diff options
263 files changed, 6979 insertions, 3676 deletions
diff --git a/configs/default b/configs/default index 9af816cf9bc..15f9285dbf6 100644 --- a/configs/default +++ b/configs/default @@ -9,7 +9,7 @@ CONFIG_NAME = default # Version info MESA_MAJOR=7 -MESA_MINOR=7 +MESA_MINOR=8 MESA_TINY=0 MESA_VERSION = $(MESA_MAJOR).$(MESA_MINOR).$(MESA_TINY) diff --git a/docs/relnotes-7.8.html b/docs/relnotes-7.8.html new file mode 100644 index 00000000000..dc003566c76 --- /dev/null +++ b/docs/relnotes-7.8.html @@ -0,0 +1,53 @@ +<HTML> + +<TITLE>Mesa Release Notes</TITLE> + +<head><link rel="stylesheet" type="text/css" href="mesa.css"></head> + +<BODY> + +<body bgcolor="#eeeeee"> + +<H1>Mesa 7.8 Release Notes / date TBD</H1> + +<p> +Mesa 7.8 is a new development release. +People who are concerned with stability and reliability should stick +with a previous release or wait for Mesa 7.8.1. +</p> +<p> +Mesa 7.8 implements the OpenGL 2.1 API, but the version reported by +glGetString(GL_VERSION) depends on the particular driver being used. +Some drivers don't support all the features required in OpenGL 2.1. +</p> +<p> +See the <a href="install.html">Compiling/Installing page</a> for prerequisites +for DRI hardware acceleration. +</p> + + +<h2>MD5 checksums</h2> +<pre> +tbd +</pre> + + +<h2>New features</h2> +<ul> +<li>TBD +</ul> + + +<h2>Bug fixes</h2> +<ul> +<li>TBD +</ul> + + +<h2>Changes</h2> +<ul> +<li>TBD +</ul> + +</body> +</html> diff --git a/docs/relnotes.html b/docs/relnotes.html index b301ee046a7..d0d9b6e5b98 100644 --- a/docs/relnotes.html +++ b/docs/relnotes.html @@ -13,6 +13,7 @@ The release notes summarize what's new or changed in each Mesa release. </p> <UL> +<LI><A HREF="relnotes-7.8.html">7.8 release notes</A> <LI><A HREF="relnotes-7.7.html">7.7 release notes</A> <LI><A HREF="relnotes-7.6.1.html">7.6.1 release notes</A> <LI><A HREF="relnotes-7.6.html">7.6 release notes</A> diff --git a/scons/llvm.py b/scons/llvm.py index 73e9310f71b..7b266502907 100644 --- a/scons/llvm.py +++ b/scons/llvm.py @@ -58,45 +58,47 @@ def generate(env): env.PrependENVPath('PATH', llvm_bin_dir) - if env['msvc']: + if env['platform'] == 'windows': # XXX: There is no llvm-config on Windows, so assume a standard layout if llvm_dir is not None: env.Prepend(CPPPATH = [os.path.join(llvm_dir, 'include')]) + env.AppendUnique(CPPDEFINES = [ + '__STDC_LIMIT_MACROS', + '__STDC_CONSTANT_MACROS', + ]) env.Prepend(LIBPATH = [os.path.join(llvm_dir, 'lib')]) env.Prepend(LIBS = [ + 'LLVMX86AsmParser', + 'LLVMX86AsmPrinter', + 'LLVMX86CodeGen', + 'LLVMX86Info', + 'LLVMLinker', + 'LLVMipo', + 'LLVMInterpreter', + 'LLVMInstrumentation', + 'LLVMJIT', + 'LLVMExecutionEngine', + 'LLVMDebugger', 'LLVMBitWriter', - 'LLVMCore', - 'LLVMSupport', - 'LLVMSystem', - 'LLVMSupport', - 'LLVMSystem', - 'LLVMCore', - 'LLVMCodeGen', + 'LLVMAsmParser', + 'LLVMArchive', + 'LLVMBitReader', 'LLVMSelectionDAG', 'LLVMAsmPrinter', - 'LLVMBitReader', - 'LLVMBitWriter', - 'LLVMTransformUtils', - 'LLVMInstrumentation', + 'LLVMCodeGen', 'LLVMScalarOpts', - 'LLVMipo', - 'LLVMHello', - 'LLVMLinker', - 'LLVMAnalysis', + 'LLVMTransformUtils', 'LLVMipa', - 'LLVMX86CodeGen', - 'LLVMX86AsmPrinter', - 'LLVMExecutionEngine', - 'LLVMInterpreter', - 'LLVMJIT', + 'LLVMAnalysis', 'LLVMTarget', - 'LLVMAsmParser', - 'LLVMDebugger', - 'LLVMArchive', + 'LLVMMC', + 'LLVMCore', + 'LLVMSupport', + 'LLVMSystem', 'imagehlp', 'psapi', ]) - env['LLVM_VERSION'] = '2.5' + env['LLVM_VERSION'] = '2.6' return elif env.Detect('llvm-config'): version = env.backtick('llvm-config --version').rstrip() diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index 4f13b3e2bad..b6f7b883224 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -50,20 +50,35 @@ struct cso_context { struct { void *samplers[PIPE_MAX_SAMPLERS]; unsigned nr_samplers; + + void *vertex_samplers[PIPE_MAX_VERTEX_SAMPLERS]; + unsigned nr_vertex_samplers; } hw; void *samplers[PIPE_MAX_SAMPLERS]; unsigned nr_samplers; + void *vertex_samplers[PIPE_MAX_VERTEX_SAMPLERS]; + unsigned nr_vertex_samplers; + unsigned nr_samplers_saved; void *samplers_saved[PIPE_MAX_SAMPLERS]; + unsigned nr_vertex_samplers_saved; + void *vertex_samplers_saved[PIPE_MAX_VERTEX_SAMPLERS]; + struct pipe_texture *textures[PIPE_MAX_SAMPLERS]; uint nr_textures; + struct pipe_texture *vertex_textures[PIPE_MAX_VERTEX_SAMPLERS]; + uint nr_vertex_textures; + uint nr_textures_saved; struct pipe_texture *textures_saved[PIPE_MAX_SAMPLERS]; + uint nr_vertex_textures_saved; + struct pipe_texture *vertex_textures_saved[PIPE_MAX_SAMPLERS]; + /** Current and saved state. * The saved state is used as a 1-deep stack. */ @@ -244,7 +259,8 @@ void cso_release_all( struct cso_context *ctx ) if (ctx->pipe) { ctx->pipe->bind_blend_state( ctx->pipe, NULL ); ctx->pipe->bind_rasterizer_state( ctx->pipe, NULL ); - ctx->pipe->bind_sampler_states( ctx->pipe, 0, NULL ); + ctx->pipe->bind_fragment_sampler_states( ctx->pipe, 0, NULL ); + ctx->pipe->bind_vertex_sampler_states(ctx->pipe, 0, NULL); ctx->pipe->bind_depth_stencil_alpha_state( ctx->pipe, NULL ); ctx->pipe->bind_fs_state( ctx->pipe, NULL ); ctx->pipe->bind_vs_state( ctx->pipe, NULL ); @@ -255,6 +271,11 @@ void cso_release_all( struct cso_context *ctx ) pipe_texture_reference(&ctx->textures_saved[i], NULL); } + for (i = 0; i < PIPE_MAX_VERTEX_SAMPLERS; i++) { + pipe_texture_reference(&ctx->vertex_textures[i], NULL); + pipe_texture_reference(&ctx->vertex_textures_saved[i], NULL); + } + free_framebuffer_state(&ctx->fb); free_framebuffer_state(&ctx->fb_saved); @@ -378,6 +399,46 @@ enum pipe_error cso_single_sampler(struct cso_context *ctx, return PIPE_OK; } +enum pipe_error +cso_single_vertex_sampler(struct cso_context *ctx, + unsigned idx, + const struct pipe_sampler_state *templ) +{ + void *handle = NULL; + + if (templ != NULL) { + unsigned hash_key = cso_construct_key((void*)templ, sizeof(struct pipe_sampler_state)); + struct cso_hash_iter iter = cso_find_state_template(ctx->cache, + hash_key, CSO_SAMPLER, + (void*)templ); + + if (cso_hash_iter_is_null(iter)) { + struct cso_sampler *cso = MALLOC(sizeof(struct cso_sampler)); + if (!cso) + return PIPE_ERROR_OUT_OF_MEMORY; + + memcpy(&cso->state, templ, sizeof(*templ)); + cso->data = ctx->pipe->create_sampler_state(ctx->pipe, &cso->state); + cso->delete_state = (cso_state_callback)ctx->pipe->delete_sampler_state; + cso->context = ctx->pipe; + + iter = cso_insert_state(ctx->cache, hash_key, CSO_SAMPLER, cso); + if (cso_hash_iter_is_null(iter)) { + FREE(cso); + return PIPE_ERROR_OUT_OF_MEMORY; + } + + handle = cso->data; + } + else { + handle = ((struct cso_sampler *)cso_hash_iter_data(iter))->data; + } + } + + ctx->vertex_samplers[idx] = handle; + return PIPE_OK; +} + void cso_single_sampler_done( struct cso_context *ctx ) { unsigned i; @@ -398,7 +459,36 @@ void cso_single_sampler_done( struct cso_context *ctx ) memcpy(ctx->hw.samplers, ctx->samplers, ctx->nr_samplers * sizeof(void *)); ctx->hw.nr_samplers = ctx->nr_samplers; - ctx->pipe->bind_sampler_states(ctx->pipe, ctx->nr_samplers, ctx->samplers); + ctx->pipe->bind_fragment_sampler_states(ctx->pipe, ctx->nr_samplers, ctx->samplers); + } +} + +void +cso_single_vertex_sampler_done(struct cso_context *ctx) +{ + unsigned i; + + /* find highest non-null sampler */ + for (i = PIPE_MAX_VERTEX_SAMPLERS; i > 0; i--) { + if (ctx->vertex_samplers[i - 1] != NULL) + break; + } + + ctx->nr_vertex_samplers = i; + + if (ctx->hw.nr_vertex_samplers != ctx->nr_vertex_samplers || + memcmp(ctx->hw.vertex_samplers, + ctx->vertex_samplers, + ctx->nr_vertex_samplers * sizeof(void *)) != 0) + { + memcpy(ctx->hw.vertex_samplers, + ctx->vertex_samplers, + ctx->nr_vertex_samplers * sizeof(void *)); + ctx->hw.nr_vertex_samplers = ctx->nr_vertex_samplers; + + ctx->pipe->bind_vertex_sampler_states(ctx->pipe, + ctx->nr_vertex_samplers, + ctx->vertex_samplers); } } @@ -447,6 +537,21 @@ void cso_restore_samplers(struct cso_context *ctx) cso_single_sampler_done( ctx ); } +void +cso_save_vertex_samplers(struct cso_context *ctx) +{ + ctx->nr_vertex_samplers_saved = ctx->nr_vertex_samplers; + memcpy(ctx->vertex_samplers_saved, ctx->vertex_samplers, sizeof(ctx->vertex_samplers)); +} + +void +cso_restore_vertex_samplers(struct cso_context *ctx) +{ + ctx->nr_vertex_samplers = ctx->nr_vertex_samplers_saved; + memcpy(ctx->vertex_samplers, ctx->vertex_samplers_saved, sizeof(ctx->vertex_samplers)); + cso_single_vertex_sampler_done(ctx); +} + enum pipe_error cso_set_sampler_textures( struct cso_context *ctx, uint count, @@ -461,7 +566,7 @@ enum pipe_error cso_set_sampler_textures( struct cso_context *ctx, for ( ; i < PIPE_MAX_SAMPLERS; i++) pipe_texture_reference(&ctx->textures[i], NULL); - ctx->pipe->set_sampler_textures(ctx->pipe, count, textures); + ctx->pipe->set_fragment_sampler_textures(ctx->pipe, count, textures); return PIPE_OK; } @@ -491,13 +596,71 @@ void cso_restore_sampler_textures( struct cso_context *ctx ) for ( ; i < PIPE_MAX_SAMPLERS; i++) pipe_texture_reference(&ctx->textures[i], NULL); - ctx->pipe->set_sampler_textures(ctx->pipe, ctx->nr_textures, ctx->textures); + ctx->pipe->set_fragment_sampler_textures(ctx->pipe, ctx->nr_textures, ctx->textures); ctx->nr_textures_saved = 0; } +enum pipe_error +cso_set_vertex_sampler_textures(struct cso_context *ctx, + uint count, + struct pipe_texture **textures) +{ + uint i; + + ctx->nr_vertex_textures = count; + + for (i = 0; i < count; i++) { + pipe_texture_reference(&ctx->vertex_textures[i], textures[i]); + } + for ( ; i < PIPE_MAX_VERTEX_SAMPLERS; i++) { + pipe_texture_reference(&ctx->vertex_textures[i], NULL); + } + + ctx->pipe->set_vertex_sampler_textures(ctx->pipe, count, textures); + + return PIPE_OK; +} + +void +cso_save_vertex_sampler_textures(struct cso_context *ctx) +{ + uint i; + + ctx->nr_vertex_textures_saved = ctx->nr_vertex_textures; + for (i = 0; i < ctx->nr_vertex_textures; i++) { + assert(!ctx->vertex_textures_saved[i]); + pipe_texture_reference(&ctx->vertex_textures_saved[i], ctx->vertex_textures[i]); + } +} + +void +cso_restore_vertex_sampler_textures(struct cso_context *ctx) +{ + uint i; + + ctx->nr_vertex_textures = ctx->nr_vertex_textures_saved; + + for (i = 0; i < ctx->nr_vertex_textures; i++) { + pipe_texture_reference(&ctx->vertex_textures[i], NULL); + ctx->vertex_textures[i] = ctx->vertex_textures_saved[i]; + ctx->vertex_textures_saved[i] = NULL; + } + for ( ; i < PIPE_MAX_VERTEX_SAMPLERS; i++) { + pipe_texture_reference(&ctx->vertex_textures[i], NULL); + } + + ctx->pipe->set_vertex_sampler_textures(ctx->pipe, + ctx->nr_vertex_textures, + ctx->vertex_textures); + + ctx->nr_vertex_textures_saved = 0; +} + + + enum pipe_error cso_set_depth_stencil_alpha(struct cso_context *ctx, const struct pipe_depth_stencil_alpha_state *templ) { diff --git a/src/gallium/auxiliary/cso_cache/cso_context.h b/src/gallium/auxiliary/cso_cache/cso_context.h index 69630e98bae..e5b92177cfd 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.h +++ b/src/gallium/auxiliary/cso_cache/cso_context.h @@ -84,6 +84,20 @@ enum pipe_error cso_single_sampler( struct cso_context *cso, void cso_single_sampler_done( struct cso_context *cso ); +void +cso_save_vertex_samplers(struct cso_context *cso); + +void +cso_restore_vertex_samplers(struct cso_context *cso); + +enum pipe_error +cso_single_vertex_sampler(struct cso_context *cso, + unsigned nr, + const struct pipe_sampler_state *states); + +void +cso_single_vertex_sampler_done(struct cso_context *cso); + enum pipe_error cso_set_sampler_textures( struct cso_context *cso, @@ -94,6 +108,17 @@ void cso_restore_sampler_textures( struct cso_context *cso ); +enum pipe_error +cso_set_vertex_sampler_textures(struct cso_context *cso, + uint count, + struct pipe_texture **textures); +void +cso_save_vertex_sampler_textures(struct cso_context *cso); +void +cso_restore_vertex_sampler_textures(struct cso_context *cso); + + + /* These aren't really sensible -- most of the time the api provides * object semantics for shaders anyway, and the cases where it doesn't * (eg mesa's internall-generated texenv programs), it will be up to diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c index 9f956715a22..ca04223e3d6 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aaline.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c @@ -398,9 +398,9 @@ aaline_create_texture(struct aaline_stage *aaline) texTemp.target = PIPE_TEXTURE_2D; texTemp.format = PIPE_FORMAT_A8_UNORM; /* XXX verify supported by driver! */ texTemp.last_level = MAX_TEXTURE_LEVEL; - texTemp.width[0] = 1 << MAX_TEXTURE_LEVEL; - texTemp.height[0] = 1 << MAX_TEXTURE_LEVEL; - texTemp.depth[0] = 1; + texTemp.width0 = 1 << MAX_TEXTURE_LEVEL; + texTemp.height0 = 1 << MAX_TEXTURE_LEVEL; + texTemp.depth0 = 1; pf_get_block(texTemp.format, &texTemp.block); aaline->texture = screen->texture_create(screen, &texTemp); @@ -413,11 +413,11 @@ aaline_create_texture(struct aaline_stage *aaline) */ for (level = 0; level <= MAX_TEXTURE_LEVEL; level++) { struct pipe_transfer *transfer; - const uint size = aaline->texture->width[level]; + const uint size = u_minify(aaline->texture->width0, level); ubyte *data; uint i, j; - assert(aaline->texture->width[level] == aaline->texture->height[level]); + assert(aaline->texture->width0 == aaline->texture->height0); /* This texture is new, no need to flush. */ @@ -896,16 +896,16 @@ draw_install_aaline_stage(struct draw_context *draw, struct pipe_context *pipe) aaline->driver_bind_fs_state = pipe->bind_fs_state; aaline->driver_delete_fs_state = pipe->delete_fs_state; - aaline->driver_bind_sampler_states = pipe->bind_sampler_states; - aaline->driver_set_sampler_textures = pipe->set_sampler_textures; + aaline->driver_bind_sampler_states = pipe->bind_fragment_sampler_states; + aaline->driver_set_sampler_textures = pipe->set_fragment_sampler_textures; /* override the driver's functions */ pipe->create_fs_state = aaline_create_fs_state; pipe->bind_fs_state = aaline_bind_fs_state; pipe->delete_fs_state = aaline_delete_fs_state; - pipe->bind_sampler_states = aaline_bind_sampler_states; - pipe->set_sampler_textures = aaline_set_sampler_textures; + pipe->bind_fragment_sampler_states = aaline_bind_sampler_states; + pipe->set_fragment_sampler_textures = aaline_set_sampler_textures; /* Install once everything is known to be OK: */ diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c index 283502cdf3e..2fc0b4de556 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c @@ -427,9 +427,9 @@ pstip_create_texture(struct pstip_stage *pstip) texTemp.target = PIPE_TEXTURE_2D; texTemp.format = PIPE_FORMAT_A8_UNORM; /* XXX verify supported by driver! */ texTemp.last_level = 0; - texTemp.width[0] = 32; - texTemp.height[0] = 32; - texTemp.depth[0] = 1; + texTemp.width0 = 32; + texTemp.height0 = 32; + texTemp.depth0 = 1; pf_get_block(texTemp.format, &texTemp.block); pstip->texture = screen->texture_create(screen, &texTemp); @@ -754,8 +754,8 @@ draw_install_pstipple_stage(struct draw_context *draw, pstip->driver_bind_fs_state = pipe->bind_fs_state; pstip->driver_delete_fs_state = pipe->delete_fs_state; - pstip->driver_bind_sampler_states = pipe->bind_sampler_states; - pstip->driver_set_sampler_textures = pipe->set_sampler_textures; + pstip->driver_bind_sampler_states = pipe->bind_fragment_sampler_states; + pstip->driver_set_sampler_textures = pipe->set_fragment_sampler_textures; pstip->driver_set_polygon_stipple = pipe->set_polygon_stipple; /* override the driver's functions */ @@ -763,8 +763,8 @@ draw_install_pstipple_stage(struct draw_context *draw, pipe->bind_fs_state = pstip_bind_fs_state; pipe->delete_fs_state = pstip_delete_fs_state; - pipe->bind_sampler_states = pstip_bind_sampler_states; - pipe->set_sampler_textures = pstip_set_sampler_textures; + pipe->bind_fragment_sampler_states = pstip_bind_sampler_states; + pipe->set_fragment_sampler_textures = pstip_set_sampler_textures; pipe->set_polygon_stipple = pstip_set_polygon_stipple; return TRUE; diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index d3f179ced18..757c4874545 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -346,7 +346,8 @@ vcache_check_run( struct draw_pt_front_end *frontend, vcache->fetch_max, draw_count); - if (max_index == 0xffffffff || + if (max_index >= DRAW_PIPE_MAX_VERTICES || + fetch_count >= UNDEFINED_VERTEX_ID || fetch_count > draw_count) { if (0) debug_printf("fail\n"); goto fail; diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c index 2ef4293d4d7..2f973684f67 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c @@ -1,6 +1,6 @@ /************************************************************************** * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 2007-2009 VMware, Inc. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -18,7 +18,7 @@ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. @@ -80,11 +80,27 @@ struct fenced_buffer_list */ struct fenced_buffer { + /* + * Immutable members. + */ + struct pb_buffer base; - struct pb_buffer *buffer; + struct fenced_buffer_list *list; + + /** + * Protected by fenced_buffer_list::mutex + */ + struct list_head head; - /* FIXME: protect access with mutex */ + /** + * Following members are mutable and protected by this mutex. + * + * You may lock this mutex alone, or lock it with fenced_buffer_list::mutex + * held, but in order to prevent deadlocks you must never lock + * fenced_buffer_list::mutex with this mutex held. + */ + pipe_mutex mutex; /** * A bitmask of PIPE_BUFFER_USAGE_CPU/GPU_READ/WRITE describing the current @@ -96,9 +112,6 @@ struct fenced_buffer struct pb_validate *vl; unsigned validation_flags; struct pipe_fence_handle *fence; - - struct list_head head; - struct fenced_buffer_list *list; }; @@ -110,15 +123,24 @@ fenced_buffer(struct pb_buffer *buf) } +/** + * Add the buffer to the fenced list. + * + * fenced_buffer_list::mutex and fenced_buffer::mutex must be held, in this + * order before calling this function. + * + * Reference count should be incremented before calling this function. + */ static INLINE void -_fenced_buffer_add(struct fenced_buffer *fenced_buf) +fenced_buffer_add_locked(struct fenced_buffer_list *fenced_list, + struct fenced_buffer *fenced_buf) { - struct fenced_buffer_list *fenced_list = fenced_buf->list; - assert(pipe_is_referenced(&fenced_buf->base.base.reference)); assert(fenced_buf->flags & PIPE_BUFFER_USAGE_GPU_READ_WRITE); assert(fenced_buf->fence); + /* TODO: Move the reference count increment here */ + #ifdef DEBUG LIST_DEL(&fenced_buf->head); assert(fenced_list->numUnfenced); @@ -130,32 +152,16 @@ _fenced_buffer_add(struct fenced_buffer *fenced_buf) /** - * Actually destroy the buffer. + * Remove the buffer from the fenced list. + * + * fenced_buffer_list::mutex and fenced_buffer::mutex must be held, in this + * order before calling this function. + * + * Reference count should be decremented after calling this function. */ static INLINE void -_fenced_buffer_destroy(struct fenced_buffer *fenced_buf) -{ - struct fenced_buffer_list *fenced_list = fenced_buf->list; - - assert(!pipe_is_referenced(&fenced_buf->base.base.reference)); - assert(!fenced_buf->fence); -#ifdef DEBUG - assert(fenced_buf->head.prev); - assert(fenced_buf->head.next); - LIST_DEL(&fenced_buf->head); - assert(fenced_list->numUnfenced); - --fenced_list->numUnfenced; -#else - (void)fenced_list; -#endif - pb_reference(&fenced_buf->buffer, NULL); - FREE(fenced_buf); -} - - -static INLINE void -_fenced_buffer_remove(struct fenced_buffer_list *fenced_list, - struct fenced_buffer *fenced_buf) +fenced_buffer_remove_locked(struct fenced_buffer_list *fenced_list, + struct fenced_buffer *fenced_buf) { struct pb_fence_ops *ops = fenced_list->ops; @@ -177,37 +183,53 @@ _fenced_buffer_remove(struct fenced_buffer_list *fenced_list, ++fenced_list->numUnfenced; #endif - /** - * FIXME!!! - */ - - if(!pipe_is_referenced(&fenced_buf->base.base.reference)) - _fenced_buffer_destroy(fenced_buf); + /* TODO: Move the reference count decrement and destruction here */ } +/** + * Wait for the fence to expire, and remove it from the fenced list. + * + * fenced_buffer::mutex must be held. fenced_buffer_list::mutex must not be + * held -- it will + */ static INLINE enum pipe_error -_fenced_buffer_finish(struct fenced_buffer *fenced_buf) +fenced_buffer_finish_locked(struct fenced_buffer_list *fenced_list, + struct fenced_buffer *fenced_buf) { - struct fenced_buffer_list *fenced_list = fenced_buf->list; struct pb_fence_ops *ops = fenced_list->ops; + enum pipe_error ret = PIPE_ERROR; #if 0 debug_warning("waiting for GPU"); #endif + assert(pipe_is_referenced(&fenced_buf->base.base.reference)); assert(fenced_buf->fence); + + /* Acquire the global lock */ + pipe_mutex_unlock(fenced_buf->mutex); + pipe_mutex_lock(fenced_list->mutex); + pipe_mutex_lock(fenced_buf->mutex); + if(fenced_buf->fence) { - if(ops->fence_finish(ops, fenced_buf->fence, 0) != 0) { - return PIPE_ERROR; + if(ops->fence_finish(ops, fenced_buf->fence, 0) == 0) { + /* Remove from the fenced list */ + /* TODO: remove consequents */ + fenced_buffer_remove_locked(fenced_list, fenced_buf); + + p_atomic_dec(&fenced_buf->base.base.reference.count); + assert(pipe_is_referenced(&fenced_buf->base.base.reference)); + + fenced_buf->flags &= ~PIPE_BUFFER_USAGE_GPU_READ_WRITE; + + ret = PIPE_OK; } - /* Remove from the fenced list */ - /* TODO: remove consequents */ - _fenced_buffer_remove(fenced_list, fenced_buf); } - fenced_buf->flags &= ~PIPE_BUFFER_USAGE_GPU_READ_WRITE; - return PIPE_OK; + pipe_mutex_unlock(fenced_list->mutex); + + return ret; } @@ -215,7 +237,7 @@ _fenced_buffer_finish(struct fenced_buffer *fenced_buf) * Free as many fenced buffers from the list head as possible. */ static void -_fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, +fenced_buffer_list_check_free_locked(struct fenced_buffer_list *fenced_list, int wait) { struct pb_fence_ops *ops = fenced_list->ops; @@ -228,21 +250,28 @@ _fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, while(curr != &fenced_list->delayed) { fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head); + pipe_mutex_lock(fenced_buf->mutex); + if(fenced_buf->fence != prev_fence) { int signaled; if (wait) signaled = ops->fence_finish(ops, fenced_buf->fence, 0); else signaled = ops->fence_signalled(ops, fenced_buf->fence, 0); - if (signaled != 0) + if (signaled != 0) { + pipe_mutex_unlock(fenced_buf->mutex); break; + } prev_fence = fenced_buf->fence; } else { assert(ops->fence_signalled(ops, fenced_buf->fence, 0) == 0); } - _fenced_buffer_remove(fenced_list, fenced_buf); + fenced_buffer_remove_locked(fenced_list, fenced_buf); + pipe_mutex_unlock(fenced_buf->mutex); + + pb_reference((struct pb_buffer **)&fenced_buf, NULL); curr = next; next = curr->next; @@ -256,30 +285,25 @@ fenced_buffer_destroy(struct pb_buffer *buf) struct fenced_buffer *fenced_buf = fenced_buffer(buf); struct fenced_buffer_list *fenced_list = fenced_buf->list; - pipe_mutex_lock(fenced_list->mutex); assert(!pipe_is_referenced(&fenced_buf->base.base.reference)); - if (fenced_buf->fence) { - struct pb_fence_ops *ops = fenced_list->ops; - if(ops->fence_signalled(ops, fenced_buf->fence, 0) == 0) { - struct list_head *curr, *prev; - curr = &fenced_buf->head; - prev = curr->prev; - do { - fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head); - assert(ops->fence_signalled(ops, fenced_buf->fence, 0) == 0); - _fenced_buffer_remove(fenced_list, fenced_buf); - curr = prev; - prev = curr->prev; - } while (curr != &fenced_list->delayed); - } - else { - /* delay destruction */ - } - } - else { - _fenced_buffer_destroy(fenced_buf); - } + assert(!fenced_buf->fence); + +#ifdef DEBUG + pipe_mutex_lock(fenced_list->mutex); + assert(fenced_buf->head.prev); + assert(fenced_buf->head.next); + LIST_DEL(&fenced_buf->head); + assert(fenced_list->numUnfenced); + --fenced_list->numUnfenced; pipe_mutex_unlock(fenced_list->mutex); +#else + (void)fenced_list; +#endif + + pb_reference(&fenced_buf->buffer, NULL); + + pipe_mutex_destroy(fenced_buf->mutex); + FREE(fenced_buf); } @@ -290,24 +314,23 @@ fenced_buffer_map(struct pb_buffer *buf, struct fenced_buffer *fenced_buf = fenced_buffer(buf); struct fenced_buffer_list *fenced_list = fenced_buf->list; struct pb_fence_ops *ops = fenced_list->ops; - void *map; + void *map = NULL; + + pipe_mutex_lock(fenced_buf->mutex); assert(!(flags & PIPE_BUFFER_USAGE_GPU_READ_WRITE)); /* Serialize writes */ if((fenced_buf->flags & PIPE_BUFFER_USAGE_GPU_WRITE) || ((fenced_buf->flags & PIPE_BUFFER_USAGE_GPU_READ) && (flags & PIPE_BUFFER_USAGE_CPU_WRITE))) { - if(flags & PIPE_BUFFER_USAGE_DONTBLOCK) { + if((flags & PIPE_BUFFER_USAGE_DONTBLOCK) && + ops->fence_signalled(ops, fenced_buf->fence, 0) == 0) { /* Don't wait for the GPU to finish writing */ - if(ops->fence_signalled(ops, fenced_buf->fence, 0) == 0) - _fenced_buffer_remove(fenced_list, fenced_buf); - else - return NULL; - } - else { - /* Wait for the GPU to finish writing */ - _fenced_buffer_finish(fenced_buf); + goto finish; } + + /* Wait for the GPU to finish writing */ + fenced_buffer_finish_locked(fenced_list, fenced_buf); } #if 0 @@ -324,6 +347,9 @@ fenced_buffer_map(struct pb_buffer *buf, fenced_buf->flags |= flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE; } +finish: + pipe_mutex_unlock(fenced_buf->mutex); + return map; } @@ -332,6 +358,9 @@ static void fenced_buffer_unmap(struct pb_buffer *buf) { struct fenced_buffer *fenced_buf = fenced_buffer(buf); + + pipe_mutex_lock(fenced_buf->mutex); + assert(fenced_buf->mapcount); if(fenced_buf->mapcount) { pb_unmap(fenced_buf->buffer); @@ -339,6 +368,8 @@ fenced_buffer_unmap(struct pb_buffer *buf) if(!fenced_buf->mapcount) fenced_buf->flags &= ~PIPE_BUFFER_USAGE_CPU_READ_WRITE; } + + pipe_mutex_unlock(fenced_buf->mutex); } @@ -350,11 +381,14 @@ fenced_buffer_validate(struct pb_buffer *buf, struct fenced_buffer *fenced_buf = fenced_buffer(buf); enum pipe_error ret; + pipe_mutex_lock(fenced_buf->mutex); + if(!vl) { /* invalidate */ fenced_buf->vl = NULL; fenced_buf->validation_flags = 0; - return PIPE_OK; + ret = PIPE_OK; + goto finish; } assert(flags & PIPE_BUFFER_USAGE_GPU_READ_WRITE); @@ -362,14 +396,17 @@ fenced_buffer_validate(struct pb_buffer *buf, flags &= PIPE_BUFFER_USAGE_GPU_READ_WRITE; /* Buffer cannot be validated in two different lists */ - if(fenced_buf->vl && fenced_buf->vl != vl) - return PIPE_ERROR_RETRY; + if(fenced_buf->vl && fenced_buf->vl != vl) { + ret = PIPE_ERROR_RETRY; + goto finish; + } #if 0 /* Do not validate if buffer is still mapped */ if(fenced_buf->flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE) { /* TODO: wait for the thread that mapped the buffer to unmap it */ - return PIPE_ERROR_RETRY; + ret = PIPE_ERROR_RETRY; + goto finish; } /* Final sanity checking */ assert(!(fenced_buf->flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE)); @@ -379,17 +416,21 @@ fenced_buffer_validate(struct pb_buffer *buf, if(fenced_buf->vl == vl && (fenced_buf->validation_flags & flags) == flags) { /* Nothing to do -- buffer already validated */ - return PIPE_OK; + ret = PIPE_OK; + goto finish; } ret = pb_validate(fenced_buf->buffer, vl, flags); if (ret != PIPE_OK) - return ret; + goto finish; fenced_buf->vl = vl; fenced_buf->validation_flags |= flags; - return PIPE_OK; +finish: + pipe_mutex_unlock(fenced_buf->mutex); + + return ret; } @@ -404,29 +445,36 @@ fenced_buffer_fence(struct pb_buffer *buf, fenced_buf = fenced_buffer(buf); fenced_list = fenced_buf->list; ops = fenced_list->ops; - - if(fence == fenced_buf->fence) { - /* Nothing to do */ - return; - } - assert(fenced_buf->vl); - assert(fenced_buf->validation_flags); - pipe_mutex_lock(fenced_list->mutex); - if (fenced_buf->fence) - _fenced_buffer_remove(fenced_list, fenced_buf); - if (fence) { - ops->fence_reference(ops, &fenced_buf->fence, fence); - fenced_buf->flags |= fenced_buf->validation_flags; - _fenced_buffer_add(fenced_buf); - } - pipe_mutex_unlock(fenced_list->mutex); + pipe_mutex_lock(fenced_buf->mutex); + + assert(pipe_is_referenced(&fenced_buf->base.base.reference)); + + if(fence != fenced_buf->fence) { + assert(fenced_buf->vl); + assert(fenced_buf->validation_flags); + + if (fenced_buf->fence) { + fenced_buffer_remove_locked(fenced_list, fenced_buf); + p_atomic_dec(&fenced_buf->base.base.reference.count); + assert(pipe_is_referenced(&fenced_buf->base.base.reference)); + } + if (fence) { + ops->fence_reference(ops, &fenced_buf->fence, fence); + fenced_buf->flags |= fenced_buf->validation_flags; + p_atomic_inc(&fenced_buf->base.base.reference.count); + fenced_buffer_add_locked(fenced_list, fenced_buf); + } + + pb_fence(fenced_buf->buffer, fence); - pb_fence(fenced_buf->buffer, fence); + fenced_buf->vl = NULL; + fenced_buf->validation_flags = 0; + } - fenced_buf->vl = NULL; - fenced_buf->validation_flags = 0; + pipe_mutex_unlock(fenced_buf->mutex); + pipe_mutex_unlock(fenced_list->mutex); } @@ -436,6 +484,7 @@ fenced_buffer_get_base_buffer(struct pb_buffer *buf, pb_size *offset) { struct fenced_buffer *fenced_buf = fenced_buffer(buf); + /* NOTE: accesses immutable members only -- mutex not necessary */ pb_get_base_buffer(fenced_buf->buffer, base_buf, offset); } @@ -475,6 +524,8 @@ fenced_buffer_create(struct fenced_buffer_list *fenced_list, buf->buffer = buffer; buf->list = fenced_list; + pipe_mutex_init(buf->mutex); + #ifdef DEBUG pipe_mutex_lock(fenced_list->mutex); LIST_ADDTAIL(&buf->head, &fenced_list->unfenced); @@ -516,7 +567,7 @@ fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, int wait) { pipe_mutex_lock(fenced_list->mutex); - _fenced_buffer_list_check_free(fenced_list, wait); + fenced_buffer_list_check_free_locked(fenced_list, wait); pipe_mutex_unlock(fenced_list->mutex); } @@ -538,11 +589,13 @@ fenced_buffer_list_dump(struct fenced_buffer_list *fenced_list) next = curr->next; while(curr != &fenced_list->unfenced) { fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head); + pipe_mutex_lock(fenced_buf->mutex); assert(!fenced_buf->fence); debug_printf("%10p %7u %7u\n", (void *) fenced_buf, fenced_buf->base.base.size, p_atomic_read(&fenced_buf->base.base.reference.count)); + pipe_mutex_unlock(fenced_buf->mutex); curr = next; next = curr->next; } @@ -552,6 +605,7 @@ fenced_buffer_list_dump(struct fenced_buffer_list *fenced_list) while(curr != &fenced_list->delayed) { int signaled; fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head); + pipe_mutex_lock(fenced_buf->mutex); signaled = ops->fence_signalled(ops, fenced_buf->fence, 0); debug_printf("%10p %7u %7u %10p %s\n", (void *) fenced_buf, @@ -559,6 +613,7 @@ fenced_buffer_list_dump(struct fenced_buffer_list *fenced_list) p_atomic_read(&fenced_buf->base.base.reference.count), (void *) fenced_buf->fence, signaled == 0 ? "y" : "n"); + pipe_mutex_unlock(fenced_buf->mutex); curr = next; next = curr->next; } @@ -579,8 +634,8 @@ fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list) #if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) sched_yield(); #endif - _fenced_buffer_list_check_free(fenced_list, 1); pipe_mutex_lock(fenced_list->mutex); + fenced_buffer_list_check_free_locked(fenced_list, 1); } #ifdef DEBUG @@ -588,6 +643,7 @@ fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list) #endif pipe_mutex_unlock(fenced_list->mutex); + pipe_mutex_destroy(fenced_list->mutex); fenced_list->ops->destroy(fenced_list->ops); diff --git a/src/gallium/auxiliary/tgsi/tgsi-instruction-set.txt b/src/gallium/auxiliary/tgsi/tgsi-instruction-set.txt index eb492076b7d..080fd4c7310 100644 --- a/src/gallium/auxiliary/tgsi/tgsi-instruction-set.txt +++ b/src/gallium/auxiliary/tgsi/tgsi-instruction-set.txt @@ -1129,3 +1129,35 @@ TGSI Instruction Specification target Label of target instruction. + +3 Other tokens +=============== + + +3.1 Declaration Semantic +------------------------- + + + Follows Declaration token if Semantic bit is set. + + Since its purpose is to link a shader with other stages of the pipeline, + it is valid to follow only those Declaration tokens that declare a register + either in INPUT or OUTPUT file. + + SemanticName field contains the semantic name of the register being declared. + There is no default value. + + SemanticIndex is an optional subscript that can be used to distinguish + different register declarations with the same semantic name. The default value + is 0. + + The meanings of the individual semantic names are explained in the following + sections. + + +3.1.1 FACE + + Valid only in a fragment shader INPUT declaration. + + FACE.x is negative when the primitive is back facing. FACE.x is positive + when the primitive is front facing. diff --git a/src/gallium/auxiliary/tgsi/tgsi_build.c b/src/gallium/auxiliary/tgsi/tgsi_build.c index 4fa10e2f7e3..9791e58db3b 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_build.c +++ b/src/gallium/auxiliary/tgsi/tgsi_build.c @@ -419,6 +419,7 @@ tgsi_default_instruction( void ) instruction.NrTokens = 1; instruction.Opcode = TGSI_OPCODE_MOV; instruction.Saturate = TGSI_SAT_NONE; + instruction.Predicate = 0; instruction.NumDstRegs = 1; instruction.NumSrcRegs = 1; instruction.Padding = 0; @@ -428,12 +429,12 @@ tgsi_default_instruction( void ) } struct tgsi_instruction -tgsi_build_instruction( - unsigned opcode, - unsigned saturate, - unsigned num_dst_regs, - unsigned num_src_regs, - struct tgsi_header *header ) +tgsi_build_instruction(unsigned opcode, + unsigned saturate, + unsigned predicate, + unsigned num_dst_regs, + unsigned num_src_regs, + struct tgsi_header *header) { struct tgsi_instruction instruction; @@ -445,6 +446,7 @@ tgsi_build_instruction( instruction = tgsi_default_instruction(); instruction.Opcode = opcode; instruction.Saturate = saturate; + instruction.Predicate = predicate; instruction.NumDstRegs = num_dst_regs; instruction.NumSrcRegs = num_src_regs; @@ -472,9 +474,9 @@ tgsi_default_full_instruction( void ) unsigned i; full_instruction.Instruction = tgsi_default_instruction(); + full_instruction.InstructionPredicate = tgsi_default_instruction_predicate(); full_instruction.InstructionExtLabel = tgsi_default_instruction_ext_label(); full_instruction.InstructionExtTexture = tgsi_default_instruction_ext_texture(); - full_instruction.InstructionExtPredicate = tgsi_default_instruction_ext_predicate(); for( i = 0; i < TGSI_FULL_MAX_DST_REGISTERS; i++ ) { full_instruction.FullDstRegisters[i] = tgsi_default_full_dst_register(); } @@ -504,14 +506,34 @@ tgsi_build_full_instruction( instruction = (struct tgsi_instruction *) &tokens[size]; size++; - *instruction = tgsi_build_instruction( - full_inst->Instruction.Opcode, - full_inst->Instruction.Saturate, - full_inst->Instruction.NumDstRegs, - full_inst->Instruction.NumSrcRegs, - header ); + *instruction = tgsi_build_instruction(full_inst->Instruction.Opcode, + full_inst->Instruction.Saturate, + full_inst->Instruction.Predicate, + full_inst->Instruction.NumDstRegs, + full_inst->Instruction.NumSrcRegs, + header); prev_token = (struct tgsi_token *) instruction; + if (full_inst->Instruction.Predicate) { + struct tgsi_instruction_predicate *instruction_predicate; + + if (maxsize <= size) { + return 0; + } + instruction_predicate = (struct tgsi_instruction_predicate *)&tokens[size]; + size++; + + *instruction_predicate = + tgsi_build_instruction_predicate(full_inst->InstructionPredicate.Index, + full_inst->InstructionPredicate.Negate, + full_inst->InstructionPredicate.SwizzleX, + full_inst->InstructionPredicate.SwizzleY, + full_inst->InstructionPredicate.SwizzleZ, + full_inst->InstructionPredicate.SwizzleW, + instruction, + header); + } + if( tgsi_compare_instruction_ext_label( full_inst->InstructionExtLabel, tgsi_default_instruction_ext_label() ) ) { @@ -550,29 +572,6 @@ tgsi_build_full_instruction( prev_token = (struct tgsi_token *) instruction_ext_texture; } - if (tgsi_compare_instruction_ext_predicate(full_inst->InstructionExtPredicate, - tgsi_default_instruction_ext_predicate())) { - struct tgsi_instruction_ext_predicate *instruction_ext_predicate; - - if (maxsize <= size) { - return 0; - } - instruction_ext_predicate = (struct tgsi_instruction_ext_predicate *)&tokens[size]; - size++; - - *instruction_ext_predicate = - tgsi_build_instruction_ext_predicate(full_inst->InstructionExtPredicate.SrcIndex, - full_inst->InstructionExtPredicate.Negate, - full_inst->InstructionExtPredicate.SwizzleX, - full_inst->InstructionExtPredicate.SwizzleY, - full_inst->InstructionExtPredicate.SwizzleZ, - full_inst->InstructionExtPredicate.SwizzleW, - prev_token, - instruction, - header); - prev_token = (struct tgsi_token *)instruction_ext_predicate; - } - for( i = 0; i < full_inst->Instruction.NumDstRegs; i++ ) { const struct tgsi_full_dst_register *reg = &full_inst->FullDstRegisters[i]; struct tgsi_dst_register *dst_register; @@ -746,6 +745,47 @@ tgsi_build_full_instruction( return size; } +struct tgsi_instruction_predicate +tgsi_default_instruction_predicate(void) +{ + struct tgsi_instruction_predicate instruction_predicate; + + instruction_predicate.SwizzleX = TGSI_SWIZZLE_X; + instruction_predicate.SwizzleY = TGSI_SWIZZLE_Y; + instruction_predicate.SwizzleZ = TGSI_SWIZZLE_Z; + instruction_predicate.SwizzleW = TGSI_SWIZZLE_W; + instruction_predicate.Negate = 0; + instruction_predicate.Index = 0; + instruction_predicate.Padding = 0; + + return instruction_predicate; +} + +struct tgsi_instruction_predicate +tgsi_build_instruction_predicate(int index, + unsigned negate, + unsigned swizzleX, + unsigned swizzleY, + unsigned swizzleZ, + unsigned swizzleW, + struct tgsi_instruction *instruction, + struct tgsi_header *header) +{ + struct tgsi_instruction_predicate instruction_predicate; + + instruction_predicate = tgsi_default_instruction_predicate(); + instruction_predicate.SwizzleX = swizzleX; + instruction_predicate.SwizzleY = swizzleY; + instruction_predicate.SwizzleZ = swizzleZ; + instruction_predicate.SwizzleW = swizzleW; + instruction_predicate.Negate = negate; + instruction_predicate.Index = index; + + instruction_grow(instruction, header); + + return instruction_predicate; +} + /** test for inequality of 32-bit values pointed to by a and b */ static INLINE boolean compare32(const void *a, const void *b) @@ -835,60 +875,6 @@ tgsi_build_instruction_ext_texture( return instruction_ext_texture; } -struct tgsi_instruction_ext_predicate -tgsi_default_instruction_ext_predicate(void) -{ - struct tgsi_instruction_ext_predicate instruction_ext_predicate; - - instruction_ext_predicate.Type = TGSI_INSTRUCTION_EXT_TYPE_PREDICATE; - instruction_ext_predicate.SwizzleX = TGSI_SWIZZLE_X; - instruction_ext_predicate.SwizzleY = TGSI_SWIZZLE_Y; - instruction_ext_predicate.SwizzleZ = TGSI_SWIZZLE_Z; - instruction_ext_predicate.SwizzleW = TGSI_SWIZZLE_W; - instruction_ext_predicate.Negate = 0; - instruction_ext_predicate.SrcIndex = 0; - instruction_ext_predicate.Padding = 0; - instruction_ext_predicate.Extended = 0; - - return instruction_ext_predicate; -} - -unsigned -tgsi_compare_instruction_ext_predicate(struct tgsi_instruction_ext_predicate a, - struct tgsi_instruction_ext_predicate b) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return compare32(&a, &b); -} - -struct tgsi_instruction_ext_predicate -tgsi_build_instruction_ext_predicate(unsigned index, - unsigned negate, - unsigned swizzleX, - unsigned swizzleY, - unsigned swizzleZ, - unsigned swizzleW, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header) -{ - struct tgsi_instruction_ext_predicate instruction_ext_predicate; - - instruction_ext_predicate = tgsi_default_instruction_ext_predicate(); - instruction_ext_predicate.SwizzleX = swizzleX; - instruction_ext_predicate.SwizzleY = swizzleY; - instruction_ext_predicate.SwizzleZ = swizzleZ; - instruction_ext_predicate.SwizzleW = swizzleW; - instruction_ext_predicate.Negate = negate; - instruction_ext_predicate.SrcIndex = index; - - prev_token->Extended = 1; - instruction_grow(instruction, header); - - return instruction_ext_predicate; -} - struct tgsi_src_register tgsi_default_src_register( void ) { diff --git a/src/gallium/auxiliary/tgsi/tgsi_build.h b/src/gallium/auxiliary/tgsi/tgsi_build.h index 669712eb8f9..0fe5f229d37 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_build.h +++ b/src/gallium/auxiliary/tgsi/tgsi_build.h @@ -143,6 +143,7 @@ struct tgsi_instruction tgsi_build_instruction( unsigned opcode, unsigned saturate, + unsigned predicate, unsigned num_dst_regs, unsigned num_src_regs, struct tgsi_header *header ); @@ -157,6 +158,19 @@ tgsi_build_full_instruction( struct tgsi_header *header, unsigned maxsize ); +struct tgsi_instruction_predicate +tgsi_default_instruction_predicate(void); + +struct tgsi_instruction_predicate +tgsi_build_instruction_predicate(int index, + unsigned negate, + unsigned swizzleX, + unsigned swizzleY, + unsigned swizzleZ, + unsigned swizzleW, + struct tgsi_instruction *instruction, + struct tgsi_header *header); + struct tgsi_instruction_ext_label tgsi_default_instruction_ext_label( void ); @@ -187,24 +201,6 @@ tgsi_build_instruction_ext_texture( struct tgsi_instruction *instruction, struct tgsi_header *header ); -struct tgsi_instruction_ext_predicate -tgsi_default_instruction_ext_predicate(void); - -unsigned -tgsi_compare_instruction_ext_predicate(struct tgsi_instruction_ext_predicate a, - struct tgsi_instruction_ext_predicate b); - -struct tgsi_instruction_ext_predicate -tgsi_build_instruction_ext_predicate(unsigned index, - unsigned negate, - unsigned swizzleX, - unsigned swizzleY, - unsigned swizzleZ, - unsigned swizzleW, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header); - struct tgsi_src_register tgsi_default_src_register( void ); diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c index b7569e74d4b..e22a1643c88 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c @@ -62,9 +62,6 @@ #define FAST_MATH 1 -/** for tgsi_full_instruction::Flags */ -#define SOA_DEPENDENCY_FLAG 0x1 - #define TILE_TOP_LEFT 0 #define TILE_TOP_RIGHT 1 #define TILE_BOTTOM_LEFT 2 @@ -332,20 +329,6 @@ tgsi_exec_machine_bind_shader( maxInstructions += 10; } - if (tgsi_check_soa_dependencies(&parse.FullToken.FullInstruction)) { - uint opcode = parse.FullToken.FullInstruction.Instruction.Opcode; - parse.FullToken.FullInstruction.Flags = SOA_DEPENDENCY_FLAG; - /* XXX we only handle SOA dependencies properly for MOV/SWZ - * at this time! - */ - if (opcode != TGSI_OPCODE_MOV) { - debug_printf("Warning: SOA dependency in instruction" - " is not handled:\n"); - tgsi_dump_instruction(&parse.FullToken.FullInstruction, - numInstructions); - } - } - memcpy(instructions + numInstructions, &parse.FullToken.FullInstruction, sizeof(instructions[0])); @@ -386,6 +369,7 @@ tgsi_exec_machine_create( void ) memset(mach, 0, sizeof(*mach)); mach->Addrs = &mach->Temps[TGSI_EXEC_TEMP_ADDR]; + mach->Predicates = &mach->Temps[TGSI_EXEC_TEMP_P0]; /* Setup constants. */ for( i = 0; i < 4; i++ ) { @@ -517,7 +501,7 @@ micro_ddy( dst->f[0] = dst->f[1] = dst->f[2] = - dst->f[3] = src->f[TILE_TOP_LEFT] - src->f[TILE_BOTTOM_LEFT]; + dst->f[3] = src->f[TILE_BOTTOM_LEFT] - src->f[TILE_TOP_LEFT]; } static void @@ -595,6 +579,24 @@ micro_exp2( dst->f[2] = util_fast_exp2( src->f[2] ); dst->f[3] = util_fast_exp2( src->f[3] ); #else + +#if DEBUG + /* Inf is okay for this instruction, so clamp it to silence assertions. */ + uint i; + union tgsi_exec_channel clamped; + + for (i = 0; i < 4; i++) { + if (src->f[i] > 127.99999f) { + clamped.f[i] = 127.99999f; + } else if (src->f[i] < -126.99999f) { + clamped.f[i] = -126.99999f; + } else { + clamped.f[i] = src->f[i]; + } + } + src = &clamped; +#endif + dst->f[0] = powf( 2.0f, src->f[0] ); dst->f[1] = powf( 2.0f, src->f[1] ); dst->f[2] = powf( 2.0f, src->f[2] ); @@ -1193,10 +1195,10 @@ fetch_src_file_channel( assert(index->i[1] < TGSI_EXEC_NUM_PREDS); assert(index->i[2] < TGSI_EXEC_NUM_PREDS); assert(index->i[3] < TGSI_EXEC_NUM_PREDS); - chan->u[0] = mach->Addrs[0].xyzw[swizzle].u[0]; - chan->u[1] = mach->Addrs[0].xyzw[swizzle].u[1]; - chan->u[2] = mach->Addrs[0].xyzw[swizzle].u[2]; - chan->u[3] = mach->Addrs[0].xyzw[swizzle].u[3]; + chan->u[0] = mach->Predicates[0].xyzw[swizzle].u[0]; + chan->u[1] = mach->Predicates[0].xyzw[swizzle].u[1]; + chan->u[2] = mach->Predicates[0].xyzw[swizzle].u[2]; + chan->u[3] = mach->Predicates[0].xyzw[swizzle].u[3]; break; case TGSI_FILE_OUTPUT: @@ -1478,10 +1480,17 @@ store_dest( dst = &mach->Addrs[index].xyzw[chan_index]; break; + case TGSI_FILE_LOOP: + assert(reg->DstRegister.Index == 0); + assert(mach->LoopCounterStackTop > 0); + assert(chan_index == CHAN_X); + dst = &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[chan_index]; + break; + case TGSI_FILE_PREDICATE: index = reg->DstRegister.Index; assert(index < TGSI_EXEC_NUM_PREDS); - dst = &mach->Addrs[index].xyzw[chan_index]; + dst = &mach->Predicates[index].xyzw[chan_index]; break; default: @@ -1489,6 +1498,47 @@ store_dest( return; } + if (inst->Instruction.Predicate) { + uint swizzle; + union tgsi_exec_channel *pred; + + switch (chan_index) { + case CHAN_X: + swizzle = inst->InstructionPredicate.SwizzleX; + break; + case CHAN_Y: + swizzle = inst->InstructionPredicate.SwizzleY; + break; + case CHAN_Z: + swizzle = inst->InstructionPredicate.SwizzleZ; + break; + case CHAN_W: + swizzle = inst->InstructionPredicate.SwizzleW; + break; + default: + assert(0); + return; + } + + assert(inst->InstructionPredicate.Index == 0); + + pred = &mach->Predicates[inst->InstructionPredicate.Index].xyzw[swizzle]; + + if (inst->InstructionPredicate.Negate) { + for (i = 0; i < QUAD_SIZE; i++) { + if (pred->u[i]) { + execmask &= ~(1 << i); + } + } + } else { + for (i = 0; i < QUAD_SIZE; i++) { + if (!pred->u[i]) { + execmask &= ~(1 << i); + } + } + } + } + switch (inst->Instruction.Saturate) { case TGSI_SAT_NONE: for (i = 0; i < QUAD_SIZE; i++) @@ -1715,6 +1765,64 @@ exec_tex(struct tgsi_exec_machine *mach, } } +static void +exec_txd(struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst) +{ + const uint unit = inst->FullSrcRegisters[3].SrcRegister.Index; + union tgsi_exec_channel r[4]; + uint chan_index; + + /* + * XXX: This is fake TXD -- the derivatives are not taken into account, yet. + */ + + switch (inst->InstructionExtTexture.Texture) { + case TGSI_TEXTURE_1D: + case TGSI_TEXTURE_SHADOW1D: + + FETCH(&r[0], 0, CHAN_X); + + fetch_texel(mach->Samplers[unit], + &r[0], &ZeroVec, &ZeroVec, 0.0f, /* S, T, P, BIAS */ + &r[0], &r[1], &r[2], &r[3]); /* R, G, B, A */ + break; + + case TGSI_TEXTURE_2D: + case TGSI_TEXTURE_RECT: + case TGSI_TEXTURE_SHADOW2D: + case TGSI_TEXTURE_SHADOWRECT: + + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 0, CHAN_Z); + + fetch_texel(mach->Samplers[unit], + &r[0], &r[1], &r[2], 0.0f, /* inputs */ + &r[0], &r[1], &r[2], &r[3]); /* outputs */ + break; + + case TGSI_TEXTURE_3D: + case TGSI_TEXTURE_CUBE: + + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 0, CHAN_Z); + + fetch_texel(mach->Samplers[unit], + &r[0], &r[1], &r[2], 0.0f, + &r[0], &r[1], &r[2], &r[3]); + break; + + default: + assert(0); + } + + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&r[chan_index], 0, chan_index); + } +} + /** * Evaluate a constant-valued coefficient at the position of the @@ -1784,53 +1892,58 @@ typedef void (* eval_coef_func)( unsigned chan ); static void -exec_declaration( - struct tgsi_exec_machine *mach, - const struct tgsi_full_declaration *decl ) +exec_declaration(struct tgsi_exec_machine *mach, + const struct tgsi_full_declaration *decl) { - if( mach->Processor == TGSI_PROCESSOR_FRAGMENT ) { - if( decl->Declaration.File == TGSI_FILE_INPUT ) { - unsigned first, last, mask; - eval_coef_func eval; + if (mach->Processor == TGSI_PROCESSOR_FRAGMENT) { + if (decl->Declaration.File == TGSI_FILE_INPUT) { + uint first, last, mask; first = decl->DeclarationRange.First; last = decl->DeclarationRange.Last; mask = decl->Declaration.UsageMask; - switch( decl->Declaration.Interpolate ) { - case TGSI_INTERPOLATE_CONSTANT: - eval = eval_constant_coef; - break; - - case TGSI_INTERPOLATE_LINEAR: - eval = eval_linear_coef; - break; - - case TGSI_INTERPOLATE_PERSPECTIVE: - eval = eval_perspective_coef; - break; + if (decl->Semantic.SemanticName == TGSI_SEMANTIC_POSITION) { + assert(decl->Semantic.SemanticIndex == 0); + assert(first == last); + assert(mask = TGSI_WRITEMASK_XYZW); - default: - assert( 0 ); - return; - } + mach->Inputs[first] = mach->QuadPos; + } else if (decl->Semantic.SemanticName == TGSI_SEMANTIC_FACE) { + uint i; - if( mask == TGSI_WRITEMASK_XYZW ) { - unsigned i, j; + assert(decl->Semantic.SemanticIndex == 0); + assert(first == last); - for( i = first; i <= last; i++ ) { - for( j = 0; j < NUM_CHANNELS; j++ ) { - eval( mach, i, j ); - } + for (i = 0; i < QUAD_SIZE; i++) { + mach->Inputs[first].xyzw[0].f[i] = mach->Face; + } + } else { + eval_coef_func eval; + uint i, j; + + switch (decl->Declaration.Interpolate) { + case TGSI_INTERPOLATE_CONSTANT: + eval = eval_constant_coef; + break; + + case TGSI_INTERPOLATE_LINEAR: + eval = eval_linear_coef; + break; + + case TGSI_INTERPOLATE_PERSPECTIVE: + eval = eval_perspective_coef; + break; + + default: + assert(0); + return; } - } - else { - unsigned i, j; - for( j = 0; j < NUM_CHANNELS; j++ ) { - if( mask & (1 << j) ) { - for( i = first; i <= last; i++ ) { - eval( mach, i, j ); + for (j = 0; j < NUM_CHANNELS; j++) { + if (mask & (1 << j)) { + for (i = first; i <= last; i++) { + eval(mach, i, j); } } } @@ -1847,6 +1960,7 @@ exec_instruction( { uint chan_index; union tgsi_exec_channel r[10]; + union tgsi_exec_channel d[8]; (*pc)++; @@ -1855,42 +1969,27 @@ exec_instruction( case TGSI_OPCODE_FLR: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); - micro_flr( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); + micro_flr(&d[chan_index], &r[0]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; case TGSI_OPCODE_MOV: - if (inst->Flags & SOA_DEPENDENCY_FLAG) { - /* Do all fetches into temp regs, then do all stores to avoid - * intermediate/accidental clobbering. This could be done all the - * time for MOV but for other instructions we'll need more temps... - */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[chan_index], 0, chan_index ); - } - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[chan_index], 0, chan_index ); - } + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&d[chan_index], 0, chan_index); } - else { - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - STORE( &r[0], 0, chan_index ); - } + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE(&d[chan_index], 0, chan_index); } break; case TGSI_OPCODE_LIT: - if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); - } - if (IS_CHANNEL_ENABLED( *inst, CHAN_Y ) || IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { FETCH( &r[0], 0, CHAN_X ); if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { - micro_max( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, CHAN_Y ); + micro_max(&d[CHAN_Y], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]); } if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { @@ -1901,11 +2000,19 @@ exec_instruction( micro_min( &r[2], &r[2], &mach->Temps[TEMP_128_I].xyzw[TEMP_128_C] ); micro_max( &r[2], &r[2], &mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C] ); micro_pow( &r[1], &r[1], &r[2] ); - micro_lt( &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, CHAN_Z ); + micro_lt(&d[CHAN_Z], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]); } - } + if (IS_CHANNEL_ENABLED(*inst, CHAN_Y)) { + STORE(&d[CHAN_Y], 0, CHAN_Y); + } + if (IS_CHANNEL_ENABLED(*inst, CHAN_Z)) { + STORE(&d[CHAN_Z], 0, CHAN_Z); + } + } + if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); + } if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); } @@ -1973,14 +2080,13 @@ exec_instruction( break; case