diff options
Diffstat (limited to 'src/mesa/drivers/dri')
50 files changed, 2555 insertions, 622 deletions
diff --git a/src/mesa/drivers/dri/ffb/ffb_vbtmp.h b/src/mesa/drivers/dri/ffb/ffb_vbtmp.h index 0495d0e276..c548ef3ad5 100644 --- a/src/mesa/drivers/dri/ffb/ffb_vbtmp.h +++ b/src/mesa/drivers/dri/ffb/ffb_vbtmp.h @@ -38,11 +38,11 @@ static void TAG(emit)(GLcontext *ctx, GLuint start, GLuint end) #endif #if (IND & (FFB_VB_RGBA_BIT)) - col0 = VB->ColorPtr[0]->data; - col0_stride = VB->ColorPtr[0]->stride; + col0 = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col0_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; #if (IND & (FFB_VB_TWOSIDE_BIT)) - col1 = VB->ColorPtr[1]->data; - col1_stride = VB->ColorPtr[1]->stride; + col1 = VB->BackfaceColorPtr->data; + col1_stride = VB->BackfaceColorPtr->stride; #endif #endif diff --git a/src/mesa/drivers/dri/gamma/gamma_render.c b/src/mesa/drivers/dri/gamma/gamma_render.c index 1b9fd169f4..a03a93d132 100644 --- a/src/mesa/drivers/dri/gamma/gamma_render.c +++ b/src/mesa/drivers/dri/gamma/gamma_render.c @@ -53,13 +53,13 @@ static void gamma_emit( GLcontext *ctx, GLuint start, GLuint end) GLfloat (*tc0)[4] = 0; GLuint tc0_size = 0; - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; if (ctx->Texture.Unit[0]._ReallyEnabled) { - tc0_stride = VB->TexCoordPtr[0]->stride; - tc0 = VB->TexCoordPtr[0]->data; - tc0_size = VB->TexCoordPtr[0]->size; + tc0_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0]->stride; + tc0 = VB->AttribPtr[_TNL_ATTRIB_TEX0]->data; + tc0_size = VB->AttribPtr[_TNL_ATTRIB_TEX0]->size; coord = VB->ClipPtr->data; coord_stride = VB->ClipPtr->stride; } else { diff --git a/src/mesa/drivers/dri/i915/i830_texstate.c b/src/mesa/drivers/dri/i915/i830_texstate.c index f4bbb53b86..c62281d341 100644 --- a/src/mesa/drivers/dri/i915/i830_texstate.c +++ b/src/mesa/drivers/dri/i915/i830_texstate.c @@ -27,6 +27,7 @@ #include "main/mtypes.h" #include "main/enums.h" +#include "main/colormac.h" #include "intel_mipmap_tree.h" #include "intel_tex.h" @@ -311,11 +312,10 @@ i830_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3) CLAMPED_FLOAT_TO_UBYTE(border[2], tObj->BorderColor[2]); CLAMPED_FLOAT_TO_UBYTE(border[3], tObj->BorderColor[3]); - state[I830_TEXREG_TM0S4] = INTEL_PACKCOLOR8888(border[0], - border[1], - border[2], - border[3]); - + state[I830_TEXREG_TM0S4] = PACK_COLOR_8888(border[3], + border[0], + border[1], + border[2]); I830_ACTIVESTATE(i830, I830_UPLOAD_TEX(unit), GL_TRUE); /* memcmp was already disabled, but definitely won't work as the diff --git a/src/mesa/drivers/dri/i915/i830_vtbl.c b/src/mesa/drivers/dri/i915/i830_vtbl.c index a6f554701e..c05c7759ac 100644 --- a/src/mesa/drivers/dri/i915/i830_vtbl.c +++ b/src/mesa/drivers/dri/i915/i830_vtbl.c @@ -126,7 +126,7 @@ i830_render_start(struct intel_context *intel) for (i = 0; i < I830_TEX_UNITS; i++) { if (RENDERINPUTS_TEST(index_bitset, _TNL_ATTRIB_TEX(i))) { - GLuint sz = VB->TexCoordPtr[i]->size; + GLuint sz = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->size; GLuint emit; GLuint mcs = (i830->state.Tex[i][I830_TEXREG_MCS] & ~TEXCOORDTYPE_MASK); @@ -714,9 +714,6 @@ i830_new_batch(struct intel_context *intel) { struct i830_context *i830 = i830_context(&intel->ctx); i830->state.emitted = 0; - - /* Check that we didn't just wrap our batchbuffer at a bad time. */ - assert(!intel->no_batch_wrap); } static void diff --git a/src/mesa/drivers/dri/i915/i915_fragprog.c b/src/mesa/drivers/dri/i915/i915_fragprog.c index d9c61446f5..9e4d318036 100644 --- a/src/mesa/drivers/dri/i915/i915_fragprog.c +++ b/src/mesa/drivers/dri/i915/i915_fragprog.c @@ -1301,7 +1301,7 @@ i915ValidateFragmentProgram(struct i915_context *i915) for (i = 0; i < p->ctx->Const.MaxTextureCoordUnits; i++) { if (inputsRead & FRAG_BIT_TEX(i)) { - int sz = VB->TexCoordPtr[i]->size; + int sz = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->size; s2 &= ~S2_TEXCOORD_FMT(i, S2_TEXCOORD_FMT0_MASK); s2 |= S2_TEXCOORD_FMT(i, SZ_TO_HW(sz)); diff --git a/src/mesa/drivers/dri/i915/i915_texstate.c b/src/mesa/drivers/dri/i915/i915_texstate.c index d6689af53f..1bacd51aec 100644 --- a/src/mesa/drivers/dri/i915/i915_texstate.c +++ b/src/mesa/drivers/dri/i915/i915_texstate.c @@ -28,6 +28,7 @@ #include "main/mtypes.h" #include "main/enums.h" #include "main/macros.h" +#include "main/colormac.h" #include "intel_mipmap_tree.h" #include "intel_tex.h" @@ -363,15 +364,15 @@ i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3) * R channel, while the hardware uses A. Spam R into all the channels * for safety. */ - state[I915_TEXREG_SS4] = INTEL_PACKCOLOR8888(border[0], - border[0], - border[0], - border[0]); + state[I915_TEXREG_SS4] = PACK_COLOR_8888(border[0], + border[0], + border[0], + border[0]); } else { - state[I915_TEXREG_SS4] = INTEL_PACKCOLOR8888(border[0], - border[1], - border[2], - border[3]); + state[I915_TEXREG_SS4] = PACK_COLOR_8888(border[3], + border[0], + border[1], + border[2]); } diff --git a/src/mesa/drivers/dri/i915/i915_vtbl.c b/src/mesa/drivers/dri/i915/i915_vtbl.c index 77ba8d5581..3e7b5101cc 100644 --- a/src/mesa/drivers/dri/i915/i915_vtbl.c +++ b/src/mesa/drivers/dri/i915/i915_vtbl.c @@ -667,9 +667,6 @@ i915_new_batch(struct intel_context *intel) * difficulties associated with them (physical address requirements). */ i915->state.emitted = 0; - - /* Check that we didn't just wrap our batchbuffer at a bad time. */ - assert(!intel->no_batch_wrap); } static void diff --git a/src/mesa/drivers/dri/i915/intel_tris.c b/src/mesa/drivers/dri/i915/intel_tris.c index bc527aae47..8a3ab39bc2 100644 --- a/src/mesa/drivers/dri/i915/intel_tris.c +++ b/src/mesa/drivers/dri/i915/intel_tris.c @@ -1250,81 +1250,6 @@ union fi GLint i; }; - -/**********************************************************************/ -/* Used only with the metaops callbacks. */ -/**********************************************************************/ -static void -intel_meta_draw_poly(struct intel_context *intel, - GLuint n, - GLfloat xy[][2], - GLfloat z, GLuint color, GLfloat tex[][2]) -{ - union fi *vb; - GLint i; - unsigned int saved_vertex_size = intel->vertex_size; - - LOCK_HARDWARE(intel); - - intel->vertex_size = 6; - - /* All 3d primitives should be emitted with LOOP_CLIPRECTS, - * otherwise the drawing origin (DR4) might not be set correctly. - */ - intel_set_prim(intel, PRIM3D_TRIFAN); - vb = (union fi *) intel_get_prim_space(intel, n); - - for (i = 0; i < n; i++) { - vb[0].f = xy[i][0]; - vb[1].f = xy[i][1]; - vb[2].f = z; - vb[3].i = color; - vb[4].f = tex[i][0]; - vb[5].f = tex[i][1]; - vb += 6; - } - - INTEL_FIREVERTICES(intel); - - intel->vertex_size = saved_vertex_size; - - UNLOCK_HARDWARE(intel); -} - -static void -intel_meta_draw_quad(struct intel_context *intel, - GLfloat x0, GLfloat x1, - GLfloat y0, GLfloat y1, - GLfloat z, - GLuint color, - GLfloat s0, GLfloat s1, GLfloat t0, GLfloat t1) -{ - GLfloat xy[4][2]; - GLfloat tex[4][2]; - - xy[0][0] = x0; - xy[0][1] = y0; - xy[1][0] = x1; - xy[1][1] = y0; - xy[2][0] = x1; - xy[2][1] = y1; - xy[3][0] = x0; - xy[3][1] = y1; - - tex[0][0] = s0; - tex[0][1] = t0; - tex[1][0] = s1; - tex[1][1] = t0; - tex[2][0] = s1; - tex[2][1] = t1; - tex[3][0] = s0; - tex[3][1] = t1; - - intel_meta_draw_poly(intel, 4, xy, z, color, tex); -} - - - /**********************************************************************/ /* Initialization. */ /**********************************************************************/ @@ -1333,7 +1258,6 @@ intel_meta_draw_quad(struct intel_context *intel, void intelInitTriFuncs(GLcontext * ctx) { - struct intel_context *intel = intel_context(ctx); TNLcontext *tnl = TNL_CONTEXT(ctx); static int firsttime = 1; @@ -1350,6 +1274,4 @@ intelInitTriFuncs(GLcontext * ctx) tnl->Driver.Render.BuildVertices = _tnl_build_vertices; tnl->Driver.Render.CopyPV = _tnl_copy_pv; tnl->Driver.Render.Interp = _tnl_interp; - - intel->vtbl.meta_draw_quad = intel_meta_draw_quad; } diff --git a/src/mesa/drivers/dri/i965/brw_context.h b/src/mesa/drivers/dri/i965/brw_context.h index fded47aa2f..e73e21433c 100644 --- a/src/mesa/drivers/dri/i965/brw_context.h +++ b/src/mesa/drivers/dri/i965/brw_context.h @@ -172,8 +172,8 @@ struct brw_fragment_program { GLuint id; /**< serial no. to identify frag progs, never re-used */ GLboolean isGLSL; /**< really, any IF/LOOP/CONT/BREAK instructions */ - dri_bo *const_buffer; /** Program constant buffer/surface */ GLboolean use_const_buffer; + dri_bo *const_buffer; /** Program constant buffer/surface */ /** for debugging, which texture units are referenced */ GLbitfield tex_units_used; @@ -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 8bcb6083f7..7ad860898f 100644 --- a/src/mesa/drivers/dri/i965/brw_draw.c +++ b/src/mesa/drivers/dri/i965/brw_draw.c @@ -145,7 +145,7 @@ static void brw_emit_prim(struct brw_context *brw, prim_packet.base_vert_location = prim->basevertex; /* Can't wrap here, since we rely on the validated state. */ - brw->no_batch_wrap = GL_TRUE; + intel->no_batch_wrap = GL_TRUE; /* If we're set to always flush, do it before and after the primitive emit. * We want to catch both missed flushes that hurt instruction/state cache @@ -163,7 +163,7 @@ static void brw_emit_prim(struct brw_context *brw, intel_batchbuffer_emit_mi_flush(intel->batch); } - brw->no_batch_wrap = GL_FALSE; + intel->no_batch_wrap = GL_FALSE; } static void brw_merge_inputs( struct brw_context *brw, diff --git a/src/mesa/drivers/dri/i965/brw_draw_upload.c b/src/mesa/drivers/dri/i965/brw_draw_upload.c index 271a88dae0..7c796dae93 100644 --- a/src/mesa/drivers/dri/i965/brw_draw_upload.c +++ b/src/mesa/drivers/dri/i965/brw_draw_upload.c @@ -243,14 +243,6 @@ static void wrap_buffers( struct brw_context *brw, dri_bo_unreference(brw->vb.upload.bo); brw->vb.upload.bo = dri_bo_alloc(brw->intel.bufmgr, "temporary VBO", size, 1); - - /* Set the internal VBO\ to no-backing-store. We only use them as a - * temporary within a brw_try_draw_prims while the lock is held. - */ - /* DON'T DO THIS AS IF WE HAVE TO RE-ORG MEMORY WE NEED SOMEWHERE WITH - FAKE TO PUSH THIS STUFF */ -// if (!brw->intel.ttm) -// dri_bo_fake_disable_backing_store(brw->vb.upload.bo, NULL, NULL); } static void get_space( struct brw_context *brw, diff --git a/src/mesa/drivers/dri/i965/brw_vtbl.c b/src/mesa/drivers/dri/i965/brw_vtbl.c index 114e6bd018..dc47f08dd4 100644 --- a/src/mesa/drivers/dri/i965/brw_vtbl.c +++ b/src/mesa/drivers/dri/i965/brw_vtbl.c @@ -144,9 +144,6 @@ 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 9dcb6e14bb..b9b987ea70 100644 --- a/src/mesa/drivers/dri/i965/brw_wm.h +++ b/src/mesa/drivers/dri/i965/brw_wm.h @@ -76,10 +76,10 @@ struct brw_wm_prog_key { GLushort tex_swizzles[BRW_MAX_TEX_UNIT]; - GLuint program_string_id:32; GLushort origin_x, origin_y; GLushort drawable_height; 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 7d03179588..3737faf26f 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_fp.c +++ b/src/mesa/drivers/dri/i965/brw_wm_fp.c @@ -138,7 +138,6 @@ static struct prog_dst_register dst_reg(GLuint file, GLuint idx) reg.CondMask = COND_TR; reg.CondSwizzle = 0; reg.CondSrc = 0; - reg.pad = 0; return reg; } diff --git a/src/mesa/drivers/dri/intel/intel_batchbuffer.c b/src/mesa/drivers/dri/intel/intel_batchbuffer.c index ca6e2fa5b1..2eae9b66d8 100644 --- a/src/mesa/drivers/dri/intel/intel_batchbuffer.c +++ b/src/mesa/drivers/dri/intel/intel_batchbuffer.c @@ -80,7 +80,7 @@ intel_batchbuffer_reset(struct intel_batchbuffer *batch) batch->buf = NULL; } - if (!batch->buffer && intel->ttm == GL_TRUE) + if (!batch->buffer) batch->buffer = malloc (intel->maxBatchSize); batch->buf = dri_bo_alloc(intel->bufmgr, "batchbuffer", @@ -212,7 +212,7 @@ _intel_batchbuffer_flush(struct intel_batchbuffer *batch, const char *file, batch->reserved_space = 0; /* Emit a flush if the bufmgr doesn't do it for us. */ - if (intel->always_flush_cache || !intel->ttm) { + if (intel->always_flush_cache) { intel_batchbuffer_emit_mi_flush(batch); used = batch->ptr - batch->map; } @@ -244,6 +244,9 @@ _intel_batchbuffer_flush(struct intel_batchbuffer *batch, const char *file, if (intel->vtbl.finish_batch) intel->vtbl.finish_batch(intel); + /* Check that we didn't just wrap our batchbuffer at a bad time. */ + assert(!intel->no_batch_wrap); + batch->reserved_space = BATCH_RESERVED; /* TODO: Just pass the relocation list and dma buffer up to the diff --git a/src/mesa/drivers/dri/intel/intel_blit.c b/src/mesa/drivers/dri/intel/intel_blit.c index 817223da41..f14854602b 100644 --- a/src/mesa/drivers/dri/intel/intel_blit.c +++ b/src/mesa/drivers/dri/intel/intel_blit.c @@ -499,10 +499,11 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask) switch (irb->texformat) { case MESA_FORMAT_ARGB8888: case MESA_FORMAT_XRGB8888: - clearVal = intel->ClearColor8888; + clearVal = PACK_COLOR_8888(clear[3], clear[0], + clear[1], clear[2]); break; case MESA_FORMAT_RGB565: - clearVal = intel->ClearColor565; + clearVal = PACK_COLOR_565(clear[0], clear[1], clear[2]); break; case MESA_FORMAT_ARGB4444: clearVal = PACK_COLOR_4444(clear[3], clear[0], diff --git a/src/mesa/drivers/dri/intel/intel_context.c b/src/mesa/drivers/dri/intel/intel_context.c index 2aeca6b81b..1434ae530b 100644 --- a/src/mesa/drivers/dri/intel/intel_context.c +++ b/src/mesa/drivers/dri/intel/intel_context.c @@ -176,9 +176,7 @@ intelGetString(GLcontext * ctx, GLenum name) break; } - (void) driGetRendererString(buffer, chipset, - (intel->ttm) ? DRIVER_DATE_GEM : DRIVER_DATE, - 0); + (void) driGetRendererString(buffer, chipset, DRIVER_DATE_GEM, 0); return (GLubyte *) buffer; default: @@ -601,6 +599,7 @@ intelInitContext(struct intel_context *intel, __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; int fthrottle_mode; + int bo_reuse_mode; if (!