diff options
Diffstat (limited to 'src/mesa/drivers/dri')
24 files changed, 321 insertions, 412 deletions
diff --git a/src/mesa/drivers/dri/common/extension_helper.h b/src/mesa/drivers/dri/common/extension_helper.h index 10f75edaaa..bf103a3931 100644 --- a/src/mesa/drivers/dri/common/extension_helper.h +++ b/src/mesa/drivers/dri/common/extension_helper.h @@ -1478,9 +1478,10 @@ static const char ImageTransformParameterfvHP_names[] = ""; #endif -#if defined(need_GL_ARB_vertex_program) +#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) static const char VertexAttrib4ivARB_names[] = "ip\0" /* Parameter signature */ + "glVertexAttrib4iv\0" "glVertexAttrib4ivARB\0" ""; #endif @@ -1586,9 +1587,10 @@ static const char PixelTransformParameterfvEXT_names[] = ""; #endif -#if defined(need_GL_ARB_vertex_program) +#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) static const char VertexAttrib4bvARB_names[] = "ip\0" /* Parameter signature */ + "glVertexAttrib4bv\0" "glVertexAttrib4bvARB\0" ""; #endif @@ -2391,9 +2393,10 @@ static const char GetAttribLocationARB_names[] = ""; #endif -#if defined(need_GL_ARB_vertex_program) +#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) static const char VertexAttrib4ubvARB_names[] = "ip\0" /* Parameter signature */ + "glVertexAttrib4ubv\0" "glVertexAttrib4ubvARB\0" ""; #endif @@ -2910,9 +2913,10 @@ static const char ReplacementCodeuiColor4ubVertex3fSUN_names[] = ""; #endif -#if defined(need_GL_ARB_vertex_program) +#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) static const char VertexAttrib4usvARB_names[] = "ip\0" /* Parameter signature */ + "glVertexAttrib4usv\0" "glVertexAttrib4usvARB\0" ""; #endif @@ -4393,9 +4397,10 @@ static const char WindowPos4iMESA_names[] = ""; #endif -#if defined(need_GL_ARB_vertex_program) +#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) static const char VertexAttrib4uivARB_names[] = "ip\0" /* Parameter signature */ + "glVertexAttrib4uiv\0" "glVertexAttrib4uivARB\0" ""; #endif @@ -6257,6 +6262,8 @@ static const struct dri_extension_function GL_VERSION_2_0_functions[] = { { GetVertexAttribivARB_names, GetVertexAttribivARB_remap_index, -1 }, { CreateProgram_names, CreateProgram_remap_index, -1 }, { StencilFuncSeparate_names, StencilFuncSeparate_remap_index, -1 }, + { VertexAttrib4ivARB_names, VertexAttrib4ivARB_remap_index, -1 }, + { VertexAttrib4bvARB_names, VertexAttrib4bvARB_remap_index, -1 }, { VertexAttrib3dARB_names, VertexAttrib3dARB_remap_index, -1 }, { VertexAttrib4fARB_names, VertexAttrib4fARB_remap_index, -1 }, { VertexAttrib4fvARB_names, VertexAttrib4fvARB_remap_index, -1 }, @@ -6270,6 +6277,7 @@ static const struct dri_extension_function GL_VERSION_2_0_functions[] = { { VertexAttrib1dvARB_names, VertexAttrib1dvARB_remap_index, -1 }, { GetVertexAttribfvARB_names, GetVertexAttribfvARB_remap_index, -1 }, { GetAttribLocationARB_names, GetAttribLocationARB_remap_index, -1 }, + { VertexAttrib4ubvARB_names, VertexAttrib4ubvARB_remap_index, -1 }, { Uniform3ivARB_names, Uniform3ivARB_remap_index, -1 }, { VertexAttrib4sARB_names, VertexAttrib4sARB_remap_index, -1 }, { VertexAttrib2dvARB_names, VertexAttrib2dvARB_remap_index, -1 }, @@ -6282,6 +6290,7 @@ static const struct dri_extension_function GL_VERSION_2_0_functions[] = { { VertexAttrib4NuivARB_names, VertexAttrib4NuivARB_remap_index, -1 }, { Uniform4fARB_names, Uniform4fARB_remap_index, -1 }, { VertexAttrib1dARB_names, VertexAttrib1dARB_remap_index, -1 }, + { VertexAttrib4usvARB_names, VertexAttrib4usvARB_remap_index, -1 }, { LinkProgramARB_names, LinkProgramARB_remap_index, -1 }, { ShaderSourceARB_names, ShaderSourceARB_remap_index, -1 }, { VertexAttrib3svARB_names, VertexAttrib3svARB_remap_index, -1 }, @@ -6311,6 +6320,7 @@ static const struct dri_extension_function GL_VERSION_2_0_functions[] = { { DrawBuffersARB_names, DrawBuffersARB_remap_index, -1 }, { Uniform1fvARB_names, Uniform1fvARB_remap_index, -1 }, { EnableVertexAttribArrayARB_names, EnableVertexAttribArrayARB_remap_index, -1 }, + { VertexAttrib4uivARB_names, VertexAttrib4uivARB_remap_index, -1 }, { VertexAttrib4svARB_names, VertexAttrib4svARB_remap_index, -1 }, { GetShaderiv_names, GetShaderiv_remap_index, -1 }, { VertexAttrib2svARB_names, VertexAttrib2svARB_remap_index, -1 }, diff --git a/src/mesa/drivers/dri/i915/intel_pixel.c b/src/mesa/drivers/dri/i915/intel_pixel.c index 535cbfcb26..d175870a0c 100644 --- a/src/mesa/drivers/dri/i915/intel_pixel.c +++ b/src/mesa/drivers/dri/i915/intel_pixel.c @@ -439,10 +439,26 @@ intelDrawPixels( GLcontext *ctx, if (INTEL_DEBUG & DEBUG_PIXEL) fprintf(stderr, "%s\n", __FUNCTION__); - if (!intelTryDrawPixels( ctx, x, y, width, height, format, type, - unpack, pixels )) + if (intelTryDrawPixels( ctx, x, y, width, height, format, type, + unpack, pixels )) + return; + + if (ctx->FragmentProgram._Current == ctx->FragmentProgram._TexEnvProgram) { + /* + * We don't want the i915 texenv program to be applied to DrawPixels. + * This is really just a performance optimization (mesa will other- + * wise happily run the fragment program on each pixel in the image). + */ + struct gl_fragment_program *fpSave = ctx->FragmentProgram._Current; + ctx->FragmentProgram._Current = NULL; + _swrast_DrawPixels( ctx, x, y, width, height, format, type, + unpack, pixels ); + ctx->FragmentProgram._Current = fpSave; + } + else { _swrast_DrawPixels( ctx, x, y, width, height, format, type, - unpack, pixels ); + unpack, pixels ); + } } diff --git a/src/mesa/drivers/dri/i915tex/i830_vtbl.c b/src/mesa/drivers/dri/i915tex/i830_vtbl.c index dd0670dec3..e432648ada 100644 --- a/src/mesa/drivers/dri/i915tex/i830_vtbl.c +++ b/src/mesa/drivers/dri/i915tex/i830_vtbl.c @@ -490,11 +490,13 @@ i830_emit_state(struct intel_context *intel) DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, state->tex_offset[i] | TM0S0_USE_FENCE); } - else { - assert(i == 0); - assert(state == &i830->meta); - OUT_BATCH(0); - } + else if (state == &i830->meta) { + assert(i == 0); + OUT_BATCH(0); + } + else { + OUT_BATCH(state->tex_offset[i]); + } OUT_BATCH(state->Tex[i][I830_TEXREG_TM0S1]); OUT_BATCH(state->Tex[i][I830_TEXREG_TM0S2]); diff --git a/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c b/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c index 10a079896a..77c67c821e 100644 --- a/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c +++ b/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c @@ -363,5 +363,20 @@ intelDrawPixels(GLcontext * ctx, if (INTEL_DEBUG & DEBUG_PIXEL) _mesa_printf("%s: fallback to swrast\n", __FUNCTION__); - _swrast_DrawPixels(ctx, x, y, width, height, format, type, unpack, pixels); + if (ctx->FragmentProgram._Current == ctx->FragmentProgram._TexEnvProgram) { + /* + * We don't want the i915 texenv program to be applied to DrawPixels. + * This is really just a performance optimization (mesa will other- + * wise happily run the fragment program on each pixel in the image). + */ + struct gl_fragment_program *fpSave = ctx->FragmentProgram._Current; + ctx->FragmentProgram._Current = NULL; + _swrast_DrawPixels( ctx, x, y, width, height, format, type, + unpack, pixels ); + ctx->FragmentProgram._Current = fpSave; + } + else { + _swrast_DrawPixels( ctx, x, y, width, height, format, type, + unpack, pixels ); + } } diff --git a/src/mesa/drivers/dri/nouveau/nouveau_bufferobj.c b/src/mesa/drivers/dri/nouveau/nouveau_bufferobj.c index 684ed7b017..fc14060c04 100644 --- a/src/mesa/drivers/dri/nouveau/nouveau_bufferobj.c +++ b/src/mesa/drivers/dri/nouveau/nouveau_bufferobj.c @@ -41,7 +41,7 @@ nouveau_bo_download_from_screen(GLcontext *ctx, GLuint offset, GLuint size, DEBUG("..sys_mem\n"); in_mem = nouveau_mem_alloc(ctx, NOUVEAU_MEM_AGP, size, 0); if (in_mem) { - DEBUG("....via AGP\n"); + DEBUG("....via GART\n"); /* otherwise, try blitting to faster memory and * copying from there */ @@ -86,7 +86,7 @@ nouveau_bo_upload_to_screen(GLcontext *ctx, GLuint offset, GLuint size, NOUVEAU_MEM_MAPPED, size, 0); if (out_mem) { - DEBUG("....via AGP\n"); + DEBUG("....via GART\n"); _mesa_memcpy(out_mem->map, nbo->cpu_mem_sys + offset, size); nouveau_memformat_flat_emit(ctx, nbo->gpu_mem, out_mem, @@ -511,7 +511,7 @@ nouveauBufferData(GLcontext *ctx, GLenum target, GLsizeiptrARB size, gpu_flags = 0; break; default: - gpu_flags = NOUVEAU_BO_VRAM_OK | NOUVEAU_BO_AGP_OK; + gpu_flags = NOUVEAU_BO_VRAM_OK | NOUVEAU_BO_GART_OK; break; } nouveau_bo_init_storage(ctx, gpu_flags, size, data, usage, obj); diff --git a/src/mesa/drivers/dri/nouveau/nouveau_bufferobj.h b/src/mesa/drivers/dri/nouveau/nouveau_bufferobj.h index 932450fd87..3439a35e7c 100644 --- a/src/mesa/drivers/dri/nouveau/nouveau_bufferobj.h +++ b/src/mesa/drivers/dri/nouveau/nouveau_bufferobj.h @@ -5,7 +5,7 @@ #include "nouveau_buffers.h" #define NOUVEAU_BO_VRAM_OK (NOUVEAU_MEM_FB | NOUVEAU_MEM_FB_ACCEPTABLE) -#define NOUVEAU_BO_AGP_OK (NOUVEAU_MEM_AGP | NOUVEAU_MEM_AGP_ACCEPTABLE) +#define NOUVEAU_BO_GART_OK (NOUVEAU_MEM_AGP | NOUVEAU_MEM_AGP_ACCEPTABLE) typedef struct nouveau_bufferobj_region_t { uint32_t start; diff --git a/src/mesa/drivers/dri/nouveau/nouveau_buffers.c b/src/mesa/drivers/dri/nouveau/nouveau_buffers.c index b54f68f402..857cd30584 100644 --- a/src/mesa/drivers/dri/nouveau/nouveau_buffers.c +++ b/src/mesa/drivers/dri/nouveau/nouveau_buffers.c @@ -32,8 +32,8 @@ nouveau_memformat_flat_emit(GLcontext *ctx, return GL_FALSE; } - src_handle = (src->type & NOUVEAU_MEM_FB) ? NvDmaFB : NvDmaAGP; - dst_handle = (dst->type & NOUVEAU_MEM_FB) ? NvDmaFB : NvDmaAGP; + src_handle = (src->type & NOUVEAU_MEM_FB) ? NvDmaFB : NvDmaTT; + dst_handle = (dst->type & NOUVEAU_MEM_FB) ? NvDmaFB : NvDmaTT; src_offset += nouveau_mem_gpu_offset_get(ctx, src); dst_offset += nouveau_mem_gpu_offset_get(ctx, dst); @@ -138,7 +138,7 @@ nouveau_mem_gpu_offset_get(GLcontext *ctx, nouveau_mem *mem) if (mem->type & NOUVEAU_MEM_FB) return (uint32_t)mem->offset - nmesa->vram_phys; else if (mem->type & NOUVEAU_MEM_AGP) - return (uint32_t)mem->offset - nmesa->agp_phys; + return (uint32_t)mem->offset - nmesa->gart_phys; else return 0xDEADF00D; } @@ -299,6 +299,8 @@ nouveau_cliprects_drawable_set(nouveauContextPtr nmesa, nmesa->pClipRects = dPriv->pClipRects; nmesa->drawX = dPriv->x; nmesa->drawY = dPriv->y; + nmesa->drawW = dPriv->w; + nmesa->drawH = dPriv->h; } static void @@ -313,6 +315,8 @@ nouveau_cliprects_renderbuffer_set(nouveauContextPtr nmesa, nmesa->osClipRect.y2 = nrb->mesa.Height; nmesa->drawX = 0; nmesa->drawY = 0; + nmesa->drawW = nrb->mesa.Width; + nmesa->drawH = nrb->mesa.Height; } void diff --git a/src/mesa/drivers/dri/nouveau/nouveau_context.c b/src/mesa/drivers/dri/nouveau/nouveau_context.c index 8e11eb6134..d96b00242c 100644 --- a/src/mesa/drivers/dri/nouveau/nouveau_context.c +++ b/src/mesa/drivers/dri/nouveau/nouveau_context.c @@ -145,10 +145,10 @@ GLboolean nouveauCreateContext( const __GLcontextModes *glVisual, &nmesa->vram_size)) return GL_FALSE; if (!nouveauDRMGetParam(nmesa, NOUVEAU_GETPARAM_AGP_PHYSICAL, - &nmesa->agp_phys)) + &nmesa->gart_phys)) return GL_FALSE; if (!nouveauDRMGetParam(nmesa, NOUVEAU_GETPARAM_AGP_SIZE, - &nmesa->agp_size)) + &nmesa->gart_size)) return GL_FALSE; if (!nouveauFifoInit(nmesa)) return GL_FALSE; diff --git a/src/mesa/drivers/dri/nouveau/nouveau_context.h b/src/mesa/drivers/dri/nouveau/nouveau_context.h index 87e4479da3..10d2ed6e17 100644 --- a/src/mesa/drivers/dri/nouveau/nouveau_context.h +++ b/src/mesa/drivers/dri/nouveau/nouveau_context.h @@ -99,19 +99,22 @@ typedef struct nouveau_context { /* The read-only regs */ volatile unsigned char* mmio; + /* The per-channel notifier block */ + volatile void *notifier_block; + /* Physical addresses of AGP/VRAM apertures */ uint64_t vram_phys; uint64_t vram_size; - uint64_t agp_phys; - uint64_t agp_size; + uint64_t gart_phys; + uint64_t gart_size; /* Channel synchronisation */ - nouveau_notifier *syncNotifier; + drm_nouveau_notifier_alloc_t *syncNotifier; /* ARB_occlusion_query / EXT_timer_query */ GLuint query_object_max; GLboolean * query_alloc; - nouveau_notifier *queryNotifier; + drm_nouveau_notifier_alloc_t *queryNotifier; /* Additional hw-specific functions */ nouveau_hw_func hw_func; @@ -150,7 +153,7 @@ typedef struct nouveau_context { GLuint numClipRects; drm_clip_rect_t *pClipRects; drm_clip_rect_t osClipRect; - GLuint drawX, drawY; + GLuint drawX, drawY, drawW, drawH; /* The rendering context information */ GLenum current_primitive; /* the current primitive enum */ diff --git a/src/mesa/drivers/dri/nouveau/nouveau_fifo.c b/src/mesa/drivers/dri/nouveau/nouveau_fifo.c index bd2b2eddd0..e9320918f9 100644 --- a/src/mesa/drivers/dri/nouveau/nouveau_fifo.c +++ b/src/mesa/drivers/dri/nouveau/nouveau_fifo.c @@ -99,13 +99,14 @@ void nouveauWaitForIdle(nouveauContextPtr nmesa) GLboolean nouveauFifoInit(nouveauContextPtr