Driver.UpdateState() has to call FLUSH_VERTICES(). (fixed?) */ - out: new_state = ctx->NewState | new_prog_state; ctx->NewState = 0; ctx->Driver.UpdateState(ctx, new_state); diff --git a/src/mesa/main/texcompress_s3tc.c b/src/mesa/main/texcompress_s3tc.c index d17e18da6b5..a1c0f18f36f 100644 --- a/src/mesa/main/texcompress_s3tc.c +++ b/src/mesa/main/texcompress_s3tc.c @@ -205,7 +205,7 @@ texstore_rgb_dxt1(TEXSTORE_PARAMS) dst, dstRowStride); } else { - _mesa_warning(ctx, "external dxt library not available"); + _mesa_warning(ctx, "external dxt library not available: texstore_rgb_dxt1"); } if (tempImage) @@ -267,7 +267,7 @@ texstore_rgba_dxt1(TEXSTORE_PARAMS) dst, dstRowStride); } else { - _mesa_warning(ctx, "external dxt library not available"); + _mesa_warning(ctx, "external dxt library not available: texstore_rgba_dxt1"); } if (tempImage) @@ -328,7 +328,7 @@ texstore_rgba_dxt3(TEXSTORE_PARAMS) dst, dstRowStride); } else { - _mesa_warning(ctx, "external dxt library not available"); + _mesa_warning(ctx, "external dxt library not available: texstore_rgba_dxt3"); } if (tempImage) @@ -389,7 +389,7 @@ texstore_rgba_dxt5(TEXSTORE_PARAMS) dst, dstRowStride); } else { - _mesa_warning(ctx, "external dxt library not available"); + _mesa_warning(ctx, "external dxt library not available: texstore_rgba_dxt5"); } if (tempImage) @@ -410,7 +410,7 @@ fetch_texel_2d_rgb_dxt1( const struct gl_texture_image *texImage, (GLubyte *)(texImage)->Data, i, j, texel); } else - _mesa_debug(NULL, "attempted to decode s3tc texture without library available\n"); + _mesa_debug(NULL, "attempted to decode s3tc texture without library available: fetch_texel_2d_rgb_dxt1"); } @@ -438,7 +438,7 @@ fetch_texel_2d_rgba_dxt1( const struct gl_texture_image *texImage, (GLubyte *)(texImage)->Data, i, j, texel); } else - _mesa_debug(NULL, "attempted to decode s3tc texture without library available\n"); + _mesa_debug(NULL, "attempted to decode s3tc texture without library available: fetch_texel_2d_rgba_dxt1\n"); } @@ -467,7 +467,7 @@ fetch_texel_2d_rgba_dxt3( const struct gl_texture_image *texImage, i, j, texel); } else - _mesa_debug(NULL, "attempted to decode s3tc texture without library available\n"); + _mesa_debug(NULL, "attempted to decode s3tc texture without library available: fetch_texel_2d_rgba_dxt3\n"); } @@ -495,7 +495,7 @@ fetch_texel_2d_rgba_dxt5( const struct gl_texture_image *texImage, i, j, texel); } else - _mesa_debug(NULL, "attempted to decode s3tc texture without library available\n"); + _mesa_debug(NULL, "attempted to decode s3tc texture without library available: fetch_texel_2d_rgba_dxt5\n"); } diff --git a/src/mesa/main/texenvprogram.c b/src/mesa/main/texenvprogram.c index 4a124bf27e2..a70d069bd9c 100644 --- a/src/mesa/main/texenvprogram.c +++ b/src/mesa/main/texenvprogram.c @@ -663,9 +663,8 @@ static void emit_arg( struct prog_src_register *reg, reg->File = ureg.file; reg->Index = ureg.idx; reg->Swizzle = ureg.swz; - reg->NegateBase = ureg.negatebase ? 0xf : 0x0; + reg->Negate = ureg.negatebase ? NEGATE_XYZW : NEGATE_NONE; reg->Abs = ureg.abs; - reg->NegateAbs = ureg.negateabs; } static void emit_dst( struct prog_dst_register *dst, diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c index 35253daa2e6..b47bf360cf5 100644 --- a/src/mesa/shader/arbprogparse.c +++ b/src/mesa/shader/arbprogparse.c @@ -2669,7 +2669,7 @@ parse_vector_src_reg(GLcontext *ctx, const GLubyte **inst, reg->File = file; reg->Index = index; reg->Swizzle = MAKE_SWIZZLE4(swizzle[0], swizzle[1], swizzle[2], swizzle[3]); - reg->NegateBase = negateMask; + reg->Negate = negateMask; reg->RelAddr = isRelOffset; return 0; } @@ -2703,7 +2703,7 @@ parse_scalar_src_reg(GLcontext *ctx, const GLubyte **inst, reg->File = file; reg->Index = index; reg->Swizzle = (swizzle[0] << 0); - reg->NegateBase = negateMask; + reg->Negate = negateMask; reg->RelAddr = isRelOffset; return 0; } @@ -3019,7 +3019,7 @@ parse_fp_instruction (GLcontext * ctx, const GLubyte ** inst, parse_extended_swizzle_mask(inst, swizzle, &negateMask); fp->SrcReg[0].File = file; fp->SrcReg[0].Index = index; - fp->SrcReg[0].NegateBase = negateMask; + fp->SrcReg[0].Negate = negateMask; fp->SrcReg[0].Swizzle = MAKE_SWIZZLE4(swizzle[0], swizzle[1], swizzle[2], @@ -3363,7 +3363,7 @@ parse_vp_instruction (GLcontext * ctx, const GLubyte ** inst, parse_extended_swizzle_mask (inst, swizzle, &negateMask); vp->SrcReg[0].File = file; vp->SrcReg[0].Index = index; - vp->SrcReg[0].NegateBase = negateMask; + vp->SrcReg[0].Negate = negateMask; vp->SrcReg[0].Swizzle = MAKE_SWIZZLE4(swizzle[0], swizzle[1], swizzle[2], diff --git a/src/mesa/shader/arbprogram.c b/src/mesa/shader/arbprogram.c index 329c0ea0b03..981565ab8f1 100644 --- a/src/mesa/shader/arbprogram.c +++ b/src/mesa/shader/arbprogram.c @@ -489,7 +489,7 @@ _mesa_ProgramEnvParameter4fARB(GLenum target, GLuint index, GET_CURRENT_CONTEXT(ctx); ASSERT_OUTSIDE_BEGIN_END(ctx); - FLUSH_VERTICES(ctx, _NEW_PROGRAM); + FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS); if (target == GL_FRAGMENT_PROGRAM_ARB && ctx->Extensions.ARB_fragment_program) { @@ -537,7 +537,7 @@ _mesa_ProgramEnvParameters4fvEXT(GLenum target, GLuint index, GLsizei count, GLfloat * dest; ASSERT_OUTSIDE_BEGIN_END(ctx); - FLUSH_VERTICES(ctx, _NEW_PROGRAM); + FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS); if (count <= 0) { _mesa_error(ctx, GL_INVALID_VALUE, "glProgramEnvParameters4fv(count)"); @@ -595,8 +595,6 @@ _mesa_GetProgramEnvParameterfvARB(GLenum target, GLuint index, { GET_CURRENT_CONTEXT(ctx); - FLUSH_VERTICES(ctx, _NEW_PROGRAM); - ASSERT_OUTSIDE_BEGIN_END(ctx); if (target == GL_FRAGMENT_PROGRAM_ARB @@ -633,7 +631,7 @@ _mesa_ProgramLocalParameter4fARB(GLenum target, GLuint index, struct gl_program *prog; ASSERT_OUTSIDE_BEGIN_END(ctx); - FLUSH_VERTICES(ctx, _NEW_PROGRAM); + FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS); if ((target == GL_FRAGMENT_PROGRAM_NV && ctx->Extensions.NV_fragment_program) || @@ -687,7 +685,7 @@ _mesa_ProgramLocalParameters4fvEXT(GLenum target, GLuint index, GLsizei count, GLint i; ASSERT_OUTSIDE_BEGIN_END(ctx); - FLUSH_VERTICES(ctx, _NEW_PROGRAM); + FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS); if (count <= 0) { _mesa_error(ctx, GL_INVALID_VALUE, "glProgramLocalParameters4fv(count)"); diff --git a/src/mesa/shader/nvfragparse.c b/src/mesa/shader/nvfragparse.c index b935cb562a6..0fd55524abf 100644 --- a/src/mesa/shader/nvfragparse.c +++ b/src/mesa/shader/nvfragparse.c @@ -43,6 +43,7 @@ #include "main/macros.h" #include "program.h" #include "prog_parameter.h" +#include "prog_print.h" #include "prog_instruction.h" #include "nvfragparse.h" @@ -385,10 +386,6 @@ static const char *InputRegisters[MAX_NV_FRAGMENT_PROGRAM_INPUTS + 1] = { }; -static const char *OutputRegisters[MAX_NV_FRAGMENT_PROGRAM_OUTPUTS + 1] = { - "DEPR", "COLR", "DATA0", NULL -}; - /**********************************************************************/ @@ -960,6 +957,7 @@ Parse_VectorSrc(struct parse_state *parseState, GLfloat sign = 1.0F; GLubyte token[100]; GLint idx; + GLuint negateBase, negateAbs; /* * First, take care of +/- and absolute value stuff. @@ -971,21 +969,23 @@ Parse_VectorSrc(struct parse_state *parseState, if (Parse_String(parseState, "|")) { srcReg->Abs = GL_TRUE; - srcReg->NegateAbs = (sign < 0.0F) ? GL_TRUE : GL_FALSE; + negateAbs = (sign < 0.0F) ? NEGATE_XYZW : NEGATE_NONE; if (Parse_String(parseState, "-")) - srcReg->NegateBase = NEGATE_XYZW; + negateBase = NEGATE_XYZW; else if (Parse_String(parseState, "+")) - srcReg->NegateBase = NEGATE_NONE; + negateBase = NEGATE_NONE; else - srcReg->NegateBase = NEGATE_NONE; + negateBase = NEGATE_NONE; } else { srcReg->Abs = GL_FALSE; - srcReg->NegateAbs = GL_FALSE; - srcReg->NegateBase = (sign < 0.0F) ? NEGATE_XYZW : NEGATE_NONE; + negateAbs = NEGATE_NONE; + negateBase = (sign < 0.0F) ? NEGATE_XYZW : NEGATE_NONE; } + srcReg->Negate = srcReg->Abs ? negateAbs : negateBase; + /* This should be the real src vector/register name */ if (!Peek_Token(parseState, token)) RETURN_ERROR; @@ -1086,6 +1086,7 @@ Parse_ScalarSrcReg(struct parse_state *parseState, GLfloat sign = 1.0F; GLboolean needSuffix = GL_TRUE; GLint idx; + GLuint negateBase, negateAbs; /* * First, take care of +/- and absolute value stuff. @@ -1097,21 +1098,23 @@ Parse_ScalarSrcReg(struct parse_state *parseState, if (Parse_String(parseState, "|")) { srcReg->Abs = GL_TRUE; - srcReg->NegateAbs = (sign < 0.0F) ? GL_TRUE : GL_FALSE; + negateAbs = (sign < 0.0F) ? NEGATE_XYZW : NEGATE_NONE; if (Parse_String(parseState, "-")) - srcReg->NegateBase = NEGATE_XYZW; + negateBase = NEGATE_XYZW; else if (Parse_String(parseState, "+")) - srcReg->NegateBase = NEGATE_NONE; + negateBase = NEGATE_NONE; else - srcReg->NegateBase = NEGATE_NONE; + negateBase = NEGATE_NONE; } else { srcReg->Abs = GL_FALSE; - srcReg->NegateAbs = GL_FALSE; - srcReg->NegateBase = (sign < 0.0F) ? NEGATE_XYZW : NEGATE_NONE; + negateAbs = NEGATE_NONE; + negateBase = (sign < 0.0F) ? NEGATE_XYZW : NEGATE_NONE; } + srcReg->Negate = srcReg->Abs ? negateAbs : negateBase; + if (!Peek_Token(parseState, token)) RETURN_ERROR; @@ -1250,9 +1253,8 @@ Parse_PrintInstruction(struct parse_state *parseState, } inst->SrcReg[0].Swizzle = SWIZZLE_NOOP; - inst->SrcReg[0].NegateBase = NEGATE_NONE; inst->SrcReg[0].Abs = GL_FALSE; - inst->SrcReg[0].NegateAbs = GL_FALSE; + inst->SrcReg[0].Negate = NEGATE_NONE; return GL_TRUE; } @@ -1559,7 +1561,7 @@ _mesa_parse_nv_fragment_program(GLcontext *ctx, GLenum dstTarget, #ifdef DEBUG_foo _mesa_printf("--- glLoadProgramNV(%d) result ---\n", program->Base.Id); - _mesa_print_nv_fragment_program(program); + _mesa_fprint_program_opt(stdout, &program->Base, PROG_PRINT_NV, 0); _mesa_printf("----------------------------------\n"); #endif } @@ -1571,243 +1573,6 @@ _mesa_parse_nv_fragment_program(GLcontext *ctx, GLenum dstTarget, } -static void -PrintSrcReg(const struct gl_fragment_program *program, - const struct prog_src_register *src) -{ - static const char comps[5] = "xyzw"; - - if (src->NegateAbs) { - _mesa_printf("-"); - } - if (src->Abs) { - _mesa_printf("|"); - } - if (src->NegateBase) { - _mesa_printf("-"); - } - if (src->File == PROGRAM_NAMED_PARAM) { - if (program->Base.Parameters->Parameters[src->Index].Type - == PROGRAM_CONSTANT) { - const GLfloat *v; - v = program->Base.Parameters->ParameterValues[src->Index]; - _mesa_printf("{%g, %g, %g, %g}", v[0], v[1], v[2], v[3]); - } - else { - ASSERT(program->Base.Parameters->Parameters[src->Index].Type - == PROGRAM_NAMED_PARAM); - _mesa_printf("%s", program->Base.Parameters->Parameters[src->Index].Name); - } - } - else if (src->File == PROGRAM_OUTPUT) { - _mesa_printf("o[%s]", OutputRegisters[src->Index]); - } - else if (src->File == PROGRAM_INPUT) { - _mesa_printf("f[%s]", InputRegisters[src->Index]); - } - else if (src->File == PROGRAM_LOCAL_PARAM) { - _mesa_printf("p[%d]", src->Index); - } - else if (src->File == PROGRAM_TEMPORARY) { - if (src->Index >= 32) - _mesa_printf("H%d", src->Index); - else - _mesa_printf("R%d", src->Index); - } - else if (src->File == PROGRAM_WRITE_ONLY) { - _mesa_printf("%cC", "HR"[src->Index]); - } - else { - _mesa_problem(NULL, "Invalid fragment register %d", src->Index); - return; - } - if (GET_SWZ(src->Swizzle, 0) == GET_SWZ(src->Swizzle, 1) && - GET_SWZ(src->Swizzle, 0) == GET_SWZ(src->Swizzle, 2) && - GET_SWZ(src->Swizzle, 0) == GET_SWZ(src->Swizzle, 3)) { - _mesa_printf(".%c", comps[GET_SWZ(src->Swizzle, 0)]); - } - else if (src->Swizzle != SWIZZLE_NOOP) { - _mesa_printf(".%c%c%c%c", - comps[GET_SWZ(src->Swizzle, 0)], - comps[GET_SWZ(src->Swizzle, 1)], - comps[GET_SWZ(src->Swizzle, 2)], - comps[GET_SWZ(src->Swizzle, 3)]); - } - if (src->Abs) { - _mesa_printf("|"); - } -} - -static void -PrintTextureSrc(const struct prog_instruction *inst) -{ - _mesa_printf("TEX%d, ", inst->TexSrcUnit); - switch (inst->TexSrcTarget) { - case TEXTURE_1D_INDEX: - _mesa_printf("1D"); - break; - case TEXTURE_2D_INDEX: - _mesa_printf("2D"); - break; - case TEXTURE_3D_INDEX: - _mesa_printf("3D"); - break; - case TEXTURE_RECT_INDEX: - _mesa_printf("RECT"); - break; - case TEXTURE_CUBE_INDEX: - _mesa_printf("CUBE"); - break; - default: - _mesa_problem(NULL, "Invalid textue target in PrintTextureSrc"); - } -} - -static void -PrintCondCode(const struct prog_dst_register *dst) -{ - static const char *comps = "xyzw"; - static const char *ccString[] = { - "??", "GT", "EQ", "LT", "UN", "GE", "LE", "NE", "TR", "FL", "??" - }; - - _mesa_printf("%s", ccString[dst->CondMask]); - if (GET_SWZ(dst->CondSwizzle, 0) == GET_SWZ(dst->CondSwizzle, 1) && - GET_SWZ(dst->CondSwizzle, 0) == GET_SWZ(dst->CondSwizzle, 2) && - GET_SWZ(dst->CondSwizzle, 0) == GET_SWZ(dst->CondSwizzle, 3)) { - _mesa_printf(".%c", comps[GET_SWZ(dst->CondSwizzle, 0)]); - } - else if (dst->CondSwizzle != SWIZZLE_NOOP) { - _mesa_printf(".