_mesa_initialize_context(&intel->ctx, mesaVis, shareCtx, functions, (void *) intel)) { @@ -635,18 +634,14 @@ intelInitContext(struct intel_context *intel, intel->maxBatchSize = BATCH_SZ; intel->bufmgr = intelScreen->bufmgr; - intel->ttm = intelScreen->ttm; - if (intel->ttm) { - int bo_reuse_mode; - bo_reuse_mode = driQueryOptioni(&intel->optionCache, "bo_reuse"); - switch (bo_reuse_mode) { - case DRI_CONF_BO_REUSE_DISABLED: - break; - case DRI_CONF_BO_REUSE_ALL: - intel_bufmgr_gem_enable_reuse(intel->bufmgr); - break; - } + bo_reuse_mode = driQueryOptioni(&intel->optionCache, "bo_reuse"); + switch (bo_reuse_mode) { + case DRI_CONF_BO_REUSE_DISABLED: + break; + case DRI_CONF_BO_REUSE_ALL: + intel_bufmgr_gem_enable_reuse(intel->bufmgr); + break; } /* This doesn't yet catch all non-conformant rendering, but it's a @@ -733,12 +728,6 @@ intelInitContext(struct intel_context *intel, intel->RenderIndex = ~0; fthrottle_mode = driQueryOptioni(&intel->optionCache, "fthrottle_mode"); - intel->irqsEmitted = 0; - - intel->do_irqs = (intel->intelScreen->irq_active && - fthrottle_mode == DRI_CONF_FTHROTTLE_IRQS); - - intel->do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS); if (intel->gen >= 4 && !intel->intelScreen->irq_active) { _mesa_printf("IRQs not active. Exiting\n"); @@ -1058,21 +1047,6 @@ intelContendedLock(struct intel_context *intel, GLuint flags) sarea->ctxOwner = me; } - /* If the last consumer of the texture memory wasn't us, notify the fake - * bufmgr and record the new owner. We should have the memory shared - * between contexts of a single fake bufmgr, but this will at least make - * things correct for now. - */ - if (!intel->ttm && sarea->texAge != intel->hHWContext) { - sarea->texAge = intel->hHWContext; - intel_bufmgr_fake_contended_lock_take(intel->bufmgr); - if (INTEL_DEBUG & DEBUG_BATCH) - intel_decode_context_reset(); - if (INTEL_DEBUG & DEBUG_BUFMGR) - fprintf(stderr, "Lost Textures: sarea->texAge %x hw context %x\n", - sarea->ctxOwner, intel->hHWContext); - } - /* Drawable changed? */ if (dPriv && intel->lastStamp != dPriv->lastStamp) { diff --git a/src/mesa/drivers/dri/intel/intel_context.h b/src/mesa/drivers/dri/intel/intel_context.h index eb7be7ddd0..481202c971 100644 --- a/src/mesa/drivers/dri/intel/intel_context.h +++ b/src/mesa/drivers/dri/intel/intel_context.h @@ -135,14 +135,6 @@ struct intel_context struct intel_region * draw_region, struct intel_region * depth_region); - void (*meta_draw_quad)(struct intel_context *intel, - GLfloat x0, GLfloat x1, - GLfloat y0, GLfloat y1, - GLfloat z, - GLuint color, /* ARGB32 */ - GLfloat s0, GLfloat s1, - GLfloat t0, GLfloat t1); - void (*meta_color_mask) (struct intel_context * intel, GLboolean); void (*meta_stencil_replace) (struct intel_context * intel, @@ -189,12 +181,6 @@ struct intel_context struct intel_region *back_region; struct intel_region *depth_region; - /** - * This value indicates that the kernel memory manager is being used - * instead of the fake client-side memory manager. - */ - GLboolean ttm; - struct intel_batchbuffer *batch; drm_intel_bo *first_post_swapbuffers_batch; GLboolean no_batch_wrap; @@ -217,10 +203,6 @@ struct intel_context char *prevLockFile; int prevLockLine; - GLuint ClearColor565; - GLuint ClearColor8888; - - /* Offsets of fields within the current vertex: */ GLuint coloroffset; @@ -237,6 +219,7 @@ struct intel_context GLboolean hw_stipple; GLboolean depth_buffer_is_float; GLboolean no_rast; + GLboolean no_hw; GLboolean always_flush_batch; GLboolean always_flush_cache; @@ -302,13 +285,6 @@ struct intel_context GLboolean use_early_z; drm_clip_rect_t fboRect; /**< cliprect for FBO rendering */ - int perf_boxes; - - GLuint do_usleeps; - int do_irqs; - GLuint irqsEmitted; - - GLboolean scissor; drm_clip_rect_t draw_rect; drm_clip_rect_t scissor_rect; @@ -325,8 +301,6 @@ struct intel_context GLuint lastStamp; - GLboolean no_hw; - /** * Configuration cache */ @@ -374,29 +348,6 @@ do { \ } while (0) /* ================================================================ - * Color packing: - */ - -#define INTEL_PACKCOLOR4444(r,g,b,a) \ - ((((a) & 0xf0) << 8) | (((r) & 0xf0) << 4) | ((g) & 0xf0) | ((b) >> 4)) - -#define INTEL_PACKCOLOR1555(r,g,b,a) \ - ((((r) & 0xf8) << 7) | (((g) & 0xf8) << 2) | (((b) & 0xf8) >> 3) | \ - ((a) ? 0x8000 : 0)) - -#define INTEL_PACKCOLOR565(r,g,b) \ - ((((r) & 0xf8) << 8) | (((g) & 0xfc) << 3) | (((b) & 0xf8) >> 3)) - -#define INTEL_PACKCOLOR8888(r,g,b,a) \ - ((a<<24) | (r<<16) | (g<<8) | b) - -#define INTEL_PACKCOLOR(format, r, g, b, a) \ -(format == DV_PF_555 ? INTEL_PACKCOLOR1555(r,g,b,a) : \ - (format == DV_PF_565 ? INTEL_PACKCOLOR565(r,g,b) : \ - (format == DV_PF_8888 ? INTEL_PACKCOLOR8888(r,g,b,a) : \ - 0))) - -/* ================================================================ * From linux kernel i386 header files, copes with odd sizes better * than COPY_DWORDS would: * XXX Put this in src/mesa/main/imports.h ??? diff --git a/src/mesa/drivers/dri/intel/intel_extensions.c b/src/mesa/drivers/dri/intel/intel_extensions.c index 1682e115cc..f5fe543b5d 100644 --- a/src/mesa/drivers/dri/intel/intel_extensions.c +++ b/src/mesa/drivers/dri/intel/intel_extensions.c @@ -79,6 +79,7 @@ static const struct dri_extension card_extensions[] = { { "GL_ARB_half_float_pixel", NULL }, { "GL_ARB_map_buffer_range", GL_ARB_map_buffer_range_functions }, { "GL_ARB_multitexture", NULL }, + { "GL_ARB_pixel_buffer_object", NULL }, { "GL_ARB_point_parameters", GL_ARB_point_parameters_functions }, { "GL_ARB_point_sprite", NULL }, { "GL_ARB_shader_objects", GL_ARB_shader_objects_functions }, @@ -104,6 +105,8 @@ static const struct dri_extension card_extensions[] = { { "GL_EXT_blend_logic_op", NULL }, { "GL_EXT_blend_subtract", NULL }, { "GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions }, + { "GL_EXT_framebuffer_blit", GL_EXT_framebuffer_blit_functions }, + { "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions }, { "GL_EXT_fog_coord", GL_EXT_fog_coord_functions }, { "GL_EXT_gpu_program_parameters", GL_EXT_gpu_program_parameters_functions }, { "GL_EXT_packed_depth_stencil", NULL }, @@ -176,14 +179,6 @@ static const struct dri_extension arb_oq_extensions[] = { { NULL, NULL } }; - -static const struct dri_extension ttm_extensions[] = { - { "GL_ARB_pixel_buffer_object", NULL }, - { "GL_EXT_framebuffer_blit", GL_EXT_framebuffer_blit_functions }, - { "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions }, - { NULL, NULL } -}; - static const struct dri_extension fragment_shader_extensions[] = { { "GL_ARB_fragment_shader", NULL }, { NULL, NULL } @@ -202,9 +197,6 @@ intelInitExtensions(GLcontext *ctx) */ driInitExtensions(ctx, card_extensions, GL_FALSE); - if (intel->ttm) - driInitExtensions(ctx, ttm_extensions, GL_FALSE); - if (IS_965(intel->intelScreen->deviceID)) driInitExtensions(ctx, brw_extensions, GL_FALSE); diff --git a/src/mesa/drivers/dri/intel/intel_mipmap_tree.c b/src/mesa/drivers/dri/intel/intel_mipmap_tree.c index abb3024bfb..6a565f80cf 100644 --- a/src/mesa/drivers/dri/intel/intel_mipmap_tree.c +++ b/src/mesa/drivers/dri/intel/intel_mipmap_tree.c @@ -224,16 +224,12 @@ int intel_miptree_pitch_align (struct intel_context *intel, if (!mt->compressed) { int pitch_align; - if (intel->ttm) { - /* XXX: Align pitch to multiple of 64 bytes for now to allow - * render-to-texture to work in all cases. This should probably be - * replaced at some point by some scheme to only do this when really - * necessary. - */ - pitch_align = 64; - } else { - pitch_align = 4; - } + /* XXX: Align pitch to multiple of 64 bytes for now to allow + * render-to-texture to work in all cases. This should probably be + * replaced at some point by some scheme to only do this when really + * necessary. + */ + pitch_align = 64; if (tiling == I915_TILING_X) pitch_align = 512; diff --git a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c index 99330b6ddf..9572b67326 100644 --- a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c +++ b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c @@ -228,10 +228,9 @@ do_blit_bitmap( GLcontext *ctx, UNCLAMPED_FLOAT_TO_UBYTE(ubcolor[3], tmpColor[3]); if (dst->cpp == 2) - color = INTEL_PACKCOLOR565(ubcolor[0], ubcolor[1], ubcolor[2]); + color = PACK_COLOR_565(ubcolor[0], ubcolor[1], ubcolor[2]); else - color = INTEL_PACKCOLOR8888(ubcolor[0], ubcolor[1], - ubcolor[2], ubcolor[3]); + color = PACK_COLOR_8888(ubcolor[3], ubcolor[0], ubcolor[1], ubcolor[2]); if (!intel_check_blit_fragment_ops(ctx, tmpColor[3] == 1.0F)) return GL_FALSE; diff --git a/src/mesa/drivers/dri/intel/intel_regions.c b/src/mesa/drivers/dri/intel/intel_regions.c index 80975163d4..d6b9dc4446 100644 --- a/src/mesa/drivers/dri/intel/intel_regions.c +++ b/src/mesa/drivers/dri/intel/intel_regions.c @@ -542,55 +542,18 @@ intel_recreate_static(struct intel_context *intel, region->buffer = NULL; } - if (intel->ttm) { - assert(region_desc->bo_handle != -1); - region->buffer = intel_bo_gem_create_from_name(intel->bufmgr, - name, - region_desc->bo_handle); - - ret = dri_bo_get_tiling(region->buffer, ®ion->tiling, - ®ion->bit_6_swizzle); - if (ret != 0) { - fprintf(stderr, "Couldn't get tiling of buffer %d (%s): %s\n", - region_desc->bo_handle, name, strerror(-ret)); - intel_region_release(®ion); - return NULL; - } - } else { - if (region->classic_map != NULL) { - drmUnmap(region->classic_map, - region->pitch * region->cpp * region->height); - region->classic_map = NULL; - } - ret = drmMap(intel->driFd, region_desc->handle, - region->pitch * region->cpp * region->height, - ®ion->classic_map); - if (ret != 0) { - fprintf(stderr, "Failed to drmMap %s buffer\n", name); - free(region); - return NULL; - } - - region->buffer = intel_bo_fake_alloc_static(intel->bufmgr, + assert(region_desc->bo_handle != -1); + region->buffer = intel_bo_gem_create_from_name(intel->bufmgr, name, - region_desc->offset, - region->pitch * region->cpp * - region->height, - region->classic_map); - - /* The sarea just gives us a boolean for whether it's tiled or not, - * instead of which tiling mode it is. Guess. - */ - if (region_desc->tiled) { - if (intel->gen >= 4 && region_desc == &intelScreen->depth) - region->tiling = I915_TILING_Y; - else - region->tiling = I915_TILING_X; - } else { - region->tiling = I915_TILING_NONE; - } + region_desc->bo_handle); - region->bit_6_swizzle = I915_BIT_6_SWIZZLE_NONE; + ret = dri_bo_get_tiling(region->buffer, ®ion->tiling, + ®ion->bit_6_swizzle); + if (ret != 0) { + fprintf(stderr, "Couldn't get tiling of buffer %d (%s): %s\n", + region_desc->bo_handle, name, strerror(-ret)); + intel_region_release(®ion); + return NULL; } assert(region->buffer != NULL); diff --git a/src/mesa/drivers/dri/intel/intel_screen.c b/src/mesa/drivers/dri/intel/intel_screen.c index 789135b49f..2c5a884a9b 100644 --- a/src/mesa/drivers/dri/intel/intel_screen.c +++ b/src/mesa/drivers/dri/intel/intel_screen.c @@ -605,7 +605,6 @@ intelFillInModes(__DRIscreenPrivate *psp, static GLboolean intel_init_bufmgr(intelScreenPrivate *intelScreen) { - GLboolean gem_disable = getenv("INTEL_NO_GEM") != NULL; int gem_kernel = 0; GLboolean gem_supported; struct drm_i915_getparam gp; @@ -622,43 +621,24 @@ intel_init_bufmgr(intelScreenPrivate *intelScreen) /* If we've got a new enough DDX that's initializing GEM and giving us * object handles for the shared buffers, use that. */ - intelScreen->ttm = GL_FALSE; if (intelScreen->driScrnPriv->dri2.enabled) gem_supported = GL_TRUE; else if (intelScreen->driScrnPriv->ddx_version.minor >= 9 && gem_kernel && intelScreen->front.bo_handle != -1) gem_supported = GL_TRUE; - else - gem_supported = GL_FALSE; - - if (!gem_disable && gem_supported) { - intelScreen->bufmgr = intel_bufmgr_gem_init(spriv->fd, BATCH_SZ); - if (intelScreen->bufmgr != NULL) - intelScreen->ttm = GL_TRUE; + else { + fprintf(stderr, "[%s:%u] Error initializing GEM.\n", + __func__, __LINE__); + return GL_FALSE; } + + intelScreen->bufmgr = intel_bufmgr_gem_init(spriv->fd, BATCH_SZ); /* Otherwise, use the classic buffer manager. */ if (intelScreen->bufmgr == NULL) { - if (gem_disable) { - _mesa_warning(NULL, "GEM disabled. Using classic."); - } else { - _mesa_warning(NULL, - "Failed to initialize GEM. Falling back to classic."); - } - - if (intelScreen->tex.size == 0) { - fprintf(stderr, "[%s:%u] Error initializing buffer manager.\n", - __func__, __LINE__); - return GL_FALSE; - } - - intelScreen->bufmgr = - intel_bufmgr_fake_init(spriv->fd, - intelScreen->tex.offset, - intelScreen->tex.map, - intelScreen->tex.size, - (unsigned int * volatile) - &intelScreen->sarea->last_dispatch); + fprintf(stderr, "[%s:%u] Error initializing buffer manager.\n", + __func__, __LINE__); + return GL_FALSE; } if (intel_get_param(spriv, I915_PARAM_NUM_FENCES_AVAIL, &num_fences)) diff --git a/src/mesa/drivers/dri/intel/intel_screen.h b/src/mesa/drivers/dri/intel/intel_screen.h index a9b9e109a6..14ca0903b6 100644 --- a/src/mesa/drivers/dri/intel/intel_screen.h +++ b/src/mesa/drivers/dri/intel/intel_screen.h @@ -77,7 +77,6 @@ typedef struct GLboolean no_hw; GLboolean no_vbo; - int ttm; dri_bufmgr *bufmgr; GLboolean kernel_exec_fencing; diff --git a/src/mesa/drivers/dri/intel/intel_span.c b/src/mesa/drivers/dri/intel/intel_span.c index 3607c7dded..2c89a66a95 100644 --- a/src/mesa/drivers/dri/intel/intel_span.c +++ b/src/mesa/drivers/dri/intel/intel_span.c @@ -613,15 +613,7 @@ intel_set_span_functions(struct intel_context *intel, struct gl_renderbuffer *rb) { struct intel_renderbuffer *irb = (struct intel_renderbuffer *) rb; - uint32_t tiling; - - /* If in GEM mode, we need to do the tile address swizzling ourselves, - * instead of the fence registers handling it. - */ - if (intel->ttm) - tiling = irb->region->tiling; - else - tiling = I915_TILING_NONE; + uint32_t tiling = irb->region->tiling; if (intel->intelScreen->kernel_exec_fencing) { switch (irb->texformat) { @@ -673,6 +665,9 @@ intel_set_span_functions(struct intel_context *intel, return; } + /* If in GEM mode, we need to do the tile address swizzling ourselves, + * instead of the fence registers handling it. + */ switch (irb->texformat) { case MESA_FORMAT_RGB565: switch (tiling) { diff --git a/src/mesa/drivers/dri/intel/intel_state.c b/src/mesa/drivers/dri/intel/intel_state.c index 4ee742377d..aefae53eb2 100644 --- a/src/mesa/drivers/dri/intel/intel_state.c +++ b/src/mesa/drivers/dri/intel/intel_state.c @@ -196,25 +196,6 @@ intel_translate_logic_op(GLenum opcode) } } - -static void -intelClearColor(GLcontext *ctx, const GLfloat color[4]) -{ - struct intel_context *intel = intel_context(ctx); - GLubyte clear[4]; - - CLAMPED_FLOAT_TO_UBYTE(clear[0], color[0]); - CLAMPED_FLOAT_TO_UBYTE(clear[1], color[1]); - CLAMPED_FLOAT_TO_UBYTE(clear[2], color[2]); - CLAMPED_FLOAT_TO_UBYTE(clear[3], color[3]); - - /* compute both 32 and 16-bit clear values */ - intel->ClearColor8888 = INTEL_PACKCOLOR8888(clear[0], clear[1], - clear[2], clear[3]); - intel->ClearColor565 = INTEL_PACKCOLOR565(clear[0], clear[1], clear[2]); -} - - /* Fallback to swrast for select and feedback. */ static void @@ -229,5 +210,4 @@ void intelInitStateFuncs(struct dd_function_table *functions) { functions->RenderMode = intelRenderMode; - functions->ClearColor = intelClearColor; } diff --git a/src/mesa/drivers/dri/mach64/mach64_native_vb.c b/src/mesa/drivers/dri/mach64/mach64_native_vb.c index 99f1a14e17..816682ec5f 100644 --- a/src/mesa/drivers/dri/mach64/mach64_native_vb.c +++ b/src/mesa/drivers/dri/mach64/mach64_native_vb.c @@ -207,19 +207,19 @@ INTERP_QUALIFIER void TAG(interp_extras)( GLcontext *ctx, LOCALVARS struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; - if (VB->ColorPtr[1]) { - assert(VB->ColorPtr[1]->stride == 4 * sizeof(GLfloat)); + if (VB->BackfaceColorPtr) { + assert(VB->BackfaceColorPtr->stride == 4 * sizeof(GLfloat)); INTERP_4F( t, - GET_COLOR(VB->ColorPtr[1], dst), - GET_COLOR(VB->ColorPtr[1], out), - GET_COLOR(VB->ColorPtr[1], in) ); + GET_COLOR(VB->BackfaceColorPtr, dst), + GET_COLOR(VB->BackfaceColorPtr, out), + GET_COLOR(VB->BackfaceColorPtr, in) ); - if (VB->SecondaryColorPtr[1]) { + if (VB->BackfaceSecondaryColorPtr) { INTERP_3F( t, - GET_COLOR(VB->SecondaryColorPtr[1], dst), - GET_COLOR(VB->SecondaryColorPtr[1], out), - GET_COLOR(VB->SecondaryColorPtr[1], in) ); + GET_COLOR(VB->BackfaceSecondaryColorPtr, dst), + GET_COLOR(VB->BackfaceSecondaryColorPtr, out), + GET_COLOR(VB->BackfaceSecondaryColorPtr, in) ); } } @@ -236,13 +236,13 @@ INTERP_QUALIFIER void TAG(copy_pv_extras)( GLcontext *ctx, LOCALVARS struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; - if (VB->ColorPtr[1]) { - COPY_4FV( GET_COLOR(VB->ColorPtr[1], dst), - GET_COLOR(VB->ColorPtr[1], src) ); + if (VB->BackfaceColorPtr) { + COPY_4FV( GET_COLOR(VB->BackfaceColorPtr, dst), + GET_COLOR(VB->BackfaceColorPtr, src) ); - if (VB->SecondaryColorPtr[1]) { - COPY_4FV( GET_COLOR(VB->SecondaryColorPtr[1], dst), - GET_COLOR(VB->SecondaryColorPtr[1], src) ); + if (VB->BackfaceSecondaryColorPtr) { + COPY_4FV( GET_COLOR(VB->BackfaceSecondaryColorPtr, dst), + GET_COLOR(VB->BackfaceSecondaryColorPtr, src) ); } } diff --git a/src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h