nmesa) { drm_nouveau_fifo_alloc_t fifo_init; - int i; + int i, ret; #ifdef NOUVEAU_RING_DEBUG return GL_TRUE; #endif - int ret; + fifo_init.fb_ctxdma_handle = NvDmaFB; + fifo_init.tt_ctxdma_handle = NvDmaTT; ret=drmCommandWriteRead(nmesa->driFd, DRM_NOUVEAU_FIFO_ALLOC, &fifo_init, sizeof(fifo_init)); if (ret) { FATAL("Fifo initialization ioctl failed (returned %d)\n",ret); @@ -117,12 +118,21 @@ GLboolean nouveauFifoInit(nouveauContextPtr nmesa) FATAL("Unable to map the fifo (returned %d)\n",ret); return GL_FALSE; } + ret = drmMap(nmesa->driFd, fifo_init.ctrl, fifo_init.ctrl_size, &nmesa->fifo.mmio); if (ret) { FATAL("Unable to map the control regs (returned %d)\n",ret); return GL_FALSE; } + ret = drmMap(nmesa->driFd, fifo_init.notifier, + fifo_init.notifier_size, + &nmesa->notifier_block); + if (ret) { + FATAL("Unable to map the notifier block (returned %d)\n",ret); + return GL_FALSE; + } + /* Setup our initial FIFO tracking params */ nmesa->fifo.channel = fifo_init.channel; nmesa->fifo.put_base = fifo_init.put_base; diff --git a/src/mesa/drivers/dri/nouveau/nouveau_object.c b/src/mesa/drivers/dri/nouveau/nouveau_object.c index b71acff430..69f8dbf794 100644 --- a/src/mesa/drivers/dri/nouveau/nouveau_object.c +++ b/src/mesa/drivers/dri/nouveau/nouveau_object.c @@ -7,61 +7,18 @@ GLboolean nouveauCreateContextObject(nouveauContextPtr nmesa, uint32_t handle, int class) { - drm_nouveau_object_init_t cto; + drm_nouveau_grobj_alloc_t cto; int ret; cto.channel = nmesa->fifo.channel; cto.handle = handle; cto.class = class; - ret = drmCommandWrite(nmesa->driFd, DRM_NOUVEAU_OBJECT_INIT, &cto, sizeof(cto)); + ret = drmCommandWrite(nmesa->driFd, DRM_NOUVEAU_GROBJ_ALLOC, + &cto, sizeof(cto)); return ret == 0; } -GLboolean nouveauCreateDmaObject(nouveauContextPtr nmesa, - uint32_t handle, - int class, - uint32_t offset, - uint32_t size, - int target, - int access) -{ - drm_nouveau_dma_object_init_t dma; - int ret; - - dma.channel = nmesa->fifo.channel; - dma.class = class; - dma.handle = handle; - dma.target = target; - dma.access = access; - dma.offset = offset; - dma.size = size; - ret = drmCommandWriteRead(nmesa->driFd, DRM_NOUVEAU_DMA_OBJECT_INIT, - &dma, sizeof(dma)); - return ret == 0; -} - -GLboolean nouveauCreateDmaObjectFromMem(nouveauContextPtr nmesa, - uint32_t handle, int class, - nouveau_mem *mem, - int access) -{ - uint32_t offset = mem->offset; - int target = mem->type & (NOUVEAU_MEM_FB | NOUVEAU_MEM_AGP); - - if (!target) - return GL_FALSE; - - if (target & NOUVEAU_MEM_FB) - offset -= nmesa->vram_phys; - else if (target & NOUVEAU_MEM_AGP) - offset -= nmesa->agp_phys; - - return nouveauCreateDmaObject(nmesa, handle, class, - offset, mem->size, - target, access); -} - void nouveauObjectOnSubchannel(nouveauContextPtr nmesa, int subchannel, int handle) { BEGIN_RING_SIZE(subchannel, 0, 1); @@ -74,16 +31,6 @@ void nouveauObjectInit(nouveauContextPtr nmesa) return; #endif -/* We need to know vram size.. and AGP size (and even if the card is AGP..) */ - nouveauCreateDmaObject( nmesa, NvDmaFB, NV_DMA_IN_MEMORY, - 0, nmesa->vram_size, - NOUVEAU_MEM_FB, - NOUVEAU_MEM_ACCESS_RW); - nouveauCreateDmaObject( nmesa, NvDmaAGP, NV_DMA_IN_MEMORY, - 0, nmesa->agp_size, - NOUVEAU_MEM_AGP, - NOUVEAU_MEM_ACCESS_RW); - nouveauCreateContextObject(nmesa, Nv3D, nmesa->screen->card->class_3d); if (nmesa->screen->card->type>=NV_10) { nouveauCreateContextObject(nmesa, NvCtxSurf2D, NV10_CONTEXT_SURFACES_2D); diff --git a/src/mesa/drivers/dri/nouveau/nouveau_object.h b/src/mesa/drivers/dri/nouveau/nouveau_object.h index 0be9b4309c..8c72d014da 100644 --- a/src/mesa/drivers/dri/nouveau/nouveau_object.h +++ b/src/mesa/drivers/dri/nouveau/nouveau_object.h @@ -14,7 +14,7 @@ enum DMAObjects { NvMemFormat = 0x80000022, NvCtxSurf3D = 0x80000023, NvDmaFB = 0xD0FB0001, - NvDmaAGP = 0xD0AA0001, + NvDmaTT = 0xD0AA0001, NvSyncNotify = 0xD0000001, NvQueryNotify = 0xD0000002 }; @@ -31,17 +31,5 @@ extern void nouveauObjectOnSubchannel(nouveauContextPtr nmesa, int subchannel, i extern GLboolean nouveauCreateContextObject(nouveauContextPtr nmesa, uint32_t handle, int class); -extern GLboolean nouveauCreateDmaObject(nouveauContextPtr nmesa, - uint32_t handle, - int class, - uint32_t offset, - uint32_t size, - int target, - int access); -extern GLboolean nouveauCreateDmaObjectFromMem(nouveauContextPtr nmesa, - uint32_t handle, - int class, - nouveau_mem *mem, - int access); #endif diff --git a/src/mesa/drivers/dri/nouveau/nouveau_query.c b/src/mesa/drivers/dri/nouveau/nouveau_query.c index de3f5b0378..0154140069 100644 --- a/src/mesa/drivers/dri/nouveau/nouveau_query.c +++ b/src/mesa/drivers/dri/nouveau/nouveau_query.c @@ -68,7 +68,7 @@ nouveauBeginQuery(GLcontext *ctx, GLenum target, struct gl_query_object *q) nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx); nouveau_query_object *nq = (nouveau_query_object *)q; - nouveau_notifier_reset(nmesa->queryNotifier, nq->notifier_id); + nouveau_notifier_reset(ctx, nmesa->queryNotifier, nq->notifier_id); switch (nmesa->screen->card->type) { case NV_20: @@ -105,12 +105,13 @@ nouveauUpdateQuery(GLcontext *ctx, GLenum target, struct gl_query_object *q) nouveau_query_object *nq = (nouveau_query_object *)q; int status; - status = nouveau_notifier_status(nmesa->queryNotifier, + status = nouveau_notifier_status(ctx, nmesa->queryNotifier, nq->notifier_id); q->Ready = (status == NV_NOTIFY_STATE_STATUS_COMPLETED); if (q->Ready) - q->Result = nouveau_notifier_return_val(nmesa->queryNotifier, + q->Result = nouveau_notifier_return_val(ctx, + nmesa->queryNotifier, nq->notifier_id); } @@ -120,7 +121,7 @@ nouveauWaitQueryResult(GLcontext *ctx, GLenum target, struct gl_query_object *q) nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx); nouveau_query_object *nq = (nouveau_query_object *)q; - nouveau_notifier_wait_status(nmesa->queryNotifier, nq->notifier_id, + nouveau_notifier_wait_status(ctx, nmesa->queryNotifier, nq->notifier_id, NV_NOTIFY_STATE_STATUS_COMPLETED, 0); nouveauUpdateQuery(ctx, target, q); } diff --git a/src/mesa/drivers/dri/nouveau/nouveau_screen.c b/src/mesa/drivers/dri/nouveau/nouveau_screen.c index 7a4b9f1cd0..bc7f39b042 100644 --- a/src/mesa/drivers/dri/nouveau/nouveau_screen.c +++ b/src/mesa/drivers/dri/nouveau/nouveau_screen.c @@ -328,7 +328,7 @@ void * __driCreateNewScreen_20050727( __DRInativeDisplay *dpy, int scrn, __DRIsc static const __DRIversion