%c%c%c%c", - comps[GET_SWZ(dst->CondSwizzle, 0)], - comps[GET_SWZ(dst->CondSwizzle, 1)], - comps[GET_SWZ(dst->CondSwizzle, 2)], - comps[GET_SWZ(dst->CondSwizzle, 3)]); - } -} - - -static void -PrintDstReg(const struct prog_dst_register *dst) -{ - if (dst->File == PROGRAM_OUTPUT) { - _mesa_printf("o[%s]", OutputRegisters[dst->Index]); - } - else if (dst->File == PROGRAM_TEMPORARY) { - if (dst->Index >= 32) - _mesa_printf("H%d", dst->Index); - else - _mesa_printf("R%d", dst->Index); - } - else if (dst->File == PROGRAM_LOCAL_PARAM) { - _mesa_printf("p[%d]", dst->Index); - } - else if (dst->File == PROGRAM_WRITE_ONLY) { - _mesa_printf("%cC", "HR"[dst->Index]); - } - else { - _mesa_printf("???"); - } - - if (dst->WriteMask != 0 && dst->WriteMask != WRITEMASK_XYZW) { - _mesa_printf("."); - if (dst->WriteMask & WRITEMASK_X) - _mesa_printf("x"); - if (dst->WriteMask & WRITEMASK_Y) - _mesa_printf("y"); - if (dst->WriteMask & WRITEMASK_Z) - _mesa_printf("z"); - if (dst->WriteMask & WRITEMASK_W) - _mesa_printf("w"); - } - - if (dst->CondMask != COND_TR || - dst->CondSwizzle != SWIZZLE_NOOP) { - _mesa_printf(" ("); - PrintCondCode(dst); - _mesa_printf(")"); - } -} - - -/** - * Print (unparse) the given vertex program. Just for debugging. - */ -void -_mesa_print_nv_fragment_program(const struct gl_fragment_program *program) -{ - const struct prog_instruction *inst; - - for (inst = program->Base.Instructions; inst->Opcode != OPCODE_END; inst++) { - int i; - for (i = 0; Instructions[i].name; i++) { - if (inst->Opcode == Instructions[i].opcode) { - /* print instruction name */ - _mesa_printf("%s", Instructions[i].name); - if (inst->Precision == FLOAT16) - _mesa_printf("H"); - else if (inst->Precision == FIXED12) - _mesa_printf("X"); - if (inst->CondUpdate) - _mesa_printf("C"); - if (inst->SaturateMode == SATURATE_ZERO_ONE) - _mesa_printf("_SAT"); - _mesa_printf(" "); - - if (Instructions[i].inputs == INPUT_CC) { - PrintCondCode(&inst->DstReg); - } - else if (Instructions[i].outputs == OUTPUT_V || - Instructions[i].outputs == OUTPUT_S) { - /* print dest register */ - PrintDstReg(&inst->DstReg); - _mesa_printf(", "); - } - - /* print source register(s) */ - if (Instructions[i].inputs == INPUT_1V || - Instructions[i].inputs == INPUT_1S) { - PrintSrcReg(program, &inst->SrcReg[0]); - } - else if (Instructions[i].inputs == INPUT_2V || - Instructions[i].inputs == INPUT_2S) { - PrintSrcReg(program, &inst->SrcReg[0]); - _mesa_printf(", "); - PrintSrcReg(program, &inst->SrcReg[1]); - } - else if (Instructions[i].inputs == INPUT_3V) { - PrintSrcReg(program, &inst->SrcReg[0]); - _mesa_printf(", "); - PrintSrcReg(program, &inst->SrcReg[1]); - _mesa_printf(", "); - PrintSrcReg(program, &inst->SrcReg[2]); - } - else if (Instructions[i].inputs == INPUT_1V_T) { - PrintSrcReg(program, &inst->SrcReg[0]); - _mesa_printf(", "); - PrintTextureSrc(inst); - } - else if (Instructions[i].inputs == INPUT_3V_T) { - PrintSrcReg(program, &inst->SrcReg[0]); - _mesa_printf(", "); - PrintSrcReg(program, &inst->SrcReg[1]); - _mesa_printf(", "); - PrintSrcReg(program, &inst->SrcReg[2]); - _mesa_printf(", "); - PrintTextureSrc(inst); - } - _mesa_printf(";\n"); - break; - } - } - if (!Instructions[i].name) { - _mesa_printf("Invalid opcode %d\n", inst->Opcode); - } - } - _mesa_printf("END\n"); -} - - const char * _mesa_nv_fragment_input_register_name(GLuint i) { diff --git a/src/mesa/shader/nvfragparse.h b/src/mesa/shader/nvfragparse.h index ac97921080c..544ab80c56c 100644 --- a/src/mesa/shader/nvfragparse.h +++ b/src/mesa/shader/nvfragparse.h @@ -37,10 +37,6 @@ _mesa_parse_nv_fragment_program(GLcontext *ctx, GLenum target, struct gl_fragment_program *program); -extern void -_mesa_print_nv_fragment_program(const struct gl_fragment_program *program); - - extern const char * _mesa_nv_fragment_input_register_name(GLuint i); diff --git a/src/mesa/shader/nvprogram.c b/src/mesa/shader/nvprogram.c index 50358cf1076..5142c2a4a59 100644 --- a/src/mesa/shader/nvprogram.c +++ b/src/mesa/shader/nvprogram.c @@ -706,7 +706,7 @@ _mesa_ProgramNamedParameter4fNV(GLuint id, GLsizei len, const GLubyte *name, GET_CURRENT_CONTEXT(ctx); ASSERT_OUTSIDE_BEGIN_END(ctx); - FLUSH_VERTICES(ctx, _NEW_PROGRAM); + FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS); prog = _mesa_lookup_program(ctx, id); if (!prog || prog->Target != GL_FRAGMENT_PROGRAM_NV) { diff --git a/src/mesa/shader/nvvertparse.c b/src/mesa/shader/nvvertparse.c index 268b577aec9..f5e2df26708 100644 --- a/src/mesa/shader/nvvertparse.c +++ b/src/mesa/shader/nvvertparse.c @@ -44,6 +44,7 @@ #include "nvprogram.h" #include "nvvertparse.h" #include "prog_instruction.h" +#include "prog_print.h" #include "program.h" @@ -640,12 +641,12 @@ Parse_SwizzleSrcReg(struct parse_state *parseState, struct prog_src_register *sr RETURN_ERROR; if (token[0] == '-') { (void) Parse_String(parseState, "-"); - srcReg->NegateBase = NEGATE_XYZW; + srcReg->Negate = NEGATE_XYZW; if (!Peek_Token(parseState, token)) RETURN_ERROR; } else { - srcReg->NegateBase = NEGATE_NONE; + srcReg->Negate = NEGATE_NONE; } /* Src reg can be R<n>, c[n], c[n +/- offset], or a named vertex attrib */ @@ -733,13 +734,13 @@ Parse_ScalarSrcReg(struct parse_state *parseState, struct prog_src_register *src if (!Peek_Token(parseState, token)) RETURN_ERROR; if (token[0] == '-') { - srcReg->NegateBase = NEGATE_XYZW; + srcReg->Negate = NEGATE_XYZW; (void) Parse_String(parseState, "-"); /* consume '-' */ if (!Peek_Token(parseState, token)) RETURN_ERROR; } else { - srcReg->NegateBase = NEGATE_NONE; + srcReg->Negate = NEGATE_NONE; } /* Src reg can be R<n>, c[n], c[n +/- offset], or a named vertex attrib */ @@ -1061,7 +1062,7 @@ Parse_PrintInstruction(struct parse_state *parseState, struct prog_instruction * RETURN_ERROR; srcReg->RelAddr = GL_FALSE; - srcReg->NegateBase = NEGATE_NONE; + srcReg->Negate = NEGATE_NONE; srcReg->Swizzle = SWIZZLE_NOOP; /* Register can be R<n>, c[n], c[n +/- offset], a named vertex attrib, @@ -1394,7 +1395,7 @@ _mesa_parse_nv_vertex_program(GLcontext *ctx, GLenum dstTarget, #ifdef DEBUG_foo _mesa_printf("--- glLoadProgramNV result ---\n"); - _mesa_print_nv_vertex_program(program); + _mesa_fprint_program_opt(stdout, &program->Base, PROG_PRINT_NV, 0); _mesa_printf("------------------------------\n"); #endif } @@ -1410,161 +1411,6 @@ _mesa_parse_nv_vertex_program(GLcontext *ctx, GLenum dstTarget, } -static void -PrintSrcReg(const struct prog_src_register *src) -{ - static const char comps[5] = "xyzw"; - if (src->NegateBase) - _mesa_printf("-"); - if (src->RelAddr) { - if (src->Index > 0) - _mesa_printf("c[A0.x + %d]", src->Index); - else if (src->Index < 0) - _mesa_printf("c[A0.x - %d]", -src->Index); - else - _mesa_printf("c[A0.x]"); - } - else if (src->File == PROGRAM_OUTPUT) { - _mesa_printf("o[%s]", OutputRegisters[src->Index]); - } - else if (src->File == PROGRAM_INPUT) { - _mesa_printf("v[%s]", InputRegisters[src->Index]); - } - else if (src->File == PROGRAM_ENV_PARAM) { - _mesa_printf("c[%d]", src->Index); - } - else { - ASSERT(src->File == PROGRAM_TEMPORARY); - _mesa_printf("R%d", src->Index); - } - - if (GET_SWZ(src->Swizzle, 0) == GET_SWZ(src->Swizzle, 1) && - GET_SWZ(src->Swizzle, 0) == GET_SWZ(src->Swizzle, 2) && - GET_SWZ(src->Swizzle, 0) == GET_SWZ(src->Swizzle, 3)) { - _mesa_printf(".%c", comps[GET_SWZ(src->Swizzle, 0)]); - } - else if (src->Swizzle != SWIZZLE_NOOP) { - _mesa_printf(".%c%c%c%c", - comps[GET_SWZ(src->Swizzle, 0)], - comps[GET_SWZ(src->Swizzle, 1)], - comps[GET_SWZ(src->Swizzle, 2)], - comps[GET_SWZ(src->Swizzle, 3)]); - } -} - - -static void -PrintDstReg(const struct prog_dst_register *dst) -{ - if (dst->File == PROGRAM_OUTPUT) { - _mesa_printf("o[%s]", OutputRegisters[dst->Index]); - } - else if (dst->File == PROGRAM_INPUT) { - _mesa_printf("v[%s]", InputRegisters[dst->Index]); - } - else if (dst->File == PROGRAM_ENV_PARAM) { - _mesa_printf("c[%d]", dst->Index); - } - else { - ASSERT(dst->File == PROGRAM_TEMPORARY); - _mesa_printf("R%d", dst->Index); - } - - if (dst->WriteMask != 0 && dst->WriteMask != WRITEMASK_XYZW) { - _mesa_printf("."); - if (dst->WriteMask & WRITEMASK_X) - _mesa_printf("x"); - if (dst->WriteMask & WRITEMASK_Y) - _mesa_printf("y"); - if (dst->WriteMask & WRITEMASK_Z) - _mesa_printf("z"); - if (dst->WriteMask & WRITEMASK_W) - _mesa_printf("w"); - } -} - - -/** - * Print a single NVIDIA vertex program instruction. - */ -void -_mesa_print_nv_vertex_instruction(const struct prog_instruction *inst) -{ - GLuint i, n; - - switch (inst->Opcode) { - case OPCODE_MOV: - case OPCODE_LIT: - case OPCODE_RCP: - case OPCODE_RSQ: - case OPCODE_EXP: - case OPCODE_LOG: - case OPCODE_RCC: - case OPCODE_ABS: - case OPCODE_MUL: - case OPCODE_ADD: - case OPCODE_DP3: - case OPCODE_DP4: - case OPCODE_DST: - case OPCODE_MIN: - case OPCODE_MAX: - case OPCODE_SLT: - case OPCODE_SGE: - case OPCODE_DPH: - case OPCODE_SUB: - case OPCODE_MAD: - _mesa_printf("%s ", _mesa_opcode_string(inst->Opcode)); - PrintDstReg(&inst->DstReg); - _mesa_printf(", "); - n = _mesa_num_inst_src_regs(inst->Opcode); - for (i = 0; i < n; i++) { - PrintSrcReg(&inst->SrcReg[i]); - if (i + 1 < n) - _mesa_printf(", "); - } - _mesa_printf(";\n"); - break; - case OPCODE_ARL: - _mesa_printf("ARL A0.x, "); - PrintSrcReg(&inst->SrcReg[0]); - _mesa_printf(";\n"); - break; - case OPCODE_PRINT: - _mesa_printf("PRINT '%s'", inst->Data); - if (inst->SrcReg[0].File != PROGRAM_UNDEFINED) { - _mesa_printf(", "); - PrintSrcReg(&inst->SrcReg[0]); - _mesa_printf(";\n"); - } - else { - _mesa_printf("\n"); - } - break; - case OPCODE_END: - _mesa_printf("END\n"); - break; - default: - _mesa_printf("BAD INSTRUCTION\n"); - } -} - - -/** - * Print (unparse) the given vertex program. Just for debugging. - */ -void -_mesa_print_nv_vertex_program(const struct gl_vertex_program *program) -{ - const struct prog_instruction *inst; - - for (inst = program->Base.Instructions; ; inst++) { - _mesa_print_nv_vertex_instruction(inst); - if (inst->Opcode == OPCODE_END) - return; - } -} - - const char * _mesa_nv_vertex_input_register_name(GLuint i) { diff --git a/src/mesa/shader/nvvertparse.h b/src/mesa/shader/nvvertparse.h index 15fb03cd4e4..9919e22388d 100644 --- a/src/mesa/shader/nvvertparse.h +++ b/src/mesa/shader/nvvertparse.h @@ -35,11 +35,6 @@ _mesa_parse_nv_vertex_program(GLcontext *ctx, GLenum target, const GLubyte *str, GLsizei len, struct gl_vertex_program *program); -extern void -_mesa_print_nv_vertex_instruction(const struct prog_instruction *inst); - -extern void -_mesa_print_nv_vertex_program(const struct gl_vertex_program *program); extern const char * _mesa_nv_vertex_input_register_name(GLuint i); diff --git a/src/mesa/shader/prog_execute.c b/src/mesa/shader/prog_execute.c index bdac1d4f8a6..68a59350a13 100644 --- a/src/mesa/shader/prog_execute.c +++ b/src/mesa/shader/prog_execute.c @@ -212,19 +212,14 @@ fetch_vector4(const struct prog_src_register *source, result[3] = src[GET_SWZ(source->Swizzle, 3)]; } - if (source->NegateBase) { - result[0] = -result[0]; - result[1] = -result[1]; - result[2] = -result[2]; - result[3] = -result[3]; - } if (source->Abs) { result[0] = FABSF(result[0]); result[1] = FABSF(result[1]); result[2] = FABSF(result[2]); result[3] = FABSF(result[3]); } - if (source->NegateAbs) { + if (source->Negate) { + ASSERT(source->Negate == NEGATE_XYZW); result[0] = -result[0]; result[1] = -result[1]; result[2] = -result[2]; @@ -259,7 +254,7 @@ fetch_vector4ui(const struct prog_src_register *source, result[3] = src[GET_SWZ(source->Swizzle, 3)]; } - /* Note: no NegateBase, Abs, NegateAbs here */ + /* Note: no Negate or Abs here */ } @@ -299,19 +294,14 @@ fetch_vector4_deriv(GLcontext * ctx, result[2] = deriv[GET_SWZ(source->Swizzle, 2)]; result[3] = deriv[GET_SWZ(source->Swizzle, 3)]; - if (source->NegateBase) { - result[0] = -result[0]; - result[1] = -result[1]; - result[2] = -result[2]; - result[3] = -result[3]; - } if (source->Abs) { result[0] = FABSF(result[0]); result[1] = FABSF(result[1]); result[2] = FABSF(result[2]); result[3] = FABSF(result[3]); } - if (source->NegateAbs) { + if (source->Negate) { + ASSERT(source->Negate == NEGATE_XYZW); result[0] = -result[0]; result[1] = -result[1]; result[2] = -result[2]; @@ -336,13 +326,10 @@ fetch_vector1(const struct prog_src_register *source, result[0] = src[GET_SWZ(source->Swizzle, 0)]; - if (source->NegateBase) { - result[0] = -result[0]; - } if (source->Abs) { result[0] = FABSF(result[0]); } - if (source->NegateAbs) { + if (source->Negate) { result[0] = -result[0]; } } @@ -1514,7 +1501,7 @@ _mesa_execute_program(GLcontext * ctx, ASSERT(swz <= 3); result[i] = src[swz]; } - if (source->NegateBase & (1 << i)) + if (source->Negate & (1 << i)) result[i] = -result[i]; } store_vector4(inst, machine, result); diff --git a/src/mesa/shader/prog_instruction.h b/src/mesa/shader/prog_instruction.h index 4adce11f957..3109f6cbae5 100644 --- a/src/mesa/shader/prog_instruction.h +++ b/src/mesa/shader/prog_instruction.h @@ -261,37 +261,15 @@ struct prog_src_register GLuint Swizzle:12; GLuint RelAddr:1; - /** - * \name Source register "sign" control. - * - * The ARB and NV extensions allow varrying degrees of control over the - * sign of the source vector components. These values allow enough control - * for all flavors of the extensions. - */ - /*@{*/ - /** - * Per-component negation for the SWZ instruction. For non-SWZ - * instructions the only possible values are NEGATE_XYZW and NEGATE_NONE. - * - * \since - * ARB_vertex_program, ARB_fragment_program - */ - GLuint NegateBase:4; - - /** - * Take the component-wise absolute value. - * - * \since - * NV_fragment_program, NV_fragment_program_option, NV_vertex_program2, - * NV_vertex_program2_option. - */ + /** Take the component-wise absolute value */ GLuint Abs:1; /** - * Post-absolute value negation (all components). + * Post-Abs negation. + * This will either be NEGATE_NONE or NEGATE_XYZW, except for the SWZ + * instruction which allows per-component