diff options
Diffstat (limited to 'src/mesa/drivers')
97 files changed, 4415 insertions, 2175 deletions
diff --git a/src/mesa/drivers/common/driverfuncs.c b/src/mesa/drivers/common/driverfuncs.c index 4ca0e7bcc37..9b271f85e92 100644 --- a/src/mesa/drivers/common/driverfuncs.c +++ b/src/mesa/drivers/common/driverfuncs.c @@ -124,7 +124,6 @@ _mesa_init_driver_functions(struct dd_function_table *driver) driver->UnmapTexture = NULL; driver->TextureMemCpy = _mesa_memcpy; driver->IsTextureResident = NULL; - driver->ActiveTexture = NULL; driver->UpdateTexturePalette = NULL; /* imaging */ diff --git a/src/mesa/drivers/dri/ffb/ffb_vbtmp.h b/src/mesa/drivers/dri/ffb/ffb_vbtmp.h index 0495d0e2767..c548ef3ad58 100644 --- a/src/mesa/drivers/dri/ffb/ffb_vbtmp.h +++ b/src/mesa/drivers/dri/ffb/ffb_vbtmp.h @@ -38,11 +38,11 @@ static void TAG(emit)(GLcontext *ctx, GLuint start, GLuint end) #endif #if (IND & (FFB_VB_RGBA_BIT)) - col0 = VB->ColorPtr[0]->data; - col0_stride = VB->ColorPtr[0]->stride; + col0 = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col0_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; #if (IND & (FFB_VB_TWOSIDE_BIT)) - col1 = VB->ColorPtr[1]->data; - col1_stride = VB->ColorPtr[1]->stride; + col1 = VB->BackfaceColorPtr->data; + col1_stride = VB->BackfaceColorPtr->stride; #endif #endif diff --git a/src/mesa/drivers/dri/gamma/gamma_render.c b/src/mesa/drivers/dri/gamma/gamma_render.c index 1b9fd169f46..a03a93d132c 100644 --- a/src/mesa/drivers/dri/gamma/gamma_render.c +++ b/src/mesa/drivers/dri/gamma/gamma_render.c @@ -53,13 +53,13 @@ static void gamma_emit( GLcontext *ctx, GLuint start, GLuint end) GLfloat (*tc0)[4] = 0; GLuint tc0_size = 0; - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; if (ctx->Texture.Unit[0]._ReallyEnabled) { - tc0_stride = VB->TexCoordPtr[0]->stride; - tc0 = VB->TexCoordPtr[0]->data; - tc0_size = VB->TexCoordPtr[0]->size; + tc0_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0]->stride; + tc0 = VB->AttribPtr[_TNL_ATTRIB_TEX0]->data; + tc0_size = VB->AttribPtr[_TNL_ATTRIB_TEX0]->size; coord = VB->ClipPtr->data; coord_stride = VB->ClipPtr->stride; } else { diff --git a/src/mesa/drivers/dri/i915/i830_texstate.c b/src/mesa/drivers/dri/i915/i830_texstate.c index ce409b3a60c..27c5aa1e085 100644 --- a/src/mesa/drivers/dri/i915/i830_texstate.c +++ b/src/mesa/drivers/dri/i915/i830_texstate.c @@ -27,6 +27,7 @@ #include "main/mtypes.h" #include "main/enums.h" +#include "main/colormac.h" #include "intel_mipmap_tree.h" #include "intel_tex.h" @@ -308,11 +309,10 @@ i830_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3) CLAMPED_FLOAT_TO_UBYTE(border[2], tObj->BorderColor[2]); CLAMPED_FLOAT_TO_UBYTE(border[3], tObj->BorderColor[3]); - state[I830_TEXREG_TM0S4] = INTEL_PACKCOLOR8888(border[0], - border[1], - border[2], - border[3]); - + state[I830_TEXREG_TM0S4] = PACK_COLOR_8888(border[3], + border[0], + border[1], + border[2]); I830_ACTIVESTATE(i830, I830_UPLOAD_TEX(unit), GL_TRUE); /* memcmp was already disabled, but definitely won't work as the diff --git a/src/mesa/drivers/dri/i915/i830_vtbl.c b/src/mesa/drivers/dri/i915/i830_vtbl.c index e8c8d5a0486..1e3c8301d8d 100644 --- a/src/mesa/drivers/dri/i915/i830_vtbl.c +++ b/src/mesa/drivers/dri/i915/i830_vtbl.c @@ -126,7 +126,7 @@ i830_render_start(struct intel_context *intel) for (i = 0; i < I830_TEX_UNITS; i++) { if (RENDERINPUTS_TEST(index_bitset, _TNL_ATTRIB_TEX(i))) { - GLuint sz = VB->TexCoordPtr[i]->size; + GLuint sz = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->size; GLuint emit; GLuint mcs = (i830->state.Tex[i][I830_TEXREG_MCS] & ~TEXCOORDTYPE_MASK); @@ -714,9 +714,6 @@ i830_new_batch(struct intel_context *intel) { struct i830_context *i830 = i830_context(&intel->ctx); i830->state.emitted = 0; - - /* Check that we didn't just wrap our batchbuffer at a bad time. */ - assert(!intel->no_batch_wrap); } static void diff --git a/src/mesa/drivers/dri/i915/i915_context.c b/src/mesa/drivers/dri/i915/i915_context.c index 7d4c7cfbabb..0485be2cc1f 100644 --- a/src/mesa/drivers/dri/i915/i915_context.c +++ b/src/mesa/drivers/dri/i915/i915_context.c @@ -143,6 +143,9 @@ i915CreateContext(const __GLcontextModes * mesaVis, ctx->Const.MaxTextureImageUnits = I915_TEX_UNITS; ctx->Const.MaxTextureCoordUnits = I915_TEX_UNITS; ctx->Const.MaxVarying = I915_TEX_UNITS; + ctx->Const.MaxCombinedTextureImageUnits = + ctx->Const.MaxVertexTextureImageUnits + + ctx->Const.MaxTextureImageUnits; /* Advertise the full hardware capabilities. The new memory * manager should cope much better with overload situations: diff --git a/src/mesa/drivers/dri/i915/i915_fragprog.c b/src/mesa/drivers/dri/i915/i915_fragprog.c index d9c61446f52..9e4d3180360 100644 --- a/src/mesa/drivers/dri/i915/i915_fragprog.c +++ b/src/mesa/drivers/dri/i915/i915_fragprog.c @@ -1301,7 +1301,7 @@ i915ValidateFragmentProgram(struct i915_context *i915) for (i = 0; i < p->ctx->Const.MaxTextureCoordUnits; i++) { if (inputsRead & FRAG_BIT_TEX(i)) { - int sz = VB->TexCoordPtr[i]->size; + int sz = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->size; s2 &= ~S2_TEXCOORD_FMT(i, S2_TEXCOORD_FMT0_MASK); s2 |= S2_TEXCOORD_FMT(i, SZ_TO_HW(sz)); diff --git a/src/mesa/drivers/dri/i915/i915_texstate.c b/src/mesa/drivers/dri/i915/i915_texstate.c index de258484d6c..221bf033327 100644 --- a/src/mesa/drivers/dri/i915/i915_texstate.c +++ b/src/mesa/drivers/dri/i915/i915_texstate.c @@ -28,6 +28,7 @@ #include "main/mtypes.h" #include "main/enums.h" #include "main/macros.h" +#include "main/colormac.h" #include "intel_mipmap_tree.h" #include "intel_tex.h" @@ -356,15 +357,15 @@ i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3) * R channel, while the hardware uses A. Spam R into all the channels * for safety. */ - state[I915_TEXREG_SS4] = INTEL_PACKCOLOR8888(border[0], - border[0], - border[0], - border[0]); + state[I915_TEXREG_SS4] = PACK_COLOR_8888(border[0], + border[0], + border[0], + border[0]); } else { - state[I915_TEXREG_SS4] = INTEL_PACKCOLOR8888(border[0], - border[1], - border[2], - border[3]); + state[I915_TEXREG_SS4] = PACK_COLOR_8888(border[3], + border[0], + border[1], + border[2]); } diff --git a/src/mesa/drivers/dri/i915/i915_vtbl.c b/src/mesa/drivers/dri/i915/i915_vtbl.c index ff97e5a944f..ba6be9796e1 100644 --- a/src/mesa/drivers/dri/i915/i915_vtbl.c +++ b/src/mesa/drivers/dri/i915/i915_vtbl.c @@ -667,9 +667,6 @@ i915_new_batch(struct intel_context *intel) * difficulties associated with them (physical address requirements). */ i915->state.emitted = 0; - - /* Check that we didn't just wrap our batchbuffer at a bad time. */ - assert(!intel->no_batch_wrap); } static void diff --git a/src/mesa/drivers/dri/i915/intel_tris.c b/src/mesa/drivers/dri/i915/intel_tris.c index bc527aae47a..8a3ab39bc28 100644 --- a/src/mesa/drivers/dri/i915/intel_tris.c +++ b/src/mesa/drivers/dri/i915/intel_tris.c @@ -1250,81 +1250,6 @@ union fi GLint i; }; - -/**********************************************************************/ -/* Used only with the metaops callbacks. */ -/**********************************************************************/ -static void -intel_meta_draw_poly(struct intel_context *intel, - GLuint n, - GLfloat xy[][2], - GLfloat z, GLuint color, GLfloat tex[][2]) -{ - union fi *vb; - GLint i; - unsigned int saved_vertex_size = intel->vertex_size; - - LOCK_HARDWARE(intel); - - intel->vertex_size = 6; - - /* All 3d primitives should be emitted with LOOP_CLIPRECTS, - * otherwise the drawing origin (DR4) might not be set correctly. - */ - intel_set_prim(intel, PRIM3D_TRIFAN); - vb = (union fi *) intel_get_prim_space(intel, n); - - for (i = 0; i < n; i++) { - vb[0].f = xy[i][0]; - vb[1].f = xy[i][1]; - vb[2].f = z; - vb[3].i = color; - vb[4].f = tex[i][0]; - vb[5].f = tex[i][1]; - vb += 6; - } - - INTEL_FIREVERTICES(intel); - - intel->vertex_size = saved_vertex_size; - - UNLOCK_HARDWARE(intel); -} - -static void -intel_meta_draw_quad(struct intel_context *intel, - GLfloat x0, GLfloat x1, - GLfloat y0, GLfloat y1, - GLfloat z, - GLuint color, - GLfloat s0, GLfloat s1, GLfloat t0, GLfloat t1) -{ - GLfloat xy[4][2]; - GLfloat tex[4][2]; - - xy[0][0] = x0; - xy[0][1] = y0; - xy[1][0] = x1; - xy[1][1] = y0; - xy[2][0] = x1; - xy[2][1] = y1; - xy[3][0] = x0; - xy[3][1] = y1; - - tex[0][0] = s0; - tex[0][1] = t0; - tex[1][0] = s1; - tex[1][1] = t0; - tex[2][0] = s1; - tex[2][1] = t1; - tex[3][0] = s0; - tex[3][1] = t1; - - intel_meta_draw_poly(intel, 4, xy, z, color, tex); -} - - - /**********************************************************************/ /* Initialization. */ /**********************************************************************/ @@ -1333,7 +1258,6 @@ intel_meta_draw_quad(struct intel_context *intel, void intelInitTriFuncs(GLcontext * ctx) { - struct intel_context *intel = intel_context(ctx); TNLcontext *tnl = TNL_CONTEXT(ctx); static int firsttime = 1; @@ -1350,6 +1274,4 @@ intelInitTriFuncs(GLcontext * ctx) tnl->Driver.Render.BuildVertices = _tnl_build_vertices; tnl->Driver.Render.CopyPV = _tnl_copy_pv; tnl->Driver.Render.Interp = _tnl_interp; - - intel->vtbl.meta_draw_quad = intel_meta_draw_quad; } diff --git a/src/mesa/drivers/dri/i965/brw_context.c b/src/mesa/drivers/dri/i965/brw_context.c index 48685c087b8..78bea829493 100644 --- a/src/mesa/drivers/dri/i965/brw_context.c +++ b/src/mesa/drivers/dri/i965/brw_context.c @@ -111,6 +111,9 @@ GLboolean brwCreateContext( const __GLcontextModes *mesaVis, ctx->Const.MaxTextureUnits = MIN2(ctx->Const.MaxTextureCoordUnits, ctx->Const.MaxTextureImageUnits); ctx->Const.MaxVertexTextureImageUnits = 0; /* no vertex shader textures */ + ctx->Const.MaxCombinedTextureImageUnits = + ctx->Const.MaxVertexTextureImageUnits + + ctx->Const.MaxTextureImageUnits; /* Mesa limits textures to 4kx4k; it would be nice to fix that someday */ diff --git a/src/mesa/drivers/dri/i965/brw_context.h b/src/mesa/drivers/dri/i965/brw_context.h index fded47aa2ff..e73e21433c5 100644 --- a/src/mesa/drivers/dri/i965/brw_context.h +++ b/src/mesa/drivers/dri/i965/brw_context.h @@ -172,8 +172,8 @@ struct brw_fragment_program { GLuint id; /**< serial no. to identify frag progs, never re-used */ GLboolean isGLSL; /**< really, any IF/LOOP/CONT/BREAK instructions */ - dri_bo *const_buffer; /** Program constant buffer/surface */ GLboolean use_const_buffer; + dri_bo *const_buffer; /** Program constant buffer/surface */ /** for debugging, which texture units are referenced */ GLbitfield tex_units_used; @@ -438,7 +438,6 @@ struct brw_context GLuint primitive; GLboolean emit_state_always; - GLboolean no_batch_wrap; struct { struct brw_state_flags dirty; diff --git a/src/mesa/drivers/dri/i965/brw_draw.c b/src/mesa/drivers/dri/i965/brw_draw.c index 8bcb6083f71..7ad860898fc 100644 --- a/src/mesa/drivers/dri/i965/brw_draw.c +++ b/src/mesa/drivers/dri/i965/brw_draw.c @@ -145,7 +145,7 @@ static void brw_emit_prim(struct brw_context *brw, prim_packet.base_vert_location = prim->basevertex; /* Can't wrap here, since we rely on the validated state. */ - brw->no_batch_wrap = GL_TRUE; + intel->no_batch_wrap = GL_TRUE; /* If we're set to always flush, do it before and after the primitive emit. * We want to catch both missed flushes that hurt instruction/state cache @@ -163,7 +163,7 @@ static void brw_emit_prim(struct brw_context *brw, intel_batchbuffer_emit_mi_flush(intel->batch); } - brw->no_batch_wrap = GL_FALSE; + intel->no_batch_wrap = GL_FALSE; } static void brw_merge_inputs( struct brw_context *brw, diff --git a/src/mesa/drivers/dri/i965/brw_draw_upload.c b/src/mesa/drivers/dri/i965/brw_draw_upload.c index 271a88dae0b..7c796dae936 100644 --- a/src/mesa/drivers/dri/i965/brw_draw_upload.c +++ b/src/mesa/drivers/dri/i965/brw_draw_upload.c @@ -243,14 +243,6 @@ static void wrap_buffers( struct brw_context *brw, dri_bo_unreference(brw->vb.upload.bo); brw->vb.upload.bo = dri_bo_alloc(brw->intel.bufmgr, "temporary VBO", size, 1); - - /* Set the internal VBO\ to no-backing-store. We only use them as a - * temporary within a brw_try_draw_prims while the lock is held. - */ - /* DON'T DO THIS AS IF WE HAVE TO RE-ORG MEMORY WE NEED SOMEWHERE WITH - FAKE TO PUSH THIS STUFF */ -// if (!brw->intel.ttm) -// dri_bo_fake_disable_backing_store(brw->vb.upload.bo, NULL, NULL); } static void get_space( struct brw_context *brw, diff --git a/src/mesa/drivers/dri/i965/brw_vtbl.c b/src/mesa/drivers/dri/i965/brw_vtbl.c index 34aaea37360..72749b3859d 100644 --- a/src/mesa/drivers/dri/i965/brw_vtbl.c +++ b/src/mesa/drivers/dri/i965/brw_vtbl.c @@ -150,9 +150,6 @@ static void brw_new_batch( struct intel_context *intel ) { struct brw_context *brw = brw_context(&intel->ctx); - /* Check that we didn't just wrap our batchbuffer at a bad time. */ - assert(!brw->no_batch_wrap); - brw->curbe.need_new_bo = GL_TRUE; /* Mark all context state as needing to be re-emitted. diff --git a/src/mesa/drivers/dri/i965/brw_wm.h b/src/mesa/drivers/dri/i965/brw_wm.h index 9dcb6e14bbe..b9b987ea70a 100644 --- a/src/mesa/drivers/dri/i965/brw_wm.h +++ b/src/mesa/drivers/dri/i965/brw_wm.h @@ -76,10 +76,10 @@ struct brw_wm_prog_key { GLushort tex_swizzles[BRW_MAX_TEX_UNIT]; - GLuint program_string_id:32; GLushort origin_x, origin_y; GLushort drawable_height; GLbitfield64 vp_outputs_written; + GLuint program_string_id:32; }; diff --git a/src/mesa/drivers/dri/i965/brw_wm_fp.c b/src/mesa/drivers/dri/i965/brw_wm_fp.c index 7d031795883..3737faf26fb 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_fp.c +++ b/src/mesa/drivers/dri/i965/brw_wm_fp.c @@ -138,7 +138,6 @@ static struct prog_dst_register dst_reg(GLuint file, GLuint idx) reg.CondMask = COND_TR; reg.CondSwizzle = 0; reg.CondSrc = 0; - reg.pad = 0; return reg; } diff --git a/src/mesa/drivers/dri/intel/intel_batchbuffer.c b/src/mesa/drivers/dri/intel/intel_batchbuffer.c index ca6e2fa5b11..2eae9b66d87 100644 --- a/src/mesa/drivers/dri/intel/intel_batchbuffer.c +++ b/src/mesa/drivers/dri/intel/intel_batchbuffer.c @@ -80,7 +80,7 @@ intel_batchbuffer_reset(struct intel_batchbuffer *batch) batch->buf = NULL; } - if (!batch->buffer && intel->ttm == GL_TRUE) + if (!batch->buffer) batch->buffer = malloc (intel->maxBatchSize); batch->buf = dri_bo_alloc(intel->bufmgr, "batchbuffer", @@ -212,7 +212,7 @@ _intel_batchbuffer_flush(struct intel_batchbuffer *batch, const char *file, batch->reserved_space = 0; /* Emit a flush if the bufmgr doesn't do it for us. */ - if (intel->always_flush_cache || !intel->ttm) { + if (intel->always_flush_cache) { intel_batchbuffer_emit_mi_flush(batch); used = batch->ptr - batch->map; } @@ -244,6 +244,9 @@ _intel_batchbuffer_flush(struct intel_batchbuffer *batch, const char *file, if (intel->vtbl.finish_batch) intel->vtbl.finish_batch(intel); + /* Check that we didn't just wrap our batchbuffer at a bad time. */ + assert(!intel->no_batch_wrap); + batch->reserved_space = BATCH_RESERVED; /* TODO: Just pass the relocation list and dma buffer up to the diff --git a/src/mesa/drivers/dri/intel/intel_blit.c b/src/mesa/drivers/dri/intel/intel_blit.c index 9f638b0ef98..cdf1408cd33 100644 --- a/src/mesa/drivers/dri/intel/intel_blit.c +++ b/src/mesa/drivers/dri/intel/intel_blit.c @@ -499,10 +499,11 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask) switch (irb->Base.Format) { case MESA_FORMAT_ARGB8888: case MESA_FORMAT_XRGB8888: - clearVal = intel->ClearColor8888; + clearVal = PACK_COLOR_8888(clear[3], clear[0], + clear[1], clear[2]); break; case MESA_FORMAT_RGB565: - clearVal = intel->ClearColor565; + clearVal = PACK_COLOR_565(clear[0], clear[1], clear[2]); break; case MESA_FORMAT_ARGB4444: clearVal = PACK_COLOR_4444(clear[3], clear[0], diff --git a/src/mesa/drivers/dri/intel/intel_context.c b/src/mesa/drivers/dri/intel/intel_context.c index 2aeca6b81b0..1434ae530be 100644 --- a/src/mesa/drivers/dri/intel/intel_context.c +++ b/src/mesa/drivers/dri/intel/intel_context.c @@ -176,9 +176,7 @@ intelGetString(GLcontext * ctx, GLenum name) break; } - (void) driGetRendererString(buffer, chipset, - (intel->ttm) ? DRIVER_DATE_GEM : DRIVER_DATE, - 0); + (void) driGetRendererString(buffer, chipset, DRIVER_DATE_GEM, 0); return (GLubyte *) buffer; default: @@ -601,6 +599,7 @@ intelInitContext(struct intel_context *intel, __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; int fthrottle_mode; + int bo_reuse_mode; if (!_mesa_initialize_context(&intel->ctx, mesaVis, shareCtx, functions, (void *) intel)) { @@ -635,18 +634,14 @@ intelInitContext(struct intel_context *intel, intel->maxBatchSize = BATCH_SZ; intel->bufmgr = intelScreen->bufmgr; - intel->ttm = intelScreen->ttm; - if (intel->ttm) { - int bo_reuse_mode; - bo_reuse_mode = driQueryOptioni(&intel->optionCache, "bo_reuse"); - switch (bo_reuse_mode) { - case DRI_CONF_BO_REUSE_DISABLED: - break; - case DRI_CONF_BO_REUSE_ALL: - intel_bufmgr_gem_enable_reuse(intel->bufmgr); - break; - } + bo_reuse_mode = driQueryOptioni(&intel->optionCache, "bo_reuse"); + switch (bo_reuse_mode) { + case DRI_CONF_BO_REUSE_DISABLED: + break; + case DRI_CONF_BO_REUSE_ALL: + intel_bufmgr_gem_enable_reuse(intel->bufmgr); + break; } /* This doesn't yet catch all non-conformant rendering, but it's a @@ -733,12 +728,6 @@ intelInitContext(struct intel_context *intel, intel->RenderIndex = ~0; fthrottle_mode = driQueryOptioni(&intel->optionCache, "fthrottle_mode"); - intel->irqsEmitted = 0; - - intel->do_irqs = (intel->intelScreen->irq_active && - fthrottle_mode == DRI_CONF_FTHROTTLE_IRQS); - - intel->do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS); if (intel->gen >= 4 && !intel->intelScreen->irq_active) { _mesa_printf("IRQs not active. Exiting\n"); @@ -1058,21 +1047,6 @@ intelContendedLock(struct intel_context *intel, GLuint flags) sarea->ctxOwner = me; } - /* If the last consumer of the texture memory wasn't us, notify the fake - * bufmgr and record the new owner. We should have the memory shared - * between contexts of a single fake bufmgr, but this will at least make - * things correct for now. - */ - if (!intel->ttm && sarea->texAge != intel->hHWContext) { - sarea->texAge = intel->hHWContext; - intel_bufmgr_fake_contended_lock_take(intel->bufmgr); - if (INTEL_DEBUG & DEBUG_BATCH) - intel_decode_context_reset(); - if (INTEL_DEBUG & DEBUG_BUFMGR) - fprintf(stderr, "Lost Textures: sarea->texAge %x hw context %x\n", - sarea->ctxOwner, intel->hHWContext); - } - /* Drawable changed? */ if (dPriv && intel->lastStamp != dPriv->lastStamp) { diff --git a/src/mesa/drivers/dri/intel/intel_context.h b/src/mesa/drivers/dri/intel/intel_context.h index eb7be7ddd0f..481202c971b 100644 --- a/src/mesa/drivers/dri/intel/intel_context.h +++ b/src/mesa/drivers/dri/intel/intel_context.h @@ -135,14 +135,6 @@ struct intel_context struct intel_region * draw_region, struct intel_region * depth_region); - void (*meta_draw_quad)(struct intel_context *intel, - GLfloat x0, GLfloat x1, - GLfloat y0, GLfloat y1, - GLfloat z, - GLuint color, /* ARGB32 */ - GLfloat s0, GLfloat s1, - GLfloat t0, GLfloat t1); - void (*meta_color_mask) (struct intel_context * intel, GLboolean); void (*meta_stencil_replace) (struct intel_context * intel, @@ -189,12 +181,6 @@ struct intel_context struct intel_region *back_region; struct intel_region *depth_region; - /** - * This value indicates that the kernel memory manager is being used - * instead of the fake client-side memory manager. - */ - GLboolean ttm; - struct intel_batchbuffer *batch; drm_intel_bo *first_post_swapbuffers_batch; GLboolean no_batch_wrap; @@ -217,10 +203,6 @@ struct intel_context char *prevLockFile; int prevLockLine; - GLuint ClearColor565; - GLuint ClearColor8888; - - /* Offsets of fields within the current vertex: */ GLuint coloroffset; @@ -237,6 +219,7 @@ struct intel_context GLboolean hw_stipple; GLboolean depth_buffer_is_float; GLboolean no_rast; + GLboolean no_hw; GLboolean always_flush_batch; GLboolean always_flush_cache; @@ -302,13 +285,6 @@ struct intel_context GLboolean use_early_z; drm_clip_rect_t fboRect; /**< cliprect for FBO rendering */ - int perf_boxes; - - GLuint do_usleeps; - int do_irqs; - GLuint irqsEmitted; - - GLboolean scissor; drm_clip_rect_t draw_rect; drm_clip_rect_t scissor_rect; @@ -325,8 +301,6 @@ struct intel_context GLuint lastStamp; - GLboolean no_hw; - /** * Configuration cache */ @@ -374,29 +348,6 @@ do { \ } while (0) /* ================================================================ - * Color packing: - */ - -#define INTEL_PACKCOLOR4444(r,g,b,a) \ - ((((a) & 0xf0) << 8) | (((r) & 0xf0) << 4) | ((g) & 0xf0) | ((b) >> 4)) - -#define INTEL_PACKCOLOR1555(r,g,b,a) \ - ((((r) & 0xf8) << 7) | (((g) & 0xf8) << 2) | (((b) & 0xf8) >> 3) | \ - ((a) ? 0x8000 : 0)) - -#define INTEL_PACKCOLOR565(r,g,b) \ - ((((r) & 0xf8) << 8) | (((g) & 0xfc) << 3) | (((b) & 0xf8) >> 3)) - -#define INTEL_PACKCOLOR8888(r,g,b,a) \ - ((a<<24) | (r<<16) | (g<<8) | b) - -#define INTEL_PACKCOLOR(format, r, g, b, a) \ -(format == DV_PF_555 ? INTEL_PACKCOLOR1555(r,g,b,a) : \ - (format == DV_PF_565 ? INTEL_PACKCOLOR565(r,g,b) : \ - (format == DV_PF_8888 ? INTEL_PACKCOLOR8888(r,g,b,a) : \ - 0))) - -/* ================================================================ * From linux kernel i386 header files, copes with odd sizes better * than COPY_DWORDS would: * XXX Put this in src/mesa/main/imports.h ??? diff --git a/src/mesa/drivers/dri/intel/intel_extensions.c b/src/mesa/drivers/dri/intel/intel_extensions.c index 48cdae509e3..86dc42cc510 100644 --- a/src/mesa/drivers/dri/intel/intel_extensions.c +++ b/src/mesa/drivers/dri/intel/intel_extensions.c @@ -79,6 +79,7 @@ static const struct dri_extension card_extensions[] = { { "GL_ARB_half_float_pixel", NULL }, { "GL_ARB_map_buffer_range", GL_ARB_map_buffer_range_functions }, { "GL_ARB_multitexture", NULL }, + { "GL_ARB_pixel_buffer_object", NULL }, { "GL_ARB_point_parameters", GL_ARB_point_parameters_functions }, { "GL_ARB_point_sprite", NULL }, { "GL_ARB_shader_objects", GL_ARB_shader_objects_functions }, @@ -104,6 +105,8 @@ static const struct dri_extension card_extensions[] = { { "GL_EXT_blend_logic_op", NULL }, { "GL_EXT_blend_subtract", NULL }, { "GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions }, + { "GL_EXT_framebuffer_blit", GL_EXT_framebuffer_blit_functions }, + { "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions }, { "GL_EXT_fog_coord", GL_EXT_fog_coord_functions }, { "GL_EXT_gpu_program_parameters", GL_EXT_gpu_program_parameters_functions }, { "GL_EXT_packed_depth_stencil", NULL }, @@ -175,14 +178,6 @@ static const struct dri_extension arb_oq_extensions[] = { { NULL, NULL } }; - -static const struct dri_extension ttm_extensions[] = { - { "GL_ARB_pixel_buffer_object", NULL }, - { "GL_EXT_framebuffer_blit", GL_EXT_framebuffer_blit_functions }, - { "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions }, - { NULL, NULL } -}; - static const struct dri_extension fragment_shader_extensions[] = { { "GL_ARB_fragment_shader", NULL }, { NULL, NULL } @@ -201,9 +196,6 @@ intelInitExtensions(GLcontext *ctx) */ driInitExtensions(ctx, card_extensions, GL_FALSE); - if (intel->ttm) - driInitExtensions(ctx, ttm_extensions, GL_FALSE); - if (IS_965(intel->intelScreen->deviceID)) driInitExtensions(ctx, brw_extensions, GL_FALSE); diff --git a/src/mesa/drivers/dri/intel/intel_mipmap_tree.c b/src/mesa/drivers/dri/intel/intel_mipmap_tree.c index abb3024bfbb..6a565f80cf4 100644 --- a/src/mesa/drivers/dri/intel/intel_mipmap_tree.c +++ b/src/mesa/drivers/dri/intel/intel_mipmap_tree.c @@ -224,16 +224,12 @@ int intel_miptree_pitch_align (struct intel_context *intel, if (!mt->compressed) { int pitch_align; - if (intel->ttm) { - /* XXX: Align pitch to multiple of 64 bytes for now to allow - * render-to-texture to work in all cases. This should probably be - * replaced at some point by some scheme to only do this when really - * necessary. - */ - pitch_align = 64; - } else { - pitch_align = 4; - } + /* XXX: Align pitch to multiple of 64 bytes for now to allow + * render-to-texture to work in all cases. This should probably be + * replaced at some point by some scheme to only do this when really + * necessary. + */ + pitch_align = 64; if (tiling == I915_TILING_X) pitch_align = 512; diff --git a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c index 204a2331737..668697cb5eb 100644 --- a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c +++ b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c @@ -228,10 +228,9 @@ do_blit_bitmap( GLcontext *ctx, UNCLAMPED_FLOAT_TO_UBYTE(ubcolor[3], tmpColor[3]); if (dst->cpp == 2) - color = INTEL_PACKCOLOR565(ubcolor[0], ubcolor[1], ubcolor[2]); + color = PACK_COLOR_565(ubcolor[0], ubcolor[1], ubcolor[2]); else - color = INTEL_PACKCOLOR8888(ubcolor[0], ubcolor[1], - ubcolor[2], ubcolor[3]); + color = PACK_COLOR_8888(ubcolor[3], ubcolor[0], ubcolor[1], ubcolor[2]); if (!intel_check_blit_fragment_ops(ctx, tmpColor[3] == 1.0F)) return GL_FALSE; diff --git a/src/mesa/drivers/dri/intel/intel_regions.c b/src/mesa/drivers/dri/intel/intel_regions.c index 80975163d47..d6b9dc44466 100644 --- a/src/mesa/drivers/dri/intel/intel_regions.c +++ b/src/mesa/drivers/dri/intel/intel_regions.c @@ -542,55 +542,18 @@ intel_recreate_static(struct intel_context *intel, region->buffer = NULL; } - if (intel->ttm) { - assert(region_desc->bo_handle != -1); - region->buffer = intel_bo_gem_create_from_name(intel->bufmgr, - name, - region_desc->bo_handle); - - ret = dri_bo_get_tiling(region->buffer, ®ion->tiling, - ®ion->bit_6_swizzle); - if (ret != 0) { - fprintf(stderr, "Couldn't get tiling of buffer %d (%s): %s\n", - region_desc->bo_handle, name, strerror(-ret)); - intel_region_release(®ion); - return NULL; - } - } else { - if (region->classic_map != NULL) { - drmUnmap(region->classic_map, - region->pitch * region->cpp * region->height); - region->classic_map = NULL; - } - ret = drmMap(intel->driFd, region_desc->handle, - region->pitch * region->cpp * region->height, - ®ion->classic_map); - if (ret != 0) { - fprintf(stderr, "Failed to drmMap %s buffer\n", name); - free(region); - return NULL; - } - - region->buffer = intel_bo_fake_alloc_static(intel->bufmgr, + assert(region_desc->bo_handle != -1); + region->buffer = intel_bo_gem_create_from_name(intel->bufmgr, name, - region_desc->offset, - region->pitch * region->cpp * - region->height, - region->classic_map); - - /* The sarea just gives us a boolean for whether it's tiled or not, - * instead of which tiling mode it is. Guess. - */ - if (region_desc->tiled) { - if (intel->gen >= 4 && region_desc == &intelScreen->depth) - region->tiling = I915_TILING_Y; - else - region->tiling = I915_TILING_X; - } else { - region->tiling = I915_TILING_NONE; - } + region_desc->bo_handle); - region->bit_6_swizzle = I915_BIT_6_SWIZZLE_NONE; + ret = dri_bo_get_tiling(region->buffer, ®ion->tiling, + ®ion->bit_6_swizzle); + if (ret != 0) { + fprintf(stderr, "Couldn't get tiling of buffer %d (%s): %s\n", + region_desc->bo_handle, name, strerror(-ret)); + intel_region_release(®ion); + return NULL; } assert(region->buffer != NULL); diff --git a/src/mesa/drivers/dri/intel/intel_screen.c b/src/mesa/drivers/dri/intel/intel_screen.c index 789135b49f9..2c5a884a9b4 100644 --- a/src/mesa/drivers/dri/intel/intel_screen.c +++ b/src/mesa/drivers/dri/intel/intel_screen.c @@ -605,7 +605,6 @@ intelFillInModes(__DRIscreenPrivate *psp, static GLboolean intel_init_bufmgr(intelScreenPrivate *intelScreen) { - GLboolean gem_disable = getenv("INTEL_NO_GEM") != NULL; int gem_kernel = 0; GLboolean gem_supported; struct drm_i915_getparam gp; @@ -622,43 +621,24 @@ intel_init_bufmgr(intelScreenPrivate *intelScreen) /* If we've got a new enough DDX that's initializing GEM and giving us * object handles for the shared buffers, use that. */ - intelScreen->ttm = GL_FALSE; if (intelScreen->driScrnPriv->dri2.enabled) gem_supported = GL_TRUE; else if (intelScreen->driScrnPriv->ddx_version.minor >= 9 && gem_kernel && intelScreen->front.bo_handle != -1) gem_supported = GL_TRUE; - else - gem_supported = GL_FALSE; - - if (!gem_disable && gem_supported) { - intelScreen->bufmgr = intel_bufmgr_gem_init(spriv->fd, BATCH_SZ); - if (intelScreen->bufmgr != NULL) - intelScreen->ttm = GL_TRUE; + else { + fprintf(stderr, "[%s:%u] Error initializing GEM.\n", + __func__, __LINE__); + return GL_FALSE; } + + intelScreen->bufmgr = intel_bufmgr_gem_init(spriv->fd, BATCH_SZ); /* Otherwise, use the classic buffer manager. */ if (intelScreen->bufmgr == NULL) { - if (gem_disable) { - _mesa_warning(NULL, "GEM disabled. Using classic."); - } else { - _mesa_warning(NULL, - "Failed to initialize GEM. Falling back to classic."); - } - - if (intelScreen->tex.size == 0) { - fprintf(stderr, "[%s:%u] Error initializing buffer manager.\n", - __func__, __LINE__); - return GL_FALSE; - } - - intelScreen->bufmgr = - intel_bufmgr_fake_init(spriv->fd, - intelScreen->tex.offset, - intelScreen->tex.map, - intelScreen->tex.size, - (unsigned int * volatile) - &intelScreen->sarea->last_dispatch); + fprintf(stderr, "[%s:%u] Error initializing buffer manager.\n", + __func__, __LINE__); + return GL_FALSE; } if (intel_get_param(spriv, I915_PARAM_NUM_FENCES_AVAIL, &num_fences)) diff --git a/src/mesa/drivers/dri/intel/intel_screen.h b/src/mesa/drivers/dri/intel/intel_screen.h index a9b9e109a6a..14ca0903b6c 100644 --- a/src/mesa/drivers/dri/intel/intel_screen.h +++ b/src/mesa/drivers/dri/intel/intel_screen.h @@ -77,7 +77,6 @@ typedef struct GLboolean no_hw; GLboolean no_vbo; - int ttm; dri_bufmgr *bufmgr; GLboolean kernel_exec_fencing; diff --git a/src/mesa/drivers/dri/intel/intel_span.c b/src/mesa/drivers/dri/intel/intel_span.c index 34c3d9df74c..d1681e9088a 100644 --- a/src/mesa/drivers/dri/intel/intel_span.c +++ b/src/mesa/drivers/dri/intel/intel_span.c @@ -613,15 +613,7 @@ intel_set_span_functions(struct intel_context *intel, struct gl_renderbuffer *rb) { struct intel_renderbuffer *irb = (struct intel_renderbuffer *) rb; - uint32_t tiling; - - /* If in GEM mode, we need to do the tile address swizzling ourselves, - * instead of the fence registers handling it. - */ - if (intel->ttm) - tiling = irb->region->tiling; - else - tiling = I915_TILING_NONE; + uint32_t tiling = irb->region->tiling; if (intel->intelScreen->kernel_exec_fencing) { switch (irb->Base.Format) { @@ -667,6 +659,9 @@ intel_set_span_functions(struct intel_context *intel, return; } + /* If in GEM mode, we need to do the tile address swizzling ourselves, + * instead of the fence registers handling it. + */ switch (irb->Base.Format) { case MESA_FORMAT_RGB565: switch (tiling) { diff --git a/src/mesa/drivers/dri/intel/intel_state.c b/src/mesa/drivers/dri/intel/intel_state.c index 4ee742377d6..aefae53eb20 100644 --- a/src/mesa/drivers/dri/intel/intel_state.c +++ b/src/mesa/drivers/dri/intel/intel_state.c @@ -196,25 +196,6 @@ intel_translate_logic_op(GLenum opcode) } } - -static void -intelClearColor(GLcontext *ctx, const GLfloat color[4]) -{ - struct intel_context *intel = intel_context(ctx); - GLubyte clear[4]; - - CLAMPED_FLOAT_TO_UBYTE(clear[0], color[0]); - CLAMPED_FLOAT_TO_UBYTE(clear[1], color[1]); - CLAMPED_FLOAT_TO_UBYTE(clear[2], color[2]); - CLAMPED_FLOAT_TO_UBYTE(clear[3], color[3]); - - /* compute both 32 and 16-bit clear values */ - intel->ClearColor8888 = INTEL_PACKCOLOR8888(clear[0], clear[1], - clear[2], clear[3]); - intel->ClearColor565 = INTEL_PACKCOLOR565(clear[0], clear[1], clear[2]); -} - - /* Fallback to swrast for select and feedback. */ static void @@ -229,5 +210,4 @@ void intelInitStateFuncs(struct dd_function_table *functions) { functions->RenderMode = intelRenderMode; - functions->ClearColor = intelClearColor; } diff --git a/src/mesa/drivers/dri/mach64/mach64_native_vb.c b/src/mesa/drivers/dri/mach64/mach64_native_vb.c index 99f1a14e171..816682ec5f1 100644 --- a/src/mesa/drivers/dri/mach64/mach64_native_vb.c +++ b/src/mesa/drivers/dri/mach64/mach64_native_vb.c @@ -207,19 +207,19 @@ INTERP_QUALIFIER void TAG(interp_extras)( GLcontext *ctx, LOCALVARS struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; - if (VB->ColorPtr[1]) { - assert(VB->ColorPtr[1]->stride == 4 * sizeof(GLfloat)); + if (VB->BackfaceColorPtr) { + assert(VB->BackfaceColorPtr->stride == 4 * sizeof(GLfloat)); INTERP_4F( t, - GET_COLOR(VB->ColorPtr[1], dst), - GET_COLOR(VB->ColorPtr[1], out), - GET_COLOR(VB->ColorPtr[1], in) ); + GET_COLOR(VB->BackfaceColorPtr, dst), + GET_COLOR(VB->BackfaceColorPtr, out), + GET_COLOR(VB->BackfaceColorPtr, in) ); - if (VB->SecondaryColorPtr[1]) { + if (VB->BackfaceSecondaryColorPtr) { INTERP_3F( t, - GET_COLOR(VB->SecondaryColorPtr[1], dst), - GET_COLOR(VB->SecondaryColorPtr[1], out), - GET_COLOR(VB->SecondaryColorPtr[1], in) ); + GET_COLOR(VB->BackfaceSecondaryColorPtr, dst), + GET_COLOR(VB->BackfaceSecondaryColorPtr, out), + GET_COLOR(VB->BackfaceSecondaryColorPtr, in) ); } } @@ -236,13 +236,13 @@ INTERP_QUALIFIER void TAG(copy_pv_extras)( GLcontext *ctx, LOCALVARS struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; - if (VB->ColorPtr[1]) { - COPY_4FV( GET_COLOR(VB->ColorPtr[1], dst), - GET_COLOR(VB->ColorPtr[1], src) ); + if (VB->BackfaceColorPtr) { + COPY_4FV( GET_COLOR(VB->BackfaceColorPtr, dst), + GET_COLOR(VB->BackfaceColorPtr, src) ); - if (VB->SecondaryColorPtr[1]) { - COPY_4FV( GET_COLOR(VB->SecondaryColorPtr[1], dst), - GET_COLOR(VB->SecondaryColorPtr[1], src) ); + if (VB->BackfaceSecondaryColorPtr) { + COPY_4FV( GET_COLOR(VB->BackfaceSecondaryColorPtr, dst), + GET_COLOR(VB->BackfaceSecondaryColorPtr, src) ); } } diff --git a/src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h b/src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h index 684f2acc890..6e5fa3520e1 100644 --- a/src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h +++ b/src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h @@ -103,10 +103,10 @@ static void TAG(emit)( GLcontext *ctx, #if DO_TEX1 { const GLuint t1 = GET_TEXSOURCE(1); - tc1 = VB->TexCoordPtr[t1]->data; - tc1_stride = VB->TexCoordPtr[t1]->stride; + tc1 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->data; + tc1_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->stride; #if DO_PTEX - tc1_size = VB->TexCoordPtr[t1]->size; + tc1_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->size; #endif } #endif @@ -114,18 +114,18 @@ static void TAG(emit)( GLcontext *ctx, #if DO_TEX0 { const GLuint t0 = GET_TEXSOURCE(0); - tc0 = VB->TexCoordPtr[t0]->data; - tc0_stride = VB->TexCoordPtr[t0]->stride; + tc0 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->data; + tc0_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->stride; #if DO_PTEX - tc0_size = VB->TexCoordPtr[t0]->size; + tc0_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->size; #endif } #endif #if DO_SPEC - if (VB->SecondaryColorPtr[0]) { - spec = VB->SecondaryColorPtr[0]->data; - spec_stride = VB->SecondaryColorPtr[0]->stride; + if (VB->AttribPtr[_TNL_ATTRIB_COLOR1]) { + spec = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->data; + spec_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->stride; } else { spec = (GLfloat (*)[4])ctx->Current.Attrib[VERT_ATTRIB_COLOR1]; spec_stride = 0; @@ -133,9 +133,9 @@ static void TAG(emit)( GLcontext *ctx, #endif #if DO_FOG - if (VB->FogCoordPtr) { - fog = VB->FogCoordPtr->data; - fog_stride = VB->FogCoordPtr->stride; + if (VB->AttribPtr[_TNL_ATTRIB_FOG]) { + fog = VB->AttribPtr[_TNL_ATTRIB_FOG]->data; + fog_stride = VB->AttribPtr[_TNL_ATTRIB_FOG]->stride; } else { static GLfloat tmp[4] = {0, 0, 0, 0}; fog = &tmp; @@ -144,8 +144,8 @@ static void TAG(emit)( GLcontext *ctx, #endif #if DO_RGBA - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; #endif coord = VB->NdcPtr->data; @@ -319,8 +319,8 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) /* Force 'missing' texcoords to something valid. */ - if (DO_TEX1 && VB->TexCoordPtr[0] == 0) - VB->TexCoordPtr[0] = VB->TexCoordPtr[1]; + if (DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0] = VB->AttribPtr[_TNL_ATTRIB_TEX1]; if (DO_PTEX) return GL_TRUE; @@ -328,12 +328,12 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) /* No hardware support for projective texture. Can fake it for * TEX0 only. */ - if ((DO_TEX1 && VB->TexCoordPtr[GET_TEXSOURCE(1)]->size == 4)) { + if ((DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(1)]->size == 4)) { PTEX_FALLBACK(); return GL_FALSE; } - if (DO_TEX0 && VB->TexCoordPtr[GET_TEXSOURCE(0)]->size == 4) { + if (DO_TEX0 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(0)]->size == 4) { if (DO_TEX1) { PTEX_FALLBACK(); } diff --git a/src/mesa/drivers/dri/mach64/mach64_tex.c b/src/mesa/drivers/dri/mach64/mach64_tex.c index a757362b11d..72917ee13bf 100644 --- a/src/mesa/drivers/dri/mach64/mach64_tex.c +++ b/src/mesa/drivers/dri/mach64/mach64_tex.c @@ -565,7 +565,6 @@ void mach64InitTextureFuncs( struct dd_function_table *functions ) functions->IsTextureResident = driIsTextureResident; functions->UpdateTexturePalette = NULL; - functions->ActiveTexture = NULL; driInitTextureFormats(); } diff --git a/src/mesa/drivers/dri/mach64/mach64_vbtmp.h b/src/mesa/drivers/dri/mach64/mach64_vbtmp.h index 938804af9ec..60bfab8f6dc 100644 --- a/src/mesa/drivers/dri/mach64/mach64_vbtmp.h +++ b/src/mesa/drivers/dri/mach64/mach64_vbtmp.h @@ -156,53 +156,53 @@ static void TAG(emit)( GLcontext *ctx, if (DO_TEX3) { const GLuint t3 = GET_TEXSOURCE(3); - tc3 = VB->TexCoordPtr[t3]->data; - tc3_stride = VB->TexCoordPtr[t3]->stride; + tc3 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t3]->data; + tc3_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t3]->stride; if (DO_PTEX) - tc3_size = VB->TexCoordPtr[t3]->size; + tc3_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t3]->size; } if (DO_TEX2) { const GLuint t2 = GET_TEXSOURCE(2); - tc2 = VB->TexCoordPtr[t2]->data; - tc2_stride = VB->TexCoordPtr[t2]->stride; + tc2 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->data; + tc2_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->stride; if (DO_PTEX) - tc2_size = VB->TexCoordPtr[t2]->size; + tc2_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->size; } if (DO_TEX1) { const GLuint t1 = GET_TEXSOURCE(1); - tc1 = VB->TexCoordPtr[t1]->data; - tc1_stride = VB->TexCoordPtr[t1]->stride; + tc1 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->data; + tc1_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->stride; if (DO_PTEX) - tc1_size = VB->TexCoordPtr[t1]->size; + tc1_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->size; } if (DO_TEX0) { const GLuint t0 = GET_TEXSOURCE(0); - tc0_stride = VB->TexCoordPtr[t0]->stride; - tc0 = VB->TexCoordPtr[t0]->data; + tc0_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->stride; + tc0 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->data; if (DO_PTEX) - tc0_size = VB->TexCoordPtr[t0]->size; + tc0_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->size; } if (DO_RGBA) { - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; } if (DO_SPEC) { - spec = VB->SecondaryColorPtr[0]->data; - spec_stride = VB->SecondaryColorPtr[0]->stride; + spec = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->data; + spec_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->stride; } else { spec = (GLfloat (*)[4])ctx->Current.Attrib[VERT_ATTRIB_COLOR1]; spec_stride = 0; } if (DO_FOG) { - if (VB->FogCoordPtr) { - fog = VB->FogCoordPtr->data; - fog_stride = VB->FogCoordPtr->stride; + if (VB->AttribPtr[_TNL_ATTRIB_FOG]) { + fog = VB->AttribPtr[_TNL_ATTRIB_FOG]->data; + fog_stride = VB->AttribPtr[_TNL_ATTRIB_FOG]->stride; } else { static GLfloat tmp[4] = {0, 0, 0, 0}; fog = &tmp; @@ -384,8 +384,8 @@ static void TAG(emit)( GLcontext *ctx, GLuint start, GLuint end, ASSERT(stride == 4); - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; /* Pack what's left into a 4-dword vertex. Color is in a different * place, and there is no 'w' coordinate. @@ -432,8 +432,8 @@ static void TAG(emit)( GLcontext *ctx, GLuint start, GLuint end, GLfloat *v = (GLfloat *)dest; int i; - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; if (start) STRIDE_4F(col, col_stride * start); @@ -473,22 +473,22 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) /* Force 'missing' texcoords to something valid. */ - if (DO_TEX3 && VB->TexCoordPtr[2] == 0) - VB->TexCoordPtr[2] = VB->TexCoordPtr[3]; + if (DO_TEX3 && VB->AttribPtr[_TNL_ATTRIB_TEX2] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX2] = VB->AttribPtr[_TNL_ATTRIB_TEX3]; - if (DO_TEX2 && VB->TexCoordPtr[1] == 0) - VB->TexCoordPtr[1] = VB->TexCoordPtr[2]; + if (DO_TEX2 && VB->AttribPtr[_TNL_ATTRIB_TEX1] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX1] = VB->AttribPtr[_TNL_ATTRIB_TEX2]; - if (DO_TEX1 && VB->TexCoordPtr[0] == 0) - VB->TexCoordPtr[0] = VB->TexCoordPtr[1]; + if (DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0] = VB->AttribPtr[_TNL_ATTRIB_TEX1]; if (DO_PTEX) return GL_TRUE; - if ((DO_TEX3 && VB->TexCoordPtr[GET_TEXSOURCE(3)]->size == 4) || - (DO_TEX2 && VB->TexCoordPtr[GET_TEXSOURCE(2)]->size == 4) || - (DO_TEX1 && VB->TexCoordPtr[GET_TEXSOURCE(1)]->size == 4) || - (DO_TEX0 && VB->TexCoordPtr[GET_TEXSOURCE(0)]->size == 4)) + if ((DO_TEX3 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(3)]->size == 4) || + (DO_TEX2 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(2)]->size == 4) || + (DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(1)]->size == 4) || + (DO_TEX0 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(0)]->size == 4)) return GL_FALSE; return GL_TRUE; @@ -501,14 +501,14 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) /* Force 'missing' texcoords to something valid. */ - if (DO_TEX3 && VB->TexCoordPtr[2] == 0) - VB->TexCoordPtr[2] = VB->TexCoordPtr[3]; + if (DO_TEX3 && VB->AttribPtr[_TNL_ATTRIB_TEX2] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX2] = VB->AttribPtr[_TNL_ATTRIB_TEX3]; - if (DO_TEX2 && VB->TexCoordPtr[1] == 0) - VB->TexCoordPtr[1] = VB->TexCoordPtr[2]; + if (DO_TEX2 && VB->AttribPtr[_TNL_ATTRIB_TEX1] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX1] = VB->AttribPtr[_TNL_ATTRIB_TEX2]; - if (DO_TEX1 && VB->TexCoordPtr[0] == 0) - VB->TexCoordPtr[0] = VB->TexCoordPtr[1]; + if (DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0] = VB->AttribPtr[_TNL_ATTRIB_TEX1]; if (DO_PTEX) return GL_TRUE; @@ -516,14 +516,14 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) /* No hardware support for projective texture. Can fake it for * TEX0 only. */ - if ((DO_TEX3 && VB->TexCoordPtr[GET_TEXSOURCE(3)]->size == 4) || - (DO_TEX2 && VB->TexCoordPtr[GET_TEXSOURCE(2)]->size == 4) || - (DO_TEX1 && VB->TexCoordPtr[GET_TEXSOURCE(1)]->size == 4)) { + if ((DO_TEX3 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(3)]->size == 4) || + (DO_TEX2 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(2)]->size == 4) || + (DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(1)]->size == 4)) { PTEX_FALLBACK(); return GL_FALSE; } - if (DO_TEX0 && VB->TexCoordPtr[GET_TEXSOURCE(0)]->size == 4) { + if (DO_TEX0 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(0)]->size == 4) { if (DO_TEX1 || DO_TEX2 || DO_TEX3) { PTEX_FALLBACK(); } diff --git a/src/mesa/drivers/dri/r128/r128_tris.c b/src/mesa/drivers/dri/r128/r128_tris.c index 82d825b8d28..86d4717b050 100644 --- a/src/mesa/drivers/dri/r128/r128_tris.c +++ b/src/mesa/drivers/dri/r128/r128_tris.c @@ -651,12 +651,12 @@ static void r128RenderStart( GLcontext *ctx ) } if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX(rmesa->tmu_source[0]) )) { - if ( VB->TexCoordPtr[rmesa->tmu_source[0]]->size > 2 ) + if ( VB->AttribPtr[_TNL_ATTRIB_TEX0 + rmesa->tmu_source[0]]->size > 2 ) fallback_projtex = GL_TRUE; EMIT_ATTR( _TNL_ATTRIB_TEX0, EMIT_2F, R128_CCE_VC_FRMT_S_T, 8 ); } if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX(rmesa->tmu_source[1]) )) { - if ( VB->TexCoordPtr[rmesa->tmu_source[1]]->size > 2 ) + if ( VB->AttribPtr[_TNL_ATTRIB_TEX0 + rmesa->tmu_source[1]]->size > 2 ) fallback_projtex = GL_TRUE; EMIT_ATTR( _TNL_ATTRIB_TEX1, EMIT_2F, R128_CCE_VC_FRMT_S2_T2, 8 ); } diff --git a/src/mesa/drivers/dri/r200/r200_maos_arrays.c b/src/mesa/drivers/dri/r200/r200_maos_arrays.c index 383a0c4b0d3..249c0bbc11d 100644 --- a/src/mesa/drivers/dri/r200/r200_maos_arrays.c +++ b/src/mesa/drivers/dri/r200/r200_maos_arrays.c @@ -90,12 +90,14 @@ static void r200_emit_vecfog(GLcontext *ctx, struct radeon_aos *aos, aos->components = size; aos->count = count; + radeon_bo_map(aos->bo, 1); out = (uint32_t*)((char*)aos->bo->ptr + aos->offset); for (i = 0; i < count; i++) { out[0] = r200ComputeFogBlendFactor( ctx, *(GLfloat *)data ); out++; data += stride; } + radeon_bo_unmap(aos->bo); } /* Emit any changed arrays to new GART memory, re-emit a packet to diff --git a/src/mesa/drivers/dri/r200/r200_swtcl.c b/src/mesa/drivers/dri/r200/r200_swtcl.c index 240fb450780..4596912ddc7 100644 --- a/src/mesa/drivers/dri/r200/r200_swtcl.c +++ b/src/mesa/drivers/dri/r200/r200_swtcl.c @@ -168,7 +168,7 @@ static void r200SetVertexFormat( GLcontext *ctx ) for (i = 0; i < ctx->Const.MaxTextureUnits; i++) { if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX(i) )) { - GLuint sz = VB->TexCoordPtr[i]->size; + GLuint sz = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->size; fmt_1 |= sz << (3 * i); EMIT_ATTR( _TNL_ATTRIB_TEX0+i, EMIT_1F + sz - 1, 0 ); @@ -297,7 +297,7 @@ void r200_swtcl_flush(GLcontext *ctx, uint32_t current_offset) radeonEmitState(&rmesa->radeon); r200EmitVertexAOS( rmesa, rmesa->radeon.swtcl.vertex_size, - first_elem(&rmesa->radeon.dma.reserved)->bo, + rmesa->radeon.swtcl.bo, current_offset); diff --git a/src/mesa/drivers/dri/r200/r200_texstate.c b/src/mesa/drivers/dri/r200/r200_texstate.c index 7782404a794..e2f9cf0ea86 100644 --- a/src/mesa/drivers/dri/r200/r200_texstate.c +++ b/src/mesa/drivers/dri/r200/r200_texstate.c @@ -797,24 +797,13 @@ void r200SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_fo return; } - radeon_update_renderbuffers(pDRICtx, dPriv); - /* back & depth buffer are useless free them right away */ - rb = (void*)rfb->base.Attachment[BUFFER_DEPTH].Renderbuffer; - if (rb && rb->bo) { - radeon_bo_unref(rb->bo); - rb->bo = NULL; - } - rb = (void*)rfb->base.Attachment[BUFFER_BACK_LEFT].Renderbuffer; - if (rb && rb->bo) { - radeon_bo_unref(rb->bo); - rb->bo = NULL; - } + radeon_update_renderbuffers(pDRICtx, dPriv, GL_TRUE); rb = rfb->color_rb[0]; if (rb->bo == NULL) { /* Failed to BO for the buffer */ return; } - + _mesa_lock_texture(radeon->glCtx, texObj); if (t->bo) { radeon_bo_unref(t->bo); diff --git a/src/mesa/drivers/dri/r300/Makefile b/src/mesa/drivers/dri/r300/Makefile index f87a6dd2ddc..be005bd1641 100644 --- a/src/mesa/drivers/dri/r300/Makefile +++ b/src/mesa/drivers/dri/r300/Makefile @@ -43,13 +43,14 @@ RADEON_COMMON_SOURCES = \ DRIVER_SOURCES = \ radeon_screen.c \ + r300_blit.c \ r300_context.c \ r300_draw.c \ - r300_ioctl.c \ r300_cmdbuf.c \ r300_state.c \ r300_render.c \ r300_tex.c \ + r300_texcopy.c \ r300_texstate.c \ r300_vertprog.c \ r300_fragprog_common.c \ diff --git a/src/mesa/drivers/dri/r300/r300_blit.c b/src/mesa/drivers/dri/r300/r300_blit.c new file mode 100644 index 00000000000..ea626d942dc --- /dev/null +++ b/src/mesa/drivers/dri/r300/r300_blit.c @@ -0,0 +1,607 @@ +/* + * Copyright (C) 2009 Maciej Cencora <[email protected]> + * + * 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, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "radeon_common.h" +#include "r300_context.h" + +#include "r300_blit.h" +#include "r300_cmdbuf.h" +#include "r300_emit.h" +#include "r300_tex.h" +#include "compiler/radeon_compiler.h" +#include "compiler/radeon_opcodes.h" + +static void vp_ins_outs(struct r300_vertex_program_compiler *c) +{ + c->code->inputs[VERT_ATTRIB_POS] = 0; + c->code->inputs[VERT_ATTRIB_TEX0] = 1; + c->code->outputs[VERT_RESULT_HPOS] = 0; + c->code->outputs[VERT_RESULT_TEX0] = 1; +} + +static void fp_allocate_hw_inputs( + struct r300_fragment_program_compiler * c, + void (*allocate)(void * data, unsigned input, unsigned hwreg), + void * mydata) +{ + allocate(mydata, FRAG_ATTRIB_TEX0, 0); +} + +static void create_vertex_program(struct r300_context *r300) +{ + struct r300_vertex_program_compiler compiler; + struct rc_instruction *inst; + + rc_init(&compiler.Base); + + inst = rc_insert_new_instruction(&compiler.Base, compiler.Base.Program.Instructions.Prev); + inst->U.I.Opcode = RC_OPCODE_MOV; + inst->U.I.DstReg.File = RC_FILE_OUTPUT; + inst->U.I.DstReg.Index = VERT_RESULT_HPOS; + inst->U.I.DstReg.RelAddr = 0; + inst->U.I.DstReg.WriteMask = RC_MASK_XYZW; + inst->U.I.SrcReg[0].Abs = 0; + inst->U.I.SrcReg[0].File = RC_FILE_INPUT; + inst->U.I.SrcReg[0].Index = VERT_ATTRIB_POS; + inst->U.I.SrcReg[0].Negate = 0; + inst->U.I.SrcReg[0].RelAddr = 0; + inst->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_XYZW; + + inst = rc_insert_new_instruction(&compiler.Base, compiler.Base.Program.Instructions.Prev); + inst->U.I.Opcode = RC_OPCODE_MOV; + inst->U.I.DstReg.File = RC_FILE_OUTPUT; + inst->U.I.DstReg.Index = VERT_RESULT_TEX0; + inst->U.I.DstReg.RelAddr = 0; + inst->U.I.DstReg.WriteMask = RC_MASK_XYZW; + inst->U.I.SrcReg[0].Abs = 0; + inst->U.I.SrcReg[0].File = RC_FILE_INPUT; + inst->U.I.SrcReg[0].Index = VERT_ATTRIB_TEX0; + inst->U.I.SrcReg[0].Negate = 0; + inst->U.I.SrcReg[0].RelAddr = 0; + inst->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_XYZW; + + compiler.Base.Program.InputsRead = (1 << VERT_ATTRIB_POS) | (1 << VERT_ATTRIB_TEX0); + compiler.RequiredOutputs = compiler.Base.Program.OutputsWritten = (1 << VERT_RESULT_HPOS) | (1 << VERT_RESULT_TEX0); + compiler.SetHwInputOutput = vp_ins_outs; + compiler.code = &r300->blit.vp_code; + + r3xx_compile_vertex_program(&compiler); +} + +static void create_fragment_program(struct r300_context *r300) +{ + struct r300_fragment_program_compiler compiler; + struct rc_instruction *inst; + + rc_init(&compiler.Base); + + inst = rc_insert_new_instruction(&compiler.Base, compiler.Base.Program.Instructions.Prev); + inst->U.I.Opcode = RC_OPCODE_TEX; + inst->U.I.TexSrcTarget = RC_TEXTURE_2D; + inst->U.I.TexSrcUnit = 0; + inst->U.I.DstReg.File = RC_FILE_OUTPUT; + inst->U.I.DstReg.Index = FRAG_RESULT_COLOR; + inst->U.I.DstReg.WriteMask = RC_MASK_XYZW; + inst->U.I.SrcReg[0].Abs = 0; + inst->U.I.SrcReg[0].File = RC_FILE_INPUT; + inst->U.I.SrcReg[0].Index = FRAG_ATTRIB_TEX0; + inst->U.I.SrcReg[0].Negate = 0; + inst->U.I.SrcReg[0].RelAddr = 0; + inst->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_XYZW; + + compiler.Base.Program.InputsRead = (1 << FRAG_ATTRIB_TEX0); + compiler.OutputColor = FRAG_RESULT_COLOR; + compiler.OutputDepth = FRAG_RESULT_DEPTH; + compiler.is_r500 = (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515); + compiler.code = &r300->blit.fp_code; + compiler.AllocateHwInputs = fp_allocate_hw_inputs; + + r3xx_compile_fragment_program(&compiler); +} + +void r300_blit_init(struct r300_context *r300) +{ + create_vertex_program(r300); + create_fragment_program(r300); +} + +static void r300_emit_tx_setup(struct r300_context *r300, + gl_format mesa_format, + struct radeon_bo *bo, + intptr_t offset, + unsigned width, + unsigned height, + unsigned pitch) +{ + BATCH_LOCALS(&r300->radeon); + + assert(width <= 2048); + assert(height <= 2048); + assert(r300TranslateTexFormat(mesa_format) >= 0); + assert(offset % 32 == 0); + + BEGIN_BATCH(17); + OUT_BATCH_REGVAL(R300_TX_FILTER0_0, + (R300_TX_CLAMP_TO_EDGE << R300_TX_WRAP_S_SHIFT) | + (R300_TX_CLAMP_TO_EDGE << R300_TX_WRAP_T_SHIFT) | + (R300_TX_CLAMP_TO_EDGE << R300_TX_WRAP_R_SHIFT) | + R300_TX_MIN_FILTER_MIP_NONE | + R300_TX_MIN_FILTER_LINEAR | + R300_TX_MAG_FILTER_LINEAR | + (0 << 28)); + OUT_BATCH_REGVAL(R300_TX_FILTER1_0, 0); + OUT_BATCH_REGVAL(R300_TX_SIZE_0, + ((width-1) << R300_TX_WIDTHMASK_SHIFT) | + ((height-1) << R300_TX_HEIGHTMASK_SHIFT) | + (0 << R300_TX_DEPTHMASK_SHIFT) | + (0 << R300_TX_MAX_MIP_LEVEL_SHIFT) | + R300_TX_SIZE_TXPITCH_EN); + + OUT_BATCH_REGVAL(R300_TX_FORMAT_0, r300TranslateTexFormat(mesa_format)); + OUT_BATCH_REGVAL(R300_TX_FORMAT2_0, pitch - 1); + OUT_BATCH_REGSEQ(R300_TX_OFFSET_0, 1); + OUT_BATCH_RELOC(0, bo, offset, RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0); + + OUT_BATCH_REGSEQ(R300_TX_INVALTAGS, 2); + OUT_BATCH(0); + OUT_BATCH(1); + + END_BATCH(); +} + +#define EASY_US_FORMAT(FMT, C0, C1, C2, C3, SIGN) \ + (FMT | R500_C0_SEL_##C0 | R500_C1_SEL_##C1 | \ + R500_C2_SEL_##C2 | R500_C3_SEL_##C3 | R500_OUT_SIGN(SIGN)) + +static uint32_t mesa_format_to_us_format(gl_format mesa_format) +{ + switch(mesa_format) + { + case MESA_FORMAT_S8_Z24: + case MESA_FORMAT_X8_Z24: + case MESA_FORMAT_RGBA8888: // x + return EASY_US_FORMAT(R500_OUT_FMT_C4_8, A, B, G, R, 0); + case MESA_FORMAT_RGB565: // x + case MESA_FORMAT_ARGB1555: // x + case MESA_FORMAT_RGBA8888_REV: // x + return EASY_US_FORMAT(R500_OUT_FMT_C4_8, R, G, B, A, 0); + case MESA_FORMAT_ARGB8888: // x + return EASY_US_FORMAT(R500_OUT_FMT_C4_8, B, G, R, A, 0); + case MESA_FORMAT_ARGB8888_REV: + return EASY_US_FORMAT(R500_OUT_FMT_C4_8, A, R, G, B, 0); + case MESA_FORMAT_XRGB8888: + return EASY_US_FORMAT(R500_OUT_FMT_C4_8, A, R, G, B, 0); + + case MESA_FORMAT_RGB332: + return EASY_US_FORMAT(R500_OUT_FMT_C_3_3_2, A, R, G, B, 0); + + case MESA_FORMAT_RGBA_FLOAT32: + return EASY_US_FORMAT(R500_OUT_FMT_C4_32_FP, R, G, B, A, 0); + case MESA_FORMAT_RGBA_FLOAT16: + return EASY_US_FORMAT(R500_OUT_FMT_C4_16_FP, R, G, B, A, 0); + case MESA_FORMAT_ALPHA_FLOAT32: + return EASY_US_FORMAT(R500_OUT_FMT_C_32_FP, A, A, A, A, 0); + case MESA_FORMAT_ALPHA_FLOAT16: + return EASY_US_FORMAT(R500_OUT_FMT_C_16_FP, A, A, A, A, 0); + + case MESA_FORMAT_SIGNED_RGBA8888: + return EASY_US_FORMAT(R500_OUT_FMT_C4_8, R, G, B, A, 0xf); + case MESA_FORMAT_SIGNED_RGBA8888_REV: + return EASY_US_FORMAT(R500_OUT_FMT_C4_8, A, B, G, R, 0xf); + case MESA_FORMAT_SIGNED_RGBA_16: + return EASY_US_FORMAT(R500_OUT_FMT_C4_16, R, G, B, A, 0xf); + + default: + assert(!"Invalid format for US output\n"); + return 0; + } +} +#undef EASY_US_FORMAT + +static void r500_emit_fp_setup(struct r300_context *r300, + struct r500_fragment_program_code *fp, + gl_format dst_format) +{ + r500_emit_fp(r300, (uint32_t *)fp->inst, (fp->inst_end + 1) * 6, 0, 0, 0); + BATCH_LOCALS(&r300->radeon); + + BEGIN_BATCH(10); + OUT_BATCH_REGSEQ(R500_US_CODE_ADDR, 3); + OUT_BATCH(R500_US_CODE_START_ADDR(0) | R500_US_CODE_END_ADDR(fp->inst_end)); + OUT_BATCH(R500_US_CODE_RANGE_ADDR(0) | R500_US_CODE_RANGE_SIZE(fp->inst_end)); + OUT_BATCH(0); + OUT_BATCH_REGVAL(R500_US_CONFIG, 0); + OUT_BATCH_REGVAL(R500_US_OUT_FMT_0, mesa_format_to_us_format(dst_format)); + OUT_BATCH_REGVAL(R500_US_PIXSIZE, fp->max_temp_idx); + END_BATCH(); +} + +static void r500_emit_rs_setup(struct r300_context *r300) +{ + BATCH_LOCALS(&r300->radeon); + + BEGIN_BATCH(7); + OUT_BATCH_REGSEQ(R300_RS_COUNT, 2); + OUT_BATCH((4 << R300_IT_COUNT_SHIFT) | R300_HIRES_EN); + OUT_BATCH(0); + OUT_BATCH_REGVAL(R500_RS_INST_0, + (0 << R500_RS_INST_TEX_ID_SHIFT) | + (0 << R500_RS_INST_TEX_ADDR_SHIFT) | + R500_RS_INST_TEX_CN_WRITE | + R500_RS_INST_COL_CN_NO_WRITE); + OUT_BATCH_REGVAL(R500_RS_IP_0, + (0 << R500_RS_IP_TEX_PTR_S_SHIFT) | + (1 << R500_RS_IP_TEX_PTR_T_SHIFT) | + (2 << R500_RS_IP_TEX_PTR_R_SHIFT) | + (3 << R500_RS_IP_TEX_PTR_Q_SHIFT)); + END_BATCH(); +} + +static void r300_emit_fp_setup(struct r300_context *r300, + struct r300_fragment_program_code *code, + gl_format dst_format) +{ + unsigned i; + BATCH_LOCALS(&r300->radeon); + + BEGIN_BATCH((code->alu.length + 1) * 4 + code->tex.length + 1 + 11); + + OUT_BATCH_REGSEQ(R300_US_ALU_RGB_INST_0, code->alu.length); + for (i = 0; i < code->alu.length; i++) { + OUT_BATCH(code->alu.inst[i].rgb_inst); + } + OUT_BATCH_REGSEQ(R300_US_ALU_RGB_ADDR_0, code->alu.length); + for (i = 0; i < code->alu.length; i++) { + OUT_BATCH(code->alu.inst[i].rgb_addr); + } + OUT_BATCH_REGSEQ(R300_US_ALU_ALPHA_INST_0, code->alu.length); + for (i = 0; i < code->alu.length; i++) { + OUT_BATCH(code->alu.inst[i].alpha_inst); + } + OUT_BATCH_REGSEQ(R300_US_ALU_ALPHA_ADDR_0, code->alu.length); + for (i = 0; i < code->alu.length; i++) { + OUT_BATCH(code->alu.inst[i].alpha_addr); + } + + OUT_BATCH_REGSEQ(R300_US_TEX_INST_0, code->tex.length); + OUT_BATCH_TABLE(code->tex.inst, code->tex.length); + + OUT_BATCH_REGSEQ(R300_US_CONFIG, 3); + OUT_BATCH(R300_PFS_CNTL_FIRST_NODE_HAS_TEX); + OUT_BATCH(code->pixsize); + OUT_BATCH(code->code_offset); + OUT_BATCH_REGSEQ(R300_US_CODE_ADDR_0, 4); + OUT_BATCH_TABLE(code->code_addr, 4); + OUT_BATCH_REGVAL(R500_US_OUT_FMT_0, mesa_format_to_us_format(dst_format)); + END_BATCH(); +} + +static void r300_emit_rs_setup(struct r300_context *r300) +{ + BATCH_LOCALS(&r300->radeon); + + BEGIN_BATCH(7); + OUT_BATCH_REGSEQ(R300_RS_COUNT, 2); + OUT_BATCH((4 << R300_IT_COUNT_SHIFT) | R300_HIRES_EN); + OUT_BATCH(0); + OUT_BATCH_REGVAL(R300_RS_INST_0, + R300_RS_INST_TEX_ID(0) | + R300_RS_INST_TEX_ADDR(0) | + R300_RS_INST_TEX_CN_WRITE); + OUT_BATCH_REGVAL(R300_RS_IP_0, + R300_RS_TEX_PTR(0) | + R300_RS_SEL_S(R300_RS_SEL_C0) | + R300_RS_SEL_T(R300_RS_SEL_C1) | + R300_RS_SEL_R(R300_RS_SEL_K0) | + R300_RS_SEL_Q(R300_RS_SEL_K1)); + END_BATCH(); +} + +static void emit_pvs_setup(struct r300_context *r300, + uint32_t *vp_code, + unsigned vp_len) +{ + BATCH_LOCALS(&r300->radeon); + + r300_emit_vpu(r300, vp_code, vp_len * 4, R300_PVS_CODE_START); + + BEGIN_BATCH(4); + OUT_BATCH_REGSEQ(R300_VAP_PVS_CODE_CNTL_0, 3); + OUT_BATCH((0 << R300_PVS_FIRST_INST_SHIFT) | + ((vp_len - 1) << R300_PVS_XYZW_VALID_INST_SHIFT) | + ((vp_len - 1)<< R300_PVS_LAST_INST_SHIFT)); + OUT_BATCH(0); + OUT_BATCH((vp_len - 1) << R300_PVS_LAST_VTX_SRC_INST_SHIFT); + END_BATCH(); +} + +static void emit_vap_setup(struct r300_context *r300) +{ + BATCH_LOCALS(&r300->radeon); + + BEGIN_BATCH(12); + OUT_BATCH_REGSEQ(R300_SE_VTE_CNTL, 2); + OUT_BATCH(R300_VTX_XY_FMT | R300_VTX_Z_FMT); + OUT_BATCH(4); + + OUT_BATCH_REGVAL(R300_VAP_PSC_SGN_NORM_CNTL, 0xaaaaaaaa); + OUT_BATCH_REGVAL(R300_VAP_PROG_STREAM_CNTL_0, + ((R300_DATA_TYPE_FLOAT_2 | (0 << R300_DST_VEC_LOC_SHIFT)) << 0) | + (((1 << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_2 | R300_LAST_VEC) << 16)); + OUT_BATCH_REGVAL(R300_VAP_PROG_STREAM_CNTL_EXT_0, + ((((R300_SWIZZLE_SELECT_X << R300_SWIZZLE_SELECT_X_SHIFT) | + (R300_SWIZZLE_SELECT_Y << R300_SWIZZLE_SELECT_Y_SHIFT) | + (R300_SWIZZLE_SELECT_FP_ZERO << R300_SWIZZLE_SELECT_Z_SHIFT) | + (R300_SWIZZLE_SELECT_FP_ONE << R300_SWIZZLE_SELECT_W_SHIFT) | + (0xf << R300_WRITE_ENA_SHIFT) ) << 0) | + (((R300_SWIZZLE_SELECT_X << R300_SWIZZLE_SELECT_X_SHIFT) | + (R300_SWIZZLE_SELECT_Y << R300_SWIZZLE_SELECT_Y_SHIFT) | + (R300_SWIZZLE_SELECT_FP_ZERO << R300_SWIZZLE_SELECT_Z_SHIFT) | + (R300_SWIZZLE_SELECT_FP_ONE << R300_SWIZZLE_SELECT_W_SHIFT) | + (0xf << R300_WRITE_ENA_SHIFT) ) << 16) ) ); + OUT_BATCH_REGSEQ(R300_VAP_OUTPUT_VTX_FMT_0, 2); + OUT_BATCH(R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT); + OUT_BATCH(R300_VAP_OUTPUT_VTX_FMT_1__4_COMPONENTS); + END_BATCH(); +} + +static GLboolean validate_buffers(struct r300_context *r300, + struct radeon_bo *src_bo, + struct radeon_bo *dst_bo) +{ + int ret; + radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs, + src_bo, RADEON_GEM_DOMAIN_VRAM, 0); + + radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs, + dst_bo, 0, RADEON_GEM_DOMAIN_VRAM); + + ret = radeon_cs_space_check_with_bo(r300->radeon.cmdbuf.cs, + first_elem(&r300->radeon.dma.reserved)->bo, + RADEON_GEM_DOMAIN_GTT, 0); + if (ret) + return GL_FALSE; + + return GL_TRUE; +} + +/** + * Calculate texcoords for given image region. + * Output values are [minx, maxx, miny, maxy] + */ +static void calc_tex_coords(float img_width, float img_height, + float x, float y, + float reg_width, float reg_height, + unsigned flip_y, float *buf) +{ + buf[0] = x / img_width; + buf[1] = buf[0] + reg_width / img_width; + buf[2] = y / img_height; + buf[3] = buf[2] + reg_height / img_height; + if (flip_y) + { + float tmp = buf[2]; + buf[2] = 1.0 - buf[3]; + buf[3] = 1.0 - tmp; + } +} + +static void emit_draw_packet(struct r300_context *r300, + unsigned src_width, unsigned src_height, + unsigned src_x_offset, unsigned src_y_offset, + unsigned dst_x_offset, unsigned dst_y_offset, + unsigned reg_width, unsigned reg_height, + unsigned flip_y) +{ + float texcoords[4]; + + calc_tex_coords(src_width, src_height, + src_x_offset, src_y_offset, + reg_width, reg_height, + flip_y, texcoords); + + float verts[] = { dst_x_offset, dst_y_offset, + texcoords[0], texcoords[3], + dst_x_offset, dst_y_offset + reg_height, + texcoords[0], texcoords[2], + dst_x_offset + reg_width, dst_y_offset + reg_height, + texcoords[1], texcoords[2], + dst_x_offset + reg_width, dst_y_offset, + texcoords[1], texcoords[3] }; + + BATCH_LOCALS(&r300->radeon); + + BEGIN_BATCH(19); + OUT_BATCH_PACKET3(R300_PACKET3_3D_DRAW_IMMD_2, 16); + OUT_BATCH(R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_EMBEDDED | + (4 << 16) | R300_VAP_VF_CNTL__PRIM_QUADS); + OUT_BATCH_TABLE(verts, 16); + END_BATCH(); +} + +static void other_stuff(struct r300_context *r300) +{ + BATCH_LOCALS(&r300->radeon); + + BEGIN_BATCH(15); + OUT_BATCH_REGVAL(R300_GA_POLY_MODE, + R300_GA_POLY_MODE_FRONT_PTYPE_TRI | R300_GA_POLY_MODE_BACK_PTYPE_TRI); + OUT_BATCH_REGVAL(R300_SU_CULL_MODE, R300_FRONT_FACE_CCW); + OUT_BATCH_REGVAL(R300_FG_FOG_BLEND, 0); + OUT_BATCH_REGVAL(R300_FG_ALPHA_FUNC, 0); + OUT_BATCH_REGSEQ(R300_RB3D_CBLEND, 2); + OUT_BATCH(0x0); + OUT_BATCH(0x0); + OUT_BATCH_REGVAL(R300_VAP_CLIP_CNTL, R300_CLIP_DISABLE); + OUT_BATCH_REGVAL(R300_ZB_CNTL, 0); + END_BATCH(); +} + +static void emit_cb_setup(struct r300_context *r300, + struct radeon_bo *bo, + intptr_t offset, + gl_format mesa_format, + unsigned pitch, + unsigned width, + unsigned height) +{ + BATCH_LOCALS(&r300->radeon); + + unsigned x1, y1, x2, y2; + x1 = 0; + y1 = 0; + x2 = width - 1; + y2 = height - 1; + + if (r300->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV515) { + x1 += R300_SCISSORS_OFFSET; + y1 += R300_SCISSORS_OFFSET; + x2 += R300_SCISSORS_OFFSET; + y2 += R300_SCISSORS_OFFSET; + } + + r300_emit_cb_setup(r300, bo, offset, mesa_format, + _mesa_get_format_bytes(mesa_format), + _mesa_format_row_stride(mesa_format, pitch)); + + BEGIN_BATCH_NO_AUTOSTATE(5); + OUT_BATCH_REGSEQ(R300_SC_SCISSORS_TL, 2); + OUT_BATCH((x1 << R300_SCISSORS_X_SHIFT)|(y1 << R300_SCISSORS_Y_SHIFT)); + OUT_BATCH((x2 << R300_SCISSORS_X_SHIFT)|(y2 << R300_SCISSORS_Y_SHIFT)); + OUT_BATCH_REGVAL(R300_RB3D_CCTL, 0); + END_BATCH(); +} + +/** + * Copy a region of [@a width x @a height] pixels from source buffer + * to destination buffer. + * @param[in] r300 r300 context + * @param[in] src_bo source radeon buffer object + * @param[in] src_offset offset of the source image in the @a src_bo + * @param[in] src_mesaformat source image format + * @param[in] src_pitch aligned source image width + * @param[in] src_width source image width + * @param[in] src_height source image height + * @param[in] src_x_offset x offset in the source image + * @param[in] src_y_offset y offset in the source image + * @param[in] dst_bo destination radeon buffer object + * @param[in] dst_offset offset of the destination image in the @a dst_bo + * @param[in] dst_mesaformat destination image format + * @param[in] dst_pitch aligned destination image width + * @param[in] dst_width destination image width + * @param[in] dst_height destination image height + * @param[in] dst_x_offset x offset in the destination image + * @param[in] dst_y_offset y offset in the destination image + * @param[in] width region width + * @param[in] height region height + * @param[in] flip_y set if y coords of the source image need to be flipped + */ +GLboolean r300_blit(struct r300_context *r300, + struct radeon_bo *src_bo, + intptr_t src_offset, + gl_format src_mesaformat, + unsigned src_pitch, + unsigned src_width, + unsigned src_height, + unsigned src_x_offset, + unsigned src_y_offset, + struct radeon_bo *dst_bo, + intptr_t dst_offset, + gl_format dst_mesaformat, + unsigned dst_pitch, + unsigned dst_width, + unsigned dst_height, + unsigned dst_x_offset, + unsigned dst_y_offset, + unsigned reg_width, + unsigned reg_height, + unsigned flip_y) +{ + /* Need to clamp the region size to make sure + * we don't read outside of the source buffer + * or write outside of the destination buffer. + */ + if (reg_width + src_x_offset > src_width) + reg_width = src_width - src_x_offset; + if (reg_height + src_y_offset > src_height) + reg_height = src_height - src_y_offset; + if (reg_width + dst_x_offset > dst_width) + reg_width = dst_width - dst_x_offset; + if (reg_height + dst_y_offset > dst_height) + reg_height = dst_height - dst_y_offset; + + if (src_bo == dst_bo) { + return GL_FALSE; + } + + if (0) { + fprintf(stderr, "src: size [%d x %d], pitch %d, " + "offset [%d x %d], format %s, bo %p\n", + src_width, src_height, src_pitch, + src_offset, src_y_offset, + _mesa_get_format_name(src_mesaformat), + src_bo); + fprintf(stderr, "dst: pitch %d, offset[%d x %d], format %s, bo %p\n", + dst_pitch, dst_x_offset, dst_y_offset, + _mesa_get_format_name(dst_mesaformat), dst_bo); + fprintf(stderr, "region: %d x %d\n", reg_width, reg_height); + } + + if (!validate_buffers(r300, src_bo, dst_bo)) + return GL_FALSE; + + rcommonEnsureCmdBufSpace(&r300->radeon, 200, __FUNCTION__); + + other_stuff(r300); + + r300_emit_tx_setup(r300, src_mesaformat, src_bo, src_offset, src_width, src_height, src_pitch); + + if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) { + r500_emit_fp_setup(r300, &r300->blit.fp_code.code.r500, dst_mesaformat); + r500_emit_rs_setup(r300); + } else { + r300_emit_fp_setup(r300, &r300->blit.fp_code.code.r300, dst_mesaformat); + r300_emit_rs_setup(r300); + } + + emit_pvs_setup(r300, r300->blit.vp_code.body.d, 2); + emit_vap_setup(r300); + + emit_cb_setup(r300, dst_bo, dst_offset, dst_mesaformat, dst_pitch, dst_width, dst_height); + + emit_draw_packet(r300, src_width, src_height, + src_x_offset, src_y_offset, + dst_x_offset, dst_y_offset, + reg_width, reg_height, + flip_y); + + r300EmitCacheFlush(r300); + + radeonFlush(r300->radeon.glCtx); + + return GL_TRUE; +}
\ No newline at end of file diff --git a/src/mesa/drivers/dri/r300/r300_blit.h b/src/mesa/drivers/dri/r300/r300_blit.h new file mode 100644 index 00000000000..dc21e880982 --- /dev/null +++ b/src/mesa/drivers/dri/r300/r300_blit.h @@ -0,0 +1,54 @@ +/* + * Copyright (C) 2009 Maciej Cencora <[email protected]> + * + * 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, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef R300_BLIT_H +#define R300_BLIT_H + +void r300_blit_init(struct r300_context *r300); + +GLboolean r300_blit(struct r300_context *r300, + struct radeon_bo *src_bo, + intptr_t src_offset, + gl_format src_mesaformat, + unsigned src_pitch, + unsigned src_width, + unsigned src_height, + unsigned src_x_offset, + unsigned src_y_offset, + struct radeon_bo *dst_bo, + intptr_t dst_offset, + gl_format dst_mesaformat, + unsigned dst_pitch, + unsigned dst_width, + unsigned dst_height, + unsigned dst_x_offset, + unsigned dst_y_offset, + unsigned width, + unsigned height, + unsigned flip_y); + +#endif // R300_BLIT_H
\ No newline at end of file diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.c b/src/mesa/drivers/dri/r300/r300_cmdbuf.c index ad8db6e68e0..e1c33bbb2cf 100644 --- a/src/mesa/drivers/dri/r300/r300_cmdbuf.c +++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.c @@ -45,7 +45,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "radeon_drm.h" #include "r300_context.h" -#include "r300_ioctl.h" #include "r300_reg.h" #include "r300_cmdbuf.h" #include "r300_emit.h" @@ -72,7 +71,7 @@ static unsigned packet0_count(r300ContextPtr r300, uint32_t *pkt) #define vpu_count(ptr) (((drm_r300_cmd_header_t*)(ptr))->vpu.count) #define r500fp_count(ptr) (((drm_r300_cmd_header_t*)(ptr))->r500fp.count) -int check_vpu(GLcontext *ctx, struct radeon_state_atom *atom) +static int check_vpu(GLcontext *ctx, struct radeon_state_atom *atom) { r300ContextPtr r300 = R300_CONTEXT(ctx); int cnt; @@ -86,54 +85,73 @@ int check_vpu(GLcontext *ctx, struct radeon_state_atom *atom) return cnt ? (cnt * 4) + extra : 0; } +void r300_emit_vpu(struct r300_context *r300, + uint32_t *data, + unsigned len, + uint32_t addr) +{ + BATCH_LOCALS(&r300->radeon); + + BEGIN_BATCH_NO_AUTOSTATE(5 + len); + OUT_BATCH_REGVAL(R300_VAP_PVS_STATE_FLUSH_REG, 0); + OUT_BATCH_REGVAL(R300_VAP_PVS_VECTOR_INDX_REG, addr); + OUT_BATCH(CP_PACKET0(R300_VAP_PVS_UPLOAD_DATA, len-1) | RADEON_ONE_REG_WR); + OUT_BATCH_TABLE(data, len); + END_BATCH(); +} -void emit_vpu(GLcontext *ctx, struct radeon_state_atom * atom) +static void emit_vpu_state(GLcontext *ctx, struct radeon_state_atom * atom) { - r300ContextPtr r300 = R300_CONTEXT(ctx); - BATCH_LOCALS(&r300->radeon); - drm_r300_cmd_header_t cmd; - uint32_t addr, ndw; + r300ContextPtr r300 = R300_CONTEXT(ctx); + drm_r300_cmd_header_t cmd; + uint32_t addr, ndw; - cmd.u = atom->cmd[0]; - addr = (cmd.vpu.adrhi << 8) | cmd.vpu.adrlo; - ndw = atom->check(ctx, atom); + cmd.u = atom->cmd[0]; + addr = (cmd.vpu.adrhi << 8) | cmd.vpu.adrlo; + ndw = atom->check(ctx, atom); - BEGIN_BATCH_NO_AUTOSTATE(ndw); + r300_emit_vpu(r300, &atom->cmd[1], vpu_count(atom->cmd) * 4, addr); +} - ndw -= 5; - OUT_BATCH_REGVAL(R300_VAP_PVS_VECTOR_INDX_REG, addr); - OUT_BATCH(CP_PACKET0(R300_VAP_PVS_UPLOAD_DATA, ndw-1) | RADEON_ONE_REG_WR); - OUT_BATCH_TABLE(&atom->cmd[1], ndw); - OUT_BATCH_REGVAL(R300_VAP_PVS_STATE_FLUSH_REG, 0); - END_BATCH(); +void r500_emit_fp(struct r300_context *r300, + uint32_t *data, + unsigned len, + uint32_t addr, + unsigned type, + unsigned clamp) +{ + BATCH_LOCALS(&r300->radeon); + + addr |= (type << 16); + addr |= (clamp << 17); + + BEGIN_BATCH_NO_AUTOSTATE(len + 3); + OUT_BATCH(CP_PACKET0(R500_GA_US_VECTOR_INDEX, 0)); + OUT_BATCH(addr); + OUT_BATCH(CP_PACKET0(R500_GA_US_VECTOR_DATA, len-1) | RADEON_ONE_REG_WR); + OUT_BATCH_TABLE(data, len); + END_BATCH(); } -void emit_r500fp(GLcontext *ctx, struct radeon_state_atom * atom) +static void emit_r500fp_atom(GLcontext *ctx, struct radeon_state_atom * atom) { - r300ContextPtr r300 = R300_CONTEXT(ctx); - BATCH_LOCALS(&r300->radeon); - drm_r300_cmd_header_t cmd; - uint32_t addr, ndw, sz; - int type, clamp; - - ndw = atom->check(ctx, atom); - - cmd.u = atom->cmd[0]; - sz = cmd.r500fp.count; - addr = ((cmd.r500fp.adrhi_flags & 1) << 8) | cmd.r500fp.adrlo; - type = !!(cmd.r500fp.adrhi_flags & R500FP_CONSTANT_TYPE); - clamp = !!(cmd.r500fp.adrhi_flags & R500FP_CONSTANT_CLAMP); - - addr |= (type << 16); - addr |= (clamp << 17); - - BEGIN_BATCH_NO_AUTOSTATE(ndw); - OUT_BATCH(CP_PACKET0(R500_GA_US_VECTOR_INDEX, 0)); - OUT_BATCH(addr); - ndw-=3; - OUT_BATCH(CP_PACKET0(R500_GA_US_VECTOR_DATA, ndw-1) | RADEON_ONE_REG_WR); - OUT_BATCH_TABLE(&atom->cmd[1], ndw); - END_BATCH(); + r300ContextPtr r300 = R300_CONTEXT(ctx); + drm_r300_cmd_header_t cmd; + uint32_t addr, count; + int type, clamp; + + cmd.u = atom->cmd[0]; + addr = ((cmd.r500fp.adrhi_flags & 1) << 8) | cmd.r500fp.adrlo; + type = !!(cmd.r500fp.adrhi_flags & R500FP_CONSTANT_TYPE); + clamp = !!(cmd.r500fp.adrhi_flags & R500FP_CONSTANT_CLAMP); + + if (type) { + count = r500fp_count(atom->cmd) * 4; + } else { + count = r500fp_count(atom->cmd) * 6; + } + + r500_emit_fp(r300, &atom->cmd[1], count, addr, type, clamp); } static int check_tex_offsets(GLcontext *ctx, struct radeon_state_atom * atom) @@ -256,110 +274,136 @@ static int check_cb_offset(GLcontext *ctx, struct radeon_state_atom * atom) return dw; } -static void emit_cb_offset(GLcontext *ctx, struct radeon_state_atom * atom) +static void emit_scissor(struct r300_context *r300, + unsigned width, + unsigned height) { - r300ContextPtr r300 = R300_CONTEXT(ctx); - BATCH_LOCALS(&r300->radeon); - struct radeon_renderbuffer *rrb; - uint32_t cbpitch; - uint32_t offset = r300->radeon.state.color.draw_offset; - uint32_t dw = 6; - int i; + int i; + BATCH_LOCALS(&r300->radeon); + if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) { + BEGIN_BATCH_NO_AUTOSTATE(3); + OUT_BATCH_REGSEQ(R300_SC_SCISSORS_TL, 2); + OUT_BATCH(0); + OUT_BATCH(((width - 1) << R300_SCISSORS_X_SHIFT) | + ((height - 1) << R300_SCISSORS_Y_SHIFT)); + END_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(16); + for (i = 0; i < 4; i++) { + OUT_BATCH_REGSEQ(R300_SC_CLIPRECT_TL_0 + (i * 8), 2); + OUT_BATCH((0 << R300_CLIPRECT_X_SHIFT) | (0 << R300_CLIPRECT_Y_SHIFT)); + OUT_BATCH(((width - 1) << R300_CLIPRECT_X_SHIFT) | ((height - 1) << R300_CLIPRECT_Y_SHIFT)); + } + OUT_BATCH_REGSEQ(R300_SC_CLIP_RULE, 1); + OUT_BATCH(0xAAAA); + OUT_BATCH_REGSEQ(R300_SC_SCREENDOOR, 1); + OUT_BATCH(0xffffff); + END_BATCH(); + } else { + BEGIN_BATCH_NO_AUTOSTATE(3); + OUT_BATCH_REGSEQ(R300_SC_SCISSORS_TL, 2); + OUT_BATCH((R300_SCISSORS_OFFSET << R300_SCISSORS_X_SHIFT) | + (R300_SCISSORS_OFFSET << R300_SCISSORS_Y_SHIFT)); + OUT_BATCH(((width + R300_SCISSORS_OFFSET - 1) << R300_SCISSORS_X_SHIFT) | + ((height + R300_SCISSORS_OFFSET - 1) << R300_SCISSORS_Y_SHIFT)); + END_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(16); + for (i = 0; i < 4; i++) { + OUT_BATCH_REGSEQ(R300_SC_CLIPRECT_TL_0 + (i * 8), 2); + OUT_BATCH((R300_SCISSORS_OFFSET << R300_CLIPRECT_X_SHIFT) | (R300_SCISSORS_OFFSET << R300_CLIPRECT_Y_SHIFT)); + OUT_BATCH(((R300_SCISSORS_OFFSET + width - 1) << R300_CLIPRECT_X_SHIFT) | + ((R300_SCISSORS_OFFSET + height - 1) << R300_CLIPRECT_Y_SHIFT)); + } + OUT_BATCH_REGSEQ(R300_SC_CLIP_RULE, 1); + OUT_BATCH(0xAAAA); + OUT_BATCH_REGSEQ(R300_SC_SCREENDOOR, 1); + OUT_BATCH(0xffffff); + END_BATCH(); + } +} - rrb = radeon_get_colorbuffer(&r300->radeon); - if (!rrb || !rrb->bo) { - fprintf(stderr, "no rrb\n"); - return; - } +void r300_emit_cb_setup(struct r300_context *r300, + struct radeon_bo *bo, + uint32_t offset, + GLuint format, + unsigned cpp, + unsigned pitch) +{ + BATCH_LOCALS(&r300->radeon); + uint32_t cbpitch = pitch / cpp; + uint32_t dw = 6; - if (RADEON_DEBUG & RADEON_STATE) - fprintf(stderr,"rrb is %p %d %dx%d\n", rrb, offset, rrb->base.Width, rrb->base.Height); - cbpitch = (rrb->pitch / rrb->cpp); - if (rrb->cpp == 4) - cbpitch |= R300_COLOR_FORMAT_ARGB8888; - else switch (rrb->base.Format) { + assert(offset % 32 == 0); + + switch (format) { case MESA_FORMAT_RGB565: - assert(_mesa_little_endian()); - cbpitch |= R300_COLOR_FORMAT_RGB565; - break; + assert(_mesa_little_endian()); + cbpitch |= R300_COLOR_FORMAT_RGB565; + break; case MESA_FORMAT_RGB565_REV: - assert(!_mesa_little_endian()); - cbpitch |= R300_COLOR_FORMAT_RGB565; - break; + assert(!_mesa_little_endian()); + cbpitch |= R300_COLOR_FORMAT_RGB565; + break; case MESA_FORMAT_ARGB4444: - assert(_mesa_little_endian()); - cbpitch |= R300_COLOR_FORMAT_ARGB4444; - break; + assert(_mesa_little_endian()); + cbpitch |= R300_COLOR_FORMAT_ARGB4444; + break; case MESA_FORMAT_ARGB4444_REV: - assert(!_mesa_little_endian()); - cbpitch |= R300_COLOR_FORMAT_ARGB4444; - break; - case MESA_FORMAT_ARGB1555: - assert(_mesa_little_endian()); - cbpitch |= R300_COLOR_FORMAT_ARGB1555; - break; - case MESA_FORMAT_ARGB1555_REV: - assert(!_mesa_little_endian()); - cbpitch |= R300_COLOR_FORMAT_ARGB1555; - break; - default: - _mesa_problem(ctx, "unexpected format in emit_cb_offset()"); - } + assert(!_mesa_little_endian()); + cbpitch |= R300_COLOR_FORMAT_ARGB4444; + break; + case MESA_FORMAT_ARGB1555: + assert(_mesa_little_endian()); + cbpitch |= R300_COLOR_FORMAT_ARGB1555; + break; + case MESA_FORMAT_ARGB1555_REV: + assert(!_mesa_little_endian()); + cbpitch |= R300_COLOR_FORMAT_ARGB1555; + break; + default: + if (cpp == 4) { + cbpitch |= R300_COLOR_FORMAT_ARGB8888; + } else { + _mesa_problem(r300->radeon.glCtx, "unexpected format in emit_cb_offset()");; + } + break; + } - if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE) - cbpitch |= R300_COLOR_TILE_ENABLE; + if (bo->flags & RADEON_BO_FLAGS_MACRO_TILE) + cbpitch |= R300_COLOR_TILE_ENABLE; + + if (r300->radeon.radeonScreen->kernel_mm) + dw += 2; + + BEGIN_BATCH_NO_AUTOSTATE(dw); + OUT_BATCH_REGSEQ(R300_RB3D_COLOROFFSET0, 1); + OUT_BATCH_RELOC(offset, bo, offset, 0, RADEON_GEM_DOMAIN_VRAM, 0); + OUT_BATCH_REGSEQ(R300_RB3D_COLORPITCH0, 1); + if (!r300->radeon.radeonScreen->kernel_mm) + OUT_BATCH(cbpitch); + else + OUT_BATCH_RELOC(cbpitch, bo, cbpitch, 0, RADEON_GEM_DOMAIN_VRAM, 0); + END_BATCH(); +} + +static void emit_cb_offset_atom(GLcontext *ctx, struct radeon_state_atom * atom) +{ + r300ContextPtr r300 = R300_CONTEXT(ctx); + struct radeon_renderbuffer *rrb; + uint32_t offset = r300->radeon.state.color.draw_offset; + + rrb = radeon_get_colorbuffer(&r300->radeon); + if (!rrb || !rrb->bo) { + fprintf(stderr, "no rrb\n"); + return; + } + + if (RADEON_DEBUG & RADEON_STATE) + fprintf(stderr,"rrb is %p %d %dx%d\n", rrb, offset, rrb->base.Width, rrb->base.Height); + + r300_emit_cb_setup(r300, rrb->bo, offset, rrb->base.Format, rrb->cpp, rrb->pitch); - if (r300->radeon.radeonScreen->kernel_mm) - dw += 2; - BEGIN_BATCH_NO_AUTOSTATE(dw); - OUT_BATCH_REGSEQ(R300_RB3D_COLOROFFSET0, 1); - OUT_BATCH_RELOC(offset, rrb->bo, offset, 0, RADEON_GEM_DOMAIN_VRAM, 0); - OUT_BATCH_REGSEQ(R300_RB3D_COLORPITCH0, 1); - if (!r300->radeon.radeonScreen->kernel_mm) - OUT_BATCH(cbpitch); - else - OUT_BATCH_RELOC(cbpitch, rrb->bo, cbpitch, 0, RADEON_GEM_DOMAIN_VRAM, 0); - END_BATCH(); if (r300->radeon.radeonScreen->driScreen->dri2.enabled) { - if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) { - BEGIN_BATCH_NO_AUTOSTATE(3); - OUT_BATCH_REGSEQ(R300_SC_SCISSORS_TL, 2); - OUT_BATCH(0); - OUT_BATCH(((rrb->base.Width - 1) << R300_SCISSORS_X_SHIFT) | - ((rrb->base.Height - 1) << R300_SCISSORS_Y_SHIFT)); - END_BATCH(); - BEGIN_BATCH_NO_AUTOSTATE(16); - for (i = 0; i < 4; i++) { - OUT_BATCH_REGSEQ(R300_SC_CLIPRECT_TL_0 + (i * 8), 2); - OUT_BATCH((0 << R300_CLIPRECT_X_SHIFT) | (0 << R300_CLIPRECT_Y_SHIFT)); - OUT_BATCH(((rrb->base.Width - 1) << R300_CLIPRECT_X_SHIFT) | ((rrb->base.Height - 1) << R300_CLIPRECT_Y_SHIFT)); - } - OUT_BATCH_REGSEQ(R300_SC_CLIP_RULE, 1); - OUT_BATCH(0xAAAA); - OUT_BATCH_REGSEQ(R300_SC_SCREENDOOR, 1); - OUT_BATCH(0xffffff); - END_BATCH(); - } else { - BEGIN_BATCH_NO_AUTOSTATE(3); - OUT_BATCH_REGSEQ(R300_SC_SCISSORS_TL, 2); - OUT_BATCH((R300_SCISSORS_OFFSET << R300_SCISSORS_X_SHIFT) | - (R300_SCISSORS_OFFSET << R300_SCISSORS_Y_SHIFT)); - OUT_BATCH(((rrb->base.Width + R300_SCISSORS_OFFSET - 1) << R300_SCISSORS_X_SHIFT) | - ((rrb->base.Height + R300_SCISSORS_OFFSET - 1) << R300_SCISSORS_Y_SHIFT)); - END_BATCH(); - BEGIN_BATCH_NO_AUTOSTATE(16); - for (i = 0; i < 4; i++) { - OUT_BATCH_REGSEQ(R300_SC_CLIPRECT_TL_0 + (i * 8), 2); - OUT_BATCH((R300_SCISSORS_OFFSET << R300_CLIPRECT_X_SHIFT) | (R300_SCISSORS_OFFSET << R300_CLIPRECT_Y_SHIFT)); - OUT_BATCH(((R300_SCISSORS_OFFSET + rrb->base.Width - 1) << R300_CLIPRECT_X_SHIFT) | - ((R300_SCISSORS_OFFSET + rrb->base.Height - 1) << R300_CLIPRECT_Y_SHIFT)); - } - OUT_BATCH_REGSEQ(R300_SC_CLIP_RULE, 1); - OUT_BATCH(0xAAAA); - OUT_BATCH_REGSEQ(R300_SC_SCREENDOOR, 1); - OUT_BATCH(0xffffff); - END_BATCH(); - } + emit_scissor(r300, rrb->base.Width, rrb->base.Height); } } @@ -455,7 +499,7 @@ static int check_variable(GLcontext *ctx, struct radeon_state_atom *atom) return cnt ? cnt + 1 : 0; } -int check_r500fp(GLcontext *ctx, struct radeon_state_atom *atom) +static int check_r500fp(GLcontext *ctx, struct radeon_state_atom *atom) { int cnt; r300ContextPtr r300 = R300_CONTEXT(ctx); @@ -467,7 +511,7 @@ int check_r500fp(GLcontext *ctx, struct radeon_state_atom *atom) return cnt ? (cnt * 6) + extra : 0; } -int check_r500fp_const(GLcontext *ctx, struct radeon_state_atom *atom) +static int check_r500fp_const(GLcontext *ctx, struct radeon_state_atom *atom) { int cnt; r300ContextPtr r300 = R300_CONTEXT(ctx); @@ -644,13 +688,13 @@ void r300InitCmdBuf(r300ContextPtr r300) r300->hw.r500fp.cmd[R300_FPI_CMD_0] = cmdr500fp(r300->radeon.radeonScreen, 0, 0, 0, 0); if (r300->radeon.radeonScreen->kernel_mm) - r300->hw.r500fp.emit = emit_r500fp; + r300->hw.r500fp.emit = emit_r500fp_atom; ALLOC_STATE(r500fp_const, r500fp_const, R500_FPP_CMDSIZE, 0); r300->hw.r500fp_const.cmd[R300_FPI_CMD_0] = cmdr500fp(r300->radeon.radeonScreen, 0, 0, 1, 0); if (r300->radeon.radeonScreen->kernel_mm) - r300->hw.r500fp_const.emit = emit_r500fp; + r300->hw.r500fp_const.emit = emit_r500fp_atom; } else { ALLOC_STATE(fp, always, R300_FP_CMDSIZE, 0); r300->hw.fp.cmd[R300_FP_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_US_CONFIG, 3); @@ -694,7 +738,7 @@ void r300InitCmdBuf(r300ContextPtr r300) ALLOC_STATE(rop, always, 2, 0); r300->hw.rop.