TGSI_OPCODE_MUL: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) - { + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { FETCH(&r[0], 0, chan_index); FETCH(&r[1], 1, chan_index); - - micro_mul( &r[0], &r[0], &r[1] ); - - STORE(&r[0], 0, chan_index); + micro_mul(&d[chan_index], &r[0], &r[1]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; @@ -1988,8 +2094,10 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_add( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); + micro_add(&d[chan_index], &r[0], &r[1]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; @@ -2045,25 +2153,29 @@ exec_instruction( break; case TGSI_OPCODE_DST: - if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); - } - if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { FETCH( &r[0], 0, CHAN_Y ); FETCH( &r[1], 1, CHAN_Y); - micro_mul( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, CHAN_Y ); + micro_mul(&d[CHAN_Y], &r[0], &r[1]); } - if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - FETCH( &r[0], 0, CHAN_Z ); - STORE( &r[0], 0, CHAN_Z ); + FETCH(&d[CHAN_Z], 0, CHAN_Z); } - if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { - FETCH( &r[0], 1, CHAN_W ); - STORE( &r[0], 0, CHAN_W ); + FETCH(&d[CHAN_W], 1, CHAN_W); + } + + if (IS_CHANNEL_ENABLED(*inst, CHAN_X)) { + STORE(&mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X); + } + if (IS_CHANNEL_ENABLED(*inst, CHAN_Y)) { + STORE(&d[CHAN_Y], 0, CHAN_Y); + } + if (IS_CHANNEL_ENABLED(*inst, CHAN_Z)) { + STORE(&d[CHAN_Z], 0, CHAN_Z); + } + if (IS_CHANNEL_ENABLED(*inst, CHAN_W)) { + STORE(&d[CHAN_W], 0, CHAN_W); } break; @@ -2073,9 +2185,10 @@ exec_instruction( FETCH(&r[1], 1, chan_index); /* XXX use micro_min()?? */ - micro_lt( &r[0], &r[0], &r[1], &r[0], &r[1] ); - - STORE(&r[0], 0, chan_index); + micro_lt(&d[chan_index], &r[0], &r[1], &r[0], &r[1]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; @@ -2085,9 +2198,10 @@ exec_instruction( FETCH(&r[1], 1, chan_index); /* XXX use micro_max()?? */ - micro_lt( &r[0], &r[0], &r[1], &r[1], &r[0] ); - - STORE(&r[0], 0, chan_index ); + micro_lt(&d[chan_index], &r[0], &r[1], &r[1], &r[0] ); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; @@ -2096,8 +2210,10 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_lt( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, chan_index ); + micro_lt(&d[chan_index], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; @@ -2106,8 +2222,10 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_le( &r[0], &r[1], &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, chan_index ); + micro_le(&d[chan_index], &r[1], &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; @@ -2118,8 +2236,10 @@ exec_instruction( FETCH( &r[1], 1, chan_index ); micro_mul( &r[0], &r[0], &r[1] ); FETCH( &r[1], 2, chan_index ); - micro_add( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); + micro_add(&d[chan_index], &r[0], &r[1]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; @@ -2127,10 +2247,10 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH(&r[0], 0, chan_index); FETCH(&r[1], 1, chan_index); - - micro_sub( &r[0], &r[0], &r[1] ); - - STORE(&r[0], 0, chan_index); + micro_sub(&d[chan_index], &r[0], &r[1]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; @@ -2139,12 +2259,12 @@ exec_instruction( FETCH(&r[0], 0, chan_index); FETCH(&r[1], 1, chan_index); FETCH(&r[2], 2, chan_index); - micro_sub( &r[1], &r[1], &r[2] ); micro_mul( &r[0], &r[0], &r[1] ); - micro_add( &r[0], &r[0], &r[2] ); - - STORE(&r[0], 0, chan_index); + micro_add(&d[chan_index], &r[0], &r[2]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; @@ -2153,8 +2273,10 @@ exec_instruction( FETCH(&r[0], 0, chan_index); FETCH(&r[1], 1, chan_index); FETCH(&r[2], 2, chan_index); - micro_lt(&r[0], &mach->Temps[TEMP_HALF_I].xyzw[TEMP_HALF_C], &r[2], &r[0], &r[1]); - STORE(&r[0], 0, chan_index); + micro_lt(&d[chan_index], &mach->Temps[TEMP_HALF_I].xyzw[TEMP_HALF_C], &r[2], &r[0], &r[1]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; @@ -2179,8 +2301,10 @@ exec_instruction( case TGSI_OPCODE_FRC: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); - micro_frc( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); + micro_frc(&d[chan_index], &r[0]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; @@ -2190,8 +2314,10 @@ exec_instruction( FETCH(&r[1], 1, chan_index); micro_max(&r[0], &r[0], &r[1]); FETCH(&r[1], 2, chan_index); - micro_min(&r[0], &r[0], &r[1]); - STORE(&r[0], 0, chan_index); + micro_min(&d[chan_index], &r[0], &r[1]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; @@ -2199,19 +2325,17 @@ exec_instruction( case TGSI_OPCODE_ARR: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); - micro_rnd( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); + micro_rnd(&d[chan_index], &r[0]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; case TGSI_OPCODE_EX2: FETCH(&r[0], 0, CHAN_X); -#if FAST_MATH micro_exp2( &r[0], &r[0] ); -#else - micro_pow( &r[0], &mach->Temps[TEMP_2_I].xyzw[TEMP_2_C], &r[0] ); -#endif FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { STORE( &r[0], 0, chan_index ); @@ -2247,11 +2371,7 @@ exec_instruction( FETCH(&r[4], 1, CHAN_Y); micro_mul( &r[5], &r[3], &r[4] ); - micro_sub( &r[2], &r[2], &r[5] ); - - if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { - STORE( &r[2], 0, CHAN_X ); - } + micro_sub(&d[CHAN_X], &r[2], &r[5]); FETCH(&r[2], 1, CHAN_X); @@ -2260,20 +2380,21 @@ exec_instruction( FETCH(&r[5], 0, CHAN_X); micro_mul( &r[1], &r[1], &r[5] ); - micro_sub( &r[3], &r[3], &r[1] ); - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { - STORE( &r[3], 0, CHAN_Y ); - } + micro_sub(&d[CHAN_Y], &r[3], &r[1]); micro_mul( &r[5], &r[5], &r[4] ); micro_mul( &r[0], &r[0], &r[2] ); - micro_sub( &r[5], &r[5], &r[0] ); + micro_sub(&d[CHAN_Z], &r[5], &r[0]); - if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - STORE( &r[5], 0, CHAN_Z ); + if (IS_CHANNEL_ENABLED(*inst, CHAN_X)) { + STORE(&d[CHAN_X], 0, CHAN_X); + } + if (IS_CHANNEL_ENABLED(*inst, CHAN_Y)) { + STORE(&d[CHAN_Y], 0, CHAN_Y); + } + if (IS_CHANNEL_ENABLED(*inst, CHAN_Z)) { + STORE(&d[CHAN_Z], 0, CHAN_Z); } - if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); } @@ -2282,11 +2403,11 @@ exec_instruction( case TGSI_OPCODE_ABS: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH(&r[0], 0, chan_index); - - micro_abs( &r[0], &r[0] ); - - STORE(&r[0], 0, chan_index); + micro_abs(&d[chan_index], &r[0]); } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); + } break; case TGSI_OPCODE_RCC: @@ -2338,16 +2459,20 @@ exec_instruction( case TGSI_OPCODE_DDX: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); - micro_ddx( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); + micro_ddx(&d[chan_index], &r[0]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; case TGSI_OPCODE_DDY: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); - micro_ddy( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); + micro_ddy(&d[chan_index], &r[0]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; @@ -2428,10 +2553,10 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_eq( &r[0], &r[0], &r[1], - &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], - &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, chan_index ); + micro_eq(&d[chan_index], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; @@ -2445,8 +2570,10 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_le( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); - STORE( &r[0], 0, chan_index ); + micro_le(&d[chan_index], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; @@ -2462,8 +2589,10 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_le( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, chan_index ); + micro_le(&d[chan_index], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; @@ -2471,8 +2600,10 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_eq( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); - STORE( &r[0], 0, chan_index ); + micro_eq(&d[chan_index], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; @@ -2502,7 +2633,7 @@ exec_instruction( /* src[1] = d[strq]/dx */ /* src[2] = d[strq]/dy */ /* src[3] = sampler unit */ - assert (0); + exec_txd(mach, inst); break; case TGSI_OPCODE_TXL: @@ -2546,13 +2677,8 @@ exec_instruction( micro_mul(&r[3], &r[3], &r[1]); micro_add(&r[2], &r[2], &r[3]); FETCH(&r[3], 0, CHAN_X); - micro_add(&r[2], &r[2], &r[3]); - if (IS_CHANNEL_ENABLED(*inst, CHAN_X)) { - STORE(&r[2], 0, CHAN_X); - } - if (IS_CHANNEL_ENABLED(*inst, CHAN_Z)) { - STORE(&r[2], 0, CHAN_Z); - } + micro_add(&d[CHAN_X], &r[2], &r[3]); + } if (IS_CHANNEL_ENABLED(*inst, CHAN_Y) || IS_CHANNEL_ENABLED(*inst, CHAN_W)) { @@ -2562,13 +2688,20 @@ exec_instruction( micro_mul(&r[3], &r[3], &r[1]); micro_add(&r[2], &r[2], &r[3]); FETCH(&r[3], 0, CHAN_Y); - micro_add(&r[2], &r[2], &r[3]); - if (IS_CHANNEL_ENABLED(*inst, CHAN_Y)) { - STORE(&r[2], 0, CHAN_Y); - } - if (IS_CHANNEL_ENABLED(*inst, CHAN_W)) { - STORE(&r[2], 0, CHAN_W); - } + micro_add(&d[CHAN_Y], &r[2], &r[3]); + + } + if (IS_CHANNEL_ENABLED(*inst, CHAN_X)) { + STORE(&d[CHAN_X], 0, CHAN_X); + } + if (IS_CHANNEL_ENABLED(*inst, CHAN_Y)) { + STORE(&d[CHAN_Y], 0, CHAN_Y); + } + if (IS_CHANNEL_ENABLED(*inst, CHAN_Z)) { + STORE(&d[CHAN_X], 0, CHAN_Z); + } + if (IS_CHANNEL_ENABLED(*inst, CHAN_W)) { + STORE(&d[CHAN_Y], 0, CHAN_W); } break; @@ -2653,8 +2786,10 @@ exec_instruction( /* TGSI_OPCODE_SGN */ FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); - micro_sgn( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); + micro_sgn(&d[chan_index], &r[0]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; @@ -2663,10 +2798,10 @@ exec_instruction( FETCH(&r[0], 0, chan_index); FETCH(&r[1], 1, chan_index); FETCH(&r[2], 2, chan_index); - - micro_lt( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[1], &r[2] ); - - STORE(&r[0], 0, chan_index); + micro_lt(&d[chan_index], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[1], &r[2]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; @@ -2841,32 +2976,40 @@ exec_instruction( case TGSI_OPCODE_CEIL: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); - micro_ceil( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); + micro_ceil(&d[chan_index], &r[0]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; case TGSI_OPCODE_I2F: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); - micro_i2f( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); + micro_i2f(&d[chan_index], &r[0]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; case TGSI_OPCODE_NOT: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); - micro_not( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); + micro_not(&d[chan_index], &r[0]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; case TGSI_OPCODE_TRUNC: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); - micro_trunc( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); + micro_trunc(&d[chan_index], &r[0]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; @@ -2874,8 +3017,10 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_shl( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); + micro_shl(&d[chan_index], &r[0], &r[1]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; @@ -2883,8 +3028,10 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_ishr( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); + micro_ishr(&d[chan_index], &r[0], &r[1]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; @@ -2892,8 +3039,10 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_and( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); + micro_and(&d[chan_index], &r[0], &r[1]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; @@ -2901,8 +3050,10 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_or( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); + micro_or(&d[chan_index], &r[0], &r[1]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; @@ -2914,8 +3065,10 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_xor( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); + micro_xor(&d[chan_index], &r[0], &r[1]); + } + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&d[chan_index], 0, chan_index); } break; @@ -2946,8 +3099,23 @@ exec_instruction( for (chan_index = 0; chan_index < 3; chan_index++) { FETCH( &mach->LoopCounterStack[mach->LoopCounterStackTop].xyzw[chan_index], 0, chan_index ); } - STORE( &mach->LoopCounterStack[mach->LoopCounterStackTop].xyzw[CHAN_Y], 0, CHAN_X ); ++mach->LoopCounterStackTop; + STORE(&mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X], 0, CHAN_X); + /* update LoopMask */ + if (mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y].f[0] <= 0.0f) { + mach->LoopMask &= ~0x1; + } + if (mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y].f[1] <= 0.0f) { + mach->LoopMask &= ~0x2; + } + if (mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y].f[2] <= 0.0f) { + mach->LoopMask &= ~0x4; + } + if (mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y].f[3] <= 0.0f) { + mach->LoopMask &= ~0x8; + } + /* TODO: if mach->LoopMask == 0, jump to end of loop */ + UPDATE_EXEC_MASK(mach); /* fall-through (for now) */ case TGSI_OPCODE_BGNLOOP: /* push LoopMask and ContMasks */ @@ -2961,28 +3129,28 @@ exec_instruction( case TGSI_OPCODE_ENDFOR: assert(mach->LoopCounterStackTop > 0); - micro_sub( &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X], - &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X], - &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + micro_sub(&mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y], + &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y], + &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]); /* update LoopMask */ - if( mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X].f[0] <= 0) { + if (mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y].f[0] <= 0.0f) { mach->LoopMask &= ~0x1; } - if( mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X].f[1] <= 0 ) { + if (mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y].f[1] <= 0.0f) { mach->LoopMask &= ~0x2; } - if( mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X].f[2] <= 0 ) { + if (mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y].f[2] <= 0.0f) { mach->LoopMask &= ~0x4; } - if( mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X].f[3] <= 0 ) { + if (mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y].f[3] <= 0.0f) { mach->LoopMask &= ~0x8; } - micro_add( &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y], - &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y], - &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Z]); + micro_add(&mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X], + &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X], + &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Z]); assert(mach->LoopLabelStackTop > 0); inst = mach->Instructions + mach->LoopLabelStack[mach->LoopLabelStackTop - 1]; - STORE( &mach->LoopCounterStack[mach->LoopCounterStackTop].xyzw[CHAN_Y], 0, CHAN_X ); + STORE(&mach->LoopCounterStack[mach->LoopCounterStackTop].xyzw[CHAN_X], 0, CHAN_X); /* Restore ContMask, but don't pop */ assert(mach->ContStackTop > 0); mach->ContMask = mach->ContStack[mach->ContStackTop - 1]; @@ -3049,7 +3217,28 @@ exec_instruction( break; case TGSI_OPCODE_ENDSUB: - /* no-op */ + /* + * XXX: This really should be a no-op. We should never reach this opcode. + */ + + assert(mach->CallStackTop > 0); + mach->CallStackTop--; + + mach->CondStackTop = mach->CallStack[mach->CallStackTop].CondStackTop; + mach->CondMask = mach->CondStack[mach->CondStackTop]; + + mach->LoopStackTop = mach->CallStack[mach->CallStackTop].LoopStackTop; + mach->LoopMask = mach->LoopStack[mach->LoopStackTop]; + + mach->ContStackTop = mach->CallStack[mach->CallStackTop].ContStackTop; + mach->ContMask = mach->ContStack[mach->ContStackTop]; + + assert(mach->FuncStackTop > 0); + mach->FuncMask = mach->FuncStack[--mach->FuncStackTop]; + + *pc = mach->CallStack[mach->CallStackTop].ReturnAddr; + + UPDATE_EXEC_MASK(mach); break; case TGSI_OPCODE_NOP: @@ -3060,6 +3249,8 @@ exec_instruction( } } +#define DEBUG_EXECUTION 0 + /** * Run TGSI interpreter. @@ -3103,10 +3294,67 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) exec_declaration( mach, mach->Declarations+i ); } - /* execute instructions, until pc is set to -1 */ - while (pc != -1) { - assert(pc < (int) mach->NumInstructions); - exec_instruction( mach, mach->Instructions + pc, &pc ); + { +#if DEBUG_EXECUTION + struct tgsi_exec_vector temps[TGSI_EXEC_NUM_TEMPS + TGSI_EXEC_NUM_TEMP_EXTRAS]; + struct tgsi_exec_vector outputs[PIPE_MAX_ATTRIBS]; + uint inst = 1; + + memcpy(temps, mach->Temps, sizeof(temps)); + memcpy(outputs, mach->Outputs, sizeof(outputs)); +#endif + + /* execute instructions, until pc is set to -1 */ + while (pc != -1) { + +#if DEBUG_EXECUTION + uint i; + + tgsi_dump_instruction(&mach->Instructions[pc], inst++); +#endif + + assert(pc < (int) mach->NumInstructions); + exec_instruction(mach, mach->Instructions + pc, &pc); + +#if DEBUG_EXECUTION + for (i = 0; i < TGSI_EXEC_NUM_TEMPS + TGSI_EXEC_NUM_TEMP_EXTRAS; i++) { + if (memcmp(&temps[i], &mach->Temps[i], sizeof(temps[i]))) { + uint j; + + memcpy(&temps[i], &mach->Temps[i], sizeof(temps[i])); + debug_printf("TEMP[%2u] = ", i); + for (j = 0; j < 4; j++) { + if (j > 0) { + debug_printf(" "); + } + debug_printf("(%6f, %6f, %6f, %6f)\n", + temps[i].xyzw[0].f[j], + temps[i].xyzw[1].f[j], + temps[i].xyzw[2].f[j], + temps[i].xyzw[3].f[j]); + } + } + } + for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { + if (memcmp(&outputs[i], &mach->Outputs[i], sizeof(outputs[i]))) { + uint j; + + memcpy(&outputs[i], &mach->Outputs[i], sizeof(outputs[i])); + debug_printf("OUT[%2u] = ", i); + for (j = 0; j < 4; j++) { + if (j > 0) { + debug_printf(" "); + } + debug_printf("{%6f, %6f, %6f, %6f}\n", + outputs[i].xyzw[0].f[j], + outputs[i].xyzw[1].f[j], + outputs[i].xyzw[2].f[j], + outputs[i].xyzw[3].f[j]); + } + } + } +#endif + } } #if 0 @@ -3120,5 +3368,10 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) } #endif + assert(mach->CondStackTop == 0); + assert(mach->LoopStackTop == 0); + assert(mach->ContStackTop == 0); + assert(mach->CallStackTop == 0); + return ~mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0]; } diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.h b/src/gallium/auxiliary/tgsi/tgsi_exec.h index 471f591dd6e..fd94c1bc440 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.h +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.h @@ -218,6 +218,7 @@ struct tgsi_exec_machine struct tgsi_exec_vector Outputs[PIPE_MAX_ATTRIBS]; struct tgsi_exec_vector *Addrs; + struct tgsi_exec_vector *Predicates; struct tgsi_sampler **Samplers; @@ -232,6 +233,7 @@ struct tgsi_exec_machine /* FRAGMENT processor only. */ const struct tgsi_interp_coef *InterpCoefs; struct tgsi_exec_vector QuadPos; + float Face; /**< +1 if front facing, -1 if back facing */ /* Conditional execution masks */ uint CondMask; /**< For IF/ELSE/ENDIF */ @@ -252,7 +254,7 @@ struct tgsi_exec_machine uint LoopLabelStack[TGSI_EXEC_MAX_LOOP_NESTING]; int LoopLabelStackTop; - /** Loop counter stack (x = count, y = current, z = step) */ + /** Loop counter stack (x = index, y = counter, z = step) */ struct tgsi_exec_vector LoopCounterStack[TGSI_EXEC_MAX_LOOP_NESTING]; int LoopCounterStackTop; diff --git a/src/gallium/auxiliary/tgsi/tgsi_parse.c b/src/gallium/auxiliary/tgsi/tgsi_parse.c index 83f9df1183e..9ca29934528 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_parse.c +++ b/src/gallium/auxiliary/tgsi/tgsi_parse.c @@ -175,6 +175,10 @@ tgsi_parse_token( copy_token(&inst->Instruction, &token); extended = inst->Instruction.Extended; + if (inst->Instruction.Predicate) { + next_token(ctx, &inst->InstructionPredicate); + } + while( extended ) { struct tgsi_src_register_ext token; @@ -189,10 +193,6 @@ tgsi_parse_token( copy_token(&inst->InstructionExtTexture, &token); break; - case TGSI_INSTRUCTION_EXT_TYPE_PREDICATE: - copy_token(&inst->InstructionExtPredicate, &token); - break; - default: assert( 0 ); } diff --git a/src/gallium/auxiliary/tgsi/tgsi_parse.h b/src/gallium/auxiliary/tgsi/tgsi_parse.h index 76f1676d85d..cb4772ade8d 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_parse.h +++ b/src/gallium/auxiliary/tgsi/tgsi_parse.h @@ -80,9 +80,9 @@ struct tgsi_full_immediate struct tgsi_full_instruction { struct tgsi_instruction Instruction; + struct tgsi_instruction_predicate InstructionPredicate; struct tgsi_instruction_ext_label InstructionExtLabel; struct tgsi_instruction_ext_texture InstructionExtTexture; - struct tgsi_instruction_ext_predicate InstructionExtPredicate; struct tgsi_full_dst_register FullDstRegisters[TGSI_FULL_MAX_DST_REGISTERS]; struct tgsi_full_src_register FullSrcRegisters[TGSI_FULL_MAX_SRC_REGISTERS]; uint Flags; /**< user-defined usage */ diff --git a/src/gallium/auxiliary/tgsi/tgsi_sanity.c b/src/gallium/auxiliary/tgsi/tgsi_sanity.c index 36e27ea52f4..4d8145032b2 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_sanity.c +++ b/src/gallium/auxiliary/tgsi/tgsi_sanity.c @@ -34,7 +34,7 @@ typedef uint reg_flag; #define BITS_IN_REG_FLAG (sizeof( reg_flag ) * 8) -#define MAX_REGISTERS 256 +#define MAX_REGISTERS 1024 #define MAX_REG_FLAGS ((MAX_REGISTERS + BITS_IN_REG_FLAG - 1) / BITS_IN_REG_FLAG) struct sanity_check_ctx diff --git a/src/gallium/auxiliary/tgsi/tgsi_ureg.c b/src/gallium/auxiliary/tgsi/tgsi_ureg.c index 3f752e9352f..bd963267cc3 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_ureg.c +++ b/src/gallium/auxiliary/tgsi/tgsi_ureg.c @@ -47,9 +47,9 @@ union tgsi_any_token { struct tgsi_immediate imm; union tgsi_immediate_data imm_data; struct tgsi_instruction insn; + struct tgsi_instruction_predicate insn_predicate; struct tgsi_instruction_ext_label insn_ext_label; struct tgsi_instruction_ext_texture insn_ext_texture; - struct tgsi_instruction_ext_predicate insn_ext_predicate; struct tgsi_src_register src; struct tgsi_src_register_ext_mod src_ext_mod; struct tgsi_dimension dim; @@ -72,6 +72,7 @@ struct ureg_tokens { #define UREG_MAX_IMMEDIATE 32 #define UREG_MAX_TEMP 256 #define UREG_MAX_ADDR 2 +#define UREG_MAX_LOOP 1 #define UREG_MAX_PRED 1 #define DOMAIN_DECL 0 @@ -117,6 +118,7 @@ struct ureg_program unsigned nr_addrs; unsigned nr_preds; + unsigned nr_loops; unsigned nr_instructions; struct ureg_tokens domain[2]; @@ -350,6 +352,7 @@ struct ureg_src ureg_DECL_constant(struct ureg_program *ureg, i = ureg->nr_constant_ranges++; ureg->constant_range[i].first = index; ureg->constant_range[i].last = index; + goto out; } /* Collapse all ranges down to one: @@ -417,6 +420,19 @@ struct ureg_dst ureg_DECL_address( struct ureg_program *ureg ) return ureg_dst_register( TGSI_FILE_ADDRESS, 0 ); } +/* Allocate a new loop register. + */ +struct ureg_dst +ureg_DECL_loop(struct ureg_program *ureg) +{ + if (ureg->nr_loops < UREG_MAX_LOOP) { + return ureg_dst_register(TGSI_FILE_LOOP, ureg->nr_loops++); + } + + assert(0); + return ureg_dst_register(TGSI_FILE_LOOP, 0); +} + /* Allocate a new predicate register. */ struct ureg_dst @@ -661,35 +677,27 @@ ureg_emit_insn(struct ureg_program *ureg, validate( opcode, num_dst, num_src ); out = get_tokens( ureg, DOMAIN_INSN, count ); - out[0].value = 0; - out[0].insn.Type = TGSI_TOKEN_TYPE_INSTRUCTION; - out[0].insn.NrTokens = 0; + out[0].insn = tgsi_default_instruction(); out[0].insn.Opcode = opcode; out[0].insn.Saturate = saturate; out[0].insn.NumDstRegs = num_dst; out[0].insn.NumSrcRegs = num_src; - out[0].insn.Padding = 0; result.insn_token = ureg->domain[DOMAIN_INSN].count - count; + result.extended_token = result.insn_token; if (predicate) { - out[0].insn.Extended = 1; - out[1].insn_ext_predicate = tgsi_default_instruction_ext_predicate(); - out[1].insn_ext_predicate.Negate = pred_negate; - out[1].insn_ext_predicate.SwizzleX = pred_swizzle_x; - out[1].insn_ext_predicate.SwizzleY = pred_swizzle_y; - out[1].insn_ext_predicate.SwizzleZ = pred_swizzle_z; - out[1].insn_ext_predicate.SwizzleW = pred_swizzle_w; - - result.extended_token = result.insn_token + 1; - } else { - out[0].insn.Extended = 0; - - result.extended_token = result.insn_token; + out[0].insn.Predicate = 1; + out[1].insn_predicate = tgsi_default_instruction_predicate(); + out[1].insn_predicate.Negate = pred_negate; + out[1].insn_predicate.SwizzleX = pred_swizzle_x; + out[1].insn_predicate.SwizzleY = pred_swizzle_y; + out[1].insn_predicate.SwizzleZ = pred_swizzle_z; + out[1].insn_predicate.SwizzleW = pred_swizzle_w; } ureg->nr_instructions++; - + return result; } @@ -1023,6 +1031,13 @@ static void emit_decls( struct ureg_program *ureg ) 0, ureg->nr_addrs ); } + if (ureg->nr_loops) { + emit_decl_range(ureg, + TGSI_FILE_LOOP, + 0, + ureg->nr_loops); + } + if (ureg->nr_preds) { emit_decl_range(ureg, TGSI_FILE_PREDICATE, diff --git a/src/gallium/auxiliary/tgsi/tgsi_ureg.h b/src/gallium/auxiliary/tgsi/tgsi_ureg.h index dae42911947..94cc70a2082 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_ureg.h +++ b/src/gallium/auxiliary/tgsi/tgsi_ureg.h @@ -158,6 +158,9 @@ struct ureg_dst ureg_DECL_address( struct ureg_program * ); struct ureg_dst +ureg_DECL_loop( struct ureg_program * ); + +struct ureg_dst ureg_DECL_predicate(struct ureg_program *); /* Supply an index to the sampler declaration as this is the hook to diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c index 50386425995..5372df57352 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -354,9 +354,9 @@ util_blit_pixels_writemask(struct blit_state *ctx, texTemp.target = PIPE_TEXTURE_2D; texTemp.format = src->format; texTemp.last_level = 0; - texTemp.width[0] = srcW; - texTemp.height[0] = srcH; - texTemp.depth[0] = 1; + texTemp.width0 = srcW; + texTemp.height0 = srcH; + texTemp.depth0 = 1; pf_get_block(src->format, &texTemp.block); tex = screen->texture_create(screen, &texTemp); @@ -389,10 +389,10 @@ util_blit_pixels_writemask(struct blit_state *ctx, } else { pipe_texture_reference(&tex, src->texture); - s0 = srcX0 / (float)tex->width[0]; - s1 = srcX1 / (float)tex->width[0]; - t0 = srcY0 / (float)tex->height[0]; - t1 = srcY1 / (float)tex->height[0]; + s0 = srcX0 / (float)tex->width0; + s1 = srcX1 / (float)tex->width0; + t0 = srcY0 / (float)tex->height0; + t1 = srcY1 / (float)tex->height0; } @@ -518,13 +518,13 @@ util_blit_pixels_tex(struct blit_state *ctx, assert(filter == PIPE_TEX_MIPFILTER_NEAREST || filter == PIPE_TEX_MIPFILTER_LINEAR); - assert(tex->width[0] != 0); - assert(tex->height[0] != 0); + assert(tex->width0 != 0); + assert(tex->height0 != 0); - s0 = srcX0 / (float)tex->width[0]; - s1 = srcX1 / (float)tex->width[0]; - t0 = srcY0 / (float)tex->height[0]; - t1 = srcY1 / (float)tex->height[0]; + s0 = srcX0 / (float)tex->width0; + s1 = srcX1 / (float)tex->width0; + t0 = srcY0 / (float)tex->height0; + t1 = srcY1 / (float)tex->height0; assert(ctx->pipe->screen->is_format_supported(ctx->pipe->screen, dst->format, PIPE_TEXTURE_2D, diff --git a/src/gallium/auxiliary/util/u_format.csv b/src/gallium/auxiliary/util/u_format.csv index f1bf94f17dd..b9cc2aa716e 100644 --- a/src/gallium/auxiliary/util/u_format.csv +++ b/src/gallium/auxiliary/util/u_format.csv @@ -97,3 +97,13 @@ PIPE_FORMAT_B8G8R8A8_SRGB , arith , 1, 1, u8 , u8 , u8 , u8 , zyxw, PIPE_FORMAT_B8G8R8X8_SRGB , arith , 1, 1, u8 , u8 , u8 , u8 , zyx1, srgb PIPE_FORMAT_X8UB8UG8SR8S_NORM , arith , 1, 1, sn8 , sn8 , un8 , x8 , 1zyx, rgb PIPE_FORMAT_B6UG5SR5S_NORM , arith , 1, 1, sn5 , sn5 , un6 , , xyz1, rgb +PIPE_FORMAT_YCBCR , yuv , 2, 1, x32 , , , , xyz1, yuv +PIPE_FORMAT_YCBCR_REV , yuv , 2, 1, x32 , , , , xyz1, yuv +PIPE_FORMAT_DXT1_RGBA , dxt , 4, 4, x64 , , , , xyzw, rgb +PIPE_FORMAT_DXT1_RGB , dxt , 4, 4, x64 , , , , xyz1, rgb +PIPE_FORMAT_DXT3_RGBA , dxt , 4, 4, x128, , , , xyzw, rgb +PIPE_FORMAT_DXT5_RGBA , dxt , 4, 4, x128, , , , xyzw, rgb +PIPE_FORMAT_DXT1_SRGBA , dxt , 4, 4, x64 , , , , xyzw, srgb +PIPE_FORMAT_DXT1_SRGB , dxt , 4, 4, x64 , , , , xyz1, srgb +PIPE_FORMAT_DXT3_SRGBA , dxt , 4, 4, x128, , , , xyzw, srgb +PIPE_FORMAT_DXT5_SRGBA , dxt , 4, 4, x128, , , , xyzw, srgb diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index aa823aa218b..f67f1e458d4 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -45,6 +45,7 @@ #include "util/u_draw_quad.h" #include "util/u_gen_mipmap.h" #include "util/u_simple_shaders.h" +#include "util/u_math.h" #include "cso_cache/cso_context.h" @@ -1125,12 +1126,12 @@ make_1d_mipmap(struct gen_mipmap_state *ctx, srcTrans = screen->get_tex_transfer(screen, pt, face, srcLevel, zslice, PIPE_TRANSFER_READ, 0, 0, - pt->width[srcLevel], - pt->height[srcLevel]); + u_minify(pt->width0, srcLevel), + u_minify(pt->height0, srcLevel)); dstTrans = screen->get_tex_transfer(screen, pt, face, dstLevel, zslice, PIPE_TRANSFER_WRITE, 0, 0, - pt->width[dstLevel], - pt->height[dstLevel]); + u_minify(pt->width0, dstLevel), + u_minify(pt->height0, dstLevel)); srcMap = (ubyte *) screen->transfer_map(screen, srcTrans); dstMap = (ubyte *) screen->transfer_map(screen, dstTrans); @@ -1168,12 +1169,12 @@ make_2d_mipmap(struct gen_mipmap_state *ctx, srcTrans = screen->get_tex_transfer(screen, pt, face, srcLevel, zslice, PIPE_TRANSFER_READ, 0, 0, - pt->width[srcLevel], - pt->height[srcLevel]); + u_minify(pt->width0, srcLevel), + u_minify(pt->height0, srcLevel)); dstTrans = screen->get_tex_transfer(screen, pt, face, dstLevel, zslice, PIPE_TRANSFER_WRITE, 0, 0, - pt->width[dstLevel], - pt->height[dstLevel]); + u_minify(pt->width0, dstLevel), + u_minify(pt->height0, dstLevel)); srcMap = (ubyte *) screen->transfer_map(screen, srcTrans); dstMap = (ubyte *) screen->transfer_map(screen, dstTrans); @@ -1213,12 +1214,12 @@ make_3d_mipmap(struct gen_mipmap_state *ctx, srcTrans = screen->get_tex_transfer(screen, pt, face, srcLevel, zslice, PIPE_TRANSFER_READ, 0, 0, - pt->width[srcLevel], - pt->height[srcLevel]); + u_minify(pt->width0, srcLevel), + u_minify(pt->height0, srcLevel)); dstTrans = screen->get_tex_transfer(screen, pt, face, dstLevel, zslice, PIPE_TRANSFER_WRITE, 0, 0, - pt->width[dstLevel], - pt->height[dstLevel]); + u_minify(pt->width0, dstLevel), + u_minify(pt->height0, dstLevel)); srcMap = (ubyte *) screen->transfer_map(screen, srcTrans); dstMap = (ubyte *) screen->transfer_map(screen, dstTrans); @@ -1575,8 +1576,8 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, * Setup framebuffer / dest surface */ fb.cbufs[0] = surf; - fb.width = pt->width[dstLevel]; - fb.height = pt->height[dstLevel]; + fb.width = u_minify(pt->width0, dstLevel); + fb.height = u_minify(pt->height0, dstLevel); cso_set_framebuffer(ctx->cso, &fb); /* @@ -1597,8 +1598,8 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, offset = set_vertex_data(ctx, pt->target, face, - (float) pt->width[dstLevel], - (float) pt->height[dstLevel]); + (float) u_minify(pt->width0, dstLevel), + (float) u_minify(pt->height0, dstLevel)); util_draw_vertex_buffer(ctx->pipe, ctx->vbuf, diff --git a/src/gallium/auxiliary/util/u_math.h b/src/gallium/auxiliary/util/u_math.h index 75b075f160d..a5cd6574c0e 100644 --- a/src/gallium/auxiliary/util/u_math.h +++ b/src/gallium/auxiliary/util/u_math.h @@ -491,6 +491,47 @@ util_next_power_of_two(unsigned x) /** + * Return number of bits set in n. + */ +static INLINE unsigned +util_bitcount(unsigned n) +{ +#if defined(PIPE_CC_GCC) + return __builtin_popcount(n); +#else + /* K&R classic bitcount. + * + * For each iteration, clear the LSB from the bitfield. + * Requires only one iteration per set bit, instead of + * one iteration per bit less than highest set bit. + */ + unsigned bits = 0; + for (bits; n; bits++) { + n &= n - 1; + } + return bits; +#endif +} + + +/** + * Reverse byte order of a 32 bit word. + */ +static INLINE uint32_t +util_bswap32(uint32_t n) +{ +#if defined(PIPE_CC_GCC) + return __builtin_bswap32(n); +#else + return (n >> 24) | + ((n >> 8) & 0x0000ff00) | + ((n << 8) & 0x00ff0000) | + (n << 24); +#endif +} + + +/** * Clamp X to [MIN, MAX]. * This is a macro to allow float, int, uint, etc. types. */ @@ -499,6 +540,9 @@ util_next_power_of_two(unsigned x) #define MIN2( A, B ) ( (A)<(B) ? (A) : (B) ) #define MAX2( A, B ) ( (A)>(B) ? (A) : (B) ) +#define MIN3( A, B, C ) MIN2( MIN2( A, B ), C ) +#define MAX3( A, B, C ) MAX2( MAX2( A, B ), C ) + static INLINE int align(int value, int alignment) @@ -507,9 +551,9 @@ align(int value, int alignment) } static INLINE unsigned -minify(unsigned value) +u_minify(unsigned value, unsigned levels) { - return MAX2(1, value >> 1); + return MAX2(1, value >> levels); } #ifndef COPY_4V diff --git a/src/gallium/auxiliary/util/u_surface.c b/src/gallium/auxiliary/util/u_surface.c index 85e443204e3..de8c266db83 100644 --- a/src/gallium/auxiliary/util/u_surface.c +++ b/src/gallium/auxiliary/util/u_surface.c @@ -79,9 +79,9 @@ util_create_rgba_surface(struct pipe_screen *screen, templ.target = target; templ.format = format; templ.last_level = 0; - templ.width[0] = width; - templ.height[0] = height; - templ.depth[0] = 1; + templ.width0 = width; + templ.height0 = height; + templ.depth0 = 1; pf_get_block(format, &templ.block); templ.tex_usage = usage; diff --git a/src/gallium/auxiliary/vl/vl_compositor.c b/src/gallium/auxiliary/vl/vl_compositor.c index cda6dc134a0..8f52170f90a 100644 --- a/src/gallium/auxiliary/vl/vl_compositor.c +++ b/src/gallium/auxiliary/vl/vl_compositor.c @@ -455,8 +455,8 @@ void vl_compositor_render(struct vl_compositor *compositor, assert(dst_area); assert(picture_type == PIPE_MPEG12_PICTURE_TYPE_FRAME); - compositor->fb_state.width = dst_surface->width[0]; - compositor->fb_state.height = dst_surface->height[0]; + compositor->fb_state.width = dst_surface->width0; + compositor->fb_state.height = dst_surface->height0; compositor->fb_state.cbufs[0] = compositor->pipe->screen->get_tex_surface ( compositor->pipe->screen, @@ -479,8 +479,8 @@ void vl_compositor_render(struct vl_compositor *compositor, compositor->pipe->set_framebuffer_state(compositor->pipe, &compositor->fb_state); compositor->pipe->set_viewport_state(compositor->pipe, &compositor->viewport); compositor->pipe->set_scissor_state(compositor->pipe, &compositor->scissor); - compositor->pipe->bind_sampler_states(compositor->pipe, 1, &compositor->sampler); - compositor->pipe->set_sampler_textures(compositor->pipe, 1, &src_surface); + compositor->pipe->bind_fragment_sampler_states(compositor->pipe, 1, &compositor->sampler); + compositor->pipe->set_fragment_sampler_textures(compositor->pipe, 1, &src_surface); compositor->pipe->bind_vs_state(compositor->pipe, compositor->vertex_shader); compositor->pipe->bind_fs_state(compositor->pipe, compositor->fragment_shader); compositor->pipe->set_vertex_buffers(compositor->pipe, 2, compositor->vertex_bufs); @@ -504,12 +504,12 @@ void vl_compositor_render(struct vl_compositor *compositor, vs_consts->dst_trans.z = 0; vs_consts->dst_trans.w = 0; - vs_consts->src_scale.x = src_area->w / (float)src_surface->width[0]; - vs_consts->src_scale.y = src_area->h / (float)src_surface->height[0]; + vs_consts->src_scale.x = src_area->w / (float)src_surface->width0; + vs_consts->src_scale.y = src_area->h / (float)src_surface->height0; vs_consts->src_scale.z = 1; vs_consts->src_scale.w = 1; - vs_consts->src_trans.x = src_area->x / (float)src_surface->width[0]; - vs_consts->src_trans.y = src_area->y / (float)src_surface->height[0]; + vs_consts->src_trans.x = src_area->x / (float)src_surface->width0; + vs_consts->src_trans.y = src_area->y / (float)src_surface->height0; vs_consts->src_trans.z = 0; vs_consts->src_trans.w = 0; diff --git a/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c b/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c index c4ba69817f9..93da67b984c 100644 --- a/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c +++ b/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c @@ -681,7 +681,7 @@ xfer_buffers_map(struct vl_mpeg12_mc_renderer *r) ( r->pipe->screen, r->textures.all[i], 0, 0, 0, PIPE_TRANSFER_WRITE, 0, 0, - r->textures.all[i]->width[0], r->textures.all[i]->height[0] + r->textures.all[i]->width0, r->textures.all[i]->height0 ); r->texels[i] = r->pipe->screen->transfer_map(r->pipe->screen, r->tex_transfer[i]); @@ -835,26 +835,26 @@ init_buffers(struct vl_mpeg12_mc_renderer *r) /* TODO: Accomodate HW that can't do this and also for cases when this isn't precise enough */ template.format = PIPE_FORMAT_R16_SNORM; template.last_level = 0; - template.width[0] = r->pot_buffers ? + template.width0 = r->pot_buffers ? util_next_power_of_two(r->picture_width) : r->picture_width; - template.height[0] = r->pot_buffers ? + template.height0 = r->pot_buffers ? util_next_power_of_two(r->picture_height) : r->picture_height; - template.depth[0] = 1; + template.depth0 = 1; pf_get_block(template.format, &template.block); template.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER | PIPE_TEXTURE_USAGE_DYNAMIC; r->textures.individual.y = r->pipe->screen->texture_create(r->pipe->screen, &template); if (r->chroma_format == PIPE_VIDEO_CHROMA_FORMAT_420) { - template.width[0] = r->pot_buffers ? + template.width0 = r->pot_buffers ? util_next_power_of_two(r->picture_width / 2) : r->picture_width / 2; - template.height[0] = r->pot_buffers ? + template.height0 = r->pot_buffers ? util_next_power_of_two(r->picture_height / 2) : r->picture_height / 2; } else if (r->chroma_format == PIPE_VIDEO_CHROMA_FORMAT_422) - template.height[0] = r->pot_buffers ? + template.height0 = r->pot_buffers ? util_next_power_of_two(r->picture_height / 2) : r->picture_height / 2; @@ -1283,8 +1283,8 @@ flush(struct vl_mpeg12_mc_renderer *r) PIPE_BUFFER_USAGE_CPU_WRITE | PIPE_BUFFER_USAGE_DISCARD ); - vs_consts->denorm.x = r->surface->width[0]; - vs_consts->denorm.y = r->surface->height[0]; + vs_consts->denorm.x = r->surface->width0; + vs_consts->denorm.y = r->surface->height0; pipe_buffer_unmap(r->pipe->screen, r->vs_const_buf.buffer); @@ -1296,8 +1296,8 @@ flush(struct vl_mpeg12_mc_renderer *r) if (num_macroblocks[MACROBLOCK_TYPE_INTRA] > 0) { r->pipe->set_vertex_buffers(r->pipe, 1, r->vertex_bufs.all); r->pipe->set_vertex_elements(r->pipe, 4, r->vertex_elems); - r->pipe->set_sampler_textures(r->pipe, 3, r->textures.all); - r->pipe->bind_sampler_states(r->pipe, 3, r->samplers.all); + r->pipe->set_fragment_sampler_textures(r->pipe, 3, r->textures.all); + r->pipe->bind_fragment_sampler_states(r->pipe, 3, r->samplers.all); r->pipe->bind_vs_state(r->pipe, r->i_vs); r->pipe->bind_fs_state(r->pipe, r->i_fs); @@ -1310,8 +1310,8 @@ flush(struct vl_mpeg12_mc_renderer *r) r->pipe->set_vertex_buffers(r->pipe, 2, r->vertex_bufs.all); r->pipe->set_vertex_elements(r->pipe, 6, r->vertex_elems); r->textures.individual.ref[0] = r->past; - r->pipe->set_sampler_textures(r->pipe, 4, r->textures.all); - r->pipe->bind_sampler_states(r->pipe, 4, r->samplers.all); + r->pipe->set_fragment_sampler_textures(r->pipe, 4, r->textures.all); + r->pipe->bind_fragment_sampler_states(r->pipe, 4, r->samplers.all); r->pipe->bind_vs_state(r->pipe, r->p_vs[0]); r->pipe->bind_fs_state(r->pipe, r->p_fs[0]); @@ -1324,8 +1324,8 @@ flush(struct vl_mpeg12_mc_renderer *r) r->pipe->set_vertex_buffers(r->pipe, 2, r->vertex_bufs.all); r->pipe->set_vertex_elements(r->pipe, 6, r->vertex_elems); r->textures.individual.ref[0] = r->past; - r->pipe->set_sampler_textures(r->pipe, 4, r->textures.all); - r->pipe->bind_sampler_states(r->pipe, 4, r->samplers.all); + r->pipe->set_fragment_sampler_textures(r->pipe, 4, r->textures.all); + r->pipe->bind_fragment_sampler_states(r->pipe, 4, r->samplers.all); r->pipe->bind_vs_state(r->pipe, r->p_vs[1]); r->pipe->bind_fs_state(r->pipe, r->p_fs[1]); @@ -1338,8 +1338,8 @@ flush(struct vl_mpeg12_mc_renderer *r) r->pipe->set_vertex_buffers(r->pipe, 2, r->vertex_bufs.all); r->pipe->set_vertex_elements(r->pipe, 6, r->vertex_elems); r->textures.individual.ref[0] = r->future; - r->pipe->set_sampler_textures(r->pipe, 4, r->textures.all); - r->pipe->bind_sampler_states(r->pipe, 4, r->samplers.all); + r->pipe->set_fragment_sampler_textures(r->pipe, 4, r->textures.all); + r->pipe->bind_fragment_sampler_states(r->pipe, 4, r->samplers.all); r->pipe->bind_vs_state(r->pipe, r->p_vs[0]); r->pipe->bind_fs_state(r->pipe, r->p_fs[0]); @@ -1352,8 +1352,8 @@ flush(struct vl_mpeg12_mc_renderer *r) r->pipe->set_vertex_buffers(r->pipe, 2, r->vertex_bufs.all); r->pipe->set_vertex_elements(r->pipe, 6, r->vertex_elems); r->textures.individual.ref[0] = r->future; - r->pipe->set_sampler_textures(r->pipe, 4, r->textures.all); - r->pipe->bind_sampler_states(r->pipe, 4, r->samplers.all); + r->pipe->set_fragment_sampler_textures(r->pipe, 4, r->textures.all); + r->pipe->bind_fragment_sampler_states(r->pipe, 4, r->samplers.all); r->pipe->bind_vs_state(r->pipe, r->p_vs[1]); r->pipe->bind_fs_state(r->pipe, r->p_fs[1]); @@ -1367,8 +1367,8 @@ flush(struct vl_mpeg12_mc_renderer *r) r->pipe->set_vertex_elements(r->pipe, 8, r->vertex_elems); r->textures.individual.ref[0] = r->past; r->textures.individual.ref[1] = r->future; - r->pipe->set_sampler_textures(r->pipe, 5, r->textures.all); - r->pipe->bind_sampler_states(r->pipe, 5, r->samplers.all); + r->pipe->set_fragment_sampler_textures(r->pipe, 5, r->textures.all); + r->pipe->bind_fragment_sampler_states(r->pipe, 5, r->samplers.all); r->pipe->bind_vs_state(r->pipe, r->b_vs[0]); r->pipe->bind_fs_state(r->pipe, r->b_fs[0]); @@ -1382,8 +1382,8 @@ flush(struct vl_mpeg12_mc_renderer *r) r->pipe->set_vertex_elements(r->pipe, 8, r->vertex_elems); r->textures.individual.ref[0] = r->past; r->textures.individual.ref[1] = r->future; - r->pipe->set_sampler_textures(r->pipe, 5, r->textures.all); - r->pipe->bind_sampler_states(r->pipe, 5, r->samplers.all); + r->pipe->set_fragment_sampler_textures(r->pipe, 5, r->textures.all); + r->pipe->bind_fragment_sampler_states(r->pipe, 5, r->samplers.all); r->pipe->bind_vs_state(r->pipe, r->b_vs[1]); r->pipe->bind_fs_state(r->pipe, r->b_fs[1]); @@ -1633,8 +1633,8 @@ vl_mpeg12_mc_renderer_render_macroblocks(struct vl_mpeg12_mc_renderer renderer->past = past; renderer->future = future; renderer->fence = fence; - renderer->surface_tex_inv_size.x = 1.0f / surface->width[0]; - renderer->surface_tex_inv_size.y = 1.0f / surface->height[0]; + renderer->surface_tex_inv_size.x = 1.0f / surface->width0; + renderer->surface_tex_inv_size.y = 1.0f / surface->height0; } while (num_macroblocks) { diff --git a/src/gallium/drivers/cell/ppu/cell_pipe_state.c b/src/gallium/drivers/cell/ppu/cell_pipe_state.c index ccd0fef6e84..c18a5d0635e 100644 --- a/src/gallium/drivers/cell/ppu/cell_pipe_state.c +++ b/src/gallium/drivers/cell/ppu/cell_pipe_state.c @@ -383,10 +383,10 @@ cell_init_state_functions(struct cell_context *cell) cell->pipe.delete_blend_state = cell_delete_blend_state; cell->pipe.create_sampler_state = cell_create_sampler_state; - cell->pipe.bind_sampler_states = cell_bind_sampler_states; + cell->pipe.bind_fragment_sampler_states = cell_bind_sampler_states; cell->pipe.delete_sampler_state = cell_delete_sampler_state; - cell->pipe.set_sampler_textures = cell_set_sampler_textures; + cell->pipe.set_fragment_sampler_textures = cell_set_sampler_textures; cell->pipe.create_depth_stencil_alpha_state = cell_create_depth_stencil_alpha_state; cell->pipe.bind_depth_stencil_alpha_state = cell_bind_depth_stencil_alpha_state; diff --git a/src/gallium/drivers/cell/ppu/cell_state_emit.c b/src/gallium/drivers/cell/ppu/cell_state_emit.c index 9479c0898fd..ac5fafec1ad 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_emit.c +++ b/src/gallium/drivers/cell/ppu/cell_state_emit.c @@ -27,6 +27,7 @@ #include "pipe/p_inlines.h" #include "util/u_memory.h" +#include "util/u_math.h" #include "cell_context.h" #include "cell_gen_fragment.h" #include "cell_state.h" @@ -299,9 +300,9 @@ cell_emit_state(struct cell_context *cell) for (level = 0; level < CELL_MAX_TEXTURE_LEVELS; level++) { texture->start[level] = (ct->mapped + ct->level_offset[level]); - texture->width[level] = ct->base.width[level]; - texture->height[level] = ct->base.height[level]; - texture->depth[level] = ct->base.depth[level]; + texture->width[level] = u_minify(ct->base.width0, level); + texture->height[level] = u_minify(ct->base.height0, level); + texture->depth[level] = u_minify(ct->base.depth0, level); } texture->target = ct->base.target; } diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c index ae4c61efb3b..e6b8a870452 100644 --- a/src/gallium/drivers/cell/ppu/cell_texture.c +++ b/src/gallium/drivers/cell/ppu/cell_texture.c @@ -49,9 +49,9 @@ cell_texture_layout(struct cell_texture *ct) { struct pipe_texture *pt = &ct->base; unsigned level; - unsigned width = pt->width[0]; - unsigned height = pt->height[0]; - unsigned depth = pt->depth[0]; + unsigned width = pt->width0; + unsigned height = pt->height0; + unsigned depth = pt->depth0; ct->buffer_size = 0; @@ -65,9 +65,6 @@ cell_texture_layout(struct cell_texture *ct) w_tile = align(width, TILE_SIZE); h_tile = align(height, TILE_SIZE); - pt->width[level] = width; - pt->height[level] = height; - pt->depth[level] = depth; pt->nblocksx[level] = pf_get_nblocksx(&pt->block, w_tile); pt->nblocksy[level] = pf_get_nblocksy(&pt->block, h_tile); @@ -83,9 +80,9 @@ cell_texture_layout(struct cell_texture *ct) ct->buffer_size += size; - width = minify(width); - height = minify(height); - depth = minify(depth); + width = u_minify(width, 1); + height = u_minify(height, 1); + depth = u_minify(depth, 1); } } @@ -276,8 +273,8 @@ cell_get_tex_surface(struct pipe_screen *screen, pipe_reference_init(&ps->reference, 1); pipe_texture_reference(&ps->texture, pt); ps->format = pt->format; - ps->width = pt->width[level]; - ps->height = pt->height[level]; + ps->width = u_minify(pt->width0, level); + ps->height = u_minify(pt->height0, level); ps->offset = ct->level_offset[level]; /* XXX may need to override usage flags (see sp_texture.c) */ ps->usage = usage; @@ -386,8 +383,8 @@ cell_transfer_map(struct pipe_screen *screen, struct pipe_transfer *transfer) struct pipe_texture *pt = transfer->texture; struct cell_texture *ct = cell_texture(pt); const uint level = ctrans->base.level; - const uint texWidth = pt->width[level]; - const uint texHeight = pt->height[level]; + const uint texWidth = u_minify(pt->width0, level); + const uint texHeight = u_minify(pt->height0, level); const uint stride = ct->stride[level]; unsigned size; @@ -440,8 +437,8 @@ cell_transfer_unmap(struct pipe_screen *screen, struct pipe_texture *pt = transfer->texture; struct cell_texture *ct = cell_texture(pt); const uint level = ctrans->base.level; - const uint texWidth = pt->width[level]; - const uint texHeight = pt->height[level]; + const uint texWidth = u_minify(pt->width0, level); + const uint texHeight = u_minify(pt->height0, level); const uint stride = ct->stride[level]; if (!ct->mapped) { diff --git a/src/gallium/drivers/failover/fo_context.h b/src/gallium/drivers/failover/fo_context.h index 9ba86ba8664..149393712a3 100644 --- a/src/gallium/drivers/failover/fo_context.h +++ b/src/gallium/drivers/failover/fo_context.h @@ -72,6 +72,7 @@ struct failover_context { */ const struct fo_state *blend; const struct fo_state *sampler[PIPE_MAX_SAMPLERS]; + const struct fo_state *vertex_samplers[PIPE_MAX_VERTEX_SAMPLERS]; const struct fo_state *depth_stencil; const struct fo_state *rasterizer; const struct fo_state *fragment_shader; @@ -83,6 +84,7 @@ struct failover_context { struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; struct pipe_texture *texture[PIPE_MAX_SAMPLERS]; + struct pipe_texture *vertex_textures[PIPE_MAX_VERTEX_SAMPLERS]; struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffers[PIPE_MAX_ATTRIBS]; struct pipe_vertex_element vertex_elements[PIPE_MAX_ATTRIBS]; @@ -92,11 +94,15 @@ struct failover_context { void *sw_sampler_state[PIPE_MAX_SAMPLERS]; void *hw_sampler_state[PIPE_MAX_SAMPLERS]; + void *sw_vertex_sampler_state[PIPE_MAX_VERTEX_SAMPLERS]; + void *hw_vertex_sampler_state[PIPE_MAX_VERTEX_SAMPLERS]; unsigned dirty; unsigned num_samplers; + unsigned num_vertex_samplers; unsigned num_textures; + unsigned num_vertex_textures; unsigned mode; struct pipe_context *hw; diff --git a/src/gallium/drivers/failover/fo_state.c b/src/gallium/drivers/failover/fo_state.c index c8eb9262994..3f5f5560323 100644 --- a/src/gallium/drivers/failover/fo_state.c +++ b/src/gallium/drivers/failover/fo_state.c @@ -322,8 +322,9 @@ failover_create_sampler_state(struct pipe_context *pipe, } static void -failover_bind_sampler_states(struct pipe_context *pipe, - unsigned num, void **sampler) +failover_bind_fragment_sampler_states(struct pipe_context *pipe, + unsigned num, + void **sampler) { struct failover_context *failover = failover_context(pipe); struct fo_state *state = (struct fo_state*)sampler; @@ -339,10 +340,40 @@ failover_bind_sampler_states(struct pipe_context *pipe, } failover->dirty |= FO_NEW_SAMPLER; failover->num_samplers = num; - failover->sw->bind_sampler_states(failover->sw, num, - failover->sw_sampler_state); - failover->hw->bind_sampler_states(failover->hw, num, - failover->hw_sampler_state); + failover->sw->bind_fragment_sampler_states(failover->sw, num, + failover->sw_sampler_state); + failover->hw->bind_fragment_sampler_states(failover->hw, num, + failover->hw_sampler_state); +} + +static void +failover_bind_vertex_sampler_states(struct pipe_context *pipe, + unsigned num_samplers, + void **samplers) +{ + struct failover_context *failover = failover_context(pipe); + struct fo_state *state = (struct fo_state*)samplers; + uint i; + + assert(num_samplers <= PIPE_MAX_VERTEX_SAMPLERS); + + /* Check for no-op */ + if (num_samplers == failover->num_vertex_samplers && + !memcmp(failover->vertex_samplers, samplers, num_samplers * sizeof(void *))) { + return; + } + for (i = 0; i < PIPE_MAX_VERTEX_SAMPLERS; i++) { + failover->sw_vertex_sampler_state[i] = i < num_samplers ? state[i].sw_state : NULL; + failover->hw_vertex_sampler_state[i] = i < num_samplers ? state[i].hw_state : NULL; + } + failover->dirty |= FO_NEW_SAMPLER; + failover->num_vertex_samplers = num_samplers; + failover->sw->bind_vertex_sampler_states(failover->sw, + num_samplers, + failover->sw_vertex_sampler_state); + failover->hw->bind_vertex_sampler_states(failover->hw, + num_samplers, + failover->hw_vertex_sampler_state); } static void @@ -360,9 +391,9 @@ failover_delete_sampler_state(struct pipe_context *pipe, void *sampler) static void -failover_set_sampler_textures(struct pipe_context *pipe, - unsigned num, - struct pipe_texture **texture) +failover_set_fragment_sampler_textures(struct pipe_context *pipe, + unsigned num, + struct pipe_texture **texture) { struct failover_context *failover = failover_context(pipe); uint i; @@ -381,8 +412,38 @@ failover_set_sampler_textures(struct pipe_context *pipe, NULL); failover->dirty |= FO_NEW_TEXTURE; failover->num_textures = num; - failover->sw->set_sampler_textures( failover->sw, num, texture ); - failover->hw->set_sampler_textures( failover->hw, num, texture ); + failover->sw->set_fragment_sampler_textures( failover->sw, num, texture ); + failover->hw->set_fragment_sampler_textures( failover->hw, num, texture ); +} + + +static void +failover_set_vertex_sampler_textures(struct pipe_context *pipe, + unsigned num_textures, + struct pipe_texture **textures) +{ + struct failover_context *failover = failover_context(pipe); + uint i; + + assert(num_textures <= PIPE_MAX_VERTEX_SAMPLERS); + + /* Check for no-op */ + if (num_textures == failover->num_vertex_textures && + !memcmp(failover->vertex_textures, textures, num_textures * sizeof(struct pipe_texture *))) { + return; + } + for (i = 0; i < num_textures; i++) { + pipe_texture_reference((struct pipe_texture **)&failover->vertex_textures[i], + textures[i]); + } + for (i = num_textures; i < failover->num_vertex_textures; i++) { + pipe_texture_reference((struct pipe_texture **)&failover->vertex_textures[i], + NULL); + } + failover->dirty |= FO_NEW_TEXTURE; + failover->num_vertex_textures = num_textures; + failover->sw->set_vertex_sampler_textures(failover->sw, num_textures, textures); + failover->hw->set_vertex_sampler_textures(failover->hw, num_textures, textures); } @@ -453,7 +514,8 @@ failover_init_state_functions( struct failover_context *failover ) failover->pipe.bind_blend_state = failover_bind_blend_state; failover->pipe.delete_blend_state = failover_delete_blend_state; failover->pipe.create_sampler_state = failover_create_sampler_state; - failover->pipe.bind_sampler_states = failover_bind_sampler_states; + failover->pipe.bind_fragment_sampler_states = failover_bind_fragment_sampler_states; + failover->pipe.bind_vertex_sampler_states = failover_bind_vertex_sampler_states; failover->pipe.delete_sampler_state = failover_delete_sampler_state; failover->pipe.create_depth_stencil_alpha_state = failover_create_depth_stencil_state; failover->pipe.bind_depth_stencil_alpha_state = failover_bind_depth_stencil_state; @@ -473,7 +535,8 @@ failover_init_state_functions( struct failover_context *failover ) failover->pipe.set_framebuffer_state = failover_set_framebuffer_state; failover->pipe.set_polygon_stipple = failover_set_polygon_stipple; failover->pipe.set_scissor_state = failover_set_scissor_state; - failover->pipe.set_sampler_textures = failover_set_sampler_textures; + failover->pipe.set_fragment_sampler_textures = failover_set_fragment_sampler_textures; + failover->pipe.set_vertex_sampler_textures = failover_set_vertex_sampler_textures; failover->pipe.set_viewport_state = failover_set_viewport_state; failover->pipe.set_vertex_buffers = failover_set_vertex_buffers; failover->pipe.set_vertex_elements = failover_set_vertex_elements; diff --git a/src/gallium/drivers/failover/fo_state_emit.c b/src/gallium/drivers/failover/fo_state_emit.c index bd4fce9d209..a3341e33f80 100644 --- a/src/gallium/drivers/failover/fo_state_emit.c +++ b/src/gallium/drivers/failover/fo_state_emit.c @@ -92,13 +92,19 @@ failover_state_emit( struct failover_context *failover ) failover->sw->set_viewport_state( failover->sw, &failover->viewport ); if (failover->dirty & FO_NEW_SAMPLER) { - failover->sw->bind_sampler_states( failover->sw, failover->num_samplers, - failover->sw_sampler_state ); + failover->sw->bind_fragment_sampler_states( failover->sw, failover->num_samplers, + failover->sw_sampler_state ); + failover->sw->bind_vertex_sampler_states(failover->sw, + failover->num_vertex_samplers, + failover->sw_vertex_sampler_state); } if (failover->dirty & FO_NEW_TEXTURE) { - failover->sw->set_sampler_textures( failover->sw, failover->num_textures, - failover->texture ); + failover->sw->set_fragment_sampler_textures( failover->sw, failover->num_textures, + failover->texture ); + failover->sw->set_vertex_sampler_textures(failover->sw, + failover->num_vertex_textures, + failover->vertex_textures); } if (failover->dirty & FO_NEW_VERTEX_BUFFER) { diff --git a/src/gallium/drivers/i915/i915_state.c b/src/gallium/drivers/i915/i915_state.c index 71f00bc346a..9103847f1c5 100644 --- a/src/gallium/drivers/i915/i915_state.c +++ b/src/gallium/drivers/i915/i915_state.c @@ -767,7 +767,7 @@ i915_init_state_functions( struct i915_context *i915 ) i915->base.delete_blend_state = i915_delete_blend_state; i915->base.create_sampler_state = i915_create_sampler_state; - i915->base.bind_sampler_states = i915_bind_sampler_states; + i915->base.bind_fragment_sampler_states = i915_bind_sampler_states; i915->base.delete_sampler_state = i915_delete_sampler_state; i915->base.create_depth_stencil_alpha_state = i915_create_depth_stencil_state; @@ -791,7 +791,7 @@ i915_init_state_functions( struct i915_context *i915 ) i915->base.set_polygon_stipple = i915_set_polygon_stipple; i915->base.set_scissor_state = i915_set_scissor_state; - i915->base.set_sampler_textures = i915_set_sampler_textures; + i915->base.set_fragment_sampler_textures = i915_set_sampler_textures; i915->base.set_viewport_state = i915_set_viewport_state; i915->base.set_vertex_buffers = i915_set_vertex_buffers; i915->base.set_vertex_elements = i915_set_vertex_elements; diff --git a/src/gallium/drivers/i915/i915_state_sampler.c b/src/gallium/drivers/i915/i915_state_sampler.c index c5e9084d12e..cbac4175c8f 100644 --- a/src/gallium/drivers/i915/i915_state_sampler.c +++ b/src/gallium/drivers/i915/i915_state_sampler.c @@ -231,7 +231,7 @@ i915_update_texture(struct i915_context *i915, { const struct pipe_texture *pt = &tex->base; uint format, pitch; - const uint width = pt->width[0], height = pt->height[0], depth = pt->depth[0]; + const uint width = pt->width0, height = pt->height0, depth = pt->depth0; const uint num_levels = pt->last_level; unsigned max_lod = num_levels * 4; unsigned tiled = MS3_USE_FENCE_REGS; diff --git a/src/gallium/drivers/i915/i915_texture.c b/src/gallium/drivers/i915/i915_texture.c index 286c9ace8e5..c7b86dd4c57 100644 --- a/src/gallium/drivers/i915/i915_texture.c +++ b/src/gallium/drivers/i915/i915_texture.c @@ -105,10 +105,6 @@ i915_miptree_set_level_info(struct i915_texture *tex, assert(level < PIPE_MAX_TEXTURE_LEVELS); - pt->width[level] = w; - pt->height[level] = h; - pt->depth[level] = d; - pt->nblocksx[level] = pf_get_nblocksx(&pt->block, w); pt->nblocksy[level] = pf_get_nblocksy(&pt->block, h); @@ -168,16 +164,16 @@ i915_scanout_layout(struct i915_texture *tex) return FALSE; i915_miptree_set_level_info(tex, 0, 1, - tex->base.width[0], - tex->base.height[0], + tex->base.width0, + tex->base.height0, 1); i915_miptree_set_image_offset(tex, 0, 0, 0, 0); - if (tex->base.width[0] >= 240) { + if (tex->base.width0 >= 240) { tex->stride = power_of_two(tex->base.nblocksx[0] * pt->block.size); tex->total_nblocksy = round_up(tex->base.nblocksy[0], 8); tex->hw_tiled = INTEL_TILE_X; - } else if (tex->base.width[0] == 64 && tex->base.height[0] == 64) { + } else if (tex->base.width0 == 64 && tex->base.height0 == 64) { tex->stride = power_of_two(tex->base.nblocksx[0] * pt->block.size); tex->total_nblocksy = round_up(tex->base.nblocksy[0], 8); } else { @@ -185,7 +181,7 @@ i915_scanout_layout(struct i915_texture *tex) } debug_printf("%s size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, - tex->base.width[0], tex->base.height[0], pt->block.size, + tex->base.width0, tex->base.height0, pt->block.size, tex->stride, tex->total_nblocksy, tex->stride * tex->total_nblocksy); return TRUE; @@ -203,12 +199,12 @@ i915_display_target_layout(struct i915_texture *tex) return FALSE; /* fallback to normal textures for small textures */ - if (tex->base.width[0] < 240) + if (tex->base.width0 < 240) return FALSE; i915_miptree_set_level_info(tex, 0, 1, - tex->base.width[0], - tex->base.height[0], + tex->base.width0, + tex->base.height0, 1); i915_miptree_set_image_offset(tex, 0, 0, 0, 0); @@ -217,7 +213,7 @@ i915_display_target_layout(struct i915_texture *tex) tex->hw_tiled = INTEL_TILE_X; debug_printf("%s size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, - tex->base.width[0], tex->base.height[0], pt->block.size, + tex->base.width0, tex->base.height0, pt->block.size, tex->stride, tex->total_nblocksy, tex->stride * tex->total_nblocksy); return TRUE; @@ -228,8 +224,8 @@ i915_miptree_layout_2d(struct i915_texture *tex) { struct pipe_texture *pt = &tex->base; unsigned level; - unsigned width = pt->width[0]; - unsigned height = pt->height[0]; + unsigned width = pt->width0; + unsigned height = pt->height0; unsigned nblocksx = pt->nblocksx[0]; unsigned nblocksy = pt->nblocksy[0]; @@ -254,8 +250,8 @@ i915_miptree_layout_2d(struct i915_texture *tex) tex->total_nblocksy += nblocksy; - width = minify(width); - height = minify(height); + width = u_minify(width, 1); + height = u_minify(height, 1); nblocksx = pf_get_nblocksx(&pt->block, width); nblocksy = pf_get_nblocksy(&pt->block, height); } @@ -267,9 +263,9 @@ i915_miptree_layout_3d(struct i915_texture *tex) struct pipe_texture *pt = &tex->base; unsigned level; - unsigned width = pt->width[0]; - unsigned height = pt->height[0]; - unsigned depth = pt->depth[0]; + unsigned width = pt->width0; + unsigned height = pt->height0; + unsigned depth = pt->depth0; unsigned nblocksx = pt->nblocksx[0]; unsigned nblocksy = pt->nblocksy[0]; unsigned stack_nblocksy = 0; @@ -285,36 +281,34 @@ i915_miptree_layout_3d(struct i915_texture *tex) stack_nblocksy += MAX2(2, nblocksy); - width = minify(width); - height = minify(height); - depth = minify(depth); + width = u_minify(width, 1); + height = u_minify(height, 1); nblocksx = pf_get_nblocksx(&pt->block, width); nblocksy = pf_get_nblocksy(&pt->block, height); } /* Fixup depth image_offsets: */ - depth = pt->depth[0]; for (level = 0; level <= pt->last_level; level++) { unsigned i; for (i = 0; i < depth; i++) i915_miptree_set_image_offset(tex, level, i, 0, i * stack_nblocksy); - depth = minify(depth); + depth = u_minify(depth, 1); } /* Multiply slice size by texture depth for total size. It's * remarkable how wasteful of memory the i915 texture layouts * are. They are largely fixed in the i945. */ - tex->total_nblocksy = stack_nblocksy * pt->depth[0]; + tex->total_nblocksy = stack_nblocksy * pt->depth0; } static void i915_miptree_layout_cube(struct i915_texture *tex) { struct pipe_texture *pt = &tex->base; - unsigned width = pt->width[0], height = pt->height[0]; + unsigned width = pt->width0, height = pt->height0; const unsigned nblocks = pt->nblocksx[0]; unsigned level; unsigned face; @@ -383,8 +377,8 @@ i945_miptree_layout_2d(struct i915_texture *tex) unsigned level; unsigned x = 0; unsigned y = 0; - unsigned width = pt->width[0]; - unsigned height = pt->height[0]; + unsigned width = pt->width0; + unsigned height = pt->height0; unsigned nblocksx = pt->nblocksx[0]; unsigned nblocksy = pt->nblocksy[0]; @@ -407,8 +401,8 @@ i945_miptree_layout_2d(struct i915_texture *tex) */ if (pt->last_level > 0) { unsigned mip1_nblocksx - = align(pf_get_nblocksx(&pt->block, minify(width)), align_x) - + pf_get_nblocksx(&pt->block, minify(minify(width))); + = align(pf_get_nblocksx(&pt->block, u_minify(width, 1)), align_x) + + pf_get_nblocksx(&pt->block, u_minify(width, 2)); if (mip1_nblocksx > nblocksx) tex->stride = mip1_nblocksx * pt->block.size; @@ -439,8 +433,8 @@ i945_miptree_layout_2d(struct i915_texture *tex) y += nblocksy; } - width = minify(width); - height = minify(height); + width = u_minify(width, 1); + height = u_minify(height, 1); nblocksx = pf_get_nblocksx(&pt->block, width); nblocksy = pf_get_nblocksy(&pt->block, height); } @@ -450,9 +444,9 @@ static void i945_miptree_layout_3d(struct i915_texture *tex) { struct pipe_texture *pt = &tex->base; - unsigned width = pt->width[0]; - unsigned height = pt->height[0]; - unsigned depth = pt->depth[0]; + unsigned width = pt->width0; + unsigned height = pt->height0; + unsigned depth = pt->depth0; unsigned nblocksx = pt->nblocksx[0]; unsigned nblocksy = pt->nblocksy[0]; unsigned pack_x_pitch, pack_x_nr; @@ -495,9 +489,9 @@ i945_miptree_layout_3d(struct i915_texture *tex) pack_y_pitch >>= 1; } - width = minify(width); - height = minify(height); - depth = minify(depth); + width = u_minify(width, 1); + height = u_minify(height, 1); + depth = u_minify(depth, 1); nblocksx = pf_get_nblocksx(&pt->block, width); nblocksy = pf_get_nblocksy(&pt->block, height); } @@ -511,11 +505,11 @@ i945_miptree_layout_cube(struct i915_texture *tex) const unsigned nblocks = pt->nblocksx[0]; unsigned face; - unsigned width = pt->width[0]; - unsigned height = pt->height[0]; + unsigned width = pt->width0; + unsigned height = pt->height0; /* - printf("%s %i, %i\n", __FUNCTION__, pt->width[0], pt->height[0]); + printf("%s %i, %i\n", __FUNCTION__, pt->width0, pt->height0); */ assert(width == height); /* cubemap images are square */ @@ -651,8 +645,8 @@ i915_texture_create(struct pipe_screen *screen, pipe_reference_init(&tex->base.reference, 1); tex->base.screen = screen; - tex->base.nblocksx[0] = pf_get_nblocksx(&tex->base.block, tex->base.width[0]); - tex->base.nblocksy[0] = pf_get_nblocksy(&tex->base.block, tex->base.height[0]); + tex->base.nblocksx[0] = pf_get_nblocksx(&tex->base.block, tex->base.width0); + tex->base.nblocksy[0] = pf_get_nblocksy(&tex->base.block, tex->base.height0); if (is->is_i945) { if (!i945_miptree_layout(tex)) @@ -667,7 +661,7 @@ i915_texture_create(struct pipe_screen *screen, /* for scanouts and cursors, cursors arn't scanouts */ - if (templat->tex_usage & PIPE_TEXTURE_USAGE_PRIMARY && templat->width[0] != 64) + if (templat->tex_usage & PIPE_TEXTURE_USAGE_PRIMARY && templat->width0 != 64) buf_usage = INTEL_NEW_SCANOUT; else buf_usage = INTEL_NEW_TEXTURE; @@ -710,7 +704,7 @@ i915_texture_blanket(struct pipe_screen * screen, /* Only supports one type */ if (base->target != PIPE_TEXTURE_2D || base->last_level != 0 || - base->depth[0] != 1) { + base->depth0 != 1) { return NULL; } @@ -724,7 +718,7 @@ i915_texture_blanket(struct pipe_screen * screen, tex->stride = stride[0]; - i915_miptree_set_level_info(tex, 0, 1, base->width[0], base->height[0], 1); + i915_miptree_set_level_info(tex, 0, 1, base->width0, base->height0, 1); i915_miptree_set_image_offset(tex, 0, 0, 0, 0); pipe_buffer_reference(&tex->buffer, buffer); @@ -788,8 +782,8 @@ i915_get_tex_surface(struct pipe_screen *screen, pipe_reference_init(&ps->reference, 1); pipe_texture_reference(&ps->texture, pt); ps->format = pt->format; - ps->width = pt->width[level]; - ps->height = pt->height[level]; + ps->width = u_minify(pt->width0, level); + ps->height = u_minify(pt->height0, level); ps->offset = offset; ps->usage = flags; } @@ -919,7 +913,7 @@ i915_texture_blanket_intel(struct pipe_screen *screen, /* Only supports one type */ if (base->target != PIPE_TEXTURE_2D || base->last_level != 0 || - base->depth[0] != 1) { + base->depth0 != 1) { return NULL; } @@ -933,7 +927,7 @@ i915_texture_blanket_intel(struct pipe_screen *screen, tex->stride = stride; - i915_miptree_set_level_info(tex, 0, 1, base->width[0], base->height[0], 1); + i915_miptree_set_level_info(tex, 0, 1, base->width0, base->height0, 1); i915_miptree_set_image_offset(tex, 0, 0, 0, 0); tex->buffer = buffer; diff --git a/src/gallium/drivers/identity/id_context.c b/src/gallium/drivers/identity/id_context.c index 4e700089e33..4509c7b1e5c 100644 --- a/src/gallium/drivers/identity/id_context.c +++ b/src/gallium/drivers/identity/id_context.c @@ -221,16 +221,29 @@ identity_create_sampler_state(struct pipe_context *_pipe, } static void -identity_bind_sampler_states(struct pipe_context *_pipe, - unsigned num, - void **samplers) +identity_bind_fragment_sampler_states(struct pipe_context *_pipe, + unsigned num_samplers, + void **samplers) { struct identity_context *id_pipe = identity_context(_pipe); struct pipe_context *pipe = id_pipe->pipe; - pipe->bind_sampler_states(pipe, - num, - samplers); + pipe->bind_fragment_sampler_states(pipe, + num_samplers, + samplers); +} + +static void +identity_bind_vertex_sampler_states(struct pipe_context *_pipe, + unsigned num_samplers, + void **samplers) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + pipe->bind_vertex_sampler_states(pipe, + num_samplers, + samplers); } static void @@ -480,9 +493,9 @@ identity_set_viewport_state(struct pipe_context *_pipe, } static void -identity_set_sampler_textures(struct pipe_context *_pipe, - unsigned num_textures, - struct pipe_texture **_textures) +identity_set_fragment_sampler_textures(struct pipe_context *_pipe, + unsigned num_textures, + struct pipe_texture **_textures) { struct identity_context *id_pipe = identity_context(_pipe); struct pipe_context *pipe = id_pipe->pipe; @@ -499,9 +512,34 @@ identity_set_sampler_textures(struct pipe_context *_pipe, textures = unwrapped_textures; } - pipe->set_sampler_textures(pipe, - num_textures, - textures); + pipe->set_fragment_sampler_textures(pipe, + num_textures, + textures); +} + +static void +identity_set_vertex_sampler_textures(struct pipe_context *_pipe, + unsigned num_textures, + struct pipe_texture **_textures) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + struct pipe_texture *unwrapped_textures[PIPE_MAX_VERTEX_SAMPLERS]; + struct pipe_texture **textures = NULL; + unsigned i; + + if (_textures) { + for (i = 0; i < num_textures; i++) + unwrapped_textures[i] = identity_texture_unwrap(_textures[i]); + for (; i < PIPE_MAX_VERTEX_SAMPLERS; i++) + unwrapped_textures[i] = NULL; + + textures = unwrapped_textures; + } + + pipe->set_vertex_sampler_textures(pipe, + num_textures, + textures); } static void @@ -682,7 +720,8 @@ identity_context_create(struct pipe_screen *_screen, struct pipe_context *pipe) id_pipe->base.bind_blend_state = identity_bind_blend_state; id_pipe->base.delete_blend_state = identity_delete_blend_state; id_pipe->base.create_sampler_state = identity_create_sampler_state; - id_pipe->base.bind_sampler_states = identity_bind_sampler_states; + id_pipe->base.bind_fragment_sampler_states = identity_bind_fragment_sampler_states; + id_pipe->base.bind_vertex_sampler_states = identity_bind_vertex_sampler_states; id_pipe->base.delete_sampler_state = identity_delete_sampler_state; id_pipe->base.create_rasterizer_state = identity_create_rasterizer_state; id_pipe->base.bind_rasterizer_state = identity_bind_rasterizer_state; @@ -703,7 +742,8 @@ identity_context_create(struct pipe_screen *_screen, struct pipe_context *pipe) id_pipe->base.set_polygon_stipple = identity_set_polygon_stipple; id_pipe->base.set_scissor_state = identity_set_scissor_state; id_pipe->base.set_viewport_state = identity_set_viewport_state; - id_pipe->base.set_sampler_textures = identity_set_sampler_textures; + id_pipe->base.set_fragment_sampler_textures = identity_set_vertex_sampler_textures; + id_pipe->base.set_vertex_sampler_textures = identity_set_vertex_sampler_textures; id_pipe->base.set_vertex_buffers = identity_set_vertex_buffers; id_pipe->base.set_vertex_elements = identity_set_vertex_elements; id_pipe->base.surface_copy = identity_surface_copy; diff --git a/src/gallium/drivers/identity/id_objects.c b/src/gallium/drivers/identity/id_objects.c index e893e599408..bc9bc7121d5 100644 --- a/src/gallium/drivers/identity/id_objects.c +++ b/src/gallium/drivers/identity/id_objects.c @@ -180,3 +180,42 @@ identity_transfer_destroy(struct identity_transfer *id_transfer) screen->tex_transfer_destroy(id_transfer->transfer); FREE(id_transfer); } + +struct pipe_video_surface * +identity_video_surface_create(struct identity_screen *id_screen, + struct pipe_video_surface *video_surface) +{ + struct identity_video_surface *id_video_surface; + + if (!video_surface) { + goto error; + } + + assert(video_surface->screen == id_screen->screen); + + id_video_surface = CALLOC_STRUCT(identity_video_surface); + if (!id_video_surface) { + goto error; + } + + memcpy(&id_video_surface->base, + video_surface, + sizeof(struct pipe_video_surface)); + + pipe_reference_init(&id_video_surface->base.reference, 1); + id_video_surface->base.screen = &id_screen->base; + id_video_surface->video_surface = video_surface; + + return &id_video_surface->base; + +error: + pipe_video_surface_reference(&video_surface, NULL); + return NULL; +} + +void +identity_video_surface_destroy(struct identity_video_surface *id_video_surface) +{ + pipe_video_surface_reference(&id_video_surface->video_surface, NULL); + FREE(id_video_surface); +} diff --git a/src/gallium/drivers/identity/id_objects.h b/src/gallium/drivers/identity/id_objects.h index ce58faa3c7c..77cc7190798 100644 --- a/src/gallium/drivers/identity/id_objects.h +++ b/src/gallium/drivers/identity/id_objects.h @@ -31,6 +31,7 @@ #include "pipe/p_compiler.h" #include "pipe/p_state.h" +#include "pipe/p_video_state.h" #include "id_screen.h" @@ -67,6 +68,14 @@ struct identity_transfer }; +struct identity_video_surface +{ + struct pipe_video_surface base; + + struct pipe_video_surface *video_surface; +}; + + static INLINE struct identity_buffer * identity_buffer(struct pipe_buffer *_buffer) { @@ -103,6 +112,15 @@ identity_transfer(struct pipe_transfer *_transfer) return (struct identity_transfer *)_transfer; } +static INLINE struct identity_video_surface * +identity_video_surface(struct pipe_video_surface *_video_surface) +{ + if (!_video_surface) { + return NULL; + } + (void)identity_screen(_video_surface->screen); + return (struct identity_video_surface *)_video_surface; +} static INLINE struct pipe_buffer * identity_buffer_unwrap(struct pipe_buffer *_buffer) @@ -165,5 +183,12 @@ identity_transfer_create(struct identity_texture *id_texture, void identity_transfer_destroy(struct identity_transfer *id_transfer); +struct pipe_video_surface * +identity_video_surface_create(struct identity_screen *id_screen, + struct pipe_video_surface *video_surface); + +void +identity_video_surface_destroy(struct identity_video_surface *id_video_surface); + #endif /* ID_OBJECTS_H */ diff --git a/src/gallium/drivers/identity/id_public.h b/src/gallium/drivers/identity/id_public.h index cac14cfd604..3d2862eaa01 100644 --- a/src/gallium/drivers/identity/id_public.h +++ b/src/gallium/drivers/identity/id_public.h @@ -37,4 +37,4 @@ identity_screen_create(struct pipe_screen *screen); struct pipe_context * identity_context_create(struct pipe_screen *screen, struct pipe_context *pipe); -#endif /* PT_PUBLIC_H */ +#endif /* ID_PUBLIC_H */ diff --git a/src/gallium/drivers/identity/id_screen.c b/src/gallium/drivers/identity/id_screen.c index 26439637d08..53eae3ef544 100644 --- a/src/gallium/drivers/identity/id_screen.c +++ b/src/gallium/drivers/identity/id_screen.c @@ -379,6 +379,33 @@ identity_screen_buffer_destroy(struct pipe_buffer *_buffer) identity_buffer_destroy(identity_buffer(_buffer)); } +static struct pipe_video_surface * +identity_screen_video_surface_create(struct pipe_screen *_screen, + enum pipe_video_chroma_format chroma_format, + unsigned width, + unsigned height) +{ + struct identity_screen *id_screen = identity_screen(_screen); + struct pipe_screen *screen = id_screen->screen; + struct pipe_video_surface *result; + + result = screen->video_surface_create(screen, + chroma_format, + width, + height); + + if (result) { + return identity_video_surface_create(id_screen, result); + } + return NULL; +} + +static void +identity_screen_video_surface_destroy(struct pipe_video_surface *_vsfc) +{ + identity_video_surface_destroy(identity_video_surface(_vsfc)); +} + static void identity_screen_flush_frontbuffer(struct pipe_screen *_screen, struct pipe_surface *_surface, @@ -472,6 +499,12 @@ identity_screen_create(struct pipe_screen *screen) if (screen->buffer_unmap) id_screen->base.buffer_unmap = identity_screen_buffer_unmap; id_screen->base.buffer_destroy = identity_screen_buffer_destroy; + if (screen->video_surface_create) { + id_screen->base.video_surface_create = identity_screen_video_surface_create; + } + if (screen->video_surface_destroy) { + id_screen->base.video_surface_destroy = identity_screen_video_surface_destroy; + } id_screen->base.flush_frontbuffer = identity_screen_flush_frontbuffer; id_screen->base.fence_reference = identity_screen_fence_reference; id_screen->base.fence_signalled = identity_screen_fence_signalled; diff --git a/src/gallium/drivers/llvmpipe/README b/src/gallium/drivers/llvmpipe/README index 89d08834a3c..0c3f00fd58f 100644 --- a/src/gallium/drivers/llvmpipe/README +++ b/src/gallium/drivers/llvmpipe/README @@ -51,21 +51,22 @@ Requirements - Linux - - udis86, http://udis86.sourceforge.net/ . Use my repository, which decodes - opcodes not yet supported by upstream. + - A x86 or amd64 processor. 64bit mode is preferred. - git clone git://people.freedesktop.org/~jrfonseca/udis86 - cd udis86 - ./configure --with-pic - make - sudo make install + Support for sse2 is strongly encouraged. Support for ssse3, and sse4.1 will + yield the most efficient code. The less features the CPU has the more + likely is that you ran into underperforming, buggy, or incomplete code. + + See /proc/cpuinfo to know what your CPU supports. + + - LLVM 2.5 or greater. LLVM 2.6 is preferred. - - LLVM 2.5. On Debian based distributions do: + On Debian based distributions do: aptitude install llvm-dev - There is a typo in one of the llvm-dev 2.5 headers, that causes compilation - errors in the debug build: + There is a typo in one of the llvm 2.5 headers, that may cause compilation + errors. To fix it apply the change: --- /usr/include/llvm-c/Core.h.orig 2009-08-10 15:38:54.000000000 +0100 +++ /usr/include/llvm-c/Core.h 2009-08-10 15:38:25.000000000 +0100 @@ -79,12 +80,17 @@ Requirements #endif return reinterpret_cast<T**>(Vals); - - A x86 or amd64 processor with support for sse2, sse3, and sse4.1 SIMD - instructions. This is necessary because we emit several SSE intrinsics for - convenience. See /proc/cpuinfo to know what your CPU supports. - - - scons + - scons (optional) + - udis86, http://udis86.sourceforge.net/ (optional): + + git clone git://udis86.git.sourceforge.net/gitroot/udis86/udis86 + cd udis86 + ./autogen.sh + ./configure --with-pic + make + sudo make install + Building ======== diff --git a/src/gallium/drivers/llvmpipe/lp_bld_depth.c b/src/gallium/drivers/llvmpipe/lp_bld_depth.c index 98ec1cb1b9d..d438c0e63d7 100644 --- a/src/gallium/drivers/llvmpipe/lp_bld_depth.c +++ b/src/gallium/drivers/llvmpipe/lp_bld_depth.c @@ -90,7 +90,7 @@ lp_depth_type(const struct util_format_description *format_desc, if(format_desc->channel[swizzle].type == UTIL_FORMAT_TYPE_FLOAT) { type.floating = TRUE; - assert(swizzle = 0); + assert(swizzle == 0); assert(format_desc->channel[swizzle].size == format_desc->block.bits); } else if(format_desc->channel[swizzle].type == UTIL_FORMAT_TYPE_UNSIGNED) { diff --git a/src/gallium/drivers/llvmpipe/lp_bld_sample.c b/src/gallium/drivers/llvmpipe/lp_bld_sample.c index 4d272bea87e..af70ddc6ab9 100644 --- a/src/gallium/drivers/llvmpipe/lp_bld_sample.c +++ b/src/gallium/drivers/llvmpipe/lp_bld_sample.c @@ -59,9 +59,9 @@ lp_sampler_static_state(struct lp_sampler_static_state *state, state->format = texture->format; state->target = texture->target; - state->pot_width = util_is_pot(texture->width[0]); - state->pot_height = util_is_pot(texture->height[0]); - state->pot_depth = util_is_pot(texture->depth[0]); + state->pot_width = util_is_pot(texture->width0); + state->pot_height = util_is_pot(texture->height0); + state->pot_depth = util_is_pot(texture->depth0); state->wrap_s = sampler->wrap_s; state->wrap_t = sampler->wrap_t; diff --git a/src/gallium/drivers/llvmpipe/lp_context.c b/src/gallium/drivers/llvmpipe/lp_context.c index 57e71f3e986..c081f6de036 100644 --- a/src/gallium/drivers/llvmpipe/lp_context.c +++ b/src/gallium/drivers/llvmpipe/lp_context.c @@ -180,7 +180,7 @@ llvmpipe_create( struct pipe_screen *screen ) llvmpipe->pipe.delete_blend_state = llvmpipe_delete_blend_state; llvmpipe->pipe.create_sampler_state = llvmpipe_create_sampler_state; - llvmpipe->pipe.bind_sampler_states = llvmpipe_bind_sampler_states; + llvmpipe->pipe.bind_fragment_sampler_states = llvmpipe_bind_sampler_states; llvmpipe->pipe.delete_sampler_state = llvmpipe_delete_sampler_state; llvmpipe->pipe.create_depth_stencil_alpha_state = llvmpipe_create_depth_stencil_state; @@ -205,7 +205,7 @@ llvmpipe_create( struct pipe_screen *screen ) llvmpipe->pipe.set_framebuffer_state = llvmpipe_set_framebuffer_state; llvmpipe->pipe.set_polygon_stipple = llvmpipe_set_polygon_stipple; llvmpipe->pipe.set_scissor_state = llvmpipe_set_scissor_state; - llvmpipe->pipe.set_sampler_textures = llvmpipe_set_sampler_textures; + llvmpipe->pipe.set_fragment_sampler_textures = llvmpipe_set_sampler_textures; llvmpipe->pipe.set_viewport_state = llvmpipe_set_viewport_state; llvmpipe->pipe.set_vertex_buffers = llvmpipe_set_vertex_buffers; diff --git a/src/gallium/drivers/llvmpipe/lp_jit.c b/src/gallium/drivers/llvmpipe/lp_jit.c index 13535dd638e..bce3baec164 100644 --- a/src/gallium/drivers/llvmpipe/lp_jit.c +++ b/src/gallium/drivers/llvmpipe/lp_jit.c @@ -154,6 +154,7 @@ lp_jit_screen_init(struct llvmpipe_screen *screen) #if 0 /* For simulating less capable machines */ util_cpu_caps.has_sse3 = 0; + util_cpu_caps.has_ssse3 = 0; util_cpu_caps.has_sse4_1 = 0; #endif @@ -167,7 +168,7 @@ lp_jit_screen_init(struct llvmpipe_screen *screen) if (LLVMCreateJITCompiler(&screen->engine, screen->provider, 1, &error)) { _debug_printf("%s\n", error); LLVMDisposeMessage(error); - abort(); + assert(0); } screen->target = LLVMGetExecutionEngineTargetData(screen->engine); diff --git a/src/gallium/drivers/llvmpipe/lp_screen.c b/src/gallium/drivers/llvmpipe/lp_screen.c index 05189274589..0fb133486aa 100644 --- a/src/gallium/drivers/llvmpipe/lp_screen.c +++ b/src/gallium/drivers/llvmpipe/lp_screen.c @@ -27,6 +27,7 @@ #include "util/u_memory.h" +#include "util/u_format.h" #include "pipe/p_defines.h" #include "pipe/p_screen.h" @@ -131,17 +132,17 @@ llvmpipe_is_format_supported( struct pipe_screen *_screen, { struct llvmpipe_screen *screen = llvmpipe_screen(_screen); struct llvmpipe_winsys *winsys = screen->winsys; + const struct util_format_description *format_desc; + + format_desc = util_format_description(format); + if(!format_desc) + return FALSE; assert(target == PIPE_TEXTURE_1D || target == PIPE_TEXTURE_2D || target == PIPE_TEXTURE_3D || target == PIPE_TEXTURE_CUBE); - if(format == PIPE_FORMAT_Z16_UNORM) - return FALSE; - if(format == PIPE_FORMAT_S8_UNORM) - return FALSE; - switch(format) { case PIPE_FORMAT_DXT1_RGB: case PIPE_FORMAT_DXT1_RGBA: @@ -152,8 +153,51 @@ llvmpipe_is_format_supported( struct pipe_screen *_screen, break; } - if(tex_usage & PIPE_TEXTURE_USAGE_DISPLAY_TARGET) - return winsys->is_displaytarget_format_supported(winsys, format); + if(tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET) { + if(format_desc->block.width != 1 || + format_desc->block.height != 1) + return FALSE; + + if(format_desc->layout != UTIL_FORMAT_LAYOUT_SCALAR && + format_desc->layout != UTIL_FORMAT_LAYOUT_ARITH && + format_desc->layout != UTIL_FORMAT_LAYOUT_ARRAY) + return FALSE; + + if(format_desc->colorspace != UTIL_FORMAT_COLORSPACE_RGB && + format_desc->colorspace != UTIL_FORMAT_COLORSPACE_SRGB) + return FALSE; + } + + if(tex_usage & PIPE_TEXTURE_USAGE_DISPLAY_TARGET) { + if(!winsys->is_displaytarget_format_supported(winsys, format)) + return FALSE; + } + + if(tex_usage & PIPE_TEXTURE_USAGE_DEPTH_STENCIL) { + if(format_desc->colorspace != UTIL_FORMAT_COLORSPACE_ZS) + return FALSE; + + /* FIXME: Temporary restriction. See lp_state_fs.c. */ + if(format_desc->block.bits != 32) + return FALSE; + } + + /* FIXME: Temporary restrictions. See lp_bld_sample_soa.c */ + if(tex_usage & PIPE_TEXTURE_USAGE_SAMPLER) { + if(format_desc->block.width != 1 || + format_desc->block.height != 1) + return FALSE; + + if(format_desc->layout != UTIL_FORMAT_LAYOUT_SCALAR && + format_desc->layout != UTIL_FORMAT_LAYOUT_ARITH && + format_desc->layout != UTIL_FORMAT_LAYOUT_ARRAY) + return FALSE; + + if(format_desc->colorspace != UTIL_FORMAT_COLORSPACE_RGB && + format_desc->colorspace != UTIL_FORMAT_COLORSPACE_SRGB && + format_desc->colorspace != UTIL_FORMAT_COLORSPACE_ZS) + return FALSE; + } return TRUE; } diff --git a/src/gallium/drivers/llvmpipe/lp_state_fs.c b/src/gallium/drivers/llvmpipe/lp_state_fs.c index 2e9aa9fffe3..ee0f69b2af9 100644 --- a/src/gallium/drivers/llvmpipe/lp_state_fs.c +++ b/src/gallium/drivers/llvmpipe/lp_state_fs.c @@ -148,6 +148,20 @@ generate_depth(LLVMBuilderRef builder, format_desc = util_format_description(key->zsbuf_format); assert(format_desc); + /* + * Depths are expected to be between 0 and 1, even if they are stored in + * floats. Setting these bits here will ensure that the lp_build_conv() call + * below won't try to unnecessarily clamp the incoming values. + */ + if(src_type.floating) { + src_type.sign = FALSE; + src_type.norm = TRUE; + } + else { + assert(!src_type.sign); + assert(src_type.norm); + } + /* Pick the depth type. */ dst_type = lp_depth_type(format_desc, src_type.width*src_type.length); @@ -155,14 +169,11 @@ generate_depth(LLVMBuilderRef builder, assert(dst_type.width == src_type.width); assert(dst_type.length == src_type.length); -#if 1 - src = lp_build_clamped_float_to_unsigned_norm(builder, - src_type, - dst_type.width, - src); -#else lp_build_conv(builder, src_type, dst_type, &src, 1, &src, 1); -#endif + + dst_ptr = LLVMBuildBitCast(builder, + dst_ptr, + LLVMPointerType(lp_build_vec_type(dst_type), 0), ""); lp_build_depth_test(builder, &key->depth, @@ -611,10 +622,12 @@ generate_fragment(struct llvmpipe_context *lp, * Translate the LLVM IR into machine code. */ +#ifdef DEBUG if(LLVMVerifyFunction(variant->function, LLVMPrintMessageAction)) { LLVMDumpValue(variant->function); - abort(); + assert(0); } +#endif LLVMRunFunctionPassManager(screen->pass, variant->function); diff --git a/src/gallium/drivers/llvmpipe/lp_state_sampler.c b/src/gallium/drivers/llvmpipe/lp_state_sampler.c index c69d90c723a..8333805a3fd 100644 --- a/src/gallium/drivers/llvmpipe/lp_state_sampler.c +++ b/src/gallium/drivers/llvmpipe/lp_state_sampler.c @@ -102,8 +102,8 @@ llvmpipe_set_sampler_textures(struct pipe_context *pipe, if(tex) { struct llvmpipe_texture *lp_tex = llvmpipe_texture(tex); struct lp_jit_texture *jit_tex = &llvmpipe->jit_context.textures[i]; - jit_tex->width = tex->width[0]; - jit_tex->height = tex->height[0]; + jit_tex->width = tex->width0; + jit_tex->height = tex->height0; jit_tex->stride = lp_tex->stride[0]; if(!lp_tex->dt) jit_tex->data = lp_tex->data; diff --git a/src/gallium/drivers/llvmpipe/lp_state_vs.c b/src/gallium/drivers/llvmpipe/lp_state_vs.c index 15c30296144..8a761648e7e 100644 --- a/src/gallium/drivers/llvmpipe/lp_state_vs.c +++ b/src/gallium/drivers/llvmpipe/lp_state_vs.c @@ -92,5 +92,6 @@ llvmpipe_delete_vs_state(struct pipe_context *pipe, void *vs) (struct lp_vertex_shader *)vs; draw_delete_vertex_shader(llvmpipe->draw, state->draw_data); + FREE( (void *)state->shader.tokens ); FREE( state ); } diff --git a/src/gallium/drivers/llvmpipe/lp_tex_cache.c b/src/gallium/drivers/llvmpipe/lp_tex_cache.c index 773e8482425..c7c4143bc62 100644 --- a/src/gallium/drivers/llvmpipe/lp_tex_cache.c +++ b/src/gallium/drivers/llvmpipe/lp_tex_cache.c @@ -36,6 +36,7 @@ #include "util/u_memory.h" #include "util/u_tile.h" #include "util/u_format.h" +#include "util/u_math.h" #include "lp_context.h" #include "lp_surface.h" #include "lp_texture.h" @@ -270,8 +271,8 @@ lp_find_cached_tex_tile(struct llvmpipe_tex_tile_cache *tc, addr.bits.level, addr.bits.z, PIPE_TRANSFER_READ, 0, 0, - tc->texture->width[addr.bits.level], - tc->texture->height[addr.bits.level]); + u_minify(tc->texture->width0, addr.bits.level), + u_minify(tc->texture->height0, addr.bits.level)); tc->tex_trans_map = screen->transfer_map(screen, tc->tex_trans); diff --git a/src/gallium/drivers/llvmpipe/lp_tex_sample_c.c b/src/gallium/drivers/llvmpipe/lp_tex_sample_c.c index a1365a045f1..0d01c07fb5e 100644 --- a/src/gallium/drivers/llvmpipe/lp_tex_sample_c.c +++ b/src/gallium/drivers/llvmpipe/lp_tex_sample_c.c @@ -544,7 +544,7 @@ compute_lambda(struct tgsi_sampler *tgsi_sampler, float dsdy = s[QUAD_TOP_LEFT] - s[QUAD_BOTTOM_LEFT]; dsdx = fabsf(dsdx); dsdy = fabsf(dsdy); - rho = MAX2(dsdx, dsdy) * texture->width[0]; + rho = MAX2(dsdx, dsdy) * texture->width0; } if (t) { float dtdx = t[QUAD_BOTTOM_RIGHT] - t[QUAD_BOTTOM_LEFT]; @@ -552,7 +552,7 @@ compute_lambda(struct tgsi_sampler *tgsi_sampler, float max; dtdx = fabsf(dtdx); dtdy = fabsf(dtdy); - max = MAX2(dtdx, dtdy) * texture->height[0]; + max = MAX2(dtdx, dtdy) * texture->height0; rho = MAX2(rho, max); } if (p) { @@ -561,7 +561,7 @@ compute_lambda(struct tgsi_sampler *tgsi_sampler, float max; dpdx = fabsf(dpdx); dpdy = fabsf(dpdy); - max = MAX2(dpdx, dpdy) * texture->depth[0]; + max = MAX2(dpdx, dpdy) * texture->depth0; rho = MAX2(rho, max); } @@ -726,9 +726,9 @@ get_texel(const struct tgsi_sampler *tgsi_sampler, const struct pipe_texture *texture = samp->texture; const struct pipe_sampler_state *sampler = samp->sampler; - if (x < 0 || x >= (int) texture->width[level] || - y < 0 || y >= (int) texture->height[level] || - z < 0 || z >= (int) texture->depth[level]) { + if (x < 0 || x >= (int) u_minify(texture->width0, level) || + y < 0 || y >= (int) u_minify(texture->height0, level) || + z < 0 || z >= (int) u_minify(texture->depth0, level)) { rgba[0][j] = sampler->border_color[0]; rgba[1][j] = sampler->border_color[1]; rgba[2][j] = sampler->border_color[2]; @@ -1093,8 +1093,8 @@ lp_get_samples_2d_common(struct tgsi_sampler *tgsi_sampler, assert(sampler->normalized_coords); - width = texture->width[level0]; - height = texture->height[level0]; + width = u_minify(texture->width0, level0); + height = u_minify(texture->height0, level0); assert(width > 0); @@ -1250,9 +1250,9 @@ lp_get_samples_3d(struct tgsi_sampler *tgsi_sampler, assert(sampler->normalized_coords); - width = texture->width[level0]; - height = texture->height[level0]; - depth = texture->depth[level0]; + width = u_minify(texture->width0, level0); + height = u_minify(texture->height0, level0); + depth = u_minify(texture->depth0, level0); assert(width > 0); assert(height > 0); @@ -1394,8 +1394,8 @@ lp_get_samples_rect(struct tgsi_sampler *tgsi_sampler, /* texture RECTS cannot be mipmapped */ assert(level0 == level1); - width = texture->width[level0]; - height = texture->height[level0]; + width = u_minify(texture->width0, level0); + height = u_minify(texture->height0, level0); assert(width > 0); @@ -1513,8 +1513,8 @@ lp_get_samples(struct tgsi_sampler *tgsi_sampler, /* Do this elsewhere: */ - samp->xpot = util_unsigned_logbase2( samp->texture->width[0] ); - samp->ypot = util_unsigned_logbase2( samp->texture->height[0] ); + samp->xpot = util_unsigned_logbase2( samp->texture->width0 ); + samp->ypot = util_unsigned_logbase2( samp->texture->height0 ); /* Try to hook in a faster sampler. Ultimately we'll have to * code-generate these. Luckily most of this looks like it is diff --git a/src/gallium/drivers/llvmpipe/lp_texture.c b/src/gallium/drivers/llvmpipe/lp_texture.c index a00f2495dfc..65d62fd0723 100644 --- a/src/gallium/drivers/llvmpipe/lp_texture.c +++ b/src/gallium/drivers/llvmpipe/lp_texture.c @@ -57,9 +57,9 @@ llvmpipe_texture_layout(struct llvmpipe_screen *screen, { struct pipe_texture *pt = &lpt->base; unsigned level; - unsigned width = pt->width[0]; - unsigned height = pt->height[0]; - unsigned depth = pt->depth[0]; + unsigned width = pt->width0; + unsigned height = pt->height0; + unsigned depth = pt->depth0; unsigned buffer_size = 0; @@ -68,9 +68,6 @@ llvmpipe_texture_layout(struct llvmpipe_screen *screen, for (level = 0; level <= pt->last_level; level++) { unsigned nblocksx, nblocksy; - pt->width[level] = width; - pt->height[level] = height; - pt->depth[level] = depth; pt->nblocksx[level] = pf_get_nblocksx(&pt->block, width); pt->nblocksy[level] = pf_get_nblocksy(&pt->block, height); @@ -87,9 +84,9 @@ llvmpipe_texture_layout(struct llvmpipe_screen *screen, ((pt->target == PIPE_TEXTURE_CUBE) ? 6 : depth) * lpt->stride[level]); - width = minify(width); - height = minify(height); - depth = minify(depth); + width = u_minify(width, 1); + height = u_minify(height, 1); + depth = u_minify(depth, 1); } lpt->data = align_malloc(buffer_size, 16); @@ -104,13 +101,13 @@ llvmpipe_displaytarget_layout(struct llvmpipe_screen *screen, struct llvmpipe_winsys *winsys = screen->winsys; pf_get_block(lpt->base.format, &lpt->base.block); - lpt->base.nblocksx[0] = pf_get_nblocksx(&lpt->base.block, lpt->base.width[0]); - lpt->base.nblocksy[0] = pf_get_nblocksy(&lpt->base.block, lpt->base.height[0]); + lpt->base.nblocksx[0] = pf_get_nblocksx(&lpt->base.block, lpt->base.width0); + lpt->base.nblocksy[0] = pf_get_nblocksy(&lpt->base.block, lpt->base.height0); lpt->dt = winsys->displaytarget_create(winsys, lpt->base.format, - lpt->base.width[0], - lpt->base.height[0], + lpt->base.width0, + lpt->base.height0, 16, &lpt->stride[0] ); @@ -172,7 +169,7 @@ llvmpipe_texture_blanket(struct pipe_screen * screen, /* Only supports one type */ if (base->target != PIPE_TEXTURE_2D || base->last_level != 0 || - base->depth[0] != 1) { + base->depth0 != 1) { return NULL; } @@ -183,8 +180,8 @@ llvmpipe_texture_blanket(struct pipe_screen * screen, lpt->base = *base; pipe_reference_init(&lpt->base.reference, 1); lpt->base.screen = screen; - lpt->base.nblocksx[0] = pf_get_nblocksx(&lpt->base.block, lpt->base.width[0]); - lpt->base.nblocksy[0] = pf_get_nblocksy(&lpt->base.block, lpt->base.height[0]); + lpt->base.nblocksx[0] = pf_get_nblocksx(&lpt->base.block, lpt->base.width0); + lpt->base.nblocksy[0] = pf_get_nblocksy(&lpt->base.block, lpt->base.height0); lpt->stride[0] = stride[0]; pipe_buffer_reference(&lpt->buffer, buffer); @@ -229,8 +226,8 @@ llvmpipe_get_tex_surface(struct pipe_screen *screen, pipe_reference_init(&ps->reference, 1); pipe_texture_reference(&ps->texture, pt); ps->format = pt->format; - ps->width = pt->width[level]; - ps->height = pt->height[level]; + ps->width = u_minify(pt->width0, level); + ps->height = u_minify(pt->height0, level); ps->offset = lpt->level_offset[level]; ps->usage = usage; diff --git a/src/gallium/drivers/nv04/nv04_fragtex.c b/src/gallium/drivers/nv04/nv04_fragtex.c index 21f990fd536..0cce71ad1de 100644 --- a/src/gallium/drivers/nv04/nv04_fragtex.c +++ b/src/gallium/drivers/nv04/nv04_fragtex.c @@ -57,8 +57,8 @@ nv04_fragtex_build(struct nv04_context *nv04, int unit) | NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ORIGIN_FOH_CORNER | nv04_fragtex_format(pt->format) | ( (pt->last_level + 1) << NV04_DX5_TEXTURED_TRIANGLE_FORMAT_MIPMAP_LEVELS_SHIFT ) - | ( log2i(pt->width[0]) << NV04_DX5_TEXTURED_TRIANGLE_FORMAT_BASE_SIZE_U_SHIFT ) - | ( log2i(pt->height[0]) << NV04_DX5_TEXTURED_TRIANGLE_FORMAT_BASE_SIZE_V_SHIFT ) + | ( log2i(pt->width0) << NV04_DX5_TEXTURED_TRIANGLE_FORMAT_BASE_SIZE_U_SHIFT ) + | ( log2i(pt->height0) << NV04_DX5_TEXTURED_TRIANGLE_FORMAT_BASE_SIZE_V_SHIFT ) | NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSU_CLAMP_TO_EDGE | NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSV_CLAMP_TO_EDGE ; diff --git a/src/gallium/drivers/nv04/nv04_miptree.c b/src/gallium/drivers/nv04/nv04_miptree.c index 93f752faec9..4fd72c82e62 100644 --- a/src/gallium/drivers/nv04/nv04_miptree.c +++ b/src/gallium/drivers/nv04/nv04_miptree.c @@ -1,6 +1,7 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" +#include "util/u_math.h" #include "nv04_context.h" #include "nv04_screen.h" @@ -9,31 +10,29 @@ static void nv04_miptree_layout(struct nv04_miptree *nv04mt) { struct pipe_texture *pt = &nv04mt->base; - uint width = pt->width[0], height = pt->height[0]; + uint width = pt->width0, height = pt->height0; uint offset = 0; int nr_faces, l; nr_faces = 1; for (l = 0; l <= pt->last_level; l++) { - pt->width[l] = width; - pt->height[l] = height; pt->nblocksx[l] = pf_get_nblocksx(&pt->block, width); pt->nblocksy[l] = pf_get_nblocksy(&pt->block, height); - nv04mt->level[l].pitch = pt->width[0]; + nv04mt->level[l].pitch = pt->width0; nv04mt->level[l].pitch = (nv04mt->level[l].pitch + 63) & ~63; - width = MAX2(1, width >> 1); - height = MAX2(1, height >> 1); + width = u_minify(width, 1); + height = u_minify(height, 1); } for (l = 0; l <= pt->last_level; l++) { nv04mt->level[l].image_offset = CALLOC(nr_faces, sizeof(unsigned)); - offset += nv04mt->level[l].pitch * pt->height[l]; + offset += nv04mt->level[l].pitch * u_minify(pt->height0, l); } nv04mt->total_size = offset; @@ -75,7 +74,7 @@ nv04_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt, /* Only supports 2D, non-mipmapped textures for the moment */ if (pt->target != PIPE_TEXTURE_2D || pt->last_level != 0 || - pt->depth[0] != 1) + pt->depth0 != 1) return NULL; mt = CALLOC_STRUCT(nv04_miptree); @@ -120,8 +119,8 @@ nv04_miptree_surface_new(struct pipe_screen *pscreen, struct pipe_texture *pt, return NULL; pipe_texture_reference(&ns->base.texture, pt); ns->base.format = pt->format; - ns->base.width = pt->width[level]; - ns->base.height = pt->height[level]; + ns->base.width = u_minify(pt->width0, level); + ns->base.height = u_minify(pt->height0, level); ns->base.usage = flags; pipe_reference_init(&ns->base.reference, 1); ns->base.face = face; diff --git a/src/gallium/drivers/nv04/nv04_state.c b/src/gallium/drivers/nv04/nv04_state.c index d356ebd8b36..ef3005db5fc 100644 --- a/src/gallium/drivers/nv04/nv04_state.c +++ b/src/gallium/drivers/nv04/nv04_state.c @@ -425,9 +425,9 @@ nv04_init_state_functions(struct nv04_context *nv04) nv04->pipe.delete_blend_state = nv04_blend_state_delete; nv04->pipe.create_sampler_state = nv04_sampler_state_create; - nv04->pipe.bind_sampler_states = nv04_sampler_state_bind; + nv04->pipe.bind_fragment_sampler_states = nv04_sampler_state_bind; nv04->pipe.delete_sampler_state = nv04_sampler_state_delete; - nv04->pipe.set_sampler_textures = nv04_set_sampler_texture; + nv04->pipe.set_fragment_sampler_textures = nv04_set_sampler_texture; nv04->pipe.create_rasterizer_state = nv04_rasterizer_state_create; nv04->pipe.bind_rasterizer_state = nv04_rasterizer_state_bind; diff --git a/src/gallium/drivers/nv04/nv04_transfer.c b/src/gallium/drivers/nv04/nv04_transfer.c index 6618660743d..e6456429f4e 100644 --- a/src/gallium/drivers/nv04/nv04_transfer.c +++ b/src/gallium/drivers/nv04/nv04_transfer.c @@ -2,6 +2,7 @@ #include <pipe/p_defines.h> #include <pipe/p_inlines.h> #include <util/u_memory.h> +#include <util/u_math.h> #include <nouveau/nouveau_winsys.h> #include "nv04_context.h" #include "nv04_screen.h" @@ -20,9 +21,9 @@ nv04_compatible_transfer_tex(struct pipe_texture *pt, unsigned level, memset(template, 0, sizeof(struct pipe_texture)); template->target = pt->target; template->format = pt->format; - template->width[0] = pt->width[level]; - template->height[0] = pt->height[level]; - template->depth[0] = 1; + template->width0 = u_minify(pt->width0, level); + template->height0 = u_minify(pt->height0, level); + template->depth0 = 1; template->block = pt->block; template->nblocksx[0] = pt->nblocksx[level]; template->nblocksy[0] = pt->nblocksx[level]; diff --git a/src/gallium/drivers/nv10/nv10_fragtex.c b/src/gallium/drivers/nv10/nv10_fragtex.c index 27f2f875847..906fdfeeb93 100644 --- a/src/gallium/drivers/nv10/nv10_fragtex.c +++ b/src/gallium/drivers/nv10/nv10_fragtex.c @@ -62,9 +62,9 @@ nv10_fragtex_build(struct nv10_context *nv10, int unit) txf = tf->format << 8; txf |= (pt->last_level + 1) << 16; - txf |= log2i(pt->width[0]) << 20; - txf |= log2i(pt->height[0]) << 24; - txf |= log2i(pt->depth[0]) << 28; + txf |= log2i(pt->width0) << 20; + txf |= log2i(pt->height0) << 24; + txf |= log2i(pt->depth0) << 28; txf |= 8; switch (pt->target) { @@ -89,7 +89,7 @@ nv10_fragtex_build(struct nv10_context *nv10, int unit) OUT_RING (0x40000000); /* enable */ OUT_RING (txs); OUT_RING (ps->filt | 0x2000 /* magic */); - OUT_RING ((pt->width[0] << 16) | pt->height[0]); + OUT_RING ((pt->width0 << 16) | pt->height0); OUT_RING (ps->bcol); #endif } diff --git a/src/gallium/drivers/nv10/nv10_miptree.c b/src/gallium/drivers/nv10/nv10_miptree.c index 34e3c2ebd77..b2a6c59b749 100644 --- a/src/gallium/drivers/nv10/nv10_miptree.c +++ b/src/gallium/drivers/nv10/nv10_miptree.c @@ -1,6 +1,7 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" +#include "util/u_math.h" #include "nv10_context.h" #include "nv10_screen.h" @@ -10,7 +11,7 @@ nv10_miptree_layout(struct nv10_miptree *nv10mt) { struct pipe_texture *pt = &nv10mt->base; boolean swizzled = FALSE; - uint width = pt->width[0], height = pt->height[0]; + uint width = pt->width0, height = pt->height0; uint offset = 0; int nr_faces, l, f; @@ -21,8 +22,7 @@ nv10_miptree_layout(struct nv10_miptree *nv10mt) } for (l = 0; l <= pt->last_level; l++) { - pt->width[l] = width; - pt->height[l] = height; + pt->nblocksx[l] = pf_get_nblocksx(&pt->block, width); pt->nblocksy[l] = pf_get_nblocksy(&pt->block, height); @@ -35,15 +35,15 @@ nv10_miptree_layout(struct nv10_miptree *nv10mt) nv10mt->level[l].image_offset = CALLOC(nr_faces, sizeof(unsigned)); - width = MAX2(1, width >> 1); - height = MAX2(1, height >> 1); + width = u_minify(width, 1); + height = u_minify(height, 1); } for (f = 0; f < nr_faces; f++) { for (l = 0; l <= pt->last_level; l++) { nv10mt->level[l].image_offset[f] = offset; - offset += nv10mt->level[l].pitch * pt->height[l]; + offset += nv10mt->level[l].pitch * u_minify(pt->height0, l); } } @@ -58,7 +58,7 @@ nv10_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt, /* Only supports 2D, non-mipmapped textures for the moment */ if (pt->target != PIPE_TEXTURE_2D || pt->last_level != 0 || - pt->depth[0] != 1) + pt->depth0 != 1) return NULL; mt = CALLOC_STRUCT(nv10_miptree); @@ -133,8 +133,8 @@ nv10_miptree_surface_get(struct pipe_screen *screen, struct pipe_texture *pt, return NULL; pipe_texture_reference(&ns->base.texture, pt); ns->base.format = pt->format; - ns->base.width = pt->width[level]; - ns->base.height = pt->height[level]; + ns->base.width = u_minify(pt->width0, level); + ns->base.height = u_minify(pt->height0, level); ns->base.usage = flags; pipe_reference_init(&ns->base.reference, 1); ns->base.face = face; diff --git a/src/gallium/drivers/nv10/nv10_state.c b/src/gallium/drivers/nv10/nv10_state.c index 9b38219b996..ffc6be3c401 100644 --- a/src/gallium/drivers/nv10/nv10_state.c +++ b/src/gallium/drivers/nv10/nv10_state.c @@ -553,9 +553,9 @@ nv10_init_state_functions(struct nv10_context *nv10) nv10->pipe.delete_blend_state = nv10_blend_state_delete; nv10->pipe.create_sampler_state = nv10_sampler_state_create; - nv10->pipe.bind_sampler_states = nv10_sampler_state_bind; + nv10->pipe.bind_fragment_sampler_states = nv10_sampler_state_bind; nv10->pipe.delete_sampler_state = nv10_sampler_state_delete; - nv10->pipe.set_sampler_textures = nv10_set_sampler_texture; + nv10->pipe.set_fragment_sampler_textures = nv10_set_sampler_texture; nv10->pipe.create_rasterizer_state = nv10_rasterizer_state_create; nv10->pipe.bind_rasterizer_state = nv10_rasterizer_state_bind; diff --git a/src/gallium/drivers/nv10/nv10_transfer.c b/src/gallium/drivers/nv10/nv10_transfer.c index 8feb85e4bda..ec54297ab01 100644 --- a/src/gallium/drivers/nv10/nv10_transfer.c +++ b/src/gallium/drivers/nv10/nv10_transfer.c @@ -2,6 +2,7 @@ #include <pipe/p_defines.h> #include <pipe/p_inlines.h> #include <util/u_memory.h> +#include <util/u_math.h> #include <nouveau/nouveau_winsys.h> #include "nv10_context.h" #include "nv10_screen.h" @@ -20,9 +21,9 @@ nv10_compatible_transfer_tex(struct pipe_texture *pt, unsigned level, memset(template, 0, sizeof(struct pipe_texture)); template->target = pt->target; template->format = pt->format; - template->width[0] = pt->width[level]; - template->height[0] = pt->height[level]; - template->depth[0] = 1; + template->width0 = u_minify(pt->width0, level); + template->height0 = u_minify(pt->height0, level); + template->depth0 = 1; template->block = pt->block; template->nblocksx[0] = pt->nblocksx[level]; template->nblocksy[0] = pt->nblocksx[level]; diff --git a/src/gallium/drivers/nv20/nv20_fragtex.c b/src/gallium/drivers/nv20/nv20_fragtex.c index 495a7be9127..2db4a4015a6 100644 --- a/src/gallium/drivers/nv20/nv20_fragtex.c +++ b/src/gallium/drivers/nv20/nv20_fragtex.c @@ -62,9 +62,9 @@ nv20_fragtex_build(struct nv20_context *nv20, int unit) txf = tf->format << 8; txf |= (pt->last_level + 1) << 16; - txf |= log2i(pt->width[0]) << 20; - txf |= log2i(pt->height[0]) << 24; - txf |= log2i(pt->depth[0]) << 28; + txf |= log2i(pt->width0) << 20; + txf |= log2i(pt->height0) << 24; + txf |= log2i(pt->depth0) << 28; txf |= 8; switch (pt->target) { @@ -89,7 +89,7 @@ nv20_fragtex_build(struct nv20_context *nv20, int unit) OUT_RING (0x40000000); /* enable */ OUT_RING (txs); OUT_RING (ps->filt | 0x2000 /* magic */); - OUT_RING ((pt->width[0] << 16) | pt->height[0]); + OUT_RING ((pt->width0 << 16) | pt->height0); OUT_RING (ps->bcol); #endif } diff --git a/src/gallium/drivers/nv20/nv20_miptree.c b/src/gallium/drivers/nv20/nv20_miptree.c index 185fbf53e0f..554e28e47dd 100644 --- a/src/gallium/drivers/nv20/nv20_miptree.c +++ b/src/gallium/drivers/nv20/nv20_miptree.c @@ -1,6 +1,7 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" +#include "util/u_math.h" #include "nv20_context.h" #include "nv20_screen.h" @@ -9,7 +10,7 @@ static void nv20_miptree_layout(struct nv20_miptree *nv20mt) { struct pipe_texture *pt = &nv20mt->base; - uint width = pt->width[0], height = pt->height[0]; + uint width = pt->width0, height = pt->height0; uint offset = 0; int nr_faces, l, f; uint wide_pitch = pt->tex_usage & (PIPE_TEXTURE_USAGE_SAMPLER | @@ -25,21 +26,19 @@ nv20_miptree_layout(struct nv20_miptree *nv20mt) } for (l = 0; l <= pt->last_level; l++) { - pt->width[l] = width; - pt->height[l] = height; pt->nblocksx[l] = pf_get_nblocksx(&pt->block, width); pt->nblocksy[l] = pf_get_nblocksy(&pt->block, height); if (wide_pitch && (pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) - nv20mt->level[l].pitch = align(pt->width[0] * pt->block.size, 64); + nv20mt->level[l].pitch = align(pt->width0 * pt->block.size, 64); else - nv20mt->level[l].pitch = pt->width[l] * pt->block.size; + nv20mt->level[l].pitch = u_minify(pt->width0, l) * pt->block.size; nv20mt->level[l].image_offset = CALLOC(nr_faces, sizeof(unsigned)); - width = MAX2(1, width >> 1); - height = MAX2(1, height >> 1); + width = u_minify(width, 1); + height = u_minify(height, 1); } for (f = 0; f < nr_faces; f++) { @@ -47,14 +46,14 @@ nv20_miptree_layout(struct nv20_miptree *nv20mt) nv20mt->level[l].image_offset[f] = offset; if (!(pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR) && - pt->width[l + 1] > 1 && pt->height[l + 1] > 1) - offset += align(nv20mt->level[l].pitch * pt->height[l], 64); + u_minify(pt->width0, l + 1) > 1 && u_minify(pt->height0, l + 1) > 1) + offset += align(nv20mt->level[l].pitch * u_minify(pt->height0, l), 64); else - offset += nv20mt->level[l].pitch * pt->height[l]; + offset += nv20mt->level[l].pitch * u_minify(pt->height0, l); } nv20mt->level[l].image_offset[f] = offset; - offset += nv20mt->level[l].pitch * pt->height[l]; + offset += nv20mt->level[l].pitch * u_minify(pt->height0, l); } nv20mt->total_size = offset; @@ -68,7 +67,7 @@ nv20_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt, /* Only supports 2D, non-mipmapped textures for the moment */ if (pt->target != PIPE_TEXTURE_2D || pt->last_level != 0 || - pt->depth[0] != 1) + pt->depth0 != 1) return NULL; mt = CALLOC_STRUCT(nv20_miptree); @@ -100,8 +99,8 @@ nv20_miptree_create(struct pipe_screen *screen, const struct pipe_texture *pt) mt->base.screen = screen; /* Swizzled textures must be POT */ - if (pt->width[0] & (pt->width[0] - 1) || - pt->height[0] & (pt->height[0] - 1)) + if (pt->width0 & (pt->width0 - 1) || + pt->height0 & (pt->height0 - 1)) mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR; else if (pt->tex_usage & (PIPE_TEXTURE_USAGE_PRIMARY | @@ -167,8 +166,8 @@ nv20_miptree_surface_get(struct pipe_screen *screen, struct pipe_texture *pt, return NULL; pipe_texture_reference(&ns->base.texture, pt); ns->base.format = pt->format; - ns->base.width = pt->width[level]; - ns->base.height = pt->height[level]; + ns->base.width = u_minify(pt->width0, level); + ns->base.height = u_minify(pt->height0, level); ns->base.usage = flags; pipe_reference_init(&ns->base.reference, 1); ns->base.face = face; diff --git a/src/gallium/drivers/nv20/nv20_state.c b/src/gallium/drivers/nv20/nv20_state.c index ed4084980f2..3a82e63423d 100644 --- a/src/gallium/drivers/nv20/nv20_state.c +++ b/src/gallium/drivers/nv20/nv20_state.c @@ -546,9 +546,9 @@ nv20_init_state_functions(struct nv20_context *nv20) nv20->pipe.delete_blend_state = nv20_blend_state_delete; nv20->pipe.create_sampler_state = nv20_sampler_state_create; - nv20->pipe.bind_sampler_states = nv20_sampler_state_bind; + nv20->pipe.bind_fragment_sampler_states = nv20_sampler_state_bind; nv20->pipe.delete_sampler_state = nv20_sampler_state_delete; - nv20->pipe.set_sampler_textures = nv20_set_sampler_texture; + nv20->pipe.set_fragment_sampler_textures = nv20_set_sampler_texture; nv20->pipe.create_rasterizer_state = nv20_rasterizer_state_create; nv20->pipe.bind_rasterizer_state = nv20_rasterizer_state_bind; diff --git a/src/gallium/drivers/nv20/nv20_transfer.c b/src/gallium/drivers/nv20/nv20_transfer.c index 81b4f1a9177..87b5c14a3c2 100644 --- a/src/gallium/drivers/nv20/nv20_transfer.c +++ b/src/gallium/drivers/nv20/nv20_transfer.c @@ -2,6 +2,7 @@ #include <pipe/p_defines.h> #include <pipe/p_inlines.h> #include <util/u_memory.h> +#include <util/u_math.h> #include <nouveau/nouveau_winsys.h> #include "nv20_context.h" #include "nv20_screen.h" @@ -20,9 +21,9 @@ nv20_compatible_transfer_tex(struct pipe_texture *pt, unsigned level, memset(template, 0, sizeof(struct pipe_texture)); template->target = pt->target; template->format = pt->format; - template->width[0] = pt->width[level]; - template->height[0] = pt->height[level]; - template->depth[0] = 1; + template->width0 = u_minify(pt->width0, level); + template->height0 = u_minify(pt->height0, level); + template->depth0 = 1; template->block = pt->block; template->nblocksx[0] = pt->nblocksx[level]; template->nblocksy[0] = pt->nblocksx[level]; diff --git a/src/gallium/drivers/nv30/nv30_fragtex.c b/src/gallium/drivers/nv30/nv30_fragtex.c index dca760cae62..b3293ee700d 100644 --- a/src/gallium/drivers/nv30/nv30_fragtex.c +++ b/src/gallium/drivers/nv30/nv30_fragtex.c @@ -74,9 +74,9 @@ nv30_fragtex_build(struct nv30_context *nv30, int unit) txf = tf->format; txf |= ((pt->last_level>0) ? NV34TCL_TX_FORMAT_MIPMAP : 0); - txf |= log2i(pt->width[0]) << NV34TCL_TX_FORMAT_BASE_SIZE_U_SHIFT; - txf |= log2i(pt->height[0]) << NV34TCL_TX_FORMAT_BASE_SIZE_V_SHIFT; - txf |= log2i(pt->depth[0]) << NV34TCL_TX_FORMAT_BASE_SIZE_W_SHIFT; + txf |= log2i(pt->width0) << NV34TCL_TX_FORMAT_BASE_SIZE_U_SHIFT; + txf |= log2i(pt->height0) << NV34TCL_TX_FORMAT_BASE_SIZE_V_SHIFT; + txf |= log2i(pt->depth0) << NV34TCL_TX_FORMAT_BASE_SIZE_W_SHIFT; txf |= NV34TCL_TX_FORMAT_NO_BORDER | 0x10000; switch (pt->target) { @@ -115,8 +115,8 @@ nv30_fragtex_build(struct nv30_context *nv30, int unit) so_data (so, NV34TCL_TX_ENABLE_ENABLE | ps->en); so_data (so, txs); so_data (so, ps->filt | 0x2000 /*voodoo*/); - so_data (so, (pt->width[0] << NV34TCL_TX_NPOT_SIZE_W_SHIFT) | - pt->height[0]); + so_data (so, (pt->width0 << NV34TCL_TX_NPOT_SIZE_W_SHIFT) | + pt->height0); so_data (so, ps->bcol); return so; diff --git a/src/gallium/drivers/nv30/nv30_miptree.c b/src/gallium/drivers/nv30/nv30_miptree.c index 280696d4503..b4c306d1272 100644 --- a/src/gallium/drivers/nv30/nv30_miptree.c +++ b/src/gallium/drivers/nv30/nv30_miptree.c @@ -1,6 +1,7 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" +#include "util/u_math.h" #include "nv30_context.h" @@ -8,7 +9,7 @@ static void nv30_miptree_layout(struct nv30_miptree *nv30mt) { struct pipe_texture *pt = &nv30mt->base; - uint width = pt->width[0], height = pt->height[0], depth = pt->depth[0]; + uint width = pt->width0, height = pt->height0, depth = pt->depth0; uint offset = 0; int nr_faces, l, f; uint wide_pitch = pt->tex_usage & (PIPE_TEXTURE_USAGE_SAMPLER | @@ -21,29 +22,26 @@ nv30_miptree_layout(struct nv30_miptree *nv30mt) nr_faces = 6; } else if (pt->target == PIPE_TEXTURE_3D) { - nr_faces = pt->depth[0]; + nr_faces = pt->depth0; } else { nr_faces = 1; } for (l = 0; l <= pt->last_level; l++) { - pt->width[l] = width; - pt->height[l] = height; - pt->depth[l] = depth; pt->nblocksx[l] = pf_get_nblocksx(&pt->block, width); pt->nblocksy[l] = pf_get_nblocksy(&pt->block, height); if (wide_pitch && (pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) - nv30mt->level[l].pitch = align(pt->width[0] * pt->block.size, 64); + nv30mt->level[l].pitch = align(pt->width0 * pt->block.size, 64); else - nv30mt->level[l].pitch = pt->width[l] * pt->block.size; + nv30mt->level[l].pitch = u_minify(pt->width0, l) * pt->block.size; nv30mt->level[l].image_offset = CALLOC(nr_faces, sizeof(unsigned)); - width = MAX2(1, width >> 1); - height = MAX2(1, height >> 1); - depth = MAX2(1, depth >> 1); + width = u_minify(width, 1); + height = u_minify(height, 1); + depth = u_minify(depth, 1); } for (f = 0; f < nr_faces; f++) { @@ -51,14 +49,14 @@ nv30_miptree_layout(struct nv30_miptree *nv30mt) nv30mt->level[l].image_offset[f] = offset; if (!(pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR) && - pt->width[l + 1] > 1 && pt->height[l + 1] > 1) - offset += align(nv30mt->level[l].pitch * pt->height[l], 64); + u_minify(pt->width0, l + 1) > 1 && u_minify(pt->height0, l + 1) > 1) + offset += align(nv30mt->level[l].pitch * u_minify(pt->height0, l), 64); else - offset += nv30mt->level[l].pitch * pt->height[l]; + offset += nv30mt->level[l].pitch * u_minify(pt->height0, l); } nv30mt->level[l].image_offset[f] = offset; - offset += nv30mt->level[l].pitch * pt->height[l]; + offset += nv30mt->level[l].pitch * u_minify(pt->height0, l); } nv30mt->total_size = offset; @@ -79,8 +77,8 @@ nv30_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *pt) mt->base.screen = pscreen; /* Swizzled textures must be POT */ - if (pt->width[0] & (pt->width[0] - 1) || - pt->height[0] & (pt->height[0] - 1)) + if (pt->width0 & (pt->width0 - 1) || + pt->height0 & (pt->height0 - 1)) mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR; else if (pt->tex_usage & (PIPE_TEXTURE_USAGE_PRIMARY | @@ -134,7 +132,7 @@ nv30_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt, /* Only supports 2D, non-mipmapped textures for the moment */ if (pt->target != PIPE_TEXTURE_2D || pt->last_level != 0 || - pt->depth[0] != 1) + pt->depth0 != 1) return NULL; mt = CALLOC_STRUCT(nv30_miptree); @@ -182,8 +180,8 @@ nv30_miptree_surface_new(struct pipe_screen *pscreen, struct pipe_texture *pt, return NULL; pipe_texture_reference(&ns->base.texture, pt); ns->base.format = pt->format; - ns->base.width = pt->width[level]; - ns->base.height = pt->height[level]; + ns->base.width = u_minify(pt->width0, level); + ns->base.height = u_minify(pt->height0, level); ns->base.usage = flags; pipe_reference_init(&ns->base.reference, 1); ns->base.face = face; diff --git a/src/gallium/drivers/nv30/nv30_state.c b/src/gallium/drivers/nv30/nv30_state.c index b91e972c123..3f802d92416 100644 --- a/src/gallium/drivers/nv30/nv30_state.c +++ b/src/gallium/drivers/nv30/nv30_state.c @@ -690,9 +690,9 @@ nv30_init_state_functions(struct nv30_context *nv30) nv30->pipe.delete_blend_state = nv30_blend_state_delete; nv30->pipe.create_sampler_state = nv30_sampler_state_create; - nv30->pipe.bind_sampler_states = nv30_sampler_state_bind; + nv30->pipe.bind_fragment_sampler_states = nv30_sampler_state_bind; nv30->pipe.delete_sampler_state = nv30_sampler_state_delete; - nv30->pipe.set_sampler_textures = nv30_set_sampler_texture; + nv30->pipe.set_fragment_sampler_textures = nv30_set_sampler_texture; nv30->pipe.create_rasterizer_state = nv30_rasterizer_state_create; nv30->pipe.bind_rasterizer_state = nv30_rasterizer_state_bind; diff --git a/src/gallium/drivers/nv30/nv30_transfer.c b/src/gallium/drivers/nv30/nv30_transfer.c index 98011decf7c..5e429b4d85c 100644 --- a/src/gallium/drivers/nv30/nv30_transfer.c +++ b/src/gallium/drivers/nv30/nv30_transfer.c @@ -2,6 +2,7 @@ #include <pipe/p_defines.h> #include <pipe/p_inlines.h> #include <util/u_memory.h> +#include <util/u_math.h> #include <nouveau/nouveau_winsys.h> #include "nv30_context.h" #include "nv30_screen.h" @@ -20,9 +21,9 @@ nv30_compatible_transfer_tex(struct pipe_texture *pt, unsigned level, memset(template, 0, sizeof(struct pipe_texture)); template->target = pt->target; template->format = pt->format; - template->width[0] = pt->width[level]; - template->height[0] = pt->height[level]; - template->depth[0] = 1; + template->width0 = u_minify(pt->width0, level); + template->height0 = u_minify(pt->height0, level); + template->depth0 = 1; template->block = pt->block; template->nblocksx[0] = pt->nblocksx[level]; template->nblocksy[0] = pt->nblocksx[level]; diff --git a/src/gallium/drivers/nv40/nv40_fragtex.c b/src/gallium/drivers/nv40/nv40_fragtex.c index e2ec57564d1..44abc845969 100644 --- a/src/gallium/drivers/nv40/nv40_fragtex.c +++ b/src/gallium/drivers/nv40/nv40_fragtex.c @@ -117,11 +117,11 @@ nv40_fragtex_build(struct nv40_context *nv40, int unit) so_data (so, NV40TCL_TEX_ENABLE_ENABLE | ps->en); so_data (so, txs); so_data (so, ps->filt | tf->sign | 0x2000 /*voodoo*/); - so_data (so, (pt->width[0] << NV40TCL_TEX_SIZE0_W_SHIFT) | - pt->height[0]); + so_data (so, (pt->width0 << NV40TCL_TEX_SIZE0_W_SHIFT) | + pt->height0); so_data (so, ps->bcol); so_method(so, nv40->screen->curie, NV40TCL_TEX_SIZE1(unit), 1); - so_data (so, (pt->depth[0] << NV40TCL_TEX_SIZE1_DEPTH_SHIFT) | txp); + so_data (so, (pt->depth0 << NV40TCL_TEX_SIZE1_DEPTH_SHIFT) | txp); return so; } diff --git a/src/gallium/drivers/nv40/nv40_miptree.c b/src/gallium/drivers/nv40/nv40_miptree.c index 465dd3b0693..f73bedff6d8 100644 --- a/src/gallium/drivers/nv40/nv40_miptree.c +++ b/src/gallium/drivers/nv40/nv40_miptree.c @@ -1,6 +1,7 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" +#include "util/u_math.h" #include "nv40_context.h" @@ -8,7 +9,7 @@ static void nv40_miptree_layout(struct nv40_miptree *mt) { struct pipe_texture *pt = &mt->base; - uint width = pt->width[0], height = pt->height[0], depth = pt->depth[0]; + uint width = pt->width0, height = pt->height0, depth = pt->depth0; uint offset = 0; int nr_faces, l, f; uint wide_pitch = pt->tex_usage & (PIPE_TEXTURE_USAGE_SAMPLER | @@ -21,29 +22,26 @@ nv40_miptree_layout(struct nv40_miptree *mt) nr_faces = 6; } else if (pt->target == PIPE_TEXTURE_3D) { - nr_faces = pt->depth[0]; + nr_faces = pt->depth0; } else { nr_faces = 1; } for (l = 0; l <= pt->last_level; l++) { - pt->width[l] = width; - pt->height[l] = height; - pt->depth[l] = depth; pt->nblocksx[l] = pf_get_nblocksx(&pt->block, width); pt->nblocksy[l] = pf_get_nblocksy(&pt->block, height); if (wide_pitch && (pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) - mt->level[l].pitch = align(pt->width[0] * pt->block.size, 64); + mt->level[l].pitch = align(pt->width0 * pt->block.size, 64); else - mt->level[l].pitch = pt->width[l] * pt->block.size; + mt->level[l].pitch = u_minify(pt->width0, l) * pt->block.size; mt->level[l].image_offset = CALLOC(nr_faces, sizeof(unsigned)); - width = MAX2(1, width >> 1); - height = MAX2(1, height >> 1); - depth = MAX2(1, depth >> 1); + width = u_minify(width, 1); + height = u_minify(height, 1); + depth = u_minify(depth, 1); } for (f = 0; f < nr_faces; f++) { @@ -51,14 +49,14 @@ nv40_miptree_layout(struct nv40_miptree *mt) mt->level[l].image_offset[f] = offset; if (!(pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR) && - pt->width[l + 1] > 1 && pt->height[l + 1] > 1) - offset += align(mt->level[l].pitch * pt->height[l], 64); + u_minify(pt->width0, l + 1) > 1 && u_minify(pt->height0, l + 1) > 1) + offset += align(mt->level[l].pitch * u_minify(pt->height0, l), 64); else - offset += mt->level[l].pitch * pt->height[l]; + offset += mt->level[l].pitch * u_minify(pt->height0, l); } mt->level[l].image_offset[f] = offset; - offset += mt->level[l].pitch * pt->height[l]; + offset += mt->level[l].pitch * u_minify(pt->height0, l); } mt->total_size = offset; @@ -79,8 +77,8 @@ nv40_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *pt) mt->base.screen = pscreen; /* Swizzled textures must be POT */ - if (pt->width[0] & (pt->width[0] - 1) || - pt->height[0] & (pt->height[0] - 1)) + if (pt->width0 & (pt->width0 - 1) || + pt->height0 & (pt->height0 - 1)) mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR; else if (pt->tex_usage & (PIPE_TEXTURE_USAGE_PRIMARY | @@ -128,7 +126,7 @@ nv40_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt, /* Only supports 2D, non-mipmapped textures for the moment */ if (pt->target != PIPE_TEXTURE_2D || pt->last_level != 0 || - pt->depth[0] != 1) + pt->depth0 != 1) return NULL; mt = CALLOC_STRUCT(nv40_miptree); @@ -176,8 +174,8 @@ nv40_miptree_surface_new(struct pipe_screen *pscreen, struct pipe_texture *pt, return NULL; pipe_texture_reference(&ns->base.texture, pt); ns->base.format = pt->format; - ns->base.width = pt->width[level]; - ns->base.height = pt->height[level]; + ns->base.width = u_minify(pt->width0, level); + ns->base.height = u_minify(pt->height0, level); ns->base.usage = flags; pipe_reference_init(&ns->base.reference, 1); ns->base.face = face; diff --git a/src/gallium/drivers/nv40/nv40_state.c b/src/gallium/drivers/nv40/nv40_state.c index c3ee4d23453..bc34e32a4b9 100644 --- a/src/gallium/drivers/nv40/nv40_state.c +++ b/src/gallium/drivers/nv40/nv40_state.c @@ -705,9 +705,9 @@ nv40_init_state_functions(struct nv40_context *nv40) nv40->pipe.delete_blend_state = nv40_blend_state_delete; nv40->pipe.create_sampler_state = nv40_sampler_state_create; - nv40->pipe.bind_sampler_states = nv40_sampler_state_bind; + nv40->pipe.bind_fragment_sampler_states = nv40_sampler_state_bind; nv40->pipe.delete_sampler_state = nv40_sampler_state_delete; - nv40->pipe.set_sampler_textures = nv40_set_sampler_texture; + nv40->pipe.set_fragment_sampler_textures = nv40_set_sampler_texture; nv40->pipe.create_rasterizer_state = nv40_rasterizer_state_create; nv40->pipe.bind_rasterizer_state = nv40_rasterizer_state_bind; diff --git a/src/gallium/drivers/nv40/nv40_transfer.c b/src/gallium/drivers/nv40/nv40_transfer.c index 92caee6f382..36e253c96f9 100644 --- a/src/gallium/drivers/nv40/nv40_transfer.c +++ b/src/gallium/drivers/nv40/nv40_transfer.c @@ -2,6 +2,7 @@ #include <pipe/p_defines.h> #include <pipe/p_inlines.h> #include <util/u_memory.h> +#include <util/u_math.h> #include <nouveau/nouveau_winsys.h> #include "nv40_context.h" #include "nv40_screen.h" @@ -20,9 +21,9 @@ nv40_compatible_transfer_tex(struct pipe_texture *pt, unsigned level, memset(template, 0, sizeof(struct pipe_texture)); template->target = pt->target; template->format = pt->format; - template->width[0] = pt->width[level]; - template->height[0] = pt->height[level]; - template->depth[0] = 1; + template->width0 = u_minify(pt->width0, level); + template->height0 = u_minify(pt->height0, level); + template->depth0 = 1; template->block = pt->block; template->nblocksx[0] = pt->nblocksx[level]; template->nblocksy[0] = pt->nblocksx[level]; diff --git a/src/gallium/drivers/nv50/nv50_miptree.c b/src/gallium/drivers/nv50/nv50_miptree.c index 9c20c5cc282..3d58746793f 100644 --- a/src/gallium/drivers/nv50/nv50_miptree.c +++ b/src/gallium/drivers/nv50/nv50_miptree.c @@ -61,8 +61,8 @@ nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp) struct nouveau_device *dev = nouveau_screen(pscreen)->device; struct nv50_miptree *mt = CALLOC_STRUCT(nv50_miptree); struct pipe_texture *pt = &mt->base.base; - unsigned width = tmp->width[0], height = tmp->height[0]; - unsigned depth = tmp->depth[0], image_alignment; + unsigned width = tmp->width0, height = tmp->height0; + unsigned depth = tmp->depth0, image_alignment; uint32_t tile_flags; int ret, i, l; @@ -92,9 +92,6 @@ nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp) for (l = 0; l <= pt->last_level; l++) { struct nv50_miptree_level *lvl = &mt->level[l]; - pt->width[l] = width; - pt->height[l] = height; - pt->depth[l] = depth; pt->nblocksx[l] = pf_get_nblocksx(&pt->block, width); pt->nblocksy[l] = pf_get_nblocksy(&pt->block, height); @@ -102,9 +99,9 @@ nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp) lvl->pitch = align(pt->nblocksx[l] * pt->block.size, 64); lvl->tile_mode = get_tile_mode(pt->nblocksy[l], depth); - width = MAX2(1, width >> 1); - height = MAX2(1, height >> 1); - depth = MAX2(1, depth >> 1); + width = u_minify(width, 1); + height = u_minify(height, 1); + depth = u_minify(depth, 1); } image_alignment = get_tile_height(mt->level[0].tile_mode) * 64; @@ -122,7 +119,7 @@ nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp) size = lvl->pitch; size *= align(pt->nblocksy[l], tile_h); - size *= align(pt->depth[l], tile_d); + size *= align(u_minify(pt->depth0, l), tile_d); lvl->image_offset[i] = mt->total_size; @@ -151,7 +148,7 @@ nv50_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt, /* Only supports 2D, non-mipmapped textures for the moment */ if (pt->target != PIPE_TEXTURE_2D || pt->last_level != 0 || - pt->depth[0] != 1) + pt->depth0 != 1) return NULL; mt = CALLOC_STRUCT(nv50_miptree); @@ -202,8 +199,8 @@ nv50_miptree_surface_new(struct pipe_screen *pscreen, struct pipe_texture *pt, return NULL; pipe_texture_reference(&ps->texture, pt); ps->format = pt->format; - ps->width = pt->width[level]; - ps->height = pt->height[level]; + ps->width = u_minify(pt->width0, level); + ps->height = u_minify(pt->height0, level); ps->usage = flags; pipe_reference_init(&ps->reference, 1); ps->face = face; diff --git a/src/gallium/drivers/nv50/nv50_program.c b/src/gallium/drivers/nv50/nv50_program.c index bf50982dd16..bad0ace7e55 100644 --- a/src/gallium/drivers/nv50/nv50_program.c +++ b/src/gallium/drivers/nv50/nv50_program.c @@ -1440,19 +1440,25 @@ emit_ddx(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src) static void emit_ddy(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src) { + struct nv50_reg *r = src; struct nv50_program_exec *e = exec(pc); assert(src->type == P_TEMP); - if (!(src->mod & NV50_MOD_NEG)) /* ! double negation */ - emit_neg(pc, src, src); + if (!(src->mod & NV50_MOD_NEG)) { /* ! double negation */ + r = alloc_temp(pc, NULL); + emit_neg(pc, r, src); + } e->inst[0] = 0xc0150000; e->inst[1] = 0x8a400000; set_long(pc, e); set_dst(pc, dst, e); - set_src_0(pc, src, e); - set_src_2(pc, src, e); + set_src_0(pc, r, e); + set_src_2(pc, r, e); + + if (r != src) + free_temp(pc, r); emit(pc, e); } @@ -2637,7 +2643,7 @@ nv50_program_tx_prep(struct nv50_pc *pc) for (i = 0, rid = 0; i < pc->result_nr; ++i) { p->cfg.io[i].hw = rid; - p->cfg.io[i].id_vp = i; + p->cfg.io[i].id = i; for (c = 0; c < 4; ++c) { int n = i * 4 + c; @@ -2669,14 +2675,12 @@ nv50_program_tx_prep(struct nv50_pc *pc) * the lower hardware IDs, so sort them: */ for (i = 0; i < pc->attr_nr; i++) { - if (pc->interp_mode[i] == INTERP_FLAT) { - p->cfg.io[m].id_vp = i + base; - p->cfg.io[m++].id_fp = i; - } else { + if (pc->interp_mode[i] == INTERP_FLAT) + p->cfg.io[m++].id = i; + else { if (!(pc->interp_mode[i] & INTERP_PERSPECTIVE)) p->cfg.io[n].linear = TRUE; - p->cfg.io[n].id_vp = i + base; - p->cfg.io[n++].id_fp = i; + p->cfg.io[n++].id = i; } } @@ -2688,7 +2692,7 @@ nv50_program_tx_prep(struct nv50_pc *pc) for (n = 0; n < pc->attr_nr; ++n) { p->cfg.io[n].hw = rid = aid; - i = p->cfg.io[n].id_fp; + i = p->cfg.io[n].id; if (p->info.input_semantic_name[n] == TGSI_SEMANTIC_FACE) { @@ -2728,8 +2732,8 @@ nv50_program_tx_prep(struct nv50_pc *pc) for (i = 0; i < pc->attr_nr; i++) { ubyte si, sn; - sn = p->info.input_semantic_name[p->cfg.io[i].id_fp]; - si = p->info.input_semantic_index[p->cfg.io[i].id_fp]; + sn = p->info.input_semantic_name[p->cfg.io[i].id]; + si = p->info.input_semantic_index[p->cfg.io[i].id]; if (sn == TGSI_SEMANTIC_COLOR) { p->cfg.two_side[si] = p->cfg.io[i]; @@ -2914,7 +2918,7 @@ nv50_fp_move_results(struct nv50_pc *pc) static void nv50_program_fixup_insns(struct nv50_pc *pc) { - struct nv50_program_exec *e, *prev = NULL, **bra_list; + struct nv50_program_exec *e, **bra_list; unsigned i, n, pos; bra_list = CALLOC(pc->p->exec_size, sizeof(struct nv50_program_exec *)); @@ -2926,6 +2930,16 @@ nv50_program_fixup_insns(struct nv50_pc *pc) if (e->param.index >= 0 && !e->param.mask) bra_list[n++] = e; + /* last instruction must be long so it can have the exit bit set */ + if (!is_long(pc->p->exec_tail)) + convert_to_long(pc, pc->p->exec_tail); + /* set exit bit */ + pc->p->exec_tail->inst[1] |= 1; + + /* !immd on exit insn simultaneously means !join */ + assert(!is_immd(pc->p->exec_head)); + assert(!is_immd(pc->p->exec_tail)); + /* Make sure we don't have any single 32 bit instructions. */ for (e = pc->p->exec_head, pos = 0; e; e = e->next) { pos += is_long(e) ? 2 : 1; @@ -2937,22 +2951,7 @@ nv50_program_fixup_insns(struct nv50_pc *pc) convert_to_long(pc, e); ++pos; } - if (e->next) - prev = e; - } - - assert(!is_immd(pc->p->exec_head)); - assert(!is_immd(pc->p->exec_tail)); - - /* last instruction must be long so it can have the end bit set */ - if (!is_long(pc->p->exec_tail)) { - convert_to_long(pc, pc->p->exec_tail); - if (prev) - convert_to_long(pc, prev); } - assert(!(pc->p->exec_tail->inst[1] & 2)); - /* set the end-bit */ - pc->p->exec_tail->inst[1] |= 1; FREE(bra_list); } @@ -3236,15 +3235,15 @@ nv50_pntc_replace(struct nv50_context *nv50, uint32_t pntc[8], unsigned base) struct nv50_program *vp = nv50->vertprog; unsigned i, c, m = base; - /* XXX: This can't work correctly in all cases yet, we either - * have to create TGSI_SEMANTIC_PNTC or sprite_coord_mode has - * to be per FP input instead of per VP output + /* XXX: this might not work correctly in all cases yet - we'll + * just assume that an FP generic input that is not written in + * the VP is PointCoord. */ memset(pntc, 0, 8 * sizeof(uint32_t)); for (i = 0; i < fp->cfg.io_nr; i++) { uint8_t sn, si; - uint8_t j = fp->cfg.io[i].id_vp, k = fp->cfg.io[i].id_fp; + uint8_t j, k = fp->cfg.io[i].id; unsigned n = popcnt4(fp->cfg.io[i].mask); if (fp->info.input_semantic_name[k] != TGSI_SEMANTIC_GENERIC) { @@ -3252,10 +3251,16 @@ nv50_pntc_replace(struct nv50_context *nv50, uint32_t pntc[8], unsigned base) continue; } - sn = vp->info.input_semantic_name[j]; - si = vp->info.input_semantic_index[j]; + for (j = 0; j < vp->info.num_outputs; ++j) { + sn = vp->info.output_semantic_name[j]; + si = vp->info.output_semantic_index[j]; - if (j < fp->cfg.io_nr && sn == TGSI_SEMANTIC_GENERIC) { + if (sn == fp->info.input_semantic_name[k] && + si == fp->info.input_semantic_index[k]) + break; + } + + if (j < vp->info.num_outputs) { ubyte mode = nv50->rasterizer->pipe.sprite_coord_mode[si]; @@ -3343,20 +3348,24 @@ nv50_linkage_validate(struct nv50_context *nv50) reg[0] += m - 4; /* adjust FFC0 id */ reg[4] |= m << 8; /* set mid where 'normal' FP inputs start */ - i = 0; - if (fp->info.input_semantic_name[0] == TGSI_SEMANTIC_POSITION) - i = 1; - for (; i < fp->cfg.io_nr; i++) { - ubyte sn = fp->info.input_semantic_name[fp->cfg.io[i].id_fp]; - ubyte si = fp->info.input_semantic_index[fp->cfg.io[i].id_fp]; - - n = fp->cfg.io[i].id_vp; - if (n >= vp->cfg.io_nr || - vp->info.output_semantic_name[n] != sn || - vp->info.output_semantic_index[n] != si) - vpo = &dummy; - else - vpo = &vp->cfg.io[n]; + for (i = 0; i < fp->cfg.io_nr; i++) { + ubyte sn = fp->info.input_semantic_name[fp->cfg.io[i].id]; + ubyte si = fp->info.input_semantic_index[fp->cfg.io[i].id]; + + /* position must be mapped first */ + assert(i == 0 || sn != TGSI_SEMANTIC_POSITION); + + /* maybe even remove these from cfg.io */ + if (sn == TGSI_SEMANTIC_POSITION || sn == TGSI_SEMANTIC_FACE) + continue; + + /* VP outputs and vp->cfg.io are in the same order */ + for (n = 0; n < vp->info.num_outputs; ++n) { + if (vp->info.output_semantic_name[n] == sn && + vp->info.output_semantic_index[n] == si) + break; + } + vpo = (n < vp->info.num_outputs) ? &vp->cfg.io[n] : &dummy; m = nv50_sreg4_map(map, m, lin, &fp->cfg.io[i], vpo); } diff --git a/src/gallium/drivers/nv50/nv50_program.h b/src/gallium/drivers/nv50/nv50_program.h index d78dee083f1..255c7c737ef 100644 --- a/src/gallium/drivers/nv50/nv50_program.h +++ b/src/gallium/drivers/nv50/nv50_program.h @@ -17,8 +17,7 @@ struct nv50_program_exec { struct nv50_sreg4 { uint8_t hw; - uint8_t id_vp; - uint8_t id_fp; + uint8_t id; /* tgsi index, nv50 needs them sorted: flat ones last */ uint8_t mask; boolean linear; diff --git a/src/gallium/drivers/nv50/nv50_state.c b/src/gallium/drivers/nv50/nv50_state.c index ffaa5e29d1c..07318f23947 100644 --- a/src/gallium/drivers/nv50/nv50_state.c +++ b/src/gallium/drivers/nv50/nv50_state.c @@ -648,9 +648,9 @@ nv50_init_state_functions(struct nv50_context *nv50) nv50->pipe.delete_blend_state = nv50_blend_state_delete; nv50->pipe.create_sampler_state = nv50_sampler_state_create; - nv50->pipe.bind_sampler_states = nv50_sampler_state_bind; + nv50->pipe.bind_fragment_sampler_states = nv50_sampler_state_bind; nv50->pipe.delete_sampler_state = nv50_sampler_state_delete; - nv50->pipe.set_sampler_textures = nv50_set_sampler_texture; + nv50->pipe.set_fragment_sampler_textures = nv50_set_sampler_texture; nv50->pipe.create_rasterizer_state = nv50_rasterizer_state_create; nv50->pipe.bind_rasterizer_state = nv50_rasterizer_state_bind; diff --git a/src/gallium/drivers/nv50/nv50_state_validate.c b/src/gallium/drivers/nv50/nv50_state_validate.c index 799d2758fee..c871acaab8d 100644 --- a/src/gallium/drivers/nv50/nv50_state_validate.c +++ b/src/gallium/drivers/nv50/nv50_state_validate.c @@ -201,7 +201,8 @@ nv50_state_emit(struct nv50_context *nv50) so_emit(chan, nv50->state.vertprog); if (nv50->state.dirty & NV50_NEW_FRAGPROG) so_emit(chan, nv50->state.fragprog); - if (nv50->state.dirty & (NV50_NEW_FRAGPROG | NV50_NEW_VERTPROG)) + if (nv50->state.dirty & (NV50_NEW_FRAGPROG | NV50_NEW_VERTPROG | + NV50_NEW_RASTERIZER)) so_emit(chan, nv50->state.programs); if (nv50->state.dirty & NV50_NEW_RASTERIZER) so_emit(chan, nv50->state.rast); @@ -264,7 +265,8 @@ nv50_state_validate(struct nv50_context *nv50) if (nv50->dirty & (NV50_NEW_FRAGPROG | NV50_NEW_FRAGPROG_CB)) nv50_fragprog_validate(nv50); - if (nv50->dirty & (NV50_NEW_FRAGPROG | NV50_NEW_VERTPROG)) + if (nv50->dirty & (NV50_NEW_FRAGPROG | NV50_NEW_VERTPROG | + NV50_NEW_RASTERIZER)) nv50_linkage_validate(nv50); if (nv50->dirty & NV50_NEW_RASTERIZER) @@ -285,7 +287,7 @@ nv50_state_validate(struct nv50_context *nv50) so = so_new(33, 0); so_method(so, tesla, NV50TCL_POLYGON_STIPPLE_PATTERN(0), 32); for (i = 0; i < 32; i++) - so_data(so, nv50->stipple.stipple[i]); + so_data(so, util_bswap32(nv50->stipple.stipple[i])); so_ref(so, &nv50->state.stipple); so_ref(NULL, &so); } diff --git a/src/gallium/drivers/nv50/nv50_tex.c b/src/gallium/drivers/nv50/nv50_tex.c index 2813f544770..417d3679422 100644 --- a/src/gallium/drivers/nv50/nv50_tex.c +++ b/src/gallium/drivers/nv50/nv50_tex.c @@ -131,9 +131,9 @@ nv50_tex_construct(struct nv50_context *nv50, struct nouveau_stateobj *so, NOUVEAU_BO_RD, 0, 0); so_data (so, mode); so_data (so, 0x00300000); - so_data (so, mt->base.base.width[0] | (1 << 31)); + so_data (so, mt->base.base.width0 | (1 << 31)); so_data (so, (mt->base.base.last_level << 28) | - (mt->base.base.depth[0] << 16) | mt->base.base.height[0]); + (mt->base.base.depth0 << 16) | mt->base.base.height0); so_data (so, 0x03000000); so_data (so, mt->base.base.last_level << 4); diff --git a/src/gallium/drivers/nv50/nv50_transfer.c b/src/gallium/drivers/nv50/nv50_transfer.c index ea61357aaa6..39d65279fc0 100644 --- a/src/gallium/drivers/nv50/nv50_transfer.c +++ b/src/gallium/drivers/nv50/nv50_transfer.c @@ -1,6 +1,7 @@ #include "pipe/p_context.h" #include "pipe/p_inlines.h" +#include "util/u_math.h" #include "nv50_context.h" @@ -156,9 +157,9 @@ nv50_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt, tx->base.block = pt->block; if (!pt->nblocksx[level]) { tx->base.nblocksx = pf_get_nblocksx(&pt->block, - pt->width[level]); + u_minify(pt->width0, level)); tx->base.nblocksy = pf_get_nblocksy(&pt->block, - pt->height[level]); + u_minify(pt->height0, level)); } else { tx->base.nblocksx = pt->nblocksx[level]; tx->base.nblocksy = pt->nblocksy[level]; @@ -167,9 +168,9 @@ nv50_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt, tx->base.usage = usage; tx->level_pitch = lvl->pitch; - tx->level_width = mt->base.base.width[level]; - tx->level_height = mt->base.base.height[level]; - tx->level_depth = mt->base.base.depth[level]; + tx->level_width = u_minify(mt->base.base.width0, level); + tx->level_height = u_minify(mt->base.base.height0, level); + tx->level_depth = u_minify(mt->base.base.depth0, level); tx->level_offset = lvl->image_offset[image]; tx->level_tiling = lvl->tile_mode; tx->level_x = pf_get_nblocksx(&tx->base.block, x); diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c index ae23329b83f..769733b6dd9 100644 --- a/src/gallium/drivers/r300/r300_context.c +++ b/src/gallium/drivers/r300/r300_context.c @@ -69,6 +69,7 @@ static void r300_destroy_context(struct pipe_context* context) FREE(r300->blend_color_state); FREE(r300->rs_block); FREE(r300->scissor_state); + FREE(r300->vertex_info); FREE(r300->viewport_state); FREE(r300); } @@ -123,15 +124,24 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen, r300->context.clear = r300_clear; - if (r300screen->caps->has_tcl) - { + if (r300screen->caps->has_tcl) { r300->context.draw_arrays = r300_draw_arrays; r300->context.draw_elements = r300_draw_elements; r300->context.draw_range_elements = r300_draw_range_elements; - } - else - { - assert(0); + } else { + r300->context.draw_arrays = r300_swtcl_draw_arrays; + r300->context.draw_elements = r300_draw_elements; + r300->context.draw_range_elements = r300_swtcl_draw_range_elements; + + /* Create a Draw. This is used for SW TCL. */ + r300->draw = draw_create(); + /* Enable our renderer. */ + draw_set_rasterize_stage(r300->draw, r300_draw_stage(r300)); + /* Enable Draw's clipping. */ + draw_set_driver_clipping(r300->draw, FALSE); + /* Force Draw to never do viewport transform, since we can do + * transform in hardware, always. */ + draw_set_viewport_state(r300->draw, &r300_viewport_identity); } r300->context.is_texture_referenced = r300_is_texture_referenced; @@ -143,18 +153,9 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen, r300->blend_color_state = CALLOC_STRUCT(r300_blend_color_state); r300->rs_block = CALLOC_STRUCT(r300_rs_block); r300->scissor_state = CALLOC_STRUCT(r300_scissor_state); + r300->vertex_info = CALLOC_STRUCT(r300_vertex_info); r300->viewport_state = CALLOC_STRUCT(r300_viewport_state); - /* Create a Draw. This is used for vert collation and SW TCL. */ - r300->draw = draw_create(); - /* Enable our renderer. */ - draw_set_rasterize_stage(r300->draw, r300_draw_stage(r300)); - /* Disable Draw's clipping if TCL is present. */ - draw_set_driver_clipping(r300->draw, r300_screen(screen)->caps->has_tcl); - /* Force Draw to never do viewport transform, since (again) we can do - * transform in hardware, always. */ - draw_set_viewport_state(r300->draw, &r300_viewport_identity); - /* Open up the OQ BO. */ r300->oqbo = screen->buffer_create(screen, 4096, PIPE_BUFFER_USAGE_VERTEX, 4096); diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h index f954ba7f9aa..39c0914cffa 100644 --- a/src/gallium/drivers/r300/r300_context.h +++ b/src/gallium/drivers/r300/r300_context.h @@ -92,6 +92,10 @@ struct r300_sampler_state { uint32_t filter0; /* R300_TX_FILTER0: 0x4400 */ uint32_t filter1; /* R300_TX_FILTER1: 0x4440 */ uint32_t border_color; /* R300_TX_BORDER_COLOR: 0x45c0 */ + + /* Min/max LOD must be clamped to [0, last_level], thus + * it's dependent on a currently bound texture */ + unsigned min_lod, max_lod; }; struct r300_scissor_state { @@ -219,11 +223,6 @@ struct r300_texture { struct r300_vertex_info { /* Parent class */ struct vertex_info vinfo; - /* Map of vertex attributes into PVS memory for HW TCL, - * or GA memory for SW TCL. */ - int vs_tab[16]; - /* Map of rasterizer attributes from GB through RS to US. */ - int fs_tab[16]; /* R300_VAP_PROG_STREAK_CNTL_[0-7] */ uint32_t vap_prog_stream_cntl[8]; diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index eeb97a2d370..98a39390bf9 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -129,7 +129,9 @@ static const float * get_shader_constant( struct rc_constant * constant, struct r300_constant_buffer * externals) { - static const float zero[4] = { 0.0, 0.0, 0.0, 0.0 }; + static float vec[4] = { 0.0, 0.0, 0.0, 1.0 }; + struct pipe_texture *tex; + switch(constant->Type) { case RC_CONSTANT_EXTERNAL: return externals->constants[constant->u.External]; @@ -137,11 +139,31 @@ static const float * get_shader_constant( case RC_CONSTANT_IMMEDIATE: return constant->u.Immediate; + case RC_CONSTANT_STATE: + switch (constant->u.State[0]) { + /* Factor for converting rectangle coords to + * normalized coords. Should only show up on non-r500. */ + case RC_STATE_R300_TEXRECT_FACTOR: + tex = &r300->textures[constant->u.State[1]]->tex; + vec[0] = 1.0 / tex->width0; + vec[1] = 1.0 / tex->height0; + break; + + default: + debug_printf("r300: Implementation error: " + "Unknown RC_CONSTANT type %d\n", constant->u.State[0]); + } + break; + default: - debug_printf("r300: Implementation error: Unhandled constant type %i\n", - constant->Type); - return zero; + debug_printf("r300: Implementation error: " + "Unhandled constant type %d\n", constant->Type); } + + /* This should either be (0, 0, 0, 1), which should be a relatively safe + * RGBA or STRQ value, or it could be one of the RC_CONSTANT_STATE + * state factors. */ + return vec; } /* Convert a normal single-precision float into the 7.16 format @@ -561,6 +583,8 @@ void r300_emit_texture(struct r300_context* r300, unsigned offset) { uint32_t filter0 = sampler->filter0; + uint32_t format0 = tex->state.format0; + unsigned min_level, max_level; CS_LOCALS(r300); /* to emulate 1D textures through 2D ones correctly */ @@ -569,13 +593,20 @@ void r300_emit_texture(struct r300_context* r300, filter0 |= R300_TX_WRAP_T(R300_TX_CLAMP_TO_EDGE); } + /* determine min/max levels */ + /* the MAX_MIP level is the largest (finest) one */ + max_level = MIN2(sampler->max_lod, tex->tex.last_level); + min_level = MIN2(sampler->min_lod, max_level); + format0 |= R300_TX_NUM_LEVELS(max_level); + filter0 |= R300_TX_MAX_MIP_LEVEL(min_level); + BEGIN_CS(16); OUT_CS_REG(R300_TX_FILTER0_0 + (offset * 4), filter0 | (offset << 28)); OUT_CS_REG(R300_TX_FILTER1_0 + (offset * 4), sampler->filter1); OUT_CS_REG(R300_TX_BORDER_COLOR_0 + (offset * 4), sampler->border_color); - OUT_CS_REG(R300_TX_FORMAT0_0 + (offset * 4), tex->state.format0); + OUT_CS_REG(R300_TX_FORMAT0_0 + (offset * 4), format0); OUT_CS_REG(R300_TX_FORMAT1_0 + (offset * 4), tex->state.format1); OUT_CS_REG(R300_TX_FORMAT2_0 + (offset * 4), tex->state.format2); OUT_CS_REG_SEQ(R300_TX_OFFSET_0 + (offset * 4), 1); @@ -690,12 +721,22 @@ void r300_emit_vertex_format_state(struct r300_context* r300) END_CS; } + void r300_emit_vertex_program_code(struct r300_context* r300, struct r300_vertex_program_code* code) { int i; struct r300_screen* r300screen = r300_screen(r300->context.screen); unsigned instruction_count = code->length / 4; + + int vtx_mem_size = r300screen->caps->is_r500 ? 128 : 72; + int input_count = MAX2(util_bitcount(code->InputsRead), 1); + int output_count = MAX2(util_bitcount(code->OutputsWritten), 1); + int temp_count = MAX2(code->num_temporaries, 1); + int pvs_num_slots = MIN3(vtx_mem_size / input_count, + vtx_mem_size / output_count, 10); + int pvs_num_controllers = MIN2(vtx_mem_size / temp_count, 6); + CS_LOCALS(r300); if (!r300screen->caps->has_tcl) { @@ -708,8 +749,7 @@ void r300_emit_vertex_program_code(struct r300_context* r300, /* R300_VAP_PVS_CODE_CNTL_0 * R300_VAP_PVS_CONST_CNTL * R300_VAP_PVS_CODE_CNTL_1 - * See the r5xx docs for instructions on how to use these. - * XXX these could be optimized to select better values... */ + * See the r5xx docs for instructions on how to use these. */ OUT_CS_REG_SEQ(R300_VAP_PVS_CODE_CNTL_0, 3); OUT_CS(R300_PVS_FIRST_INST(0) | R300_PVS_XYZW_VALID_INST(instruction_count - 1) | @@ -722,10 +762,11 @@ void r300_emit_vertex_program_code(struct r300_context* r300, for (i = 0; i < code->length; i++) OUT_CS(code->body.d[i]); - OUT_CS_REG(R300_VAP_CNTL, R300_PVS_NUM_SLOTS(10) | - R300_PVS_NUM_CNTLRS(5) | + OUT_CS_REG(R300_VAP_CNTL, R300_PVS_NUM_SLOTS(pvs_num_slots) | + R300_PVS_NUM_CNTLRS(pvs_num_controllers) | R300_PVS_NUM_FPUS(r300screen->caps->num_vert_fpus) | - R300_PVS_VF_MAX_VTX_NUM(12)); + R300_PVS_VF_MAX_VTX_NUM(12) | + (r300screen->caps->is_r500 ? R500_TCL_STATE_OPTIMIZATION : 0)); END_CS; } @@ -790,13 +831,22 @@ void r300_emit_viewport_state(struct r300_context* r300, END_CS; } +void r300_emit_texture_count(struct r300_context* r300) +{ + CS_LOCALS(r300); + + BEGIN_CS(2); + OUT_CS_REG(R300_TX_ENABLE, (1 << r300->texture_count) - 1); + END_CS; + +} + void r300_flush_textures(struct r300_context* r300) { CS_LOCALS(r300); - BEGIN_CS(4); + BEGIN_CS(2); OUT_CS_REG(R300_TX_INVALTAGS, 0); - OUT_CS_REG(R300_TX_ENABLE, (1 << r300->texture_count) - 1); END_CS; } @@ -950,6 +1000,8 @@ validate: /* Samplers and textures are tracked separately but emitted together. */ if (r300->dirty_state & (R300_ANY_NEW_SAMPLERS | R300_ANY_NEW_TEXTURES)) { + r300_emit_texture_count(r300); + for (i = 0; i < MIN2(r300->sampler_count, r300->texture_count); i++) { if (r300->dirty_state & ((R300_NEW_SAMPLER << i) | (R300_NEW_TEXTURE << i))) { diff --git a/src/gallium/drivers/r300/r300_emit.h b/src/gallium/drivers/r300/r300_emit.h index 7c83c5166de..3797d3d332a 100644 --- a/src/gallium/drivers/r300/r300_emit.h +++ b/src/gallium/drivers/r300/r300_emit.h @@ -92,6 +92,8 @@ void r300_emit_vertex_shader(struct r300_context* r300, void r300_emit_viewport_state(struct r300_context* r300, struct r300_viewport_state* viewport); +void r300_emit_texture_count(struct r300_context* r300); + void r300_flush_textures(struct r300_context* r300); /* Emit all dirty state. */ diff --git a/src/gallium/drivers/r300/r300_fs.c b/src/gallium/drivers/r300/r300_fs.c index 29ddc84c411..79b01bb4dc2 100644 --- a/src/gallium/drivers/r300/r300_fs.c +++ b/src/gallium/drivers/r300/r300_fs.c @@ -1,6 +1,7 @@ /* * Copyright 2008 Corbin Simpson <[email protected]> * Joakim Sindholt <[email protected]> + * Copyright 2009 Marek Olšák <[email protected]> * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -31,6 +32,41 @@ #include "radeon_code.h" #include "radeon_compiler.h" +/* Convert info about FS input semantics to r300_shader_semantics. */ +static void r300_shader_read_fs_inputs(struct tgsi_shader_info* info, + struct r300_shader_semantics* fs_inputs) +{ + int i; + unsigned index; + + r300_shader_semantics_reset(fs_inputs); + + for (i = 0; i < info->num_inputs; i++) { + index = info->input_semantic_index[i]; + + switch (info->input_semantic_name[i]) { + case TGSI_SEMANTIC_COLOR: + assert(index <= ATTR_COLOR_COUNT); + fs_inputs->color[index] = i; + break; + + case TGSI_SEMANTIC_GENERIC: + assert(index <= ATTR_GENERIC_COUNT); + fs_inputs->generic[index] = i; + break; + + case TGSI_SEMANTIC_FOG: + assert(index == 0); + fs_inputs->fog = i; + break; + + default: + assert(0); + } + } +} + + static void find_output_registers(struct r300_fragment_program_compiler * compiler, struct r300_fragment_shader * fs) { @@ -58,38 +94,24 @@ static void allocate_hardware_inputs( void (*allocate)(void * data, unsigned input, unsigned hwreg), void * mydata) { - struct tgsi_shader_info* info = &((struct r300_fragment_shader*)c->UserData)->info; - int total_colors = 0; - int colors = 0; - int total_generic = 0; - int generic = 0; - int i; - - for (i = 0; i < info->num_inputs; i++) { - switch (info->input_semantic_name[i]) { - case TGSI_SEMANTIC_COLOR: - total_colors++; - break; - case TGSI_SEMANTIC_FOG: - case TGSI_SEMANTIC_GENERIC: - total_generic++; - break; + struct r300_shader_semantics* inputs = + &((struct r300_fragment_shader*)c->UserData)->inputs; + int i, reg = 0; + + /* Allocate input registers. */ + for (i = 0; i < ATTR_COLOR_COUNT; i++) { + if (inputs->color[i] != ATTR_UNUSED) { + allocate(mydata, inputs->color[i], reg++); } } - - for(i = 0; i < info->num_inputs; i++) { - switch (info->input_semantic_name[i]) { - case TGSI_SEMANTIC_COLOR: - allocate(mydata, i, colors); - colors++; - break; - case TGSI_SEMANTIC_FOG: - case TGSI_SEMANTIC_GENERIC: - allocate(mydata, i, total_colors + generic); - generic++; - break; + for (i = 0; i < ATTR_GENERIC_COUNT; i++) { + if (inputs->generic[i] != ATTR_UNUSED) { + allocate(mydata, inputs->generic[i], reg++); } } + if (inputs->fog != ATTR_UNUSED) { + allocate(mydata, inputs->fog, reg++); + } } void r300_translate_fragment_shader(struct r300_context* r300, @@ -98,6 +120,10 @@ void r300_translate_fragment_shader(struct r300_context* r300, struct r300_fragment_program_compiler compiler; struct tgsi_to_rc ttr; + /* Initialize. */ + r300_shader_read_fs_inputs(&fs->info, &fs->inputs); + + /* Setup the compiler. */ memset(&compiler, 0, sizeof(compiler)); rc_init(&compiler.Base); compiler.Base.Debug = DBG_ON(r300, DBG_FP); @@ -107,7 +133,7 @@ void r300_translate_fragment_shader(struct r300_context* r300, compiler.AllocateHwInputs = &allocate_hardware_inputs; compiler.UserData = fs; - /* TODO: Program compilation depends on texture compare modes, + /* XXX: Program compilation depends on texture compare modes, * which are sampler state. Therefore, programs need to be recompiled * depending on this state as in the classic Mesa driver. * @@ -133,6 +159,7 @@ void r300_translate_fragment_shader(struct r300_context* r300, /* XXX failover maybe? */ DBG(r300, DBG_FP, "r300: Error compiling fragment program: %s\n", compiler.Base.ErrorMsg); + assert(0); } /* And, finally... */ diff --git a/src/gallium/drivers/r300/r300_fs.h b/src/gallium/drivers/r300/r300_fs.h index e831c30301b..630e2d0c8a5 100644 --- a/src/gallium/drivers/r300/r300_fs.h +++ b/src/gallium/drivers/r300/r300_fs.h @@ -1,6 +1,7 @@ /* * Copyright 2008 Corbin Simpson <[email protected]> * Joakim Sindholt <[email protected]> + * Copyright 2009 Marek Olšák <[email protected]> * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -25,15 +26,16 @@ #define R300_FS_H #include "pipe/p_state.h" - #include "tgsi/tgsi_scan.h" - #include "radeon_code.h" +#include "r300_shader_semantics.h" struct r300_fragment_shader { /* Parent class */ struct pipe_shader_state state; + struct tgsi_shader_info info; + struct r300_shader_semantics inputs; /* Has this shader been translated yet? */ boolean translated; diff --git a/src/gallium/drivers/r300/r300_reg.h b/src/gallium/drivers/r300/r300_reg.h index 8ca785cb587..85b1ea568a3 100644 --- a/src/gallium/drivers/r300/r300_reg.h +++ b/src/gallium/drivers/r300/r300_reg.h @@ -661,20 +661,20 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R300_GB_SUPER_TILE_B (1 << 15) # define R300_GB_SUBPIXEL_1_12 (0 << 16) # define R300_GB_SUBPIXEL_1_16 (1 << 16) -# define GB_TILE_CONFIG_QUADS_PER_RAS_4 (0 << 17) -# define GB_TILE_CONFIG_QUADS_PER_RAS_8 (1 << 17) -# define GB_TILE_CONFIG_QUADS_PER_RAS_16 (2 << 17) -# define GB_TILE_CONFIG_QUADS_PER_RAS_32 (3 << 17) -# define GB_TILE_CONFIG_BB_SCAN_INTERCEPT (0 << 19) -# define GB_TILE_CONFIG_BB_SCAN_BOUND_BOX (1 << 19) -# define GB_TILE_CONFIG_ALT_SCAN_EN_LR (0 << 20) -# define GB_TILE_CONFIG_ALT_SCAN_EN_LRL (1 << 20) -# define GB_TILE_CONFIG_ALT_OFFSET (0 << 21) -# define GB_TILE_CONFIG_SUBPRECISION (0 << 22) -# define GB_TILE_CONFIG_ALT_TILING_DEF (0 << 23) -# define GB_TILE_CONFIG_ALT_TILING_3_2 (1 << 23) -# define GB_TILE_CONFIG_Z_EXTENDED_24_1 (0 << 24) -# define GB_TILE_CONFIG_Z_EXTENDED_S25_1 (1 << 24) +# define R300_GB_TILE_CONFIG_QUADS_PER_RAS_4 (0 << 17) +# define R300_GB_TILE_CONFIG_QUADS_PER_RAS_8 (1 << 17) +# define R300_GB_TILE_CONFIG_QUADS_PER_RAS_16 (2 << 17) +# define R300_GB_TILE_CONFIG_QUADS_PER_RAS_32 (3 << 17) +# define R300_GB_TILE_CONFIG_BB_SCAN_INTERCEPT (0 << 19) +# define R300_GB_TILE_CONFIG_BB_SCAN_BOUND_BOX (1 << 19) +# define R300_GB_TILE_CONFIG_ALT_SCAN_EN_LR (0 << 20) +# define R300_GB_TILE_CONFIG_ALT_SCAN_EN_LRL (1 << 20) +# define R300_GB_TILE_CONFIG_ALT_OFFSET (0 << 21) +# define R300_GB_TILE_CONFIG_SUBPRECISION (0 << 22) +# define R300_GB_TILE_CONFIG_ALT_TILING_DEF (0 << 23) +# define R300_GB_TILE_CONFIG_ALT_TILING_3_2 (1 << 23) +# define R300_GB_TILE_CONFIG_Z_EXTENDED_24_1 (0 << 24) +# define R300_GB_TILE_CONFIG_Z_EXTENDED_S25_1 (1 << 24) /* Specifies the sizes of the various FIFO`s in the sc/rs/us. This register must be the first one written */ #define R300_GB_FIFO_SIZE 0x4024 @@ -700,9 +700,9 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R300_OFIFO_HIGHWATER_SHIFT 22 /* two bits only */ # define R300_CUBE_FIFO_HIGHWATER_COL_SHIFT 24 -#define GB_Z_PEQ_CONFIG 0x4028 -# define GB_Z_PEQ_CONFIG_Z_PEQ_SIZE_4_4 (0 << 0) -# define GB_Z_PEQ_CONFIG_Z_PEQ_SIZE_8_8 (1 << 0) +#define R300_GB_Z_PEQ_CONFIG 0x4028 +# define R300_GB_Z_PEQ_CONFIG_Z_PEQ_SIZE_4_4 (0 << 0) +# define R300_GB_Z_PEQ_CONFIG_Z_PEQ_SIZE_8_8 (1 << 0) /* Specifies various polygon specific selects (fog, depth, perspective). */ #define R300_GB_SELECT 0x401c @@ -725,39 +725,39 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. /* Specifies the graphics pipeline configuration for antialiasing. */ #define R300_GB_AA_CONFIG 0x4020 -# define GB_AA_CONFIG_AA_DISABLE (0 << 0) -# define GB_AA_CONFIG_AA_ENABLE (1 << 0) -# define GB_AA_CONFIG_NUM_AA_SUBSAMPLES_2 (0 << 1) -# define GB_AA_CONFIG_NUM_AA_SUBSAMPLES_3 (1 << 1) -# define GB_AA_CONFIG_NUM_AA_SUBSAMPLES_4 (2 << 1) -# define GB_AA_CONFIG_NUM_AA_SUBSAMPLES_6 (3 << 1) +# define R300_GB_AA_CONFIG_AA_DISABLE (0 << 0) +# define R300_GB_AA_CONFIG_AA_ENABLE (1 << 0) +# define R300_GB_AA_CONFIG_NUM_AA_SUBSAMPLES_2 (0 << 1) +# define R300_GB_AA_CONFIG_NUM_AA_SUBSAMPLES_3 (1 << 1) +# define R300_GB_AA_CONFIG_NUM_AA_SUBSAMPLES_4 (2 << 1) +# define R300_GB_AA_CONFIG_NUM_AA_SUBSAMPLES_6 (3 << 1) /* Selects which of 4 pipes are active. */ -#define GB_PIPE_SELECT 0x402c -# define GB_PIPE_SELECT_PIPE0_ID_SHIFT 0 -# define GB_PIPE_SELECT_PIPE1_ID_SHIFT 2 -# define GB_PIPE_SELECT_PIPE2_ID_SHIFT 4 -# define GB_PIPE_SELECT_PIPE3_ID_SHIFT 6 -# define GB_PIPE_SELECT_PIPE_MASK_SHIFT 8 -# define GB_PIPE_SELECT_MAX_PIPE 12 -# define GB_PIPE_SELECT_BAD_PIPES 14 -# define GB_PIPE_SELECT_CONFIG_PIPES 18 +#define R300_GB_PIPE_SELECT 0x402c +# define R300_GB_PIPE_SELECT_PIPE0_ID_SHIFT 0 +# define R300_GB_PIPE_SELECT_PIPE1_ID_SHIFT 2 +# define R300_GB_PIPE_SELECT_PIPE2_ID_SHIFT 4 +# define R300_GB_PIPE_SELECT_PIPE3_ID_SHIFT 6 +# define R300_GB_PIPE_SELECT_PIPE_MASK_SHIFT 8 +# define R300_GB_PIPE_SELECT_MAX_PIPE 12 +# define R300_GB_PIPE_SELECT_BAD_PIPES 14 +# define R300_GB_PIPE_SELECT_CONFIG_PIPES 18 /* Specifies the sizes of the various FIFO`s in the sc/rs. */ -#define GB_FIFO_SIZE1 0x4070 +#define R300_GB_FIFO_SIZE1 0x4070 /* High water mark for SC input fifo */ -# define GB_FIFO_SIZE1_SC_HIGHWATER_IFIFO_SHIFT 0 -# define GB_FIFO_SIZE1_SC_HIGHWATER_IFIFO_MASK 0x0000003f +# define R300_GB_FIFO_SIZE1_SC_HIGHWATER_IFIFO_SHIFT 0 +# define R300_GB_FIFO_SIZE1_SC_HIGHWATER_IFIFO_MASK 0x0000003f /* High water mark for SC input fifo (B) */ -# define GB_FIFO_SIZE1_SC_HIGHWATER_BFIFO_SHIFT 6 -# define GB_FIFO_SIZE1_SC_HIGHWATER_BFIFO_MASK 0x00000fc0 +# define R300_GB_FIFO_SIZE1_SC_HIGHWATER_BFIFO_SHIFT 6 +# define R300_GB_FIFO_SIZE1_SC_HIGHWATER_BFIFO_MASK 0x00000fc0 /* High water mark for RS colors' fifo */ -# define GB_FIFO_SIZE1_SC_HIGHWATER_COL_SHIFT 12 -# define GB_FIFO_SIZE1_SC_HIGHWATER_COL_MASK 0x0003f000 +# define R300_GB_FIFO_SIZE1_SC_HIGHWATER_COL_SHIFT 12 +# define R300_GB_FIFO_SIZE1_SC_HIGHWATER_COL_MASK 0x0003f000 /* High water mark for RS textures' fifo */ -# define GB_FIFO_SIZE1_SC_HIGHWATER_TEX_SHIFT 18 -# define GB_FIFO_SIZE1_SC_HIGHWATER_TEX_MASK 0x00fc0000 +# define R300_GB_FIFO_SIZE1_SC_HIGHWATER_TEX_SHIFT 18 +# define R300_GB_FIFO_SIZE1_SC_HIGHWATER_TEX_MASK 0x00fc0000 /* This table specifies the source location and format for up to 16 texture * addresses (i[0]:i[15]) and four colors (c[0]:c[3]) @@ -1293,7 +1293,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R500_RS_INST_TEX_ID(x) ((x) << 0) #define R500_RS_INST_TEX_CN_WRITE (1 << 4) #define R500_RS_INST_TEX_ADDR_SHIFT 5 -# define R500_RS_INST_TEX_ADDR(x) ((x) << 0) +# define R500_RS_INST_TEX_ADDR(x) ((x) << 5) #define R500_RS_INST_COL_ID_SHIFT 12 # define R500_RS_INST_COL_ID(x) ((x) << 