negation. */ - GLuint NegateAbs:1; - /*@}*/ + GLuint Negate:4; }; diff --git a/src/mesa/shader/prog_optimize.c b/src/mesa/shader/prog_optimize.c index 5f35dbf1289..6ba2e76ff96 100644 --- a/src/mesa/shader/prog_optimize.c +++ b/src/mesa/shader/prog_optimize.c @@ -660,7 +660,8 @@ find_live_intervals(struct gl_program *prog, } -static GLuint +/** Scan the array of used register flags to find free entry */ +static GLint alloc_register(GLboolean usedRegs[MAX_PROGRAM_TEMPS]) { GLuint k; @@ -670,7 +671,7 @@ alloc_register(GLboolean usedRegs[MAX_PROGRAM_TEMPS]) return k; } } - return MAX_PROGRAM_TEMPS; + return -1; } @@ -689,7 +690,7 @@ _mesa_reallocate_registers(struct gl_program *prog) GLint registerMap[MAX_PROGRAM_TEMPS]; GLboolean usedRegs[MAX_PROGRAM_TEMPS]; GLuint i; - GLuint maxTemp = 0; + GLint maxTemp = -1; if (dbg) { _mesa_printf("Optimize: Begin live-interval register reallocation\n"); @@ -754,15 +755,15 @@ _mesa_reallocate_registers(struct gl_program *prog) /* find a free register for this live interval */ { - const GLuint k = alloc_register(usedRegs); - if (k == MAX_PROGRAM_TEMPS) { + const GLint k = alloc_register(usedRegs); + if (k < 0) { /* out of registers, give up */ return; } registerMap[live->Reg] = k; maxTemp = MAX2(maxTemp, k); if (dbg) - _mesa_printf(" remap register %d -> %d\n", live->Reg, k); + _mesa_printf(" remap register %u -> %d\n", live->Reg, k); } /* Insert this live interval into the active list which is sorted diff --git a/src/mesa/shader/prog_parameter.c b/src/mesa/shader/prog_parameter.c index e9ed3985ee2..bcd8c5d9dc0 100644 --- a/src/mesa/shader/prog_parameter.c +++ b/src/mesa/shader/prog_parameter.c @@ -327,15 +327,16 @@ _mesa_add_sampler(struct gl_program_parameter_list *paramList, else { GLuint i; const GLint size = 1; /* a sampler is basically a texture unit number */ - GLfloat value; + GLfloat value[4]; GLint numSamplers = 0; for (i = 0; i < paramList->NumParameters; i++) { if (paramList->Parameters[i].Type == PROGRAM_SAMPLER) numSamplers++; } - value = (GLfloat) numSamplers; + value[0] = (GLfloat) numSamplers; + value[1] = value[2] = value[3] = 0.0F; (void) _mesa_add_parameter(paramList, PROGRAM_SAMPLER, name, - size, datatype, &value, NULL, 0x0); + size, datatype, value, NULL, 0x0); return numSamplers; } } diff --git a/src/mesa/shader/prog_print.c b/src/mesa/shader/prog_print.c index b832ddb4774..9967f2978de 100644 --- a/src/mesa/shader/prog_print.c +++ b/src/mesa/shader/prog_print.c @@ -325,19 +325,19 @@ reg_string(gl_register_file f, GLint index, gl_prog_print_mode mode, * \param extended if true, also allow 0, 1 values */ const char * -_mesa_swizzle_string(GLuint swizzle, GLuint negateBase, GLboolean extended) +_mesa_swizzle_string(GLuint swizzle, GLuint negateMask, GLboolean extended) { static const char swz[] = "xyzw01!?"; /* See SWIZZLE_x definitions */ static char s[20]; GLuint i = 0; - if (!extended && swizzle == SWIZZLE_NOOP && negateBase == 0) + if (!extended && swizzle == SWIZZLE_NOOP && negateMask == 0) return ""; /* no swizzle/negation */ if (!extended) s[i++] = '.'; - if (negateBase & NEGATE_X) + if (negateMask & NEGATE_X) s[i++] = '-'; s[i++] = swz[GET_SWZ(swizzle, 0)]; @@ -345,7 +345,7 @@ _mesa_swizzle_string(GLuint swizzle, GLuint negateBase, GLboolean extended) s[i++] = ','; } - if (negateBase & NEGATE_Y) + if (negateMask & NEGATE_Y) s[i++] = '-'; s[i++] = swz[GET_SWZ(swizzle, 1)]; @@ -353,7 +353,7 @@ _mesa_swizzle_string(GLuint swizzle, GLuint negateBase, GLboolean extended) s[i++] = ','; } - if (negateBase & NEGATE_Z) + if (negateMask & NEGATE_Z) s[i++] = '-'; s[i++] = swz[GET_SWZ(swizzle, 2)]; @@ -361,7 +361,7 @@ _mesa_swizzle_string(GLuint swizzle, GLuint negateBase, GLboolean extended) s[i++] = ','; } - if (negateBase & NEGATE_W) + if (negateMask & NEGATE_W) s[i++] = '-'; s[i++] = swz[GET_SWZ(swizzle, 3)]; @@ -465,14 +465,14 @@ fprint_src_reg(FILE *f, reg_string((gl_register_file) srcReg->File, srcReg->Index, mode, srcReg->RelAddr, prog), _mesa_swizzle_string(srcReg->Swizzle, - srcReg->NegateBase, GL_FALSE), + srcReg->Negate, GL_FALSE), abs); #if 0 _mesa_fprintf(f, "%s[%d]%s", file_string((gl_register_file) srcReg->File, mode), srcReg->Index, _mesa_swizzle_string(srcReg->Swizzle, - srcReg->NegateBase, GL_FALSE)); + srcReg->Negate, GL_FALSE)); #endif } @@ -566,7 +566,7 @@ _mesa_fprint_instruction_opt(FILE *f, mode), inst->SrcReg[0].Index, _mesa_swizzle_string(inst->SrcReg[0].Swizzle, - inst->SrcReg[0].NegateBase, GL_FALSE)); + inst->SrcReg[0].Negate, GL_FALSE)); } if (inst->Comment) _mesa_fprintf(f, " # %s", inst->Comment); @@ -583,7 +583,7 @@ _mesa_fprint_instruction_opt(FILE *f, mode), inst->SrcReg[0].Index, _mesa_swizzle_string(inst->SrcReg[0].Swizzle, - inst->SrcReg[0].NegateBase, GL_TRUE)); + inst->SrcReg[0].Negate, GL_TRUE)); fprint_comment(f, inst); break; case OPCODE_TEX: @@ -873,6 +873,7 @@ _mesa_fprint_parameter_list(FILE *f, return; _mesa_fprintf(f, "param list %p\n", (void *) list); + _mesa_fprintf(f, "dirty state flags: 0x%x\n", list->StateFlags); for (i = 0; i < list->NumParameters; i++){ struct gl_program_parameter *param = list->Parameters + i; const GLfloat *v = list->ParameterValues[i]; diff --git a/src/mesa/shader/prog_statevars.c b/src/mesa/shader/prog_statevars.c index 37a3f1fc8ca..058d4bbafb7 100644 --- a/src/mesa/shader/prog_statevars.c +++ b/src/mesa/shader/prog_statevars.c @@ -816,7 +816,10 @@ append_token(char *dst, gl_state_index k) break; /* BEGIN internal state vars */ case STATE_INTERNAL: - append(dst, "(internal)"); + append(dst, ".internal."); + break; + case STATE_CURRENT_ATTRIB: + append(dst, "current"); break; case STATE_NORMAL_SCALE: append(dst, "normalScale"); @@ -986,7 +989,9 @@ _mesa_program_state_string(const gl_state_index state[STATE_LENGTH]) break; case STATE_INTERNAL: append_token(str, state[1]); - break; + if (state[1] == STATE_CURRENT_ATTRIB) + append_index(str, state[2]); + break; default: _mesa_problem(NULL, "Invalid state in _mesa_program_state_string"); break; diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c index 2e5632710e4..d270bf9e1c3 100644 --- a/src/mesa/shader/program.c +++ b/src/mesa/shader/program.c @@ -296,6 +296,7 @@ _mesa_new_program(GLcontext *ctx, GLenum target, GLuint id) struct gl_program *prog; switch (target) { case GL_VERTEX_PROGRAM_ARB: /* == GL_VERTEX_PROGRAM_NV */ + case GL_VERTEX_STATE_PROGRAM_NV: prog = _mesa_init_vertex_program(ctx, CALLOC_STRUCT(gl_vertex_program), target, id ); break; diff --git a/src/mesa/shader/programopt.c b/src/mesa/shader/programopt.c index e283f8933b2..ecd98dc85c5 100644 --- a/src/mesa/shader/programopt.c +++ b/src/mesa/shader/programopt.c @@ -241,7 +241,7 @@ _mesa_append_fog_code(GLcontext *ctx, struct gl_fragment_program *fprog) inst->DstReg.WriteMask = WRITEMASK_X; inst->SrcReg[0].File = PROGRAM_TEMPORARY; inst->SrcReg[0].Index = fogFactorTemp; - inst->SrcReg[0].NegateBase = NEGATE_XYZW; + inst->SrcReg[0].Negate = NEGATE_XYZW; inst->SrcReg[0].Swizzle = SWIZZLE_XXXX; inst->SaturateMode = SATURATE_ZERO_ONE; inst++; diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index 61289db2d2e..644cd39185c 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -51,11 +51,6 @@ #include "glapi/dispatch.h" -#ifndef GL_PROGRAM_BINARY_LENGTH_OES -#define GL_PROGRAM_BINARY_LENGTH_OES 0x8741 -#endif - - /** * Allocate a new gl_shader_program object, initialize it. */ @@ -1794,7 +1789,7 @@ _mesa_uniform(GLcontext *ctx, GLint location, GLsizei count, return; } - FLUSH_VERTICES(ctx, _NEW_PROGRAM); + FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS); uniform = &shProg->Uniforms->Uniforms[location]; @@ -1902,7 +1897,7 @@ set_program_uniform_matrix(GLcontext *ctx, struct gl_program *program, */ static void _mesa_uniform_matrix(GLcontext *ctx, GLint cols, GLint rows, - GLenum matrixType, GLint location, GLsizei count, + GLint location, GLsizei count, GLboolean transpose, const GLfloat *values) { struct gl_shader_program *shProg = ctx->Shader.CurrentProgram; @@ -1934,7 +1929,7 @@ _mesa_uniform_matrix(GLcontext *ctx, GLint cols, GLint rows, return; } - FLUSH_VERTICES(ctx, _NEW_PROGRAM); + FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS); uniform = &shProg->Uniforms->Uniforms[location]; diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index 8493c490fbc..3f455e0640b 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -1135,7 +1135,7 @@ emit_negation(slang_emit_info *emitInfo, slang_ir_node *n) n->Children[0]->Store, NULL, NULL); - inst->SrcReg[0].NegateBase = NEGATE_XYZW; + inst->SrcReg[0].Negate = NEGATE_XYZW; return inst; } diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index f79092291b0..ca15ce1b474 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -37,11 +37,8 @@ -/* This is used to initialize st->atoms[]. We could use this list - * directly except for a single atom, st_update_constants, which has a - * .dirty value which changes according to the parameters of the - * current fragment and vertex programs, and so cannot be a static - * value. +/** + * This is used to initialize st->atoms[]. */ static const struct st_tracked_state *atoms[] = { @@ -67,34 +64,13 @@ static const struct st_tracked_state *atoms[] = void st_init_atoms( struct st_context *st ) { - GLuint i; - - st->atoms = _mesa_malloc(sizeof(atoms)); - st->nr_atoms = sizeof(atoms)/sizeof(*atoms); - memcpy(st->atoms, atoms, sizeof(atoms)); - - /* Patch in a pointer to the dynamic state atom: - */ - for (i = 0; i < st->nr_atoms; i++) { - if (st->atoms[i] == &st_update_vs_constants) { - st->atoms[i] = &st->constants.tracked_state[PIPE_SHADER_VERTEX]; - st->atoms[i][0] = st_update_vs_constants; - } - - if (st->atoms[i] == &st_update_fs_constants) { - st->atoms[i] = &st->constants.tracked_state[PIPE_SHADER_FRAGMENT]; - st->atoms[i][0] = st_update_fs_constants; - } - } + /* no-op */ } void st_destroy_atoms( struct st_context *st ) { - if (st->atoms) { - _mesa_free(st->atoms); - st->atoms = NULL; - } + /* no-op */ } @@ -172,8 +148,8 @@ void st_validate_state( struct st_context *st ) memset(&examined, 0, sizeof(examined)); prev = *state; - for (i = 0; i < st->nr_atoms; i++) { - const struct st_tracked_state *atom = st->atoms[i]; + for (i = 0; i < Elements(atoms); i++) { + const struct st_tracked_state *atom = atoms[i]; struct st_state_flags generated; // _mesa_printf("atom %s %x/%x\n", atom->name, atom->dirty.mesa, atom->dirty.st); @@ -185,7 +161,7 @@ void st_validate_state( struct st_context *st ) } if (check_state(state, &atom->dirty)) { - st->atoms[i]->update( st ); + atoms[i]->update( st ); // _mesa_printf("after: %x\n", atom->dirty.mesa); } @@ -203,11 +179,9 @@ void st_validate_state( struct st_context *st ) } else { - const GLuint nr = st->nr_atoms; - - for (i = 0; i < nr; i++) { - if (check_state(state, &st->atoms[i]->dirty)) - st->atoms[i]->update( st ); + for (i = 0; i < Elements(atoms); i++) { + if (check_state(state, &atoms[i]->dirty)) + atoms[i]->update( st ); } } diff --git a/src/mesa/state_tracker/st_atom_constbuf.c b/src/mesa/state_tracker/st_atom_constbuf.c index fd81ac36d2c..77ecd0719e6 100644 --- a/src/mesa/state_tracker/st_atom_constbuf.c +++ b/src/mesa/state_tracker/st_atom_constbuf.c @@ -42,7 +42,7 @@ #include "st_atom.h" #include "st_atom_constbuf.h" #include "st_program.h" - +#include "st_inlines.h" /** * Pass the given program parameters to the graphics pipe as a @@ -62,12 +62,6 @@ void st_upload_constants( struct st_context *st, if (params && params->NumParameters) { const uint paramBytes = params->NumParameters * sizeof(GLfloat) * 4; - /* Update our own dependency flags. This works because this - * function will also be called whenever the program changes. - */ - st->constants.tracked_state[id].dirty.mesa = - (params->StateFlags | _NEW_PROGRAM); - _mesa_load_state_parameters(st->ctx, params); /* We always need to get a new buffer, to keep the drivers simple and @@ -86,9 +80,9 @@ void st_upload_constants( struct st_context *st, /* load Mesa constants into the constant buffer */ if (cbuf->buffer) - pipe_buffer_write(pipe->screen, cbuf->buffer, - 0, paramBytes, - params->ParameterValues); + st_no_flush_pipe_buffer_write(st, cbuf->buffer, + 0, paramBytes, + params->ParameterValues); st->pipe->set_constant_buffer(st->pipe, id, 0, cbuf); } @@ -111,7 +105,7 @@ static void update_vs_constants(struct st_context *st ) const struct st_tracked_state st_update_vs_constants = { "st_update_vs_constants", /* name */ { /* dirty */ - 0, /* set dynamically above */ /* mesa */ + _NEW_PROGRAM_CONSTANTS, ST_NEW_VERTEX_PROGRAM, /* st */ }, update_vs_constants /* update */ @@ -130,7 +124,7 @@ static void update_fs_constants(struct st_context *st ) const struct st_tracked_state st_update_fs_constants = { "st_update_fs_constants", /* name */ { /* dirty */ - 0, /* set dynamically above */ /* mesa */ + (_NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS), /* mesa */ ST_NEW_FRAGMENT_PROGRAM, /* st */ }, update_fs_constants /* update */ diff --git a/src/mesa/state_tracker/st_atom_pixeltransfer.c b/src/mesa/state_tracker/st_atom_pixeltransfer.c index 8d0029dde5f..eff3666ca88 100644 --- a/src/mesa/state_tracker/st_atom_pixeltransfer.c +++ b/src/mesa/state_tracker/st_atom_pixeltransfer.c @@ -45,6 +45,7 @@ #include "st_format.h" #include "st_program.h" #include "st_texture.h" +#include "st_inlines.h" #include "pipe/p_screen.h" #include "pipe/p_context.h" @@ -125,8 +126,7 @@ create_color_map_texture(GLcontext *ctx) /* create texture for color map/table */ pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, format, 0, - texSize, texSize, 1, 0, - PIPE_TEXTURE_USAGE_SAMPLER); + texSize, texSize, 1, PIPE_TEXTURE_USAGE_SAMPLER); return pt; } @@ -148,8 +148,9 @@ load_color_map_texture(GLcontext *ctx, struct pipe_texture *pt) uint *dest; uint i, j; - transfer = screen->get_tex_transfer(screen, pt, 0, 0, 0, PIPE_TRANSFER_WRITE, - 0, 0, texSize, texSize); + transfer = st_cond_flush_get_tex_transfer(st_context(ctx), + pt, 0, 0, 0, PIPE_TRANSFER_WRITE, + 0, 0, texSize, texSize); dest = (uint *) screen->transfer_map(screen, transfer); /* Pack four 1D maps into a 2D texture: diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c index 3f9a825a153..7f793cf08db 100644 --- a/src/mesa/state_tracker/st_cb_accum.c +++ b/src/mesa/state_tracker/st_cb_accum.c @@ -40,6 +40,8 @@ #include "st_draw.h" #include "st_public.h" #include "st_format.h" +#include "st_texture.h" +#include "st_inlines.