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_RB3D_ROPCNTL, 1); ALLOC_STATE(cb, cb_offset, R300_CB_CMDSIZE, 0); - r300->hw.cb.emit = &emit_cb_offset; + r300->hw.cb.emit = &emit_cb_offset_atom; ALLOC_STATE(rb3d_dither_ctl, always, 10, 0); r300->hw.rb3d_dither_ctl.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_RB3D_DITHER_CTL, 9); ALLOC_STATE(rb3d_aaresolve_ctl, always, 2, 0); @@ -758,20 +802,20 @@ void r300InitCmdBuf(r300ContextPtr r300) r300->hw.vpi.cmd[0] = cmdvpu(r300->radeon.radeonScreen, R300_PVS_CODE_START, 0); if (r300->radeon.radeonScreen->kernel_mm) - r300->hw.vpi.emit = emit_vpu; + r300->hw.vpi.emit = emit_vpu_state; if (is_r500) { ALLOC_STATE(vpp, vpu, R300_VPP_CMDSIZE, 0); r300->hw.vpp.cmd[0] = cmdvpu(r300->radeon.radeonScreen, R500_PVS_CONST_START, 0); if (r300->radeon.radeonScreen->kernel_mm) - r300->hw.vpp.emit = emit_vpu; + r300->hw.vpp.emit = emit_vpu_state; ALLOC_STATE(vps, vpu, R300_VPS_CMDSIZE, 0); r300->hw.vps.cmd[0] = cmdvpu(r300->radeon.radeonScreen, R500_POINT_VPORT_SCALE_OFFSET, 1); if (r300->radeon.radeonScreen->kernel_mm) - r300->hw.vps.emit = emit_vpu; + r300->hw.vps.emit = emit_vpu_state; for (i = 0; i < 6; i++) { ALLOC_STATE(vpucp[i], vpu, R300_VPUCP_CMDSIZE, 0); @@ -779,20 +823,20 @@ void r300InitCmdBuf(r300ContextPtr r300) cmdvpu(r300->radeon.radeonScreen, R500_PVS_UCP_START + i, 1); if (r300->radeon.radeonScreen->kernel_mm) - r300->hw.vpucp[i].emit = emit_vpu; + r300->hw.vpucp[i].emit = emit_vpu_state; } } else { ALLOC_STATE(vpp, vpu, R300_VPP_CMDSIZE, 0); r300->hw.vpp.cmd[0] = cmdvpu(r300->radeon.radeonScreen, R300_PVS_CONST_START, 0); if (r300->radeon.radeonScreen->kernel_mm) - r300->hw.vpp.emit = emit_vpu; + r300->hw.vpp.emit = emit_vpu_state; ALLOC_STATE(vps, vpu, R300_VPS_CMDSIZE, 0); r300->hw.vps.cmd[0] = cmdvpu(r300->radeon.radeonScreen, R300_POINT_VPORT_SCALE_OFFSET, 1); if (r300->radeon.radeonScreen->kernel_mm) - r300->hw.vps.emit = emit_vpu; + r300->hw.vps.emit = emit_vpu_state; for (i = 0; i < 6; i++) { ALLOC_STATE(vpucp[i], vpu, R300_VPUCP_CMDSIZE, 0); @@ -800,7 +844,7 @@ void r300InitCmdBuf(r300ContextPtr r300) cmdvpu(r300->radeon.radeonScreen, R300_PVS_UCP_START + i, 1); if (r300->radeon.radeonScreen->kernel_mm) - r300->hw.vpucp[i].emit = emit_vpu; + r300->hw.vpucp[i].emit = emit_vpu_state; } } } diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.h b/src/mesa/drivers/dri/r300/r300_cmdbuf.h index 1b703e518a0..0e68da928ed 100644 --- a/src/mesa/drivers/dri/r300/r300_cmdbuf.h +++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.h @@ -44,14 +44,26 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define FIREAOS_BUFSZ (3) #define SCISSORS_BUFSZ (3) -extern void r300InitCmdBuf(r300ContextPtr r300); +void r300InitCmdBuf(r300ContextPtr r300); void r300_emit_scissor(GLcontext *ctx); -void emit_vpu(GLcontext *ctx, struct radeon_state_atom * atom); -int check_vpu(GLcontext *ctx, struct radeon_state_atom *atom); +void r300_emit_vpu(struct r300_context *ctx, + uint32_t *data, + unsigned len, + uint32_t addr); -void emit_r500fp(GLcontext *ctx, struct radeon_state_atom * atom); -int check_r500fp(GLcontext *ctx, struct radeon_state_atom *atom); -int check_r500fp_const(GLcontext *ctx, struct radeon_state_atom *atom); +void r500_emit_fp(struct r300_context *r300, + uint32_t *data, + unsigned len, + uint32_t addr, + unsigned type, + unsigned clamp); -#endif /* __R300_CMDBUF_H__ */ +void r300_emit_cb_setup(struct r300_context *r300, + struct radeon_bo *bo, + uint32_t offset, + GLuint format, + unsigned cpp, + unsigned pitch); + +#endif /* __R300_CMDBUF_H__ */ diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c index 5f07b956349..3c6ec2a34a8 100644 --- a/src/mesa/drivers/dri/r300/r300_context.c +++ b/src/mesa/drivers/dri/r300/r300_context.c @@ -55,13 +55,14 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "tnl/t_vp_build.h" #include "drivers/common/driverfuncs.h" +#include "drivers/common/meta.h" #include "r300_context.h" #include "radeon_context.h" #include "radeon_span.h" +#include "r300_blit.h" #include "r300_cmdbuf.h" #include "r300_state.h" -#include "r300_ioctl.h" #include "r300_tex.h" #include "r300_emit.h" #include "r300_swtcl.h" @@ -92,6 +93,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/remap_helper.h" +void r300_init_texcopy_functions(struct dd_function_table *table); static const struct dri_extension card_extensions[] = { /* *INDENT-OFF* */ @@ -451,6 +453,13 @@ static void r300InitGLExtensions(GLcontext *ctx) } } +static void r300InitIoctlFuncs(struct dd_function_table *functions) +{ + functions->Clear = _mesa_meta_Clear; + functions->Finish = radeonFinish; + functions->Flush = radeonFlush; +} + /* Create the device specific rendering context. */ GLboolean r300CreateContext(const __GLcontextModes * glVisual, @@ -484,6 +493,10 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual, radeonInitQueryObjFunctions(&functions); radeonInitBufferObjectFuncs(&functions); + if (r300->radeon.radeonScreen->kernel_mm) { + r300_init_texcopy_functions(&functions); + } + if (!radeonInitContext(&r300->radeon, &functions, glVisual, driContextPriv, sharedContextPrivate)) { @@ -530,6 +543,7 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual, r300InitSwtcl(ctx); } + r300_blit_init(r300); radeon_fbo_init(&r300->radeon); radeonInitSpanFuncs( ctx ); r300InitCmdBuf(r300); diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h index 518d5cdbf4f..54a92a2e447 100644 --- a/src/mesa/drivers/dri/r300/r300_context.h +++ b/src/mesa/drivers/dri/r300/r300_context.h @@ -533,6 +533,11 @@ struct r300_context { uint32_t fallback; + struct { + struct r300_vertex_program_code vp_code; + struct rX00_fragment_program_code fp_code; + } blit; + DECLARE_RENDERINPUTS(render_inputs_bitset); }; @@ -549,6 +554,8 @@ extern void r300InitShaderFunctions(r300ContextPtr r300); extern void r300InitDraw(GLcontext *ctx); +extern void r300_init_texcopy_functions(struct dd_function_table *table); + #define r300PackFloat32 radeonPackFloat32 #define r300PackFloat24 radeonPackFloat24 diff --git a/src/mesa/drivers/dri/r300/r300_draw.c b/src/mesa/drivers/dri/r300/r300_draw.c index e9968f9ffef..3dcd986e226 100644 --- a/src/mesa/drivers/dri/r300/r300_draw.c +++ b/src/mesa/drivers/dri/r300/r300_draw.c @@ -100,7 +100,7 @@ static void r300FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer GLubyte *in = (GLubyte *)src_ptr; radeonAllocDmaRegion(&r300->radeon, &r300->ind_buf.bo, &r300->ind_buf.bo_offset, size, 4); - + radeon_bo_map(r300->ind_buf.bo, 1); assert(r300->ind_buf.bo->ptr != NULL); out = (GLuint *)ADD_POINTERS(r300->ind_buf.bo->ptr, r300->ind_buf.bo_offset); @@ -111,7 +111,7 @@ static void r300FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer if (i < mesa_ind_buf->count) { *out++ = in[i]; } - + radeon_bo_unmap(r300->ind_buf.bo); #if MESA_BIG_ENDIAN } else { /* if (mesa_ind_buf->type == GL_UNSIGNED_SHORT) */ GLushort *in = (GLushort *)src_ptr; @@ -120,6 +120,7 @@ static void r300FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer radeonAllocDmaRegion(&r300->radeon, &r300->ind_buf.bo, &r300->ind_buf.bo_offset, size, 4); + radeon_bo_map(r300->ind_buf.bo, 1); assert(r300->ind_buf.bo->ptr != NULL); out = (GLuint *)ADD_POINTERS(r300->ind_buf.bo->ptr, r300->ind_buf.bo_offset); @@ -130,6 +131,7 @@ static void r300FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer if (i < mesa_ind_buf->count) { *out++ = in[i]; } + radeon_bo_unmap(r300->ind_buf.bo); #endif } @@ -173,10 +175,12 @@ static void r300SetupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer radeonAllocDmaRegion(&r300->radeon, &r300->ind_buf.bo, &r300->ind_buf.bo_offset, size, 4); + radeon_bo_map(r300->ind_buf.bo, 1); assert(r300->ind_buf.bo->ptr != NULL); dst_ptr = ADD_POINTERS(r300->ind_buf.bo->ptr, r300->ind_buf.bo_offset); _mesa_memcpy(dst_ptr, src_ptr, size); + radeon_bo_unmap(r300->ind_buf.bo); r300->ind_buf.is_32bit = (mesa_ind_buf->type == GL_UNSIGNED_INT); r300->ind_buf.count = mesa_ind_buf->count; @@ -242,6 +246,7 @@ static void r300ConvertAttrib(GLcontext *ctx, int count, const struct gl_client_ } radeonAllocDmaRegion(&r300->radeon, &attr->bo, &attr->bo_offset, sizeof(GLfloat) * input->Size * count, 32); + radeon_bo_map(attr->bo, 1); dst_ptr = (GLfloat *)ADD_POINTERS(attr->bo->ptr, attr->bo_offset); radeon_print(RADEON_FALLBACKS, RADEON_IMPORTANT, @@ -280,6 +285,7 @@ static void r300ConvertAttrib(GLcontext *ctx, int count, const struct gl_client_ break; } + radeon_bo_unmap(attr->bo); if (mapped_named_bo) { ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER, input->BufferObj); } @@ -294,6 +300,8 @@ static void r300AlignDataToDword(GLcontext *ctx, const struct gl_client_array *i radeonAllocDmaRegion(&r300->radeon, &attr->bo, &attr->bo_offset, size, 32); + radeon_bo_map(attr->bo, 1); + if (!input->BufferObj->Pointer) { ctx->Driver.MapBuffer(ctx, GL_ARRAY_BUFFER, GL_READ_ONLY_ARB, input->BufferObj); mapped_named_bo = GL_TRUE; @@ -317,6 +325,7 @@ static void r300AlignDataToDword(GLcontext *ctx, const struct gl_client_array *i ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER, input->BufferObj); } + radeon_bo_unmap(attr->bo); attr->stride = dst_stride; } @@ -527,6 +536,7 @@ static void r300AllocDmaRegions(GLcontext *ctx, const struct gl_client_array *in } radeonAllocDmaRegion(&r300->radeon, &vbuf->attribs[index].bo, &vbuf->attribs[index].bo_offset, size, 32); + radeon_bo_map(vbuf->attribs[index].bo, 1); assert(vbuf->attribs[index].bo->ptr != NULL); dst = (uint32_t *)ADD_POINTERS(vbuf->attribs[index].bo->ptr, vbuf->attribs[index].bo_offset); switch (vbuf->attribs[index].dwords) { @@ -536,6 +546,7 @@ static void r300AllocDmaRegions(GLcontext *ctx, const struct gl_client_array *in case 4: radeonEmitVec16(dst, input[i]->Ptr, input[i]->StrideB, local_count); break; default: assert(0); break; } + radeon_bo_unmap(vbuf->attribs[index].bo); } } diff --git a/src/mesa/drivers/dri/r300/r300_emit.c b/src/mesa/drivers/dri/r300/r300_emit.c index 07e62230874..3759ca2bea7 100644 --- a/src/mesa/drivers/dri/r300/r300_emit.c +++ b/src/mesa/drivers/dri/r300/r300_emit.c @@ -49,7 +49,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r300_context.h" #include "r300_state.h" #include "r300_emit.h" -#include "r300_ioctl.h" #include "r300_render.h" #include "r300_swtcl.h" diff --git a/src/mesa/drivers/dri/r300/r300_ioctl.c b/src/mesa/drivers/dri/r300/r300_ioctl.c deleted file mode 100644 index 5cb04e2bb6d..00000000000 --- a/src/mesa/drivers/dri/r300/r300_ioctl.c +++ /dev/null @@ -1,782 +0,0 @@ -/* -Copyright (C) The Weather Channel, Inc. 2002. -Copyright (C) 2004 Nicolai Haehnle. -All Rights Reserved. - -The Weather Channel (TM) funded Tungsten Graphics to develop the -initial release of the Radeon 8500 driver under the XFree86 license. -This notice must be preserved. - -Permission is hereby granted, free of charge, to any person obtaining -a copy of this software and associated documentation files (the -"Software"), to deal in the Software without restriction, including -without limitation the rights to use, copy, modify, merge, publish, -distribute, sublicense, and/or sell copies of the Software, and to -permit persons to whom the Software is furnished to do so, subject to -the following conditions: - -The above copyright notice and this permission notice (including the -next paragraph) shall be included in all copies or substantial -portions of the Software. - -THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, -EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF -MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. -IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE -LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION -OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION -WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - -**************************************************************************/ - -/** - * \file - * - * \author Keith Whitwell <[email protected]> - * - * \author Nicolai Haehnle <[email protected]> - */ - -#include <sched.h> -#include <errno.h> - -#include "main/glheader.h" -#include "main/imports.h" -#include "main/macros.h" -#include "main/context.h" -#include "main/simple_list.h" -#include "swrast/swrast.h" - -#include "radeon_common.h" -#include "radeon_lock.h" -#include "r300_context.h" -#include "r300_ioctl.h" -#include "r300_cmdbuf.h" -#include "r300_state.h" -#include "r300_vertprog.h" -#include "radeon_reg.h" -#include "r300_emit.h" -#include "r300_context.h" - -#include "vblank.h" - -#define R200_3D_DRAW_IMMD_2 0xC0003500 - -#define CLEARBUFFER_COLOR 0x1 -#define CLEARBUFFER_DEPTH 0x2 -#define CLEARBUFFER_STENCIL 0x4 - -#if 1 - -/** - * Fragment program helper macros - */ - -/* Produce unshifted source selectors */ -#define FP_TMP(idx) (idx) -#define FP_CONST(idx) ((idx) | (1 << 5)) - -/* Produce source/dest selector dword */ -#define FP_SELC_MASK_NO 0 -#define FP_SELC_MASK_X 1 -#define FP_SELC_MASK_Y 2 -#define FP_SELC_MASK_XY 3 -#define FP_SELC_MASK_Z 4 -#define FP_SELC_MASK_XZ 5 -#define FP_SELC_MASK_YZ 6 -#define FP_SELC_MASK_XYZ 7 - -#define FP_SELC(destidx,regmask,outmask,src0,src1,src2) \ - (((destidx) << R300_ALU_DSTC_SHIFT) | \ - (FP_SELC_MASK_##regmask << 23) | \ - (FP_SELC_MASK_##outmask << 26) | \ - ((src0) << R300_ALU_SRC0C_SHIFT) | \ - ((src1) << R300_ALU_SRC1C_SHIFT) | \ - ((src2) << R300_ALU_SRC2C_SHIFT)) - -#define FP_SELA_MASK_NO 0 -#define FP_SELA_MASK_W 1 - -#define FP_SELA(destidx,regmask,outmask,src0,src1,src2) \ - (((destidx) << R300_ALU_DSTA_SHIFT) | \ - (FP_SELA_MASK_##regmask << 23) | \ - (FP_SELA_MASK_##outmask << 24) | \ - ((src0) << R300_ALU_SRC0A_SHIFT) | \ - ((src1) << R300_ALU_SRC1A_SHIFT) | \ - ((src2) << R300_ALU_SRC2A_SHIFT)) - -/* Produce unshifted argument selectors */ -#define FP_ARGC(source) R300_ALU_ARGC_##source -#define FP_ARGA(source) R300_ALU_ARGA_##source -#define FP_ABS(arg) ((arg) | (1 << 6)) -#define FP_NEG(arg) ((arg) ^ (1 << 5)) - -/* Produce instruction dword */ -#define FP_INSTRC(opcode,arg0,arg1,arg2) \ - (R300_ALU_OUTC_##opcode | \ - ((arg0) << R300_ALU_ARG0C_SHIFT) | \ - ((arg1) << R300_ALU_ARG1C_SHIFT) | \ - ((arg2) << R300_ALU_ARG2C_SHIFT)) - -#define FP_INSTRA(opcode,arg0,arg1,arg2) \ - (R300_ALU_OUTA_##opcode | \ - ((arg0) << R300_ALU_ARG0A_SHIFT) | \ - ((arg1) << R300_ALU_ARG1A_SHIFT) | \ - ((arg2) << R300_ALU_ARG2A_SHIFT)) - -#endif - -static void r300EmitClearState(GLcontext * ctx); - -static void r300ClearBuffer(r300ContextPtr r300, int flags, - struct radeon_renderbuffer *rrb, - struct radeon_renderbuffer *rrbd) -{ - BATCH_LOCALS(&r300->radeon); - GLcontext *ctx = r300->radeon.glCtx; - __DRIdrawablePrivate *dPriv = radeon_get_drawable(&r300->radeon); - GLuint cbpitch = 0; - r300ContextPtr rmesa = r300; - - if (RADEON_DEBUG & RADEON_IOCTL) - fprintf(stderr, "%s: buffer %p (%i,%i %ix%i)\n", - __FUNCTION__, rrb, dPriv->x, dPriv->y, - dPriv->w, dPriv->h); - - if (rrb) { - cbpitch = (rrb->pitch / rrb->cpp); - if (rrb->cpp == 4) - cbpitch |= R300_COLOR_FORMAT_ARGB8888; - else - cbpitch |= R300_COLOR_FORMAT_RGB565; - - if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE){ - cbpitch |= R300_COLOR_TILE_ENABLE; - } - } - - /* TODO in bufmgr */ - cp_wait(&r300->radeon, R300_WAIT_3D | R300_WAIT_3D_CLEAN); - end_3d(&rmesa->radeon); - - if (flags & CLEARBUFFER_COLOR) { - assert(rrb != 0); - BEGIN_BATCH_NO_AUTOSTATE(6); - OUT_BATCH_REGSEQ(R300_RB3D_COLOROFFSET0, 1); - OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); - OUT_BATCH_REGVAL(R300_RB3D_COLORPITCH0, cbpitch); - END_BATCH(); - } -#if 1 - if (flags & (CLEARBUFFER_DEPTH | CLEARBUFFER_STENCIL)) { - uint32_t zbpitch = (rrbd->pitch / rrbd->cpp); - if (rrbd->bo->flags & RADEON_BO_FLAGS_MACRO_TILE){ - zbpitch |= R300_DEPTHMACROTILE_ENABLE; - } - if (rrbd->bo->flags & RADEON_BO_FLAGS_MICRO_TILE){ - zbpitch |= R300_DEPTHMICROTILE_TILED; - } - BEGIN_BATCH_NO_AUTOSTATE(6); - OUT_BATCH_REGSEQ(R300_ZB_DEPTHOFFSET, 1); - OUT_BATCH_RELOC(0, rrbd->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); - OUT_BATCH_REGSEQ(R300_ZB_DEPTHPITCH, 1); - if (!r300->radeon.radeonScreen->kernel_mm) - OUT_BATCH(zbpitch); - else - OUT_BATCH_RELOC(zbpitch, rrbd->bo, zbpitch, 0, RADEON_GEM_DOMAIN_VRAM, 0); - END_BATCH(); - } -#endif - BEGIN_BATCH_NO_AUTOSTATE(6); - OUT_BATCH_REGSEQ(RB3D_COLOR_CHANNEL_MASK, 1); - if (flags & CLEARBUFFER_COLOR) { - OUT_BATCH((ctx->Color.ColorMask[BCOMP] ? RB3D_COLOR_CHANNEL_MASK_BLUE_MASK0 : 0) | - (ctx->Color.ColorMask[GCOMP] ? RB3D_COLOR_CHANNEL_MASK_GREEN_MASK0 : 0) | - (ctx->Color.ColorMask[RCOMP] ? RB3D_COLOR_CHANNEL_MASK_RED_MASK0 : 0) | - (ctx->Color.ColorMask[ACOMP] ? RB3D_COLOR_CHANNEL_MASK_ALPHA_MASK0 : 0)); - } else { - OUT_BATCH(0); - } - - - { - uint32_t t1, t2; - - t1 = 0x0; - t2 = 0x0; - - if (flags & CLEARBUFFER_DEPTH) { - t1 |= R300_Z_ENABLE | R300_Z_WRITE_ENABLE; - t2 |= - (R300_ZS_ALWAYS << R300_Z_FUNC_SHIFT); - } - - if (flags & CLEARBUFFER_STENCIL) { - t1 |= R300_STENCIL_ENABLE; - t2 |= - (R300_ZS_ALWAYS << - R300_S_FRONT_FUNC_SHIFT) | - (R300_ZS_REPLACE << - R300_S_FRONT_SFAIL_OP_SHIFT) | - (R300_ZS_REPLACE << - R300_S_FRONT_ZPASS_OP_SHIFT) | - (R300_ZS_REPLACE << - R300_S_FRONT_ZFAIL_OP_SHIFT); - } - - OUT_BATCH_REGSEQ(R300_ZB_CNTL, 3); - OUT_BATCH(t1); - OUT_BATCH(t2); - OUT_BATCH(((ctx->Stencil.WriteMask[0] & R300_STENCILREF_MASK) << - R300_STENCILWRITEMASK_SHIFT) | - (ctx->Stencil.Clear & R300_STENCILREF_MASK)); - END_BATCH(); - } - - if (!rmesa->radeon.radeonScreen->kernel_mm) { - BEGIN_BATCH_NO_AUTOSTATE(9); - OUT_BATCH(cmdpacket3(r300->radeon.radeonScreen, R300_CMD_PACKET3_CLEAR)); - OUT_BATCH_FLOAT32(dPriv->w / 2.0); - OUT_BATCH_FLOAT32(dPriv->h / 2.0); - OUT_BATCH_FLOAT32(ctx->Depth.Clear); - OUT_BATCH_FLOAT32(1.0); - OUT_BATCH_FLOAT32(ctx->Color.ClearColor[0]); - OUT_BATCH_FLOAT32(ctx->Color.ClearColor[1]); - OUT_BATCH_FLOAT32(ctx->Color.ClearColor[2]); - OUT_BATCH_FLOAT32(ctx->Color.ClearColor[3]); - END_BATCH(); - } else { - OUT_BATCH(CP_PACKET3(R200_3D_DRAW_IMMD_2, 8)); - OUT_BATCH(R300_PRIM_TYPE_POINT | R300_PRIM_WALK_RING | - (1 << R300_PRIM_NUM_VERTICES_SHIFT)); - OUT_BATCH_FLOAT32(dPriv->w / 2.0); - OUT_BATCH_FLOAT32(dPriv->h / 2.0); - OUT_BATCH_FLOAT32(ctx->Depth.Clear); - OUT_BATCH_FLOAT32(1.0); - OUT_BATCH_FLOAT32(ctx->Color.ClearColor[0]); - OUT_BATCH_FLOAT32(ctx->Color.ClearColor[1]); - OUT_BATCH_FLOAT32(ctx->Color.ClearColor[2]); - OUT_BATCH_FLOAT32(ctx->Color.ClearColor[3]); - } - - r300EmitCacheFlush(rmesa); - cp_wait(&r300->radeon, R300_WAIT_3D | R300_WAIT_3D_CLEAN); - - R300_STATECHANGE(r300, cb); - R300_STATECHANGE(r300, cmk); - R300_STATECHANGE(r300, zs); -} - -static void r300EmitClearState(GLcontext * ctx) -{ - r300ContextPtr r300 = R300_CONTEXT(ctx); - BATCH_LOCALS(&r300->radeon); - __DRIdrawablePrivate *dPriv = radeon_get_drawable(&r300->radeon); - int i; - int has_tcl; - int is_r500 = 0; - GLuint vap_cntl; - - has_tcl = r300->options.hw_tcl_enabled; - - if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) - is_r500 = 1; - - /* State atom dirty tracking is a little subtle here. - * - * On the one hand, we need to make sure base state is emitted - * here if we start with an empty batch buffer, otherwise clear - * works incorrectly with multiple processes. Therefore, the first - * BEGIN_BATCH cannot be a BEGIN_BATCH_NO_AUTOSTATE. - * - * On the other hand, implicit state emission clears the state atom - * dirty bits, so we have to call R300_STATECHANGE later than the - * first BEGIN_BATCH. - * - * The final trickiness is that, because we change state, we need - * to ensure that any stored swtcl primitives are flushed properly - * before we start changing state. See the R300_NEWPRIM in r300Clear - * for this. - */ - BEGIN_BATCH(31); - OUT_BATCH_REGSEQ(R300_VAP_PROG_STREAM_CNTL_0, 1); - if (!has_tcl) - OUT_BATCH(((((0 << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_0_SHIFT) | - ((R300_LAST_VEC | (2 << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_1_SHIFT))); - else - OUT_BATCH(((((0 << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_0_SHIFT) | - ((R300_LAST_VEC | (1 << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_1_SHIFT))); - - OUT_BATCH_REGVAL(R300_FG_FOG_BLEND, 0); - OUT_BATCH_REGVAL(R300_VAP_PROG_STREAM_CNTL_EXT_0, - ((((R300_SWIZZLE_SELECT_X << R300_SWIZZLE_SELECT_X_SHIFT) | - (R300_SWIZZLE_SELECT_Y << R300_SWIZZLE_SELECT_Y_SHIFT) | - (R300_SWIZZLE_SELECT_Z << R300_SWIZZLE_SELECT_Z_SHIFT) | - (R300_SWIZZLE_SELECT_W << R300_SWIZZLE_SELECT_W_SHIFT) | - ((R300_WRITE_ENA_X | R300_WRITE_ENA_Y | R300_WRITE_ENA_Z | R300_WRITE_ENA_W) << R300_WRITE_ENA_SHIFT)) - << R300_SWIZZLE0_SHIFT) | - (((R300_SWIZZLE_SELECT_X << R300_SWIZZLE_SELECT_X_SHIFT) | - (R300_SWIZZLE_SELECT_Y << R300_SWIZZLE_SELECT_Y_SHIFT) | - (R300_SWIZZLE_SELECT_Z << R300_SWIZZLE_SELECT_Z_SHIFT) | - (R300_SWIZZLE_SELECT_W << R300_SWIZZLE_SELECT_W_SHIFT) | - ((R300_WRITE_ENA_X | R300_WRITE_ENA_Y | R300_WRITE_ENA_Z | R300_WRITE_ENA_W) << R300_WRITE_ENA_SHIFT)) - << R300_SWIZZLE1_SHIFT))); - - /* R300_VAP_INPUT_CNTL_0, R300_VAP_INPUT_CNTL_1 */ - OUT_BATCH_REGSEQ(R300_VAP_VTX_STATE_CNTL, 2); - OUT_BATCH((R300_SEL_USER_COLOR_0 << R300_COLOR_0_ASSEMBLY_SHIFT)); - OUT_BATCH(R300_INPUT_CNTL_POS | R300_INPUT_CNTL_COLOR | R300_INPUT_CNTL_TC0); - - /* comes from fglrx startup of clear */ - OUT_BATCH_REGSEQ(R300_SE_VTE_CNTL, 2); - OUT_BATCH(R300_VTX_W0_FMT | R300_VPORT_X_SCALE_ENA | - R300_VPORT_X_OFFSET_ENA | R300_VPORT_Y_SCALE_ENA | - R300_VPORT_Y_OFFSET_ENA | R300_VPORT_Z_SCALE_ENA | - R300_VPORT_Z_OFFSET_ENA); - OUT_BATCH(0x8); - - OUT_BATCH_REGVAL(R300_VAP_PSC_SGN_NORM_CNTL, 0xaaaaaaaa); - - OUT_BATCH_REGSEQ(R300_VAP_OUTPUT_VTX_FMT_0, 2); - OUT_BATCH(R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT | - R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT); - OUT_BATCH(0); /* no textures */ - - OUT_BATCH_REGVAL(R300_TX_ENABLE, 0); - - OUT_BATCH_REGSEQ(R300_SE_VPORT_XSCALE, 6); - OUT_BATCH_FLOAT32(1.0); - OUT_BATCH_FLOAT32(dPriv->x); - OUT_BATCH_FLOAT32(1.0); - OUT_BATCH_FLOAT32(dPriv->y); - OUT_BATCH_FLOAT32(1.0); - OUT_BATCH_FLOAT32(0.0); - - OUT_BATCH_REGVAL(R300_FG_ALPHA_FUNC, 0); - - OUT_BATCH_REGSEQ(R300_RB3D_CBLEND, 2); - OUT_BATCH(0x0); - OUT_BATCH(0x0); - END_BATCH(); - - R300_STATECHANGE(r300, vir[0]); - R300_STATECHANGE(r300, fogs); - R300_STATECHANGE(r300, vir[1]); - R300_STATECHANGE(r300, vic); - R300_STATECHANGE(r300, vte); - R300_STATECHANGE(r300, vof); - R300_STATECHANGE(r300, txe); - R300_STATECHANGE(r300, vpt); - R300_STATECHANGE(r300, at); - R300_STATECHANGE(r300, bld); - R300_STATECHANGE(r300, ps); - - if (has_tcl) { - R300_STATECHANGE(r300, vap_clip_cntl); - - BEGIN_BATCH_NO_AUTOSTATE(2); - OUT_BATCH_REGVAL(R300_VAP_CLIP_CNTL, R300_PS_UCP_MODE_CLIP_AS_TRIFAN | R300_CLIP_DISABLE); - END_BATCH(); - } - - BEGIN_BATCH_NO_AUTOSTATE(2); - OUT_BATCH_REGVAL(R300_GA_POINT_SIZE, - ((dPriv->w * 6) << R300_POINTSIZE_X_SHIFT) | - ((dPriv->h * 6) << R300_POINTSIZE_Y_SHIFT)); - END_BATCH(); - - if (!is_r500) { - R300_STATECHANGE(r300, ri); - R300_STATECHANGE(r300, rc); - R300_STATECHANGE(r300, rr); - - BEGIN_BATCH(14); - OUT_BATCH_REGSEQ(R300_RS_IP_0, 8); - for (i = 0; i < 8; ++i) - OUT_BATCH(R300_RS_SEL_T(1) | R300_RS_SEL_R(2) | R300_RS_SEL_Q(3)); - - OUT_BATCH_REGSEQ(R300_RS_COUNT, 2); - OUT_BATCH((1 << R300_IC_COUNT_SHIFT) | R300_HIRES_EN); - OUT_BATCH(0x0); - - OUT_BATCH_REGVAL(R300_RS_INST_0, R300_RS_INST_COL_CN_WRITE); - END_BATCH(); - } else { - R300_STATECHANGE(r300, ri); - R300_STATECHANGE(r300, rc); - R300_STATECHANGE(r300, rr); - - BEGIN_BATCH(14); - OUT_BATCH_REGSEQ(R500_RS_IP_0, 8); - for (i = 0; i < 8; ++i) { - OUT_BATCH((R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_S_SHIFT) | - (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_T_SHIFT) | - (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_R_SHIFT) | - (R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT)); - } - - OUT_BATCH_REGSEQ(R300_RS_COUNT, 2); - OUT_BATCH((1 << R300_IC_COUNT_SHIFT) | R300_HIRES_EN); - OUT_BATCH(0x0); - - OUT_BATCH_REGVAL(R500_RS_INST_0, R500_RS_INST_COL_CN_WRITE); - END_BATCH(); - } - - if (!is_r500) { - R300_STATECHANGE(r300, fp); - R300_STATECHANGE(r300, fpi[0]); - R300_STATECHANGE(r300, fpi[1]); - R300_STATECHANGE(r300, fpi[2]); - R300_STATECHANGE(r300, fpi[3]); - - BEGIN_BATCH(17); - OUT_BATCH_REGSEQ(R300_US_CONFIG, 3); - OUT_BATCH(0x0); - OUT_BATCH(0x0); - OUT_BATCH(0x0); - OUT_BATCH_REGSEQ(R300_US_CODE_ADDR_0, 4); - OUT_BATCH(0x0); - OUT_BATCH(0x0); - OUT_BATCH(0x0); - OUT_BATCH(R300_RGBA_OUT); - - OUT_BATCH_REGVAL(R300_US_ALU_RGB_INST_0, - FP_INSTRC(MAD, FP_ARGC(SRC0C_XYZ), FP_ARGC(ONE), FP_ARGC(ZERO))); - OUT_BATCH_REGVAL(R300_US_ALU_RGB_ADDR_0, - FP_SELC(0, NO, XYZ, FP_TMP(0), 0, 0)); - OUT_BATCH_REGVAL(R300_US_ALU_ALPHA_INST_0, - FP_INSTRA(MAD, FP_ARGA(SRC0A), FP_ARGA(ONE), FP_ARGA(ZERO))); - OUT_BATCH_REGVAL(R300_US_ALU_ALPHA_ADDR_0, - FP_SELA(0, NO, W, FP_TMP(0), 0, 0)); - END_BATCH(); - } else { - struct radeon_state_atom r500fp; - uint32_t _cmd[10]; - - R300_STATECHANGE(r300, fp); - R300_STATECHANGE(r300, r500fp); - - BEGIN_BATCH(7); - OUT_BATCH_REGSEQ(R500_US_CONFIG, 2); - OUT_BATCH(R500_ZERO_TIMES_ANYTHING_EQUALS_ZERO); - OUT_BATCH(0x0); - OUT_BATCH_REGSEQ(R500_US_CODE_ADDR, 3); - OUT_BATCH(R500_US_CODE_START_ADDR(0) | R500_US_CODE_END_ADDR(1)); - OUT_BATCH(R500_US_CODE_RANGE_ADDR(0) | R500_US_CODE_RANGE_SIZE(1)); - OUT_BATCH(R500_US_CODE_OFFSET_ADDR(0)); - END_BATCH(); - - r500fp.check = check_r500fp; - r500fp.cmd = _cmd; - r500fp.cmd[0] = cmdr500fp(r300->radeon.radeonScreen, 0, 1, 0, 0); - r500fp.cmd[1] = R500_INST_TYPE_OUT | - R500_INST_TEX_SEM_WAIT | - R500_INST_LAST | - R500_INST_RGB_OMASK_R | - R500_INST_RGB_OMASK_G | - R500_INST_RGB_OMASK_B | - R500_INST_ALPHA_OMASK | - R500_INST_RGB_CLAMP | - R500_INST_ALPHA_CLAMP; - r500fp.cmd[2] = R500_RGB_ADDR0(0) | - R500_RGB_ADDR1(0) | - R500_RGB_ADDR1_CONST | - R500_RGB_ADDR2(0) | - R500_RGB_ADDR2_CONST; - r500fp.cmd[3] = R500_ALPHA_ADDR0(0) | - R500_ALPHA_ADDR1(0) | - R500_ALPHA_ADDR1_CONST | - R500_ALPHA_ADDR2(0) | - R500_ALPHA_ADDR2_CONST; - r500fp.cmd[4] = R500_ALU_RGB_SEL_A_SRC0 | - R500_ALU_RGB_R_SWIZ_A_R | - R500_ALU_RGB_G_SWIZ_A_G | - R500_ALU_RGB_B_SWIZ_A_B | - R500_ALU_RGB_SEL_B_SRC0 | - R500_ALU_RGB_R_SWIZ_B_R | - R500_ALU_RGB_B_SWIZ_B_G | - R500_ALU_RGB_G_SWIZ_B_B; - r500fp.cmd[5] = R500_ALPHA_OP_CMP | - R500_ALPHA_SWIZ_A_A | - R500_ALPHA_SWIZ_B_A; - r500fp.cmd[6] = R500_ALU_RGBA_OP_CMP | - R500_ALU_RGBA_R_SWIZ_0 | - R500_ALU_RGBA_G_SWIZ_0 | - R500_ALU_RGBA_B_SWIZ_0 | - R500_ALU_RGBA_A_SWIZ_0; - - r500fp.cmd[7] = 0; - if (r300->radeon.radeonScreen->kernel_mm) { - emit_r500fp(ctx, &r500fp); - } else { - int dwords = r500fp.check(ctx,&r500fp); - BEGIN_BATCH_NO_AUTOSTATE(dwords); - OUT_BATCH_TABLE(r500fp.cmd, dwords); - END_BATCH(); - } - - } - - BEGIN_BATCH(2); - OUT_BATCH_REGVAL(R300_VAP_PVS_STATE_FLUSH_REG, 0); - END_BATCH(); - - if (has_tcl) { - vap_cntl = ((10 << R300_PVS_NUM_SLOTS_SHIFT) | - (5 << R300_PVS_NUM_CNTLRS_SHIFT) | - (12 << R300_VF_MAX_VTX_NUM_SHIFT)); - if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) - vap_cntl |= R500_TCL_STATE_OPTIMIZATION; - } else { - vap_cntl = ((10 << R300_PVS_NUM_SLOTS_SHIFT) | - (5 << R300_PVS_NUM_CNTLRS_SHIFT) | - (5 << R300_VF_MAX_VTX_NUM_SHIFT)); - } - - if (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV515) - vap_cntl |= (2 << R300_PVS_NUM_FPUS_SHIFT); - else if ((r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV530) || - (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV560) || - (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV570)) - vap_cntl |= (5 << R300_PVS_NUM_FPUS_SHIFT); - else if ((r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV410) || - (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_R420)) - vap_cntl |= (6 << R300_PVS_NUM_FPUS_SHIFT); - else if ((r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_R520) || - (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_R580)) - vap_cntl |= (8 << R300_PVS_NUM_FPUS_SHIFT); - else - vap_cntl |= (4 << R300_PVS_NUM_FPUS_SHIFT); - - R300_STATECHANGE(r300, vap_cntl); - - BEGIN_BATCH(2); - OUT_BATCH_REGVAL(R300_VAP_CNTL, vap_cntl); - END_BATCH(); - - if (has_tcl) { - struct radeon_state_atom vpu; - uint32_t _cmd[10]; - R300_STATECHANGE(r300, pvs); - R300_STATECHANGE(r300, vap_flush); - R300_STATECHANGE(r300, vpi); - - BEGIN_BATCH(4); - OUT_BATCH_REGSEQ(R300_VAP_PVS_CODE_CNTL_0, 3); - OUT_BATCH((0 << R300_PVS_FIRST_INST_SHIFT) | - (0 << R300_PVS_XYZW_VALID_INST_SHIFT) | - (1 << R300_PVS_LAST_INST_SHIFT)); - OUT_BATCH((0 << R300_PVS_CONST_BASE_OFFSET_SHIFT) | - (0 << R300_PVS_MAX_CONST_ADDR_SHIFT)); - OUT_BATCH(1 << R300_PVS_LAST_VTX_SRC_INST_SHIFT); - END_BATCH(); - - vpu.check = check_vpu; - vpu.cmd = _cmd; - vpu.cmd[0] = cmdvpu(r300->radeon.radeonScreen, 0, 2); - - vpu.cmd[1] = PVS_OP_DST_OPERAND(VE_ADD, GL_FALSE, GL_FALSE, - 0, 0xf, PVS_DST_REG_OUT); - vpu.cmd[2] = PVS_SRC_OPERAND(0, PVS_SRC_SELECT_X, PVS_SRC_SELECT_Y, - PVS_SRC_SELECT_Z, PVS_SRC_SELECT_W, - PVS_SRC_REG_INPUT, NEGATE_NONE); - vpu.cmd[3] = PVS_SRC_OPERAND(0, PVS_SRC_SELECT_FORCE_0, - PVS_SRC_SELECT_FORCE_0, - PVS_SRC_SELECT_FORCE_0, - PVS_SRC_SELECT_FORCE_0, - PVS_SRC_REG_INPUT, NEGATE_NONE); - vpu.cmd[4] = 0x0; - - vpu.cmd[5] = PVS_OP_DST_OPERAND(VE_ADD, GL_FALSE, GL_FALSE, 1, 0xf, - PVS_DST_REG_OUT); - vpu.cmd[6] = PVS_SRC_OPERAND(1, PVS_SRC_SELECT_X, - PVS_SRC_SELECT_Y, PVS_SRC_SELECT_Z, - PVS_SRC_SELECT_W, PVS_SRC_REG_INPUT, - NEGATE_NONE); - vpu.cmd[7] = PVS_SRC_OPERAND(1, PVS_SRC_SELECT_FORCE_0, - PVS_SRC_SELECT_FORCE_0, - PVS_SRC_SELECT_FORCE_0, - PVS_SRC_SELECT_FORCE_0, - PVS_SRC_REG_INPUT, NEGATE_NONE); - vpu.cmd[8] = 0x0; - - if (r300->radeon.radeonScreen->kernel_mm) { - int dwords = r300->hw.vap_flush.check(ctx,&r300->hw.vap_flush); - BEGIN_BATCH_NO_AUTOSTATE(dwords); - OUT_BATCH_TABLE(r300->hw.vap_flush.cmd, dwords); - END_BATCH(); - emit_vpu(ctx, &vpu); - } else { - int dwords = vpu.check(ctx,&vpu); - BEGIN_BATCH_NO_AUTOSTATE(dwords); - OUT_BATCH_TABLE(vpu.cmd, dwords); - END_BATCH(); - } - - } -} - -static int r300KernelClear(GLcontext *ctx, GLuint flags) -{ - r300ContextPtr r300 = R300_CONTEXT(ctx); - __DRIdrawablePrivate *dPriv = radeon_get_drawable(&r300->radeon); - struct radeon_framebuffer *rfb = dPriv->driverPrivate; - struct radeon_renderbuffer *rrb; - struct radeon_renderbuffer *rrbd; - int bits = 0, ret; - - /* Make sure it fits there. */ - radeon_cs_space_reset_bos(r300->radeon.cmdbuf.cs); - - if (flags & BUFFER_BIT_COLOR0) { - rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_COLOR0); - radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs, - rrb->bo, 0, RADEON_GEM_DOMAIN_VRAM); - } - - if (flags & BUFFER_BIT_FRONT_LEFT) { - rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_FRONT_LEFT); - radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs, - rrb->bo, 0, RADEON_GEM_DOMAIN_VRAM); - } - - if (flags & BUFFER_BIT_BACK_LEFT) { - rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_BACK_LEFT); - radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs, - rrb->bo, 0, RADEON_GEM_DOMAIN_VRAM); - } - - rrbd = radeon_get_renderbuffer(&rfb->base, BUFFER_DEPTH); - if (rrbd) { - radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs, - rrbd->bo, 0, RADEON_GEM_DOMAIN_VRAM); - } - - ret = radeon_cs_space_check(r300->radeon.cmdbuf.cs); - if (ret) - return -1; - - rcommonEnsureCmdBufSpace(&r300->radeon, 421 * 3, __FUNCTION__); - if (flags || bits) - r300EmitClearState(ctx); - - rrbd = radeon_get_renderbuffer(&rfb->base, BUFFER_DEPTH); - if (rrbd && (flags & BUFFER_BIT_DEPTH)) - bits |= CLEARBUFFER_DEPTH; - - if (rrbd && (flags & BUFFER_BIT_STENCIL)) - bits |= CLEARBUFFER_STENCIL; - - if (flags & BUFFER_BIT_COLOR0) { - rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_COLOR0); - r300ClearBuffer(r300, CLEARBUFFER_COLOR, rrb, NULL); - bits = 0; - } - - if (flags & BUFFER_BIT_FRONT_LEFT) { - rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_FRONT_LEFT); - r300ClearBuffer(r300, bits | CLEARBUFFER_COLOR, rrb, rrbd); - bits = 0; - } - - if (flags & BUFFER_BIT_BACK_LEFT) { - rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_BACK_LEFT); - r300ClearBuffer(r300, bits | CLEARBUFFER_COLOR, rrb, rrbd); - bits = 0; - } - - if (bits) - r300ClearBuffer(r300, bits, NULL, rrbd); - - COMMIT_BATCH(); - return 0; -} - -/** - * Buffer clear - */ -static void r300Clear(GLcontext * ctx, GLbitfield mask) -{ - r300ContextPtr r300 = R300_CONTEXT(ctx); - __DRIdrawablePrivate *dPriv = radeon_get_drawable(&r300->radeon); - const GLuint colorMask = *((GLuint *) & ctx->Color.ColorMask); - GLbitfield swrast_mask = 0, tri_mask = 0; - int i, ret; - struct gl_framebuffer *fb = ctx->DrawBuffer; - - if (RADEON_DEBUG & RADEON_IOCTL) - fprintf(stderr, "r300Clear\n"); - - if (!r300->radeon.radeonScreen->driScreen->dri2.enabled) { - LOCK_HARDWARE(&r300->radeon); - UNLOCK_HARDWARE(&r300->radeon); - if (dPriv->numClipRects == 0) - return; - } - - /* Flush swtcl vertices if necessary, because we will change hardware - * state during clear. See also the state-related comment in - * r300EmitClearState. - */ - R300_NEWPRIM(r300); - - if (colorMask == ~0) - tri_mask |= (mask & BUFFER_BITS_COLOR); - else - tri_mask |= (mask & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_BACK_LEFT)); - - - /* HW stencil */ - if (mask & BUFFER_BIT_STENCIL) { - tri_mask |= BUFFER_BIT_STENCIL; - } - - /* HW depth */ - if (mask & BUFFER_BIT_DEPTH) { - tri_mask |= BUFFER_BIT_DEPTH; - } - - /* If we're doing a tri pass for depth/stencil, include a likely color - * buffer with it. - */ - - for (i = 0; i < BUFFER_COUNT; i++) { - GLuint bufBit = 1 << i; - if ((tri_mask) & bufBit) { - if (!fb->Attachment[i].Renderbuffer->ClassID) { - tri_mask &= ~bufBit; - swrast_mask |= bufBit; - } - } - } - - /* SW fallback clearing */ - swrast_mask = mask & ~tri_mask; - - ret = 0; - if (tri_mask) { - if (r300->radeon.radeonScreen->kernel_mm) - radeonUserClear(ctx, tri_mask); - else { - /* if kernel clear fails due to size restraints fallback */ - ret = r300KernelClear(ctx, tri_mask); - if (ret < 0) - swrast_mask |= tri_mask; - } - } - - if (swrast_mask) { - if (RADEON_DEBUG & RADEON_FALLBACKS) - fprintf(stderr, "%s: swrast clear, mask: %x\n", - __FUNCTION__, swrast_mask); - _swrast_Clear(ctx, swrast_mask); - } -} - -void r300InitIoctlFuncs(struct dd_function_table *functions) -{ - functions->Clear = r300Clear; - functions->Finish = radeonFinish; - functions->Flush = radeonFlush; -} diff --git a/src/mesa/drivers/dri/r300/r300_ioctl.h b/src/mesa/drivers/dri/r300/r300_ioctl.h deleted file mode 100644 index 3abfa71a6e8..00000000000 --- a/src/mesa/drivers/dri/r300/r300_ioctl.h +++ /dev/null @@ -1,44 +0,0 @@ -/* -Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. - -The Weather Channel (TM) funded Tungsten Graphics to develop the -initial release of the Radeon 8500 driver under the XFree86 license. -This notice must be preserved. - -Permission is hereby granted, free of charge, to any person obtaining -a copy of this software and associated documentation files (the -"Software"), to deal in the Software without restriction, including -without limitation the rights to use, copy, modify, merge, publish, -distribute, sublicense, and/or sell copies of the Software, and to -permit persons to whom the Software is furnished to do so, subject to -the following conditions: - -The above copyright notice and this permission notice (including the -next paragraph) shall be included in all copies or substantial -portions of the Software. - -THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, -EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF -MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. -IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE -LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION -OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION -WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - -**************************************************************************/ - -/* - * Authors: - * Keith Whitwell <[email protected]> - * Nicolai Haehnle <[email protected]> - */ - -#ifndef __R300_IOCTL_H__ -#define __R300_IOCTL_H__ - -#include "r300_context.h" -#include "radeon_drm.h" - -extern void r300InitIoctlFuncs(struct dd_function_table *functions); - -#endif /* __R300_IOCTL_H__ */ diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c index 4ae593cbe79..02c94250a8f 100644 --- a/src/mesa/drivers/dri/r300/r300_render.c +++ b/src/mesa/drivers/dri/r300/r300_render.c @@ -68,7 +68,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #include "tnl/tnl.h" #include "tnl/t_vp_build.h" #include "r300_context.h" -#include "r300_ioctl.h" #include "r300_state.h" #include "r300_reg.h" #include "r300_tex.h" diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c index ac20c08e201..da0a9dfb4cf 100644 --- a/src/mesa/drivers/dri/r300/r300_state.c +++ b/src/mesa/drivers/dri/r300/r300_state.c @@ -55,7 +55,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "tnl/t_vp_build.h" #include "r300_context.h" -#include "r300_ioctl.h" #include "r300_state.h" #include "r300_reg.h" #include "r300_emit.h" diff --git a/src/mesa/drivers/dri/r300/r300_swtcl.c b/src/mesa/drivers/dri/r300/r300_swtcl.c index ee2c71e1a7f..383c8a274b5 100644 --- a/src/mesa/drivers/dri/r300/r300_swtcl.c +++ b/src/mesa/drivers/dri/r300/r300_swtcl.c @@ -124,7 +124,7 @@ void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead, GLuint *_ } if (ctx->Light.Enabled && ctx->Light.Model.TwoSide) { - VB->AttribPtr[VERT_ATTRIB_GENERIC0] = VB->ColorPtr[1]; + VB->AttribPtr[VERT_ATTRIB_GENERIC0] = VB->BackfaceColorPtr; OutputsWritten |= 1 << VERT_RESULT_BFC0; #if MESA_LITTLE_ENDIAN EMIT_ATTR( _TNL_ATTRIB_GENERIC0, EMIT_4UB_4F_RGBA ); @@ -134,7 +134,7 @@ void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead, GLuint *_ ADD_ATTR(VERT_ATTRIB_GENERIC0, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR2, SWIZZLE_XYZW, MASK_XYZW, 1); #endif if (fp_reads & FRAG_BIT_COL1) { - VB->AttribPtr[VERT_ATTRIB_GENERIC1] = VB->SecondaryColorPtr[1]; + VB->AttribPtr[VERT_ATTRIB_GENERIC1] = VB->BackfaceSecondaryColorPtr; GLuint