12) #define R500_RS_INST_COL_CN_NO_WRITE (0 << 16) @@ -1463,6 +1463,8 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R300_TX_MIN_FILTER_MIP_NEAREST (1 << 13) # define R300_TX_MIN_FILTER_MIP_LINEAR (2 << 13) # define R300_TX_MIN_FILTER_MIP_MASK (3 << 13) +# define R300_TX_MAX_MIP_LEVEL_SHIFT 17 +# define R300_TX_MAX_MIP_LEVEL_MASK (0xf << 17) # define R300_TX_MAX_ANISO_1_TO_1 (0 << 21) # define R300_TX_MAX_ANISO_2_TO_1 (1 << 21) # define R300_TX_MAX_ANISO_4_TO_1 (2 << 21) @@ -1471,6 +1473,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R300_TX_MAX_ANISO_MASK (7 << 21) # define R300_TX_WRAP_S(x) ((x) << 0) # define R300_TX_WRAP_T(x) ((x) << 3) +# define R300_TX_MAX_MIP_LEVEL(x) ((x) << 17) #define R300_TX_FILTER1_0 0x4440 # define R300_CHROMA_KEY_MODE_DISABLE 0 @@ -1500,8 +1503,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R300_TX_HEIGHTMASK_MASK (2047 << 11) # define R300_TX_DEPTHMASK_SHIFT 22 # define R300_TX_DEPTHMASK_MASK (0xf << 22) -# define R300_TX_MAX_MIP_LEVEL_SHIFT 26 -# define R300_TX_MAX_MIP_LEVEL_MASK (0xf << 26) # define R300_TX_SIZE_PROJECTED (1 << 30) # define R300_TX_PITCH_EN (1 << 31) # define R300_TX_WIDTH(x) ((x) << 0) diff --git a/src/gallium/drivers/r300/r300_render.c b/src/gallium/drivers/r300/r300_render.c index 62e1456ed36..4c5fb405c6a 100644 --- a/src/gallium/drivers/r300/r300_render.c +++ b/src/gallium/drivers/r300/r300_render.c @@ -183,8 +183,6 @@ boolean r300_draw_range_elements(struct pipe_context* pipe, return FALSE; } - setup_vertex_attributes(r300); - setup_index_buffer(r300, indexBuffer, indexSize); r300_emit_dirty_state(r300); @@ -226,8 +224,6 @@ boolean r300_draw_arrays(struct pipe_context* pipe, unsigned mode, return FALSE; } - setup_vertex_attributes(r300); - r300_emit_dirty_state(r300); r300_emit_aos(r300, start); diff --git a/src/gallium/drivers/r300/r300_screen.h b/src/gallium/drivers/r300/r300_screen.h index 41df31f670f..1ce5ff3904b 100644 --- a/src/gallium/drivers/r300/r300_screen.h +++ b/src/gallium/drivers/r300/r300_screen.h @@ -27,6 +27,8 @@ #include "r300_chipset.h" +struct r300_winsys; + struct r300_screen { /* Parent class */ struct pipe_screen screen; diff --git a/src/gallium/drivers/r300/r300_shader_semantics.h b/src/gallium/drivers/r300/r300_shader_semantics.h new file mode 100644 index 00000000000..85184e2cfd7 --- /dev/null +++ b/src/gallium/drivers/r300/r300_shader_semantics.h @@ -0,0 +1,64 @@ +/* + * Copyright 2009 Marek Olšák <[email protected]> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. */ + +#ifndef R300_SHADER_SEMANTICS_H +#define R300_SHADER_SEMANTICS_H + +#define ATTR_UNUSED (-1) +#define ATTR_COLOR_COUNT 2 +#define ATTR_GENERIC_COUNT 16 + +/* This structure contains information about what attributes are written by VS + * or read by FS. (but not both) It's much easier to work with than + * tgsi_shader_info. + * + * The variables contain indices to tgsi_shader_info semantics and those + * indices are nothing else than input/output register numbers. */ +struct r300_shader_semantics { + int pos; + int psize; + int color[ATTR_COLOR_COUNT]; + int bcolor[ATTR_COLOR_COUNT]; + int generic[ATTR_GENERIC_COUNT]; + int fog; +}; + +static INLINE void r300_shader_semantics_reset( + struct r300_shader_semantics* info) +{ + int i; + + info->pos = ATTR_UNUSED; + info->psize = ATTR_UNUSED; + info->fog = ATTR_UNUSED; + + for (i = 0; i < ATTR_COLOR_COUNT; i++) { + info->color[i] = ATTR_UNUSED; + info->bcolor[i] = ATTR_UNUSED; + } + + for (i = 0; i < ATTR_GENERIC_COUNT; i++) { + info->generic[i] = ATTR_UNUSED; + } +} + +#endif diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c index d1eced61db1..7505353953f 100644 --- a/src/gallium/drivers/r300/r300_state.c +++ b/src/gallium/drivers/r300/r300_state.c @@ -523,6 +523,11 @@ static void* state->mag_img_filter, state->min_mip_filter); + /* Unfortunately, r300-r500 don't support floating-point mipmap lods. */ + /* We must pass these to the emit function to clamp them properly. */ + sampler->min_lod = MAX2((unsigned)state->min_lod, 0); + sampler->max_lod = MAX2((unsigned)ceilf(state->max_lod), 0); + lod_bias = CLAMP((int)(state->lod_bias * 32), -(1 << 9), (1 << 9) - 1); sampler->filter1 |= lod_bias << R300_LOD_BIAS_SHIFT; @@ -571,6 +576,7 @@ static void r300_set_sampler_textures(struct pipe_context* pipe, struct pipe_texture** texture) { struct r300_context* r300 = r300_context(pipe); + boolean is_r500 = r300_screen(r300->context.screen)->caps->is_r500; int i; /* XXX magic num */ @@ -585,6 +591,13 @@ static void r300_set_sampler_textures(struct pipe_context* pipe, pipe_texture_reference((struct pipe_texture**)&r300->textures[i], texture[i]); r300->dirty_state |= (R300_NEW_TEXTURE << i); + + /* R300-specific - set the texrect factor in a fragment shader */ + if (!is_r500 && r300->textures[i]->is_npot) { + /* XXX It would be nice to re-emit just 1 constant, + * XXX not all of them */ + r300->dirty_state |= R300_NEW_FRAGMENT_SHADER_CONSTANTS; + } } } @@ -674,6 +687,8 @@ static void r300_set_vertex_buffers(struct pipe_context* pipe, draw_flush(r300->draw); draw_set_vertex_buffers(r300->draw, count, buffers); } + + r300->dirty_state |= R300_NEW_VERTEX_FORMAT; } static void r300_set_vertex_elements(struct pipe_context* pipe, @@ -706,9 +721,6 @@ static void* r300_create_vs_state(struct pipe_context* pipe, tgsi_scan_shader(shader->tokens, &vs->info); - /* Appease Draw. */ - vs->draw = draw_create_vertex_shader(r300->draw, shader); - return (void*)vs; } else { return draw_create_vertex_shader(r300->draw, shader); @@ -719,8 +731,6 @@ static void r300_bind_vs_state(struct pipe_context* pipe, void* shader) { struct r300_context* r300 = r300_context(pipe); - draw_flush(r300->draw); - if (r300_screen(pipe->screen)->caps->has_tcl) { struct r300_vertex_shader* vs = (struct r300_vertex_shader*)shader; @@ -731,10 +741,10 @@ static void r300_bind_vs_state(struct pipe_context* pipe, void* shader) r300_translate_vertex_shader(r300, vs); } - draw_bind_vertex_shader(r300->draw, vs->draw); r300->vs = vs; r300->dirty_state |= R300_NEW_VERTEX_SHADER | R300_NEW_VERTEX_SHADER_CONSTANTS; } else { + draw_flush(r300->draw); draw_bind_vertex_shader(r300->draw, (struct draw_vertex_shader*)shader); } @@ -748,7 +758,6 @@ static void r300_delete_vs_state(struct pipe_context* pipe, void* shader) struct r300_vertex_shader* vs = (struct r300_vertex_shader*)shader; rc_constants_destroy(&vs->code.constants); - draw_delete_vertex_shader(r300->draw, vs->draw); FREE((void*)vs->state.tokens); FREE(shader); } else { @@ -813,10 +822,10 @@ void r300_init_state_functions(struct r300_context* r300) r300->context.delete_rasterizer_state = r300_delete_rs_state; r300->context.create_sampler_state = r300_create_sampler_state; - r300->context.bind_sampler_states = r300_bind_sampler_states; + r300->context.bind_fragment_sampler_states = r300_bind_sampler_states; r300->context.delete_sampler_state = r300_delete_sampler_state; - r300->context.set_sampler_textures = r300_set_sampler_textures; + r300->context.set_fragment_sampler_textures = r300_set_sampler_textures; r300->context.set_scissor_state = r300_set_scissor_state; diff --git a/src/gallium/drivers/r300/r300_state_derived.c b/src/gallium/drivers/r300/r300_state_derived.c index 7166694edf4..cd969d633bc 100644 --- a/src/gallium/drivers/r300/r300_state_derived.c +++ b/src/gallium/drivers/r300/r300_state_derived.c @@ -1,5 +1,6 @@ /* * Copyright 2008 Corbin Simpson <[email protected]> + * Copyright 2009 Marek Olšák <[email protected]> * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -28,6 +29,7 @@ #include "r300_context.h" #include "r300_fs.h" #include "r300_screen.h" +#include "r300_shader_semantics.h" #include "r300_state_derived.h" #include "r300_state_inlines.h" #include "r300_vs.h" @@ -47,8 +49,8 @@ struct r300_shader_derived_value { unsigned r300_shader_key_hash(void* key) { struct r300_shader_key* shader_key = (struct r300_shader_key*)key; - unsigned vs = (unsigned)shader_key->vs; - unsigned fs = (unsigned)shader_key->fs; + unsigned vs = (intptr_t)shader_key->vs; + unsigned fs = (intptr_t)shader_key->fs; return (vs << 16) | (fs & 0xffff); } @@ -61,209 +63,143 @@ int r300_shader_key_compare(void* key1, void* key2) { (shader_key1->fs == shader_key2->fs); } -/* Set up the vs_tab and routes. */ -static void r300_vs_tab_routes(struct r300_context* r300, - struct r300_vertex_info* vformat) +static void r300_draw_emit_attrib(struct r300_context* r300, + enum attrib_emit emit, + enum interp_mode interp, + int index) { - struct r300_screen* r300screen = r300_screen(r300->context.screen); - struct vertex_info* vinfo = &vformat->vinfo; - int* tab = vformat->vs_tab; - boolean pos = FALSE, psize = FALSE, fog = FALSE; - int i, texs = 0, cols = 0; - struct tgsi_shader_info* info; + struct tgsi_shader_info* info = &r300->vs->info; + int output; - if (r300screen->caps->has_tcl) { - /* Use vertex shader to determine required routes. */ - info = &r300->vs->info; + output = draw_find_vs_output(r300->draw, + info->output_semantic_name[index], + info->output_semantic_index[index]); + draw_emit_vertex_attr(&r300->vertex_info->vinfo, emit, interp, output); +} + +static void r300_draw_emit_all_attribs(struct r300_context* r300) +{ + struct r300_shader_semantics* vs_outputs = &r300->vs->outputs; + int i, gen_count; + + /* Position. */ + if (vs_outputs->pos != ATTR_UNUSED) { + r300_draw_emit_attrib(r300, EMIT_4F, INTERP_PERSPECTIVE, + vs_outputs->pos); } else { - /* Use fragment shader to determine required routes. */ - info = &r300->fs->info; + assert(0); } - assert(info->num_inputs <= 16); - - if (!r300screen->caps->has_tcl || !r300->rs_state->enable_vte) - { - for (i = 0; i < info->num_inputs; i++) { - switch (r300->vs->code.inputs[i]) { - case TGSI_SEMANTIC_POSITION: - pos = TRUE; - tab[i] = 0; - break; - case TGSI_SEMANTIC_COLOR: - tab[i] = 2 + cols; - cols++; - break; - case TGSI_SEMANTIC_PSIZE: - assert(psize == FALSE); - psize = TRUE; - tab[i] = 15; - break; - case TGSI_SEMANTIC_FOG: - assert(fog == FALSE); - fog = TRUE; - /* Fall through */ - case TGSI_SEMANTIC_GENERIC: - tab[i] = 6 + texs; - texs++; - break; - default: - debug_printf("r300: Unknown vertex input %d\n", - info->input_semantic_name[i]); - break; - } - } + /* Point size. */ + if (vs_outputs->psize != ATTR_UNUSED) { + r300_draw_emit_attrib(r300, EMIT_1F_PSIZE, INTERP_POS, + vs_outputs->psize); } - else - { - /* Just copy vert attribs over as-is. */ - for (i = 0; i < info->num_inputs; i++) { - tab[i] = i; - } - for (i = 0; i < info->num_outputs; i++) { - switch (info->output_semantic_name[i]) { - case TGSI_SEMANTIC_POSITION: - pos = TRUE; - break; - case TGSI_SEMANTIC_COLOR: - cols++; - break; - case TGSI_SEMANTIC_PSIZE: - psize = TRUE; - break; - case TGSI_SEMANTIC_FOG: - fog = TRUE; - /* Fall through */ - case TGSI_SEMANTIC_GENERIC: - texs++; - break; - default: - debug_printf("r300: Unknown vertex output %d\n", - info->output_semantic_name[i]); - break; - } + /* Colors. */ + for (i = 0; i < ATTR_COLOR_COUNT; i++) { + if (vs_outputs->color[i] != ATTR_UNUSED) { + r300_draw_emit_attrib(r300, EMIT_4F, INTERP_LINEAR, + vs_outputs->color[i]); } } - /* XXX magic */ - assert(texs <= 8); + /* XXX Back-face colors. */ - /* Do the actual vertex_info setup. - * - * vertex_info has four uints of hardware-specific data in it. - * vinfo.hwfmt[0] is R300_VAP_VTX_STATE_CNTL - * vinfo.hwfmt[1] is R300_VAP_VSM_VTX_ASSM - * vinfo.hwfmt[2] is R300_VAP_OUTPUT_VTX_FMT_0 - * vinfo.hwfmt[3] is R300_VAP_OUTPUT_VTX_FMT_1 */ - - vinfo->hwfmt[0] = 0x5555; /* XXX this is classic Mesa bonghits */ - - /* We need to add vertex position attribute only for SW TCL case, - * for HW TCL case it could be generated by vertex shader */ - if (!pos && !r300screen->caps->has_tcl) { - debug_printf("r300: Forcing vertex position attribute emit...\n"); - /* Make room for the position attribute - * at the beginning of the tab. */ - for (i = 15; i > 0; i--) { - tab[i] = tab[i-1]; + /* Texture coordinates. */ + gen_count = 0; + for (i = 0; i < ATTR_GENERIC_COUNT; i++) { + if (vs_outputs->generic[i] != ATTR_UNUSED) { + r300_draw_emit_attrib(r300, EMIT_4F, INTERP_PERSPECTIVE, + vs_outputs->generic[i]); + gen_count++; } - tab[0] = 0; } - /* Position. */ - if (r300->draw) { - draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, - draw_find_vs_output(r300->draw, TGSI_SEMANTIC_POSITION, 0)); + /* Fog coordinates. */ + if (vs_outputs->fog != ATTR_UNUSED) { + r300_draw_emit_attrib(r300, EMIT_4F, INTERP_PERSPECTIVE, + vs_outputs->fog); + gen_count++; } - vinfo->hwfmt[1] |= R300_INPUT_CNTL_POS; - vinfo->hwfmt[2] |= R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT; - /* Point size. */ - if (psize) { - if (r300->draw) { - draw_emit_vertex_attr(vinfo, EMIT_1F_PSIZE, INTERP_POS, - draw_find_vs_output(r300->draw, TGSI_SEMANTIC_PSIZE, 0)); - } - vinfo->hwfmt[2] |= R300_VAP_OUTPUT_VTX_FMT_0__PT_SIZE_PRESENT; - } + /* XXX magic */ + assert(gen_count <= 8); +} - /* Colors. */ - for (i = 0; i < cols; i++) { - if (r300->draw) { - draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_LINEAR, - draw_find_vs_output(r300->draw, TGSI_SEMANTIC_COLOR, i)); - } - vinfo->hwfmt[1] |= R300_INPUT_CNTL_COLOR; - vinfo->hwfmt[2] |= (R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT << i); - } +/* Update the PSC tables. */ +static void r300_vertex_psc(struct r300_context* r300) +{ + struct r300_vertex_info *vformat = r300->vertex_info; + uint16_t type, swizzle; + enum pipe_format format; + unsigned i; - /* Init i right here, increment it if fog is enabled. - * This gets around a double-increment problem. */ - i = 0; + /* Vertex shaders have no semantics on their inputs, + * so PSC should just route stuff based on the vertex elements, + * and not on attrib information. */ + DBG(r300, DBG_DRAW, "r300: vs expects %d attribs, routing %d elements" + " in psc\n", + r300->vs->info.num_inputs, + r300->vertex_element_count); - /* Fog. This is a special-cased texcoord. */ - if (fog) { - i++; - if (r300->draw) { - draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, - draw_find_vs_output(r300->draw, TGSI_SEMANTIC_FOG, 0)); - } - vinfo->hwfmt[1] |= (R300_INPUT_CNTL_TC0 << i); - vinfo->hwfmt[3] |= (4 << (3 * i)); - } + for (i = 0; i < r300->vertex_element_count; i++) { + format = r300->vertex_element[i].src_format; + + type = r300_translate_vertex_data_type(format) | + (i << R300_DST_VEC_LOC_SHIFT); + swizzle = r300_translate_vertex_data_swizzle(format); - /* Texcoords. */ - for (; i < texs; i++) { - if (r300->draw) { - draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, - draw_find_vs_output(r300->draw, TGSI_SEMANTIC_GENERIC, i)); + if (i % 2) { + vformat->vap_prog_stream_cntl[i >> 1] |= type << 16; + vformat->vap_prog_stream_cntl_ext[i >> 1] |= swizzle << 16; + } else { + vformat->vap_prog_stream_cntl[i >> 1] |= type; + vformat->vap_prog_stream_cntl_ext[i >> 1] |= swizzle; } - vinfo->hwfmt[1] |= (R300_INPUT_CNTL_TC0 << i); - vinfo->hwfmt[3] |= (4 << (3 * i)); } - draw_compute_vertex_size(vinfo); + + assert(i <= 15); + + /* Set the last vector in the PSC. */ + if (i) { + i -= 1; + } + vformat->vap_prog_stream_cntl[i >> 1] |= + (R300_LAST_VEC << (i & 1 ? 16 : 0)); } -/* Update the PSC tables. */ -static void r300_vertex_psc(struct r300_context* r300, - struct r300_vertex_info* vformat) +/* Update the PSC tables for SW TCL, using Draw. */ +static void r300_swtcl_vertex_psc(struct r300_context* r300) { - struct r300_screen* r300screen = r300_screen(r300->context.screen); + struct r300_vertex_info *vformat = r300->vertex_info; struct vertex_info* vinfo = &vformat->vinfo; - int* tab = vformat->vs_tab; uint16_t type, swizzle; enum pipe_format format; unsigned i, attrib_count; + int* vs_output_tab = r300->vs->output_stream_loc_swtcl; - /* Vertex shaders have no semantics on their inputs, - * so PSC should just route stuff based on their info, - * and not on attrib information. */ - if (r300screen->caps->has_tcl) { - attrib_count = r300->vs->info.num_inputs; - DBG(r300, DBG_DRAW, "r300: routing %d attribs in psc for vs\n", - attrib_count); - } else { - attrib_count = vinfo->num_attribs; - DBG(r300, DBG_DRAW, "r300: attrib count: %d\n", attrib_count); - for (i = 0; i < attrib_count; i++) { - DBG(r300, DBG_DRAW, "r300: attrib: offset %d, interp %d, size %d," - " tab %d\n", vinfo->attrib[i].src_index, - vinfo->attrib[i].interp_mode, vinfo->attrib[i].emit, - tab[i]); - } + /* For each Draw attribute, route it to the fragment shader according + * to the vs_output_tab. */ + attrib_count = vinfo->num_attribs; + DBG(r300, DBG_DRAW, "r300: attrib count: %d\n", attrib_count); + for (i = 0; i < attrib_count; i++) { + DBG(r300, DBG_DRAW, "r300: attrib: offset %d, interp %d, size %d," + " vs_output_tab %d\n", vinfo->attrib[i].src_index, + vinfo->attrib[i].interp_mode, vinfo->attrib[i].emit, + vs_output_tab[i]); } for (i = 0; i < attrib_count; i++) { /* Make sure we have a proper destination for our attribute. */ - assert(tab[i] != -1); + assert(vs_output_tab[i] != -1); format = draw_translate_vinfo_format(vinfo->attrib[i].emit); /* Obtain the type of data in this attribute. */ type = r300_translate_vertex_data_type(format) | - tab[i] << R300_DST_VEC_LOC_SHIFT; + vs_output_tab[i] << R300_DST_VEC_LOC_SHIFT; /* Obtain the swizzle for this attribute. Note that the default * swizzle in the hardware is not XYZW! */ @@ -272,12 +208,10 @@ static void r300_vertex_psc(struct r300_context* r300, /* Add the attribute to the PSC table. */ if (i & 1) { vformat->vap_prog_stream_cntl[i >> 1] |= type << 16; - vformat->vap_prog_stream_cntl_ext[i >> 1] |= swizzle << 16; } else { - vformat->vap_prog_stream_cntl[i >> 1] |= type << 0; - - vformat->vap_prog_stream_cntl_ext[i >> 1] |= swizzle << 0; + vformat->vap_prog_stream_cntl[i >> 1] |= type; + vformat->vap_prog_stream_cntl_ext[i >> 1] |= swizzle; } } @@ -289,185 +223,200 @@ static void r300_vertex_psc(struct r300_context* r300, (R300_LAST_VEC << (i & 1 ? 16 : 0)); } -/* Set up the mappings from GB to US, for RS block. */ -static void r300_update_fs_tab(struct r300_context* r300, - struct r300_vertex_info* vformat) +static void r300_rs_col(struct r300_rs_block* rs, int id, int ptr, + boolean swizzle_0001) +{ + rs->ip[id] |= R300_RS_COL_PTR(ptr); + if (swizzle_0001) { + rs->ip[id] |= R300_RS_COL_FMT(R300_RS_COL_FMT_0001); + } else { + rs->ip[id] |= R300_RS_COL_FMT(R300_RS_COL_FMT_RGBA); + } + rs->inst[id] |= R300_RS_INST_COL_ID(id); +} + +static void r300_rs_col_write(struct r300_rs_block* rs, int id, int fp_offset) { - struct tgsi_shader_info* info = &r300->fs->info; - int i, cols = 0, texs = 0, cols_emitted = 0; - int* tab = vformat->fs_tab; + rs->inst[id] |= R300_RS_INST_COL_CN_WRITE | + R300_RS_INST_COL_ADDR(fp_offset); +} - for (i = 0; i < 16; i++) { - tab[i] = -1; +static void r300_rs_tex(struct r300_rs_block* rs, int id, int ptr, + boolean swizzle_X001) +{ + if (swizzle_X001) { + rs->ip[id] |= R300_RS_TEX_PTR(ptr*4) | + R300_RS_SEL_S(R300_RS_SEL_C0) | + R300_RS_SEL_T(R300_RS_SEL_K0) | + R300_RS_SEL_R(R300_RS_SEL_K0) | + R300_RS_SEL_Q(R300_RS_SEL_K1); + } else { + rs->ip[id] |= R300_RS_TEX_PTR(ptr*4) | + R300_RS_SEL_S(R300_RS_SEL_C0) | + R300_RS_SEL_T(R300_RS_SEL_C1) | + R300_RS_SEL_R(R300_RS_SEL_C2) | + R300_RS_SEL_Q(R300_RS_SEL_C3); } + rs->inst[id] |= R300_RS_INST_TEX_ID(id); +} - assert(info->num_inputs <= 16); - for (i = 0; i < info->num_inputs; i++) { - switch (info->input_semantic_name[i]) { - case TGSI_SEMANTIC_COLOR: - tab[i] = INTERP_LINEAR; - cols++; - break; - case TGSI_SEMANTIC_POSITION: - case TGSI_SEMANTIC_PSIZE: - debug_printf("r300: Implementation error: Can't use " - "pos attribs in fragshader yet!\n"); - /* Pass through for now */ - case TGSI_SEMANTIC_FOG: - case TGSI_SEMANTIC_GENERIC: - tab[i] = INTERP_PERSPECTIVE; - break; - default: - debug_printf("r300: Unknown vertex input %d\n", - info->input_semantic_name[i]); - break; - } +static void r300_rs_tex_write(struct r300_rs_block* rs, int id, int fp_offset) +{ + rs->inst[id] |= R300_RS_INST_TEX_CN_WRITE | + R300_RS_INST_TEX_ADDR(fp_offset); +} + +static void r500_rs_col(struct r300_rs_block* rs, int id, int ptr, + boolean swizzle_0001) +{ + rs->ip[id] |= R500_RS_COL_PTR(ptr); + if (swizzle_0001) { + rs->ip[id] |= R500_RS_COL_FMT(R300_RS_COL_FMT_0001); + } else { + rs->ip[id] |= R500_RS_COL_FMT(R300_RS_COL_FMT_RGBA); } + rs->inst[id] |= R500_RS_INST_COL_ID(id); +} - /* Now that we know where everything is... */ - DBG(r300, DBG_DRAW, "r300: fp input count: %d\n", info->num_inputs); - for (i = 0; i < info->num_inputs; i++) { - switch (tab[i]) { - case INTERP_LINEAR: - DBG(r300, DBG_DRAW, "r300: attrib: " - "stack offset %d, color, tab %d\n", - i, cols_emitted); - tab[i] = cols_emitted; - cols_emitted++; - break; - case INTERP_PERSPECTIVE: - DBG(r300, DBG_DRAW, "r300: attrib: " - "stack offset %d, texcoord, tab %d\n", - i, cols + texs); - tab[i] = cols + texs; - texs++; - break; - case -1: - debug_printf("r300: Implementation error: Bad fp interp!\n"); - default: - break; - } +static void r500_rs_col_write(struct r300_rs_block* rs, int id, int fp_offset) +{ + rs->inst[id] |= R500_RS_INST_COL_CN_WRITE | + R500_RS_INST_COL_ADDR(fp_offset); +} + +static void r500_rs_tex(struct r300_rs_block* rs, int id, int ptr, + boolean swizzle_X001) +{ + int rs_tex_comp = ptr*4; + + if (swizzle_X001) { + rs->ip[id] |= R500_RS_SEL_S(rs_tex_comp) | + R500_RS_SEL_T(R500_RS_IP_PTR_K0) | + R500_RS_SEL_R(R500_RS_IP_PTR_K0) | + R500_RS_SEL_Q(R500_RS_IP_PTR_K1); + } else { + rs->ip[id] |= R500_RS_SEL_S(rs_tex_comp) | + R500_RS_SEL_T(rs_tex_comp + 1) | + R500_RS_SEL_R(rs_tex_comp + 2) | + R500_RS_SEL_Q(rs_tex_comp + 3); } + rs->inst[id] |= R500_RS_INST_TEX_ID(id); +} +static void r500_rs_tex_write(struct r300_rs_block* rs, int id, int fp_offset) +{ + rs->inst[id] |= R500_RS_INST_TEX_CN_WRITE | + R500_RS_INST_TEX_ADDR(fp_offset); } -/* Set up the RS block. This is the part of the chipset that actually does - * the rasterization of vertices into fragments. This is also the part of the - * chipset that locks up if any part of it is even slightly wrong. */ +/* Set up the RS block. + * + * This is the part of the chipset that actually does the rasterization + * of vertices into fragments. This is also the part of the chipset that + * locks up if any part of it is even slightly wrong. */ static void r300_update_rs_block(struct r300_context* r300, - struct r300_rs_block* rs) + struct r300_shader_semantics* vs_outputs, + struct r300_shader_semantics* fs_inputs) { - struct tgsi_shader_info* info = &r300->fs->info; - int col_count = 0, fp_offset = 0, i, tex_count = 0; - int rs_tex_comp = 0; + struct r300_rs_block* rs = r300->rs_block; + int i, col_count = 0, tex_count = 0, fp_offset = 0; + void (*rX00_rs_col)(struct r300_rs_block*, int, int, boolean); + void (*rX00_rs_col_write)(struct r300_rs_block*, int, int); + void (*rX00_rs_tex)(struct r300_rs_block*, int, int, boolean); + void (*rX00_rs_tex_write)(struct r300_rs_block*, int, int); if (r300_screen(r300->context.screen)->caps->is_r500) { - for (i = 0; i < info->num_inputs; i++) { - switch (info->input_semantic_name[i]) { - case TGSI_SEMANTIC_COLOR: - rs->ip[col_count] |= - R500_RS_COL_PTR(col_count) | - R500_RS_COL_FMT(R300_RS_COL_FMT_RGBA); - col_count++; - break; - case TGSI_SEMANTIC_GENERIC: - rs->ip[tex_count] |= - R500_RS_SEL_S(rs_tex_comp) | - R500_RS_SEL_T(rs_tex_comp + 1) | - R500_RS_SEL_R(rs_tex_comp + 2) | - R500_RS_SEL_Q(rs_tex_comp + 3); - tex_count++; - rs_tex_comp += 4; - break; - default: - break; - } - } + rX00_rs_col = r500_rs_col; + rX00_rs_col_write = r500_rs_col_write; + rX00_rs_tex = r500_rs_tex; + rX00_rs_tex_write = r500_rs_tex_write; + } else { + rX00_rs_col = r300_rs_col; + rX00_rs_col_write = r300_rs_col_write; + rX00_rs_tex = r300_rs_tex; + rX00_rs_tex_write = r300_rs_tex_write; + } - /* Rasterize at least one color, or bad things happen. */ - if ((col_count == 0) && (tex_count == 0)) { - rs->ip[0] |= R500_RS_COL_FMT(R300_RS_COL_FMT_0001); + /* Rasterize colors. */ + for (i = 0; i < ATTR_COLOR_COUNT; i++) { + if (vs_outputs->color[i] != ATTR_UNUSED) { + /* Always rasterize if it's written by the VS, + * otherwise it locks up. */ + rX00_rs_col(rs, col_count, i, FALSE); + + /* Write it to the FS input register if it's used by the FS. */ + if (fs_inputs->color[i] != ATTR_UNUSED) { + rX00_rs_col_write(rs, col_count, fp_offset); + fp_offset++; + } col_count++; - } - - for (i = 0; i < tex_count; i++) { - rs->inst[i] |= R500_RS_INST_TEX_ID(i) | - R500_RS_INST_TEX_CN_WRITE | R500_RS_INST_TEX_ADDR(fp_offset); - fp_offset++; - } - - for (i = 0; i < col_count; i++) { - rs->inst[i] |= R500_RS_INST_COL_ID(i) | - R500_RS_INST_COL_CN_WRITE | R500_RS_INST_COL_ADDR(fp_offset); - fp_offset++; - } - } else { - for (i = 0; i < info->num_inputs; i++) { - switch (info->input_semantic_name[i]) { - case TGSI_SEMANTIC_COLOR: - rs->ip[col_count] |= - R300_RS_COL_PTR(col_count) | - R300_RS_COL_FMT(R300_RS_COL_FMT_RGBA); - col_count++; - break; - case TGSI_SEMANTIC_GENERIC: - rs->ip[tex_count] |= - R300_RS_TEX_PTR(rs_tex_comp) | - R300_RS_SEL_S(R300_RS_SEL_C0) | - R300_RS_SEL_T(R300_RS_SEL_C1) | - R300_RS_SEL_R(R300_RS_SEL_C2) | - R300_RS_SEL_Q(R300_RS_SEL_C3); - tex_count++; - rs_tex_comp+=4; - break; - default: - break; + } else { + /* Skip the FS input register, leave it uninitialized. */ + /* If we try to set it to (0,0,0,1), it will lock up. */ + if (fs_inputs->color[i] != ATTR_UNUSED) { + fp_offset++; } } + } - if (col_count == 0) { - rs->ip[0] |= R300_RS_COL_FMT(R300_RS_COL_FMT_0001); + /* Rasterize texture coordinates. */ + for (i = 0; i < ATTR_GENERIC_COUNT; i++) { + if (vs_outputs->generic[i] != ATTR_UNUSED) { + /* Always rasterize if it's written by the VS, + * otherwise it locks up. */ + rX00_rs_tex(rs, tex_count, tex_count, FALSE); + + /* Write it to the FS input register if it's used by the FS. */ + if (fs_inputs->generic[i] != ATTR_UNUSED) { + rX00_rs_tex_write(rs, tex_count, fp_offset); + fp_offset++; + } + tex_count++; + } else { + /* Skip the FS input register, leave it uninitialized. */ + /* If we try to set it to (0,0,0,1), it will lock up. */ + if (fs_inputs->generic[i] != ATTR_UNUSED) { + fp_offset++; + } } + } - if (tex_count == 0) { - rs->ip[0] |= - R300_RS_SEL_S(R300_RS_SEL_K0) | - R300_RS_SEL_T(R300_RS_SEL_K0) | - R300_RS_SEL_R(R300_RS_SEL_K0) | - R300_RS_SEL_Q(R300_RS_SEL_K1); - } + /* Rasterize fog coordinates. */ + if (vs_outputs->fog != ATTR_UNUSED) { + /* Always rasterize if it's written by the VS, + * otherwise it locks up. */ + rX00_rs_tex(rs, tex_count, tex_count, TRUE); - /* Rasterize at least one color, or bad things happen. */ - if ((col_count == 0) && (tex_count == 0)) { - col_count++; - } - - for (i = 0; i < tex_count; i++) { - rs->inst[i] |= R300_RS_INST_TEX_ID(i) | - R300_RS_INST_TEX_CN_WRITE | R300_RS_INST_TEX_ADDR(fp_offset); + /* Write it to the FS input register if it's used by the FS. */ + if (fs_inputs->fog != ATTR_UNUSED) { + rX00_rs_tex_write(rs, tex_count, fp_offset); fp_offset++; } - - for (i = 0; i < col_count; i++) { - rs->inst[i] |= R300_RS_INST_COL_ID(i) | - R300_RS_INST_COL_CN_WRITE | R300_RS_INST_COL_ADDR(fp_offset); + tex_count++; + } else { + /* Skip the FS input register, leave it uninitialized. */ + /* If we try to set it to (0,0,0,1), it will lock up. */ + if (fs_inputs->fog != ATTR_UNUSED) { fp_offset++; } } - rs->count = (rs_tex_comp) | (col_count << R300_IC_COUNT_SHIFT) | + /* Rasterize at least one color, or bad things happen. */ + if (col_count == 0 && tex_count == 0) { + rX00_rs_col(rs, 0, 0, TRUE); + col_count++; + } + + rs->count = (tex_count*4) | (col_count << R300_IC_COUNT_SHIFT) | R300_HIRES_EN; - rs->inst_count = MAX2(MAX2(col_count - 1, tex_count - 1), 0); + rs->inst_count = MAX3(col_count - 1, tex_count - 1, 0); } /* Update the vertex format. */ static void r300_update_derived_shader_state(struct r300_context* r300) { struct r300_screen* r300screen = r300_screen(r300->context.screen); - struct r300_vertex_info* vformat; - struct r300_rs_block* rs_block; - int i; /* struct r300_shader_key* key; @@ -495,27 +444,22 @@ static void r300_update_derived_shader_state(struct r300_context* r300) (void*)key, (void*)value); } */ - /* XXX This will be refactored ASAP. */ - vformat = CALLOC_STRUCT(r300_vertex_info); - rs_block = CALLOC_STRUCT(r300_rs_block); + /* Reset structures */ + memset(r300->rs_block, 0, sizeof(struct r300_rs_block)); + memset(r300->vertex_info, 0, sizeof(struct r300_vertex_info)); + memcpy(r300->vertex_info->vinfo.hwfmt, r300->vs->hwfmt, sizeof(uint)*4); - for (i = 0; i < 16; i++) { - vformat->vs_tab[i] = -1; - vformat->fs_tab[i] = -1; - } - - r300_vs_tab_routes(r300, vformat); - r300_vertex_psc(r300, vformat); - r300_update_fs_tab(r300, vformat); + r300_update_rs_block(r300, &r300->vs->outputs, &r300->fs->inputs); - r300_update_rs_block(r300, rs_block); - - FREE(r300->vertex_info); - FREE(r300->rs_block); + if (r300screen->caps->has_tcl) { + r300_vertex_psc(r300); + } else { + r300_draw_emit_all_attribs(r300); + draw_compute_vertex_size(&r300->vertex_info->vinfo); + r300_swtcl_vertex_psc(r300); + } - r300->vertex_info = vformat; - r300->rs_block = rs_block; - r300->dirty_state |= (R300_NEW_VERTEX_FORMAT | R300_NEW_RS_BLOCK); + r300->dirty_state |= R300_NEW_RS_BLOCK; } static void r300_update_ztop(struct r300_context* r300) @@ -553,9 +497,9 @@ static void r300_update_ztop(struct r300_context* r300) void r300_update_derived_state(struct r300_context* r300) { - /* XXX */ - if (TRUE || r300->dirty_state & - (R300_NEW_FRAGMENT_SHADER | R300_NEW_VERTEX_SHADER)) { + if (r300->dirty_state & + (R300_NEW_FRAGMENT_SHADER | R300_NEW_VERTEX_SHADER | + R300_NEW_VERTEX_FORMAT)) { r300_update_derived_shader_state(r300); } diff --git a/src/gallium/drivers/r300/r300_state_invariant.c b/src/gallium/drivers/r300/r300_state_invariant.c index c07e6ae676d..46d1cb39b54 100644 --- a/src/gallium/drivers/r300/r300_state_invariant.c +++ b/src/gallium/drivers/r300/r300_state_invariant.c @@ -84,7 +84,7 @@ void r300_emit_invariant_state(struct r300_context* r300) END_CS; /* XXX unsorted stuff from surface_fill */ - BEGIN_CS(60 + (caps->has_tcl ? 5 : 0) + (caps->is_r500 ? 4 : 0)); + BEGIN_CS(56 + (caps->has_tcl ? 5 : 0) + (caps->is_r500 ? 4 : 0)); /* Flush PVS. */ OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG, 0x0); @@ -135,8 +135,6 @@ void r300_emit_invariant_state(struct r300_context* r300) OUT_CS_REG(R300_ZB_DEPTHCLEARVALUE, 0x00000000); OUT_CS_REG(R300_ZB_HIZ_OFFSET, 0x00000000); OUT_CS_REG(R300_ZB_HIZ_PITCH, 0x00000000); - OUT_CS_REG(R300_VAP_VTX_STATE_CNTL, 0x1); - OUT_CS_REG(R300_VAP_VSM_VTX_ASSM, 0x405); OUT_CS_REG(R300_SE_VTE_CNTL, 0x0000043F); /* XXX */ diff --git a/src/gallium/drivers/r300/r300_texture.c b/src/gallium/drivers/r300/r300_texture.c index aea25cf71dd..093a21ebe24 100644 --- a/src/gallium/drivers/r300/r300_texture.c +++ b/src/gallium/drivers/r300/r300_texture.c @@ -34,8 +34,8 @@ static void r300_setup_texture_state(struct r300_texture* tex, boolean is_r500) struct r300_texture_state* state = &tex->state; struct pipe_texture *pt = &tex->tex; - state->format0 = R300_TX_WIDTH((pt->width[0] - 1) & 0x7ff) | - R300_TX_HEIGHT((pt->height[0] - 1) & 0x7ff); + state->format0 = R300_TX_WIDTH((pt->width0 - 1) & 0x7ff) | + R300_TX_HEIGHT((pt->height0 - 1) & 0x7ff); if (tex->is_npot) { /* rectangles love this */ @@ -43,8 +43,7 @@ static void r300_setup_texture_state(struct r300_texture* tex, boolean is_r500) state->format2 = (tex->pitch[0] - 1) & 0x1fff; } else { /* power of two textures (3D, mipmaps, and no pitch) */ - state->format0 |= R300_TX_DEPTH(util_logbase2(pt->depth[0]) & 0xf) | - R300_TX_NUM_LEVELS(pt->last_level & 0xf); + state->format0 |= R300_TX_DEPTH(util_logbase2(pt->depth0) & 0xf); } state->format1 = r300_translate_texformat(pt->format); @@ -58,17 +57,17 @@ static void r300_setup_texture_state(struct r300_texture* tex, boolean is_r500) /* large textures on r500 */ if (is_r500) { - if (pt->width[0] > 2048) { + if (pt->width0 > 2048) { state->format2 |= R500_TXWIDTH_BIT11; } - if (pt->height[0] > 2048) { + if (pt->height0 > 2048) { state->format2 |= R500_TXHEIGHT_BIT11; } } - assert(is_r500 || (pt->width[0] <= 2048 && pt->height[0] <= 2048)); + assert(is_r500 || (pt->width0 <= 2048 && pt->height0 <= 2048)); debug_printf("r300: Set texture state (%dx%d, %d levels)\n", - pt->width[0], pt->height[0], pt->last_level); + pt->width0, pt->height0, pt->last_level); } unsigned r300_texture_get_offset(struct r300_texture* tex, unsigned level, @@ -106,7 +105,7 @@ unsigned r300_texture_get_stride(struct r300_texture* tex, unsigned level) return 0; } - return align(pf_get_stride(&tex->tex.block, tex->tex.width[level]), 32); + return align(pf_get_stride(&tex->tex.block, u_minify(tex->tex.width0, level)), 32); } static void r300_setup_miptree(struct r300_texture* tex) @@ -116,14 +115,8 @@ static void r300_setup_miptree(struct r300_texture* tex) int i; for (i = 0; i <= base->last_level; i++) { - if (i > 0) { - base->width[i] = minify(base->width[i-1]); - base->height[i] = minify(base->height[i-1]); - base->depth[i] = minify(base->depth[i-1]); - } - - base->nblocksx[i] = pf_get_nblocksx(&base->block, base->width[i]); - base->nblocksy[i] = pf_get_nblocksy(&base->block, base->height[i]); + base->nblocksx[i] = pf_get_nblocksx(&base->block, u_minify(base->width0, i)); + base->nblocksy[i] = pf_get_nblocksy(&base->block, u_minify(base->height0, i)); stride = r300_texture_get_stride(tex, i); layer_size = stride * base->nblocksy[i]; @@ -131,7 +124,7 @@ static void r300_setup_miptree(struct r300_texture* tex) if (base->target == PIPE_TEXTURE_CUBE) size = layer_size * 6; else - size = layer_size * base->depth[i]; + size = layer_size * u_minify(base->depth0, i); tex->offset[i] = align(tex->size, 32); tex->size = tex->offset[i] + size; @@ -140,15 +133,15 @@ static void r300_setup_miptree(struct r300_texture* tex) debug_printf("r300: Texture miptree: Level %d " "(%dx%dx%d px, pitch %d bytes)\n", - i, base->width[i], base->height[i], base->depth[i], - stride); + i, u_minify(base->width0, i), u_minify(base->height0, i), + u_minify(base->depth0, i), stride); } } static void r300_setup_flags(struct r300_texture* tex) { - tex->is_npot = !util_is_power_of_two(tex->tex.width[0]) || - !util_is_power_of_two(tex->tex.height[0]); + tex->is_npot = !util_is_power_of_two(tex->tex.width0) || + !util_is_power_of_two(tex->tex.height0); } /* Create a new texture. */ @@ -208,8 +201,8 @@ static struct pipe_surface* r300_get_tex_surface(struct pipe_screen* screen, pipe_reference_init(&surface->reference, 1); pipe_texture_reference(&surface->texture, texture); surface->format = texture->format; - surface->width = texture->width[level]; - surface->height = texture->height[level]; + surface->width = u_minify(texture->width0, level); + surface->height = u_minify(texture->height0, level); surface->offset = offset; surface->usage = flags; surface->zslice = zslice; @@ -237,7 +230,7 @@ static struct pipe_texture* /* Support only 2D textures without mipmaps */ if (base->target != PIPE_TEXTURE_2D || - base->depth[0] != 1 || + base->depth0 != 1 || base->last_level != 0) { return NULL; } @@ -287,9 +280,9 @@ r300_video_surface_create(struct pipe_screen *screen, template.target = PIPE_TEXTURE_2D; template.format = PIPE_FORMAT_X8R8G8B8_UNORM; template.last_level = 0; - template.width[0] = util_next_power_of_two(width); - template.height[0] = util_next_power_of_two(height); - template.depth[0] = 1; + template.width0 = util_next_power_of_two(width); + template.height0 = util_next_power_of_two(height); + template.depth0 = 1; pf_get_block(template.format, &template.block); template.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER | PIPE_TEXTURE_USAGE_RENDER_TARGET; diff --git a/src/gallium/drivers/r300/r300_vbo.c b/src/gallium/drivers/r300/r300_vbo.c index a6a159667a3..6ebaf715dc5 100644 --- a/src/gallium/drivers/r300/r300_vbo.c +++ b/src/gallium/drivers/r300/r300_vbo.c @@ -34,52 +34,6 @@ #include "r300_reg.h" #include "r300_winsys.h" -static INLINE void setup_vertex_attribute(struct r300_vertex_info *vinfo, - struct pipe_vertex_element *vert_elem, - unsigned attr_num) -{ - uint16_t hw_fmt1, hw_fmt2; - - hw_fmt1 = r300_translate_vertex_data_type(vert_elem->src_format) | - (attr_num << R300_DST_VEC_LOC_SHIFT); - hw_fmt2 = r300_translate_vertex_data_swizzle(vert_elem->src_format); - - if (attr_num % 2 == 0) - { - vinfo->vap_prog_stream_cntl[attr_num >> 1] = hw_fmt1; - vinfo->vap_prog_stream_cntl_ext[attr_num >> 1] = hw_fmt2; - } - else - { - vinfo->vap_prog_stream_cntl[attr_num >> 1] |= hw_fmt1 << 16; - vinfo->vap_prog_stream_cntl_ext[attr_num >> 1] |= hw_fmt2 << 16; - } -} - -static void finish_vertex_attribs_setup(struct r300_vertex_info *vinfo, - unsigned attribs_num) -{ - uint32_t last_vec_bit = (attribs_num % 2 == 0) ? - (R300_LAST_VEC << 16) : R300_LAST_VEC; - - assert(attribs_num > 0 && attribs_num <= 16); - vinfo->vap_prog_stream_cntl[(attribs_num - 1) >> 1] |= last_vec_bit; -} - -void setup_vertex_attributes(struct r300_context *r300) -{ - struct pipe_vertex_element *vert_elem; - int i; - - for (i = 0; i < r300->vertex_element_count; i++) { - vert_elem = &r300->vertex_element[i]; - setup_vertex_attribute(r300->vertex_info, vert_elem, i); - } - - finish_vertex_attribs_setup(r300->vertex_info, - r300->vertex_element_count); -} - static INLINE int get_buffer_offset(struct r300_context *r300, unsigned int buf_nr, unsigned int elem_offset) diff --git a/src/gallium/drivers/r300/r300_vs.c b/src/gallium/drivers/r300/r300_vs.c index 74ef416dc14..31248346bc6 100644 --- a/src/gallium/drivers/r300/r300_vs.c +++ b/src/gallium/drivers/r300/r300_vs.c @@ -1,5 +1,6 @@ /* * Copyright 2009 Corbin Simpson <[email protected]> + * Copyright 2009 Marek Olšák <[email protected]> * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -23,89 +24,226 @@ #include "r300_vs.h" #include "r300_context.h" +#include "r300_screen.h" #include "r300_tgsi_to_rc.h" +#include "r300_reg.h" #include "tgsi/tgsi_dump.h" #include "tgsi/tgsi_parse.h" #include "radeon_compiler.h" - -static void set_vertex_inputs_outputs(struct r300_vertex_program_compiler * c) +/* Convert info about VS output semantics into r300_shader_semantics. */ +static void r300_shader_read_vs_outputs( + struct tgsi_shader_info* info, + struct r300_shader_semantics* vs_outputs) { - struct r300_vertex_shader * vs = c->UserData; - struct tgsi_shader_info* info = &vs->info; - struct tgsi_parse_context parser; - struct tgsi_full_declaration * decl; - boolean pointsize = FALSE; - int out_colors = 0; - int colors = 0; - int out_generic = 0; - int generic = 0; int i; + unsigned index; - /* Fill in the input mapping */ - for (i = 0; i < info->num_inputs; i++) - c->code->inputs[i] = i; + r300_shader_semantics_reset(vs_outputs); - /* Fill in the output mapping */ for (i = 0; i < info->num_outputs; i++) { + index = info->output_semantic_index[i]; + switch (info->output_semantic_name[i]) { + case TGSI_SEMANTIC_POSITION: + assert(index == 0); + vs_outputs->pos = i; + break; + case TGSI_SEMANTIC_PSIZE: - pointsize = TRUE; + assert(index == 0); + vs_outputs->psize = i; break; + case TGSI_SEMANTIC_COLOR: - out_colors++; + assert(index <= ATTR_COLOR_COUNT); + vs_outputs->color[index] = i; break; - case TGSI_SEMANTIC_FOG: + + case TGSI_SEMANTIC_BCOLOR: + assert(index <= ATTR_COLOR_COUNT); + vs_outputs->bcolor[index] = i; + break; + case TGSI_SEMANTIC_GENERIC: - out_generic++; + assert(index <= ATTR_GENERIC_COUNT); + vs_outputs->generic[index] = i; break; + + case TGSI_SEMANTIC_FOG: + assert(index == 0); + vs_outputs->fog = i; + break; + + default: + assert(0); } } +} - tgsi_parse_init(&parser, vs->state.tokens); +static void r300_shader_vap_output_fmt( + struct r300_shader_semantics* vs_outputs, + uint* hwfmt) +{ + int i, gen_count; - while (!tgsi_parse_end_of_tokens(&parser)) { - tgsi_parse_token(&parser); + /* Do the actual vertex_info setup. + * + * vertex_info has four uints of hardware-specific data in it. + * vinfo.hwfmt[0] is R300_VAP_VTX_STATE_CNTL + * vinfo.hwfmt[1] is R300_VAP_VSM_VTX_ASSM + * vinfo.hwfmt[2] is R300_VAP_OUTPUT_VTX_FMT_0 + * vinfo.hwfmt[3] is R300_VAP_OUTPUT_VTX_FMT_1 */ - if (parser.FullToken.Token.Type != TGSI_TOKEN_TYPE_DECLARATION) - continue; + hwfmt[0] = 0x5555; /* XXX this is classic Mesa bonghits */ - decl = &parser.FullToken.FullDeclaration; + /* Position. */ + if (vs_outputs->pos != ATTR_UNUSED) { + hwfmt[1] |= R300_INPUT_CNTL_POS; + hwfmt[2] |= R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT; + } else { + assert(0); + } - if (decl->Declaration.File != TGSI_FILE_OUTPUT) - continue; + /* Point size. */ + if (vs_outputs->psize != ATTR_UNUSED) { + hwfmt[2] |= R300_VAP_OUTPUT_VTX_FMT_0__PT_SIZE_PRESENT; + } - switch (decl->Semantic.SemanticName) { - case TGSI_SEMANTIC_POSITION: - c->code->outputs[decl->DeclarationRange.First] = 0; - break; - case TGSI_SEMANTIC_PSIZE: - c->code->outputs[decl->DeclarationRange.First] = 1; - break; - case TGSI_SEMANTIC_COLOR: - c->code->outputs[decl->DeclarationRange.First] = 1 + - (pointsize ? 1 : 0) + - colors++; - break; - case TGSI_SEMANTIC_FOG: - case TGSI_SEMANTIC_GENERIC: - c->code->outputs[decl->DeclarationRange.First] = 1 + - (pointsize ? 1 : 0) + - out_colors + - generic++; - break; - default: - debug_printf("r300: vs: Bad semantic declaration %d\n", - decl->Semantic.SemanticName); - break; + /* Colors. */ + for (i = 0; i < ATTR_COLOR_COUNT; i++) { + if (vs_outputs->color[i] != ATTR_UNUSED) { + hwfmt[1] |= R300_INPUT_CNTL_COLOR; + hwfmt[2] |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT << i; + } + } + + /* XXX Back-face colors. */ + + /* Texture coordinates. */ + gen_count = 0; + for (i = 0; i < ATTR_GENERIC_COUNT; i++) { + if (vs_outputs->generic[i] != ATTR_UNUSED) { + hwfmt[1] |= (R300_INPUT_CNTL_TC0 << gen_count); + hwfmt[3] |= (4 << (3 * gen_count)); + gen_count++; + } + } + + /* Fog coordinates. */ + if (vs_outputs->fog != ATTR_UNUSED) { + hwfmt[1] |= (R300_INPUT_CNTL_TC0 << gen_count); + hwfmt[3] |= (4 << (3 * gen_count)); + gen_count++; + } + + /* XXX magic */ + assert(gen_count <= 8); +} + +/* Set VS output stream locations for SWTCL. */ +static void r300_stream_locations_swtcl( + struct r300_shader_semantics* vs_outputs, + int* output_stream_loc) +{ + int i, tabi = 0, gen_count; + + /* XXX Check whether the numbers (0, 1, 2+i, etc.) are correct. + * These should go to VAP_PROG_STREAM_CNTL/DST_VEC_LOC. */ + + /* Position. */ + output_stream_loc[tabi++] = 0; + + /* Point size. */ + if (vs_outputs->psize != ATTR_UNUSED) { + output_stream_loc[tabi++] = 1; + } + + /* Colors. */ + for (i = 0; i < ATTR_COLOR_COUNT; i++) { + if (vs_outputs->color[i] != ATTR_UNUSED) { + output_stream_loc[tabi++] = 2 + i; + } + } + + /* Back-face colors. */ + for (i = 0; i < ATTR_COLOR_COUNT; i++) { + if (vs_outputs->bcolor[i] != ATTR_UNUSED) { + output_stream_loc[tabi++] = 4 + i; + } + } + + /* Texture coordinates. */ + gen_count = 0; + for (i = 0; i < ATTR_GENERIC_COUNT; i++) { + if (vs_outputs->bcolor[i] != ATTR_UNUSED) { + assert(tabi < 16); + output_stream_loc[tabi++] = 6 + gen_count; + gen_count++; } } - tgsi_parse_free(&parser); + /* Fog coordinates. */ + if (vs_outputs->fog != ATTR_UNUSED) { + assert(tabi < 16); + output_stream_loc[tabi++] = 6 + gen_count; + gen_count++; + } + + /* XXX magic */ + assert(gen_count <= 8); + + for (; tabi < 16;) { + output_stream_loc[tabi++] = -1; + } } +static void set_vertex_inputs_outputs(struct r300_vertex_program_compiler * c) +{ + struct r300_vertex_shader * vs = c->UserData; + struct r300_shader_semantics* outputs = &vs->outputs; + struct tgsi_shader_info* info = &vs->info; + int i, reg = 0; + + /* Fill in the input mapping */ + for (i = 0; i < info->num_inputs; i++) + c->code->inputs[i] = i; + + /* Position. */ + if (outputs->pos != ATTR_UNUSED) { + c->code->outputs[outputs->pos] = reg++; + } else { + assert(0); + } + + /* Point size. */ + if (outputs->psize != ATTR_UNUSED) { + c->code->outputs[outputs->psize] = reg++; + } + + /* Colors. */ + for (i = 0; i < ATTR_COLOR_COUNT; i++) { + if (outputs->color[i] != ATTR_UNUSED) { + c->code->outputs[outputs->color[i]] = reg++; + } + } + + /* XXX Back-face colors. */ + + /* Texture coordinates. */ + for (i = 0; i < ATTR_GENERIC_COUNT; i++) { + if (outputs->generic[i] != ATTR_UNUSED) { + c->code->outputs[outputs->generic[i]] = reg++; + } + } + + /* Fog coordinates. */ + if (outputs->fog != ATTR_UNUSED) { + c->code->outputs[outputs->fog] = reg++; + } +} void r300_translate_vertex_shader(struct r300_context* r300, struct r300_vertex_shader* vs) @@ -113,6 +251,14 @@ void r300_translate_vertex_shader(struct r300_context* r300, struct r300_vertex_program_compiler compiler; struct tgsi_to_rc ttr; + /* Initialize. */ + r300_shader_read_vs_outputs(&vs->info, &vs->outputs); + r300_shader_vap_output_fmt(&vs->outputs, vs->hwfmt); + + if (!r300_screen(r300->context.screen)->caps->has_tcl) { + r300_stream_locations_swtcl(&vs->outputs, vs->output_stream_loc_swtcl); + } + /* Setup the compiler */ rc_init(&compiler.Base); @@ -137,7 +283,7 @@ void r300_translate_vertex_shader(struct r300_context* r300, /* Invoke the compiler */ r3xx_compile_vertex_program(&compiler); if (compiler.Base.Error) { - /* Todo: Fail gracefully */ + /* XXX Fail gracefully */ fprintf(stderr, "r300 VP: Compiler error\n"); abort(); } diff --git a/src/gallium/drivers/r300/r300_vs.h b/src/gallium/drivers/r300/r300_vs.h index 2a4ce315e32..283dd5a9e83 100644 --- a/src/gallium/drivers/r300/r300_vs.h +++ b/src/gallium/drivers/r300/r300_vs.h @@ -1,5 +1,6 @@ /* * Copyright 2009 Corbin Simpson <[email protected]> + * Copyright 2009 Marek Olšák <[email protected]> * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -25,18 +26,20 @@ #include "pipe/p_state.h" #include "tgsi/tgsi_scan.h" - #include "radeon_code.h" +#include "r300_shader_semantics.h" + struct r300_context; struct r300_vertex_shader { /* Parent class */ struct pipe_shader_state state; - struct tgsi_shader_info info; - /* Fallback shader, because Draw has issues */ - struct draw_vertex_shader* draw; + struct tgsi_shader_info info; + struct r300_shader_semantics outputs; + int output_stream_loc_swtcl[16]; + uint hwfmt[4]; /* Has this shader been translated yet? */ boolean translated; @@ -45,9 +48,6 @@ struct r300_vertex_shader { struct r300_vertex_program_code code; }; - -extern struct r300_vertex_program_code r300_passthrough_vertex_shader; - void r300_translate_vertex_shader(struct r300_context* r300, struct r300_vertex_shader* vs); diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index 5f60139968a..f8bf3e9974e 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -90,14 +90,15 @@ softpipe_destroy( struct pipe_context *pipe ) if (softpipe->draw) draw_destroy( softpipe->draw ); - softpipe->quad.shade->destroy( softpipe->quad.shade ); - softpipe->quad.depth_test->destroy( softpipe->quad.depth_test ); - softpipe->quad.blend->destroy( softpipe->quad.blend ); + softpipe->quad.shade->destroy( softpipe->quad.shade ); + softpipe->quad.depth_test->destroy( softpipe->quad.depth_test ); + softpipe->quad.blend->destroy( softpipe->quad.blend ); for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) { sp_destroy_tile_cache(softpipe->cbuf_cache[i]); pipe_surface_reference(&softpipe->framebuffer.cbufs[i], NULL); } + sp_destroy_tile_cache(softpipe->zsbuf_cache); pipe_surface_reference(&softpipe->framebuffer.zsbuf, NULL); @@ -106,6 +107,11 @@ softpipe_destroy( struct pipe_context *pipe ) pipe_texture_reference(&softpipe->texture[i], NULL); } + for (i = 0; i < PIPE_MAX_VERTEX_SAMPLERS; i++) { + sp_destroy_tex_tile_cache(softpipe->vertex_tex_cache[i]); + pipe_texture_reference(&softpipe->vertex_textures[i], NULL); + } + for (i = 0; i < Elements(softpipe->constants); i++) { if (softpipe->constants[i].buffer) { pipe_buffer_reference(&softpipe->constants[i].buffer, NULL); @@ -152,6 +158,11 @@ softpipe_is_texture_referenced( struct pipe_context *pipe, softpipe->tex_cache[i]->texture == texture) return PIPE_REFERENCED_FOR_READ; } + for (i = 0; i < PIPE_MAX_VERTEX_SAMPLERS; i++) { + if (softpipe->vertex_tex_cache[i] && + softpipe->vertex_tex_cache[i]->texture == texture) + return PIPE_REFERENCED_FOR_READ; + } return PIPE_UNREFERENCED; } @@ -191,7 +202,8 @@ softpipe_create( struct pipe_screen *screen ) softpipe->pipe.delete_blend_state = softpipe_delete_blend_state; softpipe->pipe.create_sampler_state = softpipe_create_sampler_state; - softpipe->pipe.bind_sampler_states = softpipe_bind_sampler_states; + softpipe->pipe.bind_fragment_sampler_states = softpipe_bind_sampler_states; + softpipe->pipe.bind_vertex_sampler_states = softpipe_bind_vertex_sampler_states; softpipe->pipe.delete_sampler_state = softpipe_delete_sampler_state; softpipe->pipe.create_depth_stencil_alpha_state = softpipe_create_depth_stencil_state; @@ -216,7 +228,8 @@ softpipe_create( struct pipe_screen *screen ) softpipe->pipe.set_framebuffer_state = softpipe_set_framebuffer_state; softpipe->pipe.set_polygon_stipple = softpipe_set_polygon_stipple; softpipe->pipe.set_scissor_state = softpipe_set_scissor_state; - softpipe->pipe.set_sampler_textures = softpipe_set_sampler_textures; + softpipe->pipe.set_fragment_sampler_textures = softpipe_set_sampler_textures; + softpipe->pipe.set_vertex_sampler_textures = softpipe_set_vertex_sampler_textures; softpipe->pipe.set_viewport_state = softpipe_set_viewport_state; softpipe->pipe.set_vertex_buffers = softpipe_set_vertex_buffers; @@ -246,7 +259,9 @@ softpipe_create( struct pipe_screen *screen ) for (i = 0; i < PIPE_MAX_SAMPLERS; i++) softpipe->tex_cache[i] = sp_create_tex_tile_cache( screen ); - + for (i = 0; i < PIPE_MAX_VERTEX_SAMPLERS; i++) { + softpipe->vertex_tex_cache[i] = sp_create_tex_tile_cache(screen); + } /* setup quad rendering stages */ softpipe->quad.shade = sp_quad_shade_stage(softpipe); @@ -262,7 +277,7 @@ softpipe_create( struct pipe_screen *screen ) goto fail; draw_texture_samplers(softpipe->draw, - PIPE_MAX_SAMPLERS, + PIPE_MAX_VERTEX_SAMPLERS, (struct tgsi_sampler **) softpipe->tgsi.vert_samplers_list); diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h index a735573d6fb..8ce20c5744c 100644 --- a/src/gallium/drivers/softpipe/sp_context.h +++ b/src/gallium/drivers/softpipe/sp_context.h @@ -53,6 +53,7 @@ struct softpipe_context { /** Constant state objects */ struct pipe_blend_state *blend; struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; + struct pipe_sampler_state *vertex_samplers[PIPE_MAX_VERTEX_SAMPLERS]; struct pipe_depth_stencil_alpha_state *depth_stencil; struct pipe_rasterizer_state *rasterizer; struct sp_fragment_shader *fs; @@ -66,12 +67,15 @@ struct softpipe_context { struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; struct pipe_texture *texture[PIPE_MAX_SAMPLERS]; + struct pipe_texture *vertex_textures[PIPE_MAX_VERTEX_SAMPLERS]; struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS]; struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS]; unsigned num_samplers; unsigned num_textures; + unsigned num_vertex_samplers; + unsigned num_vertex_textures; unsigned num_vertex_elements; unsigned num_vertex_buffers; @@ -121,7 +125,7 @@ struct softpipe_context { /** TGSI exec things */ struct { - struct sp_sampler_varient *vert_samplers_list[PIPE_MAX_SAMPLERS]; + struct sp_sampler_varient *vert_samplers_list[PIPE_MAX_VERTEX_SAMPLERS]; struct sp_sampler_varient *frag_samplers_list[PIPE_MAX_SAMPLERS]; } tgsi; @@ -139,6 +143,7 @@ struct softpipe_context { unsigned tex_timestamp; struct softpipe_tex_tile_cache *tex_cache[PIPE_MAX_SAMPLERS]; + struct softpipe_tex_tile_cache *vertex_tex_cache[PIPE_MAX_VERTEX_SAMPLERS]; unsigned use_sse : 1; unsigned dump_fs : 1; diff --git a/src/gallium/drivers/softpipe/sp_flush.c b/src/gallium/drivers/softpipe/sp_flush.c