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" @@ -118,9 +120,10 @@ st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) const GLint height = ctx->DrawBuffer->_Ymax - ypos; GLubyte *map; - acc_pt = screen->get_tex_transfer(screen, acc_strb->texture, 0, 0, 0, - PIPE_TRANSFER_WRITE, xpos, ypos, - width, height); + acc_pt = st_cond_flush_get_tex_transfer(st_context(ctx), acc_strb->texture, + 0, 0, 0, + PIPE_TRANSFER_WRITE, xpos, ypos, + width, height); map = screen->transfer_map(screen, acc_pt); /* note acc_strb->format might not equal acc_pt->format */ @@ -163,9 +166,11 @@ accum_mad(GLcontext *ctx, GLfloat scale, GLfloat bias, struct pipe_transfer *acc_pt; GLubyte *map; - acc_pt = screen->get_tex_transfer(screen, acc_strb->texture, 0, 0, 0, - PIPE_TRANSFER_READ_WRITE, xpos, ypos, - width, height); + acc_pt = st_cond_flush_get_tex_transfer(st_context(ctx), acc_strb->texture, + 0, 0, 0, + PIPE_TRANSFER_READ_WRITE, + xpos, ypos, + width, height); map = screen->transfer_map(screen, acc_pt); /* note acc_strb->format might not equal acc_pt->format */ @@ -192,23 +197,25 @@ accum_mad(GLcontext *ctx, GLfloat scale, GLfloat bias, static void -accum_accum(struct pipe_context *pipe, GLfloat value, +accum_accum(struct st_context *st, GLfloat value, GLint xpos, GLint ypos, GLint width, GLint height, struct st_renderbuffer *acc_strb, struct st_renderbuffer *color_strb) { + struct pipe_context *pipe = st->pipe; struct pipe_screen *screen = pipe->screen; struct pipe_transfer *acc_trans, *color_trans; GLfloat *colorBuf, *accBuf; GLint i; - acc_trans = screen->get_tex_transfer(screen, acc_strb->texture, 0, 0, 0, - PIPE_TRANSFER_READ, xpos, ypos, - width, height); + acc_trans = st_cond_flush_get_tex_transfer(st, acc_strb->texture, 0, 0, 0, + PIPE_TRANSFER_READ, xpos, ypos, + width, height); - color_trans = screen->get_tex_transfer(screen, color_strb->texture, 0, 0, 0, - PIPE_TRANSFER_READ, xpos, ypos, - width, height); + color_trans = st_cond_flush_get_tex_transfer(st, color_strb->texture, + 0, 0, 0, + PIPE_TRANSFER_READ, xpos, ypos, + width, height); colorBuf = (GLfloat *) _mesa_malloc(width * height * 4 * sizeof(GLfloat)); accBuf = (GLfloat *) _mesa_malloc(width * height * 4 * sizeof(GLfloat)); @@ -221,9 +228,9 @@ accum_accum(struct pipe_context *pipe, GLfloat value, } screen->tex_transfer_destroy(acc_trans); - acc_trans = screen->get_tex_transfer(screen, acc_strb->texture, 0, 0, 0, - PIPE_TRANSFER_WRITE, xpos, ypos, - width, height); + acc_trans = st_no_flush_get_tex_transfer(st, acc_strb->texture, 0, 0, 0, + PIPE_TRANSFER_WRITE, xpos, ypos, + width, height); acc_put_tile_rgba(pipe, acc_trans, 0, 0, width, height, accBuf); @@ -235,23 +242,25 @@ accum_accum(struct pipe_context *pipe, GLfloat value, static void -accum_load(struct pipe_context *pipe, GLfloat value, +accum_load(struct st_context *st, GLfloat value, GLint xpos, GLint ypos, GLint width, GLint height, struct st_renderbuffer *acc_strb, struct st_renderbuffer *color_strb) { + struct pipe_context *pipe = st->pipe; struct pipe_screen *screen = pipe->screen; struct pipe_transfer *acc_trans, *color_trans; GLfloat *buf; GLint i; - acc_trans = screen->get_tex_transfer(screen, acc_strb->texture, 0, 0, 0, - PIPE_TRANSFER_WRITE, xpos, ypos, - width, height); + acc_trans = st_cond_flush_get_tex_transfer(st, acc_strb->texture, 0, 0, 0, + PIPE_TRANSFER_WRITE, xpos, ypos, + width, height); - color_trans = screen->get_tex_transfer(screen, color_strb->texture, 0, 0, 0, - PIPE_TRANSFER_READ, xpos, ypos, - width, height); + color_trans = st_cond_flush_get_tex_transfer(st, color_strb->texture, + 0, 0, 0, + PIPE_TRANSFER_READ, xpos, ypos, + width, height); buf = (GLfloat *) _mesa_malloc(width * height * 4 * sizeof(GLfloat)); @@ -284,13 +293,16 @@ accum_return(GLcontext *ctx, GLfloat value, abuf = (GLfloat *) _mesa_malloc(width * height * 4 * sizeof(GLfloat)); - acc_trans = screen->get_tex_transfer(screen, acc_strb->texture, 0, 0, 0, - PIPE_TRANSFER_READ, xpos, ypos, - width, height); + acc_trans = st_cond_flush_get_tex_transfer(st_context(ctx), + acc_strb->texture, 0, 0, 0, + PIPE_TRANSFER_READ, xpos, ypos, + width, height); - color_trans = screen->get_tex_transfer(screen, color_strb->texture, 0, 0, 0, - PIPE_TRANSFER_READ_WRITE, xpos, ypos, - width, height); + color_trans = st_cond_flush_get_tex_transfer(st_context(ctx), + color_strb->texture, 0, 0, 0, + PIPE_TRANSFER_READ_WRITE, + xpos, ypos, + width, height); acc_get_tile_rgba(pipe, acc_trans, 0, 0, width, height, abuf); @@ -325,7 +337,6 @@ static void st_Accum(GLcontext *ctx, GLenum op, GLfloat value) { struct st_context *st = ctx->st; - struct pipe_context *pipe = st->pipe; struct st_renderbuffer *acc_strb = st_renderbuffer(ctx->DrawBuffer->Attachment[BUFFER_ACCUM].Renderbuffer); struct st_renderbuffer *color_strb @@ -352,11 +363,11 @@ st_Accum(GLcontext *ctx, GLenum op, GLfloat value) break; case GL_ACCUM: if (value != 0.0F) { - accum_accum(pipe, value, xpos, ypos, width, height, acc_strb, color_strb); + accum_accum(st, value, xpos, ypos, width, height, acc_strb, color_strb); } break; case GL_LOAD: - accum_load(pipe, value, xpos, ypos, width, height, acc_strb, color_strb); + accum_load(st, value, xpos, ypos, width, height, acc_strb, color_strb); break; case GL_RETURN: accum_return(ctx, value, xpos, ypos, width, height, acc_strb, color_strb); diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 2d547dd0722..8709633557c 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -47,6 +47,8 @@ #include "st_cb_program.h" #include "st_mesa_to_tgsi.h" #include "st_texture.h" +#include "st_inlines.h" + #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" @@ -147,7 +149,7 @@ make_bitmap_fragment_program(GLcontext *ctx, GLuint samplerIndex) p->Instructions[ic].SrcReg[0].Swizzle = SWIZZLE_XXXX; p->Instructions[ic].SrcReg[0].Index = 0; - p->Instructions[ic].SrcReg[0].NegateBase = NEGATE_XYZW; + p->Instructions[ic].SrcReg[0].Negate = NEGATE_XYZW; ic++; /* END; */ @@ -330,15 +332,16 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, * Create texture to hold bitmap pattern. */ pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, ctx->st->bitmap.tex_format, - 0, width, height, 1, 0, + 0, width, height, 1, PIPE_TEXTURE_USAGE_SAMPLER); if (!pt) { _mesa_unmap_bitmap_pbo(ctx, unpack); return NULL; } - transfer = screen->get_tex_transfer(screen, pt, 0, 0, 0, PIPE_TRANSFER_WRITE, - 0, 0, width, height); + transfer = st_no_flush_get_tex_transfer(st_context(ctx), pt, 0, 0, 0, + PIPE_TRANSFER_WRITE, + 0, 0, width, height); dest = screen->transfer_map(screen, transfer); @@ -425,11 +428,11 @@ setup_bitmap_vertex_data(struct st_context *st, } /* put vertex data into vbuf */ - pipe_buffer_write(pipe->screen, - st->bitmap.vbuf, - st->bitmap.vbuf_slot * sizeof st->bitmap.vertices, - sizeof st->bitmap.vertices, - st->bitmap.vertices); + st_no_flush_pipe_buffer_write(st, + st->bitmap.vbuf, + st->bitmap.vbuf_slot * sizeof st->bitmap.vertices, + sizeof st->bitmap.vertices, + st->bitmap.vertices); return st->bitmap.vbuf_slot++ * sizeof st->bitmap.vertices; } @@ -570,8 +573,10 @@ reset_cache(struct st_context *st) cache->ymin = 1000000; cache->ymax = -1000000; - if (cache->trans) + if (cache->trans) { screen->tex_transfer_destroy(cache->trans); + cache->trans = NULL; + } assert(!cache->texture); @@ -579,16 +584,27 @@ reset_cache(struct st_context *st) cache->texture = st_texture_create(st, PIPE_TEXTURE_2D, st->bitmap.tex_format, 0, BITMAP_CACHE_WIDTH, BITMAP_CACHE_HEIGHT, - 1, 0, - PIPE_TEXTURE_USAGE_SAMPLER); + 1, PIPE_TEXTURE_USAGE_SAMPLER); + +} + +static void +create_cache_trans(struct st_context *st) +{ + struct pipe_context *pipe = st->pipe; + struct pipe_screen *screen = pipe->screen; + struct bitmap_cache *cache = st->bitmap.cache; + + if (cache->trans) + return; /* Map the texture transfer. * Subsequent glBitmap calls will write into the texture image. */ - cache->trans = screen->get_tex_transfer(screen, cache->texture, 0, 0, 0, - PIPE_TRANSFER_WRITE, 0, 0, - BITMAP_CACHE_WIDTH, - BITMAP_CACHE_HEIGHT); + cache->trans = st_no_flush_get_tex_transfer(st, cache->texture, 0, 0, 0, + PIPE_TRANSFER_WRITE, 0, 0, + BITMAP_CACHE_WIDTH, + BITMAP_CACHE_HEIGHT); cache->buffer = screen->transfer_map(screen, cache->trans); /* init image to all 0xff */ @@ -620,11 +636,13 @@ st_flush_bitmap_cache(struct st_context *st) /* The texture transfer has been mapped until now. * So unmap and release the texture transfer before drawing. */ - screen->transfer_unmap(screen, cache->trans); - cache->buffer = NULL; + if (cache->trans) { + screen->transfer_unmap(screen, cache->trans); + cache->buffer = NULL; - screen->tex_transfer_destroy(cache->trans); - cache->trans = NULL; + screen->tex_transfer_destroy(cache->trans); + cache->trans = NULL; + } draw_bitmap_quad(st->ctx, cache->xpos, @@ -709,6 +727,9 @@ accum_bitmap(struct st_context *st, if (y + height > cache->ymax) cache->ymax = y + height; + /* create the transfer if needed */ + create_cache_trans(st); + unpack_bitmap(st, px, py, width, height, unpack, bitmap, cache->buffer, BITMAP_CACHE_WIDTH); @@ -821,8 +842,7 @@ st_destroy_bitmap(struct st_context *st) struct pipe_screen *screen = pipe->screen; struct bitmap_cache *cache = st->bitmap.cache; - screen->transfer_unmap(screen, cache->trans); - screen->tex_transfer_destroy(cache->trans); + if (st->bitmap.vs) { cso_delete_vertex_shader(st->cso_context, st->bitmap.vs); @@ -834,7 +854,11 @@ st_destroy_bitmap(struct st_context *st) st->bitmap.vbuf = NULL; } - if (st->bitmap.cache) { + if (cache) { + if (cache->trans) { + screen->transfer_unmap(screen, cache->trans); + screen->tex_transfer_destroy(cache->trans); + } pipe_texture_reference(&st->bitmap.cache->texture, NULL); _mesa_free(st->bitmap.cache); st->bitmap.cache = NULL; diff --git a/src/mesa/state_tracker/st_cb_bufferobjects.c b/src/mesa/state_tracker/st_cb_bufferobjects.c index 3651e4ae7dc..a94e11fff12 100644 --- a/src/mesa/state_tracker/st_cb_bufferobjects.c +++ b/src/mesa/state_tracker/st_cb_bufferobjects.c @@ -30,6 +30,7 @@ #include "main/mtypes.h" #include "main/bufferobj.h" +#include "st_inlines.h" #include "st_context.h" #include "st_cb_bufferobjects.h" @@ -97,13 +98,13 @@ st_bufferobj_subdata(GLcontext *ctx, GLsizeiptrARB size, const GLvoid * data, struct gl_buffer_object *obj) { - struct pipe_context *pipe = st_context(ctx)->pipe; struct st_buffer_object *st_obj = st_buffer_object(obj); if (offset >= st_obj->size || size > (st_obj->size - offset)) return; - pipe_buffer_write(pipe->screen, st_obj->buffer, offset, size, data); + st_cond_flush_pipe_buffer_write(st_context(ctx), st_obj->buffer, + offset, size, data); } @@ -117,13 +118,13 @@ st_bufferobj_get_subdata(GLcontext *ctx, GLsizeiptrARB size, GLvoid * data, struct gl_buffer_object *obj) { - struct pipe_context *pipe = st_context(ctx)->pipe; struct st_buffer_object *st_obj = st_buffer_object(obj); if (offset >= st_obj->size || size > (st_obj->size - offset)) return; - pipe_buffer_read(pipe->screen, st_obj->buffer, offset, size, data); + st_cond_flush_pipe_buffer_read(st_context(ctx), st_obj->buffer, + offset, size, data); } @@ -168,10 +169,16 @@ st_bufferobj_data(GLcontext *ctx, st_obj->buffer = pipe_buffer_create( pipe->screen, 32, buffer_usage, size ); + if (!st_obj->buffer) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "glBufferDataARB"); + return; + } + st_obj->size = size; if (data) - st_bufferobj_subdata(ctx, target, 0, size, data, obj); + st_no_flush_pipe_buffer_write(st_context(ctx), st_obj->buffer, 0, + size, data); } @@ -182,7 +189,6 @@ static void * st_bufferobj_map(GLcontext *ctx, GLenum target, GLenum access, struct gl_buffer_object *obj) { - struct pipe_context *pipe = st_context(ctx)->pipe; struct st_buffer_object *st_obj = st_buffer_object(obj); GLuint flags; @@ -200,7 +206,9 @@ st_bufferobj_map(GLcontext *ctx, GLenum target, GLenum access, break; } - obj->Pointer = pipe_buffer_map(pipe->screen, st_obj->buffer, flags); + obj->Pointer = st_cond_flush_pipe_buffer_map(st_context(ctx), + st_obj->buffer, + flags); if(obj->Pointer) { obj->Offset = 0; obj->Length = obj->Size; diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 5bdc6a13309..880e83108c1 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -45,6 +45,7 @@ #include "st_program.h" #include "st_public.h" #include "st_mesa_to_tgsi.h" +#include "st_inlines.h" #include "pipe/p_context.h" #include "pipe/p_inlines.h" @@ -166,10 +167,10 @@ draw_quad(GLcontext *ctx, } /* put vertex data into vbuf */ - pipe_buffer_write(pipe->screen, st->clear.vbuf, - st->clear.vbuf_slot * sizeof(st->clear.vertices), - sizeof(st->clear.vertices), - st->clear.vertices); + st_no_flush_pipe_buffer_write(st, st->clear.vbuf, + st->clear.vbuf_slot * sizeof(st->clear.vertices), + sizeof(st->clear.vertices), + st->clear.vertices); /* draw */ util_draw_vertex_buffer(pipe, diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index ebb1d1142aa..08dc7c930e2 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -53,6 +53,8 @@ #include "st_format.h" #include "st_mesa_to_tgsi.h" #include "st_texture.h" +#include "st_inlines.h" + #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" @@ -144,6 +146,8 @@ combined_drawpix_fragment_program(GLcontext *ctx) st->pixel_xfer.xfer_prog_sn = st->pixel_xfer.program->serialNo; st->pixel_xfer.user_prog_sn = st->fp->serialNo; st->pixel_xfer.combined_prog_sn = stfp->serialNo; + /* can't reference new program directly, already have a reference on it */ + st_reference_fragprog(st, &st->pixel_xfer.combined_prog, NULL); st->pixel_xfer.combined_prog = stfp; } @@ -351,8 +355,7 @@ make_texture(struct st_context *st, if (!pixels) return NULL; - pt = st_texture_create(st, PIPE_TEXTURE_2D, pipeFormat, 0, width, height, - 1, 0, + pt = st_texture_create(st, PIPE_TEXTURE_2D, pipeFormat, 0, width, height, 1, PIPE_TEXTURE_USAGE_SAMPLER); if (!pt) { _mesa_unmap_drawpix_pbo(ctx, unpack); @@ -369,9 +372,9 @@ make_texture(struct st_context *st, /* we'll do pixel transfer in a fragment shader */ ctx->_ImageTransferState = 0x0; - transfer = screen->get_tex_transfer(screen, pt, 0, 0, 0, - PIPE_TRANSFER_WRITE, 0, 0, - width, height); + transfer = st_no_flush_get_tex_transfer(st, pt, 0, 0, 0, + PIPE_TRANSFER_WRITE, 0, 0, + width, height); /* map texture transfer */ dest = screen->transfer_map(screen, transfer); @@ -491,7 +494,7 @@ draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, /* allocate/load buffer object with vertex data */ buf = pipe_buffer_create(pipe->screen, 32, PIPE_BUFFER_USAGE_VERTEX, sizeof(verts)); - pipe_buffer_write(pipe->screen, buf, 0, sizeof(verts), verts); + st_no_flush_pipe_buffer_write(st, buf, 0, sizeof(verts), verts); util_draw_vertex_buffer(pipe, buf, 0, PIPE_PRIM_QUADS, @@ -632,8 +635,6 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y, GLint skipPixels; ubyte *stmap; - pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL); - strb = st_renderbuffer(ctx->DrawBuffer-> Attachment[BUFFER_STENCIL].Renderbuffer); @@ -641,9 +642,9 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y, y = ctx->DrawBuffer->Height - y - height; } - pt = screen->get_tex_transfer(screen, strb->texture, 0, 0, 0, - PIPE_TRANSFER_WRITE, x, y, - width, height); + pt = st_cond_flush_get_tex_transfer(st_context(ctx), strb->texture, 0, 0, 0, + PIPE_TRANSFER_WRITE, x, y, + width, height); stmap = screen->transfer_map(screen, pt); @@ -826,9 +827,10 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy, GL_STENCIL_INDEX, GL_UNSIGNED_BYTE, &ctx->DefaultPacking, buffer); - ptDraw = screen->get_tex_transfer(screen, rbDraw->texture, 0, 0, 0, - PIPE_TRANSFER_WRITE, dstx, dsty, - width, height); + ptDraw = st_cond_flush_get_tex_transfer(st_context(ctx), + rbDraw->texture, 0, 0, 0, + PIPE_TRANSFER_WRITE, dstx, dsty, + width, height); assert(ptDraw->block.width == 1); assert(ptDraw->block.height == 1); @@ -904,7 +906,6 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, GLfloat *color; enum pipe_format srcFormat, texFormat; - /* make sure rendering has completed */ pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL); st_validate_state(st); @@ -951,7 +952,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, } pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, texFormat, 0, - width, height, 1, 0, + width, height, 1, PIPE_TEXTURE_USAGE_SAMPLER); if (!pt) return; @@ -978,13 +979,13 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, else { /* CPU-based fallback/conversion */ struct pipe_transfer *ptRead = - screen->get_tex_transfer(screen, rbRead->texture, 0, 0, 0, - PIPE_TRANSFER_READ, srcx, srcy, width, - height); + st_cond_flush_get_tex_transfer(st, rbRead->texture, 0, 0, 0, + PIPE_TRANSFER_READ, srcx, srcy, width, + height); struct pipe_transfer *ptTex = - screen->get_tex_transfer(screen, pt, 0, 0, 0, PIPE_TRANSFER_WRITE, - 0, 0, width, height); + st_cond_flush_get_tex_transfer(st, pt, 0, 0, 0, PIPE_TRANSFER_WRITE, + 0, 0, width, height); if (type == GL_COLOR) { /* alternate path using get/put_tile() */ diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index f74d0d46d01..1590f275e2a 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -96,29 +96,22 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, pipe_surface_reference( &strb->surface, NULL ); pipe_texture_reference( &strb->texture, NULL ); - + /* Setup new texture template. + */ memset(&template, 0, sizeof(template)); - + template.target = PIPE_TEXTURE_2D; if (strb->format != PIPE_FORMAT_NONE) { template.format = strb->format; } else { template.format = st_choose_renderbuffer_format(pipe, internalFormat); } - - strb->Base.Width = width; - strb->Base.Height = height; - init_renderbuffer_bits(strb, template.format); - - template.target = PIPE_TEXTURE_2D; - template.compressed = 0; pf_get_block(template.format, &template.block); template.width[0] = width; template.height[0] = height; template.depth[0] = 1; template.last_level = 0; template.nr_samples = rb->NumSamples; - if (pf_is_depth_stencil(template.format)) { template.tex_usage = PIPE_TEXTURE_USAGE_DEPTH_STENCIL; } @@ -127,6 +120,10 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, PIPE_TEXTURE_USAGE_RENDER_TARGET); } + /* init renderbuffer fields */ + strb->Base.Width = width; + strb->Base.Height = height; + init_renderbuffer_bits(strb, template.format); /* Probably need dedicated flags for surface usage too: */ diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index 9ce5f3fe84c..7a4bbf5ce3a 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -42,13 +42,15 @@ #include "pipe/p_defines.h" #include "pipe/p_inlines.h" #include "util/u_tile.h" + #include "st_context.h" #include "st_cb_bitmap.h" #include "st_cb_readpixels.h" #include "st_cb_fbo.h" #include "st_format.h" #include "st_public.h" - +#include "st_texture.h" +#include "st_inlines.h" /** * Special case for reading stencil buffer. @@ -73,8 +75,11 @@ st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y, } /* Create a read transfer from the renderbuffer's texture */ - pt = screen->get_tex_transfer(screen, strb->texture, 0, 0, 0, - PIPE_TRANSFER_READ, x, y, width, height); + + pt = st_cond_flush_get_tex_transfer(st_context(ctx), strb->texture, + 0, 0, 0, + PIPE_TRANSFER_READ, x, y, + width, height); /* map the stencil buffer */ stmap = screen->transfer_map(screen, pt); @@ -240,8 +245,10 @@ st_fast_readpixels(GLcontext *ctx, struct st_renderbuffer *strb, y = strb->texture->height[0] - y - height; } - trans = screen->get_tex_transfer(screen, strb->texture, 0, 0, 0, - PIPE_TRANSFER_READ, x, y, width, height); + trans = st_cond_flush_get_tex_transfer(st_context(ctx), strb->texture, + 0, 0, 0, + PIPE_TRANSFER_READ, x, y, + width, height); if (!trans) { return GL_FALSE; } @@ -350,7 +357,6 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, if (!dest) return; - /* make sure rendering has completed */ st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL); if (format == GL_STENCIL_INDEX || @@ -395,8 +401,10 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, } /* Create a read transfer from the renderbuffer's texture */ - trans = screen->get_tex_transfer(screen, strb->texture, 0, 0, 0, - PIPE_TRANSFER_READ, x, y, width, height); + trans = st_cond_flush_get_tex_transfer(st_context(ctx), strb->texture, + 0, 0, 0, + PIPE_TRANSFER_READ, x, y, + width, height); /* determine bottom-to-top vs. top-to-bottom order */ if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) { diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 8013e69e8e1..c3e990e0775 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -26,6 +26,7 @@ **************************************************************************/ #include "main/mfeatures.h" +#include "main/bufferobj.h" #if FEATURE_convolve #include "main/convolve.h" #endif @@ -49,12 +50,14 @@ #include "state_tracker/st_public.h" #include "state_tracker/st_texture.h" #include "state_tracker/st_gen_mipmap.h" +#include "state_tracker/st_inlines.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" #include "util/u_tile.h" #include "util/u_blit.h" +#include "util/u_surface.h" #define DBG if (0) printf @@ -88,7 +91,7 @@ gl_target_to_pipe(GLenum target) * Return nominal bytes per texel for a compressed format, 0 for non-compressed * format. */ -static int +static GLuint compressed_num_bytes(GLuint mesaFormat) { switch(mesaFormat) { @@ -110,6 +113,25 @@ compressed_num_bytes(GLuint mesaFormat) } +static GLboolean +is_compressed_mesa_format(const struct gl_texture_format *format) +{ + switch (format->MesaFormat) { + case MESA_FORMAT_RGB_DXT1: + case MESA_FORMAT_RGBA_DXT1: + case MESA_FORMAT_RGBA_DXT3: + case MESA_FORMAT_RGBA_DXT5: + case MESA_FORMAT_SRGB_DXT1: + case MESA_FORMAT_SRGBA_DXT1: + case MESA_FORMAT_SRGBA_DXT3: + case MESA_FORMAT_SRGBA_DXT5: + return GL_TRUE; + default: + return GL_FALSE; + } +} + + /** called via ctx->Driver.NewTextureImage() */ static struct gl_texture_image * st_NewTextureImage(GLcontext * ctx) @@ -169,7 +191,7 @@ st_FreeTextureImageData(GLcontext * ctx, struct gl_texture_image *texImage) * than COPY_DWORDS would: * XXX Put this in src/mesa/main/imports.h ??? */ -#if defined(i386) || defined(__i386__) +#if defined(PIPE_CC_GCC) && defined(PIPE_ARCH_X86) static INLINE void * __memcpy(void *to, const void *from, size_t n) { @@ -227,6 +249,21 @@ logbase2(int n) /** + * Return default texture usage bitmask for the given texture format. + */ +static GLuint +default_usage(enum pipe_format fmt) +{ + GLuint usage = PIPE_TEXTURE_USAGE_SAMPLER; + if (pf_is_depth_stencil(fmt)) + usage |= PIPE_TEXTURE_USAGE_DEPTH_STENCIL; + else + usage |= PIPE_TEXTURE_USAGE_RENDER_TARGET; + return usage; +} + + +/** * Allocate a pipe_texture object for the given st_texture_object using * the given st_texture_image to guess the mipmap size/levels. * @@ -250,7 +287,7 @@ guess_and_alloc_texture(struct st_context *st, GLuint width = stImage->base.Width2; /* size w/out border */ GLuint height = stImage->base.Height2; GLuint depth = stImage->base.Depth2; - GLuint i, comp_byte = 0; + GLuint i, usage; enum pipe_format fmt; DBG("%s\n", __FUNCTION__); @@ -308,10 +345,10 @@ guess_and_alloc_texture(struct st_context *st, lastLevel = firstLevel + MAX2(MAX2(l2width, l2height), l2depth); } - if (stImage->base.IsCompressed) - comp_byte = compressed_num_bytes(stImage->base.TexFormat->MesaFormat); - fmt = st_mesa_format_to_pipe_format(stImage->base.TexFormat->MesaFormat); + + usage = default_usage(fmt); + stObj->pt = st_texture_create(st, gl_target_to_pipe(stObj->base.Target), fmt, @@ -319,11 +356,7 @@ guess_and_alloc_texture(struct st_context *st, width, height, depth, - comp_byte, - ( (pf_is_depth_stencil(fmt) ? - PIPE_TEXTURE_USAGE_DEPTH_STENCIL : - PIPE_TEXTURE_USAGE_RENDER_TARGET) | - PIPE_TEXTURE_USAGE_SAMPLER )); + usage); DBG("%s - success\n", __FUNCTION__); } @@ -368,6 +401,110 @@ strip_texture_border(GLint border, /** + * Try to do texture compression via rendering. If the Gallium driver + * can render into a compressed surface this will allow us to do texture + * compression. + * \return GL_TRUE for success, GL_FALSE for failure + */ +static GLboolean +compress_with_blit(GLcontext * ctx, + GLenum target, GLint level, + GLint xoffset, GLint yoffset, GLint zoffset, + GLint width, GLint height, GLint depth, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *unpack, + struct gl_texture_image *texImage) +{ + const GLuint dstImageOffsets[1] = {0}; + struct st_texture_image *stImage = st_texture_image(texImage); + struct pipe_screen *screen = ctx->st->pipe->screen; + const GLuint face = _mesa_tex_target_to_face(target); + const struct gl_texture_format *mesa_format; + struct pipe_texture templ; + struct pipe_texture *src_tex; + struct pipe_surface *dst_surface; + struct pipe_transfer *tex_xfer; + void *map; + + + if (!stImage->pt) { + /* XXX: Can this happen? Should we assert? */ + return GL_FALSE; + } + + /* get destination surface (in the compressed texture) */ + dst_surface = screen->get_tex_surface(screen, stImage->pt, + stImage->face, stImage->level, 0, + PIPE_BUFFER_USAGE_GPU_WRITE); + if (!dst_surface) { + /* can't render into this format (or other problem) */ + return GL_FALSE; + } + + /* Choose format for the temporary RGBA texture image. + */ + mesa_format = st_ChooseTextureFormat(ctx, GL_RGBA, format, type); + assert(mesa_format); + if (!mesa_format) + return GL_FALSE; + + /* Create the temporary source texture + */ + memset(&templ, 0, sizeof(templ)); + templ.target = PIPE_TEXTURE_2D; + templ.format = st_mesa_format_to_pipe_format(mesa_format->MesaFormat); + pf_get_block(templ.format, &templ.block); + templ.width[0] = width; + templ.height[0] = height; + templ.depth[0] = 1; + templ.last_level = 0; + templ.