ddx_expected = { 1, 2, 0 }; static const __DRIversion dri_expected = { 4, 0, 0 }; static const __DRIversion drm_expected = { 0, 0, NOUVEAU_DRM_HEADER_PATCHLEVEL }; -#if NOUVEAU_DRM_HEADER_PATCHLEVEL != 6 +#if NOUVEAU_DRM_HEADER_PATCHLEVEL != 7 #error nouveau_drm.h version doesn't match expected version #endif dri_interface = interface; diff --git a/src/mesa/drivers/dri/nouveau/nouveau_span.c b/src/mesa/drivers/dri/nouveau/nouveau_span.c index 74dec66afc..6e3f9fadf4 100644 --- a/src/mesa/drivers/dri/nouveau/nouveau_span.c +++ b/src/mesa/drivers/dri/nouveau/nouveau_span.c @@ -37,6 +37,8 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #define HAVE_HW_STENCIL_SPANS 0 #define HAVE_HW_STENCIL_PIXELS 0 +static char *fake_span[1280*1024*4]; + #define HW_CLIPLOOP() \ do { \ int _nc = nmesa->numClipRects; \ @@ -52,6 +54,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. GLuint height = nrb->mesa.Height; \ GLubyte *map = (GLubyte *)(nrb->map ? nrb->map : nrb->mem->map) + \ (nmesa->drawY * nrb->pitch) + (nmesa->drawX * nrb->cpp); \ + map = fake_span; \ GLuint p; \ (void) p; @@ -120,6 +123,6 @@ nouveauSpanSetFunctions(nouveau_renderbuffer *nrb, const GLvisual *vis) { if (nrb->mesa._ActualFormat == GL_RGBA8) nouveauInitPointers_ARGB8888(&nrb->mesa); - else if (nrb->mesa._ActualFormat == GL_RGB5) + else // if (nrb->mesa._ActualFormat == GL_RGB5) nouveauInitPointers_RGB565(&nrb->mesa); } diff --git a/src/mesa/drivers/dri/nouveau/nouveau_state.c b/src/mesa/drivers/dri/nouveau/nouveau_state.c index e9fd188d73..7cb805902a 100644 --- a/src/mesa/drivers/dri/nouveau/nouveau_state.c +++ b/src/mesa/drivers/dri/nouveau/nouveau_state.c @@ -60,14 +60,14 @@ static void nouveauCalcViewport(GLcontext *ctx) nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx); const GLfloat *v = ctx->Viewport._WindowMap.m; GLfloat *m = nmesa->viewport.m; - GLfloat xoffset = nmesa->drawX, yoffset = nmesa->drawY; + GLfloat xoffset = nmesa->drawX, yoffset = nmesa->drawY + nmesa->drawH; nmesa->depth_scale = 1.0 / ctx->DrawBuffer->_DepthMaxF; m[MAT_SX] = v[MAT_SX]; m[MAT_TX] = v[MAT_TX] + xoffset + SUBPIXEL_X; m[MAT_SY] = - v[MAT_SY]; - m[MAT_TY] = v[MAT_TY] + yoffset + SUBPIXEL_Y; + m[MAT_TY] = (-v[MAT_TY]) + yoffset + SUBPIXEL_Y; m[MAT_SZ] = v[MAT_SZ] * nmesa->depth_scale; m[MAT_TZ] = v[MAT_TZ] * nmesa->depth_scale; diff --git a/src/mesa/drivers/dri/nouveau/nouveau_sync.c b/src/mesa/drivers/dri/nouveau/nouveau_sync.c index 30e6696269..1d1eeede18 100644 --- a/src/mesa/drivers/dri/nouveau/nouveau_sync.c +++ b/src/mesa/drivers/dri/nouveau/nouveau_sync.c @@ -35,53 +35,51 @@ #include "nouveau_msg.h" #include "nouveau_sync.h" -nouveau_notifier * +#define NOTIFIER(__v) \ + nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx); \ + volatile uint32_t *__v = (void*)nmesa->notifier_block + notifier->offset + +drm_nouveau_notifier_alloc_t * nouveau_notifier_new(GLcontext *ctx, GLuint handle, GLuint count) { nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx); - nouveau_notifier *notifier; + drm_nouveau_notifier_alloc_t *notifier; + int ret; #ifdef NOUVEAU_RING_DEBUG return NULL; #endif - notifier = CALLOC_STRUCT(nouveau_notifier_t); + notifier = CALLOC_STRUCT(drm_nouveau_notifier_alloc); if (!notifier) return NULL; - notifier->mem = nouveau_mem_alloc(ctx, - NOUVEAU_MEM_FB | NOUVEAU_MEM_MAPPED, - count * NV_NOTIFIER_SIZE, - 0); - if (!notifier->mem) { - FREE(notifier); - return NULL; - } - - if (!nouveauCreateDmaObjectFromMem(nmesa, handle, NV_DMA_IN_MEMORY, - notifier->mem, - NOUVEAU_MEM_ACCESS_RW)) { - nouveau_mem_free(ctx, notifier->mem); + notifier->channel = nmesa->fifo.channel; + notifier->handle = handle; + notifier->count = count; + ret = drmCommandWriteRead(nmesa->driFd, DRM_NOUVEAU_NOTIFIER_ALLOC, + notifier, sizeof(*notifier)); + if (ret) { + MESSAGE("Failed to create notifier 0x%08x: %d\n", handle, ret); FREE(notifier); return NULL; } - notifier->handle = handle; return notifier; } void -nouveau_notifier_destroy(GLcontext *ctx, nouveau_notifier *notifier) +nouveau_notifier_destroy(GLcontext *ctx, drm_nouveau_notifier_alloc_t *notifier) { - /*XXX: free DMA object.. */ - nouveau_mem_free(ctx, notifier->mem); + /*XXX: free notifier object.. */ FREE(notifier); } void -nouveau_notifier_reset(nouveau_notifier *notifier, GLuint id) +nouveau_notifier_reset(GLcontext *ctx, drm_nouveau_notifier_alloc_t *notifier, + GLuint id) { - volatile GLuint *n = notifier->mem->map + (id * NV_NOTIFIER_SIZE); + NOTIFIER(n); #ifdef NOUVEAU_RING_DEBUG return; @@ -95,26 +93,29 @@ nouveau_notifier_reset(nouveau_notifier *notifier, GLuint id) } GLuint -nouveau_notifier_status(nouveau_notifier *notifier, GLuint id) +nouveau_notifier_status(GLcontext *ctx, drm_nouveau_notifier_alloc_t *notifier, + GLuint id) { - volatile GLuint *n = notifier->mem->map + (id * NV_NOTIFIER_SIZE); + NOTIFIER(n); return n[NV_NOTIFY_STATE/4] >> NV_NOTIFY_STATE_STATUS_SHIFT; } GLuint -nouveau_notifier_return_val(nouveau_notifier *notifier, GLuint id) +nouveau_notifier_return_val(GLcontext *ctx, + drm_nouveau_notifier_alloc_t *notifier, GLuint id) { - volatile GLuint *n = notifier->mem->map + (id * NV_NOTIFIER_SIZE); + NOTIFIER(n); return n[NV_NOTIFY_RETURN_VALUE/4]; } GLboolean -nouveau_notifier_wait_status(nouveau_notifier *notifier, GLuint id, +nouveau_notifier_wait_status(GLcontext *ctx, + drm_nouveau_notifier_alloc_t *notifier, GLuint id, GLuint status, GLuint timeout) { - volatile GLuint *n = notifier->mem->map + (id * NV_NOTIFIER_SIZE); + NOTIFIER(n); unsigned int time = 0; #ifdef NOUVEAU_RING_DEBUG @@ -144,13 +145,13 @@ nouveau_notifier_wait_status(nouveau_notifier *notifier, GLuint id, } void -nouveau_notifier_wait_nop(GLcontext *ctx, nouveau_notifier *notifier, - GLuint subc) +nouveau_notifier_wait_nop(GLcontext *ctx, + drm_nouveau_notifier_alloc_t *notifier, GLuint subc) { - nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx); + NOTIFIER(n); GLboolean ret; - nouveau_notifier_reset(notifier, 0); + nouveau_notifier_reset(ctx, notifier, 0); BEGIN_RING_SIZE(subc, NV_NOTIFY, 1); OUT_RING (NV_NOTIFY_STYLE_WRITE_ONLY); @@ -158,7 +159,7 @@ nouveau_notifier_wait_nop(GLcontext *ctx, nouveau_notifier *notifier, OUT_RING (0); FIRE_RING(); - ret = nouveau_notifier_wait_status(notifier, 0, + ret = nouveau_notifier_wait_status(ctx, notifier, 0, NV_NOTIFY_STATE_STATUS_COMPLETED, 0 /* no timeout */); if (ret == GL_FALSE) MESSAGE("wait on notifier