index e38b767cf2c..75dac810a12 100644 --- a/src/gallium/drivers/softpipe/sp_flush.c +++ b/src/gallium/drivers/softpipe/sp_flush.c @@ -55,6 +55,9 @@ softpipe_flush( struct pipe_context *pipe, for (i = 0; i < softpipe->num_textures; i++) { sp_flush_tex_tile_cache(softpipe->tex_cache[i]); } + for (i = 0; i < softpipe->num_vertex_textures; i++) { + sp_flush_tex_tile_cache(softpipe->vertex_tex_cache[i]); + } } if (flags & PIPE_FLUSH_SWAPBUFFERS) { diff --git a/src/gallium/drivers/softpipe/sp_fs_exec.c b/src/gallium/drivers/softpipe/sp_fs_exec.c index 4076114d392..a8999ed3479 100644 --- a/src/gallium/drivers/softpipe/sp_fs_exec.c +++ b/src/gallium/drivers/softpipe/sp_fs_exec.c @@ -126,7 +126,13 @@ exec_run( const struct sp_fragment_shader *base, setup_pos_vector(quad->posCoef, (float)quad->input.x0, (float)quad->input.y0, &machine->QuadPos); - + + if (quad->input.facing) { + machine->Face = -1.0f; + } else { + machine->Face = 1.0f; + } + quad->inout.mask &= tgsi_exec_machine_run( machine ); if (quad->inout.mask == 0) return FALSE; diff --git a/src/gallium/drivers/softpipe/sp_screen.c b/src/gallium/drivers/softpipe/sp_screen.c index 81fb7aa20c6..6bf3df8e6ad 100644 --- a/src/gallium/drivers/softpipe/sp_screen.c +++ b/src/gallium/drivers/softpipe/sp_screen.c @@ -58,7 +58,9 @@ softpipe_get_param(struct pipe_screen *screen, int param) case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: return PIPE_MAX_SAMPLERS; case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS: - return PIPE_MAX_SAMPLERS; + return PIPE_MAX_VERTEX_SAMPLERS; + case PIPE_CAP_MAX_COMBINED_SAMPLERS: + return PIPE_MAX_SAMPLERS + PIPE_MAX_VERTEX_SAMPLERS; case PIPE_CAP_NPOT_TEXTURES: return 1; case PIPE_CAP_TWO_SIDED_STENCIL: @@ -143,6 +145,10 @@ softpipe_is_format_supported( struct pipe_screen *screen, case PIPE_FORMAT_DXT3_RGBA: case PIPE_FORMAT_DXT5_RGBA: case PIPE_FORMAT_Z32_FLOAT: + case PIPE_FORMAT_R8G8_SNORM: + case PIPE_FORMAT_B6UG5SR5S_NORM: + case PIPE_FORMAT_X8UB8UG8SR8S_NORM: + case PIPE_FORMAT_A8B8G8R8_SNORM: return FALSE; default: return TRUE; diff --git a/src/gallium/drivers/softpipe/sp_state.h b/src/gallium/drivers/softpipe/sp_state.h index 77ee3c1136b..d488fb8710b 100644 --- a/src/gallium/drivers/softpipe/sp_state.h +++ b/src/gallium/drivers/softpipe/sp_state.h @@ -104,6 +104,10 @@ void * softpipe_create_sampler_state(struct pipe_context *, const struct pipe_sampler_state *); void softpipe_bind_sampler_states(struct pipe_context *, unsigned, void **); +void +softpipe_bind_vertex_sampler_states(struct pipe_context *, + unsigned num_samplers, + void **samplers); void softpipe_delete_sampler_state(struct pipe_context *, void *); void * @@ -150,6 +154,11 @@ void softpipe_set_sampler_textures( struct pipe_context *, unsigned num, struct pipe_texture ** ); +void +softpipe_set_vertex_sampler_textures(struct pipe_context *, + unsigned num_textures, + struct pipe_texture **); + void softpipe_set_viewport_state( struct pipe_context *, const struct pipe_viewport_state * ); diff --git a/src/gallium/drivers/softpipe/sp_state_derived.c b/src/gallium/drivers/softpipe/sp_state_derived.c index 3bc96b95385..c24a737d07b 100644 --- a/src/gallium/drivers/softpipe/sp_state_derived.c +++ b/src/gallium/drivers/softpipe/sp_state_derived.c @@ -213,6 +213,19 @@ update_tgsi_samplers( struct softpipe_context *softpipe ) } } } + + for (i = 0; i < PIPE_MAX_VERTEX_SAMPLERS; i++) { + struct softpipe_tex_tile_cache *tc = softpipe->vertex_tex_cache[i]; + + if (tc->texture) { + struct softpipe_texture *spt = softpipe_texture(tc->texture); + + if (spt->timestamp != tc->timestamp) { + sp_tex_tile_cache_validate_texture(tc); + tc->timestamp = spt->timestamp; + } + } + } } diff --git a/src/gallium/drivers/softpipe/sp_state_sampler.c b/src/gallium/drivers/softpipe/sp_state_sampler.c index db0b8ab76b1..ceb4e338f1a 100644 --- a/src/gallium/drivers/softpipe/sp_state_sampler.c +++ b/src/gallium/drivers/softpipe/sp_state_sampler.c @@ -94,6 +94,34 @@ softpipe_bind_sampler_states(struct pipe_context *pipe, void +softpipe_bind_vertex_sampler_states(struct pipe_context *pipe, + unsigned num_samplers, + void **samplers) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + unsigned i; + + assert(num_samplers <= PIPE_MAX_VERTEX_SAMPLERS); + + /* Check for no-op */ + if (num_samplers == softpipe->num_vertex_samplers && + !memcmp(softpipe->vertex_samplers, samplers, num_samplers * sizeof(void *))) + return; + + draw_flush(softpipe->draw); + + for (i = 0; i < num_samplers; ++i) + softpipe->vertex_samplers[i] = samplers[i]; + for (i = num_samplers; i < PIPE_MAX_VERTEX_SAMPLERS; ++i) + softpipe->vertex_samplers[i] = NULL; + + softpipe->num_vertex_samplers = num_samplers; + + softpipe->dirty |= SP_NEW_SAMPLER; +} + + +void softpipe_set_sampler_textures(struct pipe_context *pipe, unsigned num, struct pipe_texture **texture) { @@ -122,6 +150,37 @@ softpipe_set_sampler_textures(struct pipe_context *pipe, } +void +softpipe_set_vertex_sampler_textures(struct pipe_context *pipe, + unsigned num_textures, + struct pipe_texture **textures) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + uint i; + + assert(num_textures <= PIPE_MAX_VERTEX_SAMPLERS); + + /* Check for no-op */ + if (num_textures == softpipe->num_vertex_textures && + !memcmp(softpipe->vertex_textures, textures, num_textures * sizeof(struct pipe_texture *))) { + return; + } + + draw_flush(softpipe->draw); + + for (i = 0; i < PIPE_MAX_VERTEX_SAMPLERS; i++) { + struct pipe_texture *tex = i < num_textures ? textures[i] : NULL; + + pipe_texture_reference(&softpipe->vertex_textures[i], tex); + sp_tex_tile_cache_set_texture(softpipe->vertex_tex_cache[i], tex); + } + + softpipe->num_vertex_textures = num_textures; + + softpipe->dirty |= SP_NEW_TEXTURE; +} + + /** * Find/create an sp_sampler_varient object for sampling the given texture, * sampler and tex unit. @@ -185,16 +244,16 @@ softpipe_reset_sampler_varients(struct softpipe_context *softpipe) * fragment programs. */ for (i = 0; i <= softpipe->vs->max_sampler; i++) { - if (softpipe->sampler[i]) { + if (softpipe->vertex_samplers[i]) { softpipe->tgsi.vert_samplers_list[i] = get_sampler_varient( i, - sp_sampler(softpipe->sampler[i]), - softpipe->texture[i], + sp_sampler(softpipe->vertex_samplers[i]), + softpipe->vertex_textures[i], TGSI_PROCESSOR_VERTEX ); sp_sampler_varient_bind_texture( softpipe->tgsi.vert_samplers_list[i], - softpipe->tex_cache[i], - softpipe->texture[i] ); + softpipe->vertex_tex_cache[i], + softpipe->vertex_textures[i] ); } } diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c index c22ee86b66c..e26153b1d90 100644 --- a/src/gallium/drivers/softpipe/sp_tex_sample.c +++ b/src/gallium/drivers/softpipe/sp_tex_sample.c @@ -521,7 +521,7 @@ compute_lambda_1d(const struct sp_sampler_varient *samp, const struct pipe_sampler_state *sampler = samp->sampler; float dsdx = fabsf(s[QUAD_BOTTOM_RIGHT] - s[QUAD_BOTTOM_LEFT]); float dsdy = fabsf(s[QUAD_TOP_LEFT] - s[QUAD_BOTTOM_LEFT]); - float rho = MAX2(dsdx, dsdy) * texture->width[0]; + float rho = MAX2(dsdx, dsdy) * texture->width0; float lambda; lambda = util_fast_log2(rho); @@ -545,8 +545,8 @@ compute_lambda_2d(const struct sp_sampler_varient *samp, float dsdy = fabsf(s[QUAD_TOP_LEFT] - s[QUAD_BOTTOM_LEFT]); float dtdx = fabsf(t[QUAD_BOTTOM_RIGHT] - t[QUAD_BOTTOM_LEFT]); float dtdy = fabsf(t[QUAD_TOP_LEFT] - t[QUAD_BOTTOM_LEFT]); - float maxx = MAX2(dsdx, dsdy) * texture->width[0]; - float maxy = MAX2(dtdx, dtdy) * texture->height[0]; + float maxx = MAX2(dsdx, dsdy) * texture->width0; + float maxy = MAX2(dtdx, dtdy) * texture->height0; float rho = MAX2(maxx, maxy); float lambda; @@ -573,9 +573,9 @@ compute_lambda_3d(const struct sp_sampler_varient *samp, float dtdy = fabsf(t[QUAD_TOP_LEFT] - t[QUAD_BOTTOM_LEFT]); float dpdx = fabsf(p[QUAD_BOTTOM_RIGHT] - p[QUAD_BOTTOM_LEFT]); float dpdy = fabsf(p[QUAD_TOP_LEFT] - p[QUAD_BOTTOM_LEFT]); - float maxx = MAX2(dsdx, dsdy) * texture->width[0]; - float maxy = MAX2(dtdx, dtdy) * texture->height[0]; - float maxz = MAX2(dpdx, dpdy) * texture->depth[0]; + float maxx = MAX2(dsdx, dsdy) * texture->width0; + float maxy = MAX2(dtdx, dtdy) * texture->height0; + float maxz = MAX2(dpdx, dpdy) * texture->depth0; float rho, lambda; rho = MAX2(maxx, maxy); @@ -644,8 +644,8 @@ get_texel_2d(const struct sp_sampler_varient *samp, const struct pipe_texture *texture = samp->texture; unsigned level = addr.bits.level; - if (x < 0 || x >= (int) texture->width[level] || - y < 0 || y >= (int) texture->height[level]) { + if (x < 0 || x >= (int) u_minify(texture->width0, level) || + y < 0 || y >= (int) u_minify(texture->height0, level)) { return samp->sampler->border_color; } else { @@ -737,9 +737,9 @@ get_texel_3d(const struct sp_sampler_varient *samp, const struct pipe_texture *texture = samp->texture; unsigned level = addr.bits.level; - if (x < 0 || x >= (int) texture->width[level] || - y < 0 || y >= (int) texture->height[level] || - z < 0 || z >= (int) texture->depth[level]) { + if (x < 0 || x >= (int) u_minify(texture->width0, level) || + y < 0 || y >= (int) u_minify(texture->height0, level) || + z < 0 || z >= (int) u_minify(texture->depth0, level)) { return samp->sampler->border_color; } else { @@ -925,7 +925,7 @@ img_filter_1d_nearest(struct tgsi_sampler *tgsi_sampler, union tex_tile_address addr; level0 = samp->level; - width = texture->width[level0]; + width = u_minify(texture->width0, level0); assert(width > 0); @@ -961,8 +961,8 @@ img_filter_2d_nearest(struct tgsi_sampler *tgsi_sampler, level0 = samp->level; - width = texture->width[level0]; - height = texture->height[level0]; + width = u_minify(texture->width0, level0); + height = u_minify(texture->height0, level0); assert(width > 0); assert(height > 0); @@ -1008,8 +1008,8 @@ img_filter_cube_nearest(struct tgsi_sampler *tgsi_sampler, union tex_tile_address addr; level0 = samp->level; - width = texture->width[level0]; - height = texture->height[level0]; + width = u_minify(texture->width0, level0); + height = u_minify(texture->height0, level0); assert(width > 0); assert(height > 0); @@ -1046,9 +1046,9 @@ img_filter_3d_nearest(struct tgsi_sampler *tgsi_sampler, union tex_tile_address addr; level0 = samp->level; - width = texture->width[level0]; - height = texture->height[level0]; - depth = texture->depth[level0]; + width = u_minify(texture->width0, level0); + height = u_minify(texture->height0, level0); + depth = u_minify(texture->depth0, level0); assert(width > 0); assert(height > 0); @@ -1088,7 +1088,7 @@ img_filter_1d_linear(struct tgsi_sampler *tgsi_sampler, union tex_tile_address addr; level0 = samp->level; - width = texture->width[level0]; + width = u_minify(texture->width0, level0); assert(width > 0); @@ -1127,8 +1127,8 @@ img_filter_2d_linear(struct tgsi_sampler *tgsi_sampler, union tex_tile_address addr; level0 = samp->level; - width = texture->width[level0]; - height = texture->height[level0]; + width = u_minify(texture->width0, level0); + height = u_minify(texture->height0, level0); assert(width > 0); assert(height > 0); @@ -1174,8 +1174,8 @@ img_filter_cube_linear(struct tgsi_sampler *tgsi_sampler, union tex_tile_address addr; level0 = samp->level; - width = texture->width[level0]; - height = texture->height[level0]; + width = u_minify(texture->width0, level0); + height = u_minify(texture->height0, level0); assert(width > 0); assert(height > 0); @@ -1221,9 +1221,9 @@ img_filter_3d_linear(struct tgsi_sampler *tgsi_sampler, union tex_tile_address addr; level0 = samp->level; - width = texture->width[level0]; - height = texture->height[level0]; - depth = texture->depth[level0]; + width = u_minify(texture->width0, level0); + height = u_minify(texture->height0, level0); + depth = u_minify(texture->depth0, level0); addr.value = 0; addr.bits.level = level0; @@ -1778,8 +1778,8 @@ sp_sampler_varient_bind_texture( struct sp_sampler_varient *samp, samp->texture = texture; samp->cache = tex_cache; - samp->xpot = util_unsigned_logbase2( texture->width[0] ); - samp->ypot = util_unsigned_logbase2( texture->height[0] ); + samp->xpot = util_unsigned_logbase2( texture->width0 ); + samp->ypot = util_unsigned_logbase2( texture->height0 ); samp->level = CLAMP((int) sampler->min_lod, 0, (int) texture->last_level); } diff --git a/src/gallium/drivers/softpipe/sp_tex_tile_cache.c b/src/gallium/drivers/softpipe/sp_tex_tile_cache.c index 407a22a9f4b..e50a76a73bc 100644 --- a/src/gallium/drivers/softpipe/sp_tex_tile_cache.c +++ b/src/gallium/drivers/softpipe/sp_tex_tile_cache.c @@ -35,6 +35,7 @@ #include "pipe/p_inlines.h" #include "util/u_memory.h" #include "util/u_tile.h" +#include "util/u_math.h" #include "sp_context.h" #include "sp_surface.h" #include "sp_texture.h" @@ -246,9 +247,9 @@ sp_find_cached_tile_tex(struct softpipe_tex_tile_cache *tc, addr.bits.level, addr.bits.z, PIPE_TRANSFER_READ, 0, 0, - tc->texture->width[addr.bits.level], - tc->texture->height[addr.bits.level]); - + u_minify(tc->texture->width0, addr.bits.level), + u_minify(tc->texture->height0, addr.bits.level)); + tc->tex_trans_map = screen->transfer_map(screen, tc->tex_trans); tc->tex_face = addr.bits.face; diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index 7caf2928b4b..ac5f61e46f4 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -52,16 +52,17 @@ softpipe_texture_layout(struct pipe_screen *screen, { struct pipe_texture *pt = &spt->base; unsigned level; - unsigned width = pt->width[0]; - unsigned height = pt->height[0]; - unsigned depth = pt->depth[0]; + unsigned width = pt->width0; + unsigned height = pt->height0; + unsigned depth = pt->depth0; unsigned buffer_size = 0; + pt->width0 = width; + pt->height0 = height; + pt->depth0 = depth; + for (level = 0; level <= pt->last_level; level++) { - pt->width[level] = width; - pt->height[level] = height; - pt->depth[level] = depth; pt->nblocksx[level] = pf_get_nblocksx(&pt->block, width); pt->nblocksy[level] = pf_get_nblocksy(&pt->block, height); spt->stride[level] = pt->nblocksx[level]*pt->block.size; @@ -72,9 +73,9 @@ softpipe_texture_layout(struct pipe_screen *screen, ((pt->target == PIPE_TEXTURE_CUBE) ? 6 : depth) * spt->stride[level]); - width = minify(width); - height = minify(height); - depth = minify(depth); + width = u_minify(width, 1); + height = u_minify(height, 1); + depth = u_minify(depth, 1); } spt->buffer = screen->buffer_create(screen, 32, @@ -96,12 +97,12 @@ softpipe_displaytarget_layout(struct pipe_screen *screen, PIPE_BUFFER_USAGE_GPU_READ_WRITE); unsigned tex_usage = spt->base.tex_usage; - spt->base.nblocksx[0] = pf_get_nblocksx(&spt->base.block, spt->base.width[0]); - spt->base.nblocksy[0] = pf_get_nblocksy(&spt->base.block, spt->base.height[0]); + spt->base.nblocksx[0] = pf_get_nblocksx(&spt->base.block, spt->base.width0); + spt->base.nblocksy[0] = pf_get_nblocksy(&spt->base.block, spt->base.height0); spt->buffer = screen->surface_buffer_create( screen, - spt->base.width[0], - spt->base.height[0], + spt->base.width0, + spt->base.height0, spt->base.format, usage, tex_usage, @@ -126,9 +127,9 @@ softpipe_texture_create(struct pipe_screen *screen, pipe_reference_init(&spt->base.reference, 1); spt->base.screen = screen; - spt->pot = (util_is_power_of_two(template->width[0]) && - util_is_power_of_two(template->height[0]) && - util_is_power_of_two(template->depth[0])); + spt->pot = (util_is_power_of_two(template->width0) && + util_is_power_of_two(template->height0) && + util_is_power_of_two(template->depth0)); if (spt->base.tex_usage & (PIPE_TEXTURE_USAGE_DISPLAY_TARGET | PIPE_TEXTURE_USAGE_PRIMARY)) { @@ -163,7 +164,7 @@ softpipe_texture_blanket(struct pipe_screen * screen, /* Only supports one type */ if (base->target != PIPE_TEXTURE_2D || base->last_level != 0 || - base->depth[0] != 1) { + base->depth0 != 1) { return NULL; } @@ -174,8 +175,8 @@ softpipe_texture_blanket(struct pipe_screen * screen, spt->base = *base; pipe_reference_init(&spt->base.reference, 1); spt->base.screen = screen; - spt->base.nblocksx[0] = pf_get_nblocksx(&spt->base.block, spt->base.width[0]); - spt->base.nblocksy[0] = pf_get_nblocksy(&spt->base.block, spt->base.height[0]); + spt->base.nblocksx[0] = pf_get_nblocksx(&spt->base.block, spt->base.width0); + spt->base.nblocksy[0] = pf_get_nblocksy(&spt->base.block, spt->base.height0); spt->stride[0] = stride[0]; pipe_buffer_reference(&spt->buffer, buffer); @@ -213,8 +214,8 @@ softpipe_get_tex_surface(struct pipe_screen *screen, pipe_reference_init(&ps->reference, 1); pipe_texture_reference(&ps->texture, pt); ps->format = pt->format; - ps->width = pt->width[level]; - ps->height = pt->height[level]; + ps->width = u_minify(pt->width0, level); + ps->height = u_minify(pt->height0, level); ps->offset = spt->level_offset[level]; ps->usage = usage; @@ -434,9 +435,9 @@ softpipe_video_surface_create(struct pipe_screen *screen, template.format = PIPE_FORMAT_X8R8G8B8_UNORM; template.last_level = 0; /* vl_mpeg12_mc_renderer expects this when it's initialized with pot_buffers=true */ - template.width[0] = util_next_power_of_two(width); - template.height[0] = util_next_power_of_two(height); - template.depth[0] = 1; + template.width0 = util_next_power_of_two(width); + template.height0 = util_next_power_of_two(height); + template.depth0 = 1; pf_get_block(template.format, &template.block); template.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER | PIPE_TEXTURE_USAGE_RENDER_TARGET; diff --git a/src/gallium/drivers/svga/svga_pipe_sampler.c b/src/gallium/drivers/svga/svga_pipe_sampler.c index 3eeca6b784b..b4e57c5d15b 100644 --- a/src/gallium/drivers/svga/svga_pipe_sampler.c +++ b/src/gallium/drivers/svga/svga_pipe_sampler.c @@ -234,9 +234,9 @@ static void svga_set_sampler_textures(struct pipe_context *pipe, void svga_init_sampler_functions( struct svga_context *svga ) { svga->pipe.create_sampler_state = svga_create_sampler_state; - svga->pipe.bind_sampler_states = svga_bind_sampler_states; + svga->pipe.bind_fragment_sampler_states = svga_bind_sampler_states; svga->pipe.delete_sampler_state = svga_delete_sampler_state; - svga->pipe.set_sampler_textures = svga_set_sampler_textures; + svga->pipe.set_fragment_sampler_textures = svga_set_sampler_textures; } diff --git a/src/gallium/drivers/svga/svga_screen_texture.c b/src/gallium/drivers/svga/svga_screen_texture.c index e7301aba841..6e10d65a205 100644 --- a/src/gallium/drivers/svga/svga_screen_texture.c +++ b/src/gallium/drivers/svga/svga_screen_texture.c @@ -281,24 +281,21 @@ svga_texture_create(struct pipe_screen *screen, if(templat->last_level >= SVGA_MAX_TEXTURE_LEVELS) goto error2; - width = templat->width[0]; - height = templat->height[0]; - depth = templat->depth[0]; + width = templat->width0; + height = templat->height0; + depth = templat->depth0; for(level = 0; level <= templat->last_level; ++level) { - tex->base.width[level] = width; - tex->base.height[level] = height; - tex->base.depth[level] = depth; tex->base.nblocksx[level] = pf_get_nblocksx(&tex->base.block, width); tex->base.nblocksy[level] = pf_get_nblocksy(&tex->base.block, height); - width = minify(width); - height = minify(height); - depth = minify(depth); + width = u_minify(width, 1); + height = u_minify(height, 1); + depth = u_minify(depth, 1); } tex->key.flags = 0; - tex->key.size.width = templat->width[0]; - tex->key.size.height = templat->height[0]; - tex->key.size.depth = templat->depth[0]; + tex->key.size.width = templat->width0; + tex->key.size.height = templat->height0; + tex->key.size.depth = templat->depth0; if(templat->target == PIPE_TEXTURE_CUBE) { tex->key.flags |= SVGA3D_SURFACE_CUBEMAP; @@ -365,7 +362,7 @@ svga_texture_blanket(struct pipe_screen * screen, /* Only supports one type */ if (base->target != PIPE_TEXTURE_2D || base->last_level != 0 || - base->depth[0] != 1) { + base->depth0 != 1) { return NULL; } @@ -538,9 +535,9 @@ svga_texture_view_surface(struct pipe_context *pipe, key->flags = 0; key->format = format; key->numMipLevels = num_mip; - key->size.width = tex->base.width[start_mip]; - key->size.height = tex->base.height[start_mip]; - key->size.depth = zslice_pick < 0 ? tex->base.depth[start_mip] : 1; + key->size.width = u_minify(tex->base.width0, start_mip); + key->size.height = u_minify(tex->base.height0, start_mip); + key->size.depth = zslice_pick < 0 ? u_minify(tex->base.depth0, start_mip) : 1; key->cachable = 1; assert(key->size.depth == 1); @@ -574,7 +571,10 @@ svga_texture_view_surface(struct pipe_context *pipe, for (i = 0; i < key->numMipLevels; i++) { for (j = 0; j < key->numFaces; j++) { if(tex->defined[j + face_pick][i + start_mip]) { - unsigned depth = zslice_pick < 0 ? tex->base.depth[i + start_mip] : 1; + unsigned depth = (zslice_pick < 0 ? + u_minify(tex->base.depth0, i + start_mip) : + 1); + svga_texture_copy_handle(svga_context(pipe), ss, tex->handle, @@ -582,8 +582,8 @@ svga_texture_view_surface(struct pipe_context *pipe, i + start_mip, j + face_pick, handle, 0, 0, 0, i, j, - tex->base.width[i + start_mip], - tex->base.height[i + start_mip], + u_minify(tex->base.width0, i + start_mip), + u_minify(tex->base.height0, i + start_mip), depth); } } @@ -612,8 +612,8 @@ svga_get_tex_surface(struct pipe_screen *screen, pipe_reference_init(&s->base.reference, 1); pipe_texture_reference(&s->base.texture, pt); s->base.format = pt->format; - s->base.width = pt->width[level]; - s->base.height = pt->height[level]; + s->base.width = u_minify(pt->width0, level); + s->base.height = u_minify(pt->height0, level); s->base.usage = flags; s->base.level = level; s->base.face = face; @@ -741,7 +741,8 @@ svga_propagate_surface(struct pipe_context *pipe, struct pipe_surface *surf) svga_texture_copy_handle(svga_context(pipe), ss, s->handle, 0, 0, 0, s->real_level, s->real_face, tex->handle, 0, 0, surf->zslice, surf->level, surf->face, - tex->base.width[surf->level], tex->base.height[surf->level], 1); + u_minify(tex->base.width0, surf->level), + u_minify(tex->base.height0, surf->level), 1); tex->defined[surf->face][surf->level] = TRUE; } } @@ -971,9 +972,9 @@ svga_get_tex_sampler_view(struct pipe_context *pipe, struct pipe_texture *pt, "svga: Sampler view: no %p, mips %u..%u, nr %u, size (%ux%ux%u), last %u\n", pt, min_lod, max_lod, max_lod - min_lod + 1, - pt->width[0], - pt->height[0], - pt->depth[0], + pt->width0, + pt->height0, + pt->depth0, pt->last_level); sv->key.cachable = 0; sws->surface_reference(sws, &sv->handle, tex->handle); @@ -984,9 +985,9 @@ svga_get_tex_sampler_view(struct pipe_context *pipe, struct pipe_texture *pt, "svga: Sampler view: yes %p, mips %u..%u, nr %u, size (%ux%ux%u), last %u\n", pt, min_lod, max_lod, max_lod - min_lod + 1, - pt->width[0], - pt->height[0], - pt->depth[0], + pt->width0, + pt->height0, + pt->depth0, pt->last_level); sv->age = tex->age; @@ -1036,9 +1037,9 @@ svga_validate_sampler_view(struct svga_context *svga, struct svga_sampler_view * svga_texture_copy_handle(svga, NULL, tex->handle, 0, 0, 0, i, k, v->handle, 0, 0, 0, i - v->min_lod, k, - tex->base.width[i], - tex->base.height[i], - tex->base.depth[i]); + u_minify(tex->base.width0, i), + u_minify(tex->base.height0, i), + u_minify(tex->base.depth0, i)); } } @@ -1068,7 +1069,7 @@ svga_screen_buffer_from_texture(struct pipe_texture *texture, svga_translate_format(texture->format), stex->handle); - *stride = pf_get_nblocksx(&texture->block, texture->width[0]) * + *stride = pf_get_nblocksx(&texture->block, texture->width0) * texture->block.size; return *buffer != NULL; diff --git a/src/gallium/drivers/svga/svga_state_constants.c b/src/gallium/drivers/svga/svga_state_constants.c index 18cce7dde1a..209ed282450 100644 --- a/src/gallium/drivers/svga/svga_state_constants.c +++ b/src/gallium/drivers/svga/svga_state_constants.c @@ -140,8 +140,8 @@ static int emit_fs_consts( struct svga_context *svga, struct pipe_texture *tex = svga->curr.texture[i]; float data[4]; - data[0] = 1.0 / (float)tex->width[0]; - data[1] = 1.0 / (float)tex->height[0]; + data[0] = 1.0 / (float)tex->width0; + data[1] = 1.0 / (float)tex->height0; data[2] = 1.0; data[3] = 1.0; diff --git a/src/gallium/drivers/trace/tr_context.c b/src/gallium/drivers/trace/tr_context.c index bf470b46ae1..2f0f063d2de 100644 --- a/src/gallium/drivers/trace/tr_context.c +++ b/src/gallium/drivers/trace/tr_context.c @@ -107,7 +107,7 @@ trace_context_set_edgeflags(struct pipe_context *_pipe, /* FIXME: we don't know how big this array is */ trace_dump_arg(ptr, bitfield); - pipe->set_edgeflags(pipe, bitfield);; + pipe->set_edgeflags(pipe, bitfield); trace_dump_call_end(); } @@ -143,10 +143,16 @@ trace_context_draw_block(struct trace_context *tr_ctx, int flag) for (k = 0; k < tr_ctx->curr.nr_cbufs; k++) if (tr_ctx->draw_rule.surf == tr_ctx->curr.cbufs[k]) block = TRUE; - if (tr_ctx->draw_rule.tex) + if (tr_ctx->draw_rule.tex) { for (k = 0; k < tr_ctx->curr.num_texs; k++) if (tr_ctx->draw_rule.tex == tr_ctx->curr.tex[k]) block = TRUE; + for (k = 0; k < tr_ctx->curr.num_vert_texs; k++) { + if (tr_ctx->draw_rule.tex == tr_ctx->curr.vert_tex[k]) { + block = TRUE; + } + } + } if (block) tr_ctx->draw_blocked |= (flag | 4); @@ -192,7 +198,7 @@ trace_context_draw_arrays(struct pipe_context *_pipe, trace_dump_arg(uint, start); trace_dump_arg(uint, count); - result = pipe->draw_arrays(pipe, mode, start, count);; + result = pipe->draw_arrays(pipe, mode, start, count); trace_dump_ret(bool, result); @@ -232,7 +238,7 @@ trace_context_draw_elements(struct pipe_context *_pipe, trace_dump_arg(uint, start); trace_dump_arg(uint, count); - result = pipe->draw_elements(pipe, indexBuffer, indexSize, mode, start, count);; + result = pipe->draw_elements(pipe, indexBuffer, indexSize, mode, start, count); trace_dump_ret(bool, result); @@ -306,7 +312,7 @@ trace_context_create_query(struct pipe_context *_pipe, trace_dump_arg(ptr, pipe); trace_dump_arg(uint, query_type); - result = pipe->create_query(pipe, query_type);; + result = pipe->create_query(pipe, query_type); trace_dump_ret(ptr, result); @@ -328,7 +334,7 @@ trace_context_destroy_query(struct pipe_context *_pipe, trace_dump_arg(ptr, pipe); trace_dump_arg(ptr, query); - pipe->destroy_query(pipe, query);; + pipe->destroy_query(pipe, query); trace_dump_call_end(); } @@ -346,7 +352,7 @@ trace_context_begin_query(struct pipe_context *_pipe, trace_dump_arg(ptr, pipe); trace_dump_arg(ptr, query); - pipe->begin_query(pipe, query);; + pipe->begin_query(pipe, query); trace_dump_call_end(); } @@ -385,7 +391,7 @@ trace_context_get_query_result(struct pipe_context *_pipe, trace_dump_arg(ptr, pipe); - _result = pipe->get_query_result(pipe, query, wait, presult);; + _result = pipe->get_query_result(pipe, query, wait, presult); result = *presult; trace_dump_arg(uint, result); @@ -410,7 +416,7 @@ trace_context_create_blend_state(struct pipe_context *_pipe, trace_dump_arg(ptr, pipe); trace_dump_arg(blend_state, state); - result = pipe->create_blend_state(pipe, state);; + result = pipe->create_blend_state(pipe, state); trace_dump_ret(ptr, result); @@ -432,7 +438,7 @@ trace_context_bind_blend_state(struct pipe_context *_pipe, trace_dump_arg(ptr, pipe); trace_dump_arg(ptr, state); - pipe->bind_blend_state(pipe, state);; + pipe->bind_blend_state(pipe, state); trace_dump_call_end(); } @@ -450,7 +456,7 @@ trace_context_delete_blend_state(struct pipe_context *_pipe, trace_dump_arg(ptr, pipe); trace_dump_arg(ptr, state); - pipe->delete_blend_state(pipe, state);; + pipe->delete_blend_state(pipe, state); trace_dump_call_end(); } @@ -469,7 +475,7 @@ trace_context_create_sampler_state(struct pipe_context *_pipe, trace_dump_arg(ptr, pipe); trace_dump_arg(sampler_state, state); - result = pipe->create_sampler_state(pipe, state);; + result = pipe->create_sampler_state(pipe, state); trace_dump_ret(ptr, result); @@ -480,19 +486,40 @@ trace_context_create_sampler_state(struct pipe_context *_pipe, static INLINE void -trace_context_bind_sampler_states(struct pipe_context *_pipe, - unsigned num_states, void **states) +trace_context_bind_fragment_sampler_states(struct pipe_context *_pipe, + unsigned num_states, + void **states) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin("pipe_context", "bind_fragment_sampler_states"); + + trace_dump_arg(ptr, pipe); + trace_dump_arg(uint, num_states); + trace_dump_arg_array(ptr, states, num_states); + + pipe->bind_fragment_sampler_states(pipe, num_states, states); + + trace_dump_call_end(); +} + + +static INLINE void +trace_context_bind_vertex_sampler_states(struct pipe_context *_pipe, + unsigned num_states, + void **states) { struct trace_context *tr_ctx = trace_context(_pipe); struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin("pipe_context", "bind_sampler_states"); + trace_dump_call_begin("pipe_context", "bind_vertex_sampler_states"); trace_dump_arg(ptr, pipe); trace_dump_arg(uint, num_states); trace_dump_arg_array(ptr, states, num_states); - pipe->bind_sampler_states(pipe, num_states, states);; + pipe->bind_vertex_sampler_states(pipe, num_states, states); trace_dump_call_end(); } @@ -510,7 +537,7 @@ trace_context_delete_sampler_state(struct pipe_context *_pipe, trace_dump_arg(ptr, pipe); trace_dump_arg(ptr, state); - pipe->delete_sampler_state(pipe, state);; + pipe->delete_sampler_state(pipe, state); trace_dump_call_end(); } @@ -529,7 +556,7 @@ trace_context_create_rasterizer_state(struct pipe_context *_pipe, trace_dump_arg(ptr, pipe); trace_dump_arg(rasterizer_state, state); - result = pipe->create_rasterizer_state(pipe, state);; + result = pipe->create_rasterizer_state(pipe, state); trace_dump_ret(ptr, result); @@ -551,7 +578,7 @@ trace_context_bind_rasterizer_state(struct pipe_context *_pipe, trace_dump_arg(ptr, pipe); trace_dump_arg(ptr, state); - pipe->bind_rasterizer_state(pipe, state);; + pipe->bind_rasterizer_state(pipe, state); trace_dump_call_end(); } @@ -569,7 +596,7 @@ trace_context_delete_rasterizer_state(struct pipe_context *_pipe, trace_dump_arg(ptr, pipe); trace_dump_arg(ptr, state); - pipe->delete_rasterizer_state(pipe, state);; + pipe->delete_rasterizer_state(pipe, state); trace_dump_call_end(); } @@ -585,7 +612,7 @@ trace_context_create_depth_stencil_alpha_state(struct pipe_context *_pipe, trace_dump_call_begin("pipe_context", "create_depth_stencil_alpha_state"); - result = pipe->create_depth_stencil_alpha_state(pipe, state);; + result = pipe->create_depth_stencil_alpha_state(pipe, state); trace_dump_arg(ptr, pipe); trace_dump_arg(depth_stencil_alpha_state, state); @@ -610,7 +637,7 @@ trace_context_bind_depth_stencil_alpha_state(struct pipe_context *_pipe, trace_dump_arg(ptr, pipe); trace_dump_arg(ptr, state); - pipe->bind_depth_stencil_alpha_state(pipe, state);; + pipe->bind_depth_stencil_alpha_state(pipe, state); trace_dump_call_end(); } @@ -628,7 +655,7 @@ trace_context_delete_depth_stencil_alpha_state(struct pipe_context *_pipe, trace_dump_arg(ptr, pipe); trace_dump_arg(ptr, state); - pipe->delete_depth_stencil_alpha_state(pipe, state);; + pipe->delete_depth_stencil_alpha_state(pipe, state); trace_dump_call_end(); } @@ -647,7 +674,7 @@ trace_context_create_fs_state(struct pipe_context *_pipe, trace_dump_arg(ptr, pipe); trace_dump_arg(shader_state, state); - result = pipe->create_fs_state(pipe, state);; + result = pipe->create_fs_state(pipe, state); trace_dump_ret(ptr, result); @@ -750,7 +777,7 @@ trace_context_bind_vs_state(struct pipe_context *_pipe, if (tr_shdr && tr_shdr->replaced) state = tr_shdr->replaced; - pipe->bind_vs_state(pipe, state);; + pipe->bind_vs_state(pipe, state); trace_dump_call_end(); } @@ -770,7 +797,7 @@ trace_context_delete_vs_state(struct pipe_context *_pipe, trace_dump_arg(ptr, pipe); trace_dump_arg(ptr, state); - pipe->delete_vs_state(pipe, state);; + pipe->delete_vs_state(pipe, state); trace_dump_call_end(); @@ -790,7 +817,7 @@ trace_context_set_blend_color(struct pipe_context *_pipe, trace_dump_arg(ptr, pipe); trace_dump_arg(blend_color, state); - pipe->set_blend_color(pipe, state);; + pipe->set_blend_color(pipe, state); trace_dump_call_end(); } @@ -808,7 +835,7 @@ trace_context_set_clip_state(struct pipe_context *_pipe, trace_dump_arg(ptr, pipe); trace_dump_arg(clip_state, state); - pipe->set_clip_state(pipe, state);; + pipe->set_clip_state(pipe, state); trace_dump_call_end(); } @@ -880,7 +907,7 @@ trace_context_set_framebuffer_state(struct pipe_context *_pipe, trace_dump_arg(ptr, pipe); trace_dump_arg(framebuffer_state, state); - pipe->set_framebuffer_state(pipe, state);; + pipe->set_framebuffer_state(pipe, state); trace_dump_call_end(); } @@ -898,7 +925,7 @@ trace_context_set_polygon_stipple(struct pipe_context *_pipe, trace_dump_arg(ptr, pipe); trace_dump_arg(poly_stipple, state); - pipe->set_polygon_stipple(pipe, state);; + pipe->set_polygon_stipple(pipe, state); trace_dump_call_end(); } @@ -916,7 +943,7 @@ trace_context_set_scissor_state(struct pipe_context *_pipe, trace_dump_arg(ptr, pipe); trace_dump_arg(scissor_state, state); - pipe->set_scissor_state(pipe, state);; + pipe->set_scissor_state(pipe, state); trace_dump_call_end(); } @@ -934,16 +961,16 @@ trace_context_set_viewport_state(struct pipe_context *_pipe, trace_dump_arg(ptr, pipe); trace_dump_arg(viewport_state, state); - pipe->set_viewport_state(pipe, state);; + pipe->set_viewport_state(pipe, state); trace_dump_call_end(); } static INLINE void -trace_context_set_sampler_textures(struct pipe_context *_pipe, - unsigned num_textures, - struct pipe_texture **textures) +trace_context_set_fragment_sampler_textures(struct pipe_context *_pipe, + unsigned num_textures, + struct pipe_texture **textures) { struct trace_context *tr_ctx = trace_context(_pipe); struct trace_texture *tr_tex; @@ -959,13 +986,44 @@ trace_context_set_sampler_textures(struct pipe_context *_pipe, } textures = unwrapped_textures; - trace_dump_call_begin("pipe_context", "set_sampler_textures"); + trace_dump_call_begin("pipe_context", "set_fragment_sampler_textures"); + + trace_dump_arg(ptr, pipe); + trace_dump_arg(uint, num_textures); + trace_dump_arg_array(ptr, textures, num_textures); + + pipe->set_fragment_sampler_textures(pipe, num_textures, textures); + + trace_dump_call_end(); +} + + +static INLINE void +trace_context_set_vertex_sampler_textures(struct pipe_context *_pipe, + unsigned num_textures, + struct pipe_texture **textures) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_texture *tr_tex; + struct pipe_context *pipe = tr_ctx->pipe; + struct pipe_texture *unwrapped_textures[PIPE_MAX_VERTEX_SAMPLERS]; + unsigned i; + + tr_ctx->curr.num_vert_texs = num_textures; + for(i = 0; i < num_textures; ++i) { + tr_tex = trace_texture(textures[i]); + tr_ctx->curr.vert_tex[i] = tr_tex; + unwrapped_textures[i] = tr_tex ? tr_tex->texture : NULL; + } + textures = unwrapped_textures; + + trace_dump_call_begin("pipe_context", "set_vertex_sampler_textures"); trace_dump_arg(ptr, pipe); trace_dump_arg(uint, num_textures); trace_dump_arg_array(ptr, textures, num_textures); - pipe->set_sampler_textures(pipe, num_textures, textures);; + pipe->set_vertex_sampler_textures(pipe, num_textures, textures); trace_dump_call_end(); } @@ -1024,7 +1082,7 @@ trace_context_set_vertex_elements(struct pipe_context *_pipe, trace_dump_struct_array(vertex_element, elements, num_elements); trace_dump_arg_end(); - pipe->set_vertex_elements(pipe, num_elements, elements);; + pipe->set_vertex_elements(pipe, num_elements, elements); trace_dump_call_end(); } @@ -1085,7 +1143,7 @@ trace_context_surface_fill(struct pipe_context *_pipe, trace_dump_arg(uint, width); trace_dump_arg(uint, height); - pipe->surface_fill(pipe, dst, dstx, dsty, width, height, value);; + pipe->surface_fill(pipe, dst, dstx, dsty, width, height, value); trace_dump_call_end(); } @@ -1128,7 +1186,7 @@ trace_context_flush(struct pipe_context *_pipe, trace_dump_arg(ptr, pipe); trace_dump_arg(uint, flags); - pipe->flush(pipe, flags, fence);; + pipe->flush(pipe, flags, fence); if(fence) trace_dump_ret(ptr, *fence); @@ -1253,7 +1311,8 @@ trace_context_create(struct pipe_screen *_screen, tr_ctx->base.bind_blend_state = trace_context_bind_blend_state; tr_ctx->base.delete_blend_state = trace_context_delete_blend_state; tr_ctx->base.create_sampler_state = trace_context_create_sampler_state; - tr_ctx->base.bind_sampler_states = trace_context_bind_sampler_states; + tr_ctx->base.bind_fragment_sampler_states = trace_context_bind_fragment_sampler_states; + tr_ctx->base.bind_vertex_sampler_states = trace_context_bind_vertex_sampler_states; tr_ctx->base.delete_sampler_state = trace_context_delete_sampler_state; tr_ctx->base.create_rasterizer_state = trace_context_create_rasterizer_state; tr_ctx->base.bind_rasterizer_state = trace_context_bind_rasterizer_state; @@ -1274,7 +1333,8 @@ trace_context_create(struct pipe_screen *_screen, tr_ctx->base.set_polygon_stipple = trace_context_set_polygon_stipple; tr_ctx->base.set_scissor_state = trace_context_set_scissor_state; tr_ctx->base.set_viewport_state = trace_context_set_viewport_state; - tr_ctx->base.set_sampler_textures = trace_context_set_sampler_textures; + tr_ctx->base.set_fragment_sampler_textures = trace_context_set_fragment_sampler_textures; + tr_ctx->base.set_vertex_sampler_textures = trace_context_set_vertex_sampler_textures; tr_ctx->base.set_vertex_buffers = trace_context_set_vertex_buffers; tr_ctx->base.set_vertex_elements = trace_context_set_vertex_elements; if (pipe->surface_copy) diff --git a/src/gallium/drivers/trace/tr_context.h b/src/gallium/drivers/trace/tr_context.h index 6febe4b4114..852b480765a 100644 --- a/src/gallium/drivers/trace/tr_context.h +++ b/src/gallium/drivers/trace/tr_context.h @@ -54,6 +54,9 @@ struct trace_context struct trace_texture *tex[PIPE_MAX_SAMPLERS]; unsigned num_texs; + struct trace_texture *vert_tex[PIPE_MAX_VERTEX_SAMPLERS]; + unsigned num_vert_texs; + unsigned nr_cbufs; struct trace_texture *cbufs[PIPE_MAX_COLOR_BUFS]; struct trace_texture *zsbuf; diff --git a/src/gallium/drivers/trace/tr_dump_state.c b/src/gallium/drivers/trace/tr_dump_state.c index bcf6751af4f..6d582092941 100644 --- a/src/gallium/drivers/trace/tr_dump_state.c +++ b/src/gallium/drivers/trace/tr_dump_state.c @@ -83,15 +83,15 @@ void trace_dump_template(const struct pipe_texture *templat) trace_dump_member(format, templat, format); trace_dump_member_begin("width"); - trace_dump_array(uint, templat->width, 1); + trace_dump_uint(templat->width0); trace_dump_member_end(); trace_dump_member_begin("height"); - trace_dump_array(uint, templat->height, 1); + trace_dump_uint(templat->height0); trace_dump_member_end(); trace_dump_member_begin("depth"); - trace_dump_array(uint, templat->depth, 1); + trace_dump_uint(templat->depth0); trace_dump_member_end(); trace_dump_member_begin("block"); diff --git a/src/gallium/drivers/trace/tr_rbug.c b/src/gallium/drivers/trace/tr_rbug.c index 81e0a6f3b00..b59458c0e37 100644 --- a/src/gallium/drivers/trace/tr_rbug.c +++ b/src/gallium/drivers/trace/tr_rbug.c @@ -200,9 +200,9 @@ trace_rbug_texture_info(struct trace_rbug *tr_rbug, struct rbug_header *header, t = tr_tex->texture; rbug_send_texture_info_reply(tr_rbug->con, serial, t->target, t->format, - t->width, t->last_level + 1, - t->height, t->last_level + 1, - t->depth, t->last_level + 1, + &t->width0, 1, + &t->height0, 1, + &t->depth0, 1, t->block.width, t->block.height, t->block.size, t->last_level, t->nr_samples, diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h index 5569001e601..f896001eb12 100644 --- a/src/gallium/include/pipe/p_context.h +++ b/src/gallium/include/pipe/p_context.h @@ -123,7 +123,12 @@ struct pipe_context { void * (*create_sampler_state)(struct pipe_context *, const struct pipe_sampler_state *); - void (*bind_sampler_states)(struct pipe_context *, unsigned num, void **); + void (*bind_fragment_sampler_states)(struct pipe_context *, + unsigned num_samplers, + void **samplers); + void (*bind_vertex_sampler_states)(struct pipe_context *, + unsigned num_samplers, + void **samplers); void (*delete_sampler_state)(struct pipe_context *, void *); void * (*create_rasterizer_state)(struct pipe_context *, @@ -173,9 +178,13 @@ struct pipe_context { void (*set_viewport_state)( struct pipe_context *, const struct pipe_viewport_state * ); - void (*set_sampler_textures)( struct pipe_context *, - unsigned num_textures, - struct pipe_texture ** ); + void (*set_fragment_sampler_textures)(struct pipe_context *, + unsigned num_textures, + struct pipe_texture **); + + void (*set_vertex_sampler_textures)(struct pipe_context *, + unsigned num_textures, + struct pipe_texture **); void (*set_vertex_buffers)( struct pipe_context *, unsigned num_buffers, diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h index fd14dc8e92d..69a0970d5f8 100644 --- a/src/gallium/include/pipe/p_defines.h +++ b/src/gallium/include/pipe/p_defines.h @@ -390,6 +390,8 @@ enum pipe_transfer_usage { #define PIPE_CAP_BLEND_EQUATION_SEPARATE 28 #define PIPE_CAP_SM3 29 /*< Shader Model 3 supported */ #define PIPE_CAP_MAX_PREDICATE_REGISTERS 30 +#define PIPE_CAP_MAX_COMBINED_SAMPLERS 31 /*< Maximum texture image units accessible from vertex + and fragment shaders combined */ /** diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h index d4c8aadaf92..c4c28522c8d 100644 --- a/src/gallium/include/pipe/p_shader_tokens.h +++ b/src/gallium/include/pipe/p_shader_tokens.h @@ -293,6 +293,8 @@ union tgsi_immediate_data * respectively. For a given operation code, those numbers are fixed and are * present here only for convenience. * + * If Predicate is TRUE, tgsi_instruction_predicate token immediately follows. + * * If Extended is TRUE, it is now executed. * * Saturate controls how are final results in destination registers modified. @@ -306,7 +308,8 @@ struct tgsi_instruction unsigned Saturate : 2; /* TGSI_SAT_ */ unsigned NumDstRegs : 2; /* UINT */ unsigned NumSrcRegs : 4; /* UINT */ - unsigned Padding : 3; + unsigned Predicate : 1; /* BOOL */ + unsigned Padding : 2; unsigned Extended : 1; /* BOOL */ }; @@ -323,7 +326,6 @@ struct tgsi_instruction #define TGSI_INSTRUCTION_EXT_TYPE_LABEL 1 #define TGSI_INSTRUCTION_EXT_TYPE_TEXTURE 2 -#define TGSI_INSTRUCTION_EXT_TYPE_PREDICATE 3 struct tgsi_instruction_ext { @@ -339,9 +341,6 @@ struct tgsi_instruction_ext * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_TEXTURE, it * should be cast to tgsi_instruction_ext_texture. * - * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_PREDICATE, it - * should be cast to tgsi_instruction_ext_predicate. - * * If tgsi_instruction_ext::Extended is TRUE, another tgsi_instruction_ext * follows. */ @@ -382,17 +381,15 @@ struct tgsi_instruction_ext_texture * For SM3, the following constraint applies. * - Swizzle is either set to identity or replicate. */ -struct tgsi_instruction_ext_predicate +struct tgsi_instruction_predicate { - unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_PREDICATE */ + int Index : 16; /* SINT */ unsigned SwizzleX : 2; /* TGSI_SWIZZLE_x */ unsigned SwizzleY : 2; /* TGSI_SWIZZLE_x */ unsigned SwizzleZ : 2; /* TGSI_SWIZZLE_x */ unsigned SwizzleW : 2; /* TGSI_SWIZZLE_x */ unsigned Negate : 1; /* BOOL */ - unsigned SrcIndex : 8; /* UINT */ - unsigned Padding : 10; - unsigned Extended : 1; /* BOOL */ + unsigned Padding : 7; }; /** diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index 287b424e4ac..6de7af6a81c 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -60,6 +60,7 @@ extern "C" { #define PIPE_MAX_COLOR_BUFS 8 #define PIPE_MAX_CONSTANT 32 #define PIPE_MAX_SAMPLERS 16 +#define PIPE_MAX_VERTEX_SAMPLERS 16 #define PIPE_MAX_SHADER_INPUTS 16 #define PIPE_MAX_SHADER_OUTPUTS 16 #define PIPE_MAX_TEXTURE_LEVELS 16 @@ -343,9 +344,9 @@ struct pipe_texture enum pipe_texture_target target; /**< PIPE_TEXTURE_x */ enum pipe_format format; /**< PIPE_FORMAT_x */ - unsigned width[PIPE_MAX_TEXTURE_LEVELS]; - unsigned height[PIPE_MAX_TEXTURE_LEVELS]; - unsigned depth[PIPE_MAX_TEXTURE_LEVELS]; + unsigned width0; + unsigned height0; + unsigned depth0; struct pipe_format_block block; unsigned nblocksx[PIPE_MAX_TEXTURE_LEVELS]; /**< allocated width in blocks */ diff --git a/src/gallium/state_trackers/dri/dri_drawable.c b/src/gallium/state_trackers/dri/dri_drawable.c index 5625ff53cfd..45a6059ea83 100644 --- a/src/gallium/state_trackers/dri/dri_drawable.c +++ b/src/gallium/state_trackers/dri/dri_drawable.c @@ -46,7 +46,7 @@ #include "util/u_memory.h" #include "util/u_rect.h" - + static struct pipe_surface * dri_surface_from_handle(struct drm_api *api, struct pipe_screen *screen, @@ -62,10 +62,10 @@ dri_surface_from_handle(struct drm_api *api, templat.tex_usage |= PIPE_TEXTURE_USAGE_RENDER_TARGET; templat.target = PIPE_TEXTURE_2D; templat.last_level = 0; - templat.depth[0] = 1; + templat.depth0 = 1; templat.format = format; - templat.width[0] = width; - templat.height[0] = height; + templat.width0 = width; + templat.height0 = height; pf_get_block(templat.format, &templat.block); texture = api->texture_from_shared_handle(api, screen, &templat, diff --git a/src/gallium/state_trackers/egl/egl_surface.c b/src/gallium/state_trackers/egl/egl_surface.c index 91615abebee..ddd9b04cd48 100644 --- a/src/gallium/state_trackers/egl/egl_surface.c +++ b/src/gallium/state_trackers/egl/egl_surface.c @@ -114,10 +114,10 @@ drm_create_texture(_EGLDisplay *dpy, templat.tex_usage |= PIPE_TEXTURE_USAGE_PRIMARY; templat.target = PIPE_TEXTURE_2D; templat.last_level = 0; - templat.depth[0] = 1; + templat.depth0 = 1; templat.format = PIPE_FORMAT_A8R8G8B8_UNORM; - templat.width[0] = w; - templat.height[0] = h; + templat.width0 = w; + templat.height0 = h; pf_get_block(templat.format, &templat.block); texture = screen->texture_create(dev->screen, diff --git a/src/gallium/state_trackers/python/p_context.i b/src/gallium/state_trackers/python/p_context.i index a40aa1e5187..9728207d9c7 100644 --- a/src/gallium/state_trackers/python/p_context.i +++ b/src/gallium/state_trackers/python/p_context.i @@ -147,9 +147,9 @@ struct st_context { if(!texture) texture = $self->default_texture; pipe_texture_reference(&$self->sampler_textures[index], texture); - $self->pipe->set_sampler_textures($self->pipe, - PIPE_MAX_SAMPLERS, - $self->sampler_textures); + $self->pipe->set_fragment_sampler_textures($self->pipe, + PIPE_MAX_SAMPLERS, + $self->sampler_textures); } void set_vertex_buffer(unsigned index, diff --git a/src/gallium/state_trackers/python/p_device.i b/src/gallium/state_trackers/python/p_device.i index f16fe5b0ff7..a83bcc71a1a 100644 --- a/src/gallium/state_trackers/python/p_device.i +++ b/src/gallium/state_trackers/python/p_device.i @@ -113,9 +113,9 @@ struct st_device { memset(&templat, 0, sizeof(templat)); templat.format = format; pf_get_block(templat.format, &templat.block); - templat.width[0] = width; - templat.height[0] = height; - templat.depth[0] = depth; + templat.width0 = width; + templat.height0 = height; + templat.depth0 = depth; templat.last_level = last_level; templat.target = target; templat.tex_usage = tex_usage; diff --git a/src/gallium/state_trackers/python/p_texture.i b/src/gallium/state_trackers/python/p_texture.i index 1d513abf3c7..5416b872f53 100644 --- a/src/gallium/state_trackers/python/p_texture.i +++ b/src/gallium/state_trackers/python/p_texture.i @@ -59,15 +59,15 @@ } unsigned get_width(unsigned level=0) { - return $self->width[level]; + return u_minify($self->width0, level); } unsigned get_height(unsigned level=0) { - return $self->height[level]; + return u_minify($self->height0, level); } unsigned get_depth(unsigned level=0) { - return $self->depth[level]; + return u_minify($self->depth0, level); } unsigned get_nblocksx(unsigned level=0) { @@ -88,7 +88,7 @@ SWIG_exception(SWIG_ValueError, "face out of bounds"); if(level > $self->last_level) SWIG_exception(SWIG_ValueError, "level out of bounds"); - if(zslice >= $self->depth[level]) + if(zslice >= u_minify($self->depth0, level)) SWIG_exception(SWIG_ValueError, "zslice out of bounds"); surface = CALLOC_STRUCT(st_surface); @@ -375,13 +375,13 @@ struct st_surface static unsigned st_surface_width_get(struct st_surface *surface) { - return surface->texture->width[surface->level]; + return u_minify(surface->texture->width0, surface->level); } static unsigned st_surface_height_get(struct st_surface *surface) { - return surface->texture->height[surface->level]; + return u_minify(surface->texture->height0, surface->level); } static unsigned diff --git a/src/gallium/state_trackers/python/retrace/interpreter.py b/src/gallium/state_trackers/python/retrace/interpreter.py index 348f2e43683..5f826b1c4c4 100755 --- a/src/gallium/state_trackers/python/retrace/interpreter.py +++ b/src/gallium/state_trackers/python/retrace/interpreter.py @@ -279,9 +279,9 @@ class Screen(Object): def texture_create(self, templat): return self.real.texture_create( format = templat.format, - width = templat.width[0], - height = templat.height[0], - depth = templat.depth[0], + width = templat.width0, + height = templat.height0, + depth = templat.depth0, last_level = templat.last_level, target = templat.target, tex_usage = templat.tex_usage, @@ -388,7 +388,7 @@ class Context(Object): def delete_sampler_state(self, state): pass - def bind_sampler_states(self, num_states, states): + def bind_fragment_sampler_states(self, num_states, states): for i in range(num_states): self.real.set_sampler(i, states[i]) @@ -486,7 +486,7 @@ class Context(Object): def set_viewport_state(self, state): self.real.set_viewport(state) - def set_sampler_textures(self, num_textures, textures): + def set_fragment_sampler_textures(self, num_textures, textures): for i in range(num_textures): self.real.set_sampler_texture(i, textures[i]) diff --git a/src/gallium/state_trackers/python/st_device.c b/src/gallium/state_trackers/python/st_device.c index ea7d18738f6..a791113abac 100644 --- a/src/gallium/state_trackers/python/st_device.c +++ b/src/gallium/state_trackers/python/st_device.c @@ -252,9 +252,9 @@ st_context_create(struct st_device *st_dev) templat.block.size = 4; templat.block.width = 1; templat.block.height = 1; - templat.width[0] = 1; - templat.height[0] = 1; - templat.depth[0] = 1; + templat.width0 = 1; + templat.height0 = 1; + templat.depth0 = 1; templat.last_level = 0; st_ctx->default_texture = screen->texture_create( screen, &templat ); @@ -264,8 +264,8 @@ st_context_create(struct st_device *st_dev) 0, 0, 0, PIPE_TRANSFER_WRITE, 0, 0, - st_ctx->default_texture->width[0], - st_ctx->default_texture->height[0]); + st_ctx->default_texture->width0, + st_ctx->default_texture->height0); if (transfer) { uint32_t *map; map = (uint32_t *) screen->transfer_map(screen, transfer); diff --git a/src/gallium/state_trackers/python/st_sample.c b/src/gallium/state_trackers/python/st_sample.c index 53a01891e12..6fee90afdaf 100644 --- a/src/gallium/state_trackers/python/st_sample.c +++ b/src/gallium/state_trackers/python/st_sample.c @@ -528,8 +528,8 @@ st_sample_surface(struct st_surface *surface, float *rgba) { struct pipe_texture *texture = surface->texture; struct pipe_screen *screen = texture->screen; - unsigned width = texture->width[surface->level]; - unsigned height = texture->height[surface->level]; + unsigned width = u_minify(texture->width0, surface->level); + unsigned height = u_minify(texture->height0, surface->level); uint rgba_stride = width * 4; struct pipe_transfer *transfer; void *raw; diff --git a/src/gallium/state_trackers/vega/api_filters.c b/src/gallium/state_trackers/vega/api_filters.c index 862cbb03c43..faf396d0877 100644 --- a/src/gallium/state_trackers/vega/api_filters.c +++ b/src/gallium/state_trackers/vega/api_filters.c @@ -68,9 +68,9 @@ static INLINE struct pipe_texture *create_texture_1d(struct vg_context *ctx, templ.target = PIPE_TEXTURE_1D; templ.format = PIPE_FORMAT_A8R8G8B8_UNORM; templ.last_level = 0; - templ.width[0] = color_data_len; - templ.height[0] = 1; - templ.depth[0] = 1; + templ.width0 = color_data_len; + templ.height0 = 1; + templ.depth0 = 1; pf_get_block(PIPE_FORMAT_A8R8G8B8_UNORM, &templ.block); templ.