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER; + src_tex = screen->texture_create(screen, &templ); + + if (!src_tex) + return GL_FALSE; + + /* Put user's tex data into the temporary texture + */ + tex_xfer = st_cond_flush_get_tex_transfer(st_context(ctx), src_tex, + face, level, 0, + PIPE_TRANSFER_WRITE, + 0, 0, width, height); /* x, y, w, h */ + map = screen->transfer_map(screen, tex_xfer); + + mesa_format->StoreImage(ctx, 2, GL_RGBA, mesa_format, + map, /* dest ptr */ + 0, 0, 0, /* dest x/y/z offset */ + tex_xfer->stride, /* dest row stride (bytes) */ + dstImageOffsets, /* image offsets (for 3D only) */ + width, height, 1, /* size */ + format, type, /* source format/type */ + pixels, /* source data */ + unpack); /* source data packing */ + + screen->transfer_unmap(screen, tex_xfer); + screen->tex_transfer_destroy(tex_xfer); + + /* copy / compress image */ + util_blit_pixels_tex(ctx->st->blit, + src_tex, /* pipe_texture (src) */ + 0, 0, /* src x0, y0 */ + width, height, /* src x1, y1 */ + dst_surface, /* pipe_surface (dst) */ + xoffset, yoffset, /* dst x0, y0 */ + xoffset + width, /* dst x1 */ + yoffset + height, /* dst y1 */ + 0.0, /* z */ + PIPE_TEX_MIPFILTER_NEAREST); + + pipe_surface_reference(&dst_surface, NULL); + pipe_texture_reference(&src_tex, NULL); + + return GL_TRUE; +} + + +/** * Do glTexImage1/2/3D(). */ static void @@ -381,8 +518,9 @@ st_TexImage(GLcontext * ctx, const struct gl_pixelstore_attrib *unpack, struct gl_texture_object *texObj, struct gl_texture_image *texImage, - GLsizei imageSize, int compressed) + GLsizei imageSize, GLboolean compressed_src) { + struct pipe_screen *screen = ctx->st->pipe->screen; struct st_texture_object *stObj = st_texture_object(texObj); struct st_texture_image *stImage = st_texture_image(texImage); GLint postConvWidth, postConvHeight; @@ -395,8 +533,7 @@ st_TexImage(GLcontext * ctx, /* gallium does not support texture borders, strip it off */ if (border) { - strip_texture_border(border, &width, &height, &depth, - unpack, &unpackNB); + strip_texture_border(border, &width, &height, &depth, unpack, &unpackNB); unpack = &unpackNB; texImage->Width = width; texImage->Height = height; @@ -512,11 +649,12 @@ st_TexImage(GLcontext * ctx, * the expectation that the texture will be set up but nothing * more will be done. This is where those calls return: */ - if (compressed) { + if (compressed_src) { pixels = _mesa_validate_pbo_compressed_teximage(ctx, imageSize, pixels, unpack, "glCompressedTexImage"); - } else { + } + else { pixels = _mesa_validate_pbo_teximage(ctx, dims, width, height, 1, format, type, pixels, unpack, "glTexImage"); @@ -524,12 +662,28 @@ st_TexImage(GLcontext * ctx, if (!pixels) return; + /* See if we can do texture compression with a blit/render. + */ + if (!compressed_src && + !ctx->Mesa_DXTn && + is_compressed_mesa_format(texImage->TexFormat) && + screen->is_format_supported(screen, + stImage->pt->format, + stImage->pt->target, + PIPE_TEXTURE_USAGE_RENDER_TARGET, 0)) { + if (compress_with_blit(ctx, target, level, 0, 0, 0, width, height, depth, + format, type, pixels, unpack, texImage)) { + return; + } + } + if (stImage->pt) { texImage->Data = st_texture_image_map(ctx->st, stImage, 0, PIPE_TRANSFER_WRITE, 0, 0, stImage->base.Width, stImage->base.Height); - dstRowStride = stImage->transfer->stride; + if(stImage->transfer) + dstRowStride = stImage->transfer->stride; } else { /* Allocate regular memory and store the image there temporarily. */ @@ -559,16 +713,16 @@ st_TexImage(GLcontext * ctx, * the blitter to copy. Or, use the hardware to do the format * conversion and copy: */ - if (compressed) { + if (compressed_src) { memcpy(texImage->Data, pixels, imageSize); } else { - GLuint srcImageStride = _mesa_image_image_stride(unpack, width, height, - format, type); - int i; + const GLuint srcImageStride = + _mesa_image_image_stride(unpack, width, height, format, type); + GLint i; const GLubyte *src = (const GLubyte *) pixels; - for (i = 0; i++ < depth;) { + for (i = 0; i < depth; i++) { if (!texImage->TexFormat->StoreImage(ctx, dims, texImage->_BaseFormat, texImage->TexFormat, @@ -581,9 +735,11 @@ st_TexImage(GLcontext * ctx, _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); } - if (stImage->pt && i < depth) { + if (stImage->pt && i + 1 < depth) { + /* unmap this slice */ st_texture_image_unmap(ctx->st, stImage); - texImage->Data = st_texture_image_map(ctx->st, stImage, i, + /* map next slice of 3D texture */ + texImage->Data = st_texture_image_map(ctx->st, stImage, i + 1, PIPE_TRANSFER_WRITE, 0, 0, stImage->base.Width, stImage->base.Height); @@ -594,7 +750,7 @@ st_TexImage(GLcontext * ctx, _mesa_unmap_teximage_pbo(ctx, unpack); - if (stImage->pt) { + if (stImage->pt && texImage->Data) { st_texture_image_unmap(ctx->st, stImage); texImage->Data = NULL; } @@ -616,9 +772,9 @@ st_TexImage3D(GLcontext * ctx, struct gl_texture_object *texObj, struct gl_texture_image *texImage) { - st_TexImage(ctx, 3, target, level, - internalFormat, width, height, depth, border, - format, type, pixels, unpack, texObj, texImage, 0, 0); + st_TexImage(ctx, 3, target, level, internalFormat, width, height, depth, + border, format, type, pixels, unpack, texObj, texImage, + 0, GL_FALSE); } @@ -632,9 +788,8 @@ st_TexImage2D(GLcontext * ctx, struct gl_texture_object *texObj, struct gl_texture_image *texImage) { - st_TexImage(ctx, 2, target, level, - internalFormat, width, height, 1, border, - format, type, pixels, unpack, texObj, texImage, 0, 0); + st_TexImage(ctx, 2, target, level, internalFormat, width, height, 1, border, + format, type, pixels, unpack, texObj, texImage, 0, GL_FALSE); } @@ -648,9 +803,8 @@ st_TexImage1D(GLcontext * ctx, struct gl_texture_object *texObj, struct gl_texture_image *texImage) { - st_TexImage(ctx, 1, target, level, - internalFormat, width, 1, 1, border, - format, type, pixels, unpack, texObj, texImage, 0, 0); + st_TexImage(ctx, 1, target, level, internalFormat, width, 1, 1, border, + format, type, pixels, unpack, texObj, texImage, 0, GL_FALSE); } @@ -662,12 +816,96 @@ st_CompressedTexImage2D(GLcontext *ctx, GLenum target, GLint level, struct gl_texture_object *texObj, struct gl_texture_image *texImage) { - st_TexImage(ctx, 2, target, level, - internalFormat, width, height, 1, border, - 0, 0, data, &ctx->Unpack, texObj, texImage, imageSize, 1); + st_TexImage(ctx, 2, target, level, internalFormat, width, height, 1, border, + 0, 0, data, &ctx->Unpack, texObj, texImage, imageSize, GL_TRUE); } + +/** + * glGetTexImage() helper: decompress a compressed texture by rendering + * a textured quad. Store the results in the user's buffer. + */ +static void +decompress_with_blit(GLcontext * ctx, GLenum target, GLint level, + GLenum format, GLenum type, GLvoid *pixels, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + struct pipe_screen *screen = ctx->st->pipe->screen; + struct st_texture_image *stImage = st_texture_image(texImage); + const GLuint width = texImage->Width; + const GLuint height = texImage->Height; + struct pipe_surface *dst_surface; + struct pipe_texture *dst_texture; + struct pipe_transfer *tex_xfer; + + /* create temp / dest surface */ + if (!util_create_rgba_surface(screen, width, height, + &dst_texture, &dst_surface)) { + _mesa_problem(ctx, "util_create_rgba_surface() failed " + "in decompress_with_blit()"); + return; + } + + /* blit/render/decompress */ + util_blit_pixels_tex(ctx->st->blit, + stImage->pt, /* pipe_texture (src) */ + 0, 0, /* src x0, y0 */ + width, height, /* src x1, y1 */ + dst_surface, /* pipe_surface (dst) */ + 0, 0, /* dst x0, y0 */ + width, height, /* dst x1, y1 */ + 0.0, /* z */ + PIPE_TEX_MIPFILTER_NEAREST); + + /* map the dst_surface so we can read from it */ + tex_xfer = st_cond_flush_get_tex_transfer(st_context(ctx), + dst_texture, 0, 0, 0, + PIPE_TRANSFER_READ, + 0, 0, width, height); + + pixels = _mesa_map_readpix_pbo(ctx, &ctx->Pack, pixels); + + /* copy/pack data into user buffer */ + if (st_equal_formats(stImage->pt->format, format, type)) { + /* memcpy */ + const uint bytesPerRow = width * pf_get_size(stImage->pt->format); + ubyte *map = screen->transfer_map(screen, tex_xfer); + GLuint row; + for (row = 0; row < height; row++) { + GLvoid *dest = _mesa_image_address2d(&ctx->Pack, pixels, width, + height, format, type, row, 0); + memcpy(dest, map, bytesPerRow); + map += tex_xfer->stride; + } + screen->transfer_unmap(screen, tex_xfer); + } + else { + /* format translation via floats */ + GLuint row; + for (row = 0; row < height; row++) { + const GLbitfield transferOps = 0x0; /* bypassed for glGetTexImage() */ + GLfloat rgba[4 * MAX_WIDTH]; + GLvoid *dest = _mesa_image_address2d(&ctx->Pack, pixels, width, + height, format, type, row, 0); + + /* get float[4] rgba row from surface */ + pipe_get_tile_rgba(tex_xfer, 0, row, width, 1, rgba); + + _mesa_pack_rgba_span_float(ctx, width, (GLfloat (*)[4]) rgba, format, + type, dest, &ctx->Pack, transferOps); + } + } + + _mesa_unmap_readpix_pbo(ctx, &ctx->Pack); + + /* destroy the temp / dest surface */ + util_destroy_rgba_surface(dst_texture, dst_surface); +} + + + /** * Need to map texture image into memory before copying image data, * then unmap it. @@ -676,22 +914,36 @@ static void st_get_tex_image(GLcontext * ctx, GLenum target, GLint level, GLenum format, GLenum type, GLvoid * pixels, struct gl_texture_object *texObj, - struct gl_texture_image *texImage, int compressed) + struct gl_texture_image *texImage, GLboolean compressed_dst) { struct st_texture_image *stImage = st_texture_image(texImage); - GLuint dstImageStride = _mesa_image_image_stride(&ctx->Pack, - texImage->Width, - texImage->Height, - format, type); - GLuint depth; - GLuint i; + const GLuint dstImageStride = + _mesa_image_image_stride(&ctx->Pack, texImage->Width, texImage->Height, + format, type); + GLuint depth, i; GLubyte *dest; + if (stImage->pt && + pf_is_compressed(stImage->pt->format) && + !compressed_dst) { + /* Need to decompress the texture. + * We'll do this by rendering a textured quad. + * Note that we only expect RGBA formats (no Z/depth formats). + */ + decompress_with_blit(ctx, target, level, format, type, pixels, + texObj, texImage); + return; + } + /* Map */ if (stImage->pt) { /* Image is stored in hardware format in a buffer managed by the * kernel. Need to explicitly map and unmap it. */ + + st_teximage_flush_before_map(ctx->st, stImage->pt, 0, level, + PIPE_TRANSFER_READ); + texImage->Data = st_texture_image_map(ctx->st, stImage, 0, PIPE_TRANSFER_READ, 0, 0, stImage->base.Width, @@ -715,18 +967,21 @@ st_get_tex_image(GLcontext * ctx, GLenum target, GLint level, dest = (GLubyte *) pixels; - for (i = 0; i++ < depth;) { - if (compressed) { + for (i = 0; i < depth; i++) { + if (compressed_dst) { _mesa_get_compressed_teximage(ctx, target, level, dest, texObj, texImage); - } else { + } + else { _mesa_get_teximage(ctx, target, level, format, type, dest, texObj, texImage); } - if (stImage->pt && i < depth) { + if (stImage->pt && i + 1 < depth) { + /* unmap this slice */ st_texture_image_unmap(ctx->st, stImage); - texImage->Data = st_texture_image_map(ctx->st, stImage, i, + /* map next slice of 3D texture */ + texImage->Data = st_texture_image_map(ctx->st, stImage, i + 1, PIPE_TRANSFER_READ, 0, 0, stImage->base.Width, stImage->base.Height); @@ -750,8 +1005,8 @@ st_GetTexImage(GLcontext * ctx, GLenum target, GLint level, struct gl_texture_object *texObj, struct gl_texture_image *texImage) { - st_get_tex_image(ctx, target, level, format, type, pixels, - texObj, texImage, 0); + st_get_tex_image(ctx, target, level, format, type, pixels, texObj, texImage, + GL_FALSE); } @@ -761,17 +1016,14 @@ st_GetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level, struct gl_texture_object *texObj, struct gl_texture_image *texImage) { - st_get_tex_image(ctx, target, level, 0, 0, pixels, - (struct gl_texture_object *) texObj, - (struct gl_texture_image *) texImage, 1); + st_get_tex_image(ctx, target, level, 0, 0, pixels, texObj, texImage, + GL_TRUE); } static void -st_TexSubimage(GLcontext * ctx, - GLint dims, - GLenum target, GLint level, +st_TexSubimage(GLcontext *ctx, GLint dims, GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLint width, GLint height, GLint depth, GLenum format, GLenum type, const void *pixels, @@ -779,11 +1031,12 @@ st_TexSubimage(GLcontext * ctx, struct gl_texture_object *texObj, struct gl_texture_image *texImage) { + struct pipe_screen *screen = ctx->st->pipe->screen; struct st_texture_image *stImage = st_texture_image(texImage); GLuint dstRowStride; - GLuint srcImageStride = _mesa_image_image_stride(packing, width, height, - format, type); - int i; + const GLuint srcImageStride = + _mesa_image_image_stride(packing, width, height, format, type); + GLint i; const GLubyte *src; DBG("%s target %s level %d offset %d,%d %dx%d\n", __FUNCTION__, @@ -796,10 +1049,28 @@ st_TexSubimage(GLcontext * ctx, if (!pixels) return; + /* See if we can do texture compression with a blit/render. + */ + if (!ctx->Mesa_DXTn && + is_compressed_mesa_format(texImage->TexFormat) && + screen->is_format_supported(screen, + stImage->pt->format, + stImage->pt->target, + PIPE_TEXTURE_USAGE_RENDER_TARGET, 0)) { + if (compress_with_blit(ctx, target, level, + xoffset, yoffset, zoffset, + width, height, depth, + format, type, pixels, packing, texImage)) { + return; + } + } + /* Map buffer if necessary. Need to lock to prevent other contexts * from uploading the buffer under us. */ if (stImage->pt) { + st_teximage_flush_before_map(ctx->st, stImage->pt, 0, level, + PIPE_TRANSFER_WRITE); texImage->Data = st_texture_image_map(ctx->st, stImage, zoffset, PIPE_TRANSFER_WRITE, xoffset, yoffset, @@ -814,7 +1085,7 @@ st_TexSubimage(GLcontext * ctx, src = (const GLubyte *) pixels; dstRowStride = stImage->transfer->stride; - for (i = 0; i++ < depth;) { + for (i = 0; i < depth; i++) { if (!texImage->TexFormat->StoreImage(ctx, dims, texImage->_BaseFormat, texImage->TexFormat, texImage->Data, @@ -826,10 +1097,12 @@ st_TexSubimage(GLcontext * ctx, _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage"); } - if (stImage->pt && i < depth) { - /* map next slice of 3D texture */ + if (stImage->pt && i + 1 < depth) { + /* unmap this slice */ st_texture_image_unmap(ctx->st, stImage); - texImage->Data = st_texture_image_map(ctx->st, stImage, zoffset + i, + /* map next slice of 3D texture */ + texImage->Data = st_texture_image_map(ctx->st, stImage, + zoffset + i + 1, PIPE_TRANSFER_WRITE, xoffset, yoffset, width, height); @@ -852,73 +1125,58 @@ st_TexSubimage(GLcontext * ctx, static void -st_TexSubImage3D(GLcontext * ctx, - GLenum target, - GLint level, - GLint xoffset, GLint yoffset, GLint zoffset, - GLsizei width, GLsizei height, GLsizei depth, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) +st_TexSubImage3D(GLcontext *ctx, GLenum target, GLint level, + GLint xoffset, GLint yoffset, GLint zoffset, + GLsizei width, GLsizei height, GLsizei depth, + GLenum format, GLenum type, const GLvoid *pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) { - st_TexSubimage(ctx, 3, target, level, - xoffset, yoffset, zoffset, - width, height, depth, - format, type, pixels, packing, texObj, texImage); + st_TexSubimage(ctx, 3, target, level, xoffset, yoffset, zoffset, + width, height, depth, format, type, + pixels, packing, texObj, texImage); } static void -st_TexSubImage2D(GLcontext * ctx, - GLenum target, - GLint level, - GLint xoffset, GLint yoffset, - GLsizei width, GLsizei height, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) +st_TexSubImage2D(GLcontext *ctx, GLenum target, GLint level, + GLint xoffset, GLint yoffset, + GLsizei width, GLsizei height, + GLenum format, GLenum type, const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) { - st_TexSubimage(ctx, 2, target, level, - xoffset, yoffset, 0, - width, height, 1, - format, type, pixels, packing, texObj, texImage); + st_TexSubimage(ctx, 2, target, level, xoffset, yoffset, 0, + width, height, 1, format, type, + pixels, packing, texObj, texImage); } static void -st_TexSubImage1D(GLcontext * ctx, - GLenum target, - GLint level, - GLint xoffset, - GLsizei width, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) +st_TexSubImage1D(GLcontext *ctx, GLenum target, GLint level, + GLint xoffset, GLsizei width, GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) { - st_TexSubimage(ctx, 1, target, level, - xoffset, 0, 0, - width, 1, 1, + st_TexSubimage(ctx, 1, target, level, xoffset, 0, 0, width, 1, 1, format, type, pixels, packing, texObj, texImage); } /** - * Do a CopyTexSubImage operation using a read transfer from the source, a write - * transfer to the destination and get_tile()/put_tile() to access the pixels/texels. + * Do a CopyTexSubImage operation using a read transfer from the source, + * a write transfer to the destination and get_tile()/put_tile() to access + * the pixels/texels. * * Note: srcY=0=TOP of renderbuffer */ static void -fallback_copy_texsubimage(GLcontext *ctx, - GLenum target, - GLint level, +fallback_copy_texsubimage(GLcontext *ctx, GLenum target, GLint level, struct st_renderbuffer *strb, struct st_texture_image *stImage, GLenum baseFormat, @@ -937,12 +1195,15 @@ fallback_copy_texsubimage(GLcontext *ctx, srcY = strb->Base.Height - srcY - height; } - src_trans = screen->get_tex_transfer( screen, - strb->texture, - 0, 0, 0, - PIPE_TRANSFER_READ, - srcX, srcY, - width, height); + src_trans = st_cond_flush_get_tex_transfer( st_context(ctx), + strb->texture, + 0, 0, 0, + PIPE_TRANSFER_READ, + srcX, srcY, + width, height); + + st_teximage_flush_before_map(ctx->st, stImage->pt, 0, 0, + PIPE_TRANSFER_WRITE); texDest = st_texture_image_map(ctx->st, stImage, 0, PIPE_TRANSFER_WRITE, destX, destY, width, height); @@ -980,8 +1241,8 @@ fallback_copy_texsubimage(GLcontext *ctx, if (tempSrc && texDest) { const GLint dims = 2; + const GLint dstRowStride = stImage->transfer->stride; struct gl_texture_image *texImage = &stImage->base; - GLint dstRowStride = stImage->transfer->stride; struct gl_pixelstore_attrib unpack = ctx->DefaultPacking; if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) { @@ -1095,7 +1356,6 @@ st_copy_texsubimage(GLcontext *ctx, if (src_format == dest_format && !do_flip) { /* use surface_copy() / blit */ - dest_surface = screen->get_tex_surface(screen, stImage->pt, stImage->face, stImage->level, destZ, @@ -1122,7 +1382,7 @@ st_copy_texsubimage(GLcontext *ctx, PIPE_TEXTURE_USAGE_RENDER_TARGET, 0)) { /* draw textured quad to do the copy */ - int srcY0, srcY1; + GLint srcY0, srcY1; dest_surface = screen->get_tex_surface(screen, stImage->pt, stImage->face, stImage->level, @@ -1179,11 +1439,6 @@ st_CopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, struct gl_texture_image *texImage = _mesa_select_tex_image(ctx, texObj, target, level); -#if 0 - if (border) - goto fail; -#endif - /* Setup or redefine the texture object, texture and texture * image. Don't populate yet. */ @@ -1274,8 +1529,8 @@ calculate_first_last_level(struct st_texture_object *stObj) * and having firstLevel and lastLevel as signed prevents the need for * extra sign checks. */ - int firstLevel; - int lastLevel; + GLint firstLevel; + GLint lastLevel; /* Yes, this looks overly complicated, but it's all needed. */ @@ -1329,16 +1584,21 @@ copy_image_data_to_texture(struct st_context *st, /* More straightforward upload. */ - st_texture_image_data(st->pipe, - stObj->pt, - stImage->face, - dstLevel, - stImage->base.Data, - stImage->base.RowStride * - stObj->pt->block.size, - stImage->base.RowStride * - stImage->base.Height * - stObj->pt->block.size); + + st_teximage_flush_before_map(st, stObj->pt, stImage->face, dstLevel, + PIPE_TRANSFER_WRITE); + + + st_texture_image_data(st, + stObj->pt, + stImage->face, + dstLevel, + stImage->base.Data, + stImage->base.RowStride * + stObj->pt->block.size, + stImage->base.RowStride * + stImage->base.Height * + stObj->pt->block.size); _mesa_align_free(stImage->base.Data); stImage->base.Data = NULL; } @@ -1360,9 +1620,7 @@ st_finalize_texture(GLcontext *ctx, { struct st_texture_object *stObj = st_texture_object(tObj); const GLuint nr_faces = (stObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; - int comp_byte = 0; - int cpp; - GLuint face; + GLuint cpp, face; struct st_texture_image *firstImage; *needFlush = GL_FALSE; @@ -1384,14 +1642,12 @@ st_finalize_texture(GLcontext *ctx, if (firstImage->pt && firstImage->pt != stObj->pt && firstImage->pt->last_level >= stObj->lastLevel) { - pipe_texture_reference(&stObj->pt, firstImage->pt); } /* FIXME: determine format block instead of cpp */ if (firstImage->base.IsCompressed) { - comp_byte = compressed_num_bytes(firstImage->base.TexFormat->MesaFormat); - cpp = comp_byte; + cpp = compressed_num_bytes(firstImage->base.TexFormat->MesaFormat); } else { cpp = firstImage->base.TexFormat->TexelBytes; @@ -1409,8 +1665,9 @@ st_finalize_texture(GLcontext *ctx, stObj->pt->width[0] != firstImage->base.Width2 || stObj->pt->height[0] != firstImage->base.Height2 || stObj->pt->depth[0] != firstImage->base.Depth2 || - stObj->pt->block.size/stObj->pt->block.width != cpp || /* Nominal bytes per pixel */ - stObj->pt->compressed != firstImage->base.IsCompressed) { + /* Nominal bytes per pixel: */ + stObj->pt->block.size / stObj->pt->block.width != cpp) + { pipe_texture_reference(&stObj->pt, NULL); ctx->st->dirty.st |= ST_NEW_FRAMEBUFFER; } @@ -1421,6 +1678,8 @@ st_finalize_texture(GLcontext *ctx, if (!stObj->pt) { const enum pipe_format fmt = st_mesa_format_to_pipe_format(firstImage->base.TexFormat->MesaFormat); + GLuint usage = default_usage(fmt); + stObj->pt = st_texture_create(ctx->st, gl_target_to_pipe(stObj->base.Target), fmt, @@ -1428,11 +1687,7 @@ st_finalize_texture(GLcontext *ctx, firstImage->base.Width2, firstImage->base.Height2, firstImage->base.Depth2, - comp_byte, - ( (pf_is_depth_stencil(fmt) ? - PIPE_TEXTURE_USAGE_DEPTH_STENCIL : - PIPE_TEXTURE_USAGE_RENDER_TARGET) | - PIPE_TEXTURE_USAGE_SAMPLER )); + usage); if (!stObj->pt) { _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index ae8c2978bf8..f840579a404 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -116,11 +116,6 @@ struct st_context char vendor[100]; char renderer[100]; - /* State to be validated: - */ - struct st_tracked_state **atoms; - GLuint nr_atoms; - struct st_state_flags dirty; GLboolean missing_textures; diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 9e2d60c9269..d507e3e58dd 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -716,3 +716,23 @@ st_ChooseTextureFormat(GLcontext *ctx, GLint internalFormat, return translate_gallium_format_to_mesa_format(pFormat); } + + +/** + * Test if a gallium format is equivalent to a GL format/type. + */ +GLboolean +st_equal_formats(enum pipe_format pFormat, GLenum format, GLenum type) +{ + switch (pFormat) { + case PIPE_FORMAT_R8G8B8A8_UNORM: + return format == GL_RGBA && type == GL_UNSIGNED_BYTE; + case PIPE_FORMAT_B8G8R8A8_UNORM: + return format == GL_BGRA && type == GL_UNSIGNED_BYTE; + case PIPE_FORMAT_R5G6B5_UNORM: + return format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5; + /* XXX more combos... */ + default: + return GL_FALSE; + } +} diff --git a/src/mesa/state_tracker/st_format.h b/src/mesa/state_tracker/st_format.h index 3f5ac3201b1..7bbbe2d5704 100644 --- a/src/mesa/state_tracker/st_format.h +++ b/src/mesa/state_tracker/st_format.h @@ -76,4 +76,8 @@ st_ChooseTextureFormat(GLcontext * ctx, GLint internalFormat, GLenum format, GLenum type); +extern GLboolean +st_equal_formats(enum pipe_format pFormat, GLenum format, GLenum type); + + #endif /* ST_CB_TEXIMAGE_H */ diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index 9cc2176d5e4..e159b4c9db5 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -47,6 +47,7 @@ #include "st_program.h" #include "st_texture.h" #include "st_cb_texture.h" +#include "st_inlines.h" /** @@ -123,14 +124,17 @@ fallback_generate_mipmap(GLcontext *ctx, GLenum target, const ubyte *srcData; ubyte *dstData; - srcTrans = screen->get_tex_transfer(screen, pt, face, srcLevel, zslice, - PIPE_TRANSFER_READ, 0, 0, - pt->width[srcLevel], - pt->height[srcLevel]); - dstTrans = screen->get_tex_transfer(screen, pt, face, dstLevel, zslice, - PIPE_TRANSFER_WRITE, 0, 0, - pt->width[dstLevel], - pt->height[dstLevel]); + srcTrans = st_cond_flush_get_tex_transfer(st_context(ctx), pt, face, + srcLevel, zslice, + PIPE_TRANSFER_READ, 0, 0, + pt->width[srcLevel], + pt->height[srcLevel]); + + dstTrans = st_cond_flush_get_tex_transfer(st_context(ctx), pt, face, + dstLevel, zslice, + PIPE_TRANSFER_WRITE, 0, 0, + pt->width[dstLevel], + pt->height[dstLevel]); srcData = (ubyte *) screen->transfer_map(screen, srcTrans); dstData = (ubyte *) screen->transfer_map(screen, dstTrans); diff --git a/src/mesa/state_tracker/st_inlines.h b/src/mesa/state_tracker/st_inlines.h new file mode 100644 index 00000000000..0322d5dfa6e --- /dev/null +++ b/src/mesa/state_tracker/st_inlines.h @@ -0,0 +1,122 @@ +#ifndef ST_INLINES_H +#define ST_INLINES_H + +#include "pipe/p_context.h" +#include "pipe/p_screen.