swiz = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ONE); OutputsWritten |= 1 << VERT_RESULT_BFC1; #if MESA_LITTLE_ENDIAN @@ -159,7 +159,7 @@ void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead, GLuint *_ int tex_id = rmesa->selected_fp->wpos_attr - FRAG_ATTRIB_TEX0; VB->AttribPtr[VERT_ATTRIB_TEX0 + tex_id] = VB->AttribPtr[VERT_ATTRIB_POS]; - VB->TexCoordPtr[tex_id] = VB->AttribPtr[VERT_ATTRIB_POS]; + VB->AttribPtr[_TNL_ATTRIB_TEX0 + tex_id] = VB->AttribPtr[VERT_ATTRIB_POS]; RENDERINPUTS_SET(tnl->render_inputs_bitset, _TNL_ATTRIB_TEX0 + tex_id); } @@ -167,7 +167,7 @@ void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead, GLuint *_ int tex_id = rmesa->selected_fp->fog_attr - FRAG_ATTRIB_TEX0; VB->AttribPtr[VERT_ATTRIB_TEX0 + tex_id] = VB->AttribPtr[VERT_ATTRIB_FOG]; - VB->TexCoordPtr[tex_id] = VB->AttribPtr[VERT_ATTRIB_FOG]; + VB->AttribPtr[_TNL_ATTRIB_TEX0 + tex_id] = VB->AttribPtr[VERT_ATTRIB_FOG]; RENDERINPUTS_SET(tnl->render_inputs_bitset, _TNL_ATTRIB_TEX0 + tex_id); } @@ -180,7 +180,7 @@ void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead, GLuint *_ GLuint swiz, format, hw_format; for (i = 0; i < ctx->Const.MaxTextureUnits; i++) { if (fp_reads & FRAG_BIT_TEX(i)) { - switch (VB->TexCoordPtr[i]->size) { + switch (VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->size) { case 1: format = EMIT_1F; hw_format = R300_DATA_TYPE_FLOAT_1; @@ -665,11 +665,11 @@ void r300_swtcl_flush(GLcontext *ctx, uint32_t current_offset) r300EmitCacheFlush(rmesa); radeonEmitState(&rmesa->radeon); - r300_emit_scissor(ctx); + r300_emit_scissor(ctx); r300EmitVertexAOS(rmesa, - rmesa->radeon.swtcl.vertex_size, - first_elem(&rmesa->radeon.dma.reserved)->bo, - current_offset); + rmesa->radeon.swtcl.vertex_size, + rmesa->radeon.swtcl.bo, + current_offset); r300EmitVbufPrim(rmesa, rmesa->radeon.swtcl.hw_primitive, diff --git a/src/mesa/drivers/dri/r300/r300_tex.c b/src/mesa/drivers/dri/r300/r300_tex.c index 726b3ff98e1..ac3d5b1bec3 100644 --- a/src/mesa/drivers/dri/r300/r300_tex.c +++ b/src/mesa/drivers/dri/r300/r300_tex.c @@ -48,7 +48,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r300_context.h" #include "r300_state.h" -#include "r300_ioctl.h" #include "radeon_mipmap_tree.h" #include "r300_tex.h" diff --git a/src/mesa/drivers/dri/r300/r300_tex.h b/src/mesa/drivers/dri/r300/r300_tex.h index 8a653ea2d11..6ede0fe25c9 100644 --- a/src/mesa/drivers/dri/r300/r300_tex.h +++ b/src/mesa/drivers/dri/r300/r300_tex.h @@ -51,4 +51,6 @@ extern GLboolean r300ValidateBuffers(GLcontext * ctx); extern void r300InitTextureFuncs(struct dd_function_table *functions); +int32_t r300TranslateTexFormat(gl_format mesaFormat); + #endif /* __r300_TEX_H__ */ diff --git a/src/mesa/drivers/dri/r300/r300_texcopy.c b/src/mesa/drivers/dri/r300/r300_texcopy.c new file mode 100644 index 00000000000..ebc9c05b8a8 --- /dev/null +++ b/src/mesa/drivers/dri/r300/r300_texcopy.c @@ -0,0 +1,168 @@ +/* + * Copyright (C) 2009 Maciej Cencora <[email protected]> + * + * 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, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "radeon_common.h" +#include "r300_context.h" + +#include "main/image.h" +#include "main/teximage.h" +#include "main/texstate.h" +#include "drivers/common/meta.h" + +#include "radeon_mipmap_tree.h" +#include "r300_blit.h" +#include <main/debug.h> + +// TODO: +// need to pass correct pitch for small dst textures! +static GLboolean +do_copy_texsubimage(GLcontext *ctx, + GLenum target, GLint level, + struct radeon_tex_obj *tobj, + radeon_texture_image *timg, + GLint dstx, GLint dsty, + GLint x, GLint y, + GLsizei width, GLsizei height) +{ + struct r300_context *r300 = R300_CONTEXT(ctx); + struct radeon_renderbuffer *rrb; + + if (_mesa_get_format_bits(timg->base.TexFormat, GL_DEPTH_BITS) > 0) { + rrb = radeon_get_depthbuffer(&r300->radeon); + } else { + rrb = radeon_get_colorbuffer(&r300->radeon); + } + + if (!timg->mt) { + radeon_validate_texture_miptree(ctx, &tobj->base); + } + + assert(rrb && rrb->bo); + assert(timg->mt->bo); + assert(timg->base.Width >= dstx + width); + assert(timg->base.Height >= dsty + height); + + intptr_t src_offset = rrb->draw_offset; + intptr_t dst_offset = radeon_miptree_image_offset(timg->mt, _mesa_tex_target_to_face(target), level); + + if (src_offset % 32 || dst_offset % 32) { + return GL_FALSE; + } + + if (0) { + fprintf(stderr, "%s: copying to face %d, level %d\n", + __FUNCTION__, _mesa_tex_target_to_face(target), level); + fprintf(stderr, "to: x %d, y %d, offset %d\n", dstx, dsty, (uint32_t) dst_offset); + fprintf(stderr, "from (%dx%d) width %d, height %d, offset %d, pitch %d\n", + x, y, rrb->base.Width, rrb->base.Height, (uint32_t) src_offset, rrb->pitch/rrb->cpp); + fprintf(stderr, "src size %d, dst size %d\n", rrb->bo->size, timg->mt->bo->size); + + } + + /* blit from src buffer to texture */ + return r300_blit(r300, rrb->bo, src_offset, rrb->base.Format, rrb->pitch/rrb->cpp, + rrb->base.Width, rrb->base.Height, x, y, + timg->mt->bo, dst_offset, timg->base.TexFormat, + timg->base.Width, timg->base.Width, timg->base.Height, + dstx, dsty, width, height, 1); +} + +static void +r300CopyTexImage2D(GLcontext *ctx, GLenum target, GLint level, + GLenum internalFormat, + GLint x, GLint y, GLsizei width, GLsizei height, + GLint border) +{ + struct gl_texture_unit *texUnit = _mesa_get_current_tex_unit(ctx); + struct gl_texture_object *texObj = + _mesa_select_tex_object(ctx, texUnit, target); + struct gl_texture_image *texImage = + _mesa_select_tex_image(ctx, texObj, target, level); + int srcx, srcy, dstx, dsty; + + if (border) + goto fail; + + /* Setup or redefine the texture object, mipmap tree and texture + * image. Don't populate yet. + */ + ctx->Driver.TexImage2D(ctx, target, level, internalFormat, + width, height, border, + GL_RGBA, GL_UNSIGNED_BYTE, NULL, + &ctx->DefaultPacking, texObj, texImage); + + srcx = x; + srcy = y; + dstx = 0; + dsty = 0; + if (!_mesa_clip_copytexsubimage(ctx, + &dstx, &dsty, + &srcx, &srcy, + &width, &height)) { + return; + } + + if (!do_copy_texsubimage(ctx, target, level, + radeon_tex_obj(texObj), (radeon_texture_image *)texImage, + 0, 0, x, y, width, height)) { + goto fail; + } + + return; + +fail: + _mesa_meta_CopyTexImage2D(ctx, target, level, internalFormat, x, y, + width, height, border); +} + +static void +r300CopyTexSubImage2D(GLcontext *ctx, GLenum target, GLint level, + GLint xoffset, GLint yoffset, + GLint x, GLint y, + GLsizei width, GLsizei height) +{ + struct gl_texture_unit *texUnit = _mesa_get_current_tex_unit(ctx); + struct gl_texture_object *texObj = _mesa_select_tex_object(ctx, texUnit, target); + struct gl_texture_image *texImage = _mesa_select_tex_image(ctx, texObj, target, level); + + if (!do_copy_texsubimage(ctx, target, level, + radeon_tex_obj(texObj), (radeon_texture_image *)texImage, + xoffset, yoffset, x, y, width, height)) { + + //DEBUG_FALLBACKS + + _mesa_meta_CopyTexSubImage2D(ctx, target, level, + xoffset, yoffset, x, y, width, height); + } +} + + +void r300_init_texcopy_functions(struct dd_function_table *table) +{ + table->CopyTexImage2D = r300CopyTexImage2D; + table->CopyTexSubImage2D = r300CopyTexSubImage2D; +}
\ No newline at end of file diff --git a/src/mesa/drivers/dri/r300/r300_texstate.c b/src/mesa/drivers/dri/r300/r300_texstate.c index bbe8b1e95f9..eea3dee7bc6 100644 --- a/src/mesa/drivers/dri/r300/r300_texstate.c +++ b/src/mesa/drivers/dri/r300/r300_texstate.c @@ -46,19 +46,10 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r300_context.h" #include "r300_state.h" -#include "r300_ioctl.h" #include "radeon_mipmap_tree.h" #include "r300_tex.h" #include "r300_reg.h" -#define VALID_FORMAT(f) ( ((f) <= MESA_FORMAT_RGBA_DXT5 \ - || ((f) >= MESA_FORMAT_RGBA_FLOAT32 && \ - (f) <= MESA_FORMAT_INTENSITY_FLOAT16)) \ - && tx_table[f].flag ) - -#define _ASSIGN(entry, format) \ - [ MESA_FORMAT_ ## entry ] = { format, 0, 1} - /* * Note that the _REV formats are the same as the non-REV formats. This is * because the REV and non-REV formats are identical as a byte string, but @@ -68,67 +59,119 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. * identically. -- paulus */ -static const struct tx_table { - GLuint format, filter, flag; -} tx_table[] = { - /* *INDENT-OFF* */ +int32_t r300TranslateTexFormat(gl_format mesaFormat) +{ + switch (mesaFormat) + { #ifdef MESA_LITTLE_ENDIAN - _ASSIGN(RGBA8888, R300_EASY_TX_FORMAT(Y, Z, W, X, W8Z8Y8X8)), - _ASSIGN(RGBA8888_REV, R300_EASY_TX_FORMAT(Z, Y, X, W, W8Z8Y8X8)), - _ASSIGN(ARGB8888, R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8)), - _ASSIGN(ARGB8888_REV, R300_EASY_TX_FORMAT(W, Z, Y, X, W8Z8Y8X8)), + case MESA_FORMAT_RGBA8888: + return R300_EASY_TX_FORMAT(Y, Z, W, X, W8Z8Y8X8); + case MESA_FORMAT_RGBA8888_REV: + return R300_EASY_TX_FORMAT(Z, Y, X, W, W8Z8Y8X8); + case MESA_FORMAT_ARGB8888: + return R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8); + case MESA_FORMAT_ARGB8888_REV: + return R300_EASY_TX_FORMAT(W, Z, Y, X, W8Z8Y8X8); #else - _ASSIGN(RGBA8888, R300_EASY_TX_FORMAT(Z, Y, X, W, W8Z8Y8X8)), - _ASSIGN(RGBA8888_REV, R300_EASY_TX_FORMAT(Y, Z, W, X, W8Z8Y8X8)), - _ASSIGN(ARGB8888, R300_EASY_TX_FORMAT(W, Z, Y, X, W8Z8Y8X8)), - _ASSIGN(ARGB8888_REV, R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8)), + case MESA_FORMAT_RGBA8888: + return R300_EASY_TX_FORMAT(Z, Y, X, W, W8Z8Y8X8); + case MESA_FORMAT_RGBA8888_REV: + return R300_EASY_TX_FORMAT(Y, Z, W, X, W8Z8Y8X8); + case MESA_FORMAT_ARGB8888: + return R300_EASY_TX_FORMAT(W, Z, Y, X, W8Z8Y8X8); + case MESA_FORMAT_ARGB8888_REV: + return R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8); #endif - _ASSIGN(XRGB8888, R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8)), - _ASSIGN(RGB888, R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8)), - _ASSIGN(RGB565, R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z5Y6X5)), - _ASSIGN(RGB565_REV, R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z5Y6X5)), - _ASSIGN(ARGB4444, R300_EASY_TX_FORMAT(X, Y, Z, W, W4Z4Y4X4)), - _ASSIGN(ARGB4444_REV, R300_EASY_TX_FORMAT(X, Y, Z, W, W4Z4Y4X4)), - _ASSIGN(ARGB1555, R300_EASY_TX_FORMAT(X, Y, Z, W, W1Z5Y5X5)), - _ASSIGN(ARGB1555_REV, R300_EASY_TX_FORMAT(X, Y, Z, W, W1Z5Y5X5)), - _ASSIGN(AL88, R300_EASY_TX_FORMAT(X, X, X, Y, Y8X8)), - _ASSIGN(AL88_REV, R300_EASY_TX_FORMAT(X, X, X, Y, Y8X8)), - _ASSIGN(RGB332, R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z3Y3X2)), - _ASSIGN(A8, R300_EASY_TX_FORMAT(ZERO, ZERO, ZERO, X, X8)), - _ASSIGN(L8, R300_EASY_TX_FORMAT(X, X, X, ONE, X8)), - _ASSIGN(I8, R300_EASY_TX_FORMAT(X, X, X, X, X8)), - _ASSIGN(CI8, R300_EASY_TX_FORMAT(X, X, X, X, X8)), - _ASSIGN(YCBCR, R300_EASY_TX_FORMAT(X, Y, Z, ONE, G8R8_G8B8) | R300_TX_FORMAT_YUV_MODE), - _ASSIGN(YCBCR_REV, R300_EASY_TX_FORMAT(X, Y, Z, ONE, G8R8_G8B8) | R300_TX_FORMAT_YUV_MODE), - _ASSIGN(RGB_DXT1, R300_EASY_TX_FORMAT(X, Y, Z, ONE, DXT1)), - _ASSIGN(RGBA_DXT1, R300_EASY_TX_FORMAT(X, Y, Z, W, DXT1)), - _ASSIGN(RGBA_DXT3, R300_EASY_TX_FORMAT(X, Y, Z, W, DXT3)), - _ASSIGN(RGBA_DXT5, R300_EASY_TX_FORMAT(Y, Z, W, X, DXT5)), - _ASSIGN(RGBA_FLOAT32, R300_EASY_TX_FORMAT(Z, Y, X, W, FL_R32G32B32A32)), - _ASSIGN(RGBA_FLOAT16, R300_EASY_TX_FORMAT(Z, Y, X, W, FL_R16G16B16A16)), - _ASSIGN(RGB_FLOAT32, 0xffffffff), - _ASSIGN(RGB_FLOAT16, 0xffffffff), - _ASSIGN(ALPHA_FLOAT32, R300_EASY_TX_FORMAT(ZERO, ZERO, ZERO, X, FL_I32)), - _ASSIGN(ALPHA_FLOAT16, R300_EASY_TX_FORMAT(ZERO, ZERO, ZERO, X, FL_I16)), - _ASSIGN(LUMINANCE_FLOAT32, R300_EASY_TX_FORMAT(X, X, X, ONE, FL_I32)), - _ASSIGN(LUMINANCE_FLOAT16, R300_EASY_TX_FORMAT(X, X, X, ONE, FL_I16)), - _ASSIGN(LUMINANCE_ALPHA_FLOAT32, R300_EASY_TX_FORMAT(X, X, X, Y, FL_I32A32)), - _ASSIGN(LUMINANCE_ALPHA_FLOAT16, R300_EASY_TX_FORMAT(X, X, X, Y, FL_I16A16)), - _ASSIGN(INTENSITY_FLOAT32, R300_EASY_TX_FORMAT(X, X, X, X, FL_I32)), - _ASSIGN(INTENSITY_FLOAT16, R300_EASY_TX_FORMAT(X, X, X, X, FL_I16)), - _ASSIGN(Z16, R300_EASY_TX_FORMAT(X, X, X, X, X16)), - _ASSIGN(Z24_S8, R300_EASY_TX_FORMAT(X, X, X, X, X24_Y8)), - _ASSIGN(S8_Z24, R300_EASY_TX_FORMAT(Y, Y, Y, Y, X24_Y8)), - _ASSIGN(Z32, R300_EASY_TX_FORMAT(X, X, X, X, X32)), - /* EXT_texture_sRGB */ - _ASSIGN(SRGBA8, R300_EASY_TX_FORMAT(Y, Z, W, X, W8Z8Y8X8) | R300_TX_FORMAT_GAMMA), - _ASSIGN(SLA8, R300_EASY_TX_FORMAT(X, X, X, Y, Y8X8) | R300_TX_FORMAT_GAMMA), - _ASSIGN(SL8, R300_EASY_TX_FORMAT(X, X, X, ONE, X8) | R300_TX_FORMAT_GAMMA), - /* *INDENT-ON* */ + case MESA_FORMAT_XRGB8888: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8); + case MESA_FORMAT_RGB888: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8); + case MESA_FORMAT_RGB565: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z5Y6X5); + case MESA_FORMAT_RGB565_REV: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z5Y6X5); + case MESA_FORMAT_ARGB4444: + return R300_EASY_TX_FORMAT(X, Y, Z, W, W4Z4Y4X4); + case MESA_FORMAT_ARGB4444_REV: + return R300_EASY_TX_FORMAT(X, Y, Z, W, W4Z4Y4X4); + case MESA_FORMAT_ARGB1555: + return R300_EASY_TX_FORMAT(X, Y, Z, W, W1Z5Y5X5); + case MESA_FORMAT_ARGB1555_REV: + return R300_EASY_TX_FORMAT(X, Y, Z, W, W1Z5Y5X5); + case MESA_FORMAT_AL88: + return R300_EASY_TX_FORMAT(X, X, X, Y, Y8X8); + case MESA_FORMAT_AL88_REV: + return R300_EASY_TX_FORMAT(X, X, X, Y, Y8X8); + case MESA_FORMAT_RGB332: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z3Y3X2); + case MESA_FORMAT_A8: + return R300_EASY_TX_FORMAT(ZERO, ZERO, ZERO, X, X8); + case MESA_FORMAT_L8: + return R300_EASY_TX_FORMAT(X, X, X, ONE, X8); + case MESA_FORMAT_I8: + return R300_EASY_TX_FORMAT(X, X, X, X, X8); + case MESA_FORMAT_CI8: + return R300_EASY_TX_FORMAT(X, X, X, X, X8); + case MESA_FORMAT_YCBCR: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, G8R8_G8B8) | R300_TX_FORMAT_YUV_MODE; + case MESA_FORMAT_YCBCR_REV: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, G8R8_G8B8) | R300_TX_FORMAT_YUV_MODE; + case MESA_FORMAT_RGB_DXT1: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, DXT1); + case MESA_FORMAT_RGBA_DXT1: + return R300_EASY_TX_FORMAT(X, Y, Z, W, DXT1); + case MESA_FORMAT_RGBA_DXT3: + return R300_EASY_TX_FORMAT(X, Y, Z, W, DXT3); + case MESA_FORMAT_RGBA_DXT5: + return R300_EASY_TX_FORMAT(Y, Z, W, X, DXT5); + case MESA_FORMAT_RGBA_FLOAT32: + return R300_EASY_TX_FORMAT(Z, Y, X, W, FL_R32G32B32A32); + case MESA_FORMAT_RGBA_FLOAT16: + return R300_EASY_TX_FORMAT(Z, Y, X, W, FL_R16G16B16A16); + case MESA_FORMAT_ALPHA_FLOAT32: + return R300_EASY_TX_FORMAT(ZERO, ZERO, ZERO, X, FL_I32); + case MESA_FORMAT_ALPHA_FLOAT16: + return R300_EASY_TX_FORMAT(ZERO, ZERO, ZERO, X, FL_I16); + case MESA_FORMAT_LUMINANCE_FLOAT32: + return R300_EASY_TX_FORMAT(X, X, X, ONE, FL_I32); + case MESA_FORMAT_LUMINANCE_FLOAT16: + return R300_EASY_TX_FORMAT(X, X, X, ONE, FL_I16); + case MESA_FORMAT_LUMINANCE_ALPHA_FLOAT32: + return R300_EASY_TX_FORMAT(X, X, X, Y, FL_I32A32); + case MESA_FORMAT_LUMINANCE_ALPHA_FLOAT16: + return R300_EASY_TX_FORMAT(X, X, X, Y, FL_I16A16); + case MESA_FORMAT_INTENSITY_FLOAT32: + return R300_EASY_TX_FORMAT(X, X, X, X, FL_I32); + case MESA_FORMAT_INTENSITY_FLOAT16: + return R300_EASY_TX_FORMAT(X, X, X, X, FL_I16); + case MESA_FORMAT_Z16: + return R300_EASY_TX_FORMAT(X, X, X, X, X16); + case MESA_FORMAT_Z24_S8: + return R300_EASY_TX_FORMAT(X, X, X, X, X24_Y8); + case MESA_FORMAT_S8_Z24: + return R300_EASY_TX_FORMAT(Y, Y, Y, Y, X24_Y8); + case MESA_FORMAT_Z32: + return R300_EASY_TX_FORMAT(X, X, X, X, X32); + /* EXT_texture_sRGB */ + case MESA_FORMAT_SRGBA8: + return R300_EASY_TX_FORMAT(Y, Z, W, X, W8Z8Y8X8) | R300_TX_FORMAT_GAMMA; + case MESA_FORMAT_SLA8: + return R300_EASY_TX_FORMAT(X, X, X, Y, Y8X8) | R300_TX_FORMAT_GAMMA; + case MESA_FORMAT_SL8: + return R300_EASY_TX_FORMAT(X, X, X, ONE, X8) | R300_TX_FORMAT_GAMMA; + case MESA_FORMAT_SRGB_DXT1: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, DXT1) | R300_TX_FORMAT_GAMMA; + case MESA_FORMAT_SRGBA_DXT1: + return R300_EASY_TX_FORMAT(X, Y, Z, W, DXT1) | R300_TX_FORMAT_GAMMA; + case MESA_FORMAT_SRGBA_DXT3: + return R300_EASY_TX_FORMAT(X, Y, Z, W, DXT3) | R300_TX_FORMAT_GAMMA; + case MESA_FORMAT_SRGBA_DXT5: + return R300_EASY_TX_FORMAT(Y, Z, W, X, DXT5) | R300_TX_FORMAT_GAMMA; + default: + return -1; + } }; -#undef _ASSIGN - void r300SetDepthTexMode(struct gl_texture_object *tObj) { static const GLuint formats[3][3] = { @@ -205,19 +248,18 @@ static void setup_hardware_state(r300ContextPtr rmesa, radeonTexObj *t) const struct gl_texture_image *firstImage; firstImage = t->base.Image[0][t->minLod]; - if (!t->image_override - && VALID_FORMAT(firstImage->TexFormat)) { + if (!t->image_override) { if (firstImage->_BaseFormat == GL_DEPTH_COMPONENT) { r300SetDepthTexMode(&t->base); } else { - t->pp_txformat = tx_table[firstImage->TexFormat].format; + int32_t txformat = r300TranslateTexFormat(firstImage->TexFormat); + if (txformat < 0) { + _mesa_problem(rmesa->radeon.glCtx, "%s: Invalid format %s", + __FUNCTION__, _mesa_get_format_name(firstImage->TexFormat)); + _mesa_exit(1); + } + t->pp_txformat = (uint32_t) txformat; } - - t->pp_txfilter |= tx_table[firstImage->TexFormat].filter; - } else if (!t->image_override) { - _mesa_problem(NULL, "unexpected texture format in %s", - __FUNCTION__); - return; } if (t->image_override && t->bo) @@ -357,18 +399,15 @@ void r300SetTexOffset(__DRIcontext * pDRICtx, GLint texname, switch (depth) { case 32: t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8); - t->pp_txfilter |= tx_table[2].filter; pitch_val /= 4; break; case 24: default: t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8); - t->pp_txfilter |= tx_table[4].filter; pitch_val /= 4; break; case 16: t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z5Y6X5); - t->pp_txfilter |= tx_table[5].filter; pitch_val /= 2; break; } @@ -409,18 +448,7 @@ void r300SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_fo return; } - radeon_update_renderbuffers(pDRICtx, dPriv); - /* back & depth buffer are useless free them right away */ - rb = (void*)rfb->base.Attachment[BUFFER_DEPTH].Renderbuffer; - if (rb && rb->bo) { - radeon_bo_unref(rb->bo); - rb->bo = NULL; - } - rb = (void*)rfb->base.Attachment[BUFFER_BACK_LEFT].Renderbuffer; - if (rb && rb->bo) { - radeon_bo_unref(rb->bo); - rb->bo = NULL; - } + radeon_update_renderbuffers(pDRICtx, dPriv, GL_TRUE); rb = rfb->color_rb[0]; if (rb->bo == NULL) { /* Failed to BO for the buffer */ @@ -458,18 +486,15 @@ void r300SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_fo t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8); else t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8); - t->pp_txfilter |= tx_table[2].filter; pitch_val /= 4; break; case 3: default: t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8); - t->pp_txfilter |= tx_table[4].filter; pitch_val /= 4; break; case 2: t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z5Y6X5); - t->pp_txfilter |= tx_table[5].filter; pitch_val /= 2; break; } diff --git a/src/mesa/drivers/dri/r600/r600_context.c b/src/mesa/drivers/dri/r600/r600_context.c index dbd233729c0..45bbc3c071e 100644 --- a/src/mesa/drivers/dri/r600/r600_context.c +++ b/src/mesa/drivers/dri/r600/r600_context.c @@ -74,6 +74,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "utils.h" #include "xmlpool.h" /* for symbolic values of enum-type options */ +#define R600_ENABLE_GLSL_TEST 1 + #define need_GL_VERSION_2_0 #define need_GL_ARB_occlusion_query #define need_GL_ARB_point_parameters @@ -109,6 +111,7 @@ static const struct dri_extension card_extensions[] = { {"GL_ARB_texture_env_crossbar", NULL}, {"GL_ARB_texture_env_dot3", NULL}, {"GL_ARB_texture_mirrored_repeat", NULL}, + {"GL_ARB_texture_non_power_of_two", NULL}, {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions}, {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions}, {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions}, @@ -155,7 +158,11 @@ static const struct dri_extension mm_extensions[] = { * functions added by GL_ATI_separate_stencil. */ static const struct dri_extension gl_20_extension[] = { +#ifdef R600_ENABLE_GLSL_TEST + {"GL_ARB_shading_language_100", GL_VERSION_2_0_functions }, +#else {"GL_VERSION_2_0", GL_VERSION_2_0_functions }, +#endif /* R600_ENABLE_GLSL_TEST */ }; static const struct tnl_pipeline_stage *r600_pipeline[] = { @@ -308,6 +315,14 @@ static void r600InitGLExtensions(GLcontext *ctx) if (r600->radeon.radeonScreen->kernel_mm) driInitExtensions(ctx, mm_extensions, GL_FALSE); +#ifdef R600_ENABLE_GLSL_TEST + driInitExtensions(ctx, gl_20_extension, GL_TRUE); + _mesa_enable_2_0_extensions(ctx); + + /* glsl compiler has problem if this is not GL_TRUE */ + ctx->Shader.EmitCondCodes = GL_TRUE; +#endif /* R600_ENABLE_GLSL_TEST */ + if (driQueryOptionb (&r600->radeon.optionCache, "disable_stencil_two_side")) _mesa_disable_extension(ctx, "GL_EXT_stencil_two_side"); diff --git a/src/mesa/drivers/dri/r600/r600_texstate.c b/src/mesa/drivers/dri/r600/r600_texstate.c index 4ec315b78c7..2a4a6e6ee14 100644 --- a/src/mesa/drivers/dri/r600/r600_texstate.c +++ b/src/mesa/drivers/dri/r600/r600_texstate.c @@ -917,18 +917,7 @@ void r600SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_fo return; } - radeon_update_renderbuffers(pDRICtx, dPriv); - /* back & depth buffer are useless free them right away */ - rb = (void*)rfb->base.Attachment[BUFFER_DEPTH].Renderbuffer; - if (rb && rb->bo) { - radeon_bo_unref(rb->bo); - rb->bo = NULL; - } - rb = (void*)rfb->base.Attachment[BUFFER_BACK_LEFT].Renderbuffer; - if (rb && rb->bo) { - radeon_bo_unref(rb->bo); - rb->bo = NULL; - } + radeon_update_renderbuffers(pDRICtx, dPriv, GL_TRUE); rb = rfb->color_rb[0]; if (rb->bo == NULL) { /* Failed to BO for the buffer */ diff --git a/src/mesa/drivers/dri/r600/r700_assembler.c b/src/mesa/drivers/dri/r600/r700_assembler.c index 67e0ee77463..51692a11ffb 100644 --- a/src/mesa/drivers/dri/r600/r700_assembler.c +++ b/src/mesa/drivers/dri/r600/r700_assembler.c @@ -32,12 +32,49 @@ #include "main/mtypes.h" #include "main/imports.h" +#include "shader/prog_parameter.h" #include "radeon_debug.h" #include "r600_context.h" #include "r700_assembler.h" +#define USE_CF_FOR_CONTINUE_BREAK 1 +#define USE_CF_FOR_POP_AFTER 1 + +struct prog_instruction noise1_insts[12] = { + {OPCODE_BGNSUB , {{13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {13, 0, 15, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_MOV , {{0, 0, 0, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {0, 0, 2, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_MOV , {{8, 0, 0, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {0, 0, 4, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_MOV , {{8, 0, 585, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {0, 0, 8, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_SGT , {{0, 0, 585, 0, 0, 0}, {8, 0, 1170, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {0, 1, 1, 0, 8, 1672, 0}, 1, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_IF , {{13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {13, 0, 15, 0, 7, 0, 0}, 0, 0, 0, 1, 0, 0, 0, 15, 0, 0, 0}, + {OPCODE_MOV , {{0, 0, 1755, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {0, 0, 1, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_RET , {{13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {13, 0, 15, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_ENDIF , {{13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {13, 0, 15, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_MOV , {{0, 0, 1170, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {0, 0, 1, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_RET , {{13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {13, 0, 15, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_ENDSUB , {{13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {13, 0, 15, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0} +}; +float noise1_const[2][4] = { + {0.300000f, 0.900000f, 0.500000f, 0.300000f} +}; + +COMPILED_SUB noise1_presub = { + &(noise1_insts[0]), + 12, + 2, + 1, + 0, + &(noise1_const[0]), + SWIZZLE_X, + SWIZZLE_X, + SWIZZLE_X, + SWIZZLE_X, + {0,0,0}, + 0 +}; + BITS addrmode_PVSDST(PVSDST * pPVSDST) { return pPVSDST->addrmode0 | ((BITS)pPVSDST->addrmode1 << 1); @@ -327,22 +364,27 @@ GLuint GetSurfaceFormat(GLenum eType, GLuint nChannels, GLuint * pClient_size) return(format); } -unsigned int r700GetNumOperands(r700_AssemblerBase* pAsm) +unsigned int r700GetNumOperands(GLuint opcode, GLuint nIsOp3) { - if(pAsm->D.dst.op3) + if(nIsOp3 > 0) { return 3; } - switch (pAsm->D.dst.opcode) + switch (opcode) { case SQ_OP2_INST_ADD: + case SQ_OP2_INST_KILLE: case SQ_OP2_INST_KILLGT: + case SQ_OP2_INST_KILLGE: + case SQ_OP2_INST_KILLNE: case SQ_OP2_INST_MUL: case SQ_OP2_INST_MAX: case SQ_OP2_INST_MIN: //case SQ_OP2_INST_MAX_DX10: //case SQ_OP2_INST_MIN_DX10: + case SQ_OP2_INST_SETE: + case SQ_OP2_INST_SETNE: case SQ_OP2_INST_SETGT: case SQ_OP2_INST_SETGE: case SQ_OP2_INST_PRED_SETE: @@ -358,6 +400,7 @@ unsigned int r700GetNumOperands(r700_AssemblerBase* pAsm) case SQ_OP2_INST_MOVA_FLOOR: case SQ_OP2_INST_FRACT: case SQ_OP2_INST_FLOOR: + case SQ_OP2_INST_TRUNC: case SQ_OP2_INST_EXP_IEEE: case SQ_OP2_INST_LOG_CLAMPED: case SQ_OP2_INST_LOG_IEEE: @@ -369,7 +412,7 @@ unsigned int r700GetNumOperands(r700_AssemblerBase* pAsm) return 1; default: radeon_error( - "Need instruction operand number for %x.\n", pAsm->D.dst.opcode); + "Need instruction operand number for %x.\n", opcode); }; return 3; @@ -383,103 +426,128 @@ int Init_r700_AssemblerBase(SHADER_PIPE_TYPE spt, r700_AssemblerBase* pAsm, R700 pAsm->pR700Shader = pShader; pAsm->currentShaderType = spt; - pAsm->cf_last_export_ptr = NULL; + pAsm->cf_last_export_ptr = NULL; + + pAsm->cf_current_export_clause_ptr = NULL; + pAsm->cf_current_alu_clause_ptr = NULL; + pAsm->cf_current_tex_clause_ptr = NULL; + pAsm->cf_current_vtx_clause_ptr = NULL; + pAsm->cf_current_cf_clause_ptr = NULL; - pAsm->cf_current_export_clause_ptr = NULL; - pAsm->cf_current_alu_clause_ptr = NULL; - pAsm->cf_current_tex_clause_ptr = NULL; - pAsm->cf_current_vtx_clause_ptr = NULL; - pAsm->cf_current_cf_clause_ptr = NULL; + // No clause has been created yet + pAsm->cf_current_clause_type = CF_EMPTY_CLAUSE; - // No clause has been created yet - pAsm->cf_current_clause_type = CF_EMPTY_CLAUSE; + pAsm->number_of_colorandz_exports = 0; + pAsm->number_of_exports = 0; + pAsm->number_of_export_opcodes = 0; - pAsm->number_of_colorandz_exports = 0; - pAsm->number_of_exports = 0; - pAsm->number_of_export_opcodes = 0; + pAsm->alu_x_opcode = 0; + pAsm->D2.bits = 0; - pAsm->D.bits = 0; - pAsm->S[0].bits = 0; - pAsm->S[1].bits = 0; - pAsm->S[2].bits = 0; + pAsm->D.bits = 0; + pAsm->S[0].bits = 0; + pAsm->S[1].bits = 0; + pAsm->S[2].bits = 0; - pAsm->uLastPosUpdate = 0; + pAsm->uLastPosUpdate = 0; - *(BITS *) &pAsm->fp_stOutFmt0 = 0; + *(BITS *) &pAsm->fp_stOutFmt0 = 0; - pAsm->uIIns = 0; - pAsm->uOIns = 0; - pAsm->number_used_registers = 0; - pAsm->uUsedConsts = 256; + pAsm->uIIns = 0; + pAsm->uOIns = 0; + pAsm->number_used_registers = 0; + pAsm->uUsedConsts = 256; - // Fragment programs - pAsm->uBoolConsts = 0; - pAsm->uIntConsts = 0; - pAsm->uInsts = 0; - pAsm->uConsts = 0; + // Fragment programs + pAsm->uBoolConsts = 0; + pAsm->uIntConsts = 0; + pAsm->uInsts = 0; + pAsm->uConsts = 0; - pAsm->FCSP = 0; - pAsm->fc_stack[0].type = FC_NONE; + pAsm->FCSP = 0; + pAsm->fc_stack[0].type = FC_NONE; - pAsm->branch_depth = 0; - pAsm->max_branch_depth = 0; + pAsm->aArgSubst[0] = + pAsm->aArgSubst[1] = + pAsm->aArgSubst[2] = + pAsm->aArgSubst[3] = (-1); - pAsm->aArgSubst[0] = - pAsm->aArgSubst[1] = - pAsm->aArgSubst[2] = - pAsm->aArgSubst[3] = (-1); + pAsm->uOutputs = 0; - pAsm->uOutputs = 0; + for (i=0; i<NUMBER_OF_OUTPUT_COLORS; i++) + { + pAsm->color_export_register_number[i] = (-1); + } - for (i=0; i<NUMBER_OF_OUTPUT_COLORS; i++) - { - pAsm->color_export_register_number[i] = (-1); - } + pAsm->depth_export_register_number = (-1); + pAsm->stencil_export_register_number = (-1); + pAsm->coverage_to_mask_export_register_number = (-1); + pAsm->mask_export_register_number = (-1); - pAsm->depth_export_register_number = (-1); - pAsm->stencil_export_register_number = (-1); - pAsm->coverage_to_mask_export_register_number = (-1); - pAsm->mask_export_register_number = (-1); + pAsm->starting_export_register_number = 0; + pAsm->starting_vfetch_register_number = 0; + pAsm->starting_temp_register_number = 0; + pAsm->uFirstHelpReg = 0; - pAsm->starting_export_register_number = 0; - pAsm->starting_vfetch_register_number = 0; - pAsm->starting_temp_register_number = 0; - pAsm->uFirstHelpReg = 0; + pAsm->input_position_is_used = GL_FALSE; + pAsm->input_normal_is_used = GL_FALSE; + for (i=0; i<NUMBER_OF_INPUT_COLORS; i++) + { + pAsm->input_color_is_used[ i ] = GL_FALSE; + } - pAsm->input_position_is_used = GL_FALSE; - pAsm->input_normal_is_used = GL_FALSE; + for (i=0; i<NUMBER_OF_TEXTURE_UNITS; i++) + { + pAsm->input_texture_unit_is_used[ i ] = GL_FALSE; + } + for (i=0; i<VERT_ATTRIB_MAX; i++) + { + pAsm->vfetch_instruction_ptr_array[ i ] = NULL; + } - for (i=0; i<NUMBER_OF_INPUT_COLORS; i++) - { - pAsm->input_color_is_used[ i ] = GL_FALSE; - } + pAsm->number_of_inputs = 0; - for (i=0; i<NUMBER_OF_TEXTURE_UNITS; i++) - { - pAsm->input_texture_unit_is_used[ i ] = GL_FALSE; - } + pAsm->is_tex = GL_FALSE; + pAsm->need_tex_barrier = GL_FALSE; - for (i=0; i<VERT_ATTRIB_MAX; i++) - { - pAsm->vfetch_instruction_ptr_array[ i ] = NULL; - } + pAsm->subs = NULL; + pAsm->unSubArraySize = 0; + pAsm->unSubArrayPointer = 0; + pAsm->callers = NULL; + pAsm->unCallerArraySize = 0; + pAsm->unCallerArrayPointer = 0; + + pAsm->CALLSP = 0; + pAsm->CALLSTACK[0].FCSP_BeforeEntry = 0; + pAsm->CALLSTACK[0].plstCFInstructions_local + = &(pAsm->pR700Shader->lstCFInstructions); - pAsm->number_of_inputs = 0; + pAsm->CALLSTACK[0].max = 0; + pAsm->CALLSTACK[0].current = 0; - pAsm->is_tex = GL_FALSE; - pAsm->need_tex_barrier = GL_FALSE; + SetActiveCFlist(pAsm->pR700Shader, pAsm->CALLSTACK[0].plstCFInstructions_local); - return 0; + pAsm->unCFflags = 0; + + pAsm->presubs = NULL; + pAsm->unPresubArraySize = 0; + pAsm->unNumPresub = 0; + pAsm->unCurNumILInsts = 0; + + pAsm->unVetTexBits = 0; + + return 0; } GLboolean IsTex(gl_inst_opcode Opcode) { - if( (OPCODE_TEX==Opcode) || (OPCODE_TXP==Opcode) || (OPCODE_TXB==Opcode) ) + if( (OPCODE_TEX==Opcode) || (OPCODE_TXP==Opcode) || (OPCODE_TXB==Opcode) || + (OPCODE_DDX==Opcode) || (OPCODE_DDY==Opcode) ) { return GL_TRUE; } @@ -592,6 +660,31 @@ int check_current_clause(r700_AssemblerBase* pAsm, return GL_TRUE; } +GLboolean add_cf_instruction(r700_AssemblerBase* pAsm) +{ + if(GL_FALSE == check_current_clause(pAsm, CF_OTHER_CLAUSE)) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr = + (R700ControlFlowGenericClause*) CALLOC_STRUCT(R700ControlFlowGenericClause); + + if (pAsm->cf_current_cf_clause_ptr != NULL) + { + Init_R700ControlFlowGenericClause(pAsm->cf_current_cf_clause_ptr); + AddCFInstruction( pAsm->pR700Shader, + (R700ControlFlowInstruction *)pAsm->cf_current_cf_clause_ptr ); + } + else + { + radeon_error("Could not allocate a new VFetch CF instruction.\n"); + return GL_FALSE; + } + + return GL_TRUE; +} + GLboolean add_vfetch_instruction(r700_AssemblerBase* pAsm, R700VertexInstruction* vertex_instruction_ptr) { @@ -987,7 +1080,8 @@ GLboolean checkop2(r700_AssemblerBase* pAsm) checkop_init(pAsm); - if( (pILInst->SrcReg[0].File == PROGRAM_CONSTANT) || + if( (pILInst->SrcReg[0].File == PROGRAM_UNIFORM) || + (pILInst->SrcReg[0].File == PROGRAM_CONSTANT) || (pILInst->SrcReg[0].File == PROGRAM_LOCAL_PARAM) || (pILInst->SrcReg[0].File == PROGRAM_ENV_PARAM) || (pILInst->SrcReg[0].File == PROGRAM_STATE_VAR) ) @@ -998,7 +1092,8 @@ GLboolean checkop2(r700_AssemblerBase* pAsm) { bSrcConst[0] = GL_FALSE; } - if( (pILInst->SrcReg[1].File == PROGRAM_CONSTANT) || + if( (pILInst->SrcReg[1].File == PROGRAM_UNIFORM) || + (pILInst->SrcReg[1].File == PROGRAM_CONSTANT) || (pILInst->SrcReg[1].File == PROGRAM_LOCAL_PARAM) || (pILInst->SrcReg[1].File == PROGRAM_ENV_PARAM) || (pILInst->SrcReg[1].File == PROGRAM_STATE_VAR) ) @@ -1031,7 +1126,8 @@ GLboolean checkop3(r700_AssemblerBase* pAsm) checkop_init(pAsm); - if( (pILInst->SrcReg[0].File == PROGRAM_CONSTANT) || + if( (pILInst->SrcReg[0].File == PROGRAM_UNIFORM) || + (pILInst->SrcReg[0].File == PROGRAM_CONSTANT) || (pILInst->SrcReg[0].File == PROGRAM_LOCAL_PARAM) || (pILInst->SrcReg[0].File == PROGRAM_ENV_PARAM) || (pILInst->SrcReg[0].File == PROGRAM_STATE_VAR) ) @@ -1042,7 +1138,8 @@ GLboolean checkop3(r700_AssemblerBase* pAsm) { bSrcConst[0] = GL_FALSE; } - if( (pILInst->SrcReg[1].File == PROGRAM_CONSTANT) || + if( (pILInst->SrcReg[1].File == PROGRAM_UNIFORM) || + (pILInst->SrcReg[1].File == PROGRAM_CONSTANT) || (pILInst->SrcReg[1].File == PROGRAM_LOCAL_PARAM) || (pILInst->SrcReg[1].File == PROGRAM_ENV_PARAM) || (pILInst->SrcReg[1].File == PROGRAM_STATE_VAR) ) @@ -1053,7 +1150,8 @@ GLboolean checkop3(r700_AssemblerBase* pAsm) { bSrcConst[1] = GL_FALSE; } - if( (pILInst->SrcReg[2].File == PROGRAM_CONSTANT) || + if( (pILInst->SrcReg[2].File == PROGRAM_UNIFORM) || + (pILInst->SrcReg[2].File == PROGRAM_CONSTANT) || (pILInst->SrcReg[2].File == PROGRAM_LOCAL_PARAM) || (pILInst->SrcReg[2].File == PROGRAM_ENV_PARAM) || (pILInst->SrcReg[2].File == PROGRAM_STATE_VAR) ) @@ -1153,6 +1251,7 @@ GLboolean assemble_src(r700_AssemblerBase *pAsm, case PROGRAM_LOCAL_PARAM: case PROGRAM_ENV_PARAM: case PROGRAM_STATE_VAR: + case PROGRAM_UNIFORM: if (1 == pILInst->SrcReg[src].RelAddr) { setaddrmode_PVSSRC(&(pAsm->S[fld].src), ADDR_RELATIVE_A0); @@ -1163,10 +1262,18 @@ GLboolean assemble_src(r700_AssemblerBase *pAsm, } pAsm->S[fld].src.rtype = SRC_REG_CONSTANT; - pAsm->S[fld].src.reg = pILInst->SrcReg[src].Index; + if(pILInst->SrcReg[src].Index < 0) + { + WARN_ONCE("Negative register offsets not supported yet!\n"); + pAsm->S[fld].src.reg = 0; + } + else + { + pAsm->S[fld].src.reg = pILInst->SrcReg[src].Index; + } break; case PROGRAM_INPUT: - setaddrmode_PVSSRC(&(pAsm->S[fld].src), ADDR_ABSOLUTE); + setaddrmode_PVSSRC(&(pAsm->S[fld].src), ADDR_ABSOLUTE); pAsm->S[fld].src.rtype = SRC_REG_INPUT; switch (pAsm->currentShaderType) { @@ -1179,7 +1286,7 @@ GLboolean assemble_src(r700_AssemblerBase *pAsm, } break; default: - radeon_error("Invalid source argument type\n"); + radeon_error("Invalid source argument type : %d \n", pILInst->SrcReg[src].File); return GL_FALSE; } } @@ -1235,6 +1342,15 @@ GLboolean assemble_dst(r700_AssemblerBase *pAsm) pAsm->D.dst.writez = (pILInst->DstReg.WriteMask >> 2) & 0x1; pAsm->D.dst.writew = (pILInst->DstReg.WriteMask >> 3) & 0x1; + if(pILInst->SaturateMode == SATURATE_ZERO_ONE) + { + pAsm->D2.dst2.SaturateMode = 1; + } + else + { + pAsm->D2.dst2.SaturateMode = 0; + } + return GL_TRUE; } @@ -1294,6 +1410,7 @@ GLboolean tex_src(r700_AssemblerBase *pAsm) else { switch (pILInst->SrcReg[0].File) { + case PROGRAM_UNIFORM: case PROGRAM_CONSTANT: case PROGRAM_LOCAL_PARAM: case PROGRAM_ENV_PARAM: @@ -1306,36 +1423,65 @@ GLboolean tex_src(r700_AssemblerBase *pAsm) pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; break; case PROGRAM_INPUT: - switch (pILInst->SrcReg[0].Index) + if(SPT_VP == pAsm->currentShaderType) { - case FRAG_ATTRIB_WPOS: - case FRAG_ATTRIB_COL0: - case FRAG_ATTRIB_COL1: - case FRAG_ATTRIB_FOGC: - case FRAG_ATTRIB_TEX0: - case FRAG_ATTRIB_TEX1: - case FRAG_ATTRIB_TEX2: - case FRAG_ATTRIB_TEX3: - case FRAG_ATTRIB_TEX4: - case FRAG_ATTRIB_TEX5: - case FRAG_ATTRIB_TEX6: - case FRAG_ATTRIB_TEX7: - bValidTexCoord = GL_TRUE; + switch (pILInst->SrcReg[0].Index) + { + case VERT_ATTRIB_TEX0: + case VERT_ATTRIB_TEX1: + case VERT_ATTRIB_TEX2: + case VERT_ATTRIB_TEX3: + case VERT_ATTRIB_TEX4: + case VERT_ATTRIB_TEX5: + case VERT_ATTRIB_TEX6: + case VERT_ATTRIB_TEX7: + bValidTexCoord = GL_TRUE; + pAsm->S[0].src.reg = + pAsm->ucVP_AttributeMap[pILInst->SrcReg[0].Index]; + pAsm->S[0].src.rtype = SRC_REG_INPUT; + break; + } + } + else + { + switch (pILInst->SrcReg[0].Index) + { + case FRAG_ATTRIB_WPOS: + case FRAG_ATTRIB_COL0: + case FRAG_ATTRIB_COL1: + case FRAG_ATTRIB_FOGC: + case FRAG_ATTRIB_TEX0: + case FRAG_ATTRIB_TEX1: + case FRAG_ATTRIB_TEX2: + case FRAG_ATTRIB_TEX3: + case FRAG_ATTRIB_TEX4: + case FRAG_ATTRIB_TEX5: + case FRAG_ATTRIB_TEX6: + case FRAG_ATTRIB_TEX7: + bValidTexCoord = GL_TRUE; + pAsm->S[0].src.reg = + pAsm->uiFP_AttributeMap[pILInst->SrcReg[0].Index]; + pAsm->S[0].src.rtype = SRC_REG_INPUT; + break; + case FRAG_ATTRIB_FACE: + fprintf(stderr, "FRAG_ATTRIB_FACE unsupported\n"); + break; + case FRAG_ATTRIB_PNTC: + fprintf(stderr, "FRAG_ATTRIB_PNTC unsupported\n"); + break; + } + + if( (pILInst->SrcReg[0].Index >= FRAG_ATTRIB_VAR0) || + (pILInst->SrcReg[0].Index < FRAG_ATTRIB_MAX) ) + { + bValidTexCoord = GL_TRUE; pAsm->S[0].src.reg = pAsm->uiFP_AttributeMap[pILInst->SrcReg[0].Index]; pAsm->S[0].src.rtype = SRC_REG_INPUT; - break; - case FRAG_ATTRIB_FACE: - fprintf(stderr, "FRAG_ATTRIB_FACE unsupported\n"); - break; - case FRAG_ATTRIB_PNTC: - fprintf(stderr, "FRAG_ATTRIB_PNTC unsupported\n"); - break; - case FRAG_ATTRIB_VAR0: - fprintf(stderr, "FRAG_ATTRIB_VAR0 unsupported\n"); - break; + } } - break; + + break; } } @@ -1380,8 +1526,17 @@ GLboolean assemble_tex_instruction(r700_AssemblerBase *pAsm, GLboolean normalize tex_instruction_ptr->m_Word0.f.tex_inst = pAsm->D.dst.opcode; tex_instruction_ptr->m_Word0.f.bc_frac_mode = 0x0; tex_instruction_ptr->m_Word0.f.fetch_whole_quad = 0x0; + tex_instruction_ptr->m_Word0.f.alt_const = 0; - tex_instruction_ptr->m_Word0.f.resource_id = texture_unit_source->reg; + if(SPT_VP == pAsm->currentShaderType) + { + tex_instruction_ptr->m_Word0.f.resource_id = texture_unit_source->reg + VERT_ATTRIB_MAX; + pAsm->unVetTexBits |= 1 << texture_unit_source->reg; + } + else + { + tex_instruction_ptr->m_Word0.f.resource_id = texture_unit_source->reg; + } tex_instruction_ptr->m_Word1.f.lod_bias = 0x0; if (normalized) { @@ -1400,7 +1555,6 @@ GLboolean assemble_tex_instruction(r700_AssemblerBase *pAsm, GLboolean normalize tex_instruction_ptr->m_Word2.f.offset_x = 0x0; tex_instruction_ptr->m_Word2.f.offset_y = 0x0; tex_instruction_ptr->m_Word2.f.offset_z = 0x0; - tex_instruction_ptr->m_Word2.f.sampler_id = texture_unit_source->reg; // dst @@ -1517,6 +1671,10 @@ GLboolean assemble_alu_src(R700ALUInstruction* alu_instruction_ptr, { src_sel = pSource->reg + CFILE_REGISTER_OFFSET; } + else if (pSource->rtype == SRC_REC_LITERAL) + { + src_sel = SQ_ALU_SRC_LITERAL; + } else { radeon_error("Source (%d) register type (%d) not one of TEMP, INPUT, or CONSTANT.