b/src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h index 684f2acc89..6e5fa3520e 100644 --- a/src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h +++ b/src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h @@ -103,10 +103,10 @@ static void TAG(emit)( GLcontext *ctx, #if DO_TEX1 { const GLuint t1 = GET_TEXSOURCE(1); - tc1 = VB->TexCoordPtr[t1]->data; - tc1_stride = VB->TexCoordPtr[t1]->stride; + tc1 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->data; + tc1_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->stride; #if DO_PTEX - tc1_size = VB->TexCoordPtr[t1]->size; + tc1_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->size; #endif } #endif @@ -114,18 +114,18 @@ static void TAG(emit)( GLcontext *ctx, #if DO_TEX0 { const GLuint t0 = GET_TEXSOURCE(0); - tc0 = VB->TexCoordPtr[t0]->data; - tc0_stride = VB->TexCoordPtr[t0]->stride; + tc0 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->data; + tc0_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->stride; #if DO_PTEX - tc0_size = VB->TexCoordPtr[t0]->size; + tc0_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->size; #endif } #endif #if DO_SPEC - if (VB->SecondaryColorPtr[0]) { - spec = VB->SecondaryColorPtr[0]->data; - spec_stride = VB->SecondaryColorPtr[0]->stride; + if (VB->AttribPtr[_TNL_ATTRIB_COLOR1]) { + spec = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->data; + spec_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->stride; } else { spec = (GLfloat (*)[4])ctx->Current.Attrib[VERT_ATTRIB_COLOR1]; spec_stride = 0; @@ -133,9 +133,9 @@ static void TAG(emit)( GLcontext *ctx, #endif #if DO_FOG - if (VB->FogCoordPtr) { - fog = VB->FogCoordPtr->data; - fog_stride = VB->FogCoordPtr->stride; + if (VB->AttribPtr[_TNL_ATTRIB_FOG]) { + fog = VB->AttribPtr[_TNL_ATTRIB_FOG]->data; + fog_stride = VB->AttribPtr[_TNL_ATTRIB_FOG]->stride; } else { static GLfloat tmp[4] = {0, 0, 0, 0}; fog = &tmp; @@ -144,8 +144,8 @@ static void TAG(emit)( GLcontext *ctx, #endif #if DO_RGBA - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; #endif coord = VB->NdcPtr->data; @@ -319,8 +319,8 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) /* Force 'missing' texcoords to something valid. */ - if (DO_TEX1 && VB->TexCoordPtr[0] == 0) - VB->TexCoordPtr[0] = VB->TexCoordPtr[1]; + if (DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0] = VB->AttribPtr[_TNL_ATTRIB_TEX1]; if (DO_PTEX) return GL_TRUE; @@ -328,12 +328,12 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) /* No hardware support for projective texture. Can fake it for * TEX0 only. */ - if ((DO_TEX1 && VB->TexCoordPtr[GET_TEXSOURCE(1)]->size == 4)) { + if ((DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(1)]->size == 4)) { PTEX_FALLBACK(); return GL_FALSE; } - if (DO_TEX0 && VB->TexCoordPtr[GET_TEXSOURCE(0)]->size == 4) { + if (DO_TEX0 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(0)]->size == 4) { if (DO_TEX1) { PTEX_FALLBACK(); } diff --git a/src/mesa/drivers/dri/mach64/mach64_vbtmp.h b/src/mesa/drivers/dri/mach64/mach64_vbtmp.h index 938804af9e..60bfab8f6d 100644 --- a/src/mesa/drivers/dri/mach64/mach64_vbtmp.h +++ b/src/mesa/drivers/dri/mach64/mach64_vbtmp.h @@ -156,53 +156,53 @@ static void TAG(emit)( GLcontext *ctx, if (DO_TEX3) { const GLuint t3 = GET_TEXSOURCE(3); - tc3 = VB->TexCoordPtr[t3]->data; - tc3_stride = VB->TexCoordPtr[t3]->stride; + tc3 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t3]->data; + tc3_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t3]->stride; if (DO_PTEX) - tc3_size = VB->TexCoordPtr[t3]->size; + tc3_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t3]->size; } if (DO_TEX2) { const GLuint t2 = GET_TEXSOURCE(2); - tc2 = VB->TexCoordPtr[t2]->data; - tc2_stride = VB->TexCoordPtr[t2]->stride; + tc2 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->data; + tc2_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->stride; if (DO_PTEX) - tc2_size = VB->TexCoordPtr[t2]->size; + tc2_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->size; } if (DO_TEX1) { const GLuint t1 = GET_TEXSOURCE(1); - tc1 = VB->TexCoordPtr[t1]->data; - tc1_stride = VB->TexCoordPtr[t1]->stride; + tc1 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->data; + tc1_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->stride; if (DO_PTEX) - tc1_size = VB->TexCoordPtr[t1]->size; + tc1_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->size; } if (DO_TEX0) { const GLuint t0 = GET_TEXSOURCE(0); - tc0_stride = VB->TexCoordPtr[t0]->stride; - tc0 = VB->TexCoordPtr[t0]->data; + tc0_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->stride; + tc0 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->data; if (DO_PTEX) - tc0_size = VB->TexCoordPtr[t0]->size; + tc0_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->size; } if (DO_RGBA) { - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; } if (DO_SPEC) { - spec = VB->SecondaryColorPtr[0]->data; - spec_stride = VB->SecondaryColorPtr[0]->stride; + spec = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->data; + spec_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->stride; } else { spec = (GLfloat (*)[4])ctx->Current.Attrib[VERT_ATTRIB_COLOR1]; spec_stride = 0; } if (DO_FOG) { - if (VB->FogCoordPtr) { - fog = VB->FogCoordPtr->data; - fog_stride = VB->FogCoordPtr->stride; + if (VB->AttribPtr[_TNL_ATTRIB_FOG]) { + fog = VB->AttribPtr[_TNL_ATTRIB_FOG]->data; + fog_stride = VB->AttribPtr[_TNL_ATTRIB_FOG]->stride; } else { static GLfloat tmp[4] = {0, 0, 0, 0}; fog = &tmp; @@ -384,8 +384,8 @@ static void TAG(emit)( GLcontext *ctx, GLuint start, GLuint end, ASSERT(stride == 4); - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; /* Pack what's left into a 4-dword vertex. Color is in a different * place, and there is no 'w' coordinate. @@ -432,8 +432,8 @@ static void TAG(emit)( GLcontext *ctx, GLuint start, GLuint end, GLfloat *v = (GLfloat *)dest; int i; - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; if (start) STRIDE_4F(col, col_stride * start); @@ -473,22 +473,22 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) /* Force 'missing' texcoords to something valid. */ - if (DO_TEX3 && VB->TexCoordPtr[2] == 0) - VB->TexCoordPtr[2] = VB->TexCoordPtr[3]; + if (DO_TEX3 && VB->AttribPtr[_TNL_ATTRIB_TEX2] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX2] = VB->AttribPtr[_TNL_ATTRIB_TEX3]; - if (DO_TEX2 && VB->TexCoordPtr[1] == 0) - VB->TexCoordPtr[1] = VB->TexCoordPtr[2]; + if (DO_TEX2 && VB->AttribPtr[_TNL_ATTRIB_TEX1] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX1] = VB->AttribPtr[_TNL_ATTRIB_TEX2]; - if (DO_TEX1 && VB->TexCoordPtr[0] == 0) - VB->TexCoordPtr[0] = VB->TexCoordPtr[1]; + if (DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0] = VB->AttribPtr[_TNL_ATTRIB_TEX1]; if (DO_PTEX) return GL_TRUE; - if ((DO_TEX3 && VB->TexCoordPtr[GET_TEXSOURCE(3)]->size == 4) || - (DO_TEX2 && VB->TexCoordPtr[GET_TEXSOURCE(2)]->size == 4) || - (DO_TEX1 && VB->TexCoordPtr[GET_TEXSOURCE(1)]->size == 4) || - (DO_TEX0 && VB->TexCoordPtr[GET_TEXSOURCE(0)]->size == 4)) + if ((DO_TEX3 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(3)]->size == 4) || + (DO_TEX2 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(2)]->size == 4) || + (DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(1)]->size == 4) || + (DO_TEX0 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(0)]->size == 4)) return GL_FALSE; return GL_TRUE; @@ -501,14 +501,14 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) /* Force 'missing' texcoords to something valid. */ - if (DO_TEX3 && VB->TexCoordPtr[2] == 0) - VB->TexCoordPtr[2] = VB->TexCoordPtr[3]; + if (DO_TEX3 && VB->AttribPtr[_TNL_ATTRIB_TEX2] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX2] = VB->AttribPtr[_TNL_ATTRIB_TEX3]; - if (DO_TEX2 && VB->TexCoordPtr[1] == 0) - VB->TexCoordPtr[1] = VB->TexCoordPtr[2]; + if (DO_TEX2 && VB->AttribPtr[_TNL_ATTRIB_TEX1] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX1] = VB->AttribPtr[_TNL_ATTRIB_TEX2]; - if (DO_TEX1 && VB->TexCoordPtr[0] == 0) - VB->TexCoordPtr[0] = VB->TexCoordPtr[1]; + if (DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0] = VB->AttribPtr[_TNL_ATTRIB_TEX1]; if (DO_PTEX) return GL_TRUE; @@ -516,14 +516,14 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) /* No hardware support for projective texture. Can fake it for * TEX0 only. */ - if ((DO_TEX3 && VB->TexCoordPtr[GET_TEXSOURCE(3)]->size == 4) || - (DO_TEX2 && VB->TexCoordPtr[GET_TEXSOURCE(2)]->size == 4) || - (DO_TEX1 && VB->TexCoordPtr[GET_TEXSOURCE(1)]->size == 4)) { + if ((DO_TEX3 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(3)]->size == 4) || + (DO_TEX2 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(2)]->size == 4) || + (DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(1)]->size == 4)) { PTEX_FALLBACK(); return GL_FALSE; } - if (DO_TEX0 && VB->TexCoordPtr[GET_TEXSOURCE(0)]->size == 4) { + if (DO_TEX0 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(0)]->size == 4) { if (DO_TEX1 || DO_TEX2 || DO_TEX3) { PTEX_FALLBACK(); } diff --git a/src/mesa/drivers/dri/r128/r128_tris.c b/src/mesa/drivers/dri/r128/r128_tris.c index 5b91271d74..448e34e047 100644 --- a/src/mesa/drivers/dri/r128/r128_tris.c +++ b/src/mesa/drivers/dri/r128/r128_tris.c @@ -650,12 +650,12 @@ static void r128RenderStart( GLcontext *ctx ) } if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX(rmesa->tmu_source[0]) )) { - if ( VB->TexCoordPtr[rmesa->tmu_source[0]]->size > 2 ) + if ( VB->AttribPtr[_TNL_ATTRIB_TEX0 + rmesa->tmu_source[0]]->size > 2 ) fallback_projtex = GL_TRUE; EMIT_ATTR( _TNL_ATTRIB_TEX0, EMIT_2F, R128_CCE_VC_FRMT_S_T, 8 ); } if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX(rmesa->tmu_source[1]) )) { - if ( VB->TexCoordPtr[rmesa->tmu_source[1]]->size > 2 ) + if ( VB->AttribPtr[_TNL_ATTRIB_TEX0 + rmesa->tmu_source[1]]->size > 2 ) fallback_projtex = GL_TRUE; EMIT_ATTR( _TNL_ATTRIB_TEX1, EMIT_2F, R128_CCE_VC_FRMT_S2_T2, 8 ); } diff --git a/src/mesa/drivers/dri/r200/r200_swtcl.c b/src/mesa/drivers/dri/r200/r200_swtcl.c index 240fb45078..fadc766b49 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 ); diff --git a/src/mesa/drivers/dri/r300/r300_swtcl.c b/src/mesa/drivers/dri/r300/r300_swtcl.c index ee2c71e1a7..99bd22edac 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; diff --git a/src/mesa/drivers/dri/r600/r600_context.c b/src/mesa/drivers/dri/r600/r600_context.c index dbd233729c..97b25ea3ff 100644 --- a/src/mesa/drivers/dri/r600/r600_context.c +++ b/src/mesa/drivers/dri/r600/r600_context.c @@ -72,7 +72,9 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "vblank.h" #include "utils.h" -#include "xmlpool.h" /* for symbolic values of enum-type options */ +#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 @@ -154,8 +156,12 @@ static const struct dri_extension mm_extensions[] = { * The GL 2.0 functions are needed to make display lists work with * functions added by GL_ATI_separate_stencil. */ -static const struct dri_extension gl_20_extension[] = { - {"GL_VERSION_2_0", GL_VERSION_2_0_functions }, +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[] = { @@ -306,7 +312,28 @@ static void r600InitGLExtensions(GLcontext *ctx) driInitExtensions(ctx, card_extensions, GL_TRUE); if (r600->radeon.radeonScreen->kernel_mm) - driInitExtensions(ctx, mm_extensions, GL_FALSE); + driInitExtensions(ctx, mm_extensions, GL_FALSE);
+
+#ifdef R600_ENABLE_GLSL_TEST
+ driInitExtensions(ctx, gl_20_extension, GL_TRUE);
+ //_mesa_enable_2_0_extensions(ctx);
+ //1.5
+ ctx->Extensions.ARB_occlusion_query = GL_TRUE;
+ ctx->Extensions.ARB_vertex_buffer_object = GL_TRUE;
+ ctx->Extensions.EXT_shadow_funcs = GL_TRUE;
+ //2.0
+ ctx->Extensions.ARB_draw_buffers = GL_TRUE;
+ ctx->Extensions.ARB_point_sprite = GL_TRUE;
+ ctx->Extensions.ARB_shader_objects = GL_TRUE;
+ ctx->Extensions.ARB_vertex_shader = GL_TRUE;
+ ctx->Extensions.ARB_fragment_shader = GL_TRUE;
+ ctx->Extensions.ARB_texture_non_power_of_two = GL_TRUE;
+ ctx->Extensions.EXT_blend_equation_separate = GL_TRUE;
+ ctx->Extensions.ATI_separate_stencil = GL_TRUE;
+
+ /* glsl compiler has problem if this is not GL_TRUE */
+ ctx->Shader.EmitCondCodes = GL_TRUE;
+#endif /* R600_ENABLE_GLSL_TEST */ if (driQueryOptionb (&r600->radeon.optionCache, "disable_stencil_two_side")) @@ -341,7 +368,7 @@ GLboolean r600CreateContext(const __GLcontextModes * glVisual, assert(glVisual); assert(driContextPriv); - assert(screen); + assert(screen);
/* Allocate the R600 context */ r600 = (context_t*) CALLOC(sizeof(*r600)); diff --git a/src/mesa/drivers/dri/r600/r700_assembler.c b/src/mesa/drivers/dri/r600/r700_assembler.c index e0d7d4fa6b..e3bc46802f 100644 --- a/src/mesa/drivers/dri/r600/r700_assembler.c +++ b/src/mesa/drivers/dri/r600/r700_assembler.c @@ -38,6 +38,9 @@ #include "r700_assembler.h" +#define USE_CF_FOR_CONTINUE_BREAK 1 +//#define USE_CF_FOR_POP_AFTER 1 + BITS addrmode_PVSDST(PVSDST * pPVSDST) { return pPVSDST->addrmode0 | ((BITS)pPVSDST->addrmode1 << 1); @@ -343,6 +346,8 @@ unsigned int r700GetNumOperands(r700_AssemblerBase* pAsm) 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: @@ -398,6 +403,9 @@ int Init_r700_AssemblerBase(SHADER_PIPE_TYPE spt, r700_AssemblerBase* pAsm, R700 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; @@ -474,6 +482,24 @@ int Init_r700_AssemblerBase(SHADER_PIPE_TYPE spt, r700_AssemblerBase* pAsm, R700 pAsm->is_tex = GL_FALSE; pAsm->need_tex_barrier = GL_FALSE; + 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->CALLSTACK[0].stackUsage.bits = 0; + + SetActiveCFlist(pAsm->pR700Shader, pAsm->CALLSTACK[0].plstCFInstructions_local); + + pAsm->unCFflags = 0; + return 0; } @@ -592,6 +618,31 @@ int check_current_clause(r700_AssemblerBase* pAsm, return GL_TRUE; } +GLboolean add_cf_instruction(r700_AssemblerBase* pAsm) +{ + if(GL_FALSE == check_current_clause(pAsm, CF_OTHER_CLAUSE)) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr = + (R700ControlFlowGenericClause*) CALLOC_STRUCT(R700ControlFlowGenericClause); + + if (pAsm->cf_current_cf_clause_ptr != NULL) + { + Init_R700ControlFlowGenericClause(pAsm->cf_current_cf_clause_ptr); + AddCFInstruction( pAsm->pR700Shader, + (R700ControlFlowInstruction *)pAsm->cf_current_cf_clause_ptr ); + } + else + { + radeon_error("Could not allocate a new VFetch CF instruction.