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER; @@ -81,7 +81,7 @@ static INLINE struct pipe_texture *create_texture_1d(struct vg_context *ctx, screen->get_tex_transfer(screen, tex, 0, 0, 0, PIPE_TRANSFER_READ_WRITE , - 0, 0, tex->width[0], tex->height[0]); + 0, 0, tex->width0, tex->height0); void *map = screen->transfer_map(screen, transfer); memcpy(map, color_data, sizeof(VGint)*color_data_len); screen->transfer_unmap(screen, transfer); diff --git a/src/gallium/state_trackers/vega/image.c b/src/gallium/state_trackers/vega/image.c index 9a722980d52..4684a5727dd 100644 --- a/src/gallium/state_trackers/vega/image.c +++ b/src/gallium/state_trackers/vega/image.c @@ -93,8 +93,8 @@ static void vg_copy_texture(struct vg_context *ctx, dst_loc[3] = height; dst_bounds[0] = 0.f; dst_bounds[1] = 0.f; - dst_bounds[2] = dst->width[0]; - dst_bounds[3] = dst->height[0]; + dst_bounds[2] = dst->width0; + dst_bounds[3] = dst->height0; src_loc[0] = sx; src_loc[1] = sy; @@ -102,8 +102,8 @@ static void vg_copy_texture(struct vg_context *ctx, src_loc[3] = height; src_bounds[0] = 0.f; src_bounds[1] = 0.f; - src_bounds[2] = src->width[0]; - src_bounds[3] = src->height[0]; + src_bounds[2] = src->width0; + src_bounds[3] = src->height0; vg_bound_rect(src_loc, src_bounds, src_shift); vg_bound_rect(dst_loc, dst_bounds, dst_shift); @@ -272,9 +272,9 @@ struct vg_image * image_create(VGImageFormat format, pt.format = pformat; pf_get_block(pformat, &pt.block); pt.last_level = 0; - pt.width[0] = width; - pt.height[0] = height; - pt.depth[0] = 1; + pt.width0 = width; + pt.height0 = height; + pt.depth0 = 1; pt.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER; newtex = screen->texture_create(screen, &pt); @@ -414,7 +414,7 @@ void image_sub_data(struct vg_image *image, { /* upload color_data */ struct pipe_transfer *transfer = screen->get_tex_transfer( screen, texture, 0, 0, 0, - PIPE_TRANSFER_WRITE, 0, 0, texture->width[0], texture->height[0]); + PIPE_TRANSFER_WRITE, 0, 0, texture->width0, texture->height0); src += (dataStride * yoffset); for (i = 0; i < height; i++) { _vega_unpack_float_span_rgba(ctx, width, xoffset, src, dataFormat, temp); diff --git a/src/gallium/state_trackers/vega/mask.c b/src/gallium/state_trackers/vega/mask.c index 24650a37d50..b84103fdbac 100644 --- a/src/gallium/state_trackers/vega/mask.c +++ b/src/gallium/state_trackers/vega/mask.c @@ -426,7 +426,7 @@ static void mask_using_texture(struct pipe_texture *texture, if (!surface) return; if (!intersect_rectangles(surface->width, surface->height, - texture->width[0], texture->height[0], + texture->width0, texture->height0, x, y, width, height, offsets, loc)) return; @@ -493,9 +493,9 @@ struct vg_mask_layer * mask_layer_create(VGint width, VGint height) pt.format = PIPE_FORMAT_A8R8G8B8_UNORM; pf_get_block(PIPE_FORMAT_A8R8G8B8_UNORM, &pt.block); pt.last_level = 0; - pt.width[0] = width; - pt.height[0] = height; - pt.depth[0] = 1; + pt.width0 = width; + pt.height0 = height; + pt.depth0 = 1; pt.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER; pt.compressed = 0; @@ -607,8 +607,8 @@ void mask_render_to(struct path *path, struct vg_mask_layer *temp_layer; VGint width, height; - width = fb_buffers->alpha_mask->width[0]; - height = fb_buffers->alpha_mask->width[0]; + width = fb_buffers->alpha_mask->width0; + height = fb_buffers->alpha_mask->width0; temp_layer = mask_layer_create(width, height); diff --git a/src/gallium/state_trackers/vega/paint.c b/src/gallium/state_trackers/vega/paint.c index 04a6ba9cdcd..e8ca7d9e89b 100644 --- a/src/gallium/state_trackers/vega/paint.c +++ b/src/gallium/state_trackers/vega/paint.c @@ -151,9 +151,9 @@ static INLINE struct pipe_texture *create_gradient_texture(struct vg_paint *p) templ.target = PIPE_TEXTURE_1D; templ.format = PIPE_FORMAT_A8R8G8B8_UNORM; templ.last_level = 0; - templ.width[0] = 1024; - templ.height[0] = 1; - templ.depth[0] = 1; + templ.width0 = 1024; + templ.height0 = 1; + templ.depth0 = 1; pf_get_block(PIPE_FORMAT_A8R8G8B8_UNORM, &templ.block); templ.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER; @@ -328,8 +328,8 @@ static INLINE void paint_pattern_buffer(struct vg_paint *paint, void *buffer) map[4] = 0.f; map[5] = 1.f; - map[6] = paint->pattern.texture->width[0]; - map[7] = paint->pattern.texture->height[0]; + map[6] = paint->pattern.texture->width0; + map[7] = paint->pattern.texture->height0; { struct matrix mat; memcpy(&mat, &ctx->state.vg.fill_paint_to_user_matrix, diff --git a/src/gallium/state_trackers/vega/renderer.c b/src/gallium/state_trackers/vega/renderer.c index 396c88aa3d2..9085ed1bfe0 100644 --- a/src/gallium/state_trackers/vega/renderer.c +++ b/src/gallium/state_trackers/vega/renderer.c @@ -230,13 +230,13 @@ void renderer_draw_texture(struct renderer *r, struct pipe_buffer *buf; VGfloat s0, t0, s1, t1; - assert(tex->width[0] != 0); - assert(tex->height[0] != 0); + assert(tex->width0 != 0); + assert(tex->height0 != 0); - s0 = x1offset / tex->width[0]; - s1 = x2offset / tex->width[0]; - t0 = y1offset / tex->height[0]; - t1 = y2offset / tex->height[0]; + s0 = x1offset / tex->width0; + s1 = x2offset / tex->width0; + t0 = y1offset / tex->height0; + t1 = y2offset / tex->height0; cso_save_vertex_shader(r->cso); /* shaders */ @@ -276,10 +276,10 @@ void renderer_copy_texture(struct renderer *ctx, struct pipe_framebuffer_state fb; float s0, t0, s1, t1; - assert(src->width[0] != 0); - assert(src->height[0] != 0); - assert(dst->width[0] != 0); - assert(dst->height[0] != 0); + assert(src->width0 != 0); + assert(src->height0 != 0); + assert(dst->width0 != 0); + assert(dst->height0 != 0); #if 0 debug_printf("copy texture [%f, %f, %f, %f], [%f, %f, %f, %f]\n", @@ -287,10 +287,10 @@ void renderer_copy_texture(struct renderer *ctx, #endif #if 1 - s0 = sx1 / src->width[0]; - s1 = sx2 / src->width[0]; - t0 = sy1 / src->height[0]; - t1 = sy2 / src->height[0]; + s0 = sx1 / src->width0; + s1 = sx2 / src->width0; + t0 = sy1 / src->height0; + t1 = sy2 / src->height0; #else s0 = 0; s1 = 1; @@ -445,9 +445,9 @@ void renderer_copy_surface(struct renderer *ctx, texTemp.target = PIPE_TEXTURE_2D; texTemp.format = src->format; texTemp.last_level = 0; - texTemp.width[0] = srcW; - texTemp.height[0] = srcH; - texTemp.depth[0] = 1; + texTemp.width0 = srcW; + texTemp.height0 = srcH; + texTemp.depth0 = 1; pf_get_block(src->format, &texTemp.block); tex = screen->texture_create(screen, &texTemp); @@ -570,13 +570,13 @@ void renderer_texture_quad(struct renderer *r, struct pipe_buffer *buf; VGfloat s0, t0, s1, t1; - assert(tex->width[0] != 0); - assert(tex->height[0] != 0); + assert(tex->width0 != 0); + assert(tex->height0 != 0); - s0 = x1offset / tex->width[0]; - s1 = x2offset / tex->width[0]; - t0 = y1offset / tex->height[0]; - t1 = y2offset / tex->height[0]; + s0 = x1offset / tex->width0; + s1 = x2offset / tex->width0; + t0 = y1offset / tex->height0; + t1 = y2offset / tex->height0; cso_save_vertex_shader(r->cso); /* shaders */ diff --git a/src/gallium/state_trackers/vega/vg_tracker.c b/src/gallium/state_trackers/vega/vg_tracker.c index c4da01e52cc..d28463dd1b8 100644 --- a/src/gallium/state_trackers/vega/vg_tracker.c +++ b/src/gallium/state_trackers/vega/vg_tracker.c @@ -51,9 +51,9 @@ create_texture(struct pipe_context *pipe, enum pipe_format format, templ.target = PIPE_TEXTURE_2D; pf_get_block(templ.format, &templ.block); - templ.width[0] = width; - templ.height[0] = height; - templ.depth[0] = 1; + templ.width0 = width; + templ.height0 = height; + templ.depth0 = 1; templ.last_level = 0; if (pf_get_component_bits(format, PIPE_FORMAT_COMP_S)) { diff --git a/src/gallium/state_trackers/xorg/xorg_composite.c b/src/gallium/state_trackers/xorg/xorg_composite.c index a5975aad515..1c248a629e6 100644 --- a/src/gallium/state_trackers/xorg/xorg_composite.c +++ b/src/gallium/state_trackers/xorg/xorg_composite.c @@ -423,8 +423,6 @@ bind_samplers(struct exa_context *exa, int op, - - static INLINE boolean matrix_from_pict_transform(PictTransform *trans, float *matrix) { if (!trans) diff --git a/src/gallium/state_trackers/xorg/xorg_crtc.c b/src/gallium/state_trackers/xorg/xorg_crtc.c index ddcaedde37e..9e8c14d741c 100644 --- a/src/gallium/state_trackers/xorg/xorg_crtc.c +++ b/src/gallium/state_trackers/xorg/xorg_crtc.c @@ -196,10 +196,10 @@ crtc_load_cursor_argb_ga3d(xf86CrtcPtr crtc, CARD32 * image) templat.tex_usage |= PIPE_TEXTURE_USAGE_PRIMARY; templat.target = PIPE_TEXTURE_2D; templat.last_level = 0; - templat.depth[0] = 1; + templat.depth0 = 1; templat.format = PIPE_FORMAT_A8R8G8B8_UNORM; - templat.width[0] = 64; - templat.height[0] = 64; + templat.width0 = 64; + templat.height0 = 64; pf_get_block(templat.format, &templat.block); crtcp->cursor_tex = ms->screen->texture_create(ms->screen, diff --git a/src/gallium/state_trackers/xorg/xorg_dri2.c b/src/gallium/state_trackers/xorg/xorg_dri2.c index 4fa47548a43..2394f004d25 100644 --- a/src/gallium/state_trackers/xorg/xorg_dri2.c +++ b/src/gallium/state_trackers/xorg/xorg_dri2.c @@ -109,9 +109,9 @@ dri2_do_create_buffer(DrawablePtr pDraw, DRI2BufferPtr buffer, unsigned int form template.format = ms->ds_depth_bits_last ? PIPE_FORMAT_S8Z24_UNORM : PIPE_FORMAT_Z24S8_UNORM; pf_get_block(template.format, &template.block); - template.width[0] = pDraw->width; - template.height[0] = pDraw->height; - template.depth[0] = 1; + template.width0 = pDraw->width; + template.height0 = pDraw->height; + template.depth0 = 1; template.last_level = 0; template.tex_usage = PIPE_TEXTURE_USAGE_DEPTH_STENCIL | PIPE_TEXTURE_USAGE_DISPLAY_TARGET; diff --git a/src/gallium/state_trackers/xorg/xorg_exa.c b/src/gallium/state_trackers/xorg/xorg_exa.c index 32485add94e..16e683019c1 100644 --- a/src/gallium/state_trackers/xorg/xorg_exa.c +++ b/src/gallium/state_trackers/xorg/xorg_exa.c @@ -274,8 +274,8 @@ ExaPrepareAccess(PixmapPtr pPix, int index) PIPE_REFERENCED_FOR_WRITE) exa->pipe->flush(exa->pipe, 0, NULL); - assert(pPix->drawable.width <= priv->tex->width[0]); - assert(pPix->drawable.height <= priv->tex->height[0]); + assert(pPix->drawable.width <= priv->tex->width0); + assert(pPix->drawable.height <= priv->tex->height0); priv->map_transfer = exa->scrn->get_tex_transfer(exa->scrn, priv->tex, 0, 0, 0, @@ -283,9 +283,13 @@ ExaPrepareAccess(PixmapPtr pPix, int index) PIPE_TRANSFER_MAP_DIRECTLY | #endif PIPE_TRANSFER_READ_WRITE, +<<<<<<< HEAD:src/gallium/state_trackers/xorg/xorg_exa.c + 0, 0, priv->tex->width0, priv->tex->height0); +======= 0, 0, pPix->drawable.width, pPix->drawable.height ); +>>>>>>> origin/mesa_7_7_branch:src/gallium/state_trackers/xorg/xorg_exa.c if (!priv->map_transfer) #ifdef EXA_MIXED_PIXMAPS return FALSE; @@ -524,8 +528,8 @@ ExaCopy(PixmapPtr pDstPixmap, int srcX, int srcY, int dstX, int dstY, dstX, dstY, srcX, srcY, width, height, - exa->copy.src_texture->width[0], - exa->copy.src_texture->height[0]); + exa->copy.src_texture->width0, + exa->copy.src_texture->height0); } } @@ -861,8 +865,8 @@ ExaModifyPixmapHeader(PixmapPtr pPixmap, int width, int height, if (priv->tex) debug_printf(" ==> old texture %dx%d\n", - priv->tex->width[0], - priv->tex->height[0]); + priv->tex->width0, + priv->tex->height0); } @@ -890,8 +894,8 @@ ExaModifyPixmapHeader(PixmapPtr pPixmap, int width, int height, /* Deal with screen resize */ if ((exa->accel || priv->flags) && (!priv->tex || - !size_match(width, priv->tex->width[0]) || - !size_match(height, priv->tex->height[0]) || + !size_match(width, priv->tex->width0) || + !size_match(height, priv->tex->height0) || priv->tex_flags != priv->flags)) { struct pipe_texture *texture = NULL; struct pipe_texture template; @@ -901,15 +905,15 @@ ExaModifyPixmapHeader(PixmapPtr pPixmap, int width, int height, exa_get_pipe_format(depth, &template.format, &bitsPerPixel, &priv->picture_format); pf_get_block(template.format, &template.block); if (ROUND_UP_TEXTURES && priv->flags == 0) { - template.width[0] = util_next_power_of_two(width); - template.height[0] = util_next_power_of_two(height); + template.width0 = util_next_power_of_two(width); + template.height0 = util_next_power_of_two(height); } else { - template.width[0] = width; - template.height[0] = height; + template.width0 = width; + template.height0 = height; } - template.depth[0] = 1; + template.depth0 = 1; template.last_level = 0; template.tex_usage = PIPE_TEXTURE_USAGE_RENDER_TARGET | priv->flags; priv->tex_flags = priv->flags; @@ -924,12 +928,12 @@ ExaModifyPixmapHeader(PixmapPtr pPixmap, int width, int height, src_surf = xorg_gpu_surface(exa->pipe->screen, priv); if (exa->pipe->surface_copy) { exa->pipe->surface_copy(exa->pipe, dst_surf, 0, 0, src_surf, - 0, 0, min(width, texture->width[0]), - min(height, texture->height[0])); + 0, 0, min(width, texture->width0), + min(height, texture->height0)); } else { util_surface_copy(exa->pipe, FALSE, dst_surf, 0, 0, src_surf, - 0, 0, min(width, texture->width[0]), - min(height, texture->height[0])); + 0, 0, min(width, texture->width0), + min(height, texture->height0)); } exa->scrn->tex_surface_destroy(dst_surf); exa->scrn->tex_surface_destroy(src_surf); @@ -962,8 +966,8 @@ xorg_exa_set_texture(PixmapPtr pPixmap, struct pipe_texture *tex) if (!priv) return FALSE; - if (pPixmap->drawable.width != tex->width[0] || - pPixmap->drawable.height != tex->height[0]) + if (pPixmap->drawable.width != tex->width0 || + pPixmap->drawable.height != tex->height0) return FALSE; pipe_texture_reference(&priv->tex, tex); @@ -986,9 +990,9 @@ xorg_exa_create_root_texture(ScrnInfoPtr pScrn, template.target = PIPE_TEXTURE_2D; exa_get_pipe_format(depth, &template.format, &bitsPerPixel, &dummy); pf_get_block(template.format, &template.block); - template.width[0] = width; - template.height[0] = height; - template.depth[0] = 1; + template.width0 = width; + template.height0 = height; + template.depth0 = 1; template.last_level = 0; template.tex_usage |= PIPE_TEXTURE_USAGE_RENDER_TARGET; template.tex_usage |= PIPE_TEXTURE_USAGE_PRIMARY; diff --git a/src/gallium/state_trackers/xorg/xorg_exa_tgsi.c b/src/gallium/state_trackers/xorg/xorg_exa_tgsi.c index 13a9840bddd..89b794a09ac 100644 --- a/src/gallium/state_trackers/xorg/xorg_exa_tgsi.c +++ b/src/gallium/state_trackers/xorg/xorg_exa_tgsi.c @@ -47,22 +47,22 @@ static void print_fs_traits(int fs_traits) { const char *strings[] = { - "FS_COMPOSITE", // = 1 << 0, - "FS_MASK", // = 1 << 1, - "FS_SOLID_FILL", // = 1 << 2, - "FS_LINGRAD_FILL", // = 1 << 3, - "FS_RADGRAD_FILL", // = 1 << 4, - "FS_CA_FULL", // = 1 << 5, /* src.rgba * mask.rgba */ - "FS_CA_SRCALPHA", // = 1 << 6, /* src.aaaa * mask.rgba */ - "FS_YUV", // = 1 << 7, - "FS_SRC_REPEAT_NONE", // = 1 << 8, - "FS_MASK_REPEAT_NONE",// = 1 << 9, - "FS_SRC_SWIZZLE_RGB", // = 1 << 10, - "FS_MASK_SWIZZLE_RGB",// = 1 << 11, - "FS_SRC_SET_ALPHA", // = 1 << 12, - "FS_MASK_SET_ALPHA", // = 1 << 13, - "FS_SRC_LUMINANCE", // = 1 << 14, - "FS_MASK_LUMINANCE", // = 1 << 15, + "FS_COMPOSITE", /* = 1 << 0 */ + "FS_MASK", /* = 1 << 1 */ + "FS_SOLID_FILL", /* = 1 << 2 */ + "FS_LINGRAD_FILL", /* = 1 << 3 */ + "FS_RADGRAD_FILL", /* = 1 << 4 */ + "FS_CA_FULL", /* = 1 << 5 - src.rgba * mask.rgba */ + "FS_CA_SRCALPHA", /* = 1 << 6 - src.aaaa * mask.rgba */ + "FS_YUV", /* = 1 << 7 */ + "FS_SRC_REPEAT_NONE", /* = 1 << 8 */ + "FS_MASK_REPEAT_NONE",/* = 1 << 9 */ + "FS_SRC_SWIZZLE_RGB", /* = 1 << 10 */ + "FS_MASK_SWIZZLE_RGB",/* = 1 << 11 */ + "FS_SRC_SET_ALPHA", /* = 1 << 12 */ + "FS_MASK_SET_ALPHA", /* = 1 << 13 */ + "FS_SRC_LUMINANCE", /* = 1 << 14 */ + "FS_MASK_LUMINANCE", /* = 1 << 15 */ }; int i, k; debug_printf("%s: ", __func__); diff --git a/src/gallium/state_trackers/xorg/xorg_renderer.c b/src/gallium/state_trackers/xorg/xorg_renderer.c index cbb84a8c0da..8f73ec5fe11 100644 --- a/src/gallium/state_trackers/xorg/xorg_renderer.c +++ b/src/gallium/state_trackers/xorg/xorg_renderer.c @@ -170,14 +170,14 @@ add_vertex_data1(struct xorg_renderer *r, map_point(src_matrix, pt3[0], pt3[1], &pt3[0], &pt3[1]); } - s0 = pt0[0] / src->width[0]; - s1 = pt1[0] / src->width[0]; - s2 = pt2[0] / src->width[0]; - s3 = pt3[0] / src->width[0]; - t0 = pt0[1] / src->height[0]; - t1 = pt1[1] / src->height[0]; - t2 = pt2[1] / src->height[0]; - t3 = pt3[1] / src->height[0]; + s0 = pt0[0] / src->width0; + s1 = pt1[0] / src->width0; + s2 = pt2[0] / src->width0; + s3 = pt3[0] / src->width0; + t0 = pt0[1] / src->height0; + t1 = pt1[1] / src->height0; + t2 = pt2[1] / src->height0; + t3 = pt3[1] / src->height0; /* 1st vertex */ add_vertex_1tex(r, dstX, dstY, s0, t0); @@ -248,15 +248,15 @@ add_vertex_data2(struct xorg_renderer *r, map_point(mask_matrix, mpt1[0], mpt1[1], &mpt1[0], &mpt1[1]); } - src_s0 = spt0[0] / src->width[0]; - src_t0 = spt0[1] / src->height[0]; - src_s1 = spt1[0] / src->width[0]; - src_t1 = spt1[1] / src->height[0]; + src_s0 = spt0[0] / src->width0; + src_t0 = spt0[1] / src->height0; + src_s1 = spt1[0] / src->width0; + src_t1 = spt1[1] / src->height0; - mask_s0 = mpt0[0] / mask->width[0]; - mask_t0 = mpt0[1] / mask->height[0]; - mask_s1 = mpt1[0] / mask->width[0]; - mask_t1 = mpt1[1] / mask->height[0]; + mask_s0 = mpt0[0] / mask->width0; + mask_t0 = mpt0[1] / mask->height0; + mask_s1 = mpt1[0] / mask->width0; + mask_t1 = mpt1[1] / mask->height0; /* 1st vertex */ add_vertex_2tex(r, dstX, dstY, @@ -286,10 +286,10 @@ setup_vertex_data_yuv(struct xorg_renderer *r, spt1[0] = srcX + srcW; spt1[1] = srcY + srcH; - s0 = spt0[0] / tex[0]->width[0]; - t0 = spt0[1] / tex[0]->height[0]; - s1 = spt1[0] / tex[0]->width[0]; - t1 = spt1[1] / tex[0]->height[0]; + s0 = spt0[0] / tex[0]->width0; + t0 = spt0[1] / tex[0]->height0; + s1 = spt1[0] / tex[0]->width0; + t1 = spt1[1] / tex[0]->height0; /* 1st vertex */ add_vertex_1tex(r, dstX, dstY, s0, t0); @@ -509,9 +509,9 @@ renderer_clone_texture(struct xorg_renderer *r, templ.target = PIPE_TEXTURE_2D; templ.format = format; templ.last_level = 0; - templ.width[0] = src->width[0]; - templ.height[0] = src->height[0]; - templ.depth[0] = 1; + templ.width0 = src->width0; + templ.height0 = src->height0; + templ.depth0 = 1; pf_get_block(format, &templ.block); templ.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER; @@ -533,13 +533,13 @@ renderer_clone_texture(struct xorg_renderer *r, ps_tex, /* dest */ 0, 0, /* destx/y */ ps_read, - 0, 0, src->width[0], src->height[0]); + 0, 0, src->width0, src->height0); } else { util_surface_copy(pipe, FALSE, ps_tex, /* dest */ 0, 0, /* destx/y */ ps_read, - 0, 0, src->width[0], src->height[0]); + 0, 0, src->width0, src->height0); } pipe_surface_reference(&ps_read, NULL); pipe_surface_reference(&ps_tex, NULL); @@ -559,6 +559,69 @@ void renderer_copy_pixmap(struct xorg_renderer *r, float s0, t0, s1, t1; float x0, y0, x1, y1; +<<<<<<< HEAD:src/gallium/state_trackers/xorg/xorg_renderer.c + if (r->pipe->is_texture_referenced(r->pipe, src, 0, 0) & + PIPE_REFERENCED_FOR_WRITE) + r->pipe->flush(r->pipe, PIPE_FLUSH_RENDER_CACHE, NULL); + + dst_loc[0] = dx; + dst_loc[1] = dy; + dst_loc[2] = width; + dst_loc[3] = height; + dst_bounds[0] = 0.f; + dst_bounds[1] = 0.f; + dst_bounds[2] = dst->width0; + dst_bounds[3] = dst->height0; + + src_loc[0] = sx; + src_loc[1] = sy; + src_loc[2] = width; + src_loc[3] = height; + src_bounds[0] = 0.f; + src_bounds[1] = 0.f; + src_bounds[2] = src->width0; + src_bounds[3] = src->height0; + + bound_rect(src_loc, src_bounds, src_shift); + bound_rect(dst_loc, dst_bounds, dst_shift); + shift[0] = src_shift[0] - dst_shift[0]; + shift[1] = src_shift[1] - dst_shift[1]; + + if (shift[0] < 0) + shift_rectx(src_loc, src_bounds, -shift[0]); + else + shift_rectx(dst_loc, dst_bounds, shift[0]); + + if (shift[1] < 0) + shift_recty(src_loc, src_bounds, -shift[1]); + else + shift_recty(dst_loc, dst_bounds, shift[1]); + + sync_size(src_loc, dst_loc); + + if (src_loc[2] >= 0 && src_loc[3] >= 0 && + dst_loc[2] >= 0 && dst_loc[3] >= 0) { + struct pipe_texture *temp_src = src; + + if (src == dst) + temp_src = create_sampler_texture(r, src); + + renderer_copy_texture(r, + temp_src, + src_loc[0], + src_loc[1], + src_loc[0] + src_loc[2], + src_loc[1] + src_loc[3], + dst, + dst_loc[0], + dst_loc[1], + dst_loc[0] + dst_loc[2], + dst_loc[1] + dst_loc[3]); + + if (src == dst) + pipe_texture_reference(&temp_src, NULL); + } +======= /* XXX: could put the texcoord scaling calculation into the vertex * shader. @@ -579,6 +642,7 @@ void renderer_copy_pixmap(struct xorg_renderer *r, add_vertex_1tex(r, x1, y0, s1, t0); add_vertex_1tex(r, x1, y1, s1, t1); add_vertex_1tex(r, x0, y1, s0, t1); +>>>>>>> origin/mesa_7_7_branch:src/gallium/state_trackers/xorg/xorg_renderer.c } diff --git a/src/gallium/state_trackers/xorg/xorg_xv.c b/src/gallium/state_trackers/xorg/xorg_xv.c index b3315dccad8..b8eca8c8176 100644 --- a/src/gallium/state_trackers/xorg/xorg_xv.c +++ b/src/gallium/state_trackers/xorg/xorg_xv.c @@ -73,10 +73,11 @@ static XF86VideoEncodingRec DummyEncoding[1] = { } }; -#define NUM_IMAGES 2 +#define NUM_IMAGES 3 static XF86ImageRec Images[NUM_IMAGES] = { XVIMAGE_UYVY, XVIMAGE_YUY2, + XVIMAGE_YV12, }; struct xorg_xv_port_priv { @@ -166,9 +167,9 @@ create_component_texture(struct pipe_context *pipe, templ.target = PIPE_TEXTURE_2D; templ.format = PIPE_FORMAT_L8_UNORM; templ.last_level = 0; - templ.width[0] = width; - templ.height[0] = height; - templ.depth[0] = 1; + templ.width0 = width; + templ.height0 = height; + templ.depth0 = 1; pf_get_block(PIPE_FORMAT_L8_UNORM, &templ.block); templ.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER; @@ -182,18 +183,18 @@ check_yuv_textures(struct xorg_xv_port_priv *priv, int width, int height) { struct pipe_texture **dst = priv->yuv[priv->current_set]; if (!dst[0] || - dst[0]->width[0] != width || - dst[0]->height[0] != height) { + dst[0]->width0 != width || + dst[0]->height0 != height) { pipe_texture_reference(&dst[0], NULL); } if (!dst[1] || - dst[1]->width[0] != width || - dst[1]->height[0] != height) { + dst[1]->width0 != width || + dst[1]->height0 != height) { pipe_texture_reference(&dst[1], NULL); } if (!dst[2] || - dst[2]->width[0] != width || - dst[2]->height[0] != height) { + dst[2]->width0 != width || + dst[2]->height0 != height) { pipe_texture_reference(&dst[2], NULL); } @@ -532,6 +533,7 @@ put_image(ScrnInfoPtr pScrn, switch (id) { case FOURCC_UYVY: case FOURCC_YUY2: + case FOURCC_YV12: default: srcPitch = width << 1; break; @@ -580,6 +582,7 @@ query_image_attributes(ScrnInfoPtr pScrn, switch (id) { case FOURCC_UYVY: case FOURCC_YUY2: + case FOURCC_YV12: default: size = *w << 1; if (pitches) diff --git a/src/gallium/state_trackers/xorg/xvmc/surface.c b/src/gallium/state_trackers/xorg/xvmc/surface.c index bf9038f356e..8cb73f48970 100644 --- a/src/gallium/state_trackers/xorg/xvmc/surface.c +++ b/src/gallium/state_trackers/xorg/xvmc/surface.c @@ -103,9 +103,9 @@ CreateOrResizeBackBuffer(struct pipe_video_context *vpipe, unsigned int width, u /* XXX: Needs to match the drawable's format? */ template.format = PIPE_FORMAT_X8R8G8B8_UNORM; template.last_level = 0; - template.width[0] = width; - template.height[0] = height; - template.depth[0] = 1; + template.width0 = width; + template.height0 = height; + template.depth0 = 1; pf_get_block(template.format, &template.block); template.tex_usage = PIPE_TEXTURE_USAGE_DISPLAY_TARGET; diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c index 317dc44d22f..d4978613247 100644 --- a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c +++ b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c @@ -24,10 +24,10 @@ dri_surface_from_handle(struct drm_api *api, struct pipe_screen *pscreen, tmpl.tex_usage = PIPE_TEXTURE_USAGE_PRIMARY; tmpl.target = PIPE_TEXTURE_2D; tmpl.last_level = 0; - tmpl.depth[0] = 1; + tmpl.depth0 = 1; tmpl.format = format; - tmpl.width[0] = width; - tmpl.height[0] = height; + tmpl.width0 = width; + tmpl.height0 = height; pf_get_block(tmpl.format, &tmpl.block); pt = api->texture_from_shared_handle(api, pscreen, &tmpl, diff --git a/src/gallium/winsys/drm/radeon/core/radeon_buffer.c b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c index 81cd9dc4fb1..74afffc9cfa 100644 --- a/src/gallium/winsys/drm/radeon/core/radeon_buffer.c +++ b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c @@ -317,9 +317,9 @@ struct pipe_surface *radeon_surface_from_handle(struct radeon_context *radeon_co memset(&tmpl, 0, sizeof(tmpl)); tmpl.tex_usage = PIPE_TEXTURE_USAGE_DISPLAY_TARGET; tmpl.target = PIPE_TEXTURE_2D; - tmpl.width[0] = w; - tmpl.height[0] = h; - tmpl.depth[0] = 1; + tmpl.width0 = w; + tmpl.height0 = h; + tmpl.depth0 = 1; tmpl.format = format; pf_get_block(tmpl.format, &tmpl.block); tmpl.nblocksx[0] = pf_get_nblocksx(&tmpl.block, w); diff --git a/src/gallium/winsys/egl_xlib/egl_xlib.c b/src/gallium/winsys/egl_xlib/egl_xlib.c index d02f8250478..599973ce127 100644 --- a/src/gallium/winsys/egl_xlib/egl_xlib.c +++ b/src/gallium/winsys/egl_xlib/egl_xlib.c @@ -41,6 +41,7 @@ #include "pipe/p_state.h" #include "pipe/internal/p_winsys_screen.h" #include "util/u_memory.h" +#include "util/u_math.h" #include "softpipe/sp_winsys.h" #include "softpipe/sp_texture.h" @@ -138,17 +139,6 @@ lookup_context(_EGLContext *ctx) } -static unsigned int -bitcount(unsigned int n) -{ - unsigned int bits; - for (bits = 0; n > 0; n = n >> 1) { - bits += (n & 1); - } - return bits; -} - - /** * Create the EGLConfigs. (one per X visual) */ @@ -174,9 +164,9 @@ create_configs(struct xlib_egl_display *xdpy, _EGLDisplay *disp) for (i = 0; i < num_visuals; i++) { _EGLConfig *config = calloc(1, sizeof(_EGLConfig)); int id = i + 1; - int rbits = bitcount(visInfo[i].red_mask); - int gbits = bitcount(visInfo[i].green_mask); - int bbits = bitcount(visInfo[i].blue_mask); + int rbits = util_bitcount(visInfo[i].red_mask); + int gbits = util_bitcount(visInfo[i].green_mask); + int bbits = util_bitcount(visInfo[i].blue_mask); int abits = bbits == 8 ? 8 : 0; int zbits = 24; int sbits = 8; diff --git a/src/mesa/drivers/dri/ffb/ffb_vbtmp.h b/src/mesa/drivers/dri/ffb/ffb_vbtmp.h index 0495d0e2767..c548ef3ad58 100644 --- a/src/mesa/drivers/dri/ffb/ffb_vbtmp.h +++ b/src/mesa/drivers/dri/ffb/ffb_vbtmp.h @@ -38,11 +38,11 @@ static void TAG(emit)(GLcontext *ctx, GLuint start, GLuint end) #endif #if (IND & (FFB_VB_RGBA_BIT)) - col0 = VB->ColorPtr[0]->data; - col0_stride = VB->ColorPtr[0]->stride; + col0 = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col0_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; #if (IND & (FFB_VB_TWOSIDE_BIT)) - col1 = VB->ColorPtr[1]->data; - col1_stride = VB->ColorPtr[1]->stride; + col1 = VB->BackfaceColorPtr->data; + col1_stride = VB->BackfaceColorPtr->stride; #endif #endif diff --git a/src/mesa/drivers/dri/gamma/gamma_render.c b/src/mesa/drivers/dri/gamma/gamma_render.c index 1b9fd169f46..a03a93d132c 100644 --- a/src/mesa/drivers/dri/gamma/gamma_render.c +++ b/src/mesa/drivers/dri/gamma/gamma_render.c @@ -53,13 +53,13 @@ static void gamma_emit( GLcontext *ctx, GLuint start, GLuint end) GLfloat (*tc0)[4] = 0; GLuint tc0_size = 0; - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; if (ctx->Texture.Unit[0]._ReallyEnabled) { - tc0_stride = VB->TexCoordPtr[0]->stride; - tc0 = VB->TexCoordPtr[0]->data; - tc0_size = VB->TexCoordPtr[0]->size; + tc0_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0]->stride; + tc0 = VB->AttribPtr[_TNL_ATTRIB_TEX0]->data; + tc0_size = VB->AttribPtr[_TNL_ATTRIB_TEX0]->size; coord = VB->ClipPtr->data; coord_stride = VB->ClipPtr->stride; } else { diff --git a/src/mesa/drivers/dri/i915/i830_texstate.c b/src/mesa/drivers/dri/i915/i830_texstate.c index f4bbb53b863..c62281d341e 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 a6f554701e6..c05c7759ac5 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 d9c61446f52..9e4d3180360 100644 --- a/src/mesa/drivers/dri/i915/i915_fragprog.c +++ b/src/mesa/drivers/dri/i915/i915_fragprog.c @@ -1301,7 +1301,7 @@ i915ValidateFragmentProgram(struct i915_context *i915) for (i = 0; i < p->ctx->Const.MaxTextureCoordUnits; i++) { if (inputsRead & FRAG_BIT_TEX(i)) { - int sz = VB->TexCoordPtr[i]->size; + int sz = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->size; s2 &= ~S2_TEXCOORD_FMT(i, S2_TEXCOORD_FMT0_MASK); s2 |= S2_TEXCOORD_FMT(i, SZ_TO_HW(sz)); diff --git a/src/mesa/drivers/dri/i915/i915_texstate.c b/src/mesa/drivers/dri/i915/i915_texstate.c index d6689af53f6..1bacd51aec5 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 77ba8d55819..3e7b5101cca 100644 --- a/src/mesa/drivers/dri/i915/i915_vtbl.c +++ b/src/mesa/drivers/dri/i915/i915_vtbl.c @@ -667,9 +667,6 @@ i915_new_batch(struct intel_context *intel) * difficulties associated with them (physical address requirements). */ i915->state.emitted = 0; - - /* Check that we didn't just wrap our batchbuffer at a bad time. */ - assert(!intel->no_batch_wrap); } static void diff --git a/src/mesa/drivers/dri/i915/intel_tris.c b/src/mesa/drivers/dri/i915/intel_tris.c index bc527aae47a..8a3ab39bc28 100644 --- a/src/mesa/drivers/dri/i915/intel_tris.c +++ b/src/mesa/drivers/dri/i915/intel_tris.c @@ -1250,81 +1250,6 @@ union fi GLint i; }; - -/**********************************************************************/ -/* Used only with the metaops callbacks. */ -/**********************************************************************/ -static void -intel_meta_draw_poly(struct intel_context *intel, - GLuint n, - GLfloat xy[][2], - GLfloat z, GLuint color, GLfloat tex[][2]) -{ - union fi *vb; - GLint i; - unsigned int saved_vertex_size = intel->vertex_size; - - LOCK_HARDWARE(intel); - - intel->vertex_size = 6; - - /* All 3d primitives should be emitted with LOOP_CLIPRECTS, - * otherwise the drawing origin (DR4) might not be set correctly. - */ - intel_set_prim(intel, PRIM3D_TRIFAN); - vb = (union fi *) intel_get_prim_space(intel, n); - - for (i = 0; i < n; i++) { - vb[0].f = xy[i][0]; - vb[1].f = xy[i][1]; - vb[2].f = z; - vb[3].i = color; - vb[4].f = tex[i][0]; - vb[5].f = tex[i][1]; - vb += 6; - } - - INTEL_FIREVERTICES(intel); - - intel->vertex_size = saved_vertex_size; - - UNLOCK_HARDWARE(intel); -} - -static void -intel_meta_draw_quad(struct intel_context *intel, - GLfloat x0, GLfloat x1, - GLfloat y0, GLfloat y1, - GLfloat z, - GLuint color, - GLfloat s0, GLfloat s1, GLfloat t0, GLfloat t1) -{ - GLfloat xy[4][2]; - GLfloat tex[4][2]; - - xy[0][0] = x0; - xy[0][1] = y0; - xy[1][0] = x1; - xy[1][1] = y0; - xy[2][0] = x1; - xy[2][1] = y1; - xy[3][0] = x0; - xy[3][1] = y1; - - tex[0][0] = s0; - tex[0][1] = t0; - tex[1][0] = s1; - tex[1][1] = t0; - tex[2][0] = s1; - tex[2][1] = t1; - tex[3][0] = s0; - tex[3][1] = t1; - - intel_meta_draw_poly(intel, 4, xy, z, color, tex); -} - - - /**********************************************************************/ /* Initialization. */ /**********************************************************************/ @@ -1333,7 +1258,6 @@ intel_meta_draw_quad(struct intel_context *intel, void intelInitTriFuncs(GLcontext * ctx) { - struct intel_context *intel = intel_context(ctx); TNLcontext *tnl = TNL_CONTEXT(ctx); static int firsttime = 1; @@ -1350,6 +1274,4 @@ intelInitTriFuncs(GLcontext * ctx) tnl->Driver.Render.BuildVertices = _tnl_build_vertices; tnl->Driver.Render.CopyPV = _tnl_copy_pv; tnl->Driver.Render.Interp = _tnl_interp; - - intel->vtbl.meta_draw_quad = intel_meta_draw_quad; } diff --git a/src/mesa/drivers/dri/i965/brw_context.c b/src/mesa/drivers/dri/i965/brw_context.c index 48685c087b8..8bdda60697b 100644 --- a/src/mesa/drivers/dri/i965/brw_context.c +++ b/src/mesa/drivers/dri/i965/brw_context.c @@ -111,6 +111,7 @@ GLboolean brwCreateContext( const __GLcontextModes *mesaVis, ctx->Const.MaxTextureUnits = MIN2(ctx->Const.MaxTextureCoordUnits, ctx->Const.MaxTextureImageUnits); ctx->Const.MaxVertexTextureImageUnits = 0; /* no vertex shader textures */ + ctx->Const.MaxCombinedTextureImageUnits = 0; /* Mesa limits textures to 4kx4k; it would be nice to fix that someday */ diff --git a/src/mesa/drivers/dri/i965/brw_context.h b/src/mesa/drivers/dri/i965/brw_context.h index fded47aa2ff..e73e21433c5 100644 --- a/src/mesa/drivers/dri/i965/brw_context.h +++ b/src/mesa/drivers/dri/i965/brw_context.h @@ -172,8 +172,8 @@ struct brw_fragment_program { GLuint id; /**< serial no. to identify frag progs, never re-used */ GLboolean isGLSL; /**< really, any IF/LOOP/CONT/BREAK instructions */ - dri_bo *const_buffer; /** Program constant buffer/surface */ GLboolean use_const_buffer; + dri_bo *const_buffer; /** Program constant buffer/surface */ /** for debugging, which texture units are referenced */ GLbitfield tex_units_used; @@ -438,7 +438,6 @@ struct brw_context GLuint primitive; GLboolean emit_state_always; - GLboolean no_batch_wrap; struct { struct brw_state_flags dirty; diff --git a/src/mesa/drivers/dri/i965/brw_draw.c b/src/mesa/drivers/dri/i965/brw_draw.c index 8bcb6083f71..7ad860898fc 100644 --- a/src/mesa/drivers/dri/i965/brw_draw.c +++ b/src/mesa/drivers/dri/i965/brw_draw.c @@ -145,7 +145,7 @@ static void brw_emit_prim(struct brw_context *brw, prim_packet.base_vert_location = prim->basevertex; /* Can't wrap here, since we rely on the validated state. */ - brw->no_batch_wrap = GL_TRUE; + intel->no_batch_wrap = GL_TRUE; /* If we're set to always flush, do it before and after the primitive emit. * We want to catch both missed flushes that hurt instruction/state cache @@ -163,7 +163,7 @@ static void brw_emit_prim(struct brw_context *brw, intel_batchbuffer_emit_mi_flush(intel->batch); } - brw->no_batch_wrap = GL_FALSE; + intel->no_batch_wrap = GL_FALSE; } static void brw_merge_inputs( struct brw_context *brw, diff --git a/src/mesa/drivers/dri/i965/brw_draw_upload.c b/src/mesa/drivers/dri/i965/brw_draw_upload.c index 271a88dae0b..7c796dae936 100644 --- a/src/mesa/drivers/dri/i965/brw_draw_upload.c +++ b/src/mesa/drivers/dri/i965/brw_draw_upload.c @@ -243,14 +243,6 @@ static void wrap_buffers( struct brw_context *brw, dri_bo_unreference(brw->vb.upload.bo); brw->vb.upload.bo = dri_bo_alloc(brw->intel.bufmgr, "temporary VBO", size, 1); - - /* Set the internal VBO\ to no-backing-store. We only use them as a - * temporary within a brw_try_draw_prims while the lock is held. - */ - /* DON'T DO THIS AS IF WE HAVE TO RE-ORG MEMORY WE NEED SOMEWHERE WITH - FAKE TO PUSH THIS STUFF */ -// if (!brw->intel.ttm) -// dri_bo_fake_disable_backing_store(brw->vb.upload.bo, NULL, NULL); } static void get_space( struct brw_context *brw, diff --git a/src/mesa/drivers/dri/i965/brw_vtbl.c b/src/mesa/drivers/dri/i965/brw_vtbl.c index 34aaea37360..72749b3859d 100644 --- a/src/mesa/drivers/dri/i965/brw_vtbl.c +++ b/src/mesa/drivers/dri/i965/brw_vtbl.c @@ -150,9 +150,6 @@ static void brw_new_batch( struct intel_context *intel ) { struct brw_context *brw = brw_context(&intel->ctx); - /* Check that we didn't just wrap our batchbuffer at a bad time. */ - assert(!brw->no_batch_wrap); - brw->curbe.need_new_bo = GL_TRUE; /* Mark all context state as needing to be re-emitted. diff --git a/src/mesa/drivers/dri/i965/brw_wm.h b/src/mesa/drivers/dri/i965/brw_wm.h index 9dcb6e14bbe..b9b987ea70a 100644 --- a/src/mesa/drivers/dri/i965/brw_wm.h +++ b/src/mesa/drivers/dri/i965/brw_wm.h @@ -76,10 +76,10 @@ struct brw_wm_prog_key { GLushort tex_swizzles[BRW_MAX_TEX_UNIT]; - GLuint program_string_id:32; GLushort origin_x, origin_y; GLushort drawable_height; GLbitfield64 vp_outputs_written; + GLuint program_string_id:32; }; diff --git a/src/mesa/drivers/dri/i965/brw_wm_fp.c b/src/mesa/drivers/dri/i965/brw_wm_fp.c index 7d031795883..3737faf26fb 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_fp.c +++ b/src/mesa/drivers/dri/i965/brw_wm_fp.c @@ -138,7 +138,6 @@ static struct prog_dst_register dst_reg(GLuint file, GLuint idx) reg.CondMask = COND_TR; reg.CondSwizzle = 0; reg.CondSrc = 0; - reg.pad = 0; return reg; } diff --git a/src/mesa/drivers/dri/intel/intel_batchbuffer.c b/src/mesa/drivers/dri/intel/intel_batchbuffer.c index ca6e2fa5b11..2eae9b66d87 100644 --- a/src/mesa/drivers/dri/intel/intel_batchbuffer.c +++ b/src/mesa/drivers/dri/intel/intel_batchbuffer.c @@ -80,7 +80,7 @@ intel_batchbuffer_reset(struct intel_batchbuffer *batch) batch->buf = NULL; } - if (!batch->buffer && intel->ttm == GL_TRUE) + if (!batch->buffer) batch->buffer = malloc (intel->maxBatchSize); batch->buf = dri_bo_alloc(intel->bufmgr, "batchbuffer", @@ -212,7 +212,7 @@ _intel_batchbuffer_flush(struct intel_batchbuffer *batch, const char *file, batch->reserved_space = 0; /* Emit a flush if the bufmgr doesn't do it for us. */ - if (intel->always_flush_cache || !intel->ttm) { + if (intel->always_flush_cache) { intel_batchbuffer_emit_mi_flush(batch); used = batch->ptr - batch->map; } @@ -244,6 +244,9 @@ _intel_batchbuffer_flush(struct intel_batchbuffer *batch, const char *file, if (intel->vtbl.finish_batch) intel->vtbl.finish_batch(intel); + /* Check that we didn't just wrap our batchbuffer at a bad time. */ + assert(!intel->no_batch_wrap); + batch->reserved_space = BATCH_RESERVED; /* TODO: Just pass the relocation list and dma buffer up to the diff --git a/src/mesa/drivers/dri/intel/intel_blit.c b/src/mesa/drivers/dri/intel/intel_blit.c index 817223da41d..f14854602b6 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 2aeca6b81b0..1434ae530be 100644 --- a/src/mesa/drivers/dri/intel/intel_context.c +++ b/src/mesa/drivers/dri/intel/intel_context.c @@ -176,9 +176,7 @@ intelGetString(GLcontext * ctx, GLenum name) break; } - (void) driGetRendererString(buffer, chipset, - (intel->ttm) ? DRIVER_DATE_GEM : DRIVER_DATE, - 0); + (void) driGetRendererString(buffer, chipset, DRIVER_DATE_GEM, 0); return (GLubyte *) buffer; default: @@ -601,6 +599,7 @@ intelInitContext(struct intel_context *intel, __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; int fthrottle_mode; + int bo_reuse_mode; if (!_mesa_initialize_context(&intel->ctx, mesaVis, shareCtx, functions, (void *) intel)) { @@ -635,18 +634,14 @@ intelInitContext(struct intel_context *intel, intel->maxBatchSize = BATCH_SZ; intel->bufmgr = intelScreen->bufmgr; - intel->ttm = intelScreen->ttm; - if (intel->ttm) { - int bo_reuse_mode; - bo_reuse_mode = driQueryOptioni(&intel->optionCache, "bo_reuse"); - switch (bo_reuse_mode) { - case DRI_CONF_BO_REUSE_DISABLED: - break; - case DRI_CONF_BO_REUSE_ALL: - intel_bufmgr_gem_enable_reuse(intel->bufmgr); - break; - } + bo_reuse_mode = driQueryOptioni(&intel->optionCache, "bo_reuse"); + switch (bo_reuse_mode) { + case DRI_CONF_BO_REUSE_DISABLED: + break; + case DRI_CONF_BO_REUSE_ALL: + intel_bufmgr_gem_enable_reuse(intel->bufmgr); + break; } /* This doesn't yet catch all non-conformant rendering, but it's a @@ -733,12 +728,6 @@ intelInitContext(struct intel_context *intel, intel->RenderIndex = ~0; fthrottle_mode = driQueryOptioni(&intel->optionCache, "fthrottle_mode"); - intel->irqsEmitted = 0; - - intel->do_irqs = (intel->intelScreen->irq_active && - fthrottle_mode == DRI_CONF_FTHROTTLE_IRQS); - - intel->do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS); if (intel->gen >= 4 && !intel->intelScreen->irq_active) { _mesa_printf("IRQs not active. Exiting\n"); @@ -1058,21 +1047,6 @@ intelContendedLock(struct intel_context *intel, GLuint flags) sarea->ctxOwner = me; } - /* If the last consumer of the texture memory wasn't us, notify the fake - * bufmgr and record the new owner. We should have the memory shared - * between contexts of a single fake bufmgr, but this will at least make - * things correct for now. - */ - if (!intel->ttm && sarea->texAge != intel->hHWContext) { - sarea->texAge = intel->hHWContext; - intel_bufmgr_fake_contended_lock_take(intel->bufmgr); - if (INTEL_DEBUG & DEBUG_BATCH) - intel_decode_context_reset(); - if (INTEL_DEBUG & DEBUG_BUFMGR) - fprintf(stderr, "Lost Textures: sarea->texAge %x hw context %x\n", - sarea->ctxOwner, intel->hHWContext); - } - /* Drawable changed? */ if (dPriv && intel->lastStamp != dPriv->lastStamp) { diff --git a/src/mesa/drivers/dri/intel/intel_context.h b/src/mesa/drivers/dri/intel/intel_context.h index eb7be7ddd0f..481202c971b 100644 --- a/src/mesa/drivers/dri/intel/intel_context.h +++ b/src/mesa/drivers/dri/intel/intel_context.h @@ -135,14 +135,6 @@ struct intel_context struct intel_region * draw_region, struct intel_region * depth_region); - void (*meta_draw_quad)(struct intel_context *intel, - GLfloat x0, GLfloat x1, - GLfloat y0, GLfloat y1, - GLfloat z, - GLuint color, /* ARGB32 */ - GLfloat s0, GLfloat s1, - GLfloat t0, GLfloat t1); - void (*meta_color_mask) (struct intel_context * intel, GLboolean); void (*meta_stencil_replace) (struct intel_context * intel, @@ -189,12 +181,6 @@ struct intel_context struct intel_region *back_region; struct intel_region *depth_region; - /** - * This value indicates that the kernel memory manager is being used - * instead of the fake client-side memory manager. - */ - GLboolean ttm; - struct intel_batchbuffer *batch; drm_intel_bo *first_post_swapbuffers_batch; GLboolean no_batch_wrap; @@ -217,10 +203,6 @@ struct intel_context char *prevLockFile; int prevLockLine; - GLuint ClearColor565; - GLuint ClearColor8888; - - /* Offsets of fields within the current vertex: */ GLuint coloroffset; @@ -237,6 +219,7 @@ struct intel_context GLboolean hw_stipple; GLboolean depth_buffer_is_float; GLboolean no_rast; + GLboolean no_hw; GLboolean always_flush_batch; GLboolean always_flush_cache; @@ -302,13 +285,6 @@ struct intel_context GLboolean use_early_z; drm_clip_rect_t fboRect; /**< cliprect for FBO rendering */ - int perf_boxes; - - GLuint do_usleeps; - int do_irqs; - GLuint irqsEmitted; - - GLboolean scissor; drm_clip_rect_t draw_rect; drm_clip_rect_t scissor_rect; @@ -325,8 +301,6 @@ struct intel_context GLuint lastStamp; - GLboolean no_hw; - /** * Configuration cache */ @@ -374,29 +348,6 @@ do { \ } while (0) /* ================================================================ - * Color packing: - */ - -#define INTEL_PACKCOLOR4444(r,g,b,a) \ - ((((a) & 0xf0) << 8) | (((r) & 0xf0) << 4) | ((g) & 0xf0) | ((b) >> 4)) - -#define INTEL_PACKCOLOR1555(r,g,b,a) \ - ((((r) & 0xf8) << 7) | (((g) & 0xf8) << 2) | (((b) & 0xf8) >> 3) | \ - ((a) ? 0x8000 : 0)) - -#define INTEL_PACKCOLOR565(r,g,b) \ - ((((r) & 0xf8) << 8) | (((g) & 0xfc) << 3) | (((b) & 0xf8) >> 3)) - -#define INTEL_PACKCOLOR8888(r,g,b,a) \ - ((a<<24) | (r<<16) | (g<<8) | b) - -#define INTEL_PACKCOLOR(format, r, g, b, a) \ -(format == DV_PF_555 ? INTEL_PACKCOLOR1555(r,g,b,a) : \ - (format == DV_PF_565 ? INTEL_PACKCOLOR565(r,g,b) : \ - (format == DV_PF_8888 ? INTEL_PACKCOLOR8888(r,g,b,a) : \ - 0))) - -/* ================================================================ * From linux kernel i386 header files, copes with odd sizes better * than COPY_DWORDS would: * XXX Put this in src/mesa/main/imports.h ??? diff --git a/src/mesa/drivers/dri/intel/intel_extensions.c b/src/mesa/drivers/dri/intel/intel_extensions.c index 1682e115cc1..f5fe543b5df 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 abb3024bfbb..6a565f80cf4 100644 --- a/src/mesa/drivers/dri/intel/intel_mipmap_tree.c +++ b/src/mesa/drivers/dri/intel/intel_mipmap_tree.c @@ -224,16 +224,12 @@ int intel_miptree_pitch_align (struct intel_context *intel, if (!mt->compressed) { int pitch_align; - if (intel->ttm) { - /* XXX: Align pitch to multiple of 64 bytes for now to allow - * render-to-texture to work in all cases. This should probably be - * replaced at some point by some scheme to only do this when really - * necessary. - */ - pitch_align = 64; - } else { - pitch_align = 4; - } + /* XXX: Align pitch to multiple of 64 bytes for now to allow + * render-to-texture to work in all cases. This should probably be + * replaced at some point by some scheme to only do this when really + * necessary. + */ + pitch_align = 64; if (tiling == I915_TILING_X) pitch_align = 512; diff --git a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c index 99330b6ddfe..9572b673269 100644 --- a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c +++ b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c @@ -228,10 +228,9 @@ do_blit_bitmap( GLcontext *ctx, UNCLAMPED_FLOAT_TO_UBYTE(ubcolor[3], tmpColor[3]); if (dst->cpp == 2) - color = INTEL_PACKCOLOR565(ubcolor[0], ubcolor[1], ubcolor[2]); + color = PACK_COLOR_565(ubcolor[0], ubcolor[1], ubcolor[2]); else - color = INTEL_PACKCOLOR8888(ubcolor[0], ubcolor[1], - ubcolor[2], ubcolor[3]); + color = PACK_COLOR_8888(ubcolor[3], ubcolor[0], ubcolor[1], ubcolor[2]); if (!intel_check_blit_fragment_ops(ctx, tmpColor[3] == 1.0F)) return GL_FALSE; diff --git a/src/mesa/drivers/dri/intel/intel_regions.c b/src/mesa/drivers/dri/intel/intel_regions.c index 80975163d47..d6b9dc44466 100644 --- a/src/mesa/drivers/dri/intel/intel_regions.c +++ b/src/mesa/drivers/dri/intel/intel_regions.c @@ -542,55 +542,18 @@ intel_recreate_static(struct intel_context *intel, region->buffer = NULL; } - if (intel->ttm) { - assert(region_desc->bo_handle != -1); - region->buffer = intel_bo_gem_create_from_name(intel->bufmgr, - name, - region_desc->bo_handle); - - ret = dri_bo_get_tiling(region->buffer, ®ion->tiling, - ®ion->bit_6_swizzle); - if (ret != 0) { - fprintf(stderr, "Couldn't get tiling of buffer %d (%s): %s\n", - region_desc->bo_handle, name, strerror(-ret)); - intel_region_release(®ion); - return NULL; - } - } else { - if (region->classic_map != NULL) { - drmUnmap(region->classic_map, - region->pitch * region->cpp * region->height); - region->classic_map = NULL; - } - ret = drmMap(intel->driFd, region_desc->handle, - region->pitch * region->cpp * region->height, - ®ion->classic_map); - if (ret != 0) { - fprintf(stderr, "Failed to drmMap %s buffer\n", name); - free(region); - return NULL; - } - - region->buffer = intel_bo_fake_alloc_static(intel->bufmgr, + assert(region_desc->bo_handle != -1); + region->buffer = intel_bo_gem_create_from_name(intel->bufmgr, name, - region_desc->offset, - region->pitch * region->cpp * - region->height, - region->classic_map); - - /* The sarea just gives us a boolean for whether it's tiled or not, - * instead of which tiling mode it is. Guess. - */ - if (region_desc->tiled) { - if (intel->gen >= 4 && region_desc == &intelScreen->depth) - region->tiling = I915_TILING_Y; - else - region->tiling = I915_TILING_X; - } else { - region->tiling = I915_TILING_NONE; - } + region_desc->bo_handle); - region->bit_6_swizzle = I915_BIT_6_SWIZZLE_NONE; + ret = dri_bo_get_tiling(region->buffer, ®ion->tiling, + ®ion->bit_6_swizzle); + if (ret != 0) { + fprintf(stderr, "Couldn't get tiling of buffer %d (%s): %s\n", + region_desc->bo_handle, name, strerror(-ret)); + intel_region_release(®ion); + return NULL; } assert(region->buffer != NULL); diff --git a/src/mesa/drivers/dri/intel/intel_screen.c b/src/mesa/drivers/dri/intel/intel_screen.c index 789135b49f9..2c5a884a9b4 100644 --- a/src/mesa/drivers/dri/intel/intel_screen.c +++ b/src/mesa/drivers/dri/intel/intel_screen.c @@ -605,7 +605,6 @@ intelFillInModes(__DRIscreenPrivate *psp, static GLboolean intel_init_bufmgr(intelScreenPrivate *intelScreen) { - GLboolean gem_disable = getenv("INTEL_NO_GEM") != NULL; int gem_kernel = 0; GLboolean gem_supported; struct drm_i915_getparam gp; @@ -622,43 +621,24 @@ intel_init_bufmgr(intelScreenPrivate *intelScreen) /* If we've got a new enough DDX that's initializing GEM and giving us * object handles for the shared buffers, use that. */ - intelScreen->ttm = GL_FALSE; if (intelScreen->driScrnPriv->dri2.enabled) gem_supported = GL_TRUE; else if (intelScreen->driScrnPriv->ddx_version.minor >= 9 && gem_kernel && intelScreen->front.bo_handle != -1) gem_supported = GL_TRUE; - else - gem_supported = GL_FALSE; - - if (!gem_disable && gem_supported) { - intelScreen->bufmgr = intel_bufmgr_gem_init(spriv->fd, BATCH_SZ); - if (intelScreen->bufmgr != NULL) - intelScreen->ttm = GL_TRUE; + else { + fprintf(stderr, "[%s:%u] Error initializing GEM.\n", + __func__, __LINE__); + return GL_FALSE; } + + intelScreen->bufmgr = intel_bufmgr_gem_init(spriv->fd, BATCH_SZ); /* Otherwise, use the classic buffer manager. */ if (intelScreen->bufmgr == NULL) { - if (gem_disable) { - _mesa_warning(NULL, "GEM disabled. Using classic."); - } else { - _mesa_warning(NULL, - "Failed to initialize GEM. Falling back to classic."); - } - - if (intelScreen->tex.size == 0) { - fprintf(stderr, "[%s:%u] Error initializing buffer manager.\n", - __func__, __LINE__); - return GL_FALSE; - } - - intelScreen->bufmgr = - intel_bufmgr_fake_init(spriv->fd, - intelScreen->tex.offset, - intelScreen->tex.map, - intelScreen->tex.size, - (unsigned int * volatile) - &intelScreen->sarea->last_dispatch); + fprintf(stderr, "[%s:%u] Error initializing buffer manager.