\n", @@ -1606,7 +1764,8 @@ GLboolean add_alu_instruction(r700_AssemblerBase* pAsm, return GL_FALSE; } - if ( pAsm->cf_current_alu_clause_ptr == NULL || + if ( pAsm->alu_x_opcode != 0 || + pAsm->cf_current_alu_clause_ptr == NULL || ( (pAsm->cf_current_alu_clause_ptr != NULL) && (pAsm->cf_current_alu_clause_ptr->m_Word1.f.count >= (GetCFMaxInstructions(pAsm->cf_current_alu_clause_ptr->m_ShaderInstType)-contiguous_slots_needed-1) ) ) ) @@ -1636,9 +1795,17 @@ GLboolean add_alu_instruction(r700_AssemblerBase* pAsm, pAsm->cf_current_alu_clause_ptr->m_Word1.f.kcache_addr0 = 0x0; pAsm->cf_current_alu_clause_ptr->m_Word1.f.kcache_addr1 = 0x0; - //cf_current_alu_clause_ptr->m_Word1.f.count = number_of_scalar_operations - 1; pAsm->cf_current_alu_clause_ptr->m_Word1.f.count = 0x0; - pAsm->cf_current_alu_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_ALU; + + if(pAsm->alu_x_opcode != 0) + { + pAsm->cf_current_alu_clause_ptr->m_Word1.f.cf_inst = pAsm->alu_x_opcode; + pAsm->alu_x_opcode = 0; + } + else + { + pAsm->cf_current_alu_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_ALU; + } pAsm->cf_current_alu_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; @@ -1646,7 +1813,7 @@ GLboolean add_alu_instruction(r700_AssemblerBase* pAsm, } else { - pAsm->cf_current_alu_clause_ptr->m_Word1.f.count++; + pAsm->cf_current_alu_clause_ptr->m_Word1.f.count += (GetInstructionSize(alu_instruction_ptr->m_ShaderInstType) / 2); } // If this clause constains any instruction that is forward dependent on a TEX instruction, @@ -1923,7 +2090,7 @@ GLboolean check_scalar(r700_AssemblerBase* pAsm, GLuint swizzle_key; - GLuint number_of_operands = r700GetNumOperands(pAsm); + GLuint number_of_operands = r700GetNumOperands(pAsm->D.dst.opcode, pAsm->D.dst.op3); for (src=0; src<number_of_operands; src++) { @@ -2012,7 +2179,7 @@ GLboolean check_vector(r700_AssemblerBase* pAsm, GLuint swizzle_key; - GLuint number_of_operands = r700GetNumOperands(pAsm); + GLuint number_of_operands = r700GetNumOperands(pAsm->D.dst.opcode, pAsm->D.dst.op3); for (src=0; src<number_of_operands; src++) { @@ -2045,7 +2212,7 @@ GLboolean check_vector(r700_AssemblerBase* pAsm, if( is_gpr(sel) ) { if( GL_FALSE == cycle_for_vector_bank_swizzle(bank_swizzle, src, &cycle) ) - { + { return GL_FALSE; } @@ -2057,7 +2224,7 @@ GLboolean check_vector(r700_AssemblerBase* pAsm, else { if( GL_FALSE == reserve_gpr(pAsm, sel, chan, cycle) ) - { + { return GL_FALSE; } } @@ -2069,7 +2236,7 @@ GLboolean check_vector(r700_AssemblerBase* pAsm, if( is_cfile(sel) ) { if( GL_FALSE == reserve_cfile(pAsm, sel, chan) ) - { + { return GL_FALSE; } } @@ -2081,6 +2248,10 @@ GLboolean check_vector(r700_AssemblerBase* pAsm, GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) { + R700ALUInstruction * alu_instruction_ptr; + R700ALUInstructionHalfLiteral * alu_instruction_ptr_hl; + R700ALUInstructionFullLiteral * alu_instruction_ptr_fl; + GLuint number_of_scalar_operations; GLboolean is_single_scalar_operation; GLuint scalar_channel_index; @@ -2089,7 +2260,7 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) int current_source_index; GLuint contiguous_slots_needed; - GLuint uNumSrc = r700GetNumOperands(pAsm); + GLuint uNumSrc = r700GetNumOperands(pAsm->D.dst.opcode, pAsm->D.dst.op3); //GLuint channel_swizzle, j; //GLuint chan_counter[4] = {0, 0, 0, 0}; //PVSSRC * pSource[3]; @@ -2146,23 +2317,44 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) contiguous_slots_needed = 0; - if(GL_TRUE == is_reduction_opcode(&(pAsm->D)) ) + if(!is_single_scalar_operation) { contiguous_slots_needed = 4; } + contiguous_slots_needed += pAsm->D2.dst2.literal_slots; + initialize(pAsm); for (scalar_channel_index=0; scalar_channel_index < number_of_scalar_operations; scalar_channel_index++) { - R700ALUInstruction* alu_instruction_ptr = (R700ALUInstruction*) CALLOC_STRUCT(R700ALUInstruction); - if (alu_instruction_ptr == NULL) - { - return GL_FALSE; - } - Init_R700ALUInstruction(alu_instruction_ptr); + if(scalar_channel_index == (number_of_scalar_operations-1)) + { + switch(pAsm->D2.dst2.literal_slots) + { + case 0: + alu_instruction_ptr = (R700ALUInstruction*) CALLOC_STRUCT(R700ALUInstruction); + Init_R700ALUInstruction(alu_instruction_ptr); + break; + case 1: + alu_instruction_ptr_hl = (R700ALUInstructionHalfLiteral*) CALLOC_STRUCT(R700ALUInstructionHalfLiteral); + Init_R700ALUInstructionHalfLiteral(alu_instruction_ptr_hl, pAsm->C[0].f, pAsm->C[1].f); + alu_instruction_ptr = (R700ALUInstruction*)alu_instruction_ptr_hl; + break; + case 2: + alu_instruction_ptr_fl = (R700ALUInstructionFullLiteral*) CALLOC_STRUCT(R700ALUInstructionFullLiteral); + Init_R700ALUInstructionFullLiteral(alu_instruction_ptr_fl,pAsm->C[0].f, pAsm->C[1].f, pAsm->C[2].f, pAsm->C[3].f); + alu_instruction_ptr = (R700ALUInstruction*)alu_instruction_ptr_fl; + break; + }; + } + else + { + alu_instruction_ptr = (R700ALUInstruction*) CALLOC_STRUCT(R700ALUInstruction); + Init_R700ALUInstruction(alu_instruction_ptr); + } //src 0 current_source_index = 0; @@ -2172,7 +2364,7 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) current_source_index, pcurrent_source, scalar_channel_index) ) - { + { return GL_FALSE; } @@ -2186,13 +2378,13 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) current_source_index, pcurrent_source, scalar_channel_index) ) - { + { return GL_FALSE; } } //other bits - alu_instruction_ptr->m_Word0.f.index_mode = SQ_INDEX_AR_X; + alu_instruction_ptr->m_Word0.f.index_mode = pAsm->D2.dst2.index_mode; if( (is_single_scalar_operation == GL_TRUE) || (GL_TRUE == bSplitInst) ) @@ -2204,9 +2396,17 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) alu_instruction_ptr->m_Word0.f.last = (scalar_channel_index == 3) ? 1 : 0; } - alu_instruction_ptr->m_Word0.f.pred_sel = 0x0; - alu_instruction_ptr->m_Word1_OP2.f.update_pred = 0x0; - alu_instruction_ptr->m_Word1_OP2.f.update_execute_mask = 0x0; + alu_instruction_ptr->m_Word0.f.pred_sel = (pAsm->D.dst.pred_inv > 0) ? 1 : 0; + if(1 == pAsm->D.dst.predicated) + { + alu_instruction_ptr->m_Word1_OP2.f.update_pred = 0x1; + alu_instruction_ptr->m_Word1_OP2.f.update_execute_mask = 0x1; + } + else + { + alu_instruction_ptr->m_Word1_OP2.f.update_pred = 0x0; + alu_instruction_ptr->m_Word1_OP2.f.update_execute_mask = 0x0; + } // dst if( (pAsm->D.dst.rtype == DST_REG_TEMPORARY) || @@ -2215,7 +2415,7 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) alu_instruction_ptr->m_Word1.f.dst_gpr = pAsm->D.dst.reg; } else - { + { radeon_error("Only temp destination registers supported for ALU dest regs.\n"); return GL_FALSE; } @@ -2245,7 +2445,7 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) alu_instruction_ptr->m_Word1.f.dst_chan = scalar_channel_index; - alu_instruction_ptr->m_Word1.f.clamp = pAsm->pILInst[pAsm->uiCurInst].SaturateMode; + alu_instruction_ptr->m_Word1.f.clamp = pAsm->D2.dst2.SaturateMode; if (pAsm->D.dst.op3) { @@ -2272,8 +2472,8 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) { alu_instruction_ptr->m_Word1_OP2.f6.alu_inst = pAsm->D.dst.opcode; - alu_instruction_ptr->m_Word1_OP2.f6.src0_abs = 0x0; - alu_instruction_ptr->m_Word1_OP2.f6.src1_abs = 0x0; + alu_instruction_ptr->m_Word1_OP2.f6.src0_abs = pAsm->S[0].src.abs; + alu_instruction_ptr->m_Word1_OP2.f6.src1_abs = pAsm->S[1].src.abs; //alu_instruction_ptr->m_Word1_OP2.f6.update_execute_mask = 0x0; //alu_instruction_ptr->m_Word1_OP2.f6.update_pred = 0x0; @@ -2301,8 +2501,8 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) { alu_instruction_ptr->m_Word1_OP2.f.alu_inst = pAsm->D.dst.opcode; - alu_instruction_ptr->m_Word1_OP2.f.src0_abs = 0x0; - alu_instruction_ptr->m_Word1_OP2.f.src1_abs = 0x0; + alu_instruction_ptr->m_Word1_OP2.f.src0_abs = pAsm->S[0].src.abs; + alu_instruction_ptr->m_Word1_OP2.f.src1_abs = pAsm->S[1].src.abs; //alu_instruction_ptr->m_Word1_OP2.f.update_execute_mask = 0x0; //alu_instruction_ptr->m_Word1_OP2.f.update_pred = 0x0; @@ -2329,7 +2529,7 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) } if(GL_FALSE == add_alu_instruction(pAsm, alu_instruction_ptr, contiguous_slots_needed) ) - { + { return GL_FALSE; } @@ -2340,19 +2540,19 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) if (is_single_scalar_operation) { if(GL_FALSE == check_scalar(pAsm, alu_instruction_ptr) ) - { + { return GL_FALSE; } } else { if(GL_FALSE == check_vector(pAsm, alu_instruction_ptr) ) - { - return 1; + { + return GL_FALSE; } } - contiguous_slots_needed = 0; + contiguous_slots_needed -= 1; } return GL_TRUE; @@ -2403,11 +2603,14 @@ GLboolean next_ins(r700_AssemblerBase *pAsm) //reset for next inst. pAsm->D.bits = 0; + pAsm->D2.bits = 0; pAsm->S[0].bits = 0; pAsm->S[1].bits = 0; pAsm->S[2].bits = 0; pAsm->is_tex = GL_FALSE; pAsm->need_tex_barrier = GL_FALSE; + pAsm->D2.bits = 0; + pAsm->C[0].bits = pAsm->C[1].bits = pAsm->C[2].bits = pAsm->C[3].bits = 0; return GL_TRUE; } @@ -2640,9 +2843,44 @@ GLboolean assemble_CMP(r700_AssemblerBase *pAsm) return GL_TRUE; } -GLboolean assemble_COS(r700_AssemblerBase *pAsm) +GLboolean assemble_TRIG(r700_AssemblerBase *pAsm, BITS opcode) { - return assemble_math_function(pAsm, SQ_OP2_INST_COS); + int tmp; + checkop1(pAsm); + + tmp = gethelpr(pAsm); + + pAsm->D.dst.opcode = SQ_OP2_INST_MUL; + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = tmp; + pAsm->D.dst.writex = 1; + + assemble_src(pAsm, 0, -1); + + pAsm->S[1].src.rtype = SRC_REC_LITERAL; + setswizzle_PVSSRC(&(pAsm->S[1].src), SQ_SEL_X); + pAsm->D2.dst2.literal_slots = 1; + pAsm->C[0].f = 1/(3.1415926535 * 2); + pAsm->C[1].f = 0.0F; + next_ins(pAsm); + + pAsm->D.dst.opcode = opcode; + pAsm->D.dst.math = 1; + + assemble_dst(pAsm); + + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; + pAsm->S[0].src.reg = tmp; + setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X); + noneg_PVSSRC(&(pAsm->S[0].src)); + + next_ins(pAsm); + + //TODO - replicate if more channels set in WriteMask + return GL_TRUE; + } GLboolean assemble_DOT(r700_AssemblerBase *pAsm) @@ -2910,13 +3148,15 @@ GLboolean assemble_FRC(r700_AssemblerBase *pAsm) return GL_TRUE; } -GLboolean assemble_KIL(r700_AssemblerBase *pAsm) -{ - /* TODO: doc says KILL has to be last(end) ALU clause */ - - checkop1(pAsm); +GLboolean assemble_KIL(r700_AssemblerBase *pAsm, GLuint opcode) +{ + struct prog_instruction *pILInst = &(pAsm->pILInst[pAsm->uiCurInst]); - pAsm->D.dst.opcode = SQ_OP2_INST_KILLGT; + if(pILInst->Opcode == OPCODE_KIL) + checkop1(pAsm); + + pAsm->D.dst.opcode = opcode; + //pAsm->D.dst.math = 1; setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); pAsm->D.dst.rtype = DST_REG_TEMPORARY; @@ -2929,21 +3169,34 @@ GLboolean assemble_KIL(r700_AssemblerBase *pAsm) setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; pAsm->S[0].src.reg = 0; - setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_0); noneg_PVSSRC(&(pAsm->S[0].src)); - if ( GL_FALSE == assemble_src(pAsm, 0, 1) ) + if(pILInst->Opcode == OPCODE_KIL_NV) { - return GL_FALSE; + setaddrmode_PVSSRC(&(pAsm->S[1].src), ADDR_ABSOLUTE); + pAsm->S[1].src.rtype = SRC_REG_TEMPORARY; + pAsm->S[1].src.reg = 0; + setswizzle_PVSSRC(&(pAsm->S[1].src), SQ_SEL_1); + neg_PVSSRC(&(pAsm->S[1].src)); } - + else + { + if( GL_FALSE == assemble_src(pAsm, 0, 1) ) + { + return GL_FALSE; + } + + } + if ( GL_FALSE == next_ins(pAsm) ) { return GL_FALSE; } + /* Doc says KILL has to be last(end) ALU clause */ pAsm->pR700Shader->killIsUsed = GL_TRUE; + pAsm->alu_x_opcode = SQ_CF_INST_ALU; return GL_TRUE; } @@ -3007,6 +3260,7 @@ GLboolean assemble_LRP(r700_AssemblerBase *pAsm) { return GL_FALSE; } + if( GL_FALSE == assemble_src(pAsm, 2, -1) ) { return GL_FALSE; @@ -3742,77 +3996,137 @@ GLboolean assemble_RSQ(r700_AssemblerBase *pAsm) return assemble_math_function(pAsm, SQ_OP2_INST_RECIPSQRT_IEEE); } -GLboolean assemble_SIN(r700_AssemblerBase *pAsm) -{ - return assemble_math_function(pAsm, SQ_OP2_INST_SIN); -} - GLboolean assemble_SCS(r700_AssemblerBase *pAsm) { BITS tmp; - checkop1(pAsm); + checkop1(pAsm); - tmp = gethelpr(pAsm); + tmp = gethelpr(pAsm); + /* tmp.x = src /2*PI */ + pAsm->D.dst.opcode = SQ_OP2_INST_MUL; + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = tmp; + pAsm->D.dst.writex = 1; - // COS tmp.x, a.x - pAsm->D.dst.opcode = SQ_OP2_INST_COS; - pAsm->D.dst.math = 1; + assemble_src(pAsm, 0, -1); - setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); - pAsm->D.dst.rtype = DST_REG_TEMPORARY; - pAsm->D.dst.reg = tmp; - pAsm->D.dst.writex = 1; + pAsm->S[1].src.rtype = SRC_REC_LITERAL; + setswizzle_PVSSRC(&(pAsm->S[1].src), SQ_SEL_X); + pAsm->D2.dst2.literal_slots = 1; + pAsm->C[0].f = 1/(3.1415926535 * 2); + pAsm->C[1].f = 0.0F; - if( GL_FALSE == assemble_src(pAsm, 0, -1) ) - { - return GL_FALSE; - } + next_ins(pAsm); - if ( GL_FALSE == next_ins(pAsm) ) - { - return GL_FALSE; - } + // COS dst.x, a.x + pAsm->D.dst.opcode = SQ_OP2_INST_COS; + pAsm->D.dst.math = 1; - // SIN tmp.y, a.x - pAsm->D.dst.opcode = SQ_OP2_INST_SIN; - pAsm->D.dst.math = 1; + assemble_dst(pAsm); + /* mask y */ + pAsm->D.dst.writey = 0; - setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); - pAsm->D.dst.rtype = DST_REG_TEMPORARY; - pAsm->D.dst.reg = tmp; - pAsm->D.dst.writey = 1; + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; + pAsm->S[0].src.reg = tmp; + setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X); + noneg_PVSSRC(&(pAsm->S[0].src)); - if( GL_FALSE == assemble_src(pAsm, 0, -1) ) - { - return GL_FALSE; - } + if ( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } - if( GL_FALSE == next_ins(pAsm) ) - { - return GL_FALSE; - } + // SIN dst.y, a.x + pAsm->D.dst.opcode = SQ_OP2_INST_SIN; + pAsm->D.dst.math = 1; - // MOV dst.mask, tmp - pAsm->D.dst.opcode = SQ_OP2_INST_MOV; + assemble_dst(pAsm); + /* mask x */ + pAsm->D.dst.writex = 0; - if( GL_FALSE == assemble_dst(pAsm) ) - { - return GL_FALSE; - } + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; + pAsm->S[0].src.reg = tmp; + setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X); + noneg_PVSSRC(&(pAsm->S[0].src)); - setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); - pAsm->S[0].src.rtype = DST_REG_TEMPORARY; - pAsm->S[0].src.reg = tmp; + if( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } - noswizzle_PVSSRC(&(pAsm->S[0].src)); - pAsm->S[0].src.swizzlez = SQ_SEL_0; - pAsm->S[0].src.swizzlew = SQ_SEL_0; + return GL_TRUE; +} - if ( GL_FALSE == next_ins(pAsm) ) - { - return GL_FALSE; - } +GLboolean assemble_LOGIC(r700_AssemblerBase *pAsm, BITS opcode) +{ + if( GL_FALSE == checkop2(pAsm) ) + { + return GL_FALSE; + } + + pAsm->D.dst.opcode = opcode; + //pAsm->D.dst.math = 1; + + if( GL_FALSE == assemble_dst(pAsm) ) + { + return GL_FALSE; + } + + if( GL_FALSE == assemble_src(pAsm, 0, -1) ) + { + return GL_FALSE; + } + + if( GL_FALSE == assemble_src(pAsm, 1, -1) ) + { + return GL_FALSE; + } + + if( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } + + return GL_TRUE; +} + +GLboolean assemble_LOGIC_PRED(r700_AssemblerBase *pAsm, BITS opcode) +{ + struct prog_instruction *pILInst = &(pAsm->pILInst[pAsm->uiCurInst]); + + pAsm->D.dst.opcode = opcode; + pAsm->D.dst.math = 1; + pAsm->D.dst.predicated = 1; + + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = pAsm->uHelpReg; + pAsm->D.dst.writex = 1; + pAsm->D.dst.writey = pAsm->D.dst.writez = pAsm->D.dst.writew = 0; + + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; + pAsm->S[0].src.reg = pAsm->last_cond_register + pAsm->starting_temp_register_number; + pAsm->S[0].src.swizzlex = pILInst->DstReg.CondSwizzle & 0x7; + noneg_PVSSRC(&(pAsm->S[0].src)); + + pAsm->S[1].src.rtype = SRC_REG_TEMPORARY; + pAsm->S[1].src.reg = pAsm->uHelpReg; + setaddrmode_PVSSRC(&(pAsm->S[1].src), ADDR_ABSOLUTE); + noneg_PVSSRC(&(pAsm->S[1].src)); + pAsm->S[1].src.swizzlex = SQ_SEL_0; + pAsm->S[1].src.swizzley = SQ_SEL_0; + pAsm->S[1].src.swizzlez = SQ_SEL_0; + pAsm->S[1].src.swizzlew = SQ_SEL_0; + + if( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } return GL_TRUE; } @@ -3895,6 +4209,7 @@ GLboolean assemble_TEX(r700_AssemblerBase *pAsm) switch (pAsm->pILInst[pAsm->uiCurInst].SrcReg[0].File) { + case PROGRAM_UNIFORM: case PROGRAM_CONSTANT: case PROGRAM_LOCAL_PARAM: case PROGRAM_ENV_PARAM: @@ -3992,24 +4307,6 @@ GLboolean assemble_TEX(r700_AssemblerBase *pAsm) return GL_FALSE; } - /* tmp1.z = ABS(tmp1.z) dont have abs support in assembler currently - * have to do explicit instruction - */ - pAsm->D.dst.opcode = SQ_OP2_INST_MAX; - setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); - pAsm->D.dst.rtype = DST_REG_TEMPORARY; - pAsm->D.dst.reg = tmp1; - pAsm->D.dst.writez = 1; - - setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); - pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; - pAsm->S[0].src.reg = tmp1; - noswizzle_PVSSRC(&(pAsm->S[0].src)); - pAsm->S[1].bits = pAsm->S[0].bits; - flipneg_PVSSRC(&(pAsm->S[1].src)); - - next_ins(pAsm); - /* tmp1.z = RCP_e(|tmp1.z|) */ pAsm->D.dst.opcode = SQ_OP2_INST_RECIP_IEEE; pAsm->D.dst.math = 1; @@ -4022,13 +4319,13 @@ GLboolean assemble_TEX(r700_AssemblerBase *pAsm) pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; pAsm->S[0].src.reg = tmp1; pAsm->S[0].src.swizzlex = SQ_SEL_Z; + pAsm->S[0].src.abs = 1; next_ins(pAsm); /* MULADD R0.x, R0.x, PS1, (0x3FC00000, 1.5f).x * MULADD R0.y, R0.y, PS1, (0x3FC00000, 1.5f).x * muladd has no writemask, have to use another temp - * also no support for imm constants, so add 1 here */ pAsm->D.dst.opcode = SQ_OP3_INST_MULADD; pAsm->D.dst.op3 = 1; @@ -4045,30 +4342,12 @@ GLboolean assemble_TEX(r700_AssemblerBase *pAsm) pAsm->S[1].src.reg = tmp1; setswizzle_PVSSRC(&(pAsm->S[1].src), SQ_SEL_Z); setaddrmode_PVSSRC(&(pAsm->S[2].src), ADDR_ABSOLUTE); - pAsm->S[2].src.rtype = SRC_REG_TEMPORARY; + /* immediate c 1.5 */ + pAsm->D2.dst2.literal_slots = 1; + pAsm->C[0].f = 1.5F; + pAsm->S[2].src.rtype = SRC_REC_LITERAL; pAsm->S[2].src.reg = tmp1; - setswizzle_PVSSRC(&(pAsm->S[2].src), SQ_SEL_1); - - next_ins(pAsm); - - /* ADD the remaining .5 */ - pAsm->D.dst.opcode = SQ_OP2_INST_ADD; - setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); - pAsm->D.dst.rtype = DST_REG_TEMPORARY; - pAsm->D.dst.reg = tmp2; - pAsm->D.dst.writex = 1; - pAsm->D.dst.writey = 1; - pAsm->D.dst.writez = 0; - pAsm->D.dst.writew = 0; - - setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); - pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; - pAsm->S[0].src.reg = tmp2; - noswizzle_PVSSRC(&(pAsm->S[0].src)); - setaddrmode_PVSSRC(&(pAsm->S[1].src), ADDR_ABSOLUTE); - pAsm->S[1].src.rtype = SRC_REG_TEMPORARY; - pAsm->S[1].src.reg = 252; // SQ_ALU_SRC_0_5 - noswizzle_PVSSRC(&(pAsm->S[1].src)); + setswizzle_PVSSRC(&(pAsm->S[2].src), SQ_SEL_X); next_ins(pAsm); @@ -4093,22 +4372,32 @@ GLboolean assemble_TEX(r700_AssemblerBase *pAsm) } - if(pAsm->pILInst[pAsm->uiCurInst].Opcode == OPCODE_TXB) + switch(pAsm->pILInst[pAsm->uiCurInst].Opcode) { - pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE_L; - } - else - { - pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE; + case OPCODE_DDX: + /* will these need WQM(1) on CF inst ? */ + pAsm->D.dst.opcode = SQ_TEX_INST_GET_GRADIENTS_H; + break; + case OPCODE_DDY: + pAsm->D.dst.opcode = SQ_TEX_INST_GET_GRADIENTS_V; + break; + case OPCODE_TXB: + pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE_L; + break; + default: + pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE; } pAsm->is_tex = GL_TRUE; if ( GL_TRUE == need_barrier ) + + pAsm->is_tex = GL_TRUE; + if ( GL_TRUE == need_barrier ) { pAsm->need_tex_barrier = GL_TRUE; } // Set src1 to tex unit id - pAsm->S[1].src.reg = pAsm->pILInst[pAsm->uiCurInst].TexSrcUnit; + pAsm->S[1].src.reg = pAsm->SamplerUnits[pAsm->pILInst[pAsm->uiCurInst].TexSrcUnit]; pAsm->S[1].src.rtype = SRC_REG_TEMPORARY; //No sw info from mesa compiler, so hard code here. @@ -4265,27 +4554,909 @@ GLboolean assemble_EXPORT(r700_AssemblerBase *pAsm) return GL_TRUE; } -GLboolean assemble_IF(r700_AssemblerBase *pAsm) +static inline void decreaseCurrent(r700_AssemblerBase *pAsm, GLuint uReason) +{ + switch (uReason) + { + case FC_PUSH_VPM: + pAsm->CALLSTACK[pAsm->CALLSP].current--; + break; + case FC_PUSH_WQM: + pAsm->CALLSTACK[pAsm->CALLSP].current -= 4; + break; + case FC_LOOP: + pAsm->CALLSTACK[pAsm->CALLSP].current -= 4; + break; + case FC_REP: + /* TODO : for 16 vp asic, should -= 2; */ + pAsm->CALLSTACK[pAsm->CALLSP].current -= 1; + break; + }; +} + +static inline void checkStackDepth(r700_AssemblerBase *pAsm, GLuint uReason, GLboolean bCheckMaxOnly) +{ + if(GL_TRUE == bCheckMaxOnly) + { + switch (uReason) + { + case FC_PUSH_VPM: + if((pAsm->CALLSTACK[pAsm->CALLSP].current + 1) + > pAsm->CALLSTACK[pAsm->CALLSP].max) + { + pAsm->CALLSTACK[pAsm->CALLSP].max = + pAsm->CALLSTACK[pAsm->CALLSP].current + 1; + } + break; + case FC_PUSH_WQM: + if((pAsm->CALLSTACK[pAsm->CALLSP].current + 4) + > pAsm->CALLSTACK[pAsm->CALLSP].max) + { + pAsm->CALLSTACK[pAsm->CALLSP].max = + pAsm->CALLSTACK[pAsm->CALLSP].current + 4; + } + break; + } + return; + } + + switch (uReason) + { + case FC_PUSH_VPM: + pAsm->CALLSTACK[pAsm->CALLSP].current++; + break; + case FC_PUSH_WQM: + pAsm->CALLSTACK[pAsm->CALLSP].current += 4; + break; + case FC_LOOP: + pAsm->CALLSTACK[pAsm->CALLSP].current += 4; + break; + case FC_REP: + /* TODO : for 16 vp asic, should += 2; */ + pAsm->CALLSTACK[pAsm->CALLSP].current += 1; + break; + }; + + if(pAsm->CALLSTACK[pAsm->CALLSP].current + > pAsm->CALLSTACK[pAsm->CALLSP].max) + { + pAsm->CALLSTACK[pAsm->CALLSP].max = + pAsm->CALLSTACK[pAsm->CALLSP].current; + } +} + +GLboolean jumpToOffest(r700_AssemblerBase *pAsm, GLuint pops, GLint offset) +{ + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = pops; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_JUMP; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->cf_current_cf_clause_ptr->m_Word0.f.addr = pAsm->cf_current_cf_clause_ptr->m_uIndex + offset; + + return GL_TRUE; +} + +GLboolean pops(r700_AssemblerBase *pAsm, GLuint pops) +{ + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = pops; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_POP; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + pAsm->cf_current_cf_clause_ptr->m_Word0.f.addr = pAsm->cf_current_cf_clause_ptr->m_uIndex + 1; + + return GL_TRUE; +} + +GLboolean assemble_IF(r700_AssemblerBase *pAsm, GLboolean bHasElse) +{ + pAsm->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; + + assemble_LOGIC_PRED(pAsm, SQ_OP2_INST_PRED_SETNE); + + + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + if(GL_TRUE != bHasElse) + { + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + } + else + { + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 0; + } + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_JUMP; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->FCSP++; + pAsm->fc_stack[pAsm->FCSP].type = FC_IF; + pAsm->fc_stack[pAsm->FCSP].mid = NULL; + pAsm->fc_stack[pAsm->FCSP].midLen= 0; + pAsm->fc_stack[pAsm->FCSP].first = pAsm->cf_current_cf_clause_ptr; + +#ifndef USE_CF_FOR_POP_AFTER + if(GL_TRUE != bHasElse) + { + pAsm->alu_x_opcode = SQ_CF_INST_ALU_POP_AFTER; + } +#endif /* USE_CF_FOR_POP_AFTER */ + + checkStackDepth(pAsm, FC_PUSH_VPM, GL_FALSE); + + return GL_TRUE; +} + +GLboolean assemble_ELSE(r700_AssemblerBase *pAsm) { + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; /// + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_ELSE; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->fc_stack[pAsm->FCSP].mid = (R700ControlFlowGenericClause **)_mesa_realloc( (void *)pAsm->fc_stack[pAsm->FCSP].mid, + 0, + sizeof(R700ControlFlowGenericClause *) ); + pAsm->fc_stack[pAsm->FCSP].mid[0] = pAsm->cf_current_cf_clause_ptr; + //pAsm->fc_stack[pAsm->FCSP].unNumMid = 1; + +#ifndef USE_CF_FOR_POP_AFTER + pAsm->alu_x_opcode = SQ_CF_INST_ALU_POP_AFTER; +#endif /* USE_CF_FOR_POP_AFTER */ + + pAsm->fc_stack[pAsm->FCSP].first->m_Word0.f.addr = pAsm->pR700Shader->plstCFInstructions_active->uNumOfNode - 1; + return GL_TRUE; } GLboolean assemble_ENDIF(r700_AssemblerBase *pAsm) { +#ifdef USE_CF_FOR_POP_AFTER + pops(pAsm, 1); +#endif /* USE_CF_FOR_POP_AFTER */ + + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + + if(NULL == pAsm->fc_stack[pAsm->FCSP].mid) + { + /* no else in between */ + pAsm->fc_stack[pAsm->FCSP].first->m_Word0.f.addr = pAsm->pR700Shader->plstCFInstructions_active->uNumOfNode; + } + else + { + pAsm->fc_stack[pAsm->FCSP].mid[0]->m_Word0.f.addr = pAsm->pR700Shader->plstCFInstructions_active->uNumOfNode; + } + + if(NULL != pAsm->fc_stack[pAsm->FCSP].mid) + { + FREE(pAsm->fc_stack[pAsm->FCSP].mid); + } + + if(pAsm->fc_stack[pAsm->FCSP].type != FC_IF) + { + radeon_error("if/endif in shader code are not paired. \n"); + return GL_FALSE; + } + + pAsm->FCSP--; + + decreaseCurrent(pAsm, FC_PUSH_VPM); + + return GL_TRUE; +} + +GLboolean assemble_BGNLOOP(r700_AssemblerBase *pAsm) +{ + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_LOOP_START_NO_AL; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->FCSP++; + pAsm->fc_stack[pAsm->FCSP].type = FC_LOOP; + pAsm->fc_stack[pAsm->FCSP].mid = NULL; + pAsm->fc_stack[pAsm->FCSP].unNumMid = 0; + pAsm->fc_stack[pAsm->FCSP].midLen = 0; + pAsm->fc_stack[pAsm->FCSP].first = pAsm->cf_current_cf_clause_ptr; + + checkStackDepth(pAsm, FC_LOOP, GL_FALSE); + + return GL_TRUE; +} + +GLboolean assemble_BRK(r700_AssemblerBase *pAsm) +{ +#ifdef USE_CF_FOR_CONTINUE_BREAK + + pAsm->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; + + assemble_LOGIC_PRED(pAsm, SQ_OP2_INST_PRED_SETNE); + + unsigned int unFCSP; + for(unFCSP=pAsm->FCSP; unFCSP>0; unFCSP--) + { + if(FC_LOOP == pAsm->fc_stack[unFCSP].type) + { + break; + } + } + if(0 == FC_LOOP) + { + radeon_error("Break is not inside loop/endloop pair.\n"); + return GL_FALSE; + } + + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_LOOP_BREAK; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->fc_stack[unFCSP].mid = (R700ControlFlowGenericClause **)_mesa_realloc( + (void *)pAsm->fc_stack[unFCSP].mid, + sizeof(R700ControlFlowGenericClause *) * pAsm->fc_stack[unFCSP].unNumMid, + sizeof(R700ControlFlowGenericClause *) * (pAsm->fc_stack[unFCSP].unNumMid + 1) ); + pAsm->fc_stack[unFCSP].mid[pAsm->fc_stack[unFCSP].unNumMid] = pAsm->cf_current_cf_clause_ptr; + pAsm->fc_stack[unFCSP].unNumMid++; + + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_POP; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + pAsm->cf_current_cf_clause_ptr->m_Word0.f.addr = pAsm->cf_current_cf_clause_ptr->m_uIndex + 1; + + checkStackDepth(pAsm, FC_PUSH_VPM, GL_TRUE); + +#endif //USE_CF_FOR_CONTINUE_BREAK return GL_TRUE; } -GLboolean AssembleInstr(GLuint uiNumberInsts, +GLboolean assemble_CONT(r700_AssemblerBase *pAsm) +{ +#ifdef USE_CF_FOR_CONTINUE_BREAK + pAsm->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; + + assemble_LOGIC_PRED(pAsm, SQ_OP2_INST_PRED_SETNE); + + unsigned int unFCSP; + for(unFCSP=pAsm->FCSP; unFCSP>0; unFCSP--) + { + if(FC_LOOP == pAsm->fc_stack[unFCSP].type) + { + break; + } + } + if(0 == FC_LOOP) + { + radeon_error("Continue is not inside loop/endloop pair.\n"); + return GL_FALSE; + } + + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_LOOP_CONTINUE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->fc_stack[unFCSP].mid = (R700ControlFlowGenericClause **)_mesa_realloc( + (void *)pAsm->fc_stack[unFCSP].mid, + sizeof(R700ControlFlowGenericClause *) * pAsm->fc_stack[unFCSP].unNumMid, + sizeof(R700ControlFlowGenericClause *) * (pAsm->fc_stack[unFCSP].unNumMid + 1) ); + pAsm->fc_stack[unFCSP].mid[pAsm->fc_stack[unFCSP].unNumMid] = pAsm->cf_current_cf_clause_ptr; + pAsm->fc_stack[unFCSP].unNumMid++; + + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_POP; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + pAsm->cf_current_cf_clause_ptr->m_Word0.f.addr = pAsm->cf_current_cf_clause_ptr->m_uIndex + 1; + + checkStackDepth(pAsm, FC_PUSH_VPM, GL_TRUE); + +#endif /* USE_CF_FOR_CONTINUE_BREAK */ + + return GL_TRUE; +} + +GLboolean assemble_ENDLOOP(r700_AssemblerBase *pAsm) +{ + GLuint i; + + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_LOOP_END; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->cf_current_cf_clause_ptr->m_Word0.f.addr = pAsm->fc_stack[pAsm->FCSP].first->m_uIndex + 1; + pAsm->fc_stack[pAsm->FCSP].first->m_Word0.f.addr = pAsm->cf_current_cf_clause_ptr->m_uIndex + 1; + +#ifdef USE_CF_FOR_CONTINUE_BREAK + for(i=0; i<pAsm->fc_stack[pAsm->FCSP].unNumMid; i++) + { + pAsm->fc_stack[pAsm->FCSP].mid[i]->m_Word0.f.addr = pAsm->cf_current_cf_clause_ptr->m_uIndex; + } + if(NULL != pAsm->fc_stack[pAsm->FCSP].mid) + { + FREE(pAsm->fc_stack[pAsm->FCSP].mid); + } +#endif + + if(pAsm->fc_stack[pAsm->FCSP].type != FC_LOOP) + { + radeon_error("loop/endloop in shader code are not paired. \n"); + return GL_FALSE; + } + + GLuint unFCSP; + GLuint unIF = 0; + if((pAsm->unCFflags & HAS_CURRENT_LOOPRET) > 0) + { + for(unFCSP=(pAsm->FCSP-1); unFCSP>pAsm->CALLSTACK[pAsm->CALLSP].FCSP_BeforeEntry; unFCSP--) + { + if(FC_LOOP == pAsm->fc_stack[unFCSP].type) + { + breakLoopOnFlag(pAsm, unFCSP); + break; + } + else if(FC_IF == pAsm->fc_stack[unFCSP].type) + { + unIF++; + } + } + if(unFCSP <= pAsm->CALLSTACK[pAsm->CALLSP].FCSP_BeforeEntry) + { +#ifdef USE_CF_FOR_POP_AFTER + returnOnFlag(pAsm, unIF); +#else + returnOnFlag(pAsm, 0); +#endif /* USE_CF_FOR_POP_AFTER */ + pAsm->unCFflags &= ~HAS_CURRENT_LOOPRET; + } + } + + pAsm->FCSP--; + + decreaseCurrent(pAsm, FC_LOOP); + + return GL_TRUE; +} + +void add_return_inst(r700_AssemblerBase *pAsm) +{ + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + //pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_RETURN; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; +} + +GLboolean assemble_BGNSUB(r700_AssemblerBase *pAsm, GLint nILindex, GLuint uiIL_Shift) +{ + /* Put in sub */ + if( (pAsm->unSubArrayPointer + 1) > pAsm->unSubArraySize ) + { + pAsm->subs = (SUB_OFFSET*)_mesa_realloc( (void *)pAsm->subs, + sizeof(SUB_OFFSET) * pAsm->unSubArraySize, + sizeof(SUB_OFFSET) * (pAsm->unSubArraySize + 10) ); + if(NULL == pAsm->subs) + { + return GL_FALSE; + } + pAsm->unSubArraySize += 10; + } + + pAsm->subs[pAsm->unSubArrayPointer].subIL_Offset = nILindex + uiIL_Shift; + pAsm->subs[pAsm->unSubArrayPointer].lstCFInstructions_local.pHead=NULL; + pAsm->subs[pAsm->unSubArrayPointer].lstCFInstructions_local.pTail=NULL; + pAsm->subs[pAsm->unSubArrayPointer].lstCFInstructions_local.uNumOfNode=0; + + pAsm->CALLSP++; + pAsm->CALLSTACK[pAsm->CALLSP].subDescIndex = pAsm->unSubArrayPointer; + pAsm->CALLSTACK[pAsm->CALLSP].FCSP_BeforeEntry = pAsm->FCSP; + pAsm->CALLSTACK[pAsm->CALLSP].plstCFInstructions_local + = &(pAsm->subs[pAsm->unSubArrayPointer].lstCFInstructions_local); + pAsm->CALLSTACK[pAsm->CALLSP].max = 0; + pAsm->CALLSTACK[pAsm->CALLSP].current = 0; + SetActiveCFlist(pAsm->pR700Shader, + pAsm->CALLSTACK[pAsm->CALLSP].plstCFInstructions_local); + + pAsm->unSubArrayPointer++; + + /* start sub */ + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + + pAsm->FCSP++; + pAsm->fc_stack[pAsm->FCSP].type = FC_REP; + + checkStackDepth(pAsm, FC_REP, GL_FALSE); + + return GL_TRUE; +} + +GLboolean assemble_ENDSUB(r700_AssemblerBase *pAsm) +{ + if(pAsm->fc_stack[pAsm->FCSP].type != FC_REP) + { + radeon_error("BGNSUB/ENDSUB in shader code are not paired. \n"); + return GL_FALSE; + } + + /* copy max to sub structure */ + pAsm->subs[pAsm->CALLSTACK[pAsm->CALLSP].subDescIndex].unStackDepthMax + = pAsm->CALLSTACK[pAsm->CALLSP].max; + + decreaseCurrent(pAsm, FC_REP); + + pAsm->CALLSP--; + SetActiveCFlist(pAsm->pR700Shader, + pAsm->CALLSTACK[pAsm->CALLSP].plstCFInstructions_local); + + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + + pAsm->FCSP--; + + return GL_TRUE; +} + +GLboolean assemble_RET(r700_AssemblerBase *pAsm) +{ + GLuint unIF = 0; + + if(pAsm->CALLSP > 0) + { /* in sub */ + GLuint unFCSP; + for(unFCSP=pAsm->FCSP; unFCSP>pAsm->CALLSTACK[pAsm->CALLSP].FCSP_BeforeEntry; unFCSP--) + { + if(FC_LOOP == pAsm->fc_stack[unFCSP].type) + { + setRetInLoopFlag(pAsm, SQ_SEL_1); + breakLoopOnFlag(pAsm, unFCSP); + pAsm->unCFflags |= LOOPRET_FLAGS; + + return GL_TRUE; + } + else if(FC_IF == pAsm->fc_stack[unFCSP].type) + { + unIF++; + } + } + } + +#ifdef USE_CF_FOR_POP_AFTER + if(unIF > 0) + { + pops(pAsm, unIF); + } +#endif /* USE_CF_FOR_POP_AFTER */ + + add_return_inst(pAsm); + + return GL_TRUE; +} + +GLboolean assemble_CAL(r700_AssemblerBase *pAsm, + GLint nILindex, + GLuint uiIL_Shift, + GLuint uiNumberInsts, + struct prog_instruction *pILInst, + PRESUB_DESC * pPresubDesc) +{ + GLint uiIL_Offset; + + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.call_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_CALL; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + /* Put in caller */ + if( (pAsm->unCallerArrayPointer + 1) > pAsm->unCallerArraySize ) + { + pAsm->callers = (CALLER_POINTER*)_mesa_realloc( (void *)pAsm->callers, + sizeof(CALLER_POINTER) * pAsm->unCallerArraySize, + sizeof(CALLER_POINTER) * (pAsm->unCallerArraySize + 10) ); + if(NULL == pAsm->callers) + { + return GL_FALSE; + } + pAsm->unCallerArraySize += 10; + } + + uiIL_Offset = nILindex + uiIL_Shift; + pAsm->callers[pAsm->unCallerArrayPointer].subIL_Offset = uiIL_Offset; + pAsm->callers[pAsm->unCallerArrayPointer].cf_ptr = pAsm->cf_current_cf_clause_ptr; + + pAsm->callers[pAsm->unCallerArrayPointer].finale_cf_ptr = NULL; + pAsm->callers[pAsm->unCallerArrayPointer].prelude_cf_ptr = NULL; + + pAsm->unCallerArrayPointer++; + + int j; + GLuint max; + GLuint unSubID; + GLboolean bRet; + for(j=0; j<pAsm->unSubArrayPointer; j++) + { + if(uiIL_Offset == pAsm->subs[j].subIL_Offset) + { /* compiled before */ + + max = pAsm->subs[j].unStackDepthMax + + pAsm->CALLSTACK[pAsm->CALLSP].current; + if(max > pAsm->CALLSTACK[pAsm->CALLSP].max) + { + pAsm->CALLSTACK[pAsm->CALLSP].max = max; + } + + pAsm->callers[pAsm->unCallerArrayPointer - 1].subDescIndex = j; + return GL_TRUE; + } + } + + pAsm->callers[pAsm->unCallerArrayPointer - 1].subDescIndex = pAsm->unSubArrayPointer; + unSubID = pAsm->unSubArrayPointer; + + bRet = AssembleInstr(nILindex, uiIL_Shift, uiNumberInsts, pILInst, pAsm); + + if(GL_TRUE == bRet) + { + max = pAsm->subs[unSubID].unStackDepthMax + + pAsm->CALLSTACK[pAsm->CALLSP].current; + if(max > pAsm->CALLSTACK[pAsm->CALLSP].max) + { + pAsm->CALLSTACK[pAsm->CALLSP].max = max; + } + + pAsm->subs[unSubID].pPresubDesc = pPresubDesc; + } + + return bRet; +} + +GLboolean setRetInLoopFlag(r700_AssemblerBase *pAsm, GLuint flagValue) +{ + GLfloat fLiteral[2] = {0.1, 0.0}; + + pAsm->D.dst.opcode = SQ_OP2_INST_MOV; + pAsm->D.dst.op3 = 0; + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = pAsm->flag_reg_index; + pAsm->D.dst.writex = 1; + pAsm->D.dst.writey = 0; + pAsm->D.dst.writez = 0; + pAsm->D.dst.writew = 0; + pAsm->D2.dst2.literal_slots = 1; + pAsm->D2.dst2.SaturateMode = SATURATE_OFF; + pAsm->D.dst.predicated = 0; + /* in reloc where dislink flag init inst, only one slot alu inst is handled. */ + pAsm->D.dst.math = 1; /* TODO : not math really, but one channel op, more generic alu assembler needed */ + pAsm->D2.dst2.index_mode = SQ_INDEX_LOOP; /* Check this ! */ +#if 0 + pAsm->S[0].src.rtype = SRC_REC_LITERAL; + //pAsm->S[0].src.reg = 0; + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + noneg_PVSSRC(&(pAsm->S[0].src)); + pAsm->S[0].src.swizzlex = SQ_SEL_X; + pAsm->S[0].src.swizzley = SQ_SEL_Y; + pAsm->S[0].src.swizzlez = SQ_SEL_Z; + pAsm->S[0].src.swizzlew = SQ_SEL_W; + + if( GL_FALSE == next_ins_literal(pAsm, &(fLiteral[0])) ) + { + return GL_FALSE; + } +#else + pAsm->S[0].src.rtype = DST_REG_TEMPORARY; + pAsm->S[0].src.reg = 0; + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + noneg_PVSSRC(&(pAsm->S[0].src)); + pAsm->S[0].src.swizzlex = flagValue; + pAsm->S[0].src.swizzley = flagValue; + pAsm->S[0].src.swizzlez = flagValue; + pAsm->S[0].src.swizzlew = flagValue; + + if( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } +#endif + + return GL_TRUE; +} + +GLboolean testFlag(r700_AssemblerBase *pAsm) +{ + GLfloat fLiteral[2] = {0.1, 0.0}; + + //Test flag + GLuint tmp = gethelpr(pAsm); + pAsm->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; + + pAsm->D.dst.opcode = SQ_OP2_INST_PRED_SETE; + pAsm->D.dst.math = 1; + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = tmp; + pAsm->D.dst.writex = 1; + pAsm->D.dst.writey = 0; + pAsm->D.dst.writez = 0; + pAsm->D.dst.writew = 0; + pAsm->D2.dst2.literal_slots = 1; + pAsm->D2.dst2.SaturateMode = SATURATE_OFF; + pAsm->D.dst.predicated = 1; + pAsm->D2.dst2.index_mode = SQ_INDEX_LOOP; /* Check this ! */ + + pAsm->S[0].src.rtype = DST_REG_TEMPORARY; + pAsm->S[0].src.reg = pAsm->flag_reg_index; + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + noneg_PVSSRC(&(pAsm->S[0].src)); + pAsm->S[0].src.swizzlex = SQ_SEL_X; + pAsm->S[0].src.swizzley = SQ_SEL_Y; + pAsm->S[0].src.swizzlez = SQ_SEL_Z; + pAsm->S[0].src.swizzlew = SQ_SEL_W; +#if 0 + pAsm->S[1].src.rtype = SRC_REC_LITERAL; + //pAsm->S[1].src.reg = 0; + setaddrmode_PVSSRC(&(pAsm->S[1].src), ADDR_ABSOLUTE); + noneg_PVSSRC(&(pAsm->S[1].src)); + pAsm->S[1].src.swizzlex = SQ_SEL_X; + pAsm->S[1].src.swizzley = SQ_SEL_Y; + pAsm->S[1].src.swizzlez = SQ_SEL_Z; + pAsm->S[1].src.swizzlew = SQ_SEL_W; + + if( GL_FALSE == next_ins_literal(pAsm, &(fLiteral[0])) ) + { + return GL_FALSE; + } +#else + pAsm->S[1].src.rtype = DST_REG_TEMPORARY; + pAsm->S[1].src.reg = 0; + setaddrmode_PVSSRC(&(pAsm->S[1].src), ADDR_ABSOLUTE); + noneg_PVSSRC(&(pAsm->S[1].src)); + pAsm->S[1].src.swizzlex = SQ_SEL_1; + pAsm->S[1].src.swizzley = SQ_SEL_1; + pAsm->S[1].src.swizzlez = SQ_SEL_1; + pAsm->S[1].src.swizzlew = SQ_SEL_1; + + if( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } +#endif + + checkStackDepth(pAsm, FC_PUSH_VPM, GL_TRUE); + + return GL_TRUE; +} + +GLboolean returnOnFlag(r700_AssemblerBase *pAsm, GLuint unIF) +{ + testFlag(pAsm); + jumpToOffest(pAsm, 1, 4); + setRetInLoopFlag(pAsm, SQ_SEL_0); + pops(pAsm, unIF + 1); + add_return_inst(pAsm); + + return GL_TRUE; +} + +GLboolean breakLoopOnFlag(r700_AssemblerBase *pAsm, GLuint unFCSP) +{ + testFlag(pAsm); + + //break + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_LOOP_BREAK; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->fc_stack[unFCSP].mid = (R700ControlFlowGenericClause **)_mesa_realloc( + (void *)pAsm->fc_stack[unFCSP].mid, + sizeof(R700ControlFlowGenericClause *) * pAsm->fc_stack[unFCSP].unNumMid, + sizeof(R700ControlFlowGenericClause *) * (pAsm->fc_stack[unFCSP].unNumMid + 1) ); + pAsm->fc_stack[unFCSP].mid[pAsm->fc_stack[unFCSP].unNumMid] = pAsm->cf_current_cf_clause_ptr; + pAsm->fc_stack[unFCSP].unNumMid++; + + pops(pAsm, 1); + + return GL_TRUE; +} + +GLboolean AssembleInstr(GLuint uiFirstInst, + GLuint uiIL_Shift, + GLuint uiNumberInsts, struct prog_instruction *pILInst, r700_AssemblerBase *pR700AsmCode) { GLuint i; pR700AsmCode->pILInst = pILInst; - for(i=0; i<uiNumberInsts; i++) + for(i=uiFirstInst; i<uiNumberInsts; i++) { pR700AsmCode->uiCurInst = i; +#ifndef USE_CF_FOR_CONTINUE_BREAK + if(OPCODE_BRK == pILInst[i+1].Opcode) + { + switch(pILInst[i].Opcode) + { + case OPCODE_SLE: + pILInst[i].Opcode = OPCODE_SGT; + break; + case OPCODE_SLT: + pILInst[i].Opcode = OPCODE_SGE; + break; + case OPCODE_SGE: + pILInst[i].Opcode = OPCODE_SLT; + break; + case OPCODE_SGT: + pILInst[i].Opcode = OPCODE_SLE; + break; + case OPCODE_SEQ: + pILInst[i].Opcode = OPCODE_SNE; + break; + case OPCODE_SNE: + pILInst[i].Opcode = OPCODE_SEQ; + break; + default: + break; + } + } +#endif + if(pILInst[i].CondUpdate == 1) + { + /* remember dest register used for cond evaluation */ + /* XXX also handle PROGRAM_OUTPUT registers here? */ + pR700AsmCode->last_cond_register = pILInst[i].DstReg.Index; + } + switch (pILInst[i].Opcode) { case OPCODE_ABS: @@ -4313,7 +5484,7 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, return GL_FALSE; break; case OPCODE_COS: - if ( GL_FALSE == assemble_COS(pR700AsmCode) ) + if ( GL_FALSE == assemble_TRIG(pR700AsmCode, SQ_OP2_INST_COS) ) return GL_FALSE; break; @@ -4342,7 +5513,8 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, if ( GL_FALSE == assemble_FLR(pR700AsmCode) ) return GL_FALSE; break; - //case OP_FLR_INT: + //case OP_FLR_INT: ; + // if ( GL_FALSE == assemble_FLR_INT() ) // return GL_FALSE; // break; @@ -4353,7 +5525,8 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, break; case OPCODE_KIL: - if ( GL_FALSE == assemble_KIL(pR700AsmCode) ) + case OPCODE_KIL_NV: + if ( GL_FALSE == assemble_KIL(pR700AsmCode, SQ_OP2_INST_KILLGT) ) return GL_FALSE; break; case OPCODE_LG2: @@ -4393,6 +5566,26 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, case OPCODE_MUL: if ( GL_FALSE == assemble_MUL(pR700AsmCode) ) return GL_FALSE; + break; + + case OPCODE_NOISE1: + { + callPreSub(pR700AsmCode, + GLSL_NOISE1, + &noise1_presub, + pILInst->DstReg.Index + pR700AsmCode->starting_temp_register_number, + 1); + radeon_error("noise1: not yet supported shader instruction\n"); + }; + break; + case OPCODE_NOISE2: + radeon_error("noise2: not yet supported shader instruction\n"); + break; + case OPCODE_NOISE3: + radeon_error("noise3: not yet supported shader instruction\n"); + break; + case OPCODE_NOISE4: + radeon_error("noise4: not yet supported shader instruction\n"); break; case OPCODE_POW: @@ -4408,22 +5601,78 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, return GL_FALSE; break; case OPCODE_SIN: - if ( GL_FALSE == assemble_SIN(pR700AsmCode) ) + if ( GL_FALSE == assemble_TRIG(pR700AsmCode, SQ_OP2_INST_SIN) ) return GL_FALSE; break; case OPCODE_SCS: if ( GL_FALSE == assemble_SCS(pR700AsmCode) ) return GL_FALSE; - break; + break; + + case OPCODE_SEQ: + if ( GL_FALSE == assemble_LOGIC(pR700AsmCode, SQ_OP2_INST_SETE) ) + { + return GL_FALSE; + } + break; + + case OPCODE_SGT: + if ( GL_FALSE == assemble_LOGIC(pR700AsmCode, SQ_OP2_INST_SETGT) ) + { + return GL_FALSE; + } + break; case OPCODE_SGE: if ( GL_FALSE == assemble_SGE(pR700AsmCode) ) + { return GL_FALSE; - break; + } + break; + + /* NO LT, LE, TODO : use GE => LE, GT => LT : reverse 2 src order would be simpliest. Or use SQ_CF_COND_FALSE for SQ_CF_COND_ACTIVE.