\n"); + return GL_FALSE; + } + + return GL_TRUE; +} + GLboolean add_vfetch_instruction(r700_AssemblerBase* pAsm, R700VertexInstruction* vertex_instruction_ptr) { @@ -1153,6 +1204,7 @@ GLboolean assemble_src(r700_AssemblerBase *pAsm, case PROGRAM_LOCAL_PARAM: case PROGRAM_ENV_PARAM: case PROGRAM_STATE_VAR: + case PROGRAM_UNIFORM: if (1 == pILInst->SrcReg[src].RelAddr) { setaddrmode_PVSSRC(&(pAsm->S[fld].src), ADDR_RELATIVE_A0); @@ -1179,7 +1231,7 @@ GLboolean assemble_src(r700_AssemblerBase *pAsm, } break; default: - radeon_error("Invalid source argument type\n"); + radeon_error("Invalid source argument type : %d \n", pILInst->SrcReg[src].File); return GL_FALSE; } } @@ -1315,7 +1367,7 @@ GLboolean tex_src(r700_AssemblerBase *pAsm) case FRAG_ATTRIB_TEX0: case FRAG_ATTRIB_TEX1: case FRAG_ATTRIB_TEX2: - case FRAG_ATTRIB_TEX3: + case FRAG_ATTRIB_TEX3: case FRAG_ATTRIB_TEX4: case FRAG_ATTRIB_TEX5: case FRAG_ATTRIB_TEX6: @@ -1335,6 +1387,16 @@ GLboolean tex_src(r700_AssemblerBase *pAsm) fprintf(stderr, "FRAG_ATTRIB_VAR0 unsupported\n"); break; } + + if( (pILInst->SrcReg[0].Index >= FRAG_ATTRIB_VAR0) || + (pILInst->SrcReg[0].Index < FRAG_ATTRIB_MAX) ) + { + bValidTexCoord = GL_TRUE; + pAsm->S[0].src.reg = + pAsm->uiFP_AttributeMap[pILInst->SrcReg[0].Index]; + pAsm->S[0].src.rtype = SRC_REG_INPUT; + } + break; } } @@ -1517,6 +1579,10 @@ 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 +1672,8 @@ 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 +1703,17 @@ GLboolean add_alu_instruction(r700_AssemblerBase* pAsm, pAsm->cf_current_alu_clause_ptr->m_Word1.f.kcache_addr0 = 0x0; pAsm->cf_current_alu_clause_ptr->m_Word1.f.kcache_addr1 = 0x0; - //cf_current_alu_clause_ptr->m_Word1.f.count = number_of_scalar_operations - 1; pAsm->cf_current_alu_clause_ptr->m_Word1.f.count = 0x0; - pAsm->cf_current_alu_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_ALU; + + if(pAsm->alu_x_opcode != 0) + { + pAsm->cf_current_alu_clause_ptr->m_Word1.f.cf_inst = pAsm->alu_x_opcode; + pAsm->alu_x_opcode = 0; + } + else + { + pAsm->cf_current_alu_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_ALU; + } pAsm->cf_current_alu_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; @@ -2358,6 +2433,506 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) return GL_TRUE; } +GLboolean assemble_alu_instruction2(r700_AssemblerBase *pAsm) +{ + GLuint number_of_scalar_operations; + GLboolean is_single_scalar_operation; + GLuint scalar_channel_index; + + PVSSRC * pcurrent_source; + int current_source_index; + GLuint contiguous_slots_needed; + + GLuint uNumSrc = r700GetNumOperands(pAsm); + + GLboolean bSplitInst = GL_FALSE; + + if (1 == pAsm->D.dst.math) + { + is_single_scalar_operation = GL_TRUE; + number_of_scalar_operations = 1; + } + else + { + is_single_scalar_operation = GL_FALSE; + number_of_scalar_operations = 4; + } + + contiguous_slots_needed = 0; + + if(GL_TRUE == is_reduction_opcode(&(pAsm->D)) ) + { + contiguous_slots_needed = 4; + } + + initialize(pAsm); + + for (scalar_channel_index=0; + scalar_channel_index < number_of_scalar_operations; + scalar_channel_index++) + { + R700ALUInstruction* alu_instruction_ptr = (R700ALUInstruction*) CALLOC_STRUCT(R700ALUInstruction); + if (alu_instruction_ptr == NULL) + { + return GL_FALSE; + } + Init_R700ALUInstruction(alu_instruction_ptr); + + //src 0 + current_source_index = 0; + pcurrent_source = &(pAsm->S[0].src); + + if (GL_FALSE == assemble_alu_src(alu_instruction_ptr, + current_source_index, + pcurrent_source, + scalar_channel_index) ) + { + return GL_FALSE; + } + + if (uNumSrc > 1) + { + // Process source 1 + current_source_index = 1; + pcurrent_source = &(pAsm->S[current_source_index].src); + + if (GL_FALSE == assemble_alu_src(alu_instruction_ptr, + current_source_index, + pcurrent_source, + scalar_channel_index) ) + { + return GL_FALSE; + } + } + + //other bits + alu_instruction_ptr->m_Word0.f.index_mode = SQ_INDEX_LOOP; + + if( (is_single_scalar_operation == GL_TRUE) + || (GL_TRUE == bSplitInst) ) + { + alu_instruction_ptr->m_Word0.f.last = 1; + } + else + { + alu_instruction_ptr->m_Word0.f.last = (scalar_channel_index == 3) ? 1 : 0; + } + + alu_instruction_ptr->m_Word0.f.pred_sel = (pAsm->D.dst.pred_inv > 0) ? 1 : 0; + if(1 == pAsm->D.dst.predicated) + { + alu_instruction_ptr->m_Word1_OP2.f.update_pred = 0x1; + alu_instruction_ptr->m_Word1_OP2.f.update_execute_mask = 0x1; + } + else + { + alu_instruction_ptr->m_Word1_OP2.f.update_pred = 0x0; + alu_instruction_ptr->m_Word1_OP2.f.update_execute_mask = 0x0; + } + + // dst + if( (pAsm->D.dst.rtype == DST_REG_TEMPORARY) || + (pAsm->D.dst.rtype == DST_REG_OUT) ) + { + alu_instruction_ptr->m_Word1.f.dst_gpr = pAsm->D.dst.reg; + } + else + { + radeon_error("Only temp destination registers supported for ALU dest regs.\n"); + return GL_FALSE; + } + + alu_instruction_ptr->m_Word1.f.dst_rel = SQ_ABSOLUTE; //D.rtype + + if ( is_single_scalar_operation == GL_TRUE ) + { + // Override scalar_channel_index since only one scalar value will be written + if(pAsm->D.dst.writex) + { + scalar_channel_index = 0; + } + else if(pAsm->D.dst.writey) + { + scalar_channel_index = 1; + } + else if(pAsm->D.dst.writez) + { + scalar_channel_index = 2; + } + else if(pAsm->D.dst.writew) + { + scalar_channel_index = 3; + } + } + + alu_instruction_ptr->m_Word1.f.dst_chan = scalar_channel_index; + + alu_instruction_ptr->m_Word1.f.clamp = pAsm->D2.dst2.SaturateMode; + + if (pAsm->D.dst.op3) + { + //op3 + + alu_instruction_ptr->m_Word1_OP3.f.alu_inst = pAsm->D.dst.opcode; + + //There's 3rd src for op3 + current_source_index = 2; + pcurrent_source = &(pAsm->S[current_source_index].src); + + if ( GL_FALSE == assemble_alu_src(alu_instruction_ptr, + current_source_index, + pcurrent_source, + scalar_channel_index) ) + { + return GL_FALSE; + } + } + else + { + //op2 + if (pAsm->bR6xx) + { + alu_instruction_ptr->m_Word1_OP2.f6.alu_inst = pAsm->D.dst.opcode; + + alu_instruction_ptr->m_Word1_OP2.f6.src0_abs = 0x0; + alu_instruction_ptr->m_Word1_OP2.f6.src1_abs = 0x0; + + //alu_instruction_ptr->m_Word1_OP2.f6.update_execute_mask = 0x0; + //alu_instruction_ptr->m_Word1_OP2.f6.update_pred = 0x0; + switch (scalar_channel_index) + { + case 0: + alu_instruction_ptr->m_Word1_OP2.f6.write_mask = pAsm->D.dst.writex; + break; + case 1: + alu_instruction_ptr->m_Word1_OP2.f6.write_mask = pAsm->D.dst.writey; + break; + case 2: + alu_instruction_ptr->m_Word1_OP2.f6.write_mask = pAsm->D.dst.writez; + break; + case 3: + alu_instruction_ptr->m_Word1_OP2.f6.write_mask = pAsm->D.dst.writew; + break; + default: + alu_instruction_ptr->m_Word1_OP2.f6.write_mask = 1; //SQ_SEL_MASK; + break; + } + alu_instruction_ptr->m_Word1_OP2.f6.omod = SQ_ALU_OMOD_OFF; + } + else + { + alu_instruction_ptr->m_Word1_OP2.f.alu_inst = pAsm->D.dst.opcode; + + alu_instruction_ptr->m_Word1_OP2.f.src0_abs = 0x0; + alu_instruction_ptr->m_Word1_OP2.f.src1_abs = 0x0; + + //alu_instruction_ptr->m_Word1_OP2.f.update_execute_mask = 0x0; + //alu_instruction_ptr->m_Word1_OP2.f.update_pred = 0x0; + switch (scalar_channel_index) + { + case 0: + alu_instruction_ptr->m_Word1_OP2.f.write_mask = pAsm->D.dst.writex; + break; + case 1: + alu_instruction_ptr->m_Word1_OP2.f.write_mask = pAsm->D.dst.writey; + break; + case 2: + alu_instruction_ptr->m_Word1_OP2.f.write_mask = pAsm->D.dst.writez; + break; + case 3: + alu_instruction_ptr->m_Word1_OP2.f.write_mask = pAsm->D.dst.writew; + break; + default: + alu_instruction_ptr->m_Word1_OP2.f.write_mask = 1; //SQ_SEL_MASK; + break; + } + alu_instruction_ptr->m_Word1_OP2.f.omod = SQ_ALU_OMOD_OFF; + } + } + + if(GL_FALSE == add_alu_instruction(pAsm, alu_instruction_ptr, contiguous_slots_needed) ) + { + return GL_FALSE; + } + + /* + * Judge the type of current instruction, is it vector or scalar + * instruction. + */ + if (is_single_scalar_operation) + { + if(GL_FALSE == check_scalar(pAsm, alu_instruction_ptr) ) + { + return GL_FALSE; + } + } + else + { + if(GL_FALSE == check_vector(pAsm, alu_instruction_ptr) ) + { + return 1; + } + } + + contiguous_slots_needed = 0; + } + + return GL_TRUE; +} + +GLboolean assemble_alu_instruction_literal(r700_AssemblerBase *pAsm, GLfloat * pLiteral) +{ + R700ALUInstruction * alu_instruction_ptr; + R700ALUInstructionHalfLiteral * alu_instruction_ptr_hl; + R700ALUInstructionFullLiteral * alu_instruction_ptr_fl; + + GLuint number_of_scalar_operations; + GLboolean is_single_scalar_operation; + GLuint scalar_channel_index; + + GLuint contiguous_slots_needed; + GLuint lastInstruction; + GLuint not_masked[4]; + + GLuint uNumSrc = r700GetNumOperands(pAsm); + + GLboolean bSplitInst = GL_FALSE; + + number_of_scalar_operations = 0; + contiguous_slots_needed = 0; + + if(1 == pAsm->D.dst.writew) + { + lastInstruction = 3; + number_of_scalar_operations++; + not_masked[3] = 1; + } + else + { + not_masked[3] = 0; + } + if(1 == pAsm->D.dst.writez) + { + lastInstruction = 2; + number_of_scalar_operations++; + not_masked[2] = 1; + } + else + { + not_masked[2] = 0; + } + if(1 == pAsm->D.dst.writey) + { + lastInstruction = 1; + number_of_scalar_operations++; + not_masked[1] = 1; + } + else + { + not_masked[1] = 0; + } + if(1 == pAsm->D.dst.writex) + { + lastInstruction = 0; + number_of_scalar_operations++; + not_masked[0] = 1; + } + else + { + not_masked[0] = 0; + } + + if(GL_TRUE == is_reduction_opcode(&(pAsm->D)) ) + { + contiguous_slots_needed = 4; + } + else + { + contiguous_slots_needed = number_of_scalar_operations; + } + + if(1 == pAsm->D2.dst2.literal) + { + contiguous_slots_needed += 1; + } + else if(2 == pAsm->D2.dst2.literal) + { + contiguous_slots_needed += 2; + } + + initialize(pAsm); + + for (scalar_channel_index=0; scalar_channel_index < 4; scalar_channel_index++) + { + if(0 == not_masked[scalar_channel_index]) + { + continue; + } + + if(scalar_channel_index == lastInstruction) + { + switch (pAsm->D2.dst2.literal) + { + case 0: + alu_instruction_ptr = (R700ALUInstruction*) CALLOC_STRUCT(R700ALUInstruction); + if (alu_instruction_ptr == NULL) + { + return GL_FALSE; + } + Init_R700ALUInstruction(alu_instruction_ptr); + break; + case 1: + alu_instruction_ptr_hl = (R700ALUInstructionHalfLiteral*) CALLOC_STRUCT(R700ALUInstructionHalfLiteral); + if (alu_instruction_ptr_hl == NULL) + { + return GL_FALSE; + } + Init_R700ALUInstructionHalfLiteral(alu_instruction_ptr_hl, pLiteral[0], pLiteral[1]); + alu_instruction_ptr = (R700ALUInstruction*)alu_instruction_ptr_hl; + break; + case 2: + alu_instruction_ptr_fl = (R700ALUInstructionFullLiteral*) CALLOC_STRUCT(R700ALUInstructionFullLiteral); + if (alu_instruction_ptr_fl == NULL) + { + return GL_FALSE; + } + Init_R700ALUInstructionFullLiteral(alu_instruction_ptr_fl, pLiteral[0], pLiteral[1], pLiteral[2], pLiteral[3]); + alu_instruction_ptr = (R700ALUInstruction*)alu_instruction_ptr_fl; + break; + default: + break; + }; + } + else + { + alu_instruction_ptr = (R700ALUInstruction*) CALLOC_STRUCT(R700ALUInstruction); + if (alu_instruction_ptr == NULL) + { + return GL_FALSE; + } + Init_R700ALUInstruction(alu_instruction_ptr); + } + + //src 0 + if (GL_FALSE == assemble_alu_src(alu_instruction_ptr, + 0, + &(pAsm->S[0].src), + scalar_channel_index) ) + { + return GL_FALSE; + } + + if (uNumSrc > 1) + { + // Process source 1 + if (GL_FALSE == assemble_alu_src(alu_instruction_ptr, + 1, + &(pAsm->S[1].src), + scalar_channel_index) ) + { + return GL_FALSE; + } + } + + //other bits + alu_instruction_ptr->m_Word0.f.index_mode = SQ_INDEX_LOOP; + + if(scalar_channel_index == lastInstruction) + { + alu_instruction_ptr->m_Word0.f.last = 1; + } + + alu_instruction_ptr->m_Word0.f.pred_sel = 0x0; + if(1 == pAsm->D.dst.predicated) + { + alu_instruction_ptr->m_Word1_OP2.f.update_pred = 0x1; + alu_instruction_ptr->m_Word1_OP2.f.update_execute_mask = 0x1; + } + else + { + alu_instruction_ptr->m_Word1_OP2.f.update_pred = 0; + alu_instruction_ptr->m_Word1_OP2.f.update_execute_mask = 0; + } + + // dst + if( (pAsm->D.dst.rtype == DST_REG_TEMPORARY) || + (pAsm->D.dst.rtype == DST_REG_OUT) ) + { + alu_instruction_ptr->m_Word1.f.dst_gpr = pAsm->D.dst.reg; + } + else + { + radeon_error("Only temp destination registers supported for ALU dest regs.\n"); + return GL_FALSE; + } + + alu_instruction_ptr->m_Word1.f.dst_rel = SQ_ABSOLUTE; //D.rtype + + alu_instruction_ptr->m_Word1.f.dst_chan = scalar_channel_index; + + alu_instruction_ptr->m_Word1.f.clamp = pAsm->D2.dst2.SaturateMode; + + if (pAsm->D.dst.op3) + { + //op3 + alu_instruction_ptr->m_Word1_OP3.f.alu_inst = pAsm->D.dst.opcode; + + //There's 3rd src for op3 + if ( GL_FALSE == assemble_alu_src(alu_instruction_ptr, + 2, + &(pAsm->S[2].src), + scalar_channel_index) ) + { + return GL_FALSE; + } + } + else + { + //op2 + if (pAsm->bR6xx) + { + alu_instruction_ptr->m_Word1_OP2.f6.alu_inst = pAsm->D.dst.opcode; + alu_instruction_ptr->m_Word1_OP2.f6.src0_abs = 0x0; + alu_instruction_ptr->m_Word1_OP2.f6.src1_abs = 0x0; + alu_instruction_ptr->m_Word1_OP2.f6.write_mask = 1; + alu_instruction_ptr->m_Word1_OP2.f6.omod = SQ_ALU_OMOD_OFF; + } + else + { + alu_instruction_ptr->m_Word1_OP2.f.alu_inst = pAsm->D.dst.opcode; + alu_instruction_ptr->m_Word1_OP2.f.src0_abs = 0x0; + alu_instruction_ptr->m_Word1_OP2.f.src1_abs = 0x0; + alu_instruction_ptr->m_Word1_OP2.f.write_mask = 1; + alu_instruction_ptr->m_Word1_OP2.f.omod = SQ_ALU_OMOD_OFF; + } + } + + if(GL_FALSE == add_alu_instruction(pAsm, alu_instruction_ptr, contiguous_slots_needed) ) + { + return GL_FALSE; + } + + if (1 == number_of_scalar_operations) + { + if(GL_FALSE == check_scalar(pAsm, alu_instruction_ptr) ) + { + return GL_FALSE; + } + } + else + { + if(GL_FALSE == check_vector(pAsm, alu_instruction_ptr) ) + { + return GL_FALSE; + } + } + + contiguous_slots_needed -= 2; + } + + return GL_TRUE; +} + GLboolean next_ins(r700_AssemblerBase *pAsm) { struct prog_instruction *pILInst = &(pAsm->pILInst[pAsm->uiCurInst]); @@ -2403,6 +2978,71 @@ GLboolean next_ins(r700_AssemblerBase *pAsm) //reset for next inst. pAsm->D.bits = 0; + pAsm->D2.bits = 0; + pAsm->S[0].bits = 0; + pAsm->S[1].bits = 0; + pAsm->S[2].bits = 0; + pAsm->is_tex = GL_FALSE; + pAsm->need_tex_barrier = GL_FALSE; + + return GL_TRUE; +} + +GLboolean next_ins2(r700_AssemblerBase *pAsm) +{ + struct prog_instruction *pILInst = &(pAsm->pILInst[pAsm->uiCurInst]); + + //ALU + if( GL_FALSE == assemble_alu_instruction2(pAsm) ) + { + radeon_error("Error assembling ALU instruction\n"); + return GL_FALSE; + } + + if(pAsm->D.dst.rtype == DST_REG_OUT) + { + if(pAsm->D.dst.op3) + { + // There is no mask for OP3 instructions, so all channels are written + pAsm->pucOutMask[pAsm->D.dst.reg - pAsm->starting_export_register_number] = 0xF; + } + else + { + pAsm->pucOutMask[pAsm->D.dst.reg - pAsm->starting_export_register_number] + |= (unsigned char)pAsm->pILInst[pAsm->uiCurInst].DstReg.WriteMask; + } + } + + //reset for next inst. + pAsm->D.bits = 0; + pAsm->D2.bits = 0; + pAsm->S[0].bits = 0; + pAsm->S[1].bits = 0; + pAsm->S[2].bits = 0; + pAsm->is_tex = GL_FALSE; + pAsm->need_tex_barrier = GL_FALSE; + + //richard nov.16 glsl + pAsm->D2.bits = 0; + + return GL_TRUE; +} + +/* not work yet */ +GLboolean next_ins_literal(r700_AssemblerBase *pAsm, GLfloat * pLiteral) +{ + struct prog_instruction *pILInst = &(pAsm->pILInst[pAsm->uiCurInst]); + + //ALU + if( GL_FALSE == assemble_alu_instruction_literal(pAsm, pLiteral) ) + { + radeon_error("Error assembling ALU instruction\n"); + return GL_FALSE; + } + + //reset for next inst. + pAsm->D.bits = 0; + pAsm->D2.bits = 0; pAsm->S[0].bits = 0; pAsm->S[1].bits = 0; pAsm->S[2].bits = 0; @@ -3816,6 +4456,74 @@ GLboolean assemble_SCS(r700_AssemblerBase *pAsm) return GL_TRUE; } + +GLboolean assemble_LOGIC(r700_AssemblerBase *pAsm, BITS opcode) +{ + if( GL_FALSE == checkop2(pAsm) ) + { + return GL_FALSE; + } + + pAsm->D.dst.opcode = opcode; + pAsm->D.dst.math = 1; + + if( GL_FALSE == assemble_dst(pAsm) ) + { + return GL_FALSE; + } + + if( GL_FALSE == assemble_src(pAsm, 0, -1) ) + { + return GL_FALSE; + } + + if( GL_FALSE == assemble_src(pAsm, 1, -1) ) + { + return GL_FALSE; + } + + if( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } + + return GL_TRUE; +} + +GLboolean assemble_LOGIC_PRED(r700_AssemblerBase *pAsm, BITS opcode) +{ + if( GL_FALSE == checkop2(pAsm) ) + { + return GL_FALSE; + } + + pAsm->D.dst.opcode = opcode; + pAsm->D.dst.math = 1; + pAsm->D.dst.predicated = 1; + pAsm->D2.dst2.SaturateMode = pAsm->pILInst[pAsm->uiCurInst].SaturateMode; + + if( GL_FALSE == assemble_dst(pAsm) ) + { + return GL_FALSE; + } + + if( GL_FALSE == assemble_src(pAsm, 0, -1) ) + { + return GL_FALSE; + } + + if( GL_FALSE == assemble_src(pAsm, 1, -1) ) + { + return GL_FALSE; + } + + if( GL_FALSE == next_ins2(pAsm) ) + { + return GL_FALSE; + } + + return GL_TRUE; +} GLboolean assemble_SGE(r700_AssemblerBase *pAsm) { @@ -4273,27 +4981,774 @@ GLboolean assemble_EXPORT(r700_AssemblerBase *pAsm) return GL_TRUE; } -GLboolean assemble_IF(r700_AssemblerBase *pAsm) +inline void checkStackDepth(r700_AssemblerBase *pAsm, GLuint uReason) +{ + switch (uReason) + { + case FC_PUSH_VPM: + pAsm->CALLSTACK[pAsm->CALLSP].stackUsage.su.pushs++; + pAsm->CALLSTACK[pAsm->CALLSP].stackUsage.su.current++; + break; + case FC_PUSH_WQM: + pAsm->CALLSTACK[pAsm->CALLSP].stackUsage.su.pushs++; + pAsm->CALLSTACK[pAsm->CALLSP].stackUsage.su.current += 4; + break; + case FC_LOOP: + pAsm->CALLSTACK[pAsm->CALLSP].stackUsage.su.pushs += 4; + break; + case FC_REP: + /* TODO : for 16 vp asic, should += 2; */ + pAsm->CALLSTACK[pAsm->CALLSP].stackUsage.su.pushs += 1; + break; + }; + + if(pAsm->CALLSTACK[pAsm->CALLSP].stackUsage.su.pushs + > pAsm->CALLSTACK[pAsm->CALLSP].stackUsage.su.max) + { + pAsm->CALLSTACK[pAsm->CALLSP].stackUsage.su.max = + pAsm->CALLSTACK[pAsm->CALLSP].stackUsage.su.pushs; + } +} + +GLboolean jumpToOffest(r700_AssemblerBase *pAsm, GLuint pops, GLint offset) +{ + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = pops; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_JUMP; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->cf_current_cf_clause_ptr->m_Word0.f.addr = pAsm->cf_current_cf_clause_ptr->m_uIndex + offset; + + return GL_TRUE; +} + +GLboolean pops(r700_AssemblerBase *pAsm, GLuint pops) +{ + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = pops; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_POP; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + pAsm->cf_current_cf_clause_ptr->m_Word0.f.addr = pAsm->cf_current_cf_clause_ptr->m_uIndex + 1; + + return GL_TRUE; +} + +GLboolean assemble_IF(r700_AssemblerBase *pAsm, GLboolean bHasElse) +{ + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + if(GL_TRUE != bHasElse) + { + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + } + else + { + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 0; + } + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_JUMP; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->FCSP++; + pAsm->fc_stack[pAsm->FCSP].type = FC_IF; + pAsm->fc_stack[pAsm->FCSP].bpush = 0; + 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 */ + + pAsm->branch_depth++; + + if(pAsm->branch_depth > pAsm->max_branch_depth) + { + pAsm->max_branch_depth = pAsm->branch_depth; + } + 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->branch_depth--; + pAsm->FCSP--; + + 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].bpush = 1; + 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; + + pAsm->branch_depth++; + + if(pAsm->branch_depth > pAsm->max_branch_depth) + { + pAsm->max_branch_depth = pAsm->branch_depth; + } + return GL_TRUE; +} + +GLboolean assemble_BRK(r700_AssemblerBase *pAsm) +{ +#ifdef USE_CF_FOR_CONTINUE_BREAK + unsigned int unFCSP; + for(unFCSP=pAsm->FCSP; unFCSP>0; unFCSP--) + { + if(FC_LOOP == pAsm->fc_stack[unFCSP].type) + { + break; + } + } + if(0 == FC_LOOP) + { + radeon_error("Break is not inside loop/endloop pair.\n"); + return GL_FALSE; + } + + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_LOOP_BREAK; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->fc_stack[unFCSP].mid = (R700ControlFlowGenericClause **)_mesa_realloc( + (void *)pAsm->fc_stack[unFCSP].mid, + sizeof(R700ControlFlowGenericClause *) * pAsm->fc_stack[unFCSP].unNumMid, + sizeof(R700ControlFlowGenericClause *) * (pAsm->fc_stack[unFCSP].unNumMid + 1) ); + pAsm->fc_stack[unFCSP].mid[pAsm->fc_stack[unFCSP].unNumMid] = pAsm->cf_current_cf_clause_ptr; + pAsm->fc_stack[unFCSP].unNumMid++; + + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_POP; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + pAsm->cf_current_cf_clause_ptr->m_Word0.f.addr = pAsm->cf_current_cf_clause_ptr->m_uIndex + 1; + +#endif //USE_CF_FOR_CONTINUE_BREAK + return GL_TRUE; +} + +GLboolean assemble_CONT(r700_AssemblerBase *pAsm) +{ +#ifdef USE_CF_FOR_CONTINUE_BREAK + unsigned int unFCSP; + for(unFCSP=pAsm->FCSP; unFCSP>0; unFCSP--) + { + if(FC_LOOP == pAsm->fc_stack[unFCSP].type) + { + break; + } + } + if(0 == FC_LOOP) + { + radeon_error("Continue is not inside loop/endloop pair.\n"); + return GL_FALSE; + } + + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_LOOP_CONTINUE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->fc_stack[unFCSP].mid = (R700ControlFlowGenericClause **)_mesa_realloc( + (void *)pAsm->fc_stack[unFCSP].mid, + sizeof(R700ControlFlowGenericClause *) * pAsm->fc_stack[unFCSP].unNumMid, + sizeof(R700ControlFlowGenericClause *) * (pAsm->fc_stack[unFCSP].unNumMid + 1) ); + pAsm->fc_stack[unFCSP].mid[pAsm->fc_stack[unFCSP].unNumMid] = pAsm->cf_current_cf_clause_ptr; + pAsm->fc_stack[unFCSP].unNumMid++; + + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_POP; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + pAsm->cf_current_cf_clause_ptr->m_Word0.f.addr = pAsm->cf_current_cf_clause_ptr->m_uIndex + 1; + +#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; + } + + unsigned int unFCSP = 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) + { + break; + } + } + if(unFCSP <= pAsm->CALLSTACK[pAsm->CALLSP].FCSP_BeforeEntry) + { + unFCSP = 0; + + returnOnFlag(pAsm); + pAsm->unCFflags &= ~HAS_CURRENT_LOOPRET; + } + } + + pAsm->branch_depth--; + pAsm->FCSP--; + + if(unFCSP > 0) + { + breakLoopOnFlag(pAsm, unFCSP); + } + + return GL_TRUE; +} + +void add_return_inst(r700_AssemblerBase *pAsm) +{ + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + //pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_RETURN; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; +} + +GLboolean assemble_BGNSUB(r700_AssemblerBase *pAsm, GLint nILindex) +{ + /* Put in sub */ + if( (pAsm->unSubArrayPointer + 1) > pAsm->unSubArraySize ) + { + pAsm->subs = (SUB_OFFSET*)_mesa_realloc( (void *)pAsm->subs, + sizeof(SUB_OFFSET) * pAsm->unSubArraySize, + sizeof(SUB_OFFSET) * (pAsm->unSubArraySize + 10) ); + if(NULL == pAsm->subs) + { + return GL_FALSE; + } + pAsm->unSubArraySize += 10; + } + + pAsm->subs[pAsm->unSubArrayPointer].subIL_Offset = nILindex; + pAsm->subs[pAsm->unSubArrayPointer].lstCFInstructions_local.pHead=NULL; + pAsm->subs[pAsm->unSubArrayPointer].lstCFInstructions_local.pTail=NULL; + pAsm->subs[pAsm->unSubArrayPointer].lstCFInstructions_local.uNumOfNode=0; + + pAsm->CALLSP++; + pAsm->CALLSTACK[pAsm->CALLSP].FCSP_BeforeEntry = pAsm->FCSP; + pAsm->CALLSTACK[pAsm->CALLSP].plstCFInstructions_local + = &(pAsm->subs[pAsm->unSubArrayPointer].lstCFInstructions_local); + pAsm->CALLSTACK[pAsm->CALLSP].stackUsage.bits = 0; + SetActiveCFlist(pAsm->pR700Shader, + pAsm->CALLSTACK[pAsm->CALLSP].plstCFInstructions_local); + + pAsm->unSubArrayPointer++; + + /* start sub */ + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + + return GL_TRUE; +} + +GLboolean assemble_ENDSUB(r700_AssemblerBase *pAsm) +{ + pAsm->CALLSP--; + SetActiveCFlist(pAsm->pR700Shader, + pAsm->CALLSTACK[pAsm->CALLSP].plstCFInstructions_local); + + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + + return GL_TRUE; +} + +GLboolean assemble_RET(r700_AssemblerBase *pAsm) +{ + if(pAsm->CALLSP > 0) + { /* in sub */ + unsigned int 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; + } + } + } + + add_return_inst(pAsm); + + return GL_TRUE; +} + +GLboolean assemble_CAL(r700_AssemblerBase *pAsm, + GLint nILindex, + GLuint uiNumberInsts, + struct prog_instruction *pILInst) +{ + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.call_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_CALL; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + /* Put in caller */ + if( (pAsm->unCallerArrayPointer + 1) > pAsm->unCallerArraySize ) + { + pAsm->callers = (CALLER_POINTER*)_mesa_realloc( (void *)pAsm->callers, + sizeof(CALLER_POINTER) * pAsm->unCallerArraySize, + sizeof(CALLER_POINTER) * (pAsm->unCallerArraySize + 10) ); + if(NULL == pAsm->callers) + { + return GL_FALSE; + } + pAsm->unCallerArraySize += 10; + } + + pAsm->callers[pAsm->unCallerArrayPointer].subIL_Offset = nILindex; + pAsm->callers[pAsm->unCallerArrayPointer].cf_ptr = pAsm->cf_current_cf_clause_ptr; + + pAsm->unCallerArrayPointer++; + + int j; + for(j=0; j<pAsm->unSubArrayPointer; j++) + { + if(nILindex == pAsm->subs[j].subIL_Offset) + { /* compiled before */ + pAsm->callers[pAsm->unCallerArrayPointer - 1].subDescIndex = j; + return GL_TRUE; + } + } + + pAsm->callers[pAsm->unCallerArrayPointer - 1].subDescIndex = pAsm->unSubArrayPointer; + + return AssembleInstr(nILindex, uiNumberInsts, pILInst, pAsm); +} + +GLboolean setRetInLoopFlag(r700_AssemblerBase *pAsm, GLuint flagValue) +{ + GLfloat fLiteral[2] = {0.1, 0.0}; + + pAsm->D.dst.opcode = SQ_OP2_INST_MOV; + pAsm->D.dst.op3 = 0; + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = pAsm->flag_reg_index; + pAsm->D.dst.writex = 1; + pAsm->D.dst.writey = 0; + pAsm->D.dst.writez = 0; + pAsm->D.dst.writew = 0; + pAsm->D2.dst2.literal = 1; + pAsm->D2.dst2.SaturateMode = SATURATE_OFF; + pAsm->D.dst.predicated = 0; +#if 0 + pAsm->S[0].src.rtype = SRC_REC_LITERAL; + //pAsm->S[0].src.reg = 0; + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + noneg_PVSSRC(&(pAsm->S[0].src)); + pAsm->S[0].src.swizzlex = SQ_SEL_X; + pAsm->S[0].src.swizzley = SQ_SEL_Y; + pAsm->S[0].src.swizzlez = SQ_SEL_Z; + pAsm->S[0].src.swizzlew = SQ_SEL_W; + + if( GL_FALSE == next_ins_literal(pAsm, &(fLiteral[0])) ) + { + return GL_FALSE; + } +#else + pAsm->S[0].src.rtype = DST_REG_TEMPORARY; + pAsm->S[0].src.reg = 0; + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + noneg_PVSSRC(&(pAsm->S[0].src)); + pAsm->S[0].src.swizzlex = flagValue; + pAsm->S[0].src.swizzley = flagValue; + pAsm->S[0].src.swizzlez = flagValue; + pAsm->S[0].src.swizzlew = flagValue; + + if( GL_FALSE == next_ins2(pAsm) ) + { + return GL_FALSE; + } +#endif + + return GL_TRUE; +} + +GLboolean testFlag(r700_AssemblerBase *pAsm) +{ + GLfloat fLiteral[2] = {0.1, 0.0}; + + //Test flag + GLuint tmp = gethelpr(pAsm); + pAsm->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; + + pAsm->D.dst.opcode = SQ_OP2_INST_PRED_SETE; + pAsm->D.dst.math = 1; + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = tmp; + pAsm->D.dst.writex = 1; + pAsm->D.dst.writey = 0; + pAsm->D.dst.writez = 0; + pAsm->D.dst.writew = 0; + pAsm->D2.dst2.literal = 1; + pAsm->D2.dst2.SaturateMode = SATURATE_OFF; + pAsm->D.dst.predicated = 1; + + pAsm->S[0].src.rtype = DST_REG_TEMPORARY; + pAsm->S[0].src.reg = pAsm->flag_reg_index; + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + noneg_PVSSRC(&(pAsm->S[0].src)); + pAsm->S[0].src.swizzlex = SQ_SEL_X; + pAsm->S[0].src.swizzley = SQ_SEL_Y; + pAsm->S[0].src.swizzlez = SQ_SEL_Z; + pAsm->S[0].src.swizzlew = SQ_SEL_W; +#if 0 + pAsm->S[1].src.rtype = SRC_REC_LITERAL; + //pAsm->S[1].src.reg = 0; + setaddrmode_PVSSRC(&(pAsm->S[1].src), ADDR_ABSOLUTE); + noneg_PVSSRC(&(pAsm->S[1].src)); + pAsm->S[1].src.swizzlex = SQ_SEL_X; + pAsm->S[1].src.swizzley = SQ_SEL_Y; + pAsm->S[1].src.swizzlez = SQ_SEL_Z; + pAsm->S[1].src.swizzlew = SQ_SEL_W; + + if( GL_FALSE == next_ins_literal(pAsm, &(fLiteral[0])) ) + { + return GL_FALSE; + } +#else + pAsm->S[1].src.rtype = DST_REG_TEMPORARY; + pAsm->S[1].src.reg = 0; + setaddrmode_PVSSRC(&(pAsm->S[1].src), ADDR_ABSOLUTE); + noneg_PVSSRC(&(pAsm->S[1].src)); + pAsm->S[1].src.swizzlex = SQ_SEL_1; + pAsm->S[1].src.swizzley = SQ_SEL_1; + pAsm->S[1].src.swizzlez = SQ_SEL_1; + pAsm->S[1].src.swizzlew = SQ_SEL_1; + + if( GL_FALSE == next_ins2(pAsm) ) + { + return GL_FALSE; + } +#endif + + checkStackDepth(pAsm, FC_PUSH_VPM); + + return GL_TRUE; +} + +GLboolean returnOnFlag(r700_AssemblerBase *pAsm) +{ + testFlag(pAsm); + jumpToOffest(pAsm, 1, 4); + setRetInLoopFlag(pAsm, SQ_SEL_0); + pops(pAsm, 1); + add_return_inst(pAsm); + return GL_TRUE; } -GLboolean AssembleInstr(GLuint uiNumberInsts, +GLboolean breakLoopOnFlag(r700_AssemblerBase *pAsm, GLuint unFCSP) +{ + testFlag(pAsm); + + //break + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_LOOP_BREAK; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->fc_stack[unFCSP].mid = (R700ControlFlowGenericClause **)_mesa_realloc( + (void *)pAsm->fc_stack[unFCSP].mid, + sizeof(R700ControlFlowGenericClause *) * pAsm->fc_stack[unFCSP].unNumMid, + sizeof(R700ControlFlowGenericClause *) * (pAsm->fc_stack[unFCSP].unNumMid + 1) ); + pAsm->fc_stack[unFCSP].mid[pAsm->fc_stack[unFCSP].unNumMid] = pAsm->cf_current_cf_clause_ptr; + pAsm->fc_stack[unFCSP].unNumMid++; + + pops(pAsm, 1); + + return GL_TRUE; +} + +GLboolean AssembleInstr(GLuint uiFirstInst, + GLuint uiNumberInsts, struct prog_instruction *pILInst, r700_AssemblerBase *pR700AsmCode) { GLuint i; pR700AsmCode->pILInst = pILInst; - for(i=0; i<uiNumberInsts; i++) + for(i=uiFirstInst; i<uiNumberInsts; i++) { pR700AsmCode->uiCurInst = i; +#ifndef USE_CF_FOR_CONTINUE_BREAK + if(OPCODE_BRK == pILInst[i+1].Opcode) + { + switch(pILInst[i].Opcode) + { + case OPCODE_SLE: + pILInst[i].Opcode = OPCODE_SGT; + break; + case OPCODE_SLT: + pILInst[i].Opcode = OPCODE_SGE; + break; + case OPCODE_SGE: + pILInst[i].Opcode = OPCODE_SLT; + break; + case OPCODE_SGT: + pILInst[i].Opcode = OPCODE_SLE; + break; + case OPCODE_SEQ: + pILInst[i].Opcode = OPCODE_SNE; + break; + case OPCODE_SNE: + pILInst[i].Opcode = OPCODE_SEQ; + break; + default: + break; + } + } +#endif + switch (pILInst[i].Opcode) { case OPCODE_ABS: @@ -4422,16 +5877,298 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, case OPCODE_SCS: if ( GL_FALSE == assemble_SCS(pR700AsmCode) ) return GL_FALSE; - break; + break; + + case OPCODE_SEQ: + if(OPCODE_IF == pILInst[i+1].Opcode) + { + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETE) ) + { + return GL_FALSE; + } + } + else if(OPCODE_BRK == pILInst[i+1].Opcode) + { +#ifdef USE_CF_FOR_CONTINUE_BREAK + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; +#else + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_BREAK; +#endif + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETE) ) + { + return GL_FALSE; + } + } + else if(OPCODE_CONT == pILInst[i+1].Opcode) + { +#ifdef USE_CF_FOR_CONTINUE_BREAK + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; +#else + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_CONTINUE; +#endif + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETE) ) + { + return GL_FALSE; + } + } + else + { + if ( GL_FALSE == assemble_LOGIC(pR700AsmCode, SQ_OP2_INST_SETE) ) + { + return GL_FALSE; + } + } + break; + + case OPCODE_SGT: + if(OPCODE_IF == pILInst[i+1].Opcode) + { + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETGT) ) + { + return GL_FALSE; + } + } + else if(OPCODE_BRK == pILInst[i+1].Opcode) + { +#ifdef USE_CF_FOR_CONTINUE_BREAK + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; +#else + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_BREAK; +#endif + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETGT) ) + { + return GL_FALSE; + } + } + else if(OPCODE_CONT == pILInst[i+1].Opcode) + { +#ifdef USE_CF_FOR_CONTINUE_BREAK + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; +#else + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_CONTINUE; +#endif + + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETGT) ) + { + return GL_FALSE; + } + } + else + { + if ( GL_FALSE == assemble_LOGIC(pR700AsmCode, SQ_OP2_INST_SETGT) ) + { + return GL_FALSE; + } + } + break; case OPCODE_SGE: - if ( GL_FALSE == assemble_SGE(pR700AsmCode) ) - return GL_FALSE; - break; + if(OPCODE_IF == pILInst[i+1].Opcode) + { + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETGE) ) + { + return GL_FALSE; + } + } + else if(OPCODE_BRK == pILInst[i+1].Opcode) + { +#ifdef USE_CF_FOR_CONTINUE_BREAK + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; +#else + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_BREAK; +#endif + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETGE) ) + { + return GL_FALSE; + } + } + else if(OPCODE_CONT == pILInst[i+1].Opcode) + { +#ifdef USE_CF_FOR_CONTINUE_BREAK + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; +#else + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_CONTINUE; +#endif + + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETGE) ) + { + return GL_FALSE; + } + } + else + { + if ( GL_FALSE == assemble_SGE(pR700AsmCode) ) + { + return GL_FALSE; + } + } + break; + + /* NO LT, LE, TODO : use GE => LE, GT => LT : reverse 2 src order would be simpliest. Or use SQ_CF_COND_FALSE for SQ_CF_COND_ACTIVE.