failed\n"); diff --git a/src/mesa/drivers/dri/nouveau/nouveau_sync.h b/src/mesa/drivers/dri/nouveau/nouveau_sync.h index 019d5f6629..b56cc5fb54 100644 --- a/src/mesa/drivers/dri/nouveau/nouveau_sync.h +++ b/src/mesa/drivers/dri/nouveau/nouveau_sync.h @@ -47,21 +47,23 @@ #define NV_NOTIFY 0x00000104 #define NV_NOTIFY_STYLE_WRITE_ONLY 0 -typedef struct nouveau_notifier_t { - GLuint handle; - nouveau_mem *mem; -} nouveau_notifier; - -extern nouveau_notifier *nouveau_notifier_new(GLcontext *, GLuint handle, - GLuint count); -extern void nouveau_notifier_destroy(GLcontext *, nouveau_notifier *); -extern void nouveau_notifier_reset(nouveau_notifier *, GLuint id); -extern GLuint nouveau_notifier_status(nouveau_notifier *, GLuint id); -extern GLuint nouveau_notifier_return_val(nouveau_notifier *, GLuint id); -extern GLboolean nouveau_notifier_wait_status(nouveau_notifier *r, GLuint id, - GLuint status, GLuint timeout); -extern void nouveau_notifier_wait_nop(GLcontext *ctx, - nouveau_notifier *, GLuint subc); +extern drm_nouveau_notifier_alloc_t * +nouveau_notifier_new(GLcontext *, GLuint handle, GLuint count); +extern void +nouveau_notifier_destroy(GLcontext *, drm_nouveau_notifier_alloc_t *); +extern void +nouveau_notifier_reset(GLcontext *, drm_nouveau_notifier_alloc_t *, GLuint id); +extern GLuint +nouveau_notifier_status(GLcontext *, drm_nouveau_notifier_alloc_t *, GLuint id); +extern GLuint +nouveau_notifier_return_val(GLcontext *, drm_nouveau_notifier_alloc_t *, + GLuint id); +extern GLboolean +nouveau_notifier_wait_status(GLcontext *, drm_nouveau_notifier_alloc_t *, + GLuint id, GLuint status, GLuint timeout); +extern void +nouveau_notifier_wait_nop(GLcontext *ctx, drm_nouveau_notifier_alloc_t *, + GLuint subc); extern GLboolean nouveauSyncInitFuncs(GLcontext *ctx); #endif diff --git a/src/mesa/drivers/dri/nouveau/nv10_swtcl.c b/src/mesa/drivers/dri/nouveau/nv10_swtcl.c index 3bc84d862d..4576c1ede4 100644 --- a/src/mesa/drivers/dri/nouveau/nv10_swtcl.c +++ b/src/mesa/drivers/dri/nouveau/nv10_swtcl.c @@ -392,15 +392,6 @@ static inline void nv10OutputVertexFormat(struct nouveau_context* nmesa) int i; int slots=0; int total_size=0; - /* t_vertex_generic dereferences a NULL pointer if we - * pass NULL as the vp transform... - */ - const GLfloat ident_vp[16] = { - 1.0, 0.0, 0.0, 0.0, - 0.0, 1.0, 0.0, 0.0, - 0.0, 0.0, 1.0, 0.0, - 0.0, 0.0, 0.0, 1.0 - }; nmesa->vertex_attr_count = 0; RENDERINPUTS_COPY(index, nmesa->render_inputs_bitset); @@ -431,28 +422,20 @@ static inline void nv10OutputVertexFormat(struct nouveau_context* nmesa) if (RENDERINPUTS_TEST(index, i)) { slots=i+1; - if (i==_TNL_ATTRIB_POS) - { - /* special-case POS */ - EMIT_ATTR(_TNL_ATTRIB_POS,EMIT_3F_VIEWPORT); - } - else + switch(attr_size[i]) { - switch(attr_size[i]) - { - case 1: - EMIT_ATTR(i,EMIT_1F); - break; - case 2: - EMIT_ATTR(i,EMIT_2F); - break; - case 3: - EMIT_ATTR(i,EMIT_3F); - break; - case 4: - EMIT_ATTR(i,EMIT_4F); - break; - } + case 1: + EMIT_ATTR(i,EMIT_1F); + break; + case 2: + EMIT_ATTR(i,EMIT_2F); + break; + case 3: + EMIT_ATTR(i,EMIT_3F); + break; + case 4: + EMIT_ATTR(i,EMIT_4F); + break; } if (i==_TNL_ATTRIB_COLOR0) nmesa->color_offset=total_size; @@ -465,7 +448,7 @@ static inline void nv10OutputVertexFormat(struct nouveau_context* nmesa) nmesa->vertex_size=_tnl_install_attrs( ctx, nmesa->vertex_attrs, nmesa->vertex_attr_count, - ident_vp, 0 ); + NULL, 0 ); assert(nmesa->vertex_size==total_size*4); /* diff --git a/src/mesa/drivers/dri/nouveau/nv30_state.c b/src/mesa/drivers/dri/nouveau/nv30_state.c index ad21fa2730..9b010954b3 100644 --- a/src/mesa/drivers/dri/nouveau/nv30_state.c +++ b/src/mesa/drivers/dri/nouveau/nv30_state.c @@ -639,25 +639,45 @@ void (*ReadBuffer)( GLcontext *ctx, GLenum buffer ); /** Set rasterization mode */ void (*RenderMode)(GLcontext *ctx, GLenum mode ); +/* Translate GL coords to window coords, clamping w/h to the + * dimensions of the window. + */ +static void nv30WindowCoords(nouveauContextPtr nmesa, + GLuint x, GLuint y, GLuint w, GLuint h, + GLuint *wX, GLuint *wY, GLuint *wW, GLuint *wH) +{ + if ((x+w) > nmesa->drawW) + w = nmesa->drawW - x; + (*wX) = x + nmesa->drawX; + (*wW) = w; + + if ((y+h) > nmesa->drawH) + h = nmesa->drawH - y; + (*wY) = (nmesa->drawH - y) - h + nmesa->drawY; + (*wH) = h; +} + /** Define the scissor box */ static void nv30Scissor(GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h) { nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx); + GLuint wX, wY, wW, wH; /* There's no scissor enable bit, so adjust the scissor to cover the * maximum draw buffer bounds */ if (!ctx->Scissor.Enabled) { - x = y = 0; - w = h = 4095; + wX = nmesa->drawX; + wY = nmesa->drawY; + wW = nmesa->drawW; + wH = nmesa->drawH; } else { - x += nmesa->drawX; - y += nmesa->drawY; + nv30WindowCoords(nmesa, x, y, w, h, &wX, &wY, &wW, &wH); } BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_SCISSOR_WIDTH_XPOS, 2); - OUT_RING_CACHE(((w) << 16) | x); - OUT_RING_CACHE(((h) << 16) | y); + OUT_RING_CACHE ((wW << 16) | wX); + OUT_RING_CACHE ((wH << 16) | wY); } /** Select flat or smooth shading */ @@ -751,19 +771,21 @@ static void nv30WindowMoved(nouveauContextPtr nmesa) { GLcontext *ctx = nmesa->glCtx; GLfloat *v = nmesa->viewport.m; - GLuint w = ctx->Viewport.Width; - GLuint h = ctx->Viewport.Height; - GLuint x = ctx->Viewport.X + nmesa->drawX; - GLuint y = ctx->Viewport.Y + nmesa->drawY; + GLuint wX, wY, wW, wH; + nv30WindowCoords(nmesa, ctx->Viewport.X, ctx->Viewport.Y, + ctx->Viewport.Width, ctx->Viewport.Height, + &wX, &wY, &wW, &wH); BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_VIEWPORT_DIMS_0, 2); - OUT_RING_CACHE((w << 16) | x); - OUT_RING_CACHE((h << 16) | y); + OUT_RING_CACHE ((wW << 16) | wX); + OUT_RING_CACHE ((wH << 16) | wY); + /* something to do with clears, possibly doesn't belong here */ BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_VIEWPORT_COLOR_BUFFER_OFS0, 2); - OUT_RING_CACHE(((w+x) << 16) | x); - OUT_RING_CACHE(((h+y) << 16) | y); + OUT_RING_CACHE(((nmesa->drawX + nmesa->drawW) << 16) | nmesa->drawX); + OUT_RING_CACHE(((nmesa->drawY + nmesa->drawH) << 16) | nmesa->drawY); + /* viewport transform */ BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_VIEWPORT_XFRM_OX, 8); OUT_RING_CACHEf (v[MAT_TX]); @@ -786,7 +808,7 @@ static GLboolean nv30InitCard(nouveauContextPtr nmesa) BEGIN_RING_SIZE(NvSub3D, NV30_TCL_PRIMITIVE_3D_SET_OBJECT1, 