*/ case OPCODE_SLT: - if ( GL_FALSE == assemble_SLT(pR700AsmCode) ) + { + struct prog_src_register SrcRegSave[2]; + SrcRegSave[0] = pILInst[i].SrcReg[0]; + SrcRegSave[1] = pILInst[i].SrcReg[1]; + pILInst[i].SrcReg[0] = SrcRegSave[1]; + pILInst[i].SrcReg[1] = SrcRegSave[0]; + if ( GL_FALSE == assemble_LOGIC(pR700AsmCode, SQ_OP2_INST_SETGT) ) + { + pILInst[i].SrcReg[0] = SrcRegSave[0]; + pILInst[i].SrcReg[1] = SrcRegSave[1]; + return GL_FALSE; + } + pILInst[i].SrcReg[0] = SrcRegSave[0]; + pILInst[i].SrcReg[1] = SrcRegSave[1]; + } + break; + + case OPCODE_SLE: + { + struct prog_src_register SrcRegSave[2]; + SrcRegSave[0] = pILInst[i].SrcReg[0]; + SrcRegSave[1] = pILInst[i].SrcReg[1]; + pILInst[i].SrcReg[0] = SrcRegSave[1]; + pILInst[i].SrcReg[1] = SrcRegSave[0]; + if ( GL_FALSE == assemble_LOGIC(pR700AsmCode, SQ_OP2_INST_SETGE) ) + { + pILInst[i].SrcReg[0] = SrcRegSave[0]; + pILInst[i].SrcReg[1] = SrcRegSave[1]; + return GL_FALSE; + } + pILInst[i].SrcReg[0] = SrcRegSave[0]; + pILInst[i].SrcReg[1] = SrcRegSave[1]; + } + break; + + case OPCODE_SNE: + if ( GL_FALSE == assemble_LOGIC(pR700AsmCode, SQ_OP2_INST_SETNE) ) + { return GL_FALSE; - break; + } + break; //case OP_STP: // if ( GL_FALSE == assemble_STP(pR700AsmCode) ) @@ -4449,7 +5698,8 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, } } break; - + case OPCODE_DDX: + case OPCODE_DDY: case OPCODE_TEX: case OPCODE_TXB: case OPCODE_TXP: @@ -4457,30 +5707,104 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, return GL_FALSE; break; + case OPCODE_TRUNC: + if ( GL_FALSE == assemble_math_function(pR700AsmCode, SQ_OP2_INST_TRUNC) ) + return GL_FALSE; + break; + case OPCODE_XPD: if ( GL_FALSE == assemble_XPD(pR700AsmCode) ) return GL_FALSE; break; case OPCODE_IF : - if ( GL_FALSE == assemble_IF(pR700AsmCode) ) - return GL_FALSE; + { + GLboolean bHasElse = GL_FALSE; + + if(pILInst[pILInst[i].BranchTarget - 1].Opcode == OPCODE_ELSE) + { + bHasElse = GL_TRUE; + } + + if ( GL_FALSE == assemble_IF(pR700AsmCode, bHasElse) ) + { + return GL_FALSE; + } + } break; + case OPCODE_ELSE : - radeon_error("Not yet implemented instruction OPCODE_ELSE \n"); - //if ( GL_FALSE == assemble_BAD("ELSE") ) + if ( GL_FALSE == assemble_ELSE(pR700AsmCode) ) return GL_FALSE; break; + case OPCODE_ENDIF: if ( GL_FALSE == assemble_ENDIF(pR700AsmCode) ) return GL_FALSE; break; + case OPCODE_BGNLOOP: + if( GL_FALSE == assemble_BGNLOOP(pR700AsmCode) ) + { + return GL_FALSE; + } + break; + + case OPCODE_BRK: + if( GL_FALSE == assemble_BRK(pR700AsmCode) ) + { + return GL_FALSE; + } + break; + + case OPCODE_CONT: + if( GL_FALSE == assemble_CONT(pR700AsmCode) ) + { + return GL_FALSE; + } + break; + + case OPCODE_ENDLOOP: + if( GL_FALSE == assemble_ENDLOOP(pR700AsmCode) ) + { + return GL_FALSE; + } + break; + + case OPCODE_BGNSUB: + if( GL_FALSE == assemble_BGNSUB(pR700AsmCode, i, uiIL_Shift) ) + { + return GL_FALSE; + } + break; + + case OPCODE_RET: + if( GL_FALSE == assemble_RET(pR700AsmCode) ) + { + return GL_FALSE; + } + break; + + case OPCODE_CAL: + if( GL_FALSE == assemble_CAL(pR700AsmCode, + pILInst[i].BranchTarget, + uiIL_Shift, + uiNumberInsts, + pILInst, + NULL) ) + { + return GL_FALSE; + } + break; + //case OPCODE_EXPORT: // if ( GL_FALSE == assemble_EXPORT() ) // return GL_FALSE; // break; + case OPCODE_ENDSUB: + return assemble_ENDSUB(pR700AsmCode); + case OPCODE_END: //pR700AsmCode->uiCurInst = i; //This is to remaind that if in later exoort there is depth/stencil @@ -4497,6 +5821,417 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, return GL_TRUE; } +GLboolean InitShaderProgram(r700_AssemblerBase * pAsm) +{ + setRetInLoopFlag(pAsm, SQ_SEL_0); + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + return GL_TRUE; +} + +GLboolean RelocProgram(r700_AssemblerBase * pAsm, struct gl_program * pILProg) +{ + GLuint i; + GLuint unCFoffset; + TypedShaderList * plstCFmain; + TypedShaderList * plstCFsub; + + R700ShaderInstruction * pInst; + R700ControlFlowGenericClause * pCFInst; + + R700ControlFlowALUClause * pCF_ALU; + R700ALUInstruction * pALU; + GLuint unConstOffset = 0; + GLuint unRegOffset; + GLuint unMinRegIndex; + + plstCFmain = pAsm->CALLSTACK[0].plstCFInstructions_local; + + /* remove flags init if they are not used */ + if((pAsm->unCFflags & HAS_LOOPRET) == 0) + { + R700ControlFlowALUClause * pCF_ALU; + pInst = plstCFmain->pHead; + while(pInst) + { + if(SIT_CF_ALU == pInst->m_ShaderInstType) + { + pCF_ALU = (R700ControlFlowALUClause *)pInst; + if(0 == pCF_ALU->m_Word1.f.count) + { + pCF_ALU->m_Word1.f.cf_inst = SQ_CF_INST_NOP; + } + else + { + R700ALUInstruction * pALU = pCF_ALU->m_pLinkedALUInstruction; + + pALU->m_pLinkedALUClause = NULL; + pALU = (R700ALUInstruction *)(pALU->pNextInst); + pALU->m_pLinkedALUClause = pCF_ALU; + pCF_ALU->m_pLinkedALUInstruction = pALU; + + pCF_ALU->m_Word1.f.count--; + } + break; + } + pInst = pInst->pNextInst; + }; + } + + if(pAsm->CALLSTACK[0].max > 0) + { + pAsm->pR700Shader->uStackSize = ((pAsm->CALLSTACK[0].max + 3)>>2) + 2; + } + + if(0 == pAsm->unSubArrayPointer) + { + return GL_TRUE; + } + + unCFoffset = plstCFmain->uNumOfNode; + + if(NULL != pILProg->Parameters) + { + unConstOffset = pILProg->Parameters->NumParameters; + } + + /* Reloc subs */ + for(i=0; i<pAsm->unSubArrayPointer; i++) + { + pAsm->subs[i].unCFoffset = unCFoffset; + plstCFsub = &(pAsm->subs[i].lstCFInstructions_local); + + pInst = plstCFsub->pHead; + + /* reloc instructions */ + while(pInst) + { + if(SIT_CF_GENERIC == pInst->m_ShaderInstType) + { + pCFInst = (R700ControlFlowGenericClause *)pInst; + + switch (pCFInst->m_Word1.f.cf_inst) + { + case SQ_CF_INST_POP: + case SQ_CF_INST_JUMP: + case SQ_CF_INST_ELSE: + case SQ_CF_INST_LOOP_END: + case SQ_CF_INST_LOOP_START: + case SQ_CF_INST_LOOP_START_NO_AL: + case SQ_CF_INST_LOOP_CONTINUE: + case SQ_CF_INST_LOOP_BREAK: + pCFInst->m_Word0.f.addr += unCFoffset; + break; + default: + break; + } + } + + pInst->m_uIndex += unCFoffset; + + pInst = pInst->pNextInst; + }; + + if(NULL != pAsm->subs[i].pPresubDesc) + { + GLuint uNumSrc; + + unMinRegIndex = pAsm->subs[i].pPresubDesc->pCompiledSub->MinRegIndex; + unRegOffset = pAsm->subs[i].pPresubDesc->maxStartReg; + unConstOffset += pAsm->subs[i].pPresubDesc->unConstantsStart; + + pInst = plstCFsub->pHead; + while(pInst) + { + if(SIT_CF_ALU == pInst->m_ShaderInstType) + { + pCF_ALU = (R700ControlFlowALUClause *)pInst; + + pALU = pCF_ALU->m_pLinkedALUInstruction; + for(int j=0; j<=pCF_ALU->m_Word1.f.count; j++) + { + pALU->m_Word1.f.dst_gpr = pALU->m_Word1.f.dst_gpr + unRegOffset - unMinRegIndex; + + if(pALU->m_Word0.f.src0_sel < SQ_ALU_SRC_GPR_SIZE) + { + pALU->m_Word0.f.src0_sel = pALU->m_Word0.f.src0_sel + unRegOffset - unMinRegIndex; + } + else if(pALU->m_Word0.f.src0_sel >= SQ_ALU_SRC_CFILE_BASE) + { + pALU->m_Word0.f.src0_sel += unConstOffset; + } + + if( ((pALU->m_Word1.val >> SQ_ALU_WORD1_OP3_ALU_INST_SHIFT) & 0x0000001F) + >= SQ_OP3_INST_MUL_LIT ) + { /* op3 : 3 srcs */ + if(pALU->m_Word1_OP3.f.src2_sel < SQ_ALU_SRC_GPR_SIZE) + { + pALU->m_Word1_OP3.f.src2_sel = pALU->m_Word1_OP3.f.src2_sel + unRegOffset - unMinRegIndex; + } + else if(pALU->m_Word1_OP3.f.src2_sel >= SQ_ALU_SRC_CFILE_BASE) + { + pALU->m_Word1_OP3.f.src2_sel += unConstOffset; + } + if(pALU->m_Word0.f.src1_sel < SQ_ALU_SRC_GPR_SIZE) + { + pALU->m_Word0.f.src1_sel = pALU->m_Word0.f.src1_sel + unRegOffset - unMinRegIndex; + } + else if(pALU->m_Word0.f.src1_sel >= SQ_ALU_SRC_CFILE_BASE) + { + pALU->m_Word0.f.src1_sel += unConstOffset; + } + } + else + { + if(pAsm->bR6xx) + { + uNumSrc = r700GetNumOperands(pALU->m_Word1_OP2.f6.alu_inst, 0); + } + else + { + uNumSrc = r700GetNumOperands(pALU->m_Word1_OP2.f.alu_inst, 0); + } + if(2 == uNumSrc) + { /* 2 srcs */ + if(pALU->m_Word0.f.src1_sel < SQ_ALU_SRC_GPR_SIZE) + { + pALU->m_Word0.f.src1_sel = pALU->m_Word0.f.src1_sel + unRegOffset - unMinRegIndex; + } + else if(pALU->m_Word0.f.src1_sel >= SQ_ALU_SRC_CFILE_BASE) + { + pALU->m_Word0.f.src1_sel += unConstOffset; + } + } + } + pALU = (R700ALUInstruction*)(pALU->pNextInst); + } + } + pInst = pInst->pNextInst; + }; + } + + /* Put sub into main */ + plstCFmain->pTail->pNextInst = plstCFsub->pHead; + plstCFmain->pTail = plstCFsub->pTail; + plstCFmain->uNumOfNode += plstCFsub->uNumOfNode; + + unCFoffset += plstCFsub->uNumOfNode; + } + + /* reloc callers */ + for(i=0; i<pAsm->unCallerArrayPointer; i++) + { + pAsm->callers[i].cf_ptr->m_Word0.f.addr + = pAsm->subs[pAsm->callers[i].subDescIndex].unCFoffset; + + if(NULL != pAsm->subs[pAsm->callers[i].subDescIndex].pPresubDesc) + { + unMinRegIndex = pAsm->subs[pAsm->callers[i].subDescIndex].pPresubDesc->pCompiledSub->MinRegIndex; + unRegOffset = pAsm->subs[pAsm->callers[i].subDescIndex].pPresubDesc->maxStartReg; + + if(NULL != pAsm->callers[i].prelude_cf_ptr) + { + pCF_ALU = (R700ControlFlowALUClause * )(pAsm->callers[i].prelude_cf_ptr); + pALU = pCF_ALU->m_pLinkedALUInstruction; + for(int j=0; j<=pCF_ALU->m_Word1.f.count; j++) + { + pALU->m_Word1.f.dst_gpr = pALU->m_Word1.f.dst_gpr + unRegOffset - unMinRegIndex; + pALU = (R700ALUInstruction*)(pALU->pNextInst); + } + } + if(NULL != pAsm->callers[i].finale_cf_ptr) + { + pCF_ALU = (R700ControlFlowALUClause * )(pAsm->callers[i].finale_cf_ptr); + pALU = pCF_ALU->m_pLinkedALUInstruction; + for(int j=0; j<=pCF_ALU->m_Word1.f.count; j++) + { + pALU->m_Word0.f.src0_sel = pALU->m_Word0.f.src0_sel + unRegOffset - unMinRegIndex; + pALU = (R700ALUInstruction*)(pALU->pNextInst); + } + } + } + } + + return GL_TRUE; +} + +GLboolean callPreSub(r700_AssemblerBase* pAsm, + LOADABLE_SCRIPT_SIGNITURE scriptSigniture, + COMPILED_SUB * pCompiledSub, + GLshort uOutReg, + GLshort uNumValidSrc) +{ + /* save assemble context */ + GLuint starting_temp_register_number_save; + GLuint number_used_registers_save; + GLuint uFirstHelpReg_save; + GLuint uHelpReg_save; + GLuint uiCurInst_save; + struct prog_instruction *pILInst_save; + PRESUB_DESC * pPresubDesc; + GLboolean bRet; + int i; + + R700ControlFlowGenericClause* prelude_cf_ptr = NULL; + + /* copy srcs to presub inputs */ + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + for(i=0; i<uNumValidSrc; i++) + { + pAsm->D.dst.opcode = SQ_OP2_INST_MOV; + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = pCompiledSub->srcRegIndex[i]; + pAsm->D.dst.writex = 1; + pAsm->D.dst.writey = 1; + pAsm->D.dst.writez = 1; + pAsm->D.dst.writew = 1; + + if( GL_FALSE == assemble_src(pAsm, i, 0) ) + { + return GL_FALSE; + } + + next_ins(pAsm); + } + if(uNumValidSrc > 0) + { + prelude_cf_ptr = pAsm->cf_current_alu_clause_ptr; + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + } + + /* browse thro existing presubs. */ + for(i=0; i<pAsm->unNumPresub; i++) + { + if(pAsm->presubs[i].sptSigniture == scriptSigniture) + { + break; + } + } + + if(i == pAsm->unNumPresub) + { /* not loaded yet */ + /* save assemble context */ + number_used_registers_save = pAsm->number_used_registers; + uFirstHelpReg_save = pAsm->uFirstHelpReg; + uHelpReg_save = pAsm->uHelpReg; + starting_temp_register_number_save = pAsm->starting_temp_register_number; + pILInst_save = pAsm->pILInst; + uiCurInst_save = pAsm->uiCurInst; + + /* alloc in presub */ + if( (pAsm->unNumPresub + 1) > pAsm->unPresubArraySize ) + { + pAsm->presubs = (PRESUB_DESC*)_mesa_realloc( (void *)pAsm->presubs, + sizeof(PRESUB_DESC) * pAsm->unPresubArraySize, + sizeof(PRESUB_DESC) * (pAsm->unPresubArraySize + 4) ); + if(NULL == pAsm->presubs) + { + radeon_error("No memeory to allocate built in shader function description structures. \n"); + return GL_FALSE; + } + pAsm->unPresubArraySize += 4; + } + + pPresubDesc = &(pAsm->presubs[i]); + pPresubDesc->sptSigniture = scriptSigniture; + + /* constants offsets need to be final resolved at reloc. */ + if(0 == pAsm->unNumPresub) + { + pPresubDesc->unConstantsStart = 0; + } + else + { + pPresubDesc->unConstantsStart = pAsm->presubs[i-1].unConstantsStart + + pAsm->presubs[i-1].pCompiledSub->NumParameters; + } + + pPresubDesc->pCompiledSub = pCompiledSub; + + pPresubDesc->subIL_Shift = pAsm->unCurNumILInsts; + pPresubDesc->maxStartReg = uFirstHelpReg_save; + pAsm->unCurNumILInsts += pCompiledSub->NumInstructions; + + pAsm->unNumPresub++; + + /* setup new assemble context */ + pAsm->starting_temp_register_number = 0; + pAsm->number_used_registers = pCompiledSub->NumTemporaries; + pAsm->uFirstHelpReg = pAsm->number_used_registers; + pAsm->uHelpReg = pAsm->uFirstHelpReg; + + bRet = assemble_CAL(pAsm, + 0, + pPresubDesc->subIL_Shift, + pCompiledSub->NumInstructions, + pCompiledSub->Instructions, + pPresubDesc); + + + pPresubDesc->number_used_registers = pAsm->number_used_registers; + + /* restore assemble context */ + pAsm->number_used_registers = number_used_registers_save; + pAsm->uFirstHelpReg = uFirstHelpReg_save; + pAsm->uHelpReg = uHelpReg_save; + pAsm->starting_temp_register_number = starting_temp_register_number_save; + pAsm->pILInst = pILInst_save; + pAsm->uiCurInst = uiCurInst_save; + } + else + { /* was loaded */ + pPresubDesc = &(pAsm->presubs[i]); + + bRet = assemble_CAL(pAsm, + 0, + pPresubDesc->subIL_Shift, + pCompiledSub->NumInstructions, + pCompiledSub->Instructions, + pPresubDesc); + } + + if(GL_FALSE == bRet) + { + radeon_error("Shader presub assemble failed. \n"); + } + else + { + /* copy presub output to real dst */ + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + pAsm->D.dst.opcode = SQ_OP2_INST_MOV; + + if( GL_FALSE == assemble_dst(pAsm) ) + { + return GL_FALSE; + } + + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; + pAsm->S[0].src.reg = pCompiledSub->dstRegIndex; + pAsm->S[0].src.swizzlex = pCompiledSub->outputSwizzleX; + pAsm->S[0].src.swizzley = pCompiledSub->outputSwizzleY; + pAsm->S[0].src.swizzlez = pCompiledSub->outputSwizzleZ; + pAsm->S[0].src.swizzlew = pCompiledSub->outputSwizzleW; + + next_ins(pAsm); + + pAsm->callers[pAsm->unCallerArrayPointer - 1].finale_cf_ptr = pAsm->cf_current_alu_clause_ptr; + pAsm->callers[pAsm->unCallerArrayPointer - 1].prelude_cf_ptr = prelude_cf_ptr; + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + } + + if( (pPresubDesc->number_used_registers + pAsm->uFirstHelpReg) > pAsm->number_used_registers ) + { + pAsm->number_used_registers = pPresubDesc->number_used_registers + pAsm->uFirstHelpReg; + } + if(pAsm->uFirstHelpReg > pPresubDesc->maxStartReg) + { + pPresubDesc->maxStartReg = pAsm->uFirstHelpReg; + } + + return bRet; +} + GLboolean Process_Export(r700_AssemblerBase* pAsm, GLuint type, GLuint export_starting_index, @@ -4791,6 +6526,25 @@ GLboolean Process_Vertex_Exports(r700_AssemblerBase *pR700AsmCode, export_starting_index++; } } + + for(i=VERT_RESULT_VAR0; i<VERT_RESULT_MAX; i++) + { + unBit = 1 << i; + if(OutputsWritten & unBit) + { + if( GL_FALSE == Process_Export(pR700AsmCode, + SQ_EXPORT_PARAM, + export_starting_index, + 1, + pR700AsmCode->ucVP_OutputMap[i], + GL_FALSE) ) + { + return GL_FALSE; + } + + export_starting_index++; + } + } // At least one param should be exported if (export_count) @@ -4825,6 +6579,21 @@ GLboolean Clean_Up_Assembler(r700_AssemblerBase *pR700AsmCode) { FREE(pR700AsmCode->pucOutMask); FREE(pR700AsmCode->pInstDeps); + + if(NULL != pR700AsmCode->subs) + { + FREE(pR700AsmCode->subs); + } + if(NULL != pR700AsmCode->callers) + { + FREE(pR700AsmCode->callers); + } + + if(NULL != pR700AsmCode->presubs) + { + FREE(pR700AsmCode->presubs); + } + return GL_TRUE; } diff --git a/src/mesa/drivers/dri/r600/r700_assembler.h b/src/mesa/drivers/dri/r600/r700_assembler.h index c66db502a17..dbd9860f7d0 100644 --- a/src/mesa/drivers/dri/r600/r700_assembler.h +++ b/src/mesa/drivers/dri/r600/r700_assembler.h @@ -34,6 +34,45 @@ #include "r700_shaderinst.h" #include "r700_shader.h" +typedef enum LOADABLE_SCRIPT_SIGNITURE +{ + GLSL_NOISE1 = 0x10000001, + GLSL_NOISE2 = 0x10000002, + GLSL_NOISE3 = 0x10000003, + GLSL_NOISE4 = 0x10000004 +}LOADABLE_SCRIPT_SIGNITURE; + +typedef struct COMPILED_SUB +{ + struct prog_instruction *Instructions; + GLuint NumInstructions; + GLuint NumTemporaries; + GLuint NumParameters; + GLuint MinRegIndex; + GLfloat (*ParameterValues)[4]; + GLbyte outputSwizzleX; + GLbyte outputSwizzleY; + GLbyte outputSwizzleZ; + GLbyte outputSwizzleW; + GLshort srcRegIndex[3]; + GLushort dstRegIndex; +}COMPILED_SUB; + +typedef struct PRESUB_DESCtag +{ + LOADABLE_SCRIPT_SIGNITURE sptSigniture; + GLint subIL_Shift; + struct prog_src_register InReg[3]; + struct prog_dst_register OutReg; + + GLushort maxStartReg; + GLushort number_used_registers; + + GLuint unConstantsStart; + + COMPILED_SUB * pCompiledSub; +} PRESUB_DESC; + typedef enum SHADER_PIPE_TYPE { SPT_VP = 0, @@ -72,7 +111,8 @@ typedef enum SrcRegisterType SRC_REG_INPUT = 1, SRC_REG_CONSTANT = 2, SRC_REG_ALT_TEMPORARY = 3, - NUMBER_OF_SRC_REG_TYPE = 4 + SRC_REC_LITERAL = 4, + NUMBER_OF_SRC_REG_TYPE = 5 } SrcRegisterType; typedef enum DstRegisterType @@ -111,16 +151,24 @@ typedef struct PVSDSTtag BITS addrmode1:1; //32 } PVSDST; +typedef struct PVSINSTtag +{ + BITS literal_slots :2; + BITS SaturateMode :2; + BITS index_mode :3; +} PVSINST; + typedef struct PVSSRCtag { - BITS rtype:4; + BITS rtype:3; BITS addrmode0:1; - BITS reg:10; //15 (8) + BITS reg:10; //14 (8) BITS swizzlex:3; BITS swizzley:3; BITS swizzlez:3; - BITS swizzlew:3; //27 + BITS swizzlew:3; //26 + BITS abs:1; BITS negx:1; BITS negy:1; BITS negz:1; @@ -148,6 +196,7 @@ typedef union PVSDWORDtag { BITS bits; PVSDST dst; + PVSINST dst2; PVSSRC src; PVSMATH math; float f; @@ -251,6 +300,8 @@ enum FC_IF = 1, FC_LOOP = 2, FC_REP = 3, + FC_PUSH_VPM = 4, + FC_PUSH_WQM = 5, COND_NONE = 0, COND_BOOL = 1, @@ -263,22 +314,56 @@ enum typedef struct FC_LEVEL { - unsigned int first; ///< first fc instruction on level (if, rep, loop) - unsigned int* mid; ///< middle instructions - else or all breaks on this level - unsigned int midLen; - unsigned int type; - unsigned int cond; - unsigned int inv; - unsigned int bpush; ///< 1 if first instruction does branch stack push - int id; ///< id of bool or int variable + R700ControlFlowGenericClause * first; + R700ControlFlowGenericClause ** mid; + unsigned int unNumMid; + unsigned int midLen; + unsigned int type; + unsigned int cond; + unsigned int inv; + int id; ///< id of bool or int variable } FC_LEVEL; typedef struct VTX_FETCH_METHOD { - GLboolean bEnableMini; - GLuint mega_fetch_remainder; + GLboolean bEnableMini; + GLuint mega_fetch_remainder; } VTX_FETCH_METHOD; +typedef struct SUB_OFFSET +{ + GLint subIL_Offset; + GLuint unCFoffset; + GLuint unStackDepthMax; + PRESUB_DESC * pPresubDesc; + TypedShaderList lstCFInstructions_local; +} SUB_OFFSET; + +typedef struct CALLER_POINTER +{ + GLint subIL_Offset; + GLint subDescIndex; + R700ControlFlowGenericClause* cf_ptr; + + R700ControlFlowGenericClause* prelude_cf_ptr; + R700ControlFlowGenericClause* finale_cf_ptr; +} CALLER_POINTER; + +#define SQ_MAX_CALL_DEPTH 0x00000020 + +typedef struct CALL_LEVEL +{ + unsigned int FCSP_BeforeEntry; + GLint subDescIndex; + GLushort current; + GLushort max; + TypedShaderList * plstCFInstructions_local; +} CALL_LEVEL; + +#define HAS_CURRENT_LOOPRET 0x1L +#define HAS_LOOPRET 0x2L +#define LOOPRET_FLAGS HAS_LOOPRET | HAS_CURRENT_LOOPRET + typedef struct r700_AssemblerBase { R700ControlFlowSXClause* cf_last_export_ptr; @@ -294,14 +379,19 @@ typedef struct r700_AssemblerBase // No clause has been created yet CF_CLAUSE_TYPE cf_current_clause_type; + BITS alu_x_opcode; + GLuint number_of_exports; GLuint number_of_colorandz_exports; GLuint number_of_export_opcodes; PVSDWORD D; + PVSDWORD D2; PVSDWORD S[3]; + PVSDWORD C[4]; unsigned int uLastPosUpdate; + unsigned int last_cond_register; OUT_FRAGMENT_FMT_0 fp_stOutFmt0; @@ -310,6 +400,8 @@ typedef struct r700_AssemblerBase unsigned int number_used_registers; unsigned int uUsedConsts; + unsigned int flag_reg_index; + // Fragment programs unsigned int uiFP_AttributeMap[FRAG_ATTRIB_MAX]; unsigned int uiFP_OutputMap[FRAG_RESULT_MAX]; @@ -330,9 +422,6 @@ typedef struct r700_AssemblerBase unsigned int FCSP; FC_LEVEL fc_stack[32]; - unsigned int branch_depth; - unsigned int max_branch_depth; - //----------------------------------------------------------------------------------- // ArgSubst used in Assemble_Source() function //----------------------------------------------------------------------------------- @@ -373,11 +462,31 @@ typedef struct r700_AssemblerBase SHADER_PIPE_TYPE currentShaderType; struct prog_instruction * pILInst; GLuint uiCurInst; + GLubyte SamplerUnits[MAX_SAMPLERS]; GLboolean bR6xx; /* helper to decide which type of instruction to assemble */ GLboolean is_tex; /* we inserted helper intructions and need barrier on next TEX ins */ GLboolean need_tex_barrier; + + SUB_OFFSET * subs; + GLuint unSubArraySize; + GLuint unSubArrayPointer; + CALLER_POINTER * callers; + GLuint unCallerArraySize; + GLuint unCallerArrayPointer; + unsigned int CALLSP; + CALL_LEVEL CALLSTACK[SQ_MAX_CALL_DEPTH]; + + GLuint unCFflags; + + PRESUB_DESC * presubs; + GLuint unPresubArraySize; + GLuint unNumPresub; + GLuint unCurNumILInsts; + + GLuint unVetTexBits; + } r700_AssemblerBase; //Internal use @@ -399,7 +508,7 @@ BITS is_depth_component_exported(OUT_FRAGMENT_FMT_0* pFPOutFmt) ; GLboolean is_reduction_opcode(PVSDWORD * dest); GLuint GetSurfaceFormat(GLenum eType, GLuint nChannels, GLuint * pClient_size); -unsigned int r700GetNumOperands(r700_AssemblerBase* pAsm); +unsigned int r700GetNumOperands(GLuint opcode, GLuint nIsOp3); GLboolean IsTex(gl_inst_opcode Opcode); GLboolean IsAlu(gl_inst_opcode Opcode); @@ -446,6 +555,10 @@ GLboolean assemble_alu_src(R700ALUInstruction* alu_instruction_ptr, GLboolean add_alu_instruction(r700_AssemblerBase* pAsm, R700ALUInstruction* alu_instruction_ptr, GLuint contiguous_slots_needed); + +GLboolean add_cf_instruction(r700_AssemblerBase* pAsm); +void add_return_inst(r700_AssemblerBase *pAsm); + void get_src_properties(R700ALUInstruction* alu_instruction_ptr, int source_index, BITS* psrc_sel, @@ -467,13 +580,20 @@ GLboolean check_vector(r700_AssemblerBase* pAsm, R700ALUInstruction* alu_instruction_ptr); GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm); GLboolean next_ins(r700_AssemblerBase *pAsm); + +GLboolean pops(r700_AssemblerBase *pAsm, GLuint pops); +GLboolean jumpToOffest(r700_AssemblerBase *pAsm, GLuint pops, GLint offset); +GLboolean setRetInLoopFlag(r700_AssemblerBase *pAsm, GLuint flagValue); +GLboolean testFlag(r700_AssemblerBase *pAsm); +GLboolean breakLoopOnFlag(r700_AssemblerBase *pAsm, GLuint unFCSP); +GLboolean returnOnFlag(r700_AssemblerBase *pAsm, GLuint unIF); + GLboolean assemble_math_function(r700_AssemblerBase* pAsm, BITS opcode); GLboolean assemble_ABS(r700_AssemblerBase *pAsm); GLboolean assemble_ADD(r700_AssemblerBase *pAsm); GLboolean assemble_ARL(r700_AssemblerBase *pAsm); GLboolean assemble_BAD(char *opcode_str); GLboolean assemble_CMP(r700_AssemblerBase *pAsm); -GLboolean assemble_COS(r700_AssemblerBase *pAsm); GLboolean assemble_DOT(r700_AssemblerBase *pAsm); GLboolean assemble_DST(r700_AssemblerBase *pAsm); GLboolean assemble_EX2(r700_AssemblerBase *pAsm); @@ -481,7 +601,7 @@ GLboolean assemble_EXP(r700_AssemblerBase *pAsm); GLboolean assemble_FLR(r700_AssemblerBase *pAsm); GLboolean assemble_FLR_INT(r700_AssemblerBase *pAsm); GLboolean assemble_FRC(r700_AssemblerBase *pAsm); -GLboolean assemble_KIL(r700_AssemblerBase *pAsm); +GLboolean assemble_KIL(r700_AssemblerBase *pAsm, GLuint opcode); GLboolean assemble_LG2(r700_AssemblerBase *pAsm); GLboolean assemble_LRP(r700_AssemblerBase *pAsm); GLboolean assemble_LOG(r700_AssemblerBase *pAsm); @@ -494,17 +614,37 @@ GLboolean assemble_MUL(r700_AssemblerBase *pAsm); GLboolean assemble_POW(r700_AssemblerBase *pAsm); GLboolean assemble_RCP(r700_AssemblerBase *pAsm); GLboolean assemble_RSQ(r700_AssemblerBase *pAsm); -GLboolean assemble_SIN(r700_AssemblerBase *pAsm); GLboolean assemble_SCS(r700_AssemblerBase *pAsm); GLboolean assemble_SGE(r700_AssemblerBase *pAsm); + +GLboolean assemble_LOGIC(r700_AssemblerBase *pAsm, BITS opcode); +GLboolean assemble_LOGIC_PRED(r700_AssemblerBase *pAsm, BITS opcode); +GLboolean assemble_TRIG(r700_AssemblerBase *pAsm, BITS opcode); + GLboolean assemble_SLT(r700_AssemblerBase *pAsm); GLboolean assemble_STP(r700_AssemblerBase *pAsm); GLboolean assemble_TEX(r700_AssemblerBase *pAsm); GLboolean assemble_XPD(r700_AssemblerBase *pAsm); GLboolean assemble_EXPORT(r700_AssemblerBase *pAsm); -GLboolean assemble_IF(r700_AssemblerBase *pAsm); +GLboolean assemble_IF(r700_AssemblerBase *pAsm, GLboolean bHasElse); +GLboolean assemble_ELSE(r700_AssemblerBase *pAsm); GLboolean assemble_ENDIF(r700_AssemblerBase *pAsm); +GLboolean assemble_BGNLOOP(r700_AssemblerBase *pAsm); +GLboolean assemble_BRK(r700_AssemblerBase *pAsm); +GLboolean assemble_COND(r700_AssemblerBase *pAsm); +GLboolean assemble_ENDLOOP(r700_AssemblerBase *pAsm); + +GLboolean assemble_BGNSUB(r700_AssemblerBase *pAsm, GLint nILindex, GLuint uiIL_Shift); +GLboolean assemble_ENDSUB(r700_AssemblerBase *pAsm); +GLboolean assemble_RET(r700_AssemblerBase *pAsm); +GLboolean assemble_CAL(r700_AssemblerBase *pAsm, + GLint nILindex, + GLuint uiIL_Offest, + GLuint uiNumberInsts, + struct prog_instruction *pILInst, + PRESUB_DESC * pPresubDesc); + GLboolean Process_Export(r700_AssemblerBase* pAsm, GLuint type, GLuint export_starting_index, @@ -514,14 +654,25 @@ GLboolean Process_Export(r700_AssemblerBase* pAsm, GLboolean Move_Depth_Exports_To_Correct_Channels(r700_AssemblerBase *pAsm, BITS depth_channel_select); +GLboolean callPreSub(r700_AssemblerBase* pAsm, + LOADABLE_SCRIPT_SIGNITURE scriptSigniture, + /* struct prog_instruction ** pILInstParent, */ + COMPILED_SUB * pCompiledSub, + GLshort uOutReg, + GLshort uNumValidSrc); //Interface -GLboolean AssembleInstr(GLuint uiNumberInsts, +GLboolean AssembleInstr(GLuint uiFirstInst, + GLuint uiIL_Shift, + GLuint uiNumberInsts, struct prog_instruction *pILInst, r700_AssemblerBase *pR700AsmCode); GLboolean Process_Fragment_Exports(r700_AssemblerBase *pR700AsmCode, GLbitfield OutputsWritten); GLboolean Process_Vertex_Exports(r700_AssemblerBase *pR700AsmCode, GLbitfield OutputsWritten); +GLboolean RelocProgram(r700_AssemblerBase * pAsm, struct gl_program * pILProg); +GLboolean InitShaderProgram(r700_AssemblerBase * pAsm); + int Init_r700_AssemblerBase(SHADER_PIPE_TYPE spt, r700_AssemblerBase* pAsm, R700_Shader* pShader); GLboolean Clean_Up_Assembler(r700_AssemblerBase *pR700AsmCode); diff --git a/src/mesa/drivers/dri/r600/r700_chip.c b/src/mesa/drivers/dri/r600/r700_chip.c index 02c56b98d9b..c124e02184f 100644 --- a/src/mesa/drivers/dri/r600/r700_chip.c +++ b/src/mesa/drivers/dri/r600/r700_chip.c @@ -45,6 +45,9 @@ static void r700SendTexState(GLcontext *ctx, struct radeon_state_atom *atom) { context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); + + struct r700_vertex_program *vp = context->selected_vp; + struct radeon_bo *bo = NULL; unsigned int i; BATCH_LOCALS(&context->radeon); @@ -52,7 +55,7 @@ static void r700SendTexState(GLcontext *ctx, struct radeon_state_atom *atom) radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); for (i = 0; i < R700_TEXTURE_NUMBERUNITS; i++) { - if (ctx->Texture.Unit[i]._ReallyEnabled) { + if (ctx->Texture.Unit[i]._ReallyEnabled) { radeonTexObj *t = r700->textures[i]; uint32_t offset; if (t) { @@ -71,7 +74,16 @@ static void r700SendTexState(GLcontext *ctx, struct radeon_state_atom *atom) BEGIN_BATCH_NO_AUTOSTATE(9 + 4); R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_RESOURCE, 7)); - R600_OUT_BATCH(i * 7); + + if( (1<<i) & vp->r700AsmCode.unVetTexBits ) + { /* vs texture */ + R600_OUT_BATCH((i + VERT_ATTRIB_MAX + SQ_FETCH_RESOURCE_VS_OFFSET) * FETCH_RESOURCE_STRIDE); + } + else + { + R600_OUT_BATCH(i * 7); + } + R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE0); R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE1); R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE2); @@ -95,21 +107,35 @@ static void r700SendTexState(GLcontext *ctx, struct radeon_state_atom *atom) } } +#define SAMPLER_STRIDE 3 + static void r700SendTexSamplerState(GLcontext *ctx, struct radeon_state_atom *atom) { context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); unsigned int i; + + struct r700_vertex_program *vp = context->selected_vp; + BATCH_LOCALS(&context->radeon); radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); for (i = 0; i < R700_TEXTURE_NUMBERUNITS; i++) { - if (ctx->Texture.Unit[i]._ReallyEnabled) { + if (ctx->Texture.Unit[i]._ReallyEnabled) { radeonTexObj *t = r700->textures[i]; if (t) { BEGIN_BATCH_NO_AUTOSTATE(5); R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_SAMPLER, 3)); - R600_OUT_BATCH(i * 3); + + if( (1<<i) & vp->r700AsmCode.unVetTexBits ) + { /* vs texture */ + R600_OUT_BATCH((i+SQ_TEX_SAMPLER_VS_OFFSET) * SAMPLER_STRIDE); //work 1 + } + else + { + R600_OUT_BATCH(i * 3); + } + R600_OUT_BATCH(r700->textures[i]->SQ_TEX_SAMPLER0); R600_OUT_BATCH(r700->textures[i]->SQ_TEX_SAMPLER1); R600_OUT_BATCH(r700->textures[i]->SQ_TEX_SAMPLER2); @@ -446,68 +472,77 @@ static void r700SendRenderTargetState(GLcontext *ctx, struct radeon_state_atom * static void r700SendPSState(GLcontext *ctx, struct radeon_state_atom *atom) { - context_t *context = R700_CONTEXT(ctx); - R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); - struct radeon_bo * pbo; - BATCH_LOCALS(&context->radeon); - radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); + context_t *context = R700_CONTEXT(ctx); + R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); + struct radeon_bo * pbo; + BATCH_LOCALS(&context->radeon); + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); - pbo = (struct radeon_bo *)r700GetActiveFpShaderBo(GL_CONTEXT(context)); + pbo = (struct radeon_bo *)r700GetActiveFpShaderBo(GL_CONTEXT(context)); - if (!pbo) - return; + if (!pbo) + return; - r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit); + r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit); - BEGIN_BATCH_NO_AUTOSTATE(3 + 2); - R600_OUT_BATCH_REGSEQ(SQ_PGM_START_PS, 1); - R600_OUT_BATCH(r700->ps.SQ_PGM_START_PS.u32All); - R600_OUT_BATCH_RELOC(r700->ps.SQ_PGM_START_PS.u32All, - pbo, - r700->ps.SQ_PGM_START_PS.u32All, - RADEON_GEM_DOMAIN_GTT, 0, 0); - END_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(3 + 2); + R600_OUT_BATCH_REGSEQ(SQ_PGM_START_PS, 1); + R600_OUT_BATCH(r700->ps.SQ_PGM_START_PS.u32All); + R600_OUT_BATCH_RELOC(r700->ps.SQ_PGM_START_PS.u32All, + pbo, + r700->ps.SQ_PGM_START_PS.u32All, + RADEON_GEM_DOMAIN_GTT, 0, 0); + END_BATCH(); - BEGIN_BATCH_NO_AUTOSTATE(9); - R600_OUT_BATCH_REGVAL(SQ_PGM_RESOURCES_PS, r700->ps.SQ_PGM_RESOURCES_PS.u32All); - R600_OUT_BATCH_REGVAL(SQ_PGM_EXPORTS_PS, r700->ps.SQ_PGM_EXPORTS_PS.u32All); - R600_OUT_BATCH_REGVAL(SQ_PGM_CF_OFFSET_PS, r700->ps.SQ_PGM_CF_OFFSET_PS.u32All); - END_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(9); + R600_OUT_BATCH_REGVAL(SQ_PGM_RESOURCES_PS, r700->ps.SQ_PGM_RESOURCES_PS.u32All); + R600_OUT_BATCH_REGVAL(SQ_PGM_EXPORTS_PS, r700->ps.SQ_PGM_EXPORTS_PS.u32All); + R600_OUT_BATCH_REGVAL(SQ_PGM_CF_OFFSET_PS, r700->ps.SQ_PGM_CF_OFFSET_PS.u32All); + END_BATCH(); - COMMIT_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(3); + R600_OUT_BATCH_REGVAL(SQ_LOOP_CONST_0, 0x01000FFF); + END_BATCH(); + + COMMIT_BATCH(); } static void r700SendVSState(GLcontext *ctx, struct radeon_state_atom *atom) { - context_t *context = R700_CONTEXT(ctx); - R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); - struct radeon_bo * pbo; - BATCH_LOCALS(&context->radeon); - radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); + context_t *context = R700_CONTEXT(ctx); + R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); + struct radeon_bo * pbo; + BATCH_LOCALS(&context->radeon); + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); - pbo = (struct radeon_bo *)r700GetActiveVpShaderBo(GL_CONTEXT(context)); + pbo = (struct radeon_bo *)r700GetActiveVpShaderBo(GL_CONTEXT(context)); - if (!pbo) - return; + if (!pbo) + return; - r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit); + r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit); - BEGIN_BATCH_NO_AUTOSTATE(3 + 2); - R600_OUT_BATCH_REGSEQ(SQ_PGM_START_VS, 1); - R600_OUT_BATCH(r700->vs.SQ_PGM_START_VS.u32All); - R600_OUT_BATCH_RELOC(r700->vs.SQ_PGM_START_VS.u32All, - pbo, - r700->vs.SQ_PGM_START_VS.u32All, - RADEON_GEM_DOMAIN_GTT, 0, 0); - END_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(3 + 2); + R600_OUT_BATCH_REGSEQ(SQ_PGM_START_VS, 1); + R600_OUT_BATCH(r700->vs.SQ_PGM_START_VS.u32All); + R600_OUT_BATCH_RELOC(r700->vs.SQ_PGM_START_VS.u32All, + pbo, + r700->vs.SQ_PGM_START_VS.u32All, + RADEON_GEM_DOMAIN_GTT, 0, 0); + END_BATCH(); - BEGIN_BATCH_NO_AUTOSTATE(6); - R600_OUT_BATCH_REGVAL(SQ_PGM_RESOURCES_VS, r700->vs.SQ_PGM_RESOURCES_VS.u32All); - R600_OUT_BATCH_REGVAL(SQ_PGM_CF_OFFSET_VS, r700->vs.SQ_PGM_CF_OFFSET_VS.u32All); - END_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(6); + R600_OUT_BATCH_REGVAL(SQ_PGM_RESOURCES_VS, r700->vs.SQ_PGM_RESOURCES_VS.u32All); + R600_OUT_BATCH_REGVAL(SQ_PGM_CF_OFFSET_VS, r700->vs.SQ_PGM_CF_OFFSET_VS.u32All); + END_BATCH(); - COMMIT_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(3); + R600_OUT_BATCH_REGVAL((SQ_LOOP_CONST_0 + 32*4), 0x0100000F); + //R600_OUT_BATCH_REGVAL((SQ_LOOP_CONST_0 + (SQ_LOOP_CONST_vs<2)), 0x0100000F); + END_BATCH(); + + COMMIT_BATCH(); } static void r700SendFSState(GLcontext *ctx, struct radeon_state_atom *atom) @@ -1309,8 +1344,8 @@ void r600InitAtoms(context_t *context) ALLOC_STATE(spi, always, (59 + R700_MAX_SHADER_EXPORTS), r700SendSPIState); ALLOC_STATE(vpt, always, 16, r700SendViewportState); ALLOC_STATE(fs, always, 18, r700SendFSState); - ALLOC_STATE(vs, always, 18, r700SendVSState); - ALLOC_STATE(ps, always, 21, r700SendPSState); + ALLOC_STATE(vs, always, 21, r700SendVSState); + ALLOC_STATE(ps, always, 24, r700SendPSState); ALLOC_STATE(vs_consts, vs_consts, (2 + (R700_MAX_DX9_CONSTS * 4)), r700SendVSConsts); ALLOC_STATE(ps_consts, ps_consts, (2 + (R700_MAX_DX9_CONSTS * 4)), r700SendPSConsts); ALLOC_STATE(vtx, vtx, (6 + (VERT_ATTRIB_MAX * 18)), r700SendVTXState); diff --git a/src/mesa/drivers/dri/r600/r700_clear.c b/src/mesa/drivers/dri/r600/r700_clear.c index c6546ab00c2..526d3843d16 100644 --- a/src/mesa/drivers/dri/r600/r700_clear.c +++ b/src/mesa/drivers/dri/r600/r700_clear.c @@ -57,6 +57,10 @@ void r700Clear(GLcontext * ctx, GLbitfield mask) radeon_print(RADEON_RENDER, RADEON_VERBOSE, "%s %x\n", __func__, mask); + if (mask & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_FRONT_RIGHT)) { + context->radeon.front_buffer_dirty = GL_TRUE; + } + if( GL_TRUE == r700ClearFast(context, mask) ) { return; diff --git a/src/mesa/drivers/dri/r600/r700_fragprog.c b/src/mesa/drivers/dri/r600/r700_fragprog.c index ccafd433bfa..ce2d9fdf79e 100644 --- a/src/mesa/drivers/dri/r600/r700_fragprog.c +++ b/src/mesa/drivers/dri/r600/r700_fragprog.c @@ -34,6 +34,7 @@ #include "main/imports.h" #include "shader/prog_parameter.h" #include "shader/prog_statevars.h" +#include "shader/program.h" #include "r600_context.h" #include "r600_cmdbuf.h" @@ -42,14 +43,68 @@ #include "r700_debug.h" +void insert_wpos_code(GLcontext *ctx, struct gl_fragment_program *fprog) +{ + static const gl_state_index winstate[STATE_LENGTH] + = { STATE_INTERNAL, STATE_FB_SIZE, 0, 0, 0}; + struct prog_instruction *newInst, *inst; + GLint win_size; /* state reference */ + GLuint wpos_temp; /* temp register */ + int i, j; + + /* PARAM win_size = STATE_FB_SIZE */ + win_size = _mesa_add_state_reference(fprog->Base.Parameters, winstate); + + wpos_temp = fprog->Base.NumTemporaries++; + + /* scan program where WPOS is used and replace with wpos_temp */ + inst = fprog->Base.Instructions; + for (i = 0; i < fprog->Base.NumInstructions; i++) { + for (j=0; j < 3; j++) { + if(inst->SrcReg[j].File == PROGRAM_INPUT && + inst->SrcReg[j].Index == FRAG_ATTRIB_WPOS) { + inst->SrcReg[j].File = PROGRAM_TEMPORARY; + inst->SrcReg[j].Index = wpos_temp; + } + } + inst++; + } + + _mesa_insert_instructions(&(fprog->Base), 0, 1); + + newInst = fprog->Base.Instructions; + /* invert wpos.y + * wpos_temp.xyzw = wpos.x-yzw + winsize.0y00 */ + newInst[0].Opcode = OPCODE_ADD; + newInst[0].DstReg.File = PROGRAM_TEMPORARY; + newInst[0].DstReg.Index = wpos_temp; + newInst[0].DstReg.WriteMask = WRITEMASK_XYZW; + + newInst[0].SrcReg[0].File = PROGRAM_INPUT; + newInst[0].SrcReg[0].Index = FRAG_ATTRIB_WPOS; + newInst[0].SrcReg[0].Swizzle = SWIZZLE_XYZW; + newInst[0].SrcReg[0].Negate = NEGATE_Y; + + newInst[0].SrcReg[1].File = PROGRAM_STATE_VAR; + newInst[0].SrcReg[1].Index = win_size; + newInst[0].SrcReg[1].Swizzle = MAKE_SWIZZLE4(SWIZZLE_ZERO, SWIZZLE_Y, SWIZZLE_ZERO, SWIZZLE_ZERO); + +} + //TODO : Validate FP input with VP output. void Map_Fragment_Program(r700_AssemblerBase *pAsm, - struct gl_fragment_program *mesa_fp) + struct gl_fragment_program *mesa_fp, + GLcontext *ctx) { unsigned int unBit; unsigned int i; GLuint ui; + /* match fp inputs with vp exports. */ + struct r700_vertex_program_cont *vpc = + (struct r700_vertex_program_cont *)ctx->VertexProgram._Current; + GLbitfield OutputsWritten = vpc->mesa_program.Base.OutputsWritten; + pAsm->number_used_registers = 0; //Input mapping : mesa_fp->Base.InputsRead set the flag, set in @@ -61,32 +116,99 @@ void Map_Fragment_Program(r700_AssemblerBase *pAsm, pAsm->uiFP_AttributeMap[FRAG_ATTRIB_WPOS] = pAsm->number_used_registers++; } - unBit = 1 << FRAG_ATTRIB_COL0; - if(mesa_fp->Base.InputsRead & unBit) + unBit = 1 << VERT_RESULT_COL0; + if(OutputsWritten & unBit) { pAsm->uiFP_AttributeMap[FRAG_ATTRIB_COL0] = pAsm->number_used_registers++; } - unBit = 1 << FRAG_ATTRIB_COL1; - if(mesa_fp->Base.InputsRead & unBit) + unBit = 1 << VERT_RESULT_COL1; + if(OutputsWritten & unBit) { pAsm->uiFP_AttributeMap[FRAG_ATTRIB_COL1] = pAsm->number_used_registers++; } - unBit = 1 << FRAG_ATTRIB_FOGC; - if(mesa_fp->Base.InputsRead & unBit) - { - pAsm->uiFP_AttributeMap[FRAG_ATTRIB_FOGC] = pAsm->number_used_registers++; - } + unBit = 1 << VERT_RESULT_FOGC; + if(OutputsWritten & unBit) + { + pAsm->uiFP_AttributeMap[FRAG_ATTRIB_FOGC] = pAsm->number_used_registers++; + } for(i=0; i<8; i++) { - unBit = 1 << (FRAG_ATTRIB_TEX0 + i); - if(mesa_fp->Base.InputsRead & unBit) + unBit = 1 << (VERT_RESULT_TEX0 + i); + if(OutputsWritten & unBit) { pAsm->uiFP_AttributeMap[FRAG_ATTRIB_TEX0 + i] = pAsm->number_used_registers++; } } + +/* order has been taken care of */ +#if 1 + for(i=VERT_RESULT_VAR0; i<VERT_RESULT_MAX; i++) + { + unBit = 1 << i; + if(OutputsWritten & unBit) + { + pAsm->uiFP_AttributeMap[i-VERT_RESULT_VAR0+FRAG_ATTRIB_VAR0] = pAsm->number_used_registers++; + } + } +#else + if( (mesa_fp->Base.InputsRead >> FRAG_ATTRIB_VAR0) > 0 ) + { + struct r700_vertex_program_cont *vpc = + (struct r700_vertex_program_cont *)ctx->VertexProgram._Current; + struct gl_program_parameter_list * VsVarying = vpc->mesa_program.Base.Varying; + struct gl_program_parameter_list * PsVarying = mesa_fp->Base.Varying; + struct gl_program_parameter * pVsParam; + struct gl_program_parameter * pPsParam; + GLuint j, k; + GLuint unMaxVarying = 0; + + for(i=0; i<VsVarying->NumParameters; i++) + { + pAsm->uiFP_AttributeMap[i + FRAG_ATTRIB_VAR0] = 0; + } + + for(i=FRAG_ATTRIB_VAR0; i<FRAG_ATTRIB_MAX; i++) + { + unBit = 1 << i; + if(mesa_fp->Base.InputsRead & unBit) + { + j = i - FRAG_ATTRIB_VAR0; + pPsParam = PsVarying->Parameters + j; + + for(k=0; k<VsVarying->NumParameters; k++) + { + pVsParam = VsVarying->Parameters + k; + + if( strcmp(pPsParam->Name, pVsParam->Name) == 0) + { + pAsm->uiFP_AttributeMap[i] = pAsm->number_used_registers + k; + if(k > unMaxVarying) + { + unMaxVarying = k; + } + break; + } + } + } + } + + pAsm->number_used_registers += unMaxVarying + 1; + } +#endif + unBit = 1 << FRAG_ATTRIB_FACE; + if(mesa_fp->Base.InputsRead & unBit) + { + pAsm->uiFP_AttributeMap[FRAG_ATTRIB_FACE] = pAsm->number_used_registers++; + } + + unBit = 1 << FRAG_ATTRIB_PNTC; + if(mesa_fp->Base.InputsRead & unBit) + { + pAsm->uiFP_AttributeMap[FRAG_ATTRIB_PNTC] = pAsm->number_used_registers++; + } /* Map temporary registers (GPRs) */ pAsm->starting_temp_register_number = pAsm->number_used_registers; @@ -127,6 +249,8 @@ void Map_Fragment_Program(r700_AssemblerBase *pAsm, pAsm->pucOutMask[ui] = 0x0; } + pAsm->flag_reg_index = pAsm->number_used_registers++; + pAsm->uFirstHelpReg = pAsm->number_used_registers; } @@ -233,22 +357,41 @@ GLboolean Find_Instruction_Dependencies_fp(struct r700_fragment_program *fp, } GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp, - struct gl_fragment_program *mesa_fp) + struct gl_fragment_program *mesa_fp, + GLcontext *ctx) { GLuint number_of_colors_exported; GLboolean z_enabled = GL_FALSE; GLuint unBit; + int i; //Init_Program Init_r700_AssemblerBase( SPT_FP, &(fp->r700AsmCode), &(fp->r700Shader) ); - Map_Fragment_Program(&(fp->r700AsmCode), mesa_fp); + + if(mesa_fp->Base.InputsRead & FRAG_BIT_WPOS) + { + insert_wpos_code(ctx, mesa_fp); + } + + Map_Fragment_Program(&(fp->r700AsmCode), mesa_fp, ctx); if( GL_FALSE == Find_Instruction_Dependencies_fp(fp, mesa_fp) ) { return GL_FALSE; } + + InitShaderProgram(&(fp->r700AsmCode)); - if( GL_FALSE == AssembleInstr(mesa_fp->Base.NumInstructions, + for(i=0; i < MAX_SAMPLERS; i++) + { + fp->r700AsmCode.SamplerUnits[i] = fp->mesa_program.Base.SamplerUnits[i]; + } + + fp->r700AsmCode.unCurNumILInsts = mesa_fp->Base.NumInstructions; + + if( GL_FALSE == AssembleInstr(0, + 0, + mesa_fp->Base.NumInstructions, &(mesa_fp->Base.Instructions[0]), &(fp->r700AsmCode)) ) { @@ -260,6 +403,11 @@ GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp, return GL_FALSE; } + if( GL_FALSE == RelocProgram(&(fp->r700AsmCode), &(mesa_fp->Base)) ) + { + return GL_FALSE; + } + fp->r700Shader.nRegs = (fp->r700AsmCode.number_used_registers == 0) ? 