*/ case OPCODE_SLT: - if ( GL_FALSE == assemble_SLT(pR700AsmCode) ) - return GL_FALSE; - break; + { + struct prog_src_register SrcRegSave[2]; + SrcRegSave[0] = pILInst[i].SrcReg[0]; + SrcRegSave[1] = pILInst[i].SrcReg[1]; + pILInst[i].SrcReg[0] = SrcRegSave[1]; + pILInst[i].SrcReg[1] = SrcRegSave[0]; + if(OPCODE_IF == pILInst[i+1].Opcode) + { + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETGT) ) + { + pILInst[i].SrcReg[0] = SrcRegSave[0]; + pILInst[i].SrcReg[1] = SrcRegSave[1]; + return GL_FALSE; + } + } + else if(OPCODE_BRK == pILInst[i+1].Opcode) + { +#ifdef USE_CF_FOR_CONTINUE_BREAK + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; +#else + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_BREAK; +#endif + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETGT) ) + { + pILInst[i].SrcReg[0] = SrcRegSave[0]; + pILInst[i].SrcReg[1] = SrcRegSave[1]; + return GL_FALSE; + } + } + else if(OPCODE_CONT == pILInst[i+1].Opcode) + { +#ifdef USE_CF_FOR_CONTINUE_BREAK + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; +#else + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_CONTINUE; +#endif + + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETGT) ) + { + pILInst[i].SrcReg[0] = SrcRegSave[0]; + pILInst[i].SrcReg[1] = SrcRegSave[1]; + return GL_FALSE; + } + } + else + { + if ( GL_FALSE == assemble_LOGIC(pR700AsmCode, SQ_OP2_INST_SETGT) ) + { + pILInst[i].SrcReg[0] = SrcRegSave[0]; + pILInst[i].SrcReg[1] = SrcRegSave[1]; + return GL_FALSE; + } + } + pILInst[i].SrcReg[0] = SrcRegSave[0]; + pILInst[i].SrcReg[1] = SrcRegSave[1]; + } + break; + + case OPCODE_SLE: + { + struct prog_src_register SrcRegSave[2]; + SrcRegSave[0] = pILInst[i].SrcReg[0]; + SrcRegSave[1] = pILInst[i].SrcReg[1]; + pILInst[i].SrcReg[0] = SrcRegSave[1]; + pILInst[i].SrcReg[1] = SrcRegSave[0]; + if(OPCODE_IF == pILInst[i+1].Opcode) + { + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETGE) ) + { + pILInst[i].SrcReg[0] = SrcRegSave[0]; + pILInst[i].SrcReg[1] = SrcRegSave[1]; + return GL_FALSE; + } + } + else if(OPCODE_BRK == pILInst[i+1].Opcode) + { +#ifdef USE_CF_FOR_CONTINUE_BREAK + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; +#else + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_BREAK; +#endif + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETGE) ) + { + pILInst[i].SrcReg[0] = SrcRegSave[0]; + pILInst[i].SrcReg[1] = SrcRegSave[1]; + return GL_FALSE; + } + } + else if(OPCODE_CONT == pILInst[i+1].Opcode) + { +#ifdef USE_CF_FOR_CONTINUE_BREAK + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; +#else + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_CONTINUE; +#endif + + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETGE) ) + { + pILInst[i].SrcReg[0] = SrcRegSave[0]; + pILInst[i].SrcReg[1] = SrcRegSave[1]; + return GL_FALSE; + } + } + else + { + if ( GL_FALSE == assemble_LOGIC(pR700AsmCode, SQ_OP2_INST_SETGE) ) + { + pILInst[i].SrcReg[0] = SrcRegSave[0]; + pILInst[i].SrcReg[1] = SrcRegSave[1]; + return GL_FALSE; + } + } + pILInst[i].SrcReg[0] = SrcRegSave[0]; + pILInst[i].SrcReg[1] = SrcRegSave[1]; + } + break; + + case OPCODE_SNE: + if(OPCODE_IF == pILInst[i+1].Opcode) + { + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETNE) ) + { + return GL_FALSE; + } + } + else if(OPCODE_BRK == pILInst[i+1].Opcode) + { +#ifdef USE_CF_FOR_CONTINUE_BREAK + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; +#else + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_BREAK; +#endif + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETNE) ) + { + return GL_FALSE; + } + } + else if(OPCODE_CONT == pILInst[i+1].Opcode) + { +#ifdef USE_CF_FOR_CONTINUE_BREAK + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; +#else + pR700AsmCode->alu_x_opcode = SQ_CF_INST_ALU_CONTINUE; +#endif + if ( GL_FALSE == assemble_LOGIC_PRED(pR700AsmCode, SQ_OP2_INST_PRED_SETNE) ) + { + return GL_FALSE; + } + } + else + { + if ( GL_FALSE == assemble_LOGIC(pR700AsmCode, SQ_OP2_INST_SETNE) ) + { + return GL_FALSE; + } + } + break; //case OP_STP: // if ( GL_FALSE == assemble_STP(pR700AsmCode) ) @@ -4471,24 +6208,91 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, break; case OPCODE_IF : - if ( GL_FALSE == assemble_IF(pR700AsmCode) ) - return GL_FALSE; + { + GLboolean bHasElse = GL_FALSE; + + if(pILInst[pILInst[i].BranchTarget - 1].Opcode == OPCODE_ELSE) + { + bHasElse = GL_TRUE; + } + + if ( GL_FALSE == assemble_IF(pR700AsmCode, bHasElse) ) + { + return GL_FALSE; + } + } break; + case OPCODE_ELSE : - radeon_error("Not yet implemented instruction OPCODE_ELSE \n"); - //if ( GL_FALSE == assemble_BAD("ELSE") ) + if ( GL_FALSE == assemble_ELSE(pR700AsmCode) ) return GL_FALSE; break; + case OPCODE_ENDIF: if ( GL_FALSE == assemble_ENDIF(pR700AsmCode) ) return GL_FALSE; break; + case OPCODE_BGNLOOP: + if( GL_FALSE == assemble_BGNLOOP(pR700AsmCode) ) + { + return GL_FALSE; + } + break; + + case OPCODE_BRK: + if( GL_FALSE == assemble_BRK(pR700AsmCode) ) + { + return GL_FALSE; + } + break; + + case OPCODE_CONT: + if( GL_FALSE == assemble_CONT(pR700AsmCode) ) + { + return GL_FALSE; + } + break; + + case OPCODE_ENDLOOP: + if( GL_FALSE == assemble_ENDLOOP(pR700AsmCode) ) + { + return GL_FALSE; + } + break; + + case OPCODE_BGNSUB: + if( GL_FALSE == assemble_BGNSUB(pR700AsmCode, i) ) + { + return GL_FALSE; + } + break; + + case OPCODE_RET: + if( GL_FALSE == assemble_RET(pR700AsmCode) ) + { + return GL_FALSE; + } + break; + + case OPCODE_CAL: + if( GL_FALSE == assemble_CAL(pR700AsmCode, + pILInst[i].BranchTarget, + uiNumberInsts, + pILInst) ) + { + return GL_FALSE; + } + break; + //case OPCODE_EXPORT: // if ( GL_FALSE == assemble_EXPORT() ) // return GL_FALSE; // break; + case OPCODE_ENDSUB: + return assemble_ENDSUB(pR700AsmCode); + case OPCODE_END: //pR700AsmCode->uiCurInst = i; //This is to remaind that if in later exoort there is depth/stencil @@ -4505,6 +6309,116 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, return GL_TRUE; } +GLboolean InitShaderProgram(r700_AssemblerBase * pAsm) +{ + setRetInLoopFlag(pAsm, SQ_SEL_0); + return GL_TRUE; +} + +GLboolean RelocProgram(r700_AssemblerBase * pAsm) +{ + GLuint i; + GLuint unCFoffset; + TypedShaderList * plstCFmain; + TypedShaderList * plstCFsub; + + R700ShaderInstruction * pInst; + R700ControlFlowGenericClause * pCFInst; + + if(0 == pAsm->unSubArrayPointer) + { + return GL_TRUE; + } + + plstCFmain = pAsm->CALLSTACK[0].plstCFInstructions_local; + unCFoffset = plstCFmain->uNumOfNode; + + /* Reloc subs */ + for(i=0; i<pAsm->unSubArrayPointer; i++) + { + pAsm->subs[i].unCFoffset = unCFoffset; + plstCFsub = &(pAsm->subs[i].lstCFInstructions_local); + + pInst = plstCFsub->pHead; + + /* reloc instructions */ + while(pInst) + { + if(SIT_CF_GENERIC == pInst->m_ShaderInstType) + { + pCFInst = (R700ControlFlowGenericClause *)pInst; + + switch (pCFInst->m_Word1.f.cf_inst) + { + case SQ_CF_INST_POP: + case SQ_CF_INST_JUMP: + case SQ_CF_INST_ELSE: + case SQ_CF_INST_LOOP_END: + case SQ_CF_INST_LOOP_START: + case SQ_CF_INST_LOOP_START_NO_AL: + case SQ_CF_INST_LOOP_CONTINUE: + case SQ_CF_INST_LOOP_BREAK: + pCFInst->m_Word0.f.addr += unCFoffset; + break; + default: + break; + } + } + + pInst->m_uIndex += unCFoffset; + + pInst = pInst->pNextInst; + }; + + /* Put sub into main */ + plstCFmain->pTail->pNextInst = plstCFsub->pHead; + plstCFmain->pTail = plstCFsub->pTail; + plstCFmain->uNumOfNode += plstCFsub->uNumOfNode; + + unCFoffset += plstCFsub->uNumOfNode; + } + + /* reloc callers */ + for(i=0; i<pAsm->unCallerArrayPointer; i++) + { + pAsm->callers[i].cf_ptr->m_Word0.f.addr + = pAsm->subs[pAsm->callers[i].subDescIndex].unCFoffset; + } + + /* 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(1 == 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; + }; + } + + return GL_TRUE; +} + GLboolean Process_Export(r700_AssemblerBase* pAsm, GLuint type, GLuint export_starting_index, @@ -4800,6 +6714,25 @@ GLboolean Process_Vertex_Exports(r700_AssemblerBase *pR700AsmCode, } } + for(i=VERT_RESULT_VAR0; i<VERT_RESULT_MAX; i++) + { + unBit = 1 << i; + if(OutputsWritten & unBit) + { + if( GL_FALSE == Process_Export(pR700AsmCode, + SQ_EXPORT_PARAM, + export_starting_index, + 1, + pR700AsmCode->ucVP_OutputMap[i], + GL_FALSE) ) + { + return GL_FALSE; + } + + export_starting_index++; + } + } + // At least one param should be exported if (export_count) { @@ -4833,6 +6766,16 @@ GLboolean Clean_Up_Assembler(r700_AssemblerBase *pR700AsmCode) { FREE(pR700AsmCode->pucOutMask); FREE(pR700AsmCode->pInstDeps); + + if(NULL != pR700AsmCode->subs) + { + FREE(pR700AsmCode->subs); + } + if(NULL != pR700AsmCode->callers) + { + FREE(pR700AsmCode->callers); + } + return GL_TRUE; } diff --git a/src/mesa/drivers/dri/r600/r700_assembler.h b/src/mesa/drivers/dri/r600/r700_assembler.h index c66db502a1..516923f67c 100644 --- a/src/mesa/drivers/dri/r600/r700_assembler.h +++ b/src/mesa/drivers/dri/r600/r700_assembler.h @@ -72,7 +72,8 @@ typedef enum SrcRegisterType SRC_REG_INPUT = 1, SRC_REG_CONSTANT = 2, SRC_REG_ALT_TEMPORARY = 3, - NUMBER_OF_SRC_REG_TYPE = 4 + SRC_REC_LITERAL = 4, + NUMBER_OF_SRC_REG_TYPE = 5 } SrcRegisterType; typedef enum DstRegisterType @@ -111,6 +112,12 @@ typedef struct PVSDSTtag BITS addrmode1:1; //32 } PVSDST; +typedef struct PVSINSTtag +{ + BITS literal :2; + BITS SaturateMode :2; +} PVSINST; + typedef struct PVSSRCtag { BITS rtype:4; @@ -148,6 +155,7 @@ typedef union PVSDWORDtag { BITS bits; PVSDST dst; + PVSINST dst2; PVSSRC src; PVSMATH math; float f; @@ -251,6 +259,8 @@ enum FC_IF = 1, FC_LOOP = 2, FC_REP = 3, + FC_PUSH_VPM = 4, + FC_PUSH_WQM = 5, COND_NONE = 0, COND_BOOL = 1, @@ -263,14 +273,15 @@ 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; + unsigned int bpush; ///< 1 if first instruction does branch stack push + int id; ///< id of bool or int variable } FC_LEVEL; typedef struct VTX_FETCH_METHOD @@ -279,6 +290,46 @@ typedef struct VTX_FETCH_METHOD GLuint mega_fetch_remainder; } VTX_FETCH_METHOD; +typedef struct SUB_OFFSET +{ + GLint subIL_Offset; + GLuint unCFoffset; + TypedShaderList lstCFInstructions_local; +} SUB_OFFSET; + +typedef struct CALLER_POINTER +{ + GLint subIL_Offset; + GLint subDescIndex; + R700ControlFlowGenericClause* cf_ptr; +} CALLER_POINTER; + +#define SQ_MAX_CALL_DEPTH 0x00000020 + +typedef struct STACK_USAGE +{ + BITS pushs :8; + BITS current :8; + BITS max :8; +} STACK_USAGE; + +typedef union STACKDWORDtag +{ + BITS bits; + STACK_USAGE su; +} STACKDWORD; + +typedef struct CALL_LEVEL +{ + unsigned int FCSP_BeforeEntry; + STACKDWORD stackUsage; + TypedShaderList * plstCFInstructions_local; +} CALL_LEVEL; + +#define HAS_CURRENT_LOOPRET 0x1L +#define HAS_LOOPRET 0x2L +#define LOOPRET_FLAGS HAS_LOOPRET | HAS_CURRENT_LOOPRET + typedef struct r700_AssemblerBase { R700ControlFlowSXClause* cf_last_export_ptr; @@ -294,11 +345,14 @@ typedef struct r700_AssemblerBase // No clause has been created yet CF_CLAUSE_TYPE cf_current_clause_type; + BITS alu_x_opcode; + GLuint number_of_exports; GLuint number_of_colorandz_exports; GLuint number_of_export_opcodes; PVSDWORD D; + PVSDWORD D2; PVSDWORD S[3]; unsigned int uLastPosUpdate; @@ -310,6 +364,8 @@ 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]; @@ -378,9 +434,22 @@ typedef struct r700_AssemblerBase GLboolean is_tex; /* we inserted helper intructions and need barrier on next TEX ins */ GLboolean need_tex_barrier; + + SUB_OFFSET * subs; + GLuint unSubArraySize; + GLuint unSubArrayPointer; + CALLER_POINTER * callers; + GLuint unCallerArraySize; + GLuint unCallerArrayPointer; + unsigned int CALLSP; + CALL_LEVEL CALLSTACK[SQ_MAX_CALL_DEPTH]; + + GLuint unCFflags; + } r700_AssemblerBase; //Internal use +inline void checkStackDepth(r700_AssemblerBase *pAsm, GLuint uReason); BITS addrmode_PVSDST(PVSDST * pPVSDST); void setaddrmode_PVSDST(PVSDST * pPVSDST, BITS addrmode); void nomask_PVSDST(PVSDST * pPVSDST); @@ -446,6 +515,10 @@ GLboolean assemble_alu_src(R700ALUInstruction* alu_instruction_ptr, GLboolean add_alu_instruction(r700_AssemblerBase* pAsm, R700ALUInstruction* alu_instruction_ptr, GLuint contiguous_slots_needed); + +GLboolean add_cf_instruction(r700_AssemblerBase* pAsm); +void add_return_inst(r700_AssemblerBase *pAsm); + void get_src_properties(R700ALUInstruction* alu_instruction_ptr, int source_index, BITS* psrc_sel, @@ -467,6 +540,21 @@ GLboolean check_vector(r700_AssemblerBase* pAsm, R700ALUInstruction* alu_instruction_ptr); GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm); GLboolean next_ins(r700_AssemblerBase *pAsm); + +GLboolean next_ins2(r700_AssemblerBase *pAsm); +GLboolean assemble_alu_instruction2(r700_AssemblerBase *pAsm); + +/* TODO : merge next_ins/2/literal, assemble_alu_instruction/2/literal */ +GLboolean next_ins_literal(r700_AssemblerBase *pAsm, GLfloat * pLiteral); +GLboolean assemble_alu_instruction_literal(r700_AssemblerBase *pAsm, GLfloat * pLiteral); + +GLboolean pops(r700_AssemblerBase *pAsm, GLuint pops); +GLboolean jumpToOffest(r700_AssemblerBase *pAsm, GLuint pops, GLint offset); +GLboolean setRetInLoopFlag(r700_AssemblerBase *pAsm, GLuint flagValue); +GLboolean testFlag(r700_AssemblerBase *pAsm); +GLboolean breakLoopOnFlag(r700_AssemblerBase *pAsm, GLuint unFCSP); +GLboolean returnOnFlag(r700_AssemblerBase *pAsm); + GLboolean assemble_math_function(r700_AssemblerBase* pAsm, BITS opcode); GLboolean assemble_ABS(r700_AssemblerBase *pAsm); GLboolean assemble_ADD(r700_AssemblerBase *pAsm); @@ -497,14 +585,32 @@ GLboolean assemble_RSQ(r700_AssemblerBase *pAsm); GLboolean assemble_SIN(r700_AssemblerBase *pAsm); GLboolean assemble_SCS(r700_AssemblerBase *pAsm); GLboolean assemble_SGE(r700_AssemblerBase *pAsm); + +GLboolean assemble_LOGIC(r700_AssemblerBase *pAsm, BITS opcode); +GLboolean assemble_LOGIC_PRED(r700_AssemblerBase *pAsm, BITS opcode); + GLboolean assemble_SLT(r700_AssemblerBase *pAsm); GLboolean assemble_STP(r700_AssemblerBase *pAsm); GLboolean assemble_TEX(r700_AssemblerBase *pAsm); GLboolean assemble_XPD(r700_AssemblerBase *pAsm); GLboolean assemble_EXPORT(r700_AssemblerBase *pAsm); -GLboolean assemble_IF(r700_AssemblerBase *pAsm); +GLboolean assemble_IF(r700_AssemblerBase *pAsm, GLboolean bHasElse); +GLboolean assemble_ELSE(r700_AssemblerBase *pAsm); GLboolean assemble_ENDIF(r700_AssemblerBase *pAsm); +GLboolean assemble_BGNLOOP(r700_AssemblerBase *pAsm); +GLboolean assemble_BRK(r700_AssemblerBase *pAsm); +GLboolean assemble_COND(r700_AssemblerBase *pAsm); +GLboolean