3); OUT_RING(NvDmaFB); - OUT_RING(NvDmaAGP); + OUT_RING(NvDmaTT); OUT_RING(NvDmaFB); BEGIN_RING_SIZE(NvSub3D, NV30_TCL_PRIMITIVE_3D_SET_OBJECT8, 1); OUT_RING(NvDmaFB); diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h index 5e07cbd88f..6615bc79fb 100644 --- a/src/mesa/drivers/dri/r300/r300_context.h +++ b/src/mesa/drivers/dri/r300/r300_context.h @@ -583,19 +583,6 @@ struct r300_vertex_shader_fragment { struct r300_vertex_shader_state { struct r300_vertex_shader_fragment program; - - struct r300_vertex_shader_fragment unknown1; - struct r300_vertex_shader_fragment unknown2; - - int program_start; - int unknown_ptr1; /* pointer within program space */ - int program_end; - - int param_offset; - int param_count; - - int unknown_ptr2; /* pointer within program space */ - int unknown_ptr3; /* pointer within program space */ }; extern int hw_tcl_on; @@ -805,10 +792,6 @@ struct r300_state { DECLARE_RENDERINPUTS(render_inputs_bitset); /* actual render inputs that R300 was configured for. They are the same as tnl->render_inputs for fixed pipeline */ - struct { - int transform_offset; /* Transform matrix offset, -1 if none */ - } vap_param; /* vertex processor parameter allocation - tells where to write parameters */ - struct r300_stencilbuffer_state stencil; }; diff --git a/src/mesa/drivers/dri/r300/r300_reg.h b/src/mesa/drivers/dri/r300/r300_reg.h index 3ce09c16d3..e59919be49 100644 --- a/src/mesa/drivers/dri/r300/r300_reg.h +++ b/src/mesa/drivers/dri/r300/r300_reg.h @@ -336,13 +336,15 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. * The meaning of the two UNKNOWN fields is obviously not known. However, * experiments so far have shown that both *must* point to an instruction * inside the vertex program, otherwise the GPU locks up. + * * fglrx usually sets CNTL_3_UNKNOWN to the end of the program and - * CNTL_1_UNKNOWN points to instruction where last write to position takes - * place. + * R300_PVS_CNTL_1_POS_END_SHIFT points to instruction where last write to + * position takes place. + * * Most likely this is used to ignore rest of the program in cases * where group of verts arent visible. For some reason this "section" * is sometimes accepted other instruction that have no relationship with - *position calculations. + * position calculations. */ #define R300_VAP_PVS_CNTL_1 0x22D0 # define R300_PVS_CNTL_1_PROGRAM_START_SHIFT 0 diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c index a723c3c530..b5cf21d644 100644 --- a/src/mesa/drivers/dri/r300/r300_state.c +++ b/src/mesa/drivers/dri/r300/r300_state.c @@ -715,7 +715,7 @@ static void r300LineWidth(GLcontext * ctx, GLfloat widthf) widthf = ctx->Line._Width; R300_STATECHANGE(r300, lcntl); - r300->hw.lcntl.cmd[1] |= + r300->hw.lcntl.cmd[1] = R300_LINE_CNT_HO | R300_LINE_CNT_VE | (int)(widthf * 6.0); } @@ -1477,9 +1477,7 @@ static void r300SetupRSUnit(GLcontext * ctx) if(_nc>_p->vpu.count)_p->vpu.count=_nc;\ }while(0) -void static inline setup_vertex_shader_fragment(r300ContextPtr r300, int dest, struct - r300_vertex_shader_fragment - *vsf) +static inline void r300SetupVertexProgramFragment(r300ContextPtr r300, int dest, struct r300_vertex_shader_fragment *vsf) { int i; @@ -1487,8 +1485,7 @@ void static inline setup_vertex_shader_fragment(r300ContextPtr r300, int dest, s return; if (vsf->length & 0x3) { - fprintf(stderr, - "VERTEX_SHADER_FRAGMENT must have length divisible by 4\n"); + fprintf(stderr, "VERTEX_SHADER_FRAGMENT must have length divisible by 4\n"); _mesa_exit(-1); } @@ -1496,147 +1493,100 @@ void static inline setup_vertex_shader_fragment(r300ContextPtr r300, int dest, s case 0: R300_STATECHANGE(r300, vpi); for (i = 0; i < vsf->length; i++) - r300->hw.vpi.cmd[R300_VPI_INSTR_0 + i + - 4 * (dest & 0xff)] = (vsf->body.d[i]); - bump_vpu_count(r300->hw.vpi.cmd, - vsf->length + 4 * (dest & 0xff)); + r300->hw.vpi.cmd[R300_VPI_INSTR_0 + i + 4 * (dest & 0xff)] = (vsf->body.d[i]); + bump_vpu_count(r300->hw.vpi.cmd, vsf->length + 4 * (dest & 0xff)); break; case 2: R300_STATECHANGE(r300, vpp); for (i = 0; i < vsf->length; i++) - r300->hw.vpp.cmd[R300_VPP_PARAM_0 + i + - 4 * (dest & 0xff)] = (vsf->body.d[i]); - bump_vpu_count(r300->hw.vpp.cmd, - vsf->length + 4 * (dest & 0xff)); + r300->hw.vpp.cmd[R300_VPP_PARAM_0 + i + 4 * (dest & 0xff)] = (vsf->body.d[i]); + bump_vpu_count(r300->hw.vpp.cmd, vsf->length + 4 * (dest & 0xff)); break; case 4: R300_STATECHANGE(r300, vps); for (i = 0; i < vsf->length; i++) - r300->hw.vps.cmd[1 + i + 4 * (dest & 0xff)] = - (vsf->body.d[i]); - bump_vpu_count(r300->hw.vps.cmd, - vsf->length + 4 * (dest & 0xff)); + r300->hw.vps.cmd[1 + i + 4 * (dest & 0xff)] = (vsf->body.d[i]); + bump_vpu_count(r300->hw.vps.cmd, vsf->length + 4 * (dest & 0xff)); break; default: - fprintf(stderr, - "%s:%s don't know how to handle dest %04x\n", - __FILE__, __FUNCTION__, dest); + fprintf(stderr, "%s:%s don't know how to handle dest %04x\n", __FILE__, __FUNCTION__, dest); _mesa_exit(-1); } } -/* just a skeleton for now.. */ - -/* Generate a vertex shader that simply transforms vertex and texture coordinates, - while leaving colors intact. Nothing fancy (like lights) - - If implementing lights make a copy first, so it is easy to switch between the two versions */ -static void r300GenerateSimpleVertexShader(r300ContextPtr r300) +static void r300SetupDefaultVertexProgram(r300ContextPtr rmesa) { - int i; + struct r300_vertex_shader_state *prog = &(rmesa->state.vertex_shader); GLuint o_reg = 0; - - /* Allocate parameters */ - r300->state.vap_param.transform_offset = 0x0; /* transform matrix */ - r300->state.vertex_shader.param_offset = 0x0; - r300->state.vertex_shader.param_count = 0x4; /* 4 vector values - 4x4 matrix */ - - r300->state.vertex_shader.program_start = 0x0; - r300->state.vertex_shader.unknown_ptr1 = 0x4; /* magic value ? */ - r300->state.vertex_shader.program_end = 0x0; - - r300->state.vertex_shader.unknown_ptr2 = 0x0; /* magic value */ - r300->state.vertex_shader.unknown_ptr3 = 0x4; /* magic value */ - - r300->state.vertex_shader.unknown1.length = 0; - r300->state.vertex_shader.unknown2.length = 0; - -#define WRITE_OP(oper,source1,source2,source3) {\ - r300->state.vertex_shader.program.body.i[r300->state.vertex_shader.program_end].op=(oper); \ - r300->state.vertex_shader.program.body.i[r300->state.vertex_shader.program_end].src[0]=(source1); \ - r300->state.vertex_shader.program.body.i[r300->state.vertex_shader.program_end].src[1]=(source2); \ - r300->state.vertex_shader.program.body.i[r300->state.vertex_shader.program_end].src[2]=(source3); \ - r300->state.vertex_shader.program_end++; \ - } - - for (i = VERT_ATTRIB_POS; i < VERT_ATTRIB_MAX; i++) - if (r300->state.sw_tcl_inputs[i] != -1) { - WRITE_OP(EASY_VSF_OP(MUL, o_reg++, ALL, RESULT), - VSF_REG(r300->state.sw_tcl_inputs[i]), - VSF_ATTR_UNITY(r300->state. - sw_tcl_inputs[i]), - VSF_UNITY(r300->state.sw_tcl_inputs[i]) - ) - + int i; + int inst_count = 0; + int param_count = 0; + int program_end = 0; + + for (i = VERT_ATTRIB_POS; i < VERT_ATTRIB_MAX; i++) { + if (rmesa->state.sw_tcl_inputs[i] != -1) { + prog->program.body.i[program_end].op = EASY_VSF_OP(MUL, o_reg++, ALL, RESULT); + prog->program.body.i[program_end].src[0] = VSF_REG(rmesa->state.sw_tcl_inputs[i]); + prog->program.body.i[program_end].src[1] = VSF_ATTR_UNITY(rmesa->state.sw_tcl_inputs[i]); + prog->program.body.i[program_end].src[2] = VSF_UNITY(rmesa->state.sw_tcl_inputs[i]); + program_end++; } + } - r300->state.vertex_shader.program_end--; /* r300 wants program length to be one more - no idea why */ - r300->state.vertex_shader.program.length = - (r300->state.vertex_shader.program_end + 1) * 4; + prog->program.length = program_end * 4; - r300->state.vertex_shader.unknown_ptr1 = r300->state.vertex_shader.program_end; /* magic value ? */ - r300->state.vertex_shader.unknown_ptr2 = r300->state.vertex_shader.program_end; /* magic value ? */ - r300->state.vertex_shader.unknown_ptr3 = r300->state.vertex_shader.program_end; /* magic value ? */ + r300SetupVertexProgramFragment(rmesa, VSF_DEST_PROGRAM, &(prog->program)); + inst_count = (prog->program.length / 4) - 1; + R300_STATECHANGE(rmesa, pvs); + rmesa->hw.pvs.cmd[R300_PVS_CNTL_1] = + (0 << R300_PVS_CNTL_1_PROGRAM_START_SHIFT) | + (inst_count << R300_PVS_CNTL_1_POS_END_SHIFT) | + (inst_count << R300_PVS_CNTL_1_PROGRAM_END_SHIFT); + rmesa->hw.pvs.cmd[R300_PVS_CNTL_2] = + (0 << R300_PVS_CNTL_2_PARAM_OFFSET_SHIFT) | + (param_count << R300_PVS_CNTL_2_PARAM_COUNT_SHIFT); + rmesa->hw.pvs.cmd[R300_PVS_CNTL_3] = + (inst_count << R300_PVS_CNTL_3_PROGRAM_UNKNOWN_SHIFT) | + (inst_count << R300_PVS_CNTL_3_PROGRAM_UNKNOWN2_SHIFT); } -static void r300SetupVertexProgram(r300ContextPtr rmesa) +static void r300SetupRealVertexProgram(r300ContextPtr rmesa) { GLcontext *ctx = rmesa->radeon.glCtx; - int inst_count; - int param_count; - struct r300_vertex_program *prog = - (struct r300_vertex_program *)CURRENT_VERTEX_SHADER(ctx); + struct r300_vertex_program *prog = (struct r300_vertex_program *)CURRENT_VERTEX_SHADER(ctx); + int inst_count = 0; + int param_count = 0; - ((drm_r300_cmd_header_t *) rmesa->hw.vpp.cmd)->vpu.count = 0; + /* FIXME: r300SetupVertexProgramFragment */ R300_STATECHANGE(rmesa, vpp); param_count = - r300VertexProgUpdateParams(ctx, (struct r300_vertex_program_cont *) - ctx->VertexProgram._Current /*prog */ , + r300VertexProgUpdateParams(ctx, + (struct r300_vertex_program_cont *) + ctx->VertexProgram._Current, (float *)&rmesa->hw.vpp. cmd[R300_VPP_PARAM_0]); bump_vpu_count(rmesa->hw.vpp.cmd, param_count); param_count /= 4; - /* Reset state, in case we don't use something */ - ((drm_r300_cmd_header_t *) rmesa->hw.vpi.cmd)->vpu.count = 0; - ((drm_r300_cmd_header_t *) rmesa->hw.vps.cmd)->vpu.count = 0; - - setup_vertex_shader_fragment(rmesa, VSF_DEST_PROGRAM, &(prog->program)); - -#if 0 - setup_vertex_shader_fragment(rmesa, VSF_DEST_UNKNOWN1, - &(rmesa->state.vertex_shader.unknown1)); - setup_vertex_shader_fragment(rmesa, VSF_DEST_UNKNOWN2, - &(rmesa->state.vertex_shader.unknown2)); -#endif - - inst_count = prog->program.length / 4 - 1; + r300SetupVertexProgramFragment(rmesa, VSF_DEST_PROGRAM, &(prog->program)); + inst_count = (prog->program.length / 4) - 1; R300_STATECHANGE(rmesa, pvs); rmesa->hw.pvs.cmd[R300_PVS_CNTL_1] = - (0 << R300_PVS_CNTL_1_PROGRAM_START_SHIFT) - | (inst_count /*pos_end */ << R300_PVS_CNTL_1_POS_END_SHIFT) - | (inst_count << R300_PVS_CNTL_1_PROGRAM_END_SHIFT); + (0 << R300_PVS_CNTL_1_PROGRAM_START_SHIFT) | + (inst_count << R300_PVS_CNTL_1_POS_END_SHIFT) | + (inst_count << R300_PVS_CNTL_1_PROGRAM_END_SHIFT); rmesa->hw.pvs.cmd[R300_PVS_CNTL_2] = - (0 << R300_PVS_CNTL_2_PARAM_OFFSET_SHIFT) - | (param_count << R300_PVS_CNTL_2_PARAM_COUNT_SHIFT); + (0 << R300_PVS_CNTL_2_PARAM_OFFSET_SHIFT) | + (param_count << R300_PVS_CNTL_2_PARAM_COUNT_SHIFT); rmesa->hw.pvs.cmd[R300_PVS_CNTL_3] = - (0 /*rmesa->state.vertex_shader.unknown_ptr2 */ << - R300_PVS_CNTL_3_PROGRAM_UNKNOWN_SHIFT) - | (inst_count /*rmesa->state.vertex_shader.unknown_ptr3 */ << - 0); - - /* This is done for vertex shader fragments, but also needs to be done for vap_pvs, - so I leave it as a reminder */ -#if 0 - reg_start(R300_VAP_PVS_WAITIDLE, 0); - e32(0x00000000); -#endif + (inst_count << R300_PVS_CNTL_3_PROGRAM_UNKNOWN_SHIFT) | + (inst_count << R300_PVS_CNTL_3_PROGRAM_UNKNOWN2_SHIFT); } -static void r300SetupVertexShader(r300ContextPtr rmesa) +static void r300SetupVertexProgram(r300ContextPtr rmesa) { GLcontext *ctx = rmesa->radeon.glCtx; @@ -1649,46 +1599,16 @@ static void r300SetupVertexShader(r300ContextPtr rmesa) 0x400 area might have something to do with pixel shaders as it appears right after pfs programming. 0x406 is set to { 0.0, 0.0, 1.0, 0.0 } most of the time but should change with smooth points and in other rare cases. */ //setup_vertex_shader_fragment(rmesa, 0x406, &unk4); - if (hw_tcl_on - && ((struct r300_vertex_program *)CURRENT_VERTEX_SHADER(ctx))-> - translated) { - r300SetupVertexProgram(rmesa); - return; + if (hw_tcl_on && ((struct r300_vertex_program *)CURRENT_VERTEX_SHADER(ctx))->translated) { + r300SetupRealVertexProgram(rmesa); + } else { + /* FIXME: This needs to be replaced by vertex shader generation code. */ + r300SetupDefaultVertexProgram(rmesa); } - /* This needs to be replaced by vertex shader generation code */ - r300GenerateSimpleVertexShader(rmesa); - - setup_vertex_shader_fragment(rmesa, VSF_DEST_PROGRAM, - &(rmesa->state.vertex_shader.program)); - -#if 0 - setup_vertex_shader_fragment(rmesa, VSF_DEST_UNKNOWN1, - &(rmesa->state.vertex_shader.unknown1)); - setup_vertex_shader_fragment(rmesa, VSF_DEST_UNKNOWN2, - &(rmesa->state.vertex_shader.unknown2)); -#endif - - R300_STATECHANGE(rmesa, pvs); - rmesa->hw.pvs.cmd[R300_PVS_CNTL_1] = - (rmesa->state.vertex_shader. - program_start << R300_PVS_CNTL_1_PROGRAM_START_SHIFT) - | (rmesa->state.vertex_shader. - unknown_ptr1 << R300_PVS_CNTL_1_POS_END_SHIFT) - | (rmesa->state.vertex_shader. - program_end << R300_PVS_CNTL_1_PROGRAM_END_SHIFT); - rmesa->hw.pvs.cmd[R300_PVS_CNTL_2] = - (rmesa->state.vertex_shader. - param_offset << R300_PVS_CNTL_2_PARAM_OFFSET_SHIFT) - | (rmesa->state.vertex_shader. - param_count << R300_PVS_CNTL_2_PARAM_COUNT_SHIFT); - rmesa->hw.pvs.cmd[R300_PVS_CNTL_3] = - (rmesa->state.vertex_shader. - unknown_ptr2 << R300_PVS_CNTL_3_PROGRAM_UNKNOWN_SHIFT) - | (rmesa->state.vertex_shader.unknown_ptr3 << 0); - /* This is done for vertex shader fragments, but also needs to be done for vap_pvs, - so I leave it as a reminder */ + /* FIXME: This is done for vertex shader fragments, but also needs to be + * done for vap_pvs, so I leave it as a reminder. */ #if 0 reg_start(R300_VAP_PVS_WAITIDLE, 0); e32(0x00000000); @@ -1941,6 +1861,8 @@ static void r300ResetHwState(r300ContextPtr r300) r300PolygonOffset(ctx, ctx->Polygon.OffsetFactor, ctx->Polygon.OffsetUnits); + r300Enable(ctx, GL_POLYGON_OFFSET_POINT, ctx->Polygon.OffsetPoint); + r300Enable(ctx, GL_POLYGON_OFFSET_LINE, ctx->Polygon.OffsetLine); r300Enable(ctx, GL_POLYGON_OFFSET_FILL, ctx->Polygon.OffsetFill); r300->hw.unk42C0.cmd[1] = 0x4B7FFFFF; @@ -2094,7 +2016,6 @@ void r300UpdateShaders(r300ContextPtr rmesa) } r300UpdateStateParameters(ctx, _NEW_PROGRAM); } - } static void r300SetupPixelShader(r300ContextPtr rmesa) @@ -2113,62 +2034,61 @@ static void r300SetupPixelShader(r300ContextPtr rmesa) __FUNCTION__); return; } -#define OUTPUT_FIELD(st, reg, field) \ - R300_STATECHANGE(rmesa, st); \ - for(i=0;i<=fp->alu_end;i++) \ - rmesa->hw.st.cmd[R300_FPI_INSTR_0+i]=fp->alu.inst[i].field;\ - rmesa->hw.st.cmd[R300_FPI_CMD_0]=cmdpacket0(reg, fp->alu_end+1); - - OUTPUT_FIELD(fpi[0], R300_PFS_INSTR0_0, inst0); - OUTPUT_FIELD(fpi[1], R300_PFS_INSTR1_0, inst1); - OUTPUT_FIELD(fpi[2], R300_PFS_INSTR2_0, inst2); - OUTPUT_FIELD(fpi[3], R300_PFS_INSTR3_0, inst3); -#undef OUTPUT_FIELD + + R300_STATECHANGE(rmesa, fpi[0]); + rmesa->hw.fpi[0].cmd[R300_FPI_CMD_0] = cmdpacket0(R300_PFS_INSTR0_0, fp->alu_end + 1); + for (i = 0; i <= fp->alu_end; i++) { + rmesa->hw.fpi[0].cmd[R300_FPI_INSTR_0 + i] = fp->alu.inst[i].inst0; + } + + R300_STATECHANGE(rmesa, fpi[1]); + rmesa->hw.fpi[1].cmd[R300_FPI_CMD_0] = cmdpacket0(R300_PFS_INSTR1_0, fp->alu_end + 1); + for (i = 0; i <= fp->alu_end; i++) { + rmesa->hw.fpi[1].cmd[R300_FPI_INSTR_0 + i] = fp->alu.inst[i].inst1; + } + + R300_STATECHANGE(rmesa, fpi[2]); + rmesa->hw.fpi[2].cmd[R300_FPI_CMD_0] = cmdpacket0(R300_PFS_INSTR2_0, fp->alu_end + 1); + for (i = 0; i <= fp->alu_end; i++) { + rmesa->hw.fpi[2].cmd[R300_FPI_INSTR_0 + i] = fp->alu.inst[i].inst2; + } + + R300_STATECHANGE(rmesa, fpi[3]); + rmesa->hw.fpi[3].cmd[R300_FPI_CMD_0] = cmdpacket0(R300_PFS_INSTR3_0, fp->alu_end + 1); + for (i = 0; i <= fp->alu_end; i++) { + rmesa->hw.fpi[3].cmd[R300_FPI_INSTR_0 + i] = fp->alu.inst[i].inst3; + } R300_STATECHANGE(rmesa, fp); + rmesa->hw.fp.cmd[R300_FP_CNTL0] = fp->cur_node | (fp->first_node_has_tex << 3); + rmesa->hw.fp.cmd[R300_FP_CNTL1] = fp->max_temp_idx; + rmesa->hw.fp.cmd[R300_FP_CNTL2] = + (fp->alu_offset << R300_PFS_CNTL_ALU_OFFSET_SHIFT) | + (fp->alu_end << R300_PFS_CNTL_ALU_END_SHIFT) | + (fp->tex_offset << R300_PFS_CNTL_TEX_OFFSET_SHIFT) | + (fp->tex_end << R300_PFS_CNTL_TEX_END_SHIFT); /* I just want to say, the way these nodes are stored.. weird.. */ for (i = 0, k = (4 - (fp->cur_node + 1)); i < 4; i++, k++) { if (i < (fp->cur_node + 1)) { rmesa->hw.fp.cmd[R300_FP_NODE0 + k] = - (fp->node[i]. - alu_offset << R300_PFS_NODE_ALU_OFFSET_SHIFT) - | (fp->node[i]. - alu_end << R300_PFS_NODE_ALU_END_SHIFT) - | (fp->node[i]. - tex_offset << R300_PFS_NODE_TEX_OFFSET_SHIFT) - | (fp->node[i]. - tex_end << R300_PFS_NODE_TEX_END_SHIFT) - | fp->node[i].flags; + (fp->node[i].alu_offset << R300_PFS_NODE_ALU_OFFSET_SHIFT) | + (fp->node[i].alu_end << R300_PFS_NODE_ALU_END_SHIFT) | + (fp->node[i].tex_offset << R300_PFS_NODE_TEX_OFFSET_SHIFT) | + (fp->node[i].tex_end << R300_PFS_NODE_TEX_END_SHIFT) | + fp->node[i].flags; } else { rmesa->hw.fp.cmd[R300_FP_NODE0 + (3 - i)] = 0; } } - /* PFS_CNTL_0 */ - rmesa->hw.fp.cmd[R300_FP_CNTL0] = - fp->cur_node | (fp->first_node_has_tex << 3); - /* PFS_CNTL_1 */ - rmesa->hw.fp.cmd[R300_FP_CNTL1] = fp->max_temp_idx; - /* PFS_CNTL_2 */ - rmesa->hw.fp.cmd[R300_FP_CNTL2] = - (fp->alu_offset << R300_PFS_CNTL_ALU_OFFSET_SHIFT) - | (fp->alu_end << R300_PFS_CNTL_ALU_END_SHIFT) - | (fp->tex_offset << R300_PFS_CNTL_TEX_OFFSET_SHIFT) - | (fp->tex_end << R300_PFS_CNTL_TEX_END_SHIFT); - R300_STATECHANGE(rmesa, fpp); + rmesa->hw.fpp.cmd[R300_FPP_CMD_0] = cmdpacket0(R300_PFS_PARAM_0_X, fp->const_nr * 4); for (i = 0; i < fp->const_nr; i++) { - rmesa->hw.fpp.cmd[R300_FPP_PARAM_0 + 4 * i + 0] = - r300PackFloat24(fp->constant[i][0]); - rmesa->hw.fpp.cmd[R300_FPP_PARAM_0 + 4 * i + 1] = - r300PackFloat24(fp->constant[i][1]); - rmesa->hw.fpp.cmd[R300_FPP_PARAM_0 + 4 * i + 2] = - r300PackFloat24(fp->constant[i][2]); - rmesa->hw.fpp.cmd[R300_FPP_PARAM_0 + 4 * i + 3] = - r300PackFloat24(fp->constant[i][3]); + rmesa->hw.fpp.cmd[R300_FPP_PARAM_0 + 4 * i + 0] = r300PackFloat24(fp->constant[i][0]); + rmesa->hw.fpp.cmd[R300_FPP_PARAM_0 + 4 * i + 1] = r300PackFloat24(fp->constant[i][1]); + rmesa->hw.fpp.cmd[R300_FPP_PARAM_0 + 4 * i + 2] = r300PackFloat24(fp->constant[i][2]); + rmesa->hw.fpp.cmd[R300_FPP_PARAM_0 + 4 * i + 3] = r300PackFloat24(fp->constant[i][3]); } - rmesa->hw.fpp.cmd[R300_FPP_CMD_0] = - cmdpacket0(R300_PFS_PARAM_0_X, fp->const_nr * 4); } void r300UpdateShaderStates(r300ContextPtr rmesa) @@ -2182,7 +2102,7 @@ void r300UpdateShaderStates(r300ContextPtr rmesa) r300SetupTextures(ctx); if ((rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) - r300SetupVertexShader(rmesa); + r300SetupVertexProgram(rmesa); r300SetupRSUnit(ctx); } diff --git a/src/mesa/drivers/dri/r300/radeon_lock.h b/src/mesa/drivers/dri/r300/radeon_lock.h index c47adc9575..a344837f47 100644 --- a/src/mesa/drivers/dri/r300/radeon_lock.h +++ b/src/mesa/drivers/dri/r300/radeon_lock.h @@ -42,9 +42,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #ifndef __RADEON_LOCK_H__ #define __RADEON_LOCK_H__ -#if 0 -#include "r200_ioctl.h" -#endif #include "radeon_context.h" extern void radeonGetLock(radeonContextPtr rmesa, GLuint flags); |