0 : (fp->r700AsmCode.number_used_registers - 1); @@ -300,7 +448,7 @@ void r700SelectFragmentShader(GLcontext *ctx) } if (GL_FALSE == fp->translated) - r700TranslateFragmentShader(fp, &(fp->mesa_program)); + r700TranslateFragmentShader(fp, &(fp->mesa_program), ctx); } void * r700GetActiveFpShaderBo(GLcontext * ctx) @@ -325,6 +473,7 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) unsigned int unNumOfReg; unsigned int unBit; GLuint exportCount; + GLboolean point_sprite = GL_FALSE; if(GL_FALSE == fp->loaded) { @@ -378,6 +527,50 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) CLEARbit(r700->SPI_INPUT_Z.u32All, PROVIDE_Z_TO_SPI_bit); } + if (mesa_fp->Base.InputsRead & (1 << FRAG_ATTRIB_FACE)) + { + ui += 1; + SETfield(r700->SPI_PS_IN_CONTROL_0.u32All, ui, NUM_INTERP_shift, NUM_INTERP_mask); + SETbit(r700->SPI_PS_IN_CONTROL_1.u32All, FRONT_FACE_ENA_bit); + SETbit(r700->SPI_PS_IN_CONTROL_1.u32All, FRONT_FACE_ALL_BITS_bit); + SETfield(r700->SPI_PS_IN_CONTROL_1.u32All, pAsm->uiFP_AttributeMap[FRAG_ATTRIB_FACE], FRONT_FACE_ADDR_shift, FRONT_FACE_ADDR_mask); + } + else + { + CLEARbit(r700->SPI_PS_IN_CONTROL_1.u32All, FRONT_FACE_ENA_bit); + } + + /* see if we need any point_sprite replacements */ + for (i = VERT_RESULT_TEX0; i<= VERT_RESULT_TEX7; i++) + { + if(ctx->Point.CoordReplace[i - VERT_RESULT_TEX0] == GL_TRUE) + point_sprite = GL_TRUE; + } + + if ((mesa_fp->Base.InputsRead & (1 << FRAG_ATTRIB_PNTC)) || point_sprite) + { + /* for FRAG_ATTRIB_PNTC we need to increase num_interp */ + if(mesa_fp->Base.InputsRead & (1 << FRAG_ATTRIB_PNTC)) + { + ui++; + SETfield(r700->SPI_PS_IN_CONTROL_0.u32All, ui, NUM_INTERP_shift, NUM_INTERP_mask); + } + SETbit(r700->SPI_INTERP_CONTROL_0.u32All, PNT_SPRITE_ENA_bit); + SETfield(r700->SPI_INTERP_CONTROL_0.u32All, SPI_PNT_SPRITE_SEL_S, PNT_SPRITE_OVRD_X_shift, PNT_SPRITE_OVRD_X_mask); + SETfield(r700->SPI_INTERP_CONTROL_0.u32All, SPI_PNT_SPRITE_SEL_T, PNT_SPRITE_OVRD_Y_shift, PNT_SPRITE_OVRD_Y_mask); + SETfield(r700->SPI_INTERP_CONTROL_0.u32All, SPI_PNT_SPRITE_SEL_0, PNT_SPRITE_OVRD_Z_shift, PNT_SPRITE_OVRD_Z_mask); + SETfield(r700->SPI_INTERP_CONTROL_0.u32All, SPI_PNT_SPRITE_SEL_1, PNT_SPRITE_OVRD_W_shift, PNT_SPRITE_OVRD_W_mask); + if(ctx->Point.SpriteOrigin == GL_LOWER_LEFT) + SETbit(r700->SPI_INTERP_CONTROL_0.u32All, PNT_SPRITE_TOP_1_bit); + else + CLEARbit(r700->SPI_INTERP_CONTROL_0.u32All, PNT_SPRITE_TOP_1_bit); + } + else + { + CLEARbit(r700->SPI_INTERP_CONTROL_0.u32All, PNT_SPRITE_ENA_bit); + } + + ui = (unNumOfReg < ui) ? ui : unNumOfReg; SETfield(r700->ps.SQ_PGM_RESOURCES_PS.u32All, ui, NUM_GPRS_shift, NUM_GPRS_mask); @@ -394,6 +587,13 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) EXPORT_MODE_shift, EXPORT_MODE_mask); // emit ps input map + struct r700_vertex_program_cont *vpc = + (struct r700_vertex_program_cont *)ctx->VertexProgram._Current; + GLbitfield OutputsWritten = vpc->mesa_program.Base.OutputsWritten; + + for(ui = 0; ui < R700_MAX_SHADER_EXPORTS; ui++) + r700->SPI_PS_INPUT_CNTL[ui].u32All = 0; + unBit = 1 << FRAG_ATTRIB_WPOS; if(mesa_fp->Base.InputsRead & unBit) { @@ -407,8 +607,8 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); } - unBit = 1 << FRAG_ATTRIB_COL0; - if(mesa_fp->Base.InputsRead & unBit) + unBit = 1 << VERT_RESULT_COL0; + if(OutputsWritten & unBit) { ui = pAsm->uiFP_AttributeMap[FRAG_ATTRIB_COL0]; SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, SEL_CENTROID_bit); @@ -420,8 +620,8 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); } - unBit = 1 << FRAG_ATTRIB_COL1; - if(mesa_fp->Base.InputsRead & unBit) + unBit = 1 << VERT_RESULT_COL1; + if(OutputsWritten & unBit) { ui = pAsm->uiFP_AttributeMap[FRAG_ATTRIB_COL1]; SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, SEL_CENTROID_bit); @@ -433,8 +633,8 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); } - unBit = 1 << FRAG_ATTRIB_FOGC; - if(mesa_fp->Base.InputsRead & unBit) + unBit = 1 << VERT_RESULT_FOGC; + if(OutputsWritten & unBit) { ui = pAsm->uiFP_AttributeMap[FRAG_ATTRIB_FOGC]; SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, SEL_CENTROID_bit); @@ -448,17 +648,67 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) for(i=0; i<8; i++) { - unBit = 1 << (FRAG_ATTRIB_TEX0 + i); - if(mesa_fp->Base.InputsRead & unBit) + unBit = 1 << (VERT_RESULT_TEX0 + i); + if(OutputsWritten & unBit) { ui = pAsm->uiFP_AttributeMap[FRAG_ATTRIB_TEX0 + i]; SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, SEL_CENTROID_bit); SETfield(r700->SPI_PS_INPUT_CNTL[ui].u32All, ui, SEMANTIC_shift, SEMANTIC_mask); CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); + /* ARB_point_sprite */ + if(ctx->Point.CoordReplace[i] == GL_TRUE) + { + SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, PT_SPRITE_TEX_bit); + } } } + unBit = 1 << FRAG_ATTRIB_FACE; + if(mesa_fp->Base.InputsRead & unBit) + { + ui = pAsm->uiFP_AttributeMap[FRAG_ATTRIB_FACE]; + SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, SEL_CENTROID_bit); + SETfield(r700->SPI_PS_INPUT_CNTL[ui].u32All, ui, + SEMANTIC_shift, SEMANTIC_mask); + if (r700->SPI_INTERP_CONTROL_0.u32All & FLAT_SHADE_ENA_bit) + SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); + else + CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); + } + unBit = 1 << FRAG_ATTRIB_PNTC; + if(mesa_fp->Base.InputsRead & unBit) + { + ui = pAsm->uiFP_AttributeMap[FRAG_ATTRIB_PNTC]; + SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, SEL_CENTROID_bit); + SETfield(r700->SPI_PS_INPUT_CNTL[ui].u32All, ui, + SEMANTIC_shift, SEMANTIC_mask); + if (r700->SPI_INTERP_CONTROL_0.u32All & FLAT_SHADE_ENA_bit) + SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); + else + CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); + SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, PT_SPRITE_TEX_bit); + } + + + + + for(i=VERT_RESULT_VAR0; i<VERT_RESULT_MAX; i++) + { + unBit = 1 << i; + if(OutputsWritten & unBit) + { + ui = pAsm->uiFP_AttributeMap[i-VERT_RESULT_VAR0+FRAG_ATTRIB_VAR0]; + SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, SEL_CENTROID_bit); + SETfield(r700->SPI_PS_INPUT_CNTL[ui].u32All, ui, + SEMANTIC_shift, SEMANTIC_mask); + if (r700->SPI_INTERP_CONTROL_0.u32All & FLAT_SHADE_ENA_bit) + SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); + else + CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); + } + } + exportCount = (r700->ps.SQ_PGM_EXPORTS_PS.u32All & EXPORT_MODE_mask) / (1 << EXPORT_MODE_shift); if (r700->CB_SHADER_CONTROL.u32All != ((1 << exportCount) - 1)) { @@ -469,7 +719,8 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) /* sent out shader constants. */ paramList = fp->mesa_program.Base.Parameters; - if(NULL != paramList) { + if(NULL != paramList) + { _mesa_load_state_parameters(ctx, paramList); if (paramList->NumParameters > R700_MAX_DX9_CONSTS) @@ -482,14 +733,33 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) unNumParamData = paramList->NumParameters; for(ui=0; ui<unNumParamData; ui++) { - r700->ps.consts[ui][0].f32All = paramList->ParameterValues[ui][0]; - r700->ps.consts[ui][1].f32All = paramList->ParameterValues[ui][1]; - r700->ps.consts[ui][2].f32All = paramList->ParameterValues[ui][2]; - r700->ps.consts[ui][3].f32All = paramList->ParameterValues[ui][3]; + r700->ps.consts[ui][0].f32All = paramList->ParameterValues[ui][0]; + r700->ps.consts[ui][1].f32All = paramList->ParameterValues[ui][1]; + r700->ps.consts[ui][2].f32All = paramList->ParameterValues[ui][2]; + r700->ps.consts[ui][3].f32All = paramList->ParameterValues[ui][3]; } } else r700->ps.num_consts = 0; + COMPILED_SUB * pCompiledSub; + GLuint uj; + GLuint unConstOffset = r700->ps.num_consts; + for(ui=0; ui<pAsm->unNumPresub; ui++) + { + pCompiledSub = pAsm->presubs[ui].pCompiledSub; + + r700->ps.num_consts += pCompiledSub->NumParameters; + + for(uj=0; uj<pCompiledSub->NumParameters; uj++) + { + r700->ps.consts[uj + unConstOffset][0].f32All = pCompiledSub->ParameterValues[uj][0]; + r700->ps.consts[uj + unConstOffset][1].f32All = pCompiledSub->ParameterValues[uj][1]; + r700->ps.consts[uj + unConstOffset][2].f32All = pCompiledSub->ParameterValues[uj][2]; + r700->ps.consts[uj + unConstOffset][3].f32All = pCompiledSub->ParameterValues[uj][3]; + } + unConstOffset += pCompiledSub->NumParameters; + } + return GL_TRUE; } diff --git a/src/mesa/drivers/dri/r600/r700_fragprog.h b/src/mesa/drivers/dri/r600/r700_fragprog.h index cbb108d2128..39c59c9201d 100644 --- a/src/mesa/drivers/dri/r600/r700_fragprog.h +++ b/src/mesa/drivers/dri/r600/r700_fragprog.h @@ -48,13 +48,17 @@ struct r700_fragment_program }; /* Internal */ +void insert_wpos_code(GLcontext *ctx, struct gl_fragment_program *fprog); + void Map_Fragment_Program(r700_AssemblerBase *pAsm, - struct gl_fragment_program *mesa_fp); + struct gl_fragment_program *mesa_fp, + GLcontext *ctx); GLboolean Find_Instruction_Dependencies_fp(struct r700_fragment_program *fp, struct gl_fragment_program *mesa_fp); GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp, - struct gl_fragment_program *mesa_vp); + struct gl_fragment_program *mesa_vp, + GLcontext *ctx); /* Interface */ extern void r700SelectFragmentShader(GLcontext *ctx); diff --git a/src/mesa/drivers/dri/r600/r700_render.c b/src/mesa/drivers/dri/r600/r700_render.c index 47f89c91f87..eab27cbd842 100644 --- a/src/mesa/drivers/dri/r600/r700_render.c +++ b/src/mesa/drivers/dri/r600/r700_render.c @@ -526,6 +526,9 @@ static void r700ConvertAttrib(GLcontext *ctx, int count, radeonAllocDmaRegion(&context->radeon, &attr->bo, &attr->bo_offset, sizeof(GLfloat) * input->Size * count, 32); + + radeon_bo_map(attr->bo, 1); + dst_ptr = (GLfloat *)ADD_POINTERS(attr->bo->ptr, attr->bo_offset); assert(src_ptr != NULL); @@ -559,6 +562,8 @@ static void r700ConvertAttrib(GLcontext *ctx, int count, break; } + radeon_bo_unmap(attr->bo); + if (mapped_named_bo) { ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER, input->BufferObj); @@ -577,6 +582,8 @@ static void r700AlignDataToDword(GLcontext *ctx, radeonAllocDmaRegion(&context->radeon, &attr->bo, &attr->bo_offset, size, 32); + radeon_bo_map(attr->bo, 1); + if (!input->BufferObj->Pointer) { ctx->Driver.MapBuffer(ctx, GL_ARRAY_BUFFER, GL_READ_ONLY_ARB, input->BufferObj); @@ -596,6 +603,7 @@ static void r700AlignDataToDword(GLcontext *ctx, } } + radeon_bo_unmap(attr->bo); if (mapped_named_bo) { ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER, input->BufferObj); @@ -664,14 +672,18 @@ static void r700SetupStreams(GLcontext *ctx, const struct gl_client_array *input radeonAllocDmaRegion(&context->radeon, &context->stream_desc[index].bo, &context->stream_desc[index].bo_offset, size, 32); + + radeon_bo_map(context->stream_desc[index].bo, 1); assert(context->stream_desc[index].bo->ptr != NULL); + + dst = (uint32_t *)ADD_POINTERS(context->stream_desc[index].bo->ptr, context->stream_desc[index].bo_offset); switch (context->stream_desc[index].dwords) { case 1: - radeonEmitVec4(dst, input[i]->Ptr, input[i]->StrideB, local_count); + radeonEmitVec4(dst, input[i]->Ptr, input[i]->StrideB, local_count); break; case 2: radeonEmitVec8(dst, input[i]->Ptr, input[i]->StrideB, local_count); @@ -686,6 +698,7 @@ static void r700SetupStreams(GLcontext *ctx, const struct gl_client_array *input assert(0); break; } + radeon_bo_unmap(context->stream_desc[index].bo); } } @@ -757,6 +770,7 @@ static void r700FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer radeonAllocDmaRegion(&context->radeon, &context->ind_buf.bo, &context->ind_buf.bo_offset, size, 4); + radeon_bo_map(context->ind_buf.bo, 1); assert(context->ind_buf.bo->ptr != NULL); out = (GLuint *)ADD_POINTERS(context->ind_buf.bo->ptr, context->ind_buf.bo_offset); @@ -770,6 +784,7 @@ static void r700FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer *out++ = in[i]; } + radeon_bo_unmap(context->ind_buf.bo); #if MESA_BIG_ENDIAN } else @@ -780,6 +795,7 @@ static void r700FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer radeonAllocDmaRegion(&context->radeon, &context->ind_buf.bo, &context->ind_buf.bo_offset, size, 4); + radeon_bo_map(context->ind_buf.bo, 1); assert(context->ind_buf.bo->ptr != NULL); out = (GLuint *)ADD_POINTERS(context->ind_buf.bo->ptr, context->ind_buf.bo_offset); @@ -792,6 +808,7 @@ static void r700FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer { *out++ = in[i]; } + radeon_bo_unmap(context->ind_buf.bo); #endif } @@ -837,11 +854,13 @@ static void r700SetupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer radeonAllocDmaRegion(&context->radeon, &context->ind_buf.bo, &context->ind_buf.bo_offset, size, 4); + radeon_bo_map(context->ind_buf.bo, 1); assert(context->ind_buf.bo->ptr != NULL); dst_ptr = ADD_POINTERS(context->ind_buf.bo->ptr, context->ind_buf.bo_offset); _mesa_memcpy(dst_ptr, src_ptr, size); + radeon_bo_unmap(context->ind_buf.bo); context->ind_buf.is_32bit = (mesa_ind_buf->type == GL_UNSIGNED_INT); context->ind_buf.count = mesa_ind_buf->count; diff --git a/src/mesa/drivers/dri/r600/r700_shader.c b/src/mesa/drivers/dri/r600/r700_shader.c index 955ea4e4e1d..2eed1acc2f5 100644 --- a/src/mesa/drivers/dri/r600/r700_shader.c +++ b/src/mesa/drivers/dri/r600/r700_shader.c @@ -159,13 +159,18 @@ void Init_R700_Shader(R700_Shader * pShader) pShader->lstVTXInstructions.uNumOfNode=0; } +void SetActiveCFlist(R700_Shader *pShader, TypedShaderList * plstCF) +{ + pShader->plstCFInstructions_active = plstCF; +} + void AddCFInstruction(R700_Shader *pShader, R700ControlFlowInstruction *pCFInst) { R700ControlFlowSXClause* pSXClause; R700ControlFlowSMXClause* pSMXClause; - pCFInst->m_uIndex = pShader->lstCFInstructions.uNumOfNode; - AddInstToList(&(pShader->lstCFInstructions), + pCFInst->m_uIndex = pShader->plstCFInstructions_active->uNumOfNode; + AddInstToList(pShader->plstCFInstructions_active, (R700ShaderInstruction*)pCFInst); pShader->uShaderBinaryDWORDSize += GetInstructionSize(pCFInst->m_ShaderInstType); diff --git a/src/mesa/drivers/dri/r600/r700_shader.h b/src/mesa/drivers/dri/r600/r700_shader.h index c6a058617ec..0599ffd901f 100644 --- a/src/mesa/drivers/dri/r600/r700_shader.h +++ b/src/mesa/drivers/dri/r600/r700_shader.h @@ -109,6 +109,7 @@ typedef struct R700_Shader GLuint uStackSize; GLuint uMaxCallDepth; + TypedShaderList * plstCFInstructions_active; TypedShaderList lstCFInstructions; TypedShaderList lstALUInstructions; TypedShaderList lstTEXInstructions; @@ -132,13 +133,13 @@ void TakeInstOutFromList(TypedShaderList * plstCFInstructions, R700ShaderInstruc void ResolveLinks(R700_Shader *pShader); void Assemble(R700_Shader *pShader); - //Interface void Init_R700_Shader(R700_Shader * pShader); void AddCFInstruction(R700_Shader *pShader, R700ControlFlowInstruction *pCFInst); void AddVTXInstruction(R700_Shader *pShader, R700VertexInstruction *pVTXInst); void AddTEXInstruction(R700_Shader *pShader, R700TextureInstruction *pTEXInst); void AddALUInstruction(R700_Shader *pShader, R700ALUInstruction *pALUInst); +void SetActiveCFlist(R700_Shader *pShader, TypedShaderList * plstCF); void LoadProgram(R700_Shader *pShader); void UpdateShaderRegisters(R700_Shader *pShader); diff --git a/src/mesa/drivers/dri/r600/r700_shaderinst.h b/src/mesa/drivers/dri/r600/r700_shaderinst.h index 2829cca0a3c..cdb9a570f7c 100644 --- a/src/mesa/drivers/dri/r600/r700_shaderinst.h +++ b/src/mesa/drivers/dri/r600/r700_shaderinst.h @@ -42,6 +42,13 @@ #define SQ_FETCH_RESOURCE_VS_OFFSET 0x000000a0 #define SQ_FETCH_RESOURCE_VS_COUNT 0x000000b0 +//richard dec.10 glsl +#define SQ_TEX_SAMPLER_PS_OFFSET 0x00000000 +#define SQ_TEX_SAMPLER_PS_COUNT 0x00000012 +#define SQ_TEX_SAMPLER_VS_OFFSET 0x00000012 +#define SQ_TEX_SAMPLER_VS_COUNT 0x00000012 +//------------------- + #define SHADERINST_TYPEMASK_CF 0x10 #define SHADERINST_TYPEMASK_ALU 0x20 #define SHADERINST_TYPEMASK_TEX 0x40 diff --git a/src/mesa/drivers/dri/r600/r700_vertprog.c b/src/mesa/drivers/dri/r600/r700_vertprog.c index ffc6068bd88..90fac078ff0 100644 --- a/src/mesa/drivers/dri/r600/r700_vertprog.c +++ b/src/mesa/drivers/dri/r600/r700_vertprog.c @@ -111,6 +111,15 @@ unsigned int Map_Vertex_Output(r700_AssemblerBase *pAsm, } } + for(i=VERT_RESULT_VAR0; i<VERT_RESULT_MAX; i++) + { + unBit = 1 << i; + if(mesa_vp->Base.OutputsWritten & unBit) + { + pAsm->ucVP_OutputMap[i] = unTotal++; + } + } + return (unTotal - unStart); } @@ -235,6 +244,8 @@ void Map_Vertex_Program(GLcontext *ctx, pAsm->number_used_registers += mesa_vp->Base.NumTemporaries; } + pAsm->flag_reg_index = pAsm->number_used_registers++; + pAsm->uFirstHelpReg = pAsm->number_used_registers; } @@ -324,7 +335,18 @@ struct r700_vertex_program* r700TranslateVertexShader(GLcontext *ctx, return NULL; } - if(GL_FALSE == AssembleInstr(vp->mesa_program->Base.NumInstructions, + InitShaderProgram(&(vp->r700AsmCode)); + + for(i=0; i < MAX_SAMPLERS; i++) + { + vp->r700AsmCode.SamplerUnits[i] = vp->mesa_program->Base.SamplerUnits[i]; + } + + vp->r700AsmCode.unCurNumILInsts = vp->mesa_program->Base.NumInstructions; + + if(GL_FALSE == AssembleInstr(0, + 0, + vp->mesa_program->Base.NumInstructions, &(vp->mesa_program->Base.Instructions[0]), &(vp->r700AsmCode)) ) { @@ -336,6 +358,11 @@ struct r700_vertex_program* r700TranslateVertexShader(GLcontext *ctx, return NULL; } + if( GL_FALSE == RelocProgram(&(vp->r700AsmCode), &(vp->mesa_program->Base)) ) + { + return GL_FALSE; + } + vp->r700Shader.nRegs = (vp->r700AsmCode.number_used_registers == 0) ? 0 : (vp->r700AsmCode.number_used_registers - 1); @@ -612,6 +639,12 @@ GLboolean r700SetupVertexProgram(GLcontext * ctx) paramList = vp->mesa_program->Base.Parameters; if(NULL != paramList) { + /* vp->mesa_program was cloned, not updated by glsl shader api. */ + /* _mesa_reference_program has already checked glsl shProg is ok and set ctx->VertexProgem._Current */ + /* so, use ctx->VertexProgem._Current */ + struct gl_program_parameter_list *paramListOrginal = + paramListOrginal = ctx->VertexProgram._Current->Base.Parameters; + _mesa_load_state_parameters(ctx, paramList); if (paramList->NumParameters > R700_MAX_DX9_CONSTS) @@ -624,13 +657,42 @@ GLboolean r700SetupVertexProgram(GLcontext * ctx) unNumParamData = paramList->NumParameters; for(ui=0; ui<unNumParamData; ui++) { - r700->vs.consts[ui][0].f32All = paramList->ParameterValues[ui][0]; - r700->vs.consts[ui][1].f32All = paramList->ParameterValues[ui][1]; - r700->vs.consts[ui][2].f32All = paramList->ParameterValues[ui][2]; - r700->vs.consts[ui][3].f32All = paramList->ParameterValues[ui][3]; + if(paramList->Parameters[ui].Type == PROGRAM_UNIFORM) + { + r700->vs.consts[ui][0].f32All = paramListOrginal->ParameterValues[ui][0]; + r700->vs.consts[ui][1].f32All = paramListOrginal->ParameterValues[ui][1]; + r700->vs.consts[ui][2].f32All = paramListOrginal->ParameterValues[ui][2]; + r700->vs.consts[ui][3].f32All = paramListOrginal->ParameterValues[ui][3]; + } + else + { + r700->vs.consts[ui][0].f32All = paramList->ParameterValues[ui][0]; + r700->vs.consts[ui][1].f32All = paramList->ParameterValues[ui][1]; + r700->vs.consts[ui][2].f32All = paramList->ParameterValues[ui][2]; + r700->vs.consts[ui][3].f32All = paramList->ParameterValues[ui][3]; + } } } else r700->vs.num_consts = 0; + COMPILED_SUB * pCompiledSub; + GLuint uj; + GLuint unConstOffset = r700->vs.num_consts; + for(ui=0; ui<vp->r700AsmCode.unNumPresub; ui++) + { + pCompiledSub = vp->r700AsmCode.presubs[ui].pCompiledSub; + + r700->vs.num_consts += pCompiledSub->NumParameters; + + for(uj=0; uj<pCompiledSub->NumParameters; uj++) + { + r700->vs.consts[uj + unConstOffset][0].f32All = pCompiledSub->ParameterValues[uj][0]; + r700->vs.consts[uj + unConstOffset][1].f32All = pCompiledSub->ParameterValues[uj][1]; + r700->vs.consts[uj + unConstOffset][2].f32All = pCompiledSub->ParameterValues[uj][2]; + r700->vs.consts[uj + unConstOffset][3].f32All = pCompiledSub->ParameterValues[uj][3]; + } + unConstOffset += pCompiledSub->NumParameters; + } + return GL_TRUE; } diff --git a/src/mesa/drivers/dri/radeon/radeon_common.c b/src/mesa/drivers/dri/radeon/radeon_common.c index 9b64c216859..2a2b16a54bd 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common.c +++ b/src/mesa/drivers/dri/radeon/radeon_common.c @@ -641,6 +641,27 @@ void radeonCopySubBuffer(__DRIdrawablePrivate * dPriv, } } +/** + * Check if we're about to draw into the front color buffer. + * If so, set the intel->front_buffer_dirty field to true. + */ +void +radeon_check_front_buffer_rendering(GLcontext *ctx) +{ + radeonContextPtr radeon = RADEON_CONTEXT(ctx); + const struct gl_framebuffer *fb = ctx->DrawBuffer; + + if (fb->Name == 0) { + /* drawing to window system buffer */ + if (fb->_NumColorDrawBuffers > 0) { + if (fb->_ColorDrawBufferIndexes[0] == BUFFER_FRONT_LEFT) { + radeon->front_buffer_dirty = GL_TRUE; + } + } + } +} + + void radeon_draw_buffer(GLcontext *ctx, struct gl_framebuffer *fb) { radeonContextPtr radeon = RADEON_CONTEXT(ctx); @@ -817,7 +838,7 @@ void radeonDrawBuffer( GLcontext *ctx, GLenum mode ) */ if (!was_front_buffer_rendering && radeon->is_front_buffer_rendering) { radeon_update_renderbuffers(radeon->dri.context, - radeon->dri.context->driDrawablePriv); + radeon->dri.context->driDrawablePriv, GL_FALSE); } } @@ -834,7 +855,7 @@ void radeonReadBuffer( GLcontext *ctx, GLenum mode ) if (!was_front_buffer_reading && rmesa->is_front_buffer_reading) { radeon_update_renderbuffers(rmesa->dri.context, - rmesa->dri.context->driReadablePriv); + rmesa->dri.context->driReadablePriv, GL_FALSE); } } /* nothing, until we implement h/w glRead/CopyPixels or CopyTexImage */ @@ -885,9 +906,9 @@ void radeon_viewport(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei he if (radeon->is_front_buffer_rendering) { ctx->Driver.Flush(ctx); } - radeon_update_renderbuffers(driContext, driContext->driDrawablePriv); + radeon_update_renderbuffers(driContext, driContext->driDrawablePriv, GL_FALSE); if (driContext->driDrawablePriv != driContext->driReadablePriv) - radeon_update_renderbuffers(driContext, driContext->driReadablePriv); + radeon_update_renderbuffers(driContext, driContext->driReadablePriv, GL_FALSE); } old_viewport = ctx->Driver.Viewport; @@ -1095,7 +1116,7 @@ void radeonFlush(GLcontext *ctx) then no point flushing anything at all. */ if (!radeon->dma.flush && !radeon->cmdbuf.cs->cdw && is_empty_list(&radeon->dma.reserved)) - return; + goto flush_front; if (radeon->dma.flush) radeon->dma.flush( ctx ); @@ -1103,6 +1124,7 @@ void radeonFlush(GLcontext *ctx) if (radeon->cmdbuf.cs->cdw) rcommonFlushCmdBuf(radeon, __FUNCTION__); +flush_front: if ((ctx->DrawBuffer->Name == 0) && radeon->front_buffer_dirty) { __DRIscreen *const screen = radeon->radeonScreen->driScreen; diff --git a/src/mesa/drivers/dri/radeon/radeon_common.h b/src/mesa/drivers/dri/radeon/radeon_common.h index 0608fe2418c..faad145cc49 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common.h +++ b/src/mesa/drivers/dri/radeon/radeon_common.h @@ -43,6 +43,8 @@ radeon_renderbuffer_set_bo(struct radeon_renderbuffer *rb, struct radeon_bo *bo); struct radeon_renderbuffer * radeon_create_renderbuffer(gl_format format, __DRIdrawablePrivate *driDrawPriv); + +void radeon_check_front_buffer_rendering(GLcontext *ctx); static inline struct radeon_renderbuffer *radeon_renderbuffer(struct gl_renderbuffer *rb) { struct radeon_renderbuffer *rrb = (struct radeon_renderbuffer *)rb; diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.c b/src/mesa/drivers/dri/radeon/radeon_common_context.c index 71f70d724b9..5c68bf5df6c 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common_context.c +++ b/src/mesa/drivers/dri/radeon/radeon_common_context.c @@ -499,7 +499,8 @@ radeon_bits_per_pixel(const struct radeon_renderbuffer *rb) } void -radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable) +radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable, + GLboolean front_only) { unsigned int attachments[10]; __DRIbuffer *buffers = NULL; @@ -525,7 +526,7 @@ radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable) struct radeon_renderbuffer *stencil_rb; i = 0; - if ((radeon->is_front_buffer_rendering || + if ((front_only || radeon->is_front_buffer_rendering || radeon->is_front_buffer_reading || !draw->color_rb[1]) && draw->color_rb[0]) { @@ -533,23 +534,25 @@ radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable) attachments[i++] = radeon_bits_per_pixel(draw->color_rb[0]); } - if (draw->color_rb[1]) { - attachments[i++] = __DRI_BUFFER_BACK_LEFT; - attachments[i++] = radeon_bits_per_pixel(draw->color_rb[1]); - } + if (!front_only) { + if (draw->color_rb[1]) { + attachments[i++] = __DRI_BUFFER_BACK_LEFT; + attachments[i++] = radeon_bits_per_pixel(draw->color_rb[1]); + } - depth_rb = radeon_get_renderbuffer(&draw->base, BUFFER_DEPTH); - stencil_rb = radeon_get_renderbuffer(&draw->base, BUFFER_STENCIL); - - if ((depth_rb != NULL) && (stencil_rb != NULL)) { - attachments[i++] = __DRI_BUFFER_DEPTH_STENCIL; - attachments[i++] = radeon_bits_per_pixel(depth_rb); - } else if (depth_rb != NULL) { - attachments[i++] = __DRI_BUFFER_DEPTH; - attachments[i++] = radeon_bits_per_pixel(depth_rb); - } else if (stencil_rb != NULL) { - attachments[i++] = __DRI_BUFFER_STENCIL; - attachments[i++] = radeon_bits_per_pixel(stencil_rb); + depth_rb = radeon_get_renderbuffer(&draw->base, BUFFER_DEPTH); + stencil_rb = radeon_get_renderbuffer(&draw->base, BUFFER_STENCIL); + + if ((depth_rb != NULL) && (stencil_rb != NULL)) { + attachments[i++] = __DRI_BUFFER_DEPTH_STENCIL; + attachments[i++] = radeon_bits_per_pixel(depth_rb); + } else if (depth_rb != NULL) { + attachments[i++] = __DRI_BUFFER_DEPTH; + attachments[i++] = radeon_bits_per_pixel(depth_rb); + } else if (stencil_rb != NULL) { + attachments[i++] = __DRI_BUFFER_STENCIL; + attachments[i++] = radeon_bits_per_pixel(stencil_rb); + } } buffers = (*screen->dri2.loader->getBuffersWithFormat)(drawable, @@ -562,12 +565,14 @@ radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable) i = 0; if (draw->color_rb[0]) attachments[i++] = __DRI_BUFFER_FRONT_LEFT; - if (draw->color_rb[1]) - attachments[i++] = __DRI_BUFFER_BACK_LEFT; - if (radeon_get_renderbuffer(&draw->base, BUFFER_DEPTH)) - attachments[i++] = __DRI_BUFFER_DEPTH; - if (radeon_get_renderbuffer(&draw->base, BUFFER_STENCIL)) - attachments[i++] = __DRI_BUFFER_STENCIL; + if (!front_only) { + if (draw->color_rb[1]) + attachments[i++] = __DRI_BUFFER_BACK_LEFT; + if (radeon_get_renderbuffer(&draw->base, BUFFER_DEPTH)) + attachments[i++] = __DRI_BUFFER_DEPTH; + if (radeon_get_renderbuffer(&draw->base, BUFFER_STENCIL)) + attachments[i++] = __DRI_BUFFER_STENCIL; + } buffers = (*screen->dri2.loader->getBuffers)(drawable, &drawable->w, @@ -735,9 +740,9 @@ GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv, readfb = driReadPriv->driverPrivate; if (driContextPriv->driScreenPriv->dri2.enabled) { - radeon_update_renderbuffers(driContextPriv, driDrawPriv); + radeon_update_renderbuffers(driContextPriv, driDrawPriv, GL_FALSE); if (driDrawPriv != driReadPriv) - radeon_update_renderbuffers(driContextPriv, driReadPriv); + radeon_update_renderbuffers(driContextPriv, driReadPriv, GL_FALSE); _mesa_reference_renderbuffer(&radeon->state.color.rb, &(radeon_get_renderbuffer(&drfb->base, BUFFER_BACK_LEFT)->base)); _mesa_reference_renderbuffer(&radeon->state.depth.rb, diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.h b/src/mesa/drivers/dri/radeon/radeon_common_context.h index 62987489428..0739496e032 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common_context.h +++ b/src/mesa/drivers/dri/radeon/radeon_common_context.h @@ -328,6 +328,7 @@ struct radeon_swtcl_info { GLuint vertex_attr_count; GLuint emit_prediction; + struct radeon_bo *bo; }; #define RADEON_MAX_AOS_ARRAYS 16 @@ -585,7 +586,8 @@ GLboolean radeonInitContext(radeonContextPtr radeon, void radeonCleanupContext(radeonContextPtr radeon); GLboolean radeonUnbindContext(__DRIcontextPrivate * driContextPriv); -void radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable); +void radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable, + GLboolean front_only); GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv, __DRIdrawablePrivate * driDrawPriv, __DRIdrawablePrivate * driReadPriv); diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.c b/src/mesa/drivers/dri/radeon/radeon_dma.c index 232972d7ec7..d31e4e47ddb 100644 --- a/src/mesa/drivers/dri/radeon/radeon_dma.c +++ b/src/mesa/drivers/dri/radeon/radeon_dma.c @@ -151,6 +151,7 @@ void rcommon_emit_vector(GLcontext * ctx, struct radeon_aos *aos, aos->components = size; aos->count = count; + radeon_bo_map(aos->bo, 1); out = (uint32_t*)((char*)aos->bo->ptr + aos->offset); switch (size) { case 1: radeonEmitVec4(out, data, stride, count); break; @@ -161,6 +162,7 @@ void rcommon_emit_vector(GLcontext * ctx, struct radeon_aos *aos, assert(0); break; } + radeon_bo_unmap(aos->bo); } void radeon_init_dma(radeonContextPtr rmesa) @@ -183,10 +185,6 @@ void radeonRefillCurrentDmaRegion(radeonContextPtr rmesa, int size) __FUNCTION__, size, rmesa->dma.minimum_size); - /* unmap old reserved bo */ - if (!is_empty_list(&rmesa->dma.reserved)) - radeon_bo_unmap(first_elem(&rmesa->dma.reserved)->bo); - if (is_empty_list(&rmesa->dma.free) || last_elem(&rmesa->dma.free)->bo->size < size) { dma_bo = CALLOC_STRUCT(radeon_dma_bo); @@ -223,8 +221,6 @@ again_alloc: /* Cmd buff have been flushed in radeon_revalidate_bos */ goto again_alloc; } - - radeon_bo_map(first_elem(&rmesa->dma.reserved)->bo, 1); } /* Allocates a region from rmesa->dma.current. If there isn't enough @@ -281,7 +277,6 @@ void radeonFreeDmaRegions(radeonContextPtr rmesa) foreach_s(dma_bo, temp, &rmesa->dma.reserved) { remove_from_list(dma_bo); - radeon_bo_unmap(dma_bo->bo); radeon_bo_unref(dma_bo->bo); FREE(dma_bo); } @@ -361,9 +356,6 @@ void radeonReleaseDmaRegions(radeonContextPtr rmesa) insert_at_tail(&rmesa->dma.free, dma_bo); } - /* unmap the last dma region */ - if (!is_empty_list(&rmesa->dma.reserved)) - radeon_bo_unmap(first_elem(&rmesa->dma.reserved)->bo); /* move reserved to wait list */ foreach_s(dma_bo, temp, &rmesa->dma.reserved) { /* free objects that are too small to be used because of large request */ @@ -397,11 +389,12 @@ void rcommon_flush_last_swtcl_prim( GLcontext *ctx ) radeonContextPtr rmesa = RADEON_CONTEXT(ctx); struct radeon_dma *dma = &rmesa->dma; - if (RADEON_DEBUG & RADEON_IOCTL) fprintf(stderr, "%s\n", __FUNCTION__); dma->flush = NULL; + radeon_bo_unmap(rmesa->swtcl.bo); + if (!is_empty_list(&dma->reserved)) { GLuint current_offset = dma->current_used; @@ -416,6 +409,8 @@ void rcommon_flush_last_swtcl_prim( GLcontext *ctx ) } rmesa->swtcl.numverts = 0; } + radeon_bo_unref(rmesa->swtcl.bo); + rmesa->swtcl.bo = NULL; } /* Alloc space in the current dma region. */ @@ -426,6 +421,7 @@ rcommonAllocDmaLowVerts( radeonContextPtr rmesa, int nverts, int vsize ) void *head; if (RADEON_DEBUG & RADEON_IOCTL) fprintf(stderr, "%s\n", __FUNCTION__); + if(is_empty_list(&rmesa->dma.reserved) ||rmesa->dma.current_vertexptr + bytes > first_elem(&rmesa->dma.reserved)->bo->size) { if (rmesa->dma.flush) { @@ -449,7 +445,13 @@ rcommonAllocDmaLowVerts( radeonContextPtr rmesa, int nverts, int vsize ) rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 == rmesa->dma.current_vertexptr ); - head = (first_elem(&rmesa->dma.reserved)->bo->ptr + rmesa->dma.current_vertexptr); + if (!rmesa->swtcl.bo) { + rmesa->swtcl.bo = first_elem(&rmesa->dma.reserved)->bo; + radeon_bo_ref(rmesa->swtcl.bo); + radeon_bo_map(rmesa->swtcl.bo, 1); + } + + head = (rmesa->swtcl.bo->ptr + rmesa->dma.current_vertexptr); rmesa->dma.current_vertexptr += bytes; rmesa->swtcl.numverts += nverts; return head; diff --git a/src/mesa/drivers/dri/radeon/radeon_fbo.c b/src/mesa/drivers/dri/radeon/radeon_fbo.c index fc21069a92c..a536436d55f 100644 --- a/src/mesa/drivers/dri/radeon/radeon_fbo.c +++ b/src/mesa/drivers/dri/radeon/radeon_fbo.c @@ -166,8 +166,9 @@ radeon_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, uint32_t size; uint32_t pitch = ((cpp * width + 63) & ~63) / cpp; - fprintf(stderr,"Allocating %d x %d radeon RBO (pitch %d)\n", width, - height, pitch); + if (RADEON_DEBUG & RADEON_MEMORY) + fprintf(stderr,"Allocating %d x %d radeon RBO (pitch %d)\n", width, + height, pitch); size = pitch * height * cpp; rrb->pitch = pitch * cpp; diff --git a/src/mesa/drivers/dri/radeon/radeon_ioctl.c b/src/mesa/drivers/dri/radeon/radeon_ioctl.c index a0106d00fa2..13fd6f99719 100644 --- a/src/mesa/drivers/dri/radeon/radeon_ioctl.c +++ b/src/mesa/drivers/dri/radeon/radeon_ioctl.c @@ -575,6 +575,10 @@ static void radeonClear( GLcontext *ctx, GLbitfield mask ) GLuint color_mask = 0; GLuint orig_mask = mask; + if (mask & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_FRONT_RIGHT)) { + rmesa->radeon.front_buffer_dirty = GL_TRUE; + } + if ( RADEON_DEBUG & RADEON_IOCTL ) { fprintf( stderr, "radeonClear\n"); } diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c b/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c index 08e1c5d00df..d810e6080eb 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c +++ b/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c @@ -76,12 +76,14 @@ static void emit_vecfog(GLcontext *ctx, struct radeon_aos *aos, /* Emit the data */ + radeon_bo_map(aos->bo, 1); out = (uint32_t*)((char*)aos->bo->ptr + aos->offset); for (i = 0; i < count; i++) { out[0] = radeonComputeFogBlendFactor( ctx, *(GLfloat *)data ); out++; data += stride; } + radeon_bo_unmap(aos->bo); } static void emit_s0_vec(uint32_t *out, GLvoid *data, int stride, int count) @@ -151,6 +153,7 @@ static void emit_tex_vector(GLcontext *ctx, struct radeon_aos *aos, /* Emit the data */ + radeon_bo_map(aos->bo, 1); out = (uint32_t*)((char*)aos->bo->ptr + aos->offset); switch (size) { case 1: @@ -170,6 +173,7 @@ static void emit_tex_vector(GLcontext *ctx, struct radeon_aos *aos, exit(1); break; } + radeon_bo_unmap(aos->bo); } @@ -196,12 +200,12 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (!rmesa->tcl.obj.buf) rcommon_emit_vector( ctx, &(rmesa->tcl.aos[nr]), - (char *)VB->ObjPtr->data, - VB->ObjPtr->size, - VB->ObjPtr->stride, + (char *)VB->AttribPtr[_TNL_ATTRIB_POS]->data, + VB->AttribPtr[_TNL_ATTRIB_POS]->size, + VB->AttribPtr[_TNL_ATTRIB_POS]->stride, count); - switch( VB->ObjPtr->size ) { + switch( VB->AttribPtr[_TNL_ATTRIB_POS]->size ) { case 4: vfmt |= RADEON_CP_VC_FRMT_W0; case 3: vfmt |= RADEON_CP_VC_FRMT_Z; case 2: vfmt |= RADEON_CP_VC_FRMT_XY; @@ -216,9 +220,9 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (!rmesa->tcl.norm.buf) rcommon_emit_vector( ctx, &(rmesa->tcl.aos[nr]), - (char *)VB->NormalPtr->data, + (char *)VB->AttribPtr[_TNL_ATTRIB_NORMAL]->data, 3, - VB->NormalPtr->stride, + VB->AttribPtr[_TNL_ATTRIB_NORMAL]->stride, count); vfmt |= RADEON_CP_VC_FRMT_N0; @@ -227,9 +231,9 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (inputs & VERT_BIT_COLOR0) { int emitsize; - if (VB->ColorPtr[0]->size == 4 && - (VB->ColorPtr[0]->stride != 0 || - VB->ColorPtr[0]->data[0][3] != 1.0)) { + if (VB->AttribPtr[_TNL_ATTRIB_COLOR0]->size == 4 && + (VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride != 0 || + VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data[0][3] != 1.0)) { vfmt |= RADEON_CP_VC_FRMT_FPCOLOR | RADEON_CP_VC_FRMT_FPALPHA; emitsize = 4; } @@ -242,9 +246,9 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (!rmesa->tcl.rgba.buf) rcommon_emit_vector( ctx, &(rmesa->tcl.aos[nr]), - (char *)VB->ColorPtr[0]->data, + (char *)VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data, emitsize, - VB->ColorPtr[0]->stride, + VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride, count); nr++; @@ -256,9 +260,9 