assemble_ENDLOOP(r700_AssemblerBase *pAsm); + +GLboolean assemble_BGNSUB(r700_AssemblerBase *pAsm, GLint nILindex); +GLboolean assemble_ENDSUB(r700_AssemblerBase *pAsm); +GLboolean assemble_RET(r700_AssemblerBase *pAsm); +GLboolean assemble_CAL(r700_AssemblerBase *pAsm, + GLint nILindex, + GLuint uiNumberInsts, + struct prog_instruction *pILInst); + GLboolean Process_Export(r700_AssemblerBase* pAsm, GLuint type, GLuint export_starting_index, @@ -516,12 +622,16 @@ GLboolean Move_Depth_Exports_To_Correct_Channels(r700_AssemblerBase *pAsm, //Interface -GLboolean AssembleInstr(GLuint uiNumberInsts, +GLboolean AssembleInstr(GLuint uiFirstInst, + GLuint uiNumberInsts, struct prog_instruction *pILInst, r700_AssemblerBase *pR700AsmCode); GLboolean Process_Fragment_Exports(r700_AssemblerBase *pR700AsmCode, GLbitfield OutputsWritten); GLboolean Process_Vertex_Exports(r700_AssemblerBase *pR700AsmCode, GLbitfield OutputsWritten); +GLboolean RelocProgram(r700_AssemblerBase * pAsm); +GLboolean InitShaderProgram(r700_AssemblerBase * pAsm); + int Init_r700_AssemblerBase(SHADER_PIPE_TYPE spt, r700_AssemblerBase* pAsm, R700_Shader* pShader); GLboolean Clean_Up_Assembler(r700_AssemblerBase *pR700AsmCode); diff --git a/src/mesa/drivers/dri/r600/r700_chip.c b/src/mesa/drivers/dri/r600/r700_chip.c index 2b2b4d748f..8126777bf4 100644 --- a/src/mesa/drivers/dri/r600/r700_chip.c +++ b/src/mesa/drivers/dri/r600/r700_chip.c @@ -446,68 +446,77 @@ static void r700SendRenderTargetState(GLcontext *ctx, struct radeon_state_atom * static void r700SendPSState(GLcontext *ctx, struct radeon_state_atom *atom) { - context_t *context = R700_CONTEXT(ctx); - R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); - struct radeon_bo * pbo; - BATCH_LOCALS(&context->radeon); - radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); + context_t *context = R700_CONTEXT(ctx); + R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); + struct radeon_bo * pbo; + BATCH_LOCALS(&context->radeon); + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); - pbo = (struct radeon_bo *)r700GetActiveFpShaderBo(GL_CONTEXT(context)); + pbo = (struct radeon_bo *)r700GetActiveFpShaderBo(GL_CONTEXT(context)); - if (!pbo) - return; + if (!pbo) + return; - r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit); + r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit); - BEGIN_BATCH_NO_AUTOSTATE(3 + 2); - R600_OUT_BATCH_REGSEQ(SQ_PGM_START_PS, 1); - R600_OUT_BATCH(r700->ps.SQ_PGM_START_PS.u32All); - R600_OUT_BATCH_RELOC(r700->ps.SQ_PGM_START_PS.u32All, - pbo, - r700->ps.SQ_PGM_START_PS.u32All, - RADEON_GEM_DOMAIN_GTT, 0, 0); - END_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(3 + 2); + R600_OUT_BATCH_REGSEQ(SQ_PGM_START_PS, 1); + R600_OUT_BATCH(r700->ps.SQ_PGM_START_PS.u32All); + R600_OUT_BATCH_RELOC(r700->ps.SQ_PGM_START_PS.u32All, + pbo, + r700->ps.SQ_PGM_START_PS.u32All, + RADEON_GEM_DOMAIN_GTT, 0, 0); + END_BATCH(); - BEGIN_BATCH_NO_AUTOSTATE(9); - R600_OUT_BATCH_REGVAL(SQ_PGM_RESOURCES_PS, r700->ps.SQ_PGM_RESOURCES_PS.u32All); - R600_OUT_BATCH_REGVAL(SQ_PGM_EXPORTS_PS, r700->ps.SQ_PGM_EXPORTS_PS.u32All); - R600_OUT_BATCH_REGVAL(SQ_PGM_CF_OFFSET_PS, r700->ps.SQ_PGM_CF_OFFSET_PS.u32All); - END_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(9); + R600_OUT_BATCH_REGVAL(SQ_PGM_RESOURCES_PS, r700->ps.SQ_PGM_RESOURCES_PS.u32All); + R600_OUT_BATCH_REGVAL(SQ_PGM_EXPORTS_PS, r700->ps.SQ_PGM_EXPORTS_PS.u32All); + R600_OUT_BATCH_REGVAL(SQ_PGM_CF_OFFSET_PS, r700->ps.SQ_PGM_CF_OFFSET_PS.u32All); + END_BATCH(); - COMMIT_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(3); + R600_OUT_BATCH_REGVAL(SQ_LOOP_CONST_0, 0x01000FFF); + END_BATCH(); + + COMMIT_BATCH(); } static void r700SendVSState(GLcontext *ctx, struct radeon_state_atom *atom) { - context_t *context = R700_CONTEXT(ctx); - R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); - struct radeon_bo * pbo; - BATCH_LOCALS(&context->radeon); - radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); + context_t *context = R700_CONTEXT(ctx); + R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); + struct radeon_bo * pbo; + BATCH_LOCALS(&context->radeon); + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); - pbo = (struct radeon_bo *)r700GetActiveVpShaderBo(GL_CONTEXT(context)); + pbo = (struct radeon_bo *)r700GetActiveVpShaderBo(GL_CONTEXT(context)); - if (!pbo) - return; + if (!pbo) + return; - r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit); + r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit); - BEGIN_BATCH_NO_AUTOSTATE(3 + 2); - R600_OUT_BATCH_REGSEQ(SQ_PGM_START_VS, 1); - R600_OUT_BATCH(r700->vs.SQ_PGM_START_VS.u32All); - R600_OUT_BATCH_RELOC(r700->vs.SQ_PGM_START_VS.u32All, - pbo, - r700->vs.SQ_PGM_START_VS.u32All, - RADEON_GEM_DOMAIN_GTT, 0, 0); - END_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(3 + 2); + R600_OUT_BATCH_REGSEQ(SQ_PGM_START_VS, 1); + R600_OUT_BATCH(r700->vs.SQ_PGM_START_VS.u32All); + R600_OUT_BATCH_RELOC(r700->vs.SQ_PGM_START_VS.u32All, + pbo, + r700->vs.SQ_PGM_START_VS.u32All, + RADEON_GEM_DOMAIN_GTT, 0, 0); + END_BATCH(); - BEGIN_BATCH_NO_AUTOSTATE(6); - R600_OUT_BATCH_REGVAL(SQ_PGM_RESOURCES_VS, r700->vs.SQ_PGM_RESOURCES_VS.u32All); - R600_OUT_BATCH_REGVAL(SQ_PGM_CF_OFFSET_VS, r700->vs.SQ_PGM_CF_OFFSET_VS.u32All); - END_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(6); + R600_OUT_BATCH_REGVAL(SQ_PGM_RESOURCES_VS, r700->vs.SQ_PGM_RESOURCES_VS.u32All); + R600_OUT_BATCH_REGVAL(SQ_PGM_CF_OFFSET_VS, r700->vs.SQ_PGM_CF_OFFSET_VS.u32All); + END_BATCH(); - COMMIT_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(3); + R600_OUT_BATCH_REGVAL((SQ_LOOP_CONST_0 + 32*4), 0x0100000F); + //R600_OUT_BATCH_REGVAL((SQ_LOOP_CONST_0 + (SQ_LOOP_CONST_vs<2)), 0x0100000F); + END_BATCH(); + + COMMIT_BATCH(); } static void r700SendFSState(GLcontext *ctx, struct radeon_state_atom *atom) @@ -1305,8 +1314,8 @@ void r600InitAtoms(context_t *context) ALLOC_STATE(spi, always, (59 + R700_MAX_SHADER_EXPORTS), r700SendSPIState); ALLOC_STATE(vpt, always, 16, r700SendViewportState); ALLOC_STATE(fs, always, 18, r700SendFSState); - ALLOC_STATE(vs, always, 18, r700SendVSState); - ALLOC_STATE(ps, always, 21, r700SendPSState); + ALLOC_STATE(vs, always, 21, r700SendVSState); + ALLOC_STATE(ps, always, 24, r700SendPSState); ALLOC_STATE(vs_consts, vs_consts, (2 + (R700_MAX_DX9_CONSTS * 4)), r700SendVSConsts); ALLOC_STATE(ps_consts, ps_consts, (2 + (R700_MAX_DX9_CONSTS * 4)), r700SendPSConsts); ALLOC_STATE(vtx, vtx, (6 + (VERT_ATTRIB_MAX * 18)), r700SendVTXState); diff --git a/src/mesa/drivers/dri/r600/r700_fragprog.c b/src/mesa/drivers/dri/r600/r700_fragprog.c index ccafd433bf..21ac46e7b8 100644 --- a/src/mesa/drivers/dri/r600/r700_fragprog.c +++ b/src/mesa/drivers/dri/r600/r700_fragprog.c @@ -73,11 +73,11 @@ void Map_Fragment_Program(r700_AssemblerBase *pAsm, pAsm->uiFP_AttributeMap[FRAG_ATTRIB_COL1] = pAsm->number_used_registers++; } - unBit = 1 << FRAG_ATTRIB_FOGC; - if(mesa_fp->Base.InputsRead & unBit) - { - pAsm->uiFP_AttributeMap[FRAG_ATTRIB_FOGC] = pAsm->number_used_registers++; - } + unBit = 1 << FRAG_ATTRIB_FOGC; + if(mesa_fp->Base.InputsRead & unBit) + { + pAsm->uiFP_AttributeMap[FRAG_ATTRIB_FOGC] = pAsm->number_used_registers++; + } for(i=0; i<8; i++) { @@ -88,6 +88,62 @@ void Map_Fragment_Program(r700_AssemblerBase *pAsm, } } +/* order has been taken care of */ +#if 1 + for(i=FRAG_ATTRIB_VAR0; i<FRAG_ATTRIB_MAX; i++) + { + unBit = 1 << i; + if(mesa_fp->Base.InputsRead & unBit) + { + pAsm->uiFP_AttributeMap[i] = pAsm->number_used_registers++; + } + } +#else + if( (mesa_fp->Base.InputsRead >> FRAG_ATTRIB_VAR0) > 0 ) + { + struct r700_vertex_program_cont *vpc = + (struct r700_vertex_program_cont *)ctx->VertexProgram._Current; + struct gl_program_parameter_list * VsVarying = vpc->mesa_program.Base.Varying; + struct gl_program_parameter_list * PsVarying = mesa_fp->Base.Varying; + struct gl_program_parameter * pVsParam; + struct gl_program_parameter * pPsParam; + GLuint j, k; + GLuint unMaxVarying = 0; + + for(i=0; i<VsVarying->NumParameters; i++) + { + pAsm->uiFP_AttributeMap[i + FRAG_ATTRIB_VAR0] = 0; + } + + for(i=FRAG_ATTRIB_VAR0; i<FRAG_ATTRIB_MAX; i++) + { + unBit = 1 << i; + if(mesa_fp->Base.InputsRead & unBit) + { + j = i - FRAG_ATTRIB_VAR0; + pPsParam = PsVarying->Parameters + j; + + for(k=0; k<VsVarying->NumParameters; k++) + { + pVsParam = VsVarying->Parameters + k; + + if( strcmp(pPsParam->Name, pVsParam->Name) == 0) + { + pAsm->uiFP_AttributeMap[i] = pAsm->number_used_registers + k; + if(k > unMaxVarying) + { + unMaxVarying = k; + } + break; + } + } + } + } + + pAsm->number_used_registers += unMaxVarying + 1; + } +#endif + /* Map temporary registers (GPRs) */ pAsm->starting_temp_register_number = pAsm->number_used_registers; @@ -127,6 +183,8 @@ void Map_Fragment_Program(r700_AssemblerBase *pAsm, pAsm->pucOutMask[ui] = 0x0; } + pAsm->flag_reg_index = pAsm->number_used_registers++; + pAsm->uFirstHelpReg = pAsm->number_used_registers; } @@ -247,8 +305,11 @@ GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp, { return GL_FALSE; } + + InitShaderProgram(&(fp->r700AsmCode)); - if( GL_FALSE == AssembleInstr(mesa_fp->Base.NumInstructions, + if( GL_FALSE == AssembleInstr(0, + mesa_fp->Base.NumInstructions, &(mesa_fp->Base.Instructions[0]), &(fp->r700AsmCode)) ) { @@ -260,6 +321,11 @@ GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp, return GL_FALSE; } + if( GL_FALSE == RelocProgram(&(fp->r700AsmCode)) ) + { + return GL_FALSE; + } + fp->r700Shader.nRegs = (fp->r700AsmCode.number_used_registers == 0) ? 0 : (fp->r700AsmCode.number_used_registers - 1); @@ -459,6 +525,22 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) } } + for(i=FRAG_ATTRIB_VAR0; i<FRAG_ATTRIB_MAX; i++) + { + unBit = 1 << i; + if(mesa_fp->Base.InputsRead & unBit) + { + ui = pAsm->uiFP_AttributeMap[i]; + SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, SEL_CENTROID_bit); + SETfield(r700->SPI_PS_INPUT_CNTL[ui].u32All, ui, + SEMANTIC_shift, SEMANTIC_mask); + if (r700->SPI_INTERP_CONTROL_0.u32All & FLAT_SHADE_ENA_bit) + SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); + else + CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); + } + } + exportCount = (r700->ps.SQ_PGM_EXPORTS_PS.u32All & EXPORT_MODE_mask) / (1 << EXPORT_MODE_shift); if (r700->CB_SHADER_CONTROL.u32All != ((1 << exportCount) - 1)) { diff --git a/src/mesa/drivers/dri/r600/r700_shader.c b/src/mesa/drivers/dri/r600/r700_shader.c index 955ea4e4e1..db951e48c4 100644 --- a/src/mesa/drivers/dri/r600/r700_shader.c +++ b/src/mesa/drivers/dri/r600/r700_shader.c @@ -140,7 +140,7 @@ void Init_R700_Shader(R700_Shader * pShader) pShader->killIsUsed = GL_FALSE; pShader->uCFOffset = 0; - pShader->uStackSize = 0; + pShader->uStackSize = 10; //richard test pShader->uMaxCallDepth = 0; pShader->bSurfAllocated = GL_FALSE; @@ -159,13 +159,18 @@ void Init_R700_Shader(R700_Shader * pShader) pShader->lstVTXInstructions.uNumOfNode=0; } +void SetActiveCFlist(R700_Shader *pShader, TypedShaderList * plstCF) +{ + pShader->plstCFInstructions_active = plstCF; +} + void AddCFInstruction(R700_Shader *pShader, R700ControlFlowInstruction *pCFInst) { R700ControlFlowSXClause* pSXClause; R700ControlFlowSMXClause* pSMXClause; - pCFInst->m_uIndex = pShader->lstCFInstructions.uNumOfNode; - AddInstToList(&(pShader->lstCFInstructions), + pCFInst->m_uIndex = pShader->plstCFInstructions_active->uNumOfNode; + AddInstToList(pShader->plstCFInstructions_active, (R700ShaderInstruction*)pCFInst); pShader->uShaderBinaryDWORDSize += GetInstructionSize(pCFInst->m_ShaderInstType); diff --git a/src/mesa/drivers/dri/r600/r700_shader.h b/src/mesa/drivers/dri/r600/r700_shader.h index c6a058617e..0599ffd901 100644 --- a/src/mesa/drivers/dri/r600/r700_shader.h +++ b/src/mesa/drivers/dri/r600/r700_shader.h @@ -109,6 +109,7 @@ typedef struct R700_Shader GLuint uStackSize; GLuint uMaxCallDepth; + TypedShaderList * plstCFInstructions_active; TypedShaderList lstCFInstructions; TypedShaderList lstALUInstructions; TypedShaderList lstTEXInstructions; @@ -132,13 +133,13 @@ void TakeInstOutFromList(TypedShaderList * plstCFInstructions, R700ShaderInstruc void ResolveLinks(R700_Shader *pShader); void Assemble(R700_Shader *pShader); - //Interface void Init_R700_Shader(R700_Shader * pShader); void AddCFInstruction(R700_Shader *pShader, R700ControlFlowInstruction *pCFInst); void AddVTXInstruction(R700_Shader *pShader, R700VertexInstruction *pVTXInst); void AddTEXInstruction(R700_Shader *pShader, R700TextureInstruction *pTEXInst); void AddALUInstruction(R700_Shader *pShader, R700ALUInstruction *pALUInst); +void SetActiveCFlist(R700_Shader *pShader, TypedShaderList * plstCF); void LoadProgram(R700_Shader *pShader); void UpdateShaderRegisters(R700_Shader *pShader); diff --git a/src/mesa/drivers/dri/r600/r700_vertprog.c b/src/mesa/drivers/dri/r600/r700_vertprog.c index ffc6068bd8..c8f72d588b 100644 --- a/src/mesa/drivers/dri/r600/r700_vertprog.c +++ b/src/mesa/drivers/dri/r600/r700_vertprog.c @@ -111,6 +111,15 @@ unsigned int Map_Vertex_Output(r700_AssemblerBase *pAsm, } } + for(i=VERT_RESULT_VAR0; i<VERT_RESULT_MAX; i++) + { + unBit = 1 << i; + if(mesa_vp->Base.OutputsWritten & unBit) + { + pAsm->ucVP_OutputMap[i] = unTotal++; + } + } + return (unTotal - unStart); } @@ -235,6 +244,8 @@ void Map_Vertex_Program(GLcontext *ctx, pAsm->number_used_registers += mesa_vp->Base.NumTemporaries; } + pAsm->flag_reg_index = pAsm->number_used_registers++; + pAsm->uFirstHelpReg = pAsm->number_used_registers; } @@ -324,7 +335,10 @@ struct r700_vertex_program* r700TranslateVertexShader(GLcontext *ctx, return NULL; } - if(GL_FALSE == AssembleInstr(vp->mesa_program->Base.NumInstructions, + InitShaderProgram(&(vp->r700AsmCode)); + + if(GL_FALSE == AssembleInstr(0, + vp->mesa_program->Base.NumInstructions, &(vp->mesa_program->Base.Instructions[0]), &(vp->r700AsmCode)) ) { @@ -336,6 +350,11 @@ struct r700_vertex_program* r700TranslateVertexShader(GLcontext *ctx, return NULL; } + if( GL_FALSE == RelocProgram(&(vp->r700AsmCode)) ) + { + return GL_FALSE; + } + vp->r700Shader.nRegs = (vp->r700AsmCode.number_used_registers == 0) ? 