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" +#include "pipe/p_state.h" + +#include "st_context.h" +#include "st_texture.h" +#include "st_public.h" + +static INLINE struct pipe_transfer * +st_cond_flush_get_tex_transfer(struct st_context *st, + struct pipe_texture *pt, + unsigned int face, + unsigned int level, + unsigned int zslice, + enum pipe_transfer_usage usage, + unsigned int x, unsigned int y, + unsigned int w, unsigned int h) +{ + struct pipe_screen *screen = st->pipe->screen; + + st_teximage_flush_before_map(st, pt, face, level, usage); + return screen->get_tex_transfer(screen, pt, face, level, zslice, usage, + x, y, w, h); +} + +static INLINE struct pipe_transfer * +st_no_flush_get_tex_transfer(struct st_context *st, + struct pipe_texture *pt, + unsigned int face, + unsigned int level, + unsigned int zslice, + enum pipe_transfer_usage usage, + unsigned int x, unsigned int y, + unsigned int w, unsigned int h) +{ + struct pipe_screen *screen = st->pipe->screen; + + return screen->get_tex_transfer(screen, pt, face, level, + zslice, usage, x, y, w, h); +} + +static INLINE void * +st_cond_flush_pipe_buffer_map(struct st_context *st, + struct pipe_buffer *buf, + unsigned int map_flags) +{ + struct pipe_context *pipe = st->pipe; + unsigned int referenced = pipe->is_buffer_referenced(pipe, buf); + + if (referenced && ((referenced & PIPE_REFERENCED_FOR_WRITE) || + (map_flags & PIPE_BUFFER_USAGE_CPU_WRITE))) + st_flush(st, PIPE_FLUSH_RENDER_CACHE, NULL); + + return pipe_buffer_map(pipe->screen, buf, map_flags); +} + +static INLINE void * +st_no_flush_pipe_buffer_map(struct st_context *st, + struct pipe_buffer *buf, + unsigned int map_flags) +{ + return pipe_buffer_map(st->pipe->screen, buf, map_flags); +} + + +static INLINE void +st_cond_flush_pipe_buffer_write(struct st_context *st, + struct pipe_buffer *buf, + unsigned int offset, + unsigned int size, + const void * data) +{ + struct pipe_context *pipe = st->pipe; + + if (pipe->is_buffer_referenced(pipe, buf)) + st_flush(st, PIPE_FLUSH_RENDER_CACHE, NULL); + + pipe_buffer_write(pipe->screen, buf, offset, size, data); +} + +static INLINE void +st_no_flush_pipe_buffer_write(struct st_context *st, + struct pipe_buffer *buf, + unsigned int offset, + unsigned int size, + const void * data) +{ + pipe_buffer_write(st->pipe->screen, buf, offset, size, data); +} + +static INLINE void +st_cond_flush_pipe_buffer_read(struct st_context *st, + struct pipe_buffer *buf, + unsigned int offset, + unsigned int size, + void * data) +{ + struct pipe_context *pipe = st->pipe; + + if (pipe->is_buffer_referenced(pipe, buf) & PIPE_REFERENCED_FOR_WRITE) + st_flush(st, PIPE_FLUSH_RENDER_CACHE, NULL); + + pipe_buffer_read(pipe->screen, buf, offset, size, data); +} + +static INLINE void +st_no_flush_pipe_buffer_read(struct st_context *st, + struct pipe_buffer *buf, + unsigned int offset, + unsigned int size, + void * data) +{ + pipe_buffer_read(st->pipe->screen, buf, offset, size, data); +} + +#endif + diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index ffa607dd87c..43c9afccc3b 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -275,8 +275,8 @@ compile_instruction( /* swizzle (ext swizzle also depends on negation) */ { GLuint swz[4]; - GLboolean extended = (inst->SrcReg[i].NegateBase != NEGATE_NONE && - inst->SrcReg[i].NegateBase != NEGATE_XYZW); + GLboolean extended = (inst->SrcReg[i].Negate != NEGATE_NONE && + inst->SrcReg[i].Negate != NEGATE_XYZW); for( j = 0; j < 4; j++ ) { swz[j] = GET_SWZ( inst->SrcReg[i].Swizzle, j ); if (swz[j] > SWIZZLE_W) @@ -296,20 +296,20 @@ compile_instruction( } } - if( inst->SrcReg[i].NegateBase == NEGATE_XYZW ) { + if( inst->SrcReg[i].Negate == NEGATE_XYZW ) { fullsrc->SrcRegister.Negate = 1; } - else if( inst->SrcReg[i].NegateBase != NEGATE_NONE ) { - if( inst->SrcReg[i].NegateBase & NEGATE_X ) { + else if( inst->SrcReg[i].Negate != NEGATE_NONE ) { + if( inst->SrcReg[i].Negate & NEGATE_X ) { fullsrc->SrcRegisterExtSwz.NegateX = 1; } - if( inst->SrcReg[i].NegateBase & NEGATE_Y ) { + if( inst->SrcReg[i].Negate & NEGATE_Y ) { fullsrc->SrcRegisterExtSwz.NegateY = 1; } - if( inst->SrcReg[i].NegateBase & NEGATE_Z ) { + if( inst->SrcReg[i].Negate & NEGATE_Z ) { fullsrc->SrcRegisterExtSwz.NegateZ = 1; } - if( inst->SrcReg[i].NegateBase & NEGATE_W ) { + if( inst->SrcReg[i].Negate & NEGATE_W ) { fullsrc->SrcRegisterExtSwz.NegateW = 1; } } @@ -318,10 +318,6 @@ compile_instruction( fullsrc->SrcRegisterExtMod.Absolute = 1; } - if( inst->SrcReg[i].NegateAbs ) { - fullsrc->SrcRegisterExtMod.Negate = 1; - } - if( inst->SrcReg[i].RelAddr ) { fullsrc->SrcRegister.Indirect = 1; diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 6348e83d8ae..2795570cf10 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -484,14 +484,14 @@ st_translate_fragment_program(struct st_context *st, /* handled above */ assert(0); break; - case FRAG_RESULT_COLOR: + default: + assert(attr == FRAG_RESULT_COLOR || + (FRAG_RESULT_DATA0 <= attr && attr < FRAG_RESULT_MAX)); fs_output_semantic_name[fs_num_outputs] = TGSI_SEMANTIC_COLOR; fs_output_semantic_index[fs_num_outputs] = numColors; outputMapping[attr] = fs_num_outputs; numColors++; break; - default: - assert(0); } output_flags[fs_num_outputs] = stfp->Base.Base.OutputFlags[attr]; diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c index 3f90ad502c2..10faa633ea8 100644 --- a/src/mesa/state_tracker/st_texture.c +++ b/src/mesa/state_tracker/st_texture.c @@ -30,6 +30,7 @@ #include "st_public.h" #include "st_texture.h" #include "st_cb_fbo.h" +#include "st_inlines.h" #include "main/enums.h" #include "main/teximage.h" #include "main/texstore.h" @@ -78,7 +79,6 @@ st_texture_create(struct st_context *st, GLuint width0, GLuint height0, GLuint depth0, - GLuint compress_byte, GLuint usage ) { struct pipe_texture pt, *newtex; @@ -101,7 +101,6 @@ st_texture_create(struct st_context *st, pt.width[0] = width0; pt.height[0] = height0; pt.depth[0] = depth0; - pt.compressed = compress_byte ? 1 : 0; pf_get_block(format, &pt.block); pt.tex_usage = usage; @@ -128,8 +127,7 @@ st_texture_match_image(const struct pipe_texture *pt, /* Check if this image's format matches the established texture's format. */ - if (st_mesa_format_to_pipe_format(image->TexFormat->MesaFormat) != pt->format || - image->IsCompressed != pt->compressed) + if (st_mesa_format_to_pipe_format(image->TexFormat->MesaFormat) != pt->format) return GL_FALSE; /* Test if this image's size matches what's expected in the @@ -191,13 +189,15 @@ st_texture_image_map(struct st_context *st, struct st_texture_image *stImage, GLuint zoffset, enum pipe_transfer_usage usage, GLuint x, GLuint y, GLuint w, GLuint h) { - struct pipe_screen *screen = st->pipe->screen; + struct pipe_context *pipe = st->pipe; + struct pipe_screen *screen = pipe->screen; struct pipe_texture *pt = stImage->pt; + DBG("%s \n", __FUNCTION__); - stImage->transfer = screen->get_tex_transfer(screen, pt, stImage->face, - stImage->level, zoffset, - usage, x, y, w, h); + stImage->transfer = st_no_flush_get_tex_transfer(st, pt, stImage->face, + stImage->level, zoffset, + usage, x, y, w, h); if (stImage->transfer) return screen->transfer_map(screen, stImage->transfer); @@ -254,13 +254,14 @@ st_surface_data(struct pipe_context *pipe, /* Upload data for a particular image. */ void -st_texture_image_data(struct pipe_context *pipe, +st_texture_image_data(struct st_context *st, struct pipe_texture *dst, GLuint face, GLuint level, void *src, GLuint src_row_stride, GLuint src_image_stride) { + struct pipe_context *pipe = st->pipe; struct pipe_screen *screen = pipe->screen; GLuint depth = dst->depth[level]; GLuint i; @@ -268,11 +269,12 @@ st_texture_image_data(struct pipe_context *pipe, struct pipe_transfer *dst_transfer; DBG("%s\n", __FUNCTION__); + for (i = 0; i < depth; i++) { - dst_transfer = screen->get_tex_transfer(screen, dst, face, level, i, - PIPE_TRANSFER_WRITE, 0, 0, - dst->width[level], - dst->height[level]); + dst_transfer = st_no_flush_get_tex_transfer(st, dst, face, level, i, + PIPE_TRANSFER_WRITE, 0, 0, + dst->width[level], + dst->height[level]); st_surface_data(pipe, dst_transfer, 0, 0, /* dstx, dsty */ @@ -484,3 +486,20 @@ st_release_teximage(struct st_framebuffer *stfb, uint surfIndex, return 1; } + +void +st_teximage_flush_before_map(struct st_context *st, + struct pipe_texture *pt, + unsigned int face, + unsigned int level, + enum pipe_transfer_usage usage) +{ + struct pipe_context *pipe = st->pipe; + unsigned referenced = + pipe->is_texture_referenced(pipe, pt, face, level); + + if (referenced && ((referenced & PIPE_REFERENCED_FOR_WRITE) || + usage == PIPE_TRANSFER_WRITE || + usage == PIPE_TRANSFER_READ_WRITE)) + st_flush(st, PIPE_FLUSH_RENDER_CACHE, NULL); +} diff --git a/src/mesa/state_tracker/st_texture.h b/src/mesa/state_tracker/st_texture.h index 28c2f580f68..b9d447cb56e 100644 --- a/src/mesa/state_tracker/st_texture.h +++ b/src/mesa/state_tracker/st_texture.h @@ -108,7 +108,6 @@ st_texture_create(struct st_context *st, GLuint width0, GLuint height0, GLuint depth0, - GLuint compress_byte, GLuint tex_usage ); @@ -157,7 +156,7 @@ st_texture_texel_offset(const struct pipe_texture * pt, /* Upload an image into a texture */ extern void -st_texture_image_data(struct pipe_context *pipe, +st_texture_image_data(struct st_context *st, struct pipe_texture *dst, GLuint face, GLuint level, void *src, GLuint src_row_pitch, GLuint src_image_pitch); @@ -171,5 +170,10 @@ st_texture_image_copy(struct pipe_context *pipe, struct pipe_texture *src, GLuint face); - +extern void +st_teximage_flush_before_map(struct st_context *st, + struct pipe_texture *pt, + unsigned int face, + unsigned int level, + enum pipe_transfer_usage usage); #endif diff --git a/src/mesa/swrast/s_context.c b/src/mesa/swrast/s_context.c index a7eaf76a0ad..f24f4fc59b2 100644 --- a/src/mesa/swrast/s_context.c +++ b/src/mesa/swrast/s_context.c @@ -264,13 +264,7 @@ _swrast_update_fragment_program(GLcontext *ctx, GLbitfield newState) { const struct gl_fragment_program *fp = ctx->FragmentProgram._Current; if (fp) { -#if 0 - /* XXX Need a way to trigger the initial loading of parameters - * even when there's no recent state changes. - */ - if (fp->Base.Parameters->StateFlags & newState) -#endif - _mesa_load_state_parameters(ctx, fp->Base.Parameters); + _mesa_load_state_parameters(ctx, fp->Base.Parameters); } } @@ -524,13 +518,6 @@ _swrast_invalidate_state( GLcontext *ctx, GLbitfield new_state ) new_state = ~0; } - { - const struct gl_fragment_program *fp = ctx->FragmentProgram._Current; - if (fp && (fp->Base.Parameters->StateFlags & new_state)) { - _mesa_load_state_parameters(ctx, fp->Base.Parameters); - } - } - if (new_state & swrast->InvalidateTriangleMask) swrast->Triangle = _swrast_validate_triangle; @@ -647,17 +634,7 @@ _swrast_validate_derived( GLcontext *ctx ) if (swrast->NewState & (_NEW_FOG | _NEW_PROGRAM)) _swrast_update_fog_state( ctx ); - if (swrast->NewState & (_NEW_MODELVIEW | - _NEW_PROJECTION | - _NEW_TEXTURE_MATRIX | - _NEW_FOG | - _NEW_LIGHT | - _NEW_LINE | - _NEW_TEXTURE | - _NEW_TRANSFORM | - _NEW_POINT | - _NEW_VIEWPORT | - _NEW_PROGRAM)) + if (swrast->NewState & (_NEW_PROGRAM_CONSTANTS | _NEW_PROGRAM)) _swrast_update_fragment_program( ctx, swrast->NewState ); if (swrast->NewState & (_NEW_TEXTURE | _NEW_PROGRAM)) { diff --git a/src/mesa/swrast/s_texcombine.c b/src/mesa/swrast/s_texcombine.c index fae7280efb4..889164b9861 100644 --- a/src/mesa/swrast/s_texcombine.c +++ b/src/mesa/swrast/s_texcombine.c @@ -49,8 +49,7 @@ typedef float (*float4_array)[4]; static INLINE float4_array get_texel_array(SWcontext *swrast, GLuint unit) { - return (float4_array) - (swrast->TexelBuffer + unit * MAX_WIDTH * 4 * sizeof(GLfloat)); + return (float4_array) (swrast->TexelBuffer + unit * MAX_WIDTH * 4); } @@ -408,17 +407,10 @@ texture_combine( GLcontext *ctx, GLuint unit, GLuint n, /* this produces a fixed rgba color, and the coord calc is done elsewhere */ for (i = 0; i < n; i++) { /* rgba result is 0,0,0,1 */ -#if CHAN_TYPE == GL_FLOAT rgba[i][RCOMP] = 0.0; rgba[i][GCOMP] = 0.0; rgba[i][BCOMP] = 0.0; rgba[i][ACOMP] = 1.0; -#else - rgba[i][RCOMP] = 0; - rgba[i][GCOMP] = 0; - rgba[i][BCOMP] = 0; - rgba[i][ACOMP] = CHAN_MAX; -#endif } return; /* no alpha processing */ default: diff --git a/src/mesa/vbo/vbo_save.h b/src/mesa/vbo/vbo_save.h index 9558f838837..86bbd24f7b1 100644 --- a/src/mesa/vbo/vbo_save.h +++ b/src/mesa/vbo/vbo_save.h @@ -64,6 +64,13 @@ struct vbo_save_vertex_list { GLubyte attrsz[VBO_ATTRIB_MAX]; GLuint vertex_size; + /* Copy of the final vertex from node->vertex_store->bufferobj. + * Keep this in regular (non-VBO) memory to avoid repeated + * map/unmap of the VBO when updating GL current data. + */ + GLfloat *current_data; + GLuint current_size; + GLuint buffer_offset; GLuint count; GLuint wrap_count; /* number of copied vertices at start */ diff --git a/src/mesa/vbo/vbo_save_api.c b/src/mesa/vbo/vbo_save_api.c index 52b6f1884ee..868226075ac 100644 --- a/src/mesa/vbo/vbo_save_api.c +++ b/src/mesa/vbo/vbo_save_api.c @@ -289,6 +289,31 @@ static void _save_compile_vertex_list( GLcontext *ctx ) node->vertex_store->refcount++; node->prim_store->refcount++; + + node->current_size = node->vertex_size - node->attrsz[0]; + node->current_data = NULL; + + if (node->current_size) { + /* If the malloc fails, we just pull the data out of the VBO + * later instead. + */ + node->current_data = MALLOC( node->current_size * sizeof(GLfloat) ); + if (node->current_data) { + const char *buffer = (const char *)save->vertex_store->buffer; + unsigned attr_offset = node->attrsz[0] * sizeof(GLfloat); + unsigned vertex_offset = 0; + + if (node->count) + vertex_offset = (node->count-1) * node->vertex_size * sizeof(GLfloat); + + memcpy( node->current_data, + buffer + node->buffer_offset + vertex_offset + attr_offset, + node->current_size * sizeof(GLfloat) ); + } + } + + + assert(node->attrsz[VBO_ATTRIB_POS] != 0 || node->count == 0); diff --git a/src/mesa/vbo/vbo_save_draw.c b/src/mesa/vbo/vbo_save_draw.c index f59e1036d01..5110648c282 100644 --- a/src/mesa/vbo/vbo_save_draw.c +++ b/src/mesa/vbo/vbo_save_draw.c @@ -46,20 +46,31 @@ static void _playback_copy_to_current( GLcontext *ctx, const struct vbo_save_vertex_list *node ) { struct vbo_context *vbo = vbo_context(ctx); - GLfloat vertex[VBO_ATTRIB_MAX * 4], *data = vertex; + GLfloat vertex[VBO_ATTRIB_MAX * 4]; + GLfloat *data; GLuint i, offset; - if (node->count) - offset = (node->buffer_offset + - (node->count-1) * node->vertex_size * sizeof(GLfloat)); - else - offset = node->buffer_offset; + if (node->current_size == 0) + return; - ctx->Driver.GetBufferSubData( ctx, 0, offset, - node->vertex_size * sizeof(GLfloat), - data, node->vertex_store->bufferobj ); + if (node->current_data) { + data = node->current_data; + } + else { + data = vertex; + + if (node->count) + offset = (node->buffer_offset + + (node->count-1) * node->vertex_size * sizeof(GLfloat)); + else + offset = node->buffer_offset; - data += node->attrsz[0]; /* skip vertex position */ + ctx->Driver.GetBufferSubData( ctx, 0, offset, + node->vertex_size * sizeof(GLfloat), + data, node->vertex_store->bufferobj ); + + data += node->attrsz[0]; /* skip vertex position */ + } for (i = VBO_ATTRIB_POS+1 ; i < VBO_ATTRIB_MAX ; i++) { if (node->attrsz[i]) { |