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) rcommon_emit_vector( ctx, &(rmesa->tcl.aos[nr]), - (char *)VB->SecondaryColorPtr[0]->data, + (char *)VB->AttribPtr[_TNL_ATTRIB_COLOR1]->data, 3, - VB->SecondaryColorPtr[0]->stride, + VB->AttribPtr[_TNL_ATTRIB_COLOR1]->stride, count); } @@ -273,8 +277,8 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (!rmesa->tcl.fog.buf) emit_vecfog( ctx, &(rmesa->tcl.aos[nr]), - (char *)VB->FogCoordPtr->data, - VB->FogCoordPtr->stride, + (char *)VB->AttribPtr[_TNL_ATTRIB_FOG]->data, + VB->AttribPtr[_TNL_ATTRIB_FOG]->stride, count); vfmt |= RADEON_CP_VC_FRMT_FPFOG; @@ -290,24 +294,24 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (!rmesa->tcl.tex[unit].buf) emit_tex_vector( ctx, &(rmesa->tcl.aos[nr]), - (char *)VB->TexCoordPtr[unit]->data, - VB->TexCoordPtr[unit]->size, - VB->TexCoordPtr[unit]->stride, + (char *)VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->data, + VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size, + VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->stride, count ); nr++; vfmt |= RADEON_ST_BIT(unit); /* assume we need the 3rd coord if texgen is active for r/q OR at least 3 coords are submitted. This may not be 100% correct */ - if (VB->TexCoordPtr[unit]->size >= 3) { + if (VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size >= 3) { vtx |= RADEON_Q_BIT(unit); vfmt |= RADEON_Q_BIT(unit); } if ( (ctx->Texture.Unit[unit].TexGenEnabled & (R_BIT | Q_BIT)) ) vtx |= RADEON_Q_BIT(unit); - else if ((VB->TexCoordPtr[unit]->size >= 3) && + else if ((VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size >= 3) && ((ctx->Texture.Unit[unit]._ReallyEnabled & (TEXTURE_CUBE_BIT)) == 0)) { - GLuint swaptexmatcol = (VB->TexCoordPtr[unit]->size - 3); + GLuint swaptexmatcol = (VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size - 3); if (((rmesa->NeedTexMatrix >> unit) & 1) && (swaptexmatcol != ((rmesa->TexMatColSwap >> unit) & 1))) radeonUploadTexMatrix( rmesa, unit, swaptexmatcol ) ; diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h b/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h index 515783135d6..d764ccb9826 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h +++ b/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h @@ -56,18 +56,18 @@ static void TAG(emit)( GLcontext *ctx, radeon_print(RADEON_SWRENDER, RADEON_VERBOSE, "%s\n", __FUNCTION__); - coord = (GLuint (*)[4])VB->ObjPtr->data; - coord_stride = VB->ObjPtr->stride; + coord = (GLuint (*)[4])VB->AttribPtr[_TNL_ATTRIB_POS]->data; + coord_stride = VB->AttribPtr[_TNL_ATTRIB_POS]->stride; if (DO_TEX2) { - if (VB->TexCoordPtr[2]) { + if (VB->AttribPtr[_TNL_ATTRIB_TEX2]) { const GLuint t2 = GET_TEXSOURCE(2); - tc2 = (GLuint (*)[4])VB->TexCoordPtr[t2]->data; - tc2_stride = VB->TexCoordPtr[t2]->stride; - if (DO_PTEX && VB->TexCoordPtr[t2]->size < 3) { + tc2 = (GLuint (*)[4])VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->data; + tc2_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->stride; + if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->size < 3) { fill_tex |= (1<<2); } - else if (DO_PTEX && VB->TexCoordPtr[t2]->size < 4) { + else if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->size < 4) { rqcoordsnoswap |= (1<<2); } } else { @@ -77,14 +77,14 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_TEX1) { - if (VB->TexCoordPtr[1]) { + if (VB->AttribPtr[_TNL_ATTRIB_TEX1]) { const GLuint t1 = GET_TEXSOURCE(1); - tc1 = (GLuint (*)[4])VB->TexCoordPtr[t1]->data; - tc1_stride = VB->TexCoordPtr[t1]->stride; - if (DO_PTEX && VB->TexCoordPtr[t1]->size < 3) { + tc1 = (GLuint (*)[4])VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->data; + tc1_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->stride; + if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->size < 3) { fill_tex |= (1<<1); } - else if (DO_PTEX && VB->TexCoordPtr[t1]->size < 4) { + else if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->size < 4) { rqcoordsnoswap |= (1<<1); } } else { @@ -94,14 +94,14 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_TEX0) { - if (VB->TexCoordPtr[0]) { + if (VB->AttribPtr[_TNL_ATTRIB_TEX0]) { const GLuint t0 = GET_TEXSOURCE(0); - tc0_stride = VB->TexCoordPtr[t0]->stride; - tc0 = (GLuint (*)[4])VB->TexCoordPtr[t0]->data; - if (DO_PTEX && VB->TexCoordPtr[t0]->size < 3) { + tc0_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->stride; + tc0 = (GLuint (*)[4])VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->data; + if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->size < 3) { fill_tex |= (1<<0); } - else if (DO_PTEX && VB->TexCoordPtr[t0]->size < 4) { + else if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->size < 4) { rqcoordsnoswap |= (1<<0); } } else { @@ -112,9 +112,9 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_NORM) { - if (VB->NormalPtr) { - norm_stride = VB->NormalPtr->stride; - norm = (GLuint (*)[4])VB->NormalPtr->data; + if (VB->AttribPtr[_TNL_ATTRIB_NORMAL]) { + norm_stride = VB->AttribPtr[_TNL_ATTRIB_NORMAL]->stride; + norm = (GLuint (*)[4])VB->AttribPtr[_TNL_ATTRIB_NORMAL]->data; } else { norm_stride = 0; norm = (GLuint (*)[4])&ctx->Current.Attrib[VERT_ATTRIB_NORMAL]; @@ -122,9 +122,9 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_RGBA) { - if (VB->ColorPtr[0]) { - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + if (VB->AttribPtr[_TNL_ATTRIB_COLOR0]) { + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; } else { col = (GLfloat (*)[4])ctx->Current.Attrib[VERT_ATTRIB_COLOR0]; col_stride = 0; @@ -132,9 +132,9 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_SPEC_OR_FOG) { - if (VB->SecondaryColorPtr[0]) { - spec = VB->SecondaryColorPtr[0]->data; - spec_stride = VB->SecondaryColorPtr[0]->stride; + if (VB->AttribPtr[_TNL_ATTRIB_COLOR1]) { + spec = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->data; + spec_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->stride; } else { spec = (GLfloat (*)[4])ctx->Current.Attrib[VERT_ATTRIB_COLOR1]; spec_stride = 0; @@ -142,9 +142,9 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_SPEC_OR_FOG) { - if (VB->FogCoordPtr) { - fog = VB->FogCoordPtr->data; - fog_stride = VB->FogCoordPtr->stride; + if (VB->AttribPtr[_TNL_ATTRIB_FOG]) { + fog = VB->AttribPtr[_TNL_ATTRIB_FOG]->data; + fog_stride = VB->AttribPtr[_TNL_ATTRIB_FOG]->stride; } else { fog = (GLfloat (*)[4])ctx->Current.Attrib[VERT_ATTRIB_FOG]; fog_stride = 0; diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c index 78ec1193026..98f96ff2a76 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c +++ b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c @@ -326,7 +326,7 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (1) { req |= RADEON_CP_VC_FRMT_Z; - if (VB->ObjPtr->size == 4) { + if (VB->AttribPtr[_TNL_ATTRIB_POS]->size == 4) { req |= RADEON_CP_VC_FRMT_W0; } } @@ -348,15 +348,15 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) req |= RADEON_ST_BIT(unit); /* assume we need the 3rd coord if texgen is active for r/q OR at least 3 coords are submitted. This may not be 100% correct */ - if (VB->TexCoordPtr[unit]->size >= 3) { + if (VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size >= 3) { req |= RADEON_Q_BIT(unit); vtx |= RADEON_Q_BIT(unit); } if ( (ctx->Texture.Unit[unit].TexGenEnabled & (R_BIT | Q_BIT)) ) vtx |= RADEON_Q_BIT(unit); - else if ((VB->TexCoordPtr[unit]->size >= 3) && + else if ((VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size >= 3) && ((ctx->Texture.Unit[unit]._ReallyEnabled & (TEXTURE_CUBE_BIT)) == 0)) { - GLuint swaptexmatcol = (VB->TexCoordPtr[unit]->size - 3); + GLuint swaptexmatcol = (VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size - 3); if (((rmesa->NeedTexMatrix >> unit) & 1) && (swaptexmatcol != ((rmesa->TexMatColSwap >> unit) & 1))) radeonUploadTexMatrix( rmesa, unit, swaptexmatcol ) ; @@ -390,19 +390,19 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) * this, add more vertex code (for obj-2, obj-3) or preferably move * to maos. */ - if (VB->ObjPtr->size < 3 || - (VB->ObjPtr->size == 3 && + if (VB->AttribPtr[_TNL_ATTRIB_POS]->size < 3 || + (VB->AttribPtr[_TNL_ATTRIB_POS]->size == 3 && (setup_tab[i].vertex_format & RADEON_CP_VC_FRMT_W0))) { _math_trans_4f( rmesa->tcl.ObjClean.data, - VB->ObjPtr->data, - VB->ObjPtr->stride, + VB->AttribPtr[_TNL_ATTRIB_POS]->data, + VB->AttribPtr[_TNL_ATTRIB_POS]->stride, GL_FLOAT, - VB->ObjPtr->size, + VB->AttribPtr[_TNL_ATTRIB_POS]->size, 0, VB->Count ); - switch (VB->ObjPtr->size) { + switch (VB->AttribPtr[_TNL_ATTRIB_POS]->size) { case 1: _mesa_vector4f_clean_elem(&rmesa->tcl.ObjClean, VB->Count, 1); case 2: @@ -416,14 +416,14 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) break; } - VB->ObjPtr = &rmesa->tcl.ObjClean; + VB->AttribPtr[_TNL_ATTRIB_POS] = &rmesa->tcl.ObjClean; } - + radeon_bo_map(rmesa->radeon.tcl.aos[0].bo, 1); setup_tab[i].emit( ctx, 0, VB->Count, rmesa->radeon.tcl.aos[0].bo->ptr + rmesa->radeon.tcl.aos[0].offset); - + radeon_bo_unmap(rmesa->radeon.tcl.aos[0].bo); // rmesa->radeon.tcl.aos[0].size = setup_tab[i].vertex_size; rmesa->radeon.tcl.aos[0].stride = setup_tab[i].vertex_size; rmesa->tcl.vertex_format = setup_tab[i].vertex_format; diff --git a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c index d7b5d71d21f..033f26db2a1 100644 --- a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c +++ b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c @@ -68,6 +68,19 @@ static unsigned get_compressed_image_size( return rowStride * ((height + blockHeight - 1) / blockHeight); } +static int find_next_power_of_two(GLuint value) +{ + int i, tmp; + + i = 0; + tmp = value - 1; + while (tmp) { + tmp >>= 1; + i++; + } + return (1 << i); +} + /** * Compute sizes and fill in offset and blit information for the given * image (determined by \p face and \p level). @@ -80,25 +93,28 @@ static void compute_tex_image_offset(radeonContextPtr rmesa, radeon_mipmap_tree { radeon_mipmap_level *lvl = &mt->levels[level]; uint32_t row_align; + GLuint height; + + height = find_next_power_of_two(lvl->height); /* Find image size in bytes */ if (_mesa_is_format_compressed(mt->mesaFormat)) { lvl->rowstride = get_aligned_compressed_row_stride(mt->mesaFormat, lvl->width, rmesa->texture_compressed_row_align); - lvl->size = get_compressed_image_size(mt->mesaFormat, lvl->rowstride, lvl->height); + lvl->size = get_compressed_image_size(mt->mesaFormat, lvl->rowstride, height); } else if (mt->target == GL_TEXTURE_RECTANGLE_NV) { row_align = rmesa->texture_rect_row_align - 1; lvl->rowstride = (_mesa_format_row_stride(mt->mesaFormat, lvl->width) + row_align) & ~row_align; - lvl->size = lvl->rowstride * lvl->height; + lvl->size = lvl->rowstride * height; } else if (mt->tilebits & RADEON_TXO_MICRO_TILE) { /* tile pattern is 16 bytes x2. mipmaps stay 32 byte aligned, * though the actual offset may be different (if texture is less than * 32 bytes width) to the untiled case */ lvl->rowstride = (_mesa_format_row_stride(mt->mesaFormat, lvl->width) * 2 + 31) & ~31; - lvl->size = lvl->rowstride * ((lvl->height + 1) / 2) * lvl->depth; + lvl->size = lvl->rowstride * ((height + 1) / 2) * lvl->depth; } else { row_align = rmesa->texture_row_align - 1; lvl->rowstride = (_mesa_format_row_stride(mt->mesaFormat, lvl->width) + row_align) & ~row_align; - lvl->size = lvl->rowstride * lvl->height * lvl->depth; + lvl->size = lvl->rowstride * height * lvl->depth; } assert(lvl->size > 0); @@ -110,7 +126,7 @@ static void compute_tex_image_offset(radeonContextPtr rmesa, radeon_mipmap_tree if (RADEON_DEBUG & RADEON_TEXTURE) fprintf(stderr, "level %d, face %d: rs:%d %dx%d at %d\n", - level, face, lvl->rowstride, lvl->width, lvl->height, lvl->faces[face].offset); + level, face, lvl->rowstride, lvl->width, height, lvl->faces[face].offset); } static GLuint minify(GLuint size, GLuint levels) diff --git a/src/mesa/drivers/dri/radeon/radeon_span.c b/src/mesa/drivers/dri/radeon/radeon_span.c index 665f2b6f516..8db3d2b143b 100644 --- a/src/mesa/drivers/dri/radeon/radeon_span.c +++ b/src/mesa/drivers/dri/radeon/radeon_span.c @@ -827,18 +827,21 @@ static void map_unmap_rb(struct gl_renderbuffer *rb, int flag) } static void -radeon_map_unmap_buffers(GLcontext *ctx, GLboolean map) +radeon_map_unmap_framebuffer(GLcontext *ctx, struct gl_framebuffer *fb, + GLboolean map) { GLuint i, j; /* color draw buffers */ for (j = 0; j < ctx->DrawBuffer->_NumColorDrawBuffers; j++) - map_unmap_rb(ctx->DrawBuffer->_ColorDrawBuffers[j], map); + map_unmap_rb(fb->_ColorDrawBuffers[j], map); + + map_unmap_rb(fb->_ColorReadBuffer, map); /* check for render to textures */ for (i = 0; i < BUFFER_COUNT; i++) { struct gl_renderbuffer_attachment *att = - ctx->DrawBuffer->Attachment + i; + fb->Attachment + i; struct gl_texture_object *tex = att->Texture; if (tex) { /* Render to texture. Note that a mipmapped texture need not @@ -854,15 +857,15 @@ radeon_map_unmap_buffers(GLcontext *ctx, GLboolean map) radeon_teximage_unmap(image); } } - - map_unmap_rb(ctx->ReadBuffer->_ColorReadBuffer, map); - + /* depth buffer (Note wrapper!) */ - if (ctx->DrawBuffer->_DepthBuffer) - map_unmap_rb(ctx->DrawBuffer->_DepthBuffer->Wrapped, map); + if (fb->_DepthBuffer) + map_unmap_rb(fb->_DepthBuffer->Wrapped, map); + + if (fb->_StencilBuffer) + map_unmap_rb(fb->_StencilBuffer->Wrapped, map); - if (ctx->DrawBuffer->_StencilBuffer) - map_unmap_rb(ctx->DrawBuffer->_StencilBuffer->Wrapped, map); + radeon_check_front_buffer_rendering(ctx); } static void radeonSpanRenderStart(GLcontext * ctx) @@ -887,23 +890,30 @@ static void radeonSpanRenderStart(GLcontext * ctx) ctx->Driver.MapTexture(ctx, ctx->Texture.Unit[i]._Current); } - radeon_map_unmap_buffers(ctx, 1); + radeon_map_unmap_framebuffer(ctx, ctx->DrawBuffer, GL_TRUE); + if (ctx->ReadBuffer != ctx->DrawBuffer) + radeon_map_unmap_framebuffer(ctx, ctx->ReadBuffer, GL_TRUE); } static void radeonSpanRenderFinish(GLcontext * ctx) { radeonContextPtr rmesa = RADEON_CONTEXT(ctx); int i; + _swrast_flush(ctx); - if (!rmesa->radeonScreen->driScreen->dri2.enabled) { - UNLOCK_HARDWARE(rmesa); - } + for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) { if (ctx->Texture.Unit[i]._ReallyEnabled) ctx->Driver.UnmapTexture(ctx, ctx->Texture.Unit[i]._Current); } - radeon_map_unmap_buffers(ctx, 0); + radeon_map_unmap_framebuffer(ctx, ctx->DrawBuffer, GL_FALSE); + if (ctx->ReadBuffer != ctx->DrawBuffer) + radeon_map_unmap_framebuffer(ctx, ctx->ReadBuffer, GL_FALSE); + + if (!rmesa->radeonScreen->driScreen->dri2.enabled) { + UNLOCK_HARDWARE(rmesa); + } } void radeonInitSpanFuncs(GLcontext * ctx) diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.c b/src/mesa/drivers/dri/radeon/radeon_swtcl.c index e61f59eaeaf..8bf1bfbc575 100644 --- a/src/mesa/drivers/dri/radeon/radeon_swtcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.c @@ -179,7 +179,7 @@ static void radeonSetVertexFormat( GLcontext *ctx ) for (i = 0; i < ctx->Const.MaxTextureUnits; i++) { if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX(i) )) { - GLuint sz = VB->TexCoordPtr[i]->size; + GLuint sz = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->size; switch (sz) { case 1: @@ -309,7 +309,7 @@ void r100_swtcl_flush(GLcontext *ctx, uint32_t current_offset) radeonEmitState(&rmesa->radeon); radeonEmitVertexAOS( rmesa, rmesa->radeon.swtcl.vertex_size, - first_elem(&rmesa->radeon.dma.reserved)->bo, + rmesa->radeon.swtcl.bo, current_offset); diff --git a/src/mesa/drivers/dri/radeon/radeon_texstate.c b/src/mesa/drivers/dri/radeon/radeon_texstate.c index 3cbe3b47254..84ddcfd4fd3 100644 --- a/src/mesa/drivers/dri/radeon/radeon_texstate.c +++ b/src/mesa/drivers/dri/radeon/radeon_texstate.c @@ -672,24 +672,13 @@ void radeonSetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_ return; } - radeon_update_renderbuffers(pDRICtx, dPriv); - /* back & depth buffer are useless free them right away */ - rb = (void*)rfb->base.Attachment[BUFFER_DEPTH].Renderbuffer; - if (rb && rb->bo) { - radeon_bo_unref(rb->bo); - rb->bo = NULL; - } - rb = (void*)rfb->base.Attachment[BUFFER_BACK_LEFT].Renderbuffer; - if (rb && rb->bo) { - radeon_bo_unref(rb->bo); - rb->bo = NULL; - } + radeon_update_renderbuffers(pDRICtx, dPriv, GL_TRUE); rb = rfb->color_rb[0]; if (rb->bo == NULL) { /* Failed to BO for the buffer */ return; } - + _mesa_lock_texture(radeon->glCtx, texObj); if (t->bo) { radeon_bo_unref(t->bo); diff --git a/src/mesa/drivers/dri/savage/savagerender.c b/src/mesa/drivers/dri/savage/savagerender.c index 32c74f9467e..8221edf3872 100644 --- a/src/mesa/drivers/dri/savage/savagerender.c +++ b/src/mesa/drivers/dri/savage/savagerender.c @@ -252,13 +252,13 @@ static GLboolean run_texnorm_stage( GLcontext *ctx, const GLboolean normalizeS = (texObj->WrapS == GL_REPEAT); const GLboolean normalizeT = (reallyEnabled & TEXTURE_2D_BIT) && (texObj->WrapT == GL_REPEAT); - const GLfloat *in = (GLfloat *)VB->TexCoordPtr[i]->data; - const GLint instride = VB->TexCoordPtr[i]->stride; + const GLfloat *in = (GLfloat *)VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->data; + const GLint instride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->stride; GLfloat (*out)[4] = store->texcoord[i].data; GLint j; if (!ctx->Texture.Unit[i]._ReallyEnabled || - VB->TexCoordPtr[i]->size == 4) + VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->size == 4) /* Never try to normalize homogenous tex coords! */ continue; @@ -297,7 +297,7 @@ static GLboolean run_texnorm_stage( GLcontext *ctx, } if (normalizeS || normalizeT) - VB->AttribPtr[VERT_ATTRIB_TEX0+i] = VB->TexCoordPtr[i] = &store->texcoord[i]; + VB->AttribPtr[_TNL_ATTRIB_TEX0 + i] = &store->texcoord[i]; } } diff --git a/src/mesa/drivers/dri/savage/savagetris.c b/src/mesa/drivers/dri/savage/savagetris.c index 0714101dc05..9a92541ef7e 100644 --- a/src/mesa/drivers/dri/savage/savagetris.c +++ b/src/mesa/drivers/dri/savage/savagetris.c @@ -880,13 +880,13 @@ static GLboolean savageCheckPTexHack( GLcontext *ctx ) RENDERINPUTS_COPY( index_bitset, tnl->render_inputs_bitset ); - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 ) && VB->TexCoordPtr[0]->size == 4) { + if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 ) && VB->AttribPtr[_TNL_ATTRIB_TEX0]->size == 4) { if (!RENDERINPUTS_TEST_RANGE( index_bitset, _TNL_ATTRIB_TEX1, _TNL_LAST_TEX )) return GL_TRUE; /* apply ptex hack */ else FALLBACK(ctx, SAVAGE_FALLBACK_PROJ_TEXTURE, GL_TRUE); } - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX1 ) && VB->TexCoordPtr[1]->size == 4) + if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX1 ) && VB->AttribPtr[_TNL_ATTRIB_TEX1]->size == 4) FALLBACK(ctx, SAVAGE_FALLBACK_PROJ_TEXTURE, GL_TRUE); return GL_FALSE; /* don't apply ptex hack */ @@ -977,13 +977,13 @@ static INLINE GLuint savageChooseVertexFormat_s3d( GLcontext *ctx ) if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 )) { if (imesa->ptexHack) EMIT_ATTR( _TNL_ATTRIB_TEX0, EMIT_3F_XYW, SAVAGE_EMIT_STQ0, SAVAGE_SKIP_ST0); - else if (VB->TexCoordPtr[0]->size == 4) + else if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size == 4) assert (0); /* should be caught by savageCheckPTexHack */ - else if (VB->TexCoordPtr[0]->size >= 2) + else if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size >= 2) /* The chromium menu emits some 3D tex coords even though no * 3D texture is enabled. Ignore the 3rd coordinate. */ EMIT_ATTR( _TNL_ATTRIB_TEX0, EMIT_2F, SAVAGE_EMIT_ST0, SAVAGE_SKIP_ST0 ); - else if (VB->TexCoordPtr[0]->size == 1) { + else if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size == 1) { EMIT_ATTR( _TNL_ATTRIB_TEX0, EMIT_1F, SAVAGE_EMIT_S0, SAVAGE_SKIP_S0 ); EMIT_PAD( 4 ); } else @@ -1026,9 +1026,9 @@ static INLINE GLuint savageChooseVertexFormat_s4( GLcontext *ctx ) if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 )) { if (imesa->ptexHack) NEED_ATTR( SAVAGE_EMIT_STQ0, SAVAGE_SKIP_ST0); - else if (VB->TexCoordPtr[0]->size == 4) + else if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size == 4) assert (0); /* should be caught by savageCheckPTexHack */ - else if (VB->TexCoordPtr[0]->size >= 2) + else if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size >= 2) /* The chromium menu emits some 3D tex coords even though no * 3D texture is enabled. Ignore the 3rd coordinate. */ NEED_ATTR( SAVAGE_EMIT_ST0, SAVAGE_SKIP_ST0 ); @@ -1036,10 +1036,10 @@ static INLINE GLuint savageChooseVertexFormat_s4( GLcontext *ctx ) NEED_ATTR( SAVAGE_EMIT_S0, SAVAGE_SKIP_S0 ); } if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX1 )) { - if (VB->TexCoordPtr[1]->size == 4) + if (VB->AttribPtr[_TNL_ATTRIB_TEX1]->size == 4) /* projective textures are not supported by the hardware */ assert (0); /* should be caught by savageCheckPTexHack */ - else if (VB->TexCoordPtr[1]->size >= 2) + else if (VB->AttribPtr[_TNL_ATTRIB_TEX1]->size >= 2) NEED_ATTR( SAVAGE_EMIT_ST1, SAVAGE_SKIP_ST1 ); else NEED_ATTR( SAVAGE_EMIT_S1, SAVAGE_SKIP_S1 ); diff --git a/src/mesa/drivers/dri/sis/sis_tris.c b/src/mesa/drivers/dri/sis/sis_tris.c index 4fa2e41e44f..4690274c3c0 100644 --- a/src/mesa/drivers/dri/sis/sis_tris.c +++ b/src/mesa/drivers/dri/sis/sis_tris.c @@ -904,14 +904,14 @@ static void sisRenderStart( GLcontext *ctx ) /* projective textures are not supported by the hardware */ if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 )) { - if (VB->TexCoordPtr[0]->size > 2) + if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size > 2) tex_fallback = GL_TRUE; EMIT_ATTR(_TNL_ATTRIB_TEX0, EMIT_2F); AGPParseSet |= SiS_PS_HAS_UV0; } /* Will only hit tex1 on SiS300 */ if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX1 )) { - if (VB->TexCoordPtr[1]->size > 2) + if (VB->AttribPtr[_TNL_ATTRIB_TEX1]->size > 2) tex_fallback = GL_TRUE; EMIT_ATTR(_TNL_ATTRIB_TEX1, EMIT_2F); AGPParseSet |= SiS_PS_HAS_UV1; diff --git a/src/mesa/drivers/dri/tdfx/tdfx_vb.c b/src/mesa/drivers/dri/tdfx/tdfx_vb.c index 49288022328..c200ba3255a 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_vb.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_vb.c @@ -69,11 +69,11 @@ static void interp_extras( GLcontext *ctx, /*fprintf(stderr, "%s\n", __FUNCTION__);*/ - if (VB->ColorPtr[1]) { + if (VB->BackfaceColorPtr) { INTERP_4F( t, - GET_COLOR(VB->ColorPtr[1], dst), - GET_COLOR(VB->ColorPtr[1], out), - GET_COLOR(VB->ColorPtr[1], in) ); + GET_COLOR(VB->BackfaceColorPtr, dst), + GET_COLOR(VB->BackfaceColorPtr, out), + GET_COLOR(VB->BackfaceColorPtr, in) ); } if (VB->EdgeFlag) { @@ -88,9 +88,9 @@ static void copy_pv_extras( GLcontext *ctx, GLuint dst, GLuint src ) { struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; - if (VB->ColorPtr[1]) { - COPY_4FV( GET_COLOR(VB->ColorPtr[1], dst), - GET_COLOR(VB->ColorPtr[1], src) ); + if (VB->BackfaceColorPtr) { + COPY_4FV( GET_COLOR(VB->BackfaceColorPtr, dst), + GET_COLOR(VB->BackfaceColorPtr, src) ); } setup_tab[TDFX_CONTEXT(ctx)->SetupIndex].copy_pv(ctx, dst, src); diff --git a/src/mesa/drivers/dri/tdfx/tdfx_vbtmp.h b/src/mesa/drivers/dri/tdfx/tdfx_vbtmp.h index 9b780761f42..19baf7d0d25 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_vbtmp.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_vbtmp.h @@ -58,32 +58,32 @@ static void TAG(emit)( GLcontext *ctx, /* fprintf(stderr, "%s\n", __FUNCTION__); */ if (IND & TDFX_TEX0_BIT) { - tc0_stride = VB->TexCoordPtr[tmu0_source]->stride; - tc0 = VB->TexCoordPtr[tmu0_source]->data; + tc0_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu0_source]->stride; + tc0 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu0_source]->data; u0scale = fxMesa->sScale0; v0scale = fxMesa->tScale0; if (IND & TDFX_PTEX_BIT) - tc0_size = VB->TexCoordPtr[tmu0_source]->size; + tc0_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu0_source]->size; } if (IND & TDFX_TEX1_BIT) { - tc1 = VB->TexCoordPtr[tmu1_source]->data; - tc1_stride = VB->TexCoordPtr[tmu1_source]->stride; + tc1 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu1_source]->data; + tc1_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu1_source]->stride; u1scale = fxMesa->sScale1; v1scale = fxMesa->tScale1; if (IND & TDFX_PTEX_BIT) - tc1_size = VB->TexCoordPtr[tmu1_source]->size; + tc1_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu1_source]->size; } if (IND & TDFX_RGBA_BIT) { - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; - col_size = VB->ColorPtr[0]->size; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; + col_size = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->size; } if (IND & TDFX_FOGC_BIT) { - fog = VB->FogCoordPtr->data; - fog_stride = VB->FogCoordPtr->stride; + fog = VB->AttribPtr[_TNL_ATTRIB_FOG]->data; + fog_stride = VB->AttribPtr[_TNL_ATTRIB_FOG]->stride; } { @@ -168,14 +168,14 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; if (IND & TDFX_TEX1_BIT) { - if (VB->TexCoordPtr[0] == 0) - VB->TexCoordPtr[0] = VB->TexCoordPtr[1]; + if (VB->AttribPtr[_TNL_ATTRIB_TEX0] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0] = VB->AttribPtr[_TNL_ATTRIB_TEX1]; - if (VB->TexCoordPtr[1]->size == 4) + if (VB->AttribPtr[_TNL_ATTRIB_TEX1]->size == 4) return GL_FALSE; } - if (VB->TexCoordPtr[0]->size == 4) + if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size == 4) return GL_FALSE; } diff --git a/src/mesa/drivers/dri/unichrome/via_tris.c b/src/mesa/drivers/dri/unichrome/via_tris.c index e2f1f02c99e..01359d51ea6 100644 --- a/src/mesa/drivers/dri/unichrome/via_tris.c +++ b/src/mesa/drivers/dri/unichrome/via_tris.c @@ -833,13 +833,13 @@ static GLboolean viaCheckPTexHack( GLcontext *ctx ) RENDERINPUTS_COPY( index_bitset, tnl->render_inputs_bitset ); - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 ) && VB->TexCoordPtr[0]->size == 4) { + if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 ) && VB->AttribPtr[_TNL_ATTRIB_TEX0]->size == 4) { if (!RENDERINPUTS_TEST_RANGE( index_bitset, _TNL_ATTRIB_TEX1, _TNL_LAST_TEX )) ptexHack = GL_TRUE; else fallback = GL_TRUE; } - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX1 ) && VB->TexCoordPtr[1]->size == 4) + if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX1 ) && VB->AttribPtr[_TNL_ATTRIB_TEX1]->size == 4) fallback = GL_TRUE; FALLBACK(VIA_CONTEXT(ctx), VIA_FALLBACK_PROJ_TEXTURE, fallback); diff --git a/src/mesa/drivers/glide/fxvb.c b/src/mesa/drivers/glide/fxvb.c index 1dc5f9891a1..cc9ad0e8b83 100644 --- a/src/mesa/drivers/glide/fxvb.c +++ b/src/mesa/drivers/glide/fxvb.c @@ -104,24 +104,24 @@ static void interp_extras( GLcontext *ctx, { struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; - if (VB->ColorPtr[1]) { - /* If stride is zero, ColorPtr[1] is constant across the VB, so + if (VB->BackfaceColorPtr) { + /* If stride is zero, BackfaceColorPtr is constant across the VB, so * there is no point interpolating between two values as they will * be identical. This case is handled in t_dd_tritmp.h */ - if (VB->ColorPtr[1]->stride) { - assert(VB->ColorPtr[1]->stride == 4 * sizeof(GLfloat)); + if (VB->BackfaceColorPtr->stride) { + assert(VB->BackfaceColorPtr->stride == 4 * sizeof(GLfloat)); INTERP_4F( t, - GET_COLOR(VB->ColorPtr[1], dst), - GET_COLOR(VB->ColorPtr[1], out), - GET_COLOR(VB->ColorPtr[1], in) ); + GET_COLOR(VB->BackfaceColorPtr, dst), + GET_COLOR(VB->BackfaceColorPtr, out), + GET_COLOR(VB->BackfaceColorPtr, in) ); } - if (VB->SecondaryColorPtr[1]) { + if (VB->BackfaceSecondaryColorPtr) { INTERP_3F( t, - GET_COLOR(VB->SecondaryColorPtr[1], dst), - GET_COLOR(VB->SecondaryColorPtr[1], out), - GET_COLOR(VB->SecondaryColorPtr[1], in) ); + GET_COLOR(VB->BackfaceSecondaryColorPtr, dst), + GET_COLOR(VB->BackfaceSecondaryColorPtr, out), + GET_COLOR(VB->BackfaceSecondaryColorPtr, in) ); } } @@ -137,13 +137,13 @@ static void copy_pv_extras( GLcontext *ctx, GLuint dst, GLuint src ) { struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; - if (VB->ColorPtr[1]) { - COPY_4FV( GET_COLOR(VB->ColorPtr[1], dst), - GET_COLOR(VB->ColorPtr[1], src) ); + if (VB->BackfaceColorPtr) { + COPY_4FV( GET_COLOR(VB->BackfaceColorPtr, dst), + GET_COLOR(VB->BackfaceColorPtr, src) ); - if (VB->SecondaryColorPtr[1]) { - COPY_3FV( GET_COLOR(VB->SecondaryColorPtr[1], dst), - GET_COLOR(VB->SecondaryColorPtr[1], src) ); + if (VB->BackfaceSecondaryColorPtr) { + COPY_3FV( GET_COLOR(VB->BackfaceSecondaryColorPtr, dst), + GET_COLOR(VB->BackfaceSecondaryColorPtr, src) ); } } diff --git a/src/mesa/drivers/glide/fxvbtmp.h b/src/mesa/drivers/glide/fxvbtmp.h index f7970c78e29..f7893c1573f 100644 --- a/src/mesa/drivers/glide/fxvbtmp.h +++ b/src/mesa/drivers/glide/fxvbtmp.h @@ -62,37 +62,37 @@ static void TAG(emit)( GLcontext *ctx, } if (IND & SETUP_TMU0) { - tc0 = VB->TexCoordPtr[tmu0_source]->data; - tc0_stride = VB->TexCoordPtr[tmu0_source]->stride; + tc0 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu0_source]->data; + tc0_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu0_source]->stride; u0scale = fxMesa->s0scale; v0scale = fxMesa->t0scale; if (IND & SETUP_PTEX) - tc0_size = VB->TexCoordPtr[tmu0_source]->size; + tc0_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu0_source]->size; } if (IND & SETUP_TMU1) { - tc1 = VB->TexCoordPtr[tmu1_source]->data; - tc1_stride = VB->TexCoordPtr[tmu1_source]->stride; + tc1 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu1_source]->data; + tc1_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu1_source]->stride; u1scale = fxMesa->s1scale; /* wrong if tmu1_source == 0, possible? */ v1scale = fxMesa->t1scale; if (IND & SETUP_PTEX) - tc1_size = VB->TexCoordPtr[tmu1_source]->size; + tc1_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu1_source]->size; } if (IND & SETUP_RGBA) { - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; - col_size = VB->ColorPtr[0]->size; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; + col_size = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->size; } if (IND & SETUP_SPEC) { - spec = VB->SecondaryColorPtr[0]->data; - spec_stride = VB->SecondaryColorPtr[0]->stride; + spec = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->data; + spec_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->stride; } if (IND & SETUP_FOGC) { - fog = VB->FogCoordPtr->data; - fog_stride = VB->FogCoordPtr->stride; + fog = VB->AttribPtr[_TNL_ATTRIB_FOG]->data; + fog_stride = VB->AttribPtr[_TNL_ATTRIB_FOG]->stride; } if (start) { @@ -220,14 +220,15 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; if (IND & SETUP_TMU1) { - if (VB->TexCoordPtr[0] == 0) - VB->TexCoordPtr[0] = VB->TexCoordPtr[1]; + if (VB->AttribPtr[_TNL_ATTRIB_TEX0] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0] = VB->AttribPtr[_TNL_ATTRIB_TEX1]; - if (VB->TexCoordPtr[1]->size == 4) + if (VB->AttribPtr[_TNL_ATTRIB_TEX1]->size == 4) return GL_FALSE; } - if (VB->TexCoordPtr[0] && VB->TexCoordPtr[0]->size == 4) + if (VB->AttribPtr[_TNL_ATTRIB_TEX0] && + VB->AttribPtr[_TNL_ATTRIB_TEX0]->size == 4) return GL_FALSE; } diff --git a/src/mesa/drivers/osmesa/Makefile b/src/mesa/drivers/osmesa/Makefile index 92d41494665..3a94cc95a13 100644 --- a/src/mesa/drivers/osmesa/Makefile +++ b/src/mesa/drivers/osmesa/Makefile @@ -21,7 +21,11 @@ INCLUDE_DIRS = \ # Standalone osmesa needs to be linked with core Mesa APIs ifeq ($(DRIVER_DIRS), osmesa) -CORE_MESA = $(TOP)/src/mesa/libmesa.a $(TOP)/src/mesa/libglapi.a +CORE_MESA = \ + $(TOP)/src/mesa/libmesa.a \ + $(TOP)/src/mesa/libglapi.a \ + $(TOP)/src/glsl/cl/libglslcl.a \ + $(TOP)/src/glsl/cl/libglslpp.a else CORE_MESA = endif diff --git a/src/mesa/drivers/windows/gldirect/dx7/gld_primitive_dx7.c b/src/mesa/drivers/windows/gldirect/dx7/gld_primitive_dx7.c index c99ba0bba55..0b373814fee 100644 --- a/src/mesa/drivers/windows/gldirect/dx7/gld_primitive_dx7.c +++ b/src/mesa/drivers/windows/gldirect/dx7/gld_primitive_dx7.c @@ -189,9 +189,9 @@ GLfloat ex,ey,fx,fy,cc; \ /* Get vars for later */ \ VB = &TNL_CONTEXT(ctx)->vb; \ - vbcolor = (GLchan (*)[4])VB->ColorPtr[1]->data; \ - if (VB->SecondaryColorPtr[1]) { \ - vbspec = (GLchan (*)[4])VB->SecondaryColorPtr[1]->data; \ + vbcolor = (GLchan (*)[4])VB->BackfaceColorPtr->data; \ + if (VB->BackfaceSecondaryColorPtr) { \ + vbspec = (GLchan (*)[4])VB->BackfaceSecondaryColorPtr->data; \ } else { \ vbspec = NULL; \ } \ @@ -241,33 +241,33 @@ DWORD dwColor; #define GLD_SETUP_3D_VERTEX(v) \ - p4f = VB->ObjPtr->data; \ + p4f = VB->AttribPtr[_TNL_ATTRIB_POS]->data; \ pV->Position.x = p4f[##v][0]; \ pV->Position.y = p4f[##v][1]; \ pV->Position.z = p4f[##v][2]; #define GLD_SETUP_SMOOTH_COLOUR_3D(v) \ - p4f = (GLfloat (*)[4])VB->ColorPtr[0]->data; \ + p4f = (GLfloat (*)[4])VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; \ pV->Diffuse = D3DCOLOR_COLORVALUE(p4f[##v][0], p4f[##v][1], p4f[##v][2], p4f[##v][3]); #define GLD_SETUP_GET_FLAT_COLOUR_3D(v) \ - p4f = (GLfloat (*)[4])VB->ColorPtr[0]->data; \ + p4f = (GLfloat (*)[4])VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; \ dwColor = D3DCOLOR_COLORVALUE(p4f[##v][0], p4f[##v][1], p4f[##v][2], p4f[##v][3]); #define GLD_SETUP_USE_FLAT_COLOUR_3D \ pV->Diffuse = dwColor; #define GLD_SETUP_TEX0_3D(v) \ - if (VB->TexCoordPtr[0]) { \ - tc = VB->TexCoordPtr[0]->data; \ + if (VB->AttribPtr[_TNL_ATTRIB_TEX0]) { \ + tc = VB->AttribPtr[_TNL_ATTRIB_TEX0]->data; \ pV->TexUnit0.x = tc[##v][0]; \ pV->TexUnit0.y = tc[##v][1]; \ } #define GLD_SETUP_TEX1_3D(v) \ - if (VB->TexCoordPtr[1]) { \ - tc = VB->TexCoordPtr[1]->data; \ + if (VB->AttribPtr[_TNL_ATTRIB_TEX1]) { \ + tc = VB->AttribPtr[_TNL_ATTRIB_TEX1]->data; \ pV->TexUnit1.x = tc[##v][0]; \ pV->TexUnit1.y = tc[##v][1]; \ } diff --git a/src/mesa/drivers/windows/gldirect/dx7/gld_vb_d3d_render_dx7.c b/src/mesa/drivers/windows/gldirect/dx7/gld_vb_d3d_render_dx7.c index a85620dde8d..c39775cad32 100644 --- a/src/mesa/drivers/windows/gldirect/dx7/gld_vb_d3d_render_dx7.c +++ b/src/mesa/drivers/windows/gldirect/dx7/gld_vb_d3d_render_dx7.c @@ -151,7 +151,7 @@ static GLboolean gld_d3d_render_stage_run( #if 0 // For debugging: Useful to see if an app passes colour data in // an unusual format. - switch (VB->ColorPtr[0]->Type) { + switch (VB->AttribPtr[_TNL_ATTRIB_COLOR0]->Type) { case GL_FLOAT: ddlogMessage(GLDLOG_SYSTEM, "ColorPtr: GL_FLOAT\n"); break; diff --git a/src/mesa/drivers/windows/gldirect/dx8/gld_primitive_dx8.c b/src/mesa/drivers/windows/gldirect/dx8/gld_primitive_dx8.c index a5b5462f039..990922580aa 100644 --- a/src/mesa/drivers/windows/gldirect/dx8/gld_primitive_dx8.c +++ b/src/mesa/drivers/windows/gldirect/dx8/gld_primitive_dx8.c @@ -189,9 +189,9 @@ GLfloat ex,ey,fx,fy,cc; \ /* Get vars for later */ \ VB = &TNL_CONTEXT(ctx)->vb; \ - vbcolor = (GLchan (*)[4])VB->ColorPtr[1]->data; \ - if (VB->SecondaryColorPtr[1]) { \ - vbspec = (GLchan (*)[4])VB->SecondaryColorPtr[1]->data; \ + vbcolor = (GLchan (*)[4])VB->BackfaceColorPtr->data; \ + if (VB->BackfaceSecondaryColorPtr) { \ + vbspec = (GLchan (*)[4])VB->BackfaceSecondaryColorPtr->data; \ } else { \ vbspec = NULL; \ } \ @@ -241,33 +241,33 @@ DWORD dwColor; #define GLD_SETUP_3D_VERTEX(v) \ - p4f = VB->ObjPtr->data; \ + p4f = VB->AttribPtr[_TNL_ATTRIB_POS]->data; \ pV->Position.x = p4f[##v][0]; \ pV->Position.y = p4f[##v][1]; \ pV->Position.z = p4f[##v][2]; #define GLD_SETUP_SMOOTH_COLOUR_3D(v) \ - p4f = (GLfloat (*)[4])VB->ColorPtr[0]->data; \ + p4f = (GLfloat (*)[4])VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; \ pV->Diffuse = D3DCOLOR_COLORVALUE(p4f[##v][0], p4f[##v][1], p4f[##v][2], p4f[##v][3]); #define GLD_SETUP_GET_FLAT_COLOUR_3D(v) \ - p4f = (GLfloat (*)[4])VB->ColorPtr[0]->data; \ + p4f = (GLfloat (*)[4])VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; \ dwColor = D3DCOLOR_COLORVALUE(p4f[##v][0], p4f[##v][1], p4f[##v][2], p4f[##v][3]); #define GLD_SETUP_USE_FLAT_COLOUR_3D \ pV->Diffuse = dwColor; #define GLD_SETUP_TEX0_3D(v) \ - if (VB->TexCoordPtr[0]) { \ - tc = VB->TexCoordPtr[0]->data; \ + if (VB->AttribPtr[_TNL_ATTRIB_TEX0]) { \ + tc = VB->TnlAttribPtr[_TNL_ATTRIB_TEX0]->data; \ pV->TexUnit0.x = tc[##v][0]; \ pV->TexUnit0.y = tc[##v][1]; \ } #define GLD_SETUP_TEX1_3D(v) \ - if (VB->TexCoordPtr[1]) { \ - tc = VB->TexCoordPtr[1]->data; \ + if (VB->TnlAttribPtr[_TNL_ATTRIB_TEX1]) { \ + tc = VB->TnlAttribPtr[_TNL_ATTRIB_TEX1]->data; \ pV->TexUnit1.x = tc[##v][0]; \ pV->TexUnit1.y = tc[##v][1]; \ } diff --git a/src/mesa/drivers/windows/gldirect/dx8/gld_vb_d3d_render_dx8.c b/src/mesa/drivers/windows/gldirect/dx8/gld_vb_d3d_render_dx8.c index cafbf4f5c50..265c81fb4a4 100644 --- a/src/mesa/drivers/windows/gldirect/dx8/gld_vb_d3d_render_dx8.c +++ b/src/mesa/drivers/windows/gldirect/dx8/gld_vb_d3d_render_dx8.c @@ -149,7 +149,7 @@ static GLboolean gld_d3d_render_stage_run( #if 0 // For debugging: Useful to see if an app passes colour data in // an unusual format. - switch (VB->ColorPtr[0]->Type) { + switch (VB->AttribPtr[_TNL_ATTRIB_COLOR0]->Type) { case GL_FLOAT: ddlogMessage(GLDLOG_SYSTEM, "ColorPtr: GL_FLOAT\n"); break; diff --git a/src/mesa/drivers/windows/gldirect/dx9/gld_primitive_dx9.c b/src/mesa/drivers/windows/gldirect/dx9/gld_primitive_dx9.c index 403a9d5f86e..fd4dd4ed751 100644 --- a/src/mesa/drivers/windows/gldirect/dx9/gld_primitive_dx9.c +++ b/src/mesa/drivers/windows/gldirect/dx9/gld_primitive_dx9.c @@ -189,9 +189,9 @@ GLfloat ex,ey,fx,fy,cc; \ /* Get vars for later */ \ VB = &TNL_CONTEXT(ctx)->vb; \ - vbcolor = (GLchan (*)[4])VB->ColorPtr[1]->data; \ - if (VB->SecondaryColorPtr[1]) { \ - vbspec = (GLchan (*)[4])VB->SecondaryColorPtr[1]->data; \ + vbcolor = (GLchan (*)[4])VB->BackfaceColorPtr->data; \ + if (VB->BackfaceSecondaryColorPtr) { \ + vbspec = (GLchan (*)[4])VB->BackfaceSecondaryColorPtr->data; \ } else { \ vbspec = NULL; \ } \ @@ -241,33 +241,33 @@ DWORD dwColor; #define GLD_SETUP_3D_VERTEX(v) \ - p4f = VB->ObjPtr->data; \ + p4f = VB->AttribPtr[_TNL_ATTRIB_POS]->data; \ pV->Position.x = p4f[##v][0]; \ pV->Position.y = p4f[##v][1]; \ pV->Position.z = p4f[##v][2]; #define GLD_SETUP_SMOOTH_COLOUR_3D(v) \ - p4f = (GLfloat (*)[4])VB->ColorPtr[0]->data; \ + p4f = (GLfloat (*)[4])VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; \ pV->Diffuse = D3DCOLOR_COLORVALUE(p4f[##v][0], p4f[##v][1], p4f[##v][2], p4f[##v][3]); #define GLD_SETUP_GET_FLAT_COLOUR_3D(v) \ - p4f = (GLfloat (*)[4])VB->ColorPtr[0]->data; \ + p4f = (GLfloat (*)[4])VB->AttribPtr[_TNL_ATTRIB_COLOR00]->data; \ dwColor = D3DCOLOR_COLORVALUE(p4f[##v][0], p4f[##v][1], p4f[##v][2], p4f[##v][3]); #define GLD_SETUP_USE_FLAT_COLOUR_3D \ pV->Diffuse = dwColor; #define GLD_SETUP_TEX0_3D(v) \ - if (VB->TexCoordPtr[0]) { \ - tc = VB->TexCoordPtr[0]->data; \ + if (VB->AttribPtr[_TNL_ATTRIB_TEX0]) { \ + tc = VB->AttribPtr[_TNL_ATTRIB_TEX0]->data; \ pV->TexUnit0.x = tc[##v][0]; \ pV->TexUnit0.y = tc[##v][1]; \ } #define GLD_SETUP_TEX1_3D(v) \ - if (VB->TexCoordPtr[1]) { \ - tc = VB->TexCoordPtr[1]->data; \ + if (VB->AttribPtr[_TNL_ATTRIB_TEX1]) { \ + tc = VB->AttribPtr[_TNL_ATTRIB_TEX1]->data; \ pV->TexUnit1.x = tc[##v][0]; \ pV->TexUnit1.y = tc[##v][1]; \ } diff --git a/src/mesa/drivers/windows/gldirect/dx9/gld_vb_d3d_render_dx9.c b/src/mesa/drivers/windows/gldirect/dx9/gld_vb_d3d_render_dx9.c index 4fa6bcaf1ab..91a68b3f2d9 100644 --- a/src/mesa/drivers/windows/gldirect/dx9/gld_vb_d3d_render_dx9.c +++ b/src/mesa/drivers/windows/gldirect/dx9/gld_vb_d3d_render_dx9.c @@ -149,7 +149,7 @@ static GLboolean gld_d3d_render_stage_run( #if 0 // For debugging: Useful to see if an app passes colour data in // an unusual format. - switch (VB->ColorPtr[0]->Type) { + switch (VB->AttribPtr[_TNL_ATTRIB_COLOR0]->Type) { case GL_FLOAT: ddlogMessage(GLDLOG_SYSTEM, "ColorPtr: GL_FLOAT\n"); break; |