0 : (vp->r700AsmCode.number_used_registers - 1); diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c b/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c index 08e1c5d00d..de18d2ddd6 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c +++ b/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c @@ -196,12 +196,12 @@ 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 +216,9 @@ 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 +227,9 @@ 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 +242,9 @@ 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 +256,9 @@ 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 +273,8 @@ 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 +290,24 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (!rmesa->tcl.tex[unit].buf) emit_tex_vector( ctx, &(rmesa->tcl.aos[nr]), - (char *)VB->TexCoordPtr[unit]->data, - VB->TexCoordPtr[unit]->size, - VB->TexCoordPtr[unit]->stride, + (char *)VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->data, + VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size, + VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->stride, count ); nr++; vfmt |= RADEON_ST_BIT(unit); /* assume we need the 3rd coord if texgen is active for r/q OR at least 3 coords are submitted. This may not be 100% correct */ - if (VB->TexCoordPtr[unit]->size >= 3) { + if (VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size >= 3) { vtx |= RADEON_Q_BIT(unit); vfmt |= RADEON_Q_BIT(unit); } if ( (ctx->Texture.Unit[unit].TexGenEnabled & (R_BIT | Q_BIT)) ) vtx |= RADEON_Q_BIT(unit); - else if ((VB->TexCoordPtr[unit]->size >= 3) && + else if ((VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size >= 3) && ((ctx->Texture.Unit[unit]._ReallyEnabled & (TEXTURE_CUBE_BIT)) == 0)) { - GLuint swaptexmatcol = (VB->TexCoordPtr[unit]->size - 3); + GLuint swaptexmatcol = (VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size - 3); if (((rmesa->NeedTexMatrix >> unit) & 1) && (swaptexmatcol != ((rmesa->TexMatColSwap >> unit) & 1))) radeonUploadTexMatrix( rmesa, unit, swaptexmatcol ) ; diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h b/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h index 515783135d..d764ccb982 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h +++ b/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h @@ -56,18 +56,18 @@ static void TAG(emit)( GLcontext *ctx, radeon_print(RADEON_SWRENDER, RADEON_VERBOSE, "%s\n", __FUNCTION__); - coord = (GLuint (*)[4])VB->ObjPtr->data; - coord_stride = VB->ObjPtr->stride; + coord = (GLuint (*)[4])VB->AttribPtr[_TNL_ATTRIB_POS]->data; + coord_stride = VB->AttribPtr[_TNL_ATTRIB_POS]->stride; if (DO_TEX2) { - if (VB->TexCoordPtr[2]) { + if (VB->AttribPtr[_TNL_ATTRIB_TEX2]) { const GLuint t2 = GET_TEXSOURCE(2); - tc2 = (GLuint (*)[4])VB->TexCoordPtr[t2]->data; - tc2_stride = VB->TexCoordPtr[t2]->stride; - if (DO_PTEX && VB->TexCoordPtr[t2]->size < 3) { + tc2 = (GLuint (*)[4])VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->data; + tc2_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->stride; + if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->size < 3) { fill_tex |= (1<<2); } - else if (DO_PTEX && VB->TexCoordPtr[t2]->size < 4) { + else if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->size < 4) { rqcoordsnoswap |= (1<<2); } } else { @@ -77,14 +77,14 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_TEX1) { - if (VB->TexCoordPtr[1]) { + if (VB->AttribPtr[_TNL_ATTRIB_TEX1]) { const GLuint t1 = GET_TEXSOURCE(1); - tc1 = (GLuint (*)[4])VB->TexCoordPtr[t1]->data; - tc1_stride = VB->TexCoordPtr[t1]->stride; - if (DO_PTEX && VB->TexCoordPtr[t1]->size < 3) { + tc1 = (GLuint (*)[4])VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->data; + tc1_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->stride; + if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->size < 3) { fill_tex |= (1<<1); } - else if (DO_PTEX && VB->TexCoordPtr[t1]->size < 4) { + else if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->size < 4) { rqcoordsnoswap |= (1<<1); } } else { @@ -94,14 +94,14 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_TEX0) { - if (VB->TexCoordPtr[0]) { + if (VB->AttribPtr[_TNL_ATTRIB_TEX0]) { const GLuint t0 = GET_TEXSOURCE(0); - tc0_stride = VB->TexCoordPtr[t0]->stride; - tc0 = (GLuint (*)[4])VB->TexCoordPtr[t0]->data; - if (DO_PTEX && VB->TexCoordPtr[t0]->size < 3) { + tc0_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->stride; + tc0 = (GLuint (*)[4])VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->data; + if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->size < 3) { fill_tex |= (1<<0); } - else if (DO_PTEX && VB->TexCoordPtr[t0]->size < 4) { + else if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->size < 4) { rqcoordsnoswap |= (1<<0); } } else { @@ -112,9 +112,9 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_NORM) { - if (VB->NormalPtr) { - norm_stride = VB->NormalPtr->stride; - norm = (GLuint (*)[4])VB->NormalPtr->data; + if (VB->AttribPtr[_TNL_ATTRIB_NORMAL]) { + norm_stride = VB->AttribPtr[_TNL_ATTRIB_NORMAL]->stride; + norm = (GLuint (*)[4])VB->AttribPtr[_TNL_ATTRIB_NORMAL]->data; } else { norm_stride = 0; norm = (GLuint (*)[4])&ctx->Current.Attrib[VERT_ATTRIB_NORMAL]; @@ -122,9 +122,9 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_RGBA) { - if (VB->ColorPtr[0]) { - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + if (VB->AttribPtr[_TNL_ATTRIB_COLOR0]) { + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; } else { col = (GLfloat (*)[4])ctx->Current.Attrib[VERT_ATTRIB_COLOR0]; col_stride = 0; @@ -132,9 +132,9 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_SPEC_OR_FOG) { - if (VB->SecondaryColorPtr[0]) { - spec = VB->SecondaryColorPtr[0]->data; - spec_stride = VB->SecondaryColorPtr[0]->stride; + if (VB->AttribPtr[_TNL_ATTRIB_COLOR1]) { + spec = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->data; + spec_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->stride; } else { spec = (GLfloat (*)[4])ctx->Current.Attrib[VERT_ATTRIB_COLOR1]; spec_stride = 0; @@ -142,9 +142,9 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_SPEC_OR_FOG) { - if (VB->FogCoordPtr) { - fog = VB->FogCoordPtr->data; - fog_stride = VB->FogCoordPtr->stride; + if (VB->AttribPtr[_TNL_ATTRIB_FOG]) { + fog = VB->AttribPtr[_TNL_ATTRIB_FOG]->data; + fog_stride = VB->AttribPtr[_TNL_ATTRIB_FOG]->stride; } else { fog = (GLfloat (*)[4])ctx->Current.Attrib[VERT_ATTRIB_FOG]; fog_stride = 0; diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c index 78ec119302..5ed11d0a9d 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: @@ -416,7 +416,7 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) break; } - VB->ObjPtr = &rmesa->tcl.ObjClean; + VB->AttribPtr[_TNL_ATTRIB_POS] = &rmesa->tcl.ObjClean; } diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.c b/src/mesa/drivers/dri/radeon/radeon_swtcl.c index e61f59eaea..6bbe8e252e 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: diff --git a/src/mesa/drivers/dri/savage/savagerender.c b/src/mesa/drivers/dri/savage/savagerender.c index 32c74f9467..8221edf387 100644 --- a/src/mesa/drivers/dri/savage/savagerender.c +++ b/src/mesa/drivers/dri/savage/savagerender.c @@ -252,13 +252,13 @@ static GLboolean run_texnorm_stage( GLcontext *ctx, const GLboolean normalizeS = (texObj->WrapS == GL_REPEAT); const GLboolean normalizeT = (reallyEnabled & TEXTURE_2D_BIT) && (texObj->WrapT == GL_REPEAT); - const GLfloat *in = (GLfloat *)VB->TexCoordPtr[i]->data; - const GLint instride = VB->TexCoordPtr[i]->stride; + const GLfloat *in = (GLfloat *)VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->data; + const GLint instride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->stride; GLfloat (*out)[4] = store->texcoord[i].data; GLint j; if (!ctx->Texture.Unit[i]._ReallyEnabled || - VB->TexCoordPtr[i]->size == 4) + VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->size == 4) /* Never try to normalize homogenous tex coords! */ continue; @@ -297,7 +297,7 @@ static GLboolean run_texnorm_stage( GLcontext *ctx, } if (normalizeS || normalizeT) - VB->AttribPtr[VERT_ATTRIB_TEX0+i] = VB->TexCoordPtr[i] = &store->texcoord[i]; + VB->AttribPtr[_TNL_ATTRIB_TEX0 + i] = &store->texcoord[i]; } } diff --git a/src/mesa/drivers/dri/savage/savagetris.c b/src/mesa/drivers/dri/savage/savagetris.c index c04763b40e..e9529d1939 100644 --- a/src/mesa/drivers/dri/savage/savagetris.c +++ b/src/mesa/drivers/dri/savage/savagetris.c @@ -879,13 +879,13 @@ static GLboolean savageCheckPTexHack( GLcontext *ctx ) RENDERINPUTS_COPY( index_bitset, tnl->render_inputs_bitset ); - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 ) && VB->TexCoordPtr[0]->size == 4) { + if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 ) && VB->AttribPtr[_TNL_ATTRIB_TEX0]->size == 4) { if (!RENDERINPUTS_TEST_RANGE( index_bitset, _TNL_ATTRIB_TEX1, _TNL_LAST_TEX )) return GL_TRUE; /* apply ptex hack */ else FALLBACK(ctx, SAVAGE_FALLBACK_PROJ_TEXTURE, GL_TRUE); } - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX1 ) && VB->TexCoordPtr[1]->size == 4) + if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX1 ) && VB->AttribPtr[_TNL_ATTRIB_TEX1]->size == 4) FALLBACK(ctx, SAVAGE_FALLBACK_PROJ_TEXTURE, GL_TRUE); return GL_FALSE; /* don't apply ptex hack */ @@ -976,13 +976,13 @@ static INLINE GLuint savageChooseVertexFormat_s3d( GLcontext *ctx ) if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 )) { if (imesa->ptexHack) EMIT_ATTR( _TNL_ATTRIB_TEX0, EMIT_3F_XYW, SAVAGE_EMIT_STQ0, SAVAGE_SKIP_ST0); - else if (VB->TexCoordPtr[0]->size == 4) + else if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size == 4) assert (0); /* should be caught by savageCheckPTexHack */ - else if (VB->TexCoordPtr[0]->size >= 2) + else if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size >= 2) /* The chromium menu emits some 3D tex coords even though no * 3D texture is enabled. Ignore the 3rd coordinate. */ EMIT_ATTR( _TNL_ATTRIB_TEX0, EMIT_2F, SAVAGE_EMIT_ST0, SAVAGE_SKIP_ST0 ); - else if (VB->TexCoordPtr[0]->size == 1) { + else if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size == 1) { EMIT_ATTR( _TNL_ATTRIB_TEX0, EMIT_1F, SAVAGE_EMIT_S0, SAVAGE_SKIP_S0 ); EMIT_PAD( 4 ); } else @@ -1025,9 +1025,9 @@ static INLINE GLuint savageChooseVertexFormat_s4( GLcontext *ctx ) if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 )) { if (imesa->ptexHack) NEED_ATTR( SAVAGE_EMIT_STQ0, SAVAGE_SKIP_ST0); - else if (VB->TexCoordPtr[0]->size == 4) + else if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size == 4) assert (0); /* should be caught by savageCheckPTexHack */ - else if (VB->TexCoordPtr[0]->size >= 2) + else if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size >= 2) /* The chromium menu emits some 3D tex coords even though no * 3D texture is enabled. Ignore the 3rd coordinate. */ NEED_ATTR( SAVAGE_EMIT_ST0, SAVAGE_SKIP_ST0 ); @@ -1035,10 +1035,10 @@ static INLINE GLuint savageChooseVertexFormat_s4( GLcontext *ctx ) NEED_ATTR( SAVAGE_EMIT_S0, SAVAGE_SKIP_S0 ); } if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX1 )) { - if (VB->TexCoordPtr[1]->size == 4) + if (VB->AttribPtr[_TNL_ATTRIB_TEX1]->size == 4) /* projective textures are not supported by the hardware */ assert (0); /* should be caught by savageCheckPTexHack */ - else if (VB->TexCoordPtr[1]->size >= 2) + else if (VB->AttribPtr[_TNL_ATTRIB_TEX1]->size >= 2) NEED_ATTR( SAVAGE_EMIT_ST1, SAVAGE_SKIP_ST1 ); else NEED_ATTR( SAVAGE_EMIT_S1, SAVAGE_SKIP_S1 ); diff --git a/src/mesa/drivers/dri/sis/sis_tris.c b/src/mesa/drivers/dri/sis/sis_tris.c index 76d12d07b3..3cf10007b5 100644 --- a/src/mesa/drivers/dri/sis/sis_tris.c +++ b/src/mesa/drivers/dri/sis/sis_tris.c @@ -903,14 +903,14 @@ static void sisRenderStart( GLcontext *ctx ) /* projective textures are not supported by the hardware */ if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 )) { - if (VB->TexCoordPtr[0]->size > 2) + if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size > 2) tex_fallback = GL_TRUE; EMIT_ATTR(_TNL_ATTRIB_TEX0, EMIT_2F); AGPParseSet |= SiS_PS_HAS_UV0; } /* Will only hit tex1 on SiS300 */ if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX1 )) { - if (VB->TexCoordPtr[1]->size > 2) + if (VB->AttribPtr[_TNL_ATTRIB_TEX1]->size > 2) tex_fallback = GL_TRUE; EMIT_ATTR(_TNL_ATTRIB_TEX1, EMIT_2F); AGPParseSet |= SiS_PS_HAS_UV1; diff --git a/src/mesa/drivers/dri/tdfx/tdfx_vb.c b/src/mesa/drivers/dri/tdfx/tdfx_vb.c index 4928802232..c200ba3255 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_vb.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_vb.c @@ -69,11 +69,11 @@ static void interp_extras( GLcontext *ctx, /*fprintf(stderr, "%s\n", __FUNCTION__);*/ - if (VB->ColorPtr[1]) { + if (VB->BackfaceColorPtr) { INTERP_4F( t, - GET_COLOR(VB->ColorPtr[1], dst), - GET_COLOR(VB->ColorPtr[1], out), - GET_COLOR(VB->ColorPtr[1], in) ); + GET_COLOR(VB->BackfaceColorPtr, dst), + GET_COLOR(VB->BackfaceColorPtr, out), + GET_COLOR(VB->BackfaceColorPtr, in) ); } if (VB->EdgeFlag) { @@ -88,9 +88,9 @@ static void copy_pv_extras( GLcontext *ctx, GLuint dst, GLuint src ) { struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; - if (VB->ColorPtr[1]) { - COPY_4FV( GET_COLOR(VB->ColorPtr[1], dst), - GET_COLOR(VB->ColorPtr[1], src) ); + if (VB->BackfaceColorPtr) { + COPY_4FV( GET_COLOR(VB->BackfaceColorPtr, dst), + GET_COLOR(VB->BackfaceColorPtr, src) ); } setup_tab[TDFX_CONTEXT(ctx)->SetupIndex].copy_pv(ctx, dst, src); diff --git a/src/mesa/drivers/dri/tdfx/tdfx_vbtmp.h b/src/mesa/drivers/dri/tdfx/tdfx_vbtmp.h index 9b780761f4..19baf7d0d2 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_vbtmp.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_vbtmp.h @@ -58,32 +58,32 @@ static void TAG(emit)( GLcontext *ctx, /* fprintf(stderr, "%s\n", __FUNCTION__); */ if (IND & TDFX_TEX0_BIT) { - tc0_stride = VB->TexCoordPtr[tmu0_source]->stride; - tc0 = VB->TexCoordPtr[tmu0_source]->data; + tc0_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu0_source]->stride; + tc0 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu0_source]->data; u0scale = fxMesa->sScale0; v0scale = fxMesa->tScale0; if (IND & TDFX_PTEX_BIT) - tc0_size = VB->TexCoordPtr[tmu0_source]->size; + tc0_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu0_source]->size; } if (IND & TDFX_TEX1_BIT) { - tc1 = VB->TexCoordPtr[tmu1_source]->data; - tc1_stride = VB->TexCoordPtr[tmu1_source]->stride; + tc1 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu1_source]->data; + tc1_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu1_source]->stride; u1scale = fxMesa->sScale1; v1scale = fxMesa->tScale1; if (IND & TDFX_PTEX_BIT) - tc1_size = VB->TexCoordPtr[tmu1_source]->size; + tc1_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu1_source]->size; } if (IND & TDFX_RGBA_BIT) { - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; - col_size = VB->ColorPtr[0]->size; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; + col_size = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->size; } if (IND & TDFX_FOGC_BIT) { - fog = VB->FogCoordPtr->data; - fog_stride = VB->FogCoordPtr->stride; + fog = VB->AttribPtr[_TNL_ATTRIB_FOG]->data; + fog_stride = VB->AttribPtr[_TNL_ATTRIB_FOG]->stride; } { @@ -168,14 +168,14 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; if (IND & TDFX_TEX1_BIT) { - if (VB->TexCoordPtr[0] == 0) - VB->TexCoordPtr[0] = VB->TexCoordPtr[1]; + if (VB->AttribPtr[_TNL_ATTRIB_TEX0] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0] = VB->AttribPtr[_TNL_ATTRIB_TEX1]; - if (VB->TexCoordPtr[1]->size == 4) + if (VB->AttribPtr[_TNL_ATTRIB_TEX1]->size == 4) return GL_FALSE; } - if (VB->TexCoordPtr[0]->size == 4) + if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size == 4) return GL_FALSE; } diff --git a/src/mesa/drivers/dri/unichrome/via_tris.c b/src/mesa/drivers/dri/unichrome/via_tris.c index 79e67620c9..ab457d41dc 100644 --- a/src/mesa/drivers/dri/unichrome/via_tris.c +++ b/src/mesa/drivers/dri/unichrome/via_tris.c @@ -832,13 +832,13 @@ static GLboolean viaCheckPTexHack( GLcontext *ctx ) RENDERINPUTS_COPY( index_bitset, tnl->render_inputs_bitset ); - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 ) && VB->TexCoordPtr[0]->size == 4) { + if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 ) && VB->AttribPtr[_TNL_ATTRIB_TEX0]->size == 4) { if (!RENDERINPUTS_TEST_RANGE( index_bitset, _TNL_ATTRIB_TEX1, _TNL_LAST_TEX )) ptexHack = GL_TRUE; else fallback = GL_TRUE; } - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX1 ) && VB->TexCoordPtr[1]->size == 4) + if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX1 ) && VB->AttribPtr[_TNL_ATTRIB_TEX1]->size == 4) fallback = GL_TRUE; FALLBACK(VIA_CONTEXT(ctx), VIA_FALLBACK_PROJ_TEXTURE, fallback); |