\n", + __func__, __LINE__); + return GL_FALSE; } if (intel_get_param(spriv, I915_PARAM_NUM_FENCES_AVAIL, &num_fences)) diff --git a/src/mesa/drivers/dri/intel/intel_screen.h b/src/mesa/drivers/dri/intel/intel_screen.h index a9b9e109a6a..14ca0903b6c 100644 --- a/src/mesa/drivers/dri/intel/intel_screen.h +++ b/src/mesa/drivers/dri/intel/intel_screen.h @@ -77,7 +77,6 @@ typedef struct GLboolean no_hw; GLboolean no_vbo; - int ttm; dri_bufmgr *bufmgr; GLboolean kernel_exec_fencing; diff --git a/src/mesa/drivers/dri/intel/intel_span.c b/src/mesa/drivers/dri/intel/intel_span.c index 3607c7ddedd..2c89a66a95f 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 4ee742377d6..aefae53eb20 100644 --- a/src/mesa/drivers/dri/intel/intel_state.c +++ b/src/mesa/drivers/dri/intel/intel_state.c @@ -196,25 +196,6 @@ intel_translate_logic_op(GLenum opcode) } } - -static void -intelClearColor(GLcontext *ctx, const GLfloat color[4]) -{ - struct intel_context *intel = intel_context(ctx); - GLubyte clear[4]; - - CLAMPED_FLOAT_TO_UBYTE(clear[0], color[0]); - CLAMPED_FLOAT_TO_UBYTE(clear[1], color[1]); - CLAMPED_FLOAT_TO_UBYTE(clear[2], color[2]); - CLAMPED_FLOAT_TO_UBYTE(clear[3], color[3]); - - /* compute both 32 and 16-bit clear values */ - intel->ClearColor8888 = INTEL_PACKCOLOR8888(clear[0], clear[1], - clear[2], clear[3]); - intel->ClearColor565 = INTEL_PACKCOLOR565(clear[0], clear[1], clear[2]); -} - - /* Fallback to swrast for select and feedback. */ static void @@ -229,5 +210,4 @@ void intelInitStateFuncs(struct dd_function_table *functions) { functions->RenderMode = intelRenderMode; - functions->ClearColor = intelClearColor; } diff --git a/src/mesa/drivers/dri/mach64/mach64_native_vb.c b/src/mesa/drivers/dri/mach64/mach64_native_vb.c index 99f1a14e171..816682ec5f1 100644 --- a/src/mesa/drivers/dri/mach64/mach64_native_vb.c +++ b/src/mesa/drivers/dri/mach64/mach64_native_vb.c @@ -207,19 +207,19 @@ INTERP_QUALIFIER void TAG(interp_extras)( GLcontext *ctx, LOCALVARS struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; - if (VB->ColorPtr[1]) { - assert(VB->ColorPtr[1]->stride == 4 * sizeof(GLfloat)); + if (VB->BackfaceColorPtr) { + assert(VB->BackfaceColorPtr->stride == 4 * sizeof(GLfloat)); INTERP_4F( t, - GET_COLOR(VB->ColorPtr[1], dst), - GET_COLOR(VB->ColorPtr[1], out), - GET_COLOR(VB->ColorPtr[1], in) ); + GET_COLOR(VB->BackfaceColorPtr, dst), + GET_COLOR(VB->BackfaceColorPtr, out), + GET_COLOR(VB->BackfaceColorPtr, in) ); - if (VB->SecondaryColorPtr[1]) { + if (VB->BackfaceSecondaryColorPtr) { INTERP_3F( t, - GET_COLOR(VB->SecondaryColorPtr[1], dst), - GET_COLOR(VB->SecondaryColorPtr[1], out), - GET_COLOR(VB->SecondaryColorPtr[1], in) ); + GET_COLOR(VB->BackfaceSecondaryColorPtr, dst), + GET_COLOR(VB->BackfaceSecondaryColorPtr, out), + GET_COLOR(VB->BackfaceSecondaryColorPtr, in) ); } } @@ -236,13 +236,13 @@ INTERP_QUALIFIER void TAG(copy_pv_extras)( GLcontext *ctx, LOCALVARS struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; - if (VB->ColorPtr[1]) { - COPY_4FV( GET_COLOR(VB->ColorPtr[1], dst), - GET_COLOR(VB->ColorPtr[1], src) ); + if (VB->BackfaceColorPtr) { + COPY_4FV( GET_COLOR(VB->BackfaceColorPtr, dst), + GET_COLOR(VB->BackfaceColorPtr, src) ); - if (VB->SecondaryColorPtr[1]) { - COPY_4FV( GET_COLOR(VB->SecondaryColorPtr[1], dst), - GET_COLOR(VB->SecondaryColorPtr[1], src) ); + if (VB->BackfaceSecondaryColorPtr) { + COPY_4FV( GET_COLOR(VB->BackfaceSecondaryColorPtr, dst), + GET_COLOR(VB->BackfaceSecondaryColorPtr, src) ); } } diff --git a/src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h b/src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h index 684f2acc890..6e5fa3520e1 100644 --- a/src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h +++ b/src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h @@ -103,10 +103,10 @@ static void TAG(emit)( GLcontext *ctx, #if DO_TEX1 { const GLuint t1 = GET_TEXSOURCE(1); - tc1 = VB->TexCoordPtr[t1]->data; - tc1_stride = VB->TexCoordPtr[t1]->stride; + tc1 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->data; + tc1_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->stride; #if DO_PTEX - tc1_size = VB->TexCoordPtr[t1]->size; + tc1_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->size; #endif } #endif @@ -114,18 +114,18 @@ static void TAG(emit)( GLcontext *ctx, #if DO_TEX0 { const GLuint t0 = GET_TEXSOURCE(0); - tc0 = VB->TexCoordPtr[t0]->data; - tc0_stride = VB->TexCoordPtr[t0]->stride; + tc0 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->data; + tc0_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->stride; #if DO_PTEX - tc0_size = VB->TexCoordPtr[t0]->size; + tc0_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->size; #endif } #endif #if DO_SPEC - if (VB->SecondaryColorPtr[0]) { - spec = VB->SecondaryColorPtr[0]->data; - spec_stride = VB->SecondaryColorPtr[0]->stride; + if (VB->AttribPtr[_TNL_ATTRIB_COLOR1]) { + spec = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->data; + spec_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->stride; } else { spec = (GLfloat (*)[4])ctx->Current.Attrib[VERT_ATTRIB_COLOR1]; spec_stride = 0; @@ -133,9 +133,9 @@ static void TAG(emit)( GLcontext *ctx, #endif #if DO_FOG - if (VB->FogCoordPtr) { - fog = VB->FogCoordPtr->data; - fog_stride = VB->FogCoordPtr->stride; + if (VB->AttribPtr[_TNL_ATTRIB_FOG]) { + fog = VB->AttribPtr[_TNL_ATTRIB_FOG]->data; + fog_stride = VB->AttribPtr[_TNL_ATTRIB_FOG]->stride; } else { static GLfloat tmp[4] = {0, 0, 0, 0}; fog = &tmp; @@ -144,8 +144,8 @@ static void TAG(emit)( GLcontext *ctx, #endif #if DO_RGBA - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; #endif coord = VB->NdcPtr->data; @@ -319,8 +319,8 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) /* Force 'missing' texcoords to something valid. */ - if (DO_TEX1 && VB->TexCoordPtr[0] == 0) - VB->TexCoordPtr[0] = VB->TexCoordPtr[1]; + if (DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0] = VB->AttribPtr[_TNL_ATTRIB_TEX1]; if (DO_PTEX) return GL_TRUE; @@ -328,12 +328,12 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) /* No hardware support for projective texture. Can fake it for * TEX0 only. */ - if ((DO_TEX1 && VB->TexCoordPtr[GET_TEXSOURCE(1)]->size == 4)) { + if ((DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(1)]->size == 4)) { PTEX_FALLBACK(); return GL_FALSE; } - if (DO_TEX0 && VB->TexCoordPtr[GET_TEXSOURCE(0)]->size == 4) { + if (DO_TEX0 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(0)]->size == 4) { if (DO_TEX1) { PTEX_FALLBACK(); } diff --git a/src/mesa/drivers/dri/mach64/mach64_vbtmp.h b/src/mesa/drivers/dri/mach64/mach64_vbtmp.h index 938804af9ec..60bfab8f6dc 100644 --- a/src/mesa/drivers/dri/mach64/mach64_vbtmp.h +++ b/src/mesa/drivers/dri/mach64/mach64_vbtmp.h @@ -156,53 +156,53 @@ static void TAG(emit)( GLcontext *ctx, if (DO_TEX3) { const GLuint t3 = GET_TEXSOURCE(3); - tc3 = VB->TexCoordPtr[t3]->data; - tc3_stride = VB->TexCoordPtr[t3]->stride; + tc3 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t3]->data; + tc3_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t3]->stride; if (DO_PTEX) - tc3_size = VB->TexCoordPtr[t3]->size; + tc3_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t3]->size; } if (DO_TEX2) { const GLuint t2 = GET_TEXSOURCE(2); - tc2 = VB->TexCoordPtr[t2]->data; - tc2_stride = VB->TexCoordPtr[t2]->stride; + tc2 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->data; + tc2_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->stride; if (DO_PTEX) - tc2_size = VB->TexCoordPtr[t2]->size; + tc2_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->size; } if (DO_TEX1) { const GLuint t1 = GET_TEXSOURCE(1); - tc1 = VB->TexCoordPtr[t1]->data; - tc1_stride = VB->TexCoordPtr[t1]->stride; + tc1 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->data; + tc1_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->stride; if (DO_PTEX) - tc1_size = VB->TexCoordPtr[t1]->size; + tc1_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->size; } if (DO_TEX0) { const GLuint t0 = GET_TEXSOURCE(0); - tc0_stride = VB->TexCoordPtr[t0]->stride; - tc0 = VB->TexCoordPtr[t0]->data; + tc0_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->stride; + tc0 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->data; if (DO_PTEX) - tc0_size = VB->TexCoordPtr[t0]->size; + tc0_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->size; } if (DO_RGBA) { - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; } if (DO_SPEC) { - spec = VB->SecondaryColorPtr[0]->data; - spec_stride = VB->SecondaryColorPtr[0]->stride; + spec = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->data; + spec_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->stride; } else { spec = (GLfloat (*)[4])ctx->Current.Attrib[VERT_ATTRIB_COLOR1]; spec_stride = 0; } if (DO_FOG) { - if (VB->FogCoordPtr) { - fog = VB->FogCoordPtr->data; - fog_stride = VB->FogCoordPtr->stride; + if (VB->AttribPtr[_TNL_ATTRIB_FOG]) { + fog = VB->AttribPtr[_TNL_ATTRIB_FOG]->data; + fog_stride = VB->AttribPtr[_TNL_ATTRIB_FOG]->stride; } else { static GLfloat tmp[4] = {0, 0, 0, 0}; fog = &tmp; @@ -384,8 +384,8 @@ static void TAG(emit)( GLcontext *ctx, GLuint start, GLuint end, ASSERT(stride == 4); - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; /* Pack what's left into a 4-dword vertex. Color is in a different * place, and there is no 'w' coordinate. @@ -432,8 +432,8 @@ static void TAG(emit)( GLcontext *ctx, GLuint start, GLuint end, GLfloat *v = (GLfloat *)dest; int i; - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; if (start) STRIDE_4F(col, col_stride * start); @@ -473,22 +473,22 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) /* Force 'missing' texcoords to something valid. */ - if (DO_TEX3 && VB->TexCoordPtr[2] == 0) - VB->TexCoordPtr[2] = VB->TexCoordPtr[3]; + if (DO_TEX3 && VB->AttribPtr[_TNL_ATTRIB_TEX2] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX2] = VB->AttribPtr[_TNL_ATTRIB_TEX3]; - if (DO_TEX2 && VB->TexCoordPtr[1] == 0) - VB->TexCoordPtr[1] = VB->TexCoordPtr[2]; + if (DO_TEX2 && VB->AttribPtr[_TNL_ATTRIB_TEX1] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX1] = VB->AttribPtr[_TNL_ATTRIB_TEX2]; - if (DO_TEX1 && VB->TexCoordPtr[0] == 0) - VB->TexCoordPtr[0] = VB->TexCoordPtr[1]; + if (DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0] = VB->AttribPtr[_TNL_ATTRIB_TEX1]; if (DO_PTEX) return GL_TRUE; - if ((DO_TEX3 && VB->TexCoordPtr[GET_TEXSOURCE(3)]->size == 4) || - (DO_TEX2 && VB->TexCoordPtr[GET_TEXSOURCE(2)]->size == 4) || - (DO_TEX1 && VB->TexCoordPtr[GET_TEXSOURCE(1)]->size == 4) || - (DO_TEX0 && VB->TexCoordPtr[GET_TEXSOURCE(0)]->size == 4)) + if ((DO_TEX3 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(3)]->size == 4) || + (DO_TEX2 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(2)]->size == 4) || + (DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(1)]->size == 4) || + (DO_TEX0 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(0)]->size == 4)) return GL_FALSE; return GL_TRUE; @@ -501,14 +501,14 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) /* Force 'missing' texcoords to something valid. */ - if (DO_TEX3 && VB->TexCoordPtr[2] == 0) - VB->TexCoordPtr[2] = VB->TexCoordPtr[3]; + if (DO_TEX3 && VB->AttribPtr[_TNL_ATTRIB_TEX2] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX2] = VB->AttribPtr[_TNL_ATTRIB_TEX3]; - if (DO_TEX2 && VB->TexCoordPtr[1] == 0) - VB->TexCoordPtr[1] = VB->TexCoordPtr[2]; + if (DO_TEX2 && VB->AttribPtr[_TNL_ATTRIB_TEX1] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX1] = VB->AttribPtr[_TNL_ATTRIB_TEX2]; - if (DO_TEX1 && VB->TexCoordPtr[0] == 0) - VB->TexCoordPtr[0] = VB->TexCoordPtr[1]; + if (DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0] = VB->AttribPtr[_TNL_ATTRIB_TEX1]; if (DO_PTEX) return GL_TRUE; @@ -516,14 +516,14 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) /* No hardware support for projective texture. Can fake it for * TEX0 only. */ - if ((DO_TEX3 && VB->TexCoordPtr[GET_TEXSOURCE(3)]->size == 4) || - (DO_TEX2 && VB->TexCoordPtr[GET_TEXSOURCE(2)]->size == 4) || - (DO_TEX1 && VB->TexCoordPtr[GET_TEXSOURCE(1)]->size == 4)) { + if ((DO_TEX3 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(3)]->size == 4) || + (DO_TEX2 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(2)]->size == 4) || + (DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(1)]->size == 4)) { PTEX_FALLBACK(); return GL_FALSE; } - if (DO_TEX0 && VB->TexCoordPtr[GET_TEXSOURCE(0)]->size == 4) { + if (DO_TEX0 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(0)]->size == 4) { if (DO_TEX1 || DO_TEX2 || DO_TEX3) { PTEX_FALLBACK(); } diff --git a/src/mesa/drivers/dri/r128/r128_tris.c b/src/mesa/drivers/dri/r128/r128_tris.c index 5b91271d740..448e34e047c 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_maos_arrays.c b/src/mesa/drivers/dri/r200/r200_maos_arrays.c index 383a0c4b0d3..249c0bbc11d 100644 --- a/src/mesa/drivers/dri/r200/r200_maos_arrays.c +++ b/src/mesa/drivers/dri/r200/r200_maos_arrays.c @@ -90,12 +90,14 @@ static void r200_emit_vecfog(GLcontext *ctx, struct radeon_aos *aos, aos->components = size; aos->count = count; + radeon_bo_map(aos->bo, 1); out = (uint32_t*)((char*)aos->bo->ptr + aos->offset); for (i = 0; i < count; i++) { out[0] = r200ComputeFogBlendFactor( ctx, *(GLfloat *)data ); out++; data += stride; } + radeon_bo_unmap(aos->bo); } /* Emit any changed arrays to new GART memory, re-emit a packet to diff --git a/src/mesa/drivers/dri/r200/r200_swtcl.c b/src/mesa/drivers/dri/r200/r200_swtcl.c index 240fb450780..4596912ddc7 100644 --- a/src/mesa/drivers/dri/r200/r200_swtcl.c +++ b/src/mesa/drivers/dri/r200/r200_swtcl.c @@ -168,7 +168,7 @@ static void r200SetVertexFormat( GLcontext *ctx ) for (i = 0; i < ctx->Const.MaxTextureUnits; i++) { if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX(i) )) { - GLuint sz = VB->TexCoordPtr[i]->size; + GLuint sz = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->size; fmt_1 |= sz << (3 * i); EMIT_ATTR( _TNL_ATTRIB_TEX0+i, EMIT_1F + sz - 1, 0 ); @@ -297,7 +297,7 @@ void r200_swtcl_flush(GLcontext *ctx, uint32_t current_offset) radeonEmitState(&rmesa->radeon); r200EmitVertexAOS( rmesa, rmesa->radeon.swtcl.vertex_size, - first_elem(&rmesa->radeon.dma.reserved)->bo, + rmesa->radeon.swtcl.bo, current_offset); diff --git a/src/mesa/drivers/dri/r300/r300_draw.c b/src/mesa/drivers/dri/r300/r300_draw.c index e9968f9ffef..3dcd986e226 100644 --- a/src/mesa/drivers/dri/r300/r300_draw.c +++ b/src/mesa/drivers/dri/r300/r300_draw.c @@ -100,7 +100,7 @@ static void r300FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer GLubyte *in = (GLubyte *)src_ptr; radeonAllocDmaRegion(&r300->radeon, &r300->ind_buf.bo, &r300->ind_buf.bo_offset, size, 4); - + radeon_bo_map(r300->ind_buf.bo, 1); assert(r300->ind_buf.bo->ptr != NULL); out = (GLuint *)ADD_POINTERS(r300->ind_buf.bo->ptr, r300->ind_buf.bo_offset); @@ -111,7 +111,7 @@ static void r300FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer if (i < mesa_ind_buf->count) { *out++ = in[i]; } - + radeon_bo_unmap(r300->ind_buf.bo); #if MESA_BIG_ENDIAN } else { /* if (mesa_ind_buf->type == GL_UNSIGNED_SHORT) */ GLushort *in = (GLushort *)src_ptr; @@ -120,6 +120,7 @@ static void r300FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer radeonAllocDmaRegion(&r300->radeon, &r300->ind_buf.bo, &r300->ind_buf.bo_offset, size, 4); + radeon_bo_map(r300->ind_buf.bo, 1); assert(r300->ind_buf.bo->ptr != NULL); out = (GLuint *)ADD_POINTERS(r300->ind_buf.bo->ptr, r300->ind_buf.bo_offset); @@ -130,6 +131,7 @@ static void r300FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer if (i < mesa_ind_buf->count) { *out++ = in[i]; } + radeon_bo_unmap(r300->ind_buf.bo); #endif } @@ -173,10 +175,12 @@ static void r300SetupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer radeonAllocDmaRegion(&r300->radeon, &r300->ind_buf.bo, &r300->ind_buf.bo_offset, size, 4); + radeon_bo_map(r300->ind_buf.bo, 1); assert(r300->ind_buf.bo->ptr != NULL); dst_ptr = ADD_POINTERS(r300->ind_buf.bo->ptr, r300->ind_buf.bo_offset); _mesa_memcpy(dst_ptr, src_ptr, size); + radeon_bo_unmap(r300->ind_buf.bo); r300->ind_buf.is_32bit = (mesa_ind_buf->type == GL_UNSIGNED_INT); r300->ind_buf.count = mesa_ind_buf->count; @@ -242,6 +246,7 @@ static void r300ConvertAttrib(GLcontext *ctx, int count, const struct gl_client_ } radeonAllocDmaRegion(&r300->radeon, &attr->bo, &attr->bo_offset, sizeof(GLfloat) * input->Size * count, 32); + radeon_bo_map(attr->bo, 1); dst_ptr = (GLfloat *)ADD_POINTERS(attr->bo->ptr, attr->bo_offset); radeon_print(RADEON_FALLBACKS, RADEON_IMPORTANT, @@ -280,6 +285,7 @@ static void r300ConvertAttrib(GLcontext *ctx, int count, const struct gl_client_ break; } + radeon_bo_unmap(attr->bo); if (mapped_named_bo) { ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER, input->BufferObj); } @@ -294,6 +300,8 @@ static void r300AlignDataToDword(GLcontext *ctx, const struct gl_client_array *i radeonAllocDmaRegion(&r300->radeon, &attr->bo, &attr->bo_offset, size, 32); + radeon_bo_map(attr->bo, 1); + if (!input->BufferObj->Pointer) { ctx->Driver.MapBuffer(ctx, GL_ARRAY_BUFFER, GL_READ_ONLY_ARB, input->BufferObj); mapped_named_bo = GL_TRUE; @@ -317,6 +325,7 @@ static void r300AlignDataToDword(GLcontext *ctx, const struct gl_client_array *i ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER, input->BufferObj); } + radeon_bo_unmap(attr->bo); attr->stride = dst_stride; } @@ -527,6 +536,7 @@ static void r300AllocDmaRegions(GLcontext *ctx, const struct gl_client_array *in } radeonAllocDmaRegion(&r300->radeon, &vbuf->attribs[index].bo, &vbuf->attribs[index].bo_offset, size, 32); + radeon_bo_map(vbuf->attribs[index].bo, 1); assert(vbuf->attribs[index].bo->ptr != NULL); dst = (uint32_t *)ADD_POINTERS(vbuf->attribs[index].bo->ptr, vbuf->attribs[index].bo_offset); switch (vbuf->attribs[index].dwords) { @@ -536,6 +546,7 @@ static void r300AllocDmaRegions(GLcontext *ctx, const struct gl_client_array *in case 4: radeonEmitVec16(dst, input[i]->Ptr, input[i]->StrideB, local_count); break; default: assert(0); break; } + radeon_bo_unmap(vbuf->attribs[index].bo); } } diff --git a/src/mesa/drivers/dri/r300/r300_swtcl.c b/src/mesa/drivers/dri/r300/r300_swtcl.c index ee2c71e1a7f..383c8a274b5 100644 --- a/src/mesa/drivers/dri/r300/r300_swtcl.c +++ b/src/mesa/drivers/dri/r300/r300_swtcl.c @@ -124,7 +124,7 @@ void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead, GLuint *_ } if (ctx->Light.Enabled && ctx->Light.Model.TwoSide) { - VB->AttribPtr[VERT_ATTRIB_GENERIC0] = VB->ColorPtr[1]; + VB->AttribPtr[VERT_ATTRIB_GENERIC0] = VB->BackfaceColorPtr; OutputsWritten |= 1 << VERT_RESULT_BFC0; #if MESA_LITTLE_ENDIAN EMIT_ATTR( _TNL_ATTRIB_GENERIC0, EMIT_4UB_4F_RGBA ); @@ -134,7 +134,7 @@ void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead, GLuint *_ ADD_ATTR(VERT_ATTRIB_GENERIC0, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR2, SWIZZLE_XYZW, MASK_XYZW, 1); #endif if (fp_reads & FRAG_BIT_COL1) { - VB->AttribPtr[VERT_ATTRIB_GENERIC1] = VB->SecondaryColorPtr[1]; + VB->AttribPtr[VERT_ATTRIB_GENERIC1] = VB->BackfaceSecondaryColorPtr; GLuint swiz = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ONE); OutputsWritten |= 1 << VERT_RESULT_BFC1; #if MESA_LITTLE_ENDIAN @@ -159,7 +159,7 @@ void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead, GLuint *_ int tex_id = rmesa->selected_fp->wpos_attr - FRAG_ATTRIB_TEX0; VB->AttribPtr[VERT_ATTRIB_TEX0 + tex_id] = VB->AttribPtr[VERT_ATTRIB_POS]; - VB->TexCoordPtr[tex_id] = VB->AttribPtr[VERT_ATTRIB_POS]; + VB->AttribPtr[_TNL_ATTRIB_TEX0 + tex_id] = VB->AttribPtr[VERT_ATTRIB_POS]; RENDERINPUTS_SET(tnl->render_inputs_bitset, _TNL_ATTRIB_TEX0 + tex_id); } @@ -167,7 +167,7 @@ void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead, GLuint *_ int tex_id = rmesa->selected_fp->fog_attr - FRAG_ATTRIB_TEX0; VB->AttribPtr[VERT_ATTRIB_TEX0 + tex_id] = VB->AttribPtr[VERT_ATTRIB_FOG]; - VB->TexCoordPtr[tex_id] = VB->AttribPtr[VERT_ATTRIB_FOG]; + VB->AttribPtr[_TNL_ATTRIB_TEX0 + tex_id] = VB->AttribPtr[VERT_ATTRIB_FOG]; RENDERINPUTS_SET(tnl->render_inputs_bitset, _TNL_ATTRIB_TEX0 + tex_id); } @@ -180,7 +180,7 @@ void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead, GLuint *_ GLuint swiz, format, hw_format; for (i = 0; i < ctx->Const.MaxTextureUnits; i++) { if (fp_reads & FRAG_BIT_TEX(i)) { - switch (VB->TexCoordPtr[i]->size) { + switch (VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->size) { case 1: format = EMIT_1F; hw_format = R300_DATA_TYPE_FLOAT_1; @@ -665,11 +665,11 @@ void r300_swtcl_flush(GLcontext *ctx, uint32_t current_offset) r300EmitCacheFlush(rmesa); radeonEmitState(&rmesa->radeon); - r300_emit_scissor(ctx); + r300_emit_scissor(ctx); r300EmitVertexAOS(rmesa, - rmesa->radeon.swtcl.vertex_size, - first_elem(&rmesa->radeon.dma.reserved)->bo, - current_offset); + rmesa->radeon.swtcl.vertex_size, + rmesa->radeon.swtcl.bo, + current_offset); r300EmitVbufPrim(rmesa, rmesa->radeon.swtcl.hw_primitive, diff --git a/src/mesa/drivers/dri/r600/r600_context.c b/src/mesa/drivers/dri/r600/r600_context.c index dbd233729c0..25314eff563 100644 --- a/src/mesa/drivers/dri/r600/r600_context.c +++ b/src/mesa/drivers/dri/r600/r600_context.c @@ -74,6 +74,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "utils.h" #include "xmlpool.h" /* for symbolic values of enum-type options */ +//#define R600_ENABLE_GLSL_TEST 1 + #define need_GL_VERSION_2_0 #define need_GL_ARB_occlusion_query #define need_GL_ARB_point_parameters @@ -109,6 +111,7 @@ static const struct dri_extension card_extensions[] = { {"GL_ARB_texture_env_crossbar", NULL}, {"GL_ARB_texture_env_dot3", NULL}, {"GL_ARB_texture_mirrored_repeat", NULL}, + {"GL_ARB_texture_non_power_of_two", NULL}, {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions}, {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions}, {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions}, @@ -155,7 +158,11 @@ static const struct dri_extension mm_extensions[] = { * functions added by GL_ATI_separate_stencil. */ static const struct dri_extension gl_20_extension[] = { +#ifdef R600_ENABLE_GLSL_TEST + {"GL_ARB_shading_language_100", GL_VERSION_2_0_functions }, +#else {"GL_VERSION_2_0", GL_VERSION_2_0_functions }, +#endif /* R600_ENABLE_GLSL_TEST */ }; static const struct tnl_pipeline_stage *r600_pipeline[] = { @@ -308,6 +315,26 @@ static void r600InitGLExtensions(GLcontext *ctx) if (r600->radeon.radeonScreen->kernel_mm) driInitExtensions(ctx, mm_extensions, GL_FALSE); +#ifdef R600_ENABLE_GLSL_TEST + driInitExtensions(ctx, gl_20_extension, GL_TRUE); + //_mesa_enable_2_0_extensions(ctx); + //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.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")) _mesa_disable_extension(ctx, "GL_EXT_stencil_two_side"); diff --git a/src/mesa/drivers/dri/r600/r700_assembler.c b/src/mesa/drivers/dri/r600/r700_assembler.c index e0d7d4fa6b7..309c90fdd0c 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); @@ -337,12 +340,17 @@ unsigned int r700GetNumOperands(r700_AssemblerBase* pAsm) switch (pAsm->D.dst.opcode) { case SQ_OP2_INST_ADD: + case SQ_OP2_INST_KILLE: case SQ_OP2_INST_KILLGT: + case SQ_OP2_INST_KILLGE: + case SQ_OP2_INST_KILLNE: case SQ_OP2_INST_MUL: case SQ_OP2_INST_MAX: case SQ_OP2_INST_MIN: //case SQ_OP2_INST_MAX_DX10: //case SQ_OP2_INST_MIN_DX10: + case SQ_OP2_INST_SETE: + case SQ_OP2_INST_SETNE: case SQ_OP2_INST_SETGT: case SQ_OP2_INST_SETGE: case SQ_OP2_INST_PRED_SETE: @@ -358,6 +366,7 @@ unsigned int r700GetNumOperands(r700_AssemblerBase* pAsm) case SQ_OP2_INST_MOVA_FLOOR: case SQ_OP2_INST_FRACT: case SQ_OP2_INST_FLOOR: + case SQ_OP2_INST_TRUNC: case SQ_OP2_INST_EXP_IEEE: case SQ_OP2_INST_LOG_CLAMPED: case SQ_OP2_INST_LOG_IEEE: @@ -383,98 +392,115 @@ int Init_r700_AssemblerBase(SHADER_PIPE_TYPE spt, r700_AssemblerBase* pAsm, R700 pAsm->pR700Shader = pShader; pAsm->currentShaderType = spt; - pAsm->cf_last_export_ptr = NULL; + pAsm->cf_last_export_ptr = NULL; + + pAsm->cf_current_export_clause_ptr = NULL; + pAsm->cf_current_alu_clause_ptr = NULL; + pAsm->cf_current_tex_clause_ptr = NULL; + pAsm->cf_current_vtx_clause_ptr = NULL; + pAsm->cf_current_cf_clause_ptr = NULL; - pAsm->cf_current_export_clause_ptr = NULL; - pAsm->cf_current_alu_clause_ptr = NULL; - pAsm->cf_current_tex_clause_ptr = NULL; - pAsm->cf_current_vtx_clause_ptr = NULL; - pAsm->cf_current_cf_clause_ptr = NULL; + // No clause has been created yet + pAsm->cf_current_clause_type = CF_EMPTY_CLAUSE; - // No clause has been created yet - pAsm->cf_current_clause_type = CF_EMPTY_CLAUSE; + pAsm->number_of_colorandz_exports = 0; + pAsm->number_of_exports = 0; + pAsm->number_of_export_opcodes = 0; - pAsm->number_of_colorandz_exports = 0; - pAsm->number_of_exports = 0; - pAsm->number_of_export_opcodes = 0; + pAsm->alu_x_opcode = 0; + pAsm->D2.bits = 0; - pAsm->D.bits = 0; - pAsm->S[0].bits = 0; - pAsm->S[1].bits = 0; - pAsm->S[2].bits = 0; + pAsm->D.bits = 0; + pAsm->S[0].bits = 0; + pAsm->S[1].bits = 0; + pAsm->S[2].bits = 0; - pAsm->uLastPosUpdate = 0; + pAsm->uLastPosUpdate = 0; - *(BITS *) &pAsm->fp_stOutFmt0 = 0; + *(BITS *) &pAsm->fp_stOutFmt0 = 0; - pAsm->uIIns = 0; - pAsm->uOIns = 0; - pAsm->number_used_registers = 0; - pAsm->uUsedConsts = 256; + pAsm->uIIns = 0; + pAsm->uOIns = 0; + pAsm->number_used_registers = 0; + pAsm->uUsedConsts = 256; - // Fragment programs - pAsm->uBoolConsts = 0; - pAsm->uIntConsts = 0; - pAsm->uInsts = 0; - pAsm->uConsts = 0; + // Fragment programs + pAsm->uBoolConsts = 0; + pAsm->uIntConsts = 0; + pAsm->uInsts = 0; + pAsm->uConsts = 0; - pAsm->FCSP = 0; - pAsm->fc_stack[0].type = FC_NONE; + pAsm->FCSP = 0; + pAsm->fc_stack[0].type = FC_NONE; - pAsm->branch_depth = 0; - pAsm->max_branch_depth = 0; + pAsm->aArgSubst[0] = + pAsm->aArgSubst[1] = + pAsm->aArgSubst[2] = + pAsm->aArgSubst[3] = (-1); - pAsm->aArgSubst[0] = - pAsm->aArgSubst[1] = - pAsm->aArgSubst[2] = - pAsm->aArgSubst[3] = (-1); + pAsm->uOutputs = 0; - pAsm->uOutputs = 0; + for (i=0; i<NUMBER_OF_OUTPUT_COLORS; i++) + { + pAsm->color_export_register_number[i] = (-1); + } - for (i=0; i<NUMBER_OF_OUTPUT_COLORS; i++) - { - pAsm->color_export_register_number[i] = (-1); - } + pAsm->depth_export_register_number = (-1); + pAsm->stencil_export_register_number = (-1); + pAsm->coverage_to_mask_export_register_number = (-1); + pAsm->mask_export_register_number = (-1); - pAsm->depth_export_register_number = (-1); - pAsm->stencil_export_register_number = (-1); - pAsm->coverage_to_mask_export_register_number = (-1); - pAsm->mask_export_register_number = (-1); + pAsm->starting_export_register_number = 0; + pAsm->starting_vfetch_register_number = 0; + pAsm->starting_temp_register_number = 0; + pAsm->uFirstHelpReg = 0; - pAsm->starting_export_register_number = 0; - pAsm->starting_vfetch_register_number = 0; - pAsm->starting_temp_register_number = 0; - pAsm->uFirstHelpReg = 0; + pAsm->input_position_is_used = GL_FALSE; + pAsm->input_normal_is_used = GL_FALSE; + for (i=0; i<NUMBER_OF_INPUT_COLORS; i++) + { + pAsm->input_color_is_used[ i ] = GL_FALSE; + } - pAsm->input_position_is_used = GL_FALSE; - pAsm->input_normal_is_used = GL_FALSE; + for (i=0; i<NUMBER_OF_TEXTURE_UNITS; i++) + { + pAsm->input_texture_unit_is_used[ i ] = GL_FALSE; + } + for (i=0; i<VERT_ATTRIB_MAX; i++) + { + pAsm->vfetch_instruction_ptr_array[ i ] = NULL; + } - for (i=0; i<NUMBER_OF_INPUT_COLORS; i++) - { - pAsm->input_color_is_used[ i ] = GL_FALSE; - } + pAsm->number_of_inputs = 0; - for (i=0; i<NUMBER_OF_TEXTURE_UNITS; i++) - { - pAsm->input_texture_unit_is_used[ i ] = GL_FALSE; - } + pAsm->is_tex = GL_FALSE; + pAsm->need_tex_barrier = GL_FALSE; - for (i=0; i<VERT_ATTRIB_MAX; i++) - { - pAsm->vfetch_instruction_ptr_array[ i ] = NULL; - } + pAsm->subs = NULL; + pAsm->unSubArraySize = 0; + pAsm->unSubArrayPointer = 0; + pAsm->callers = NULL; + pAsm->unCallerArraySize = 0; + pAsm->unCallerArrayPointer = 0; + + pAsm->CALLSP = 0; + pAsm->CALLSTACK[0].FCSP_BeforeEntry = 0; + pAsm->CALLSTACK[0].plstCFInstructions_local + = &(pAsm->pR700Shader->lstCFInstructions); - pAsm->number_of_inputs = 0; + pAsm->CALLSTACK[0].max = 0; + pAsm->CALLSTACK[0].current = 0; - pAsm->is_tex = GL_FALSE; - pAsm->need_tex_barrier = GL_FALSE; + SetActiveCFlist(pAsm->pR700Shader, pAsm->CALLSTACK[0].plstCFInstructions_local); - return 0; + pAsm->unCFflags = 0; + + return 0; } GLboolean IsTex(gl_inst_opcode Opcode) @@ -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) { @@ -987,7 +1038,8 @@ GLboolean checkop2(r700_AssemblerBase* pAsm) checkop_init(pAsm); - if( (pILInst->SrcReg[0].File == PROGRAM_CONSTANT) || + if( (pILInst->SrcReg[0].File == PROGRAM_UNIFORM) || + (pILInst->SrcReg[0].File == PROGRAM_CONSTANT) || (pILInst->SrcReg[0].File == PROGRAM_LOCAL_PARAM) || (pILInst->SrcReg[0].File == PROGRAM_ENV_PARAM) || (pILInst->SrcReg[0].File == PROGRAM_STATE_VAR) ) @@ -998,7 +1050,8 @@ GLboolean checkop2(r700_AssemblerBase* pAsm) { bSrcConst[0] = GL_FALSE; } - if( (pILInst->SrcReg[1].File == PROGRAM_CONSTANT) || + if( (pILInst->SrcReg[1].File == PROGRAM_UNIFORM) || + (pILInst->SrcReg[1].File == PROGRAM_CONSTANT) || (pILInst->SrcReg[1].File == PROGRAM_LOCAL_PARAM) || (pILInst->SrcReg[1].File == PROGRAM_ENV_PARAM) || (pILInst->SrcReg[1].File == PROGRAM_STATE_VAR) ) @@ -1031,7 +1084,8 @@ GLboolean checkop3(r700_AssemblerBase* pAsm) checkop_init(pAsm); - if( (pILInst->SrcReg[0].File == PROGRAM_CONSTANT) || + if( (pILInst->SrcReg[0].File == PROGRAM_UNIFORM) || + (pILInst->SrcReg[0].File == PROGRAM_CONSTANT) || (pILInst->SrcReg[0].File == PROGRAM_LOCAL_PARAM) || (pILInst->SrcReg[0].File == PROGRAM_ENV_PARAM) || (pILInst->SrcReg[0].File == PROGRAM_STATE_VAR) ) @@ -1042,7 +1096,8 @@ GLboolean checkop3(r700_AssemblerBase* pAsm) { bSrcConst[0] = GL_FALSE; } - if( (pILInst->SrcReg[1].File == PROGRAM_CONSTANT) || + if( (pILInst->SrcReg[1].File == PROGRAM_UNIFORM) || + (pILInst->SrcReg[1].File == PROGRAM_CONSTANT) || (pILInst->SrcReg[1].File == PROGRAM_LOCAL_PARAM) || (pILInst->SrcReg[1].File == PROGRAM_ENV_PARAM) || (pILInst->SrcReg[1].File == PROGRAM_STATE_VAR) ) @@ -1053,7 +1108,8 @@ GLboolean checkop3(r700_AssemblerBase* pAsm) { bSrcConst[1] = GL_FALSE; } - if( (pILInst->SrcReg[2].File == PROGRAM_CONSTANT) || + if( (pILInst->SrcReg[2].File == PROGRAM_UNIFORM) || + (pILInst->SrcReg[2].File == PROGRAM_CONSTANT) || (pILInst->SrcReg[2].File == PROGRAM_LOCAL_PARAM) || (pILInst->SrcReg[2].File == PROGRAM_ENV_PARAM) || (pILInst->SrcReg[2].File == PROGRAM_STATE_VAR) ) @@ -1153,6 +1209,7 @@ 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); @@ -1166,7 +1223,7 @@ GLboolean assemble_src(r700_AssemblerBase *pAsm, pAsm->S[fld].src.reg = pILInst->SrcReg[src].Index; break; case PROGRAM_INPUT: - setaddrmode_PVSSRC(&(pAsm->S[fld].src), ADDR_ABSOLUTE); + setaddrmode_PVSSRC(&(pAsm->S[fld].src), ADDR_ABSOLUTE); pAsm->S[fld].src.rtype = SRC_REG_INPUT; switch (pAsm->currentShaderType) { @@ -1179,7 +1236,7 @@ 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; } } @@ -1294,6 +1351,7 @@ GLboolean tex_src(r700_AssemblerBase *pAsm) else { switch (pILInst->SrcReg[0].File) { + case PROGRAM_UNIFORM: case PROGRAM_CONSTANT: case PROGRAM_LOCAL_PARAM: case PROGRAM_ENV_PARAM: @@ -1315,7 +1373,7 @@ 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: @@ -1331,10 +1389,17 @@ GLboolean tex_src(r700_AssemblerBase *pAsm) case FRAG_ATTRIB_PNTC: fprintf(stderr, "FRAG_ATTRIB_PNTC unsupported\n"); break; - case FRAG_ATTRIB_VAR0: - 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 +1582,10 @@ 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 +1675,8 @@ 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 +1706,17 @@ 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; @@ -2045,7 +2123,7 @@ GLboolean check_vector(r700_AssemblerBase* pAsm, if( is_gpr(sel) ) { if( GL_FALSE == cycle_for_vector_bank_swizzle(bank_swizzle, src, &cycle) ) - { + { return GL_FALSE; } @@ -2057,7 +2135,7 @@ GLboolean check_vector(r700_AssemblerBase* pAsm, else { if( GL_FALSE == reserve_gpr(pAsm, sel, chan, cycle) ) - { + { return GL_FALSE; } } @@ -2069,7 +2147,7 @@ GLboolean check_vector(r700_AssemblerBase* pAsm, if( is_cfile(sel) ) { if( GL_FALSE == reserve_cfile(pAsm, sel, chan) ) - { + { return GL_FALSE; } } @@ -2172,7 +2250,7 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) current_source_index, pcurrent_source, scalar_channel_index) ) - { + { return GL_FALSE; } @@ -2186,7 +2264,7 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) current_source_index, pcurrent_source, scalar_channel_index) ) - { + { return GL_FALSE; } } @@ -2215,7 +2293,7 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) alu_instruction_ptr->m_Word1.f.dst_gpr = pAsm->D.dst.reg; } else - { + { radeon_error("Only temp destination registers supported for ALU dest regs.\n"); return GL_FALSE; } @@ -2329,6 +2407,253 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) } 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 GL_FALSE; + } + } + + contiguous_slots_needed = 0; + } + + 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; } @@ -2348,7 +2673,7 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) { if(GL_FALSE == check_vector(pAsm, alu_instruction_ptr) ) { - return 1; + return GL_FALSE; } } @@ -2358,6 +2683,259 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) 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 +2981,70 @@ 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; + + 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; @@ -2910,13 +3552,12 @@ GLboolean assemble_FRC(r700_AssemblerBase *pAsm) return GL_TRUE; } -GLboolean assemble_KIL(r700_AssemblerBase *pAsm) -{ - /* TODO: doc says KILL has to be last(end) ALU clause */ - - checkop1(pAsm); +GLboolean assemble_KIL(r700_AssemblerBase *pAsm, GLuint opcode) +{ + checkop2(pAsm); - pAsm->D.dst.opcode = SQ_OP2_INST_KILLGT; + pAsm->D.dst.opcode = opcode; + pAsm->D.dst.math = 1; setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); pAsm->D.dst.rtype = DST_REG_TEMPORARY; @@ -2926,24 +3567,24 @@ GLboolean assemble_KIL(r700_AssemblerBase *pAsm) pAsm->D.dst.writez = 0; pAsm->D.dst.writew = 0; - setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); - pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; - pAsm->S[0].src.reg = 0; - - setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_0); - noneg_PVSSRC(&(pAsm->S[0].src)); + if( GL_FALSE == assemble_src(pAsm, 0, -1) ) + { + return GL_FALSE; + } - if ( GL_FALSE == assemble_src(pAsm, 0, 1) ) + if( GL_FALSE == assemble_src(pAsm, 1, -1) ) { return GL_FALSE; } - if ( GL_FALSE == next_ins(pAsm) ) + if ( GL_FALSE == next_ins2(pAsm) ) { return GL_FALSE; } + /* Doc says KILL has to be last(end) ALU clause */ pAsm->pR700Shader->killIsUsed = GL_TRUE; + pAsm->alu_x_opcode = SQ_CF_INST_ALU; return GL_TRUE; } @@ -3007,6 +3648,7 @@ GLboolean assemble_LRP(r700_AssemblerBase *pAsm) { return GL_FALSE; } + if( GL_FALSE == assemble_src(pAsm, 2, -1) ) { return GL_FALSE; @@ -3816,6 +4458,74 @@ 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) { @@ -3895,6 +4605,7 @@ GLboolean assemble_TEX(r700_AssemblerBase *pAsm) switch (pAsm->pILInst[pAsm->uiCurInst].SrcReg[0].File) { + case PROGRAM_UNIFORM: case PROGRAM_CONSTANT: case PROGRAM_LOCAL_PARAM: case PROGRAM_ENV_PARAM: @@ -4273,27 +4984,876 @@ GLboolean assemble_EXPORT(r700_AssemblerBase *pAsm) return GL_TRUE; } -GLboolean assemble_IF(r700_AssemblerBase *pAsm) +static inline void decreaseCurrent(r700_AssemblerBase *pAsm, GLuint uReason) +{ + switch (uReason) + { + case FC_PUSH_VPM: + pAsm->CALLSTACK[pAsm->CALLSP].current--; + break; + case FC_PUSH_WQM: + pAsm->CALLSTACK[pAsm->CALLSP].current -= 4; + break; + case FC_LOOP: + pAsm->CALLSTACK[pAsm->CALLSP].current -= 4; + break; + case FC_REP: + /* TODO : for 16 vp asic, should -= 2; */ + pAsm->CALLSTACK[pAsm->CALLSP].current -= 1; + break; + }; +} + +static inline void checkStackDepth(r700_AssemblerBase *pAsm, GLuint uReason, GLboolean bCheckMaxOnly) { + if(GL_TRUE == bCheckMaxOnly) + { + switch (uReason) + { + case FC_PUSH_VPM: + if((pAsm->CALLSTACK[pAsm->CALLSP].current + 1) + > pAsm->CALLSTACK[pAsm->CALLSP].max) + { + pAsm->CALLSTACK[pAsm->CALLSP].max = + pAsm->CALLSTACK[pAsm->CALLSP].current + 1; + } + break; + case FC_PUSH_WQM: + if((pAsm->CALLSTACK[pAsm->CALLSP].current + 4) + > pAsm->CALLSTACK[pAsm->CALLSP].max) + { + pAsm->CALLSTACK[pAsm->CALLSP].max = + pAsm->CALLSTACK[pAsm->CALLSP].current + 4; + } + break; + } + return; + } + + switch (uReason) + { + case FC_PUSH_VPM: + pAsm->CALLSTACK[pAsm->CALLSP].current++; + break; + case FC_PUSH_WQM: + pAsm->CALLSTACK[pAsm->CALLSP].current += 4; + break; + case FC_LOOP: + pAsm->CALLSTACK[pAsm->CALLSP].current += 4; + break; + case FC_REP: + /* TODO : for 16 vp asic, should += 2; */ + pAsm->CALLSTACK[pAsm->CALLSP].current += 1; + break; + }; + + if(pAsm->CALLSTACK[pAsm->CALLSP].current + > pAsm->CALLSTACK[pAsm->CALLSP].max) + { + pAsm->CALLSTACK[pAsm->CALLSP].max = + pAsm->CALLSTACK[pAsm->CALLSP].current; + } +} + +GLboolean jumpToOffest(r700_AssemblerBase *pAsm, GLuint pops, GLint offset) +{ + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = pops; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_JUMP; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->cf_current_cf_clause_ptr->m_Word0.f.addr = pAsm->cf_current_cf_clause_ptr->m_uIndex + offset; + + return GL_TRUE; +} + +GLboolean pops(r700_AssemblerBase *pAsm, GLuint pops) +{ + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = pops; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_POP; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + pAsm->cf_current_cf_clause_ptr->m_Word0.f.addr = pAsm->cf_current_cf_clause_ptr->m_uIndex + 1; + + return GL_TRUE; +} + +GLboolean assemble_IF(r700_AssemblerBase *pAsm, GLboolean bHasElse) +{ + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + if(GL_TRUE != bHasElse) + { + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + } + else + { + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 0; + } + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_JUMP; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->FCSP++; + pAsm->fc_stack[pAsm->FCSP].type = FC_IF; + pAsm->fc_stack[pAsm->FCSP].mid = NULL; + pAsm->fc_stack[pAsm->FCSP].midLen= 0; + pAsm->fc_stack[pAsm->FCSP].first = pAsm->cf_current_cf_clause_ptr; + +#ifndef USE_CF_FOR_POP_AFTER + if(GL_TRUE != bHasElse) + { + pAsm->alu_x_opcode = SQ_CF_INST_ALU_POP_AFTER; + } +#endif /* USE_CF_FOR_POP_AFTER */ + + checkStackDepth(pAsm, FC_PUSH_VPM, GL_FALSE); + + return GL_TRUE; +} + +GLboolean assemble_ELSE(r700_AssemblerBase *pAsm) +{ + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; /// + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_ELSE; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->fc_stack[pAsm->FCSP].mid = (R700ControlFlowGenericClause **)_mesa_realloc( (void *)pAsm->fc_stack[pAsm->FCSP].mid, + 0, + sizeof(R700ControlFlowGenericClause *) ); + pAsm->fc_stack[pAsm->FCSP].mid[0] = pAsm->cf_current_cf_clause_ptr; + //pAsm->fc_stack[pAsm->FCSP].unNumMid = 1; + +#ifndef USE_CF_FOR_POP_AFTER + pAsm->alu_x_opcode = SQ_CF_INST_ALU_POP_AFTER; +#endif /* USE_CF_FOR_POP_AFTER */ + + pAsm->fc_stack[pAsm->FCSP].first->m_Word0.f.addr = pAsm->pR700Shader->plstCFInstructions_active->uNumOfNode - 1; + return GL_TRUE; } GLboolean assemble_ENDIF(r700_AssemblerBase *pAsm) { +#ifdef USE_CF_FOR_POP_AFTER + pops(pAsm, 1); +#endif /* USE_CF_FOR_POP_AFTER */ + + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + + if(NULL == pAsm->fc_stack[pAsm->FCSP].mid) + { + /* no else in between */ + pAsm->fc_stack[pAsm->FCSP].first->m_Word0.f.addr = pAsm->pR700Shader->plstCFInstructions_active->uNumOfNode; + } + else + { + pAsm->fc_stack[pAsm->FCSP].mid[0]->m_Word0.f.addr = pAsm->pR700Shader->plstCFInstructions_active->uNumOfNode; + } + + if(NULL != pAsm->fc_stack[pAsm->FCSP].mid) + { + FREE(pAsm->fc_stack[pAsm->FCSP].mid); + } + + if(pAsm->fc_stack[pAsm->FCSP].type != FC_IF) + { + radeon_error("if/endif in shader code are not paired. \n"); + return GL_FALSE; + } + + pAsm->FCSP--; + + decreaseCurrent(pAsm, FC_PUSH_VPM); + + return GL_TRUE; +} + +GLboolean assemble_BGNLOOP(r700_AssemblerBase *pAsm) +{ + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_LOOP_START_NO_AL; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->FCSP++; + pAsm->fc_stack[pAsm->FCSP].type = FC_LOOP; + pAsm->fc_stack[pAsm->FCSP].mid = NULL; + pAsm->fc_stack[pAsm->FCSP].unNumMid = 0; + pAsm->fc_stack[pAsm->FCSP].midLen = 0; + pAsm->fc_stack[pAsm->FCSP].first = pAsm->cf_current_cf_clause_ptr; + + checkStackDepth(pAsm, FC_LOOP, GL_FALSE); + + return GL_TRUE; +} + +GLboolean assemble_BRK(r700_AssemblerBase *pAsm) +{ +#ifdef USE_CF_FOR_CONTINUE_BREAK + unsigned int unFCSP; + for(unFCSP=pAsm->FCSP; unFCSP>0; unFCSP--) + { + if(FC_LOOP == pAsm->fc_stack[unFCSP].type) + { + break; + } + } + if(0 == FC_LOOP) + { + radeon_error("Break is not inside loop/endloop pair.\n"); + return GL_FALSE; + } + + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_LOOP_BREAK; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->fc_stack[unFCSP].mid = (R700ControlFlowGenericClause **)_mesa_realloc( + (void *)pAsm->fc_stack[unFCSP].mid, + sizeof(R700ControlFlowGenericClause *) * pAsm->fc_stack[unFCSP].unNumMid, + sizeof(R700ControlFlowGenericClause *) * (pAsm->fc_stack[unFCSP].unNumMid + 1) ); + pAsm->fc_stack[unFCSP].mid[pAsm->fc_stack[unFCSP].unNumMid] = pAsm->cf_current_cf_clause_ptr; + pAsm->fc_stack[unFCSP].unNumMid++; + + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_POP; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + pAsm->cf_current_cf_clause_ptr->m_Word0.f.addr = pAsm->cf_current_cf_clause_ptr->m_uIndex + 1; + + checkStackDepth(pAsm, FC_PUSH_VPM, GL_TRUE); + +#endif //USE_CF_FOR_CONTINUE_BREAK + return GL_TRUE; +} + +GLboolean 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; + + checkStackDepth(pAsm, FC_PUSH_VPM, GL_TRUE); + +#endif /* USE_CF_FOR_CONTINUE_BREAK */ + + return GL_TRUE; +} + +GLboolean assemble_ENDLOOP(r700_AssemblerBase *pAsm) +{ + GLuint i; + + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_LOOP_END; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->cf_current_cf_clause_ptr->m_Word0.f.addr = pAsm->fc_stack[pAsm->FCSP].first->m_uIndex + 1; + pAsm->fc_stack[pAsm->FCSP].first->m_Word0.f.addr = pAsm->cf_current_cf_clause_ptr->m_uIndex + 1; + +#ifdef USE_CF_FOR_CONTINUE_BREAK + for(i=0; i<pAsm->fc_stack[pAsm->FCSP].unNumMid; i++) + { + pAsm->fc_stack[pAsm->FCSP].mid[i]->m_Word0.f.addr = pAsm->cf_current_cf_clause_ptr->m_uIndex; + } + if(NULL != pAsm->fc_stack[pAsm->FCSP].mid) + { + FREE(pAsm->fc_stack[pAsm->FCSP].mid); + } +#endif + + if(pAsm->fc_stack[pAsm->FCSP].type != FC_LOOP) + { + radeon_error("loop/endloop in shader code are not paired. \n"); + return GL_FALSE; + } + + GLuint unFCSP; + GLuint unIF = 0; + if((pAsm->unCFflags & HAS_CURRENT_LOOPRET) > 0) + { + for(unFCSP=(pAsm->FCSP-1); unFCSP>pAsm->CALLSTACK[pAsm->CALLSP].FCSP_BeforeEntry; unFCSP--) + { + if(FC_LOOP == pAsm->fc_stack[unFCSP].type) + { + breakLoopOnFlag(pAsm, unFCSP); + break; + } + else if(FC_IF == pAsm->fc_stack[unFCSP].type) + { + unIF++; + } + } + if(unFCSP <= pAsm->CALLSTACK[pAsm->CALLSP].FCSP_BeforeEntry) + { +#ifdef USE_CF_FOR_POP_AFTER + returnOnFlag(pAsm, unIF); +#else + returnOnFlag(pAsm, 0); +#endif /* USE_CF_FOR_POP_AFTER */ + pAsm->unCFflags &= ~HAS_CURRENT_LOOPRET; + } + } + + pAsm->FCSP--; + + decreaseCurrent(pAsm, FC_LOOP); + + return GL_TRUE; +} + +void add_return_inst(r700_AssemblerBase *pAsm) +{ + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + //pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_RETURN; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; +} + +GLboolean assemble_BGNSUB(r700_AssemblerBase *pAsm, GLint nILindex) +{ + /* 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].subDescIndex = pAsm->unSubArrayPointer; + pAsm->CALLSTACK[pAsm->CALLSP].FCSP_BeforeEntry = pAsm->FCSP; + pAsm->CALLSTACK[pAsm->CALLSP].plstCFInstructions_local + = &(pAsm->subs[pAsm->unSubArrayPointer].lstCFInstructions_local); + pAsm->CALLSTACK[pAsm->CALLSP].max = 0; + pAsm->CALLSTACK[pAsm->CALLSP].current = 0; + SetActiveCFlist(pAsm->pR700Shader, + pAsm->CALLSTACK[pAsm->CALLSP].plstCFInstructions_local); + + pAsm->unSubArrayPointer++; + + /* start sub */ + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + + pAsm->FCSP++; + pAsm->fc_stack[pAsm->FCSP].type = FC_REP; + + checkStackDepth(pAsm, FC_REP, GL_FALSE); + + return GL_TRUE; +} + +GLboolean assemble_ENDSUB(r700_AssemblerBase *pAsm) +{ + if(pAsm->fc_stack[pAsm->FCSP].type != FC_REP) + { + radeon_error("BGNSUB/ENDSUB in shader code are not paired. \n"); + return GL_FALSE; + } + + /* copy max to sub structure */ + pAsm->subs[pAsm->CALLSTACK[pAsm->CALLSP].subDescIndex].unStackDepthMax + = pAsm->CALLSTACK[pAsm->CALLSP].max; + + decreaseCurrent(pAsm, FC_REP); + + pAsm->CALLSP--; + SetActiveCFlist(pAsm->pR700Shader, + pAsm->CALLSTACK[pAsm->CALLSP].plstCFInstructions_local); + + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + + pAsm->FCSP--; + + return GL_TRUE; +} + +GLboolean assemble_RET(r700_AssemblerBase *pAsm) +{ + GLuint unIF = 0; + + if(pAsm->CALLSP > 0) + { /* in sub */ + GLuint unFCSP; + for(unFCSP=pAsm->FCSP; unFCSP>pAsm->CALLSTACK[pAsm->CALLSP].FCSP_BeforeEntry; unFCSP--) + { + if(FC_LOOP == pAsm->fc_stack[unFCSP].type) + { + setRetInLoopFlag(pAsm, SQ_SEL_1); + breakLoopOnFlag(pAsm, unFCSP); + pAsm->unCFflags |= LOOPRET_FLAGS; + + return GL_TRUE; + } + else if(FC_IF == pAsm->fc_stack[unFCSP].type) + { + unIF++; + } + } + } + +#ifdef USE_CF_FOR_POP_AFTER + if(unIF > 0) + { + pops(pAsm, unIF); + } +#endif /* USE_CF_FOR_POP_AFTER */ + + add_return_inst(pAsm); + + return GL_TRUE; +} + +GLboolean assemble_CAL(r700_AssemblerBase *pAsm, + GLint nILindex, + GLuint 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; + GLuint max; + GLuint unSubID; + GLboolean bRet; + for(j=0; j<pAsm->unSubArrayPointer; j++) + { + if(nILindex == pAsm->subs[j].subIL_Offset) + { /* compiled before */ + + max = pAsm->subs[j].unStackDepthMax + + pAsm->CALLSTACK[pAsm->CALLSP].current; + if(max > pAsm->CALLSTACK[pAsm->CALLSP].max) + { + pAsm->CALLSTACK[pAsm->CALLSP].max = max; + } + + pAsm->callers[pAsm->unCallerArrayPointer - 1].subDescIndex = j; + return GL_TRUE; + } + } + + pAsm->callers[pAsm->unCallerArrayPointer - 1].subDescIndex = pAsm->unSubArrayPointer; + unSubID = pAsm->unSubArrayPointer; + + bRet = AssembleInstr(nILindex, uiNumberInsts, pILInst, pAsm); + + if(GL_TRUE == bRet) + { + max = pAsm->subs[unSubID].unStackDepthMax + + pAsm->CALLSTACK[pAsm->CALLSP].current; + if(max > pAsm->CALLSTACK[pAsm->CALLSP].max) + { + pAsm->CALLSTACK[pAsm->CALLSP].max = max; + } + } + + return bRet; +} + +GLboolean setRetInLoopFlag(r700_AssemblerBase *pAsm, GLuint flagValue) +{ + GLfloat fLiteral[2] = {0.1, 0.0}; + + pAsm->D.dst.opcode = SQ_OP2_INST_MOV; + pAsm->D.dst.op3 = 0; + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = pAsm->flag_reg_index; + pAsm->D.dst.writex = 1; + pAsm->D.dst.writey = 0; + pAsm->D.dst.writez = 0; + pAsm->D.dst.writew = 0; + pAsm->D2.dst2.literal = 1; + pAsm->D2.dst2.SaturateMode = SATURATE_OFF; + pAsm->D.dst.predicated = 0; + /* in reloc where dislink flag init inst, only one slot alu inst is handled. */ + pAsm->D.dst.math = 1; /* TODO : not math really, but one channel op, more generic alu assembler needed */ +#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, GL_TRUE); + + return GL_TRUE; +} + +GLboolean returnOnFlag(r700_AssemblerBase *pAsm, GLuint unIF) +{ + testFlag(pAsm); + jumpToOffest(pAsm, 1, 4); + setRetInLoopFlag(pAsm, SQ_SEL_0); + pops(pAsm, unIF + 1); + add_return_inst(pAsm); + return GL_TRUE; } -GLboolean 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: @@ -4350,7 +5910,8 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, if ( GL_FALSE == assemble_FLR(pR700AsmCode) ) return GL_FALSE; break; - //case OP_FLR_INT: + //case OP_FLR_INT: ; + // if ( GL_FALSE == assemble_FLR_INT() ) // return GL_FALSE; // break; @@ -4361,8 +5922,10 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, break; case OPCODE_KIL: - if ( GL_FALSE == assemble_KIL(pR700AsmCode) ) - return GL_FALSE; + case OPCODE_KIL_NV: + /* done at OPCODE_SE/SGT...etc. */ + /* if ( GL_FALSE == assemble_KIL(pR700AsmCode) ) + return GL_FALSE; */ break; case OPCODE_LG2: if ( GL_FALSE == assemble_LG2(pR700AsmCode) ) @@ -4422,16 +5985,340 @@ 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((OPCODE_KIL == pILInst[i+1].Opcode)||(OPCODE_KIL_NV == pILInst[i+1].Opcode)) + { + if ( GL_FALSE == assemble_KIL(pR700AsmCode, SQ_OP2_INST_KILLE) ) + { + 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((OPCODE_KIL == pILInst[i+1].Opcode)||(OPCODE_KIL_NV == pILInst[i+1].Opcode)) + { + if ( GL_FALSE == assemble_KIL(pR700AsmCode, SQ_OP2_INST_KILLGT) ) + { + 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((OPCODE_KIL == pILInst[i+1].Opcode)||(OPCODE_KIL_NV == pILInst[i+1].Opcode)) + { + if ( GL_FALSE == assemble_KIL(pR700AsmCode, SQ_OP2_INST_KILLGE) ) + { + 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((OPCODE_KIL == pILInst[i+1].Opcode)||(OPCODE_KIL_NV == pILInst[i+1].Opcode)) + { + if ( GL_FALSE == assemble_KIL(pR700AsmCode, SQ_OP2_INST_KILLGT) ) + { + 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((OPCODE_KIL == pILInst[i+1].Opcode)||(OPCODE_KIL_NV == pILInst[i+1].Opcode)) + { + if ( GL_FALSE == assemble_KIL(pR700AsmCode, SQ_OP2_INST_KILLGE) ) + { + 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((OPCODE_KIL == pILInst[i+1].Opcode)||(OPCODE_KIL_NV == pILInst[i+1].Opcode)) + { + if ( GL_FALSE == assemble_KIL(pR700AsmCode, SQ_OP2_INST_KILLNE) ) + { + 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) ) @@ -4465,30 +6352,102 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, return GL_FALSE; break; + case OPCODE_TRUNC: + if ( GL_FALSE == assemble_math_function(pR700AsmCode, SQ_OP2_INST_TRUNC) ) + return GL_FALSE; + break; + case OPCODE_XPD: if ( GL_FALSE == assemble_XPD(pR700AsmCode) ) return GL_FALSE; break; case OPCODE_IF : - if ( GL_FALSE == assemble_IF(pR700AsmCode) ) - return GL_FALSE; + { + GLboolean bHasElse = GL_FALSE; + + if(pILInst[pILInst[i].BranchTarget - 1].Opcode == OPCODE_ELSE) + { + bHasElse = GL_TRUE; + } + + if ( GL_FALSE == assemble_IF(pR700AsmCode, bHasElse) ) + { + return GL_FALSE; + } + } break; + case OPCODE_ELSE : - radeon_error("Not yet implemented instruction OPCODE_ELSE \n"); - //if ( GL_FALSE == assemble_BAD("ELSE") ) + if ( GL_FALSE == assemble_ELSE(pR700AsmCode) ) return GL_FALSE; break; + case OPCODE_ENDIF: if ( GL_FALSE == assemble_ENDIF(pR700AsmCode) ) return GL_FALSE; break; + case OPCODE_BGNLOOP: + if( GL_FALSE == assemble_BGNLOOP(pR700AsmCode) ) + { + return GL_FALSE; + } + break; + + case OPCODE_BRK: + if( GL_FALSE == assemble_BRK(pR700AsmCode) ) + { + return GL_FALSE; + } + break; + + case OPCODE_CONT: + if( GL_FALSE == assemble_CONT(pR700AsmCode) ) + { + return GL_FALSE; + } + break; + + case OPCODE_ENDLOOP: + if( GL_FALSE == assemble_ENDLOOP(pR700AsmCode) ) + { + return GL_FALSE; + } + break; + + case OPCODE_BGNSUB: + if( GL_FALSE == assemble_BGNSUB(pR700AsmCode, i) ) + { + 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 +6464,123 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, return GL_TRUE; } +GLboolean InitShaderProgram(r700_AssemblerBase * pAsm) +{ + setRetInLoopFlag(pAsm, SQ_SEL_0); + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + return GL_TRUE; +} + +GLboolean RelocProgram(r700_AssemblerBase * pAsm) +{ + GLuint i; + GLuint unCFoffset; + TypedShaderList * plstCFmain; + TypedShaderList * plstCFsub; + + R700ShaderInstruction * pInst; + R700ControlFlowGenericClause * pCFInst; + + plstCFmain = pAsm->CALLSTACK[0].plstCFInstructions_local; + + /* remove flags init if they are not used */ + if((pAsm->unCFflags & HAS_LOOPRET) == 0) + { + R700ControlFlowALUClause * pCF_ALU; + pInst = plstCFmain->pHead; + while(pInst) + { + if(SIT_CF_ALU == pInst->m_ShaderInstType) + { + pCF_ALU = (R700ControlFlowALUClause *)pInst; + if(0 == pCF_ALU->m_Word1.f.count) + { + pCF_ALU->m_Word1.f.cf_inst = SQ_CF_INST_NOP; + } + else + { + R700ALUInstruction * pALU = pCF_ALU->m_pLinkedALUInstruction; + + pALU->m_pLinkedALUClause = NULL; + pALU = (R700ALUInstruction *)(pALU->pNextInst); + pALU->m_pLinkedALUClause = pCF_ALU; + pCF_ALU->m_pLinkedALUInstruction = pALU; + + pCF_ALU->m_Word1.f.count--; + } + break; + } + pInst = pInst->pNextInst; + }; + } + + if(pAsm->CALLSTACK[0].max > 0) + { + pAsm->pR700Shader->uStackSize = ((pAsm->CALLSTACK[0].max + 3)>>2) + 2; + } + + if(0 == pAsm->unSubArrayPointer) + { + return GL_TRUE; + } + + unCFoffset = plstCFmain->uNumOfNode; + + /* 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; + } + + return GL_TRUE; +} + GLboolean Process_Export(r700_AssemblerBase* pAsm, GLuint type, GLuint export_starting_index, @@ -4799,6 +6875,25 @@ GLboolean Process_Vertex_Exports(r700_AssemblerBase *pR700AsmCode, export_starting_index++; } } + + for(i=VERT_RESULT_VAR0; i<VERT_RESULT_MAX; i++) + { + unBit = 1 << i; + if(OutputsWritten & unBit) + { + if( GL_FALSE == Process_Export(pR700AsmCode, + SQ_EXPORT_PARAM, + export_starting_index, + 1, + pR700AsmCode->ucVP_OutputMap[i], + GL_FALSE) ) + { + return GL_FALSE; + } + + export_starting_index++; + } + } // At least one param should be exported if (export_count) @@ -4833,6 +6928,16 @@ 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 c66db502a17..130fc89dae1 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, @@ -263,22 +273,52 @@ enum typedef struct FC_LEVEL { - unsigned int first; ///< first fc instruction on level (if, rep, loop) - unsigned int* mid; ///< middle instructions - else or all breaks on this level - unsigned int midLen; - unsigned int type; - unsigned int cond; - unsigned int inv; - unsigned int bpush; ///< 1 if first instruction does branch stack push - int id; ///< id of bool or int variable + R700ControlFlowGenericClause * first; + R700ControlFlowGenericClause ** mid; + unsigned int unNumMid; + unsigned int midLen; + unsigned int type; + unsigned int cond; + unsigned int inv; + int id; ///< id of bool or int variable } FC_LEVEL; typedef struct VTX_FETCH_METHOD { - GLboolean bEnableMini; - GLuint mega_fetch_remainder; + GLboolean bEnableMini; + GLuint mega_fetch_remainder; } VTX_FETCH_METHOD; +typedef struct SUB_OFFSET +{ + GLint subIL_Offset; + GLuint unCFoffset; + GLuint unStackDepthMax; + 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 CALL_LEVEL +{ + unsigned int FCSP_BeforeEntry; + GLint subDescIndex; + GLushort current; + GLushort max; + TypedShaderList * plstCFInstructions_local; +} CALL_LEVEL; + +#define HAS_CURRENT_LOOPRET 0x1L +#define HAS_LOOPRET 0x2L +#define LOOPRET_FLAGS HAS_LOOPRET | HAS_CURRENT_LOOPRET + typedef struct r700_AssemblerBase { R700ControlFlowSXClause* cf_last_export_ptr; @@ -294,11 +334,14 @@ 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 +353,8 @@ typedef struct r700_AssemblerBase unsigned int number_used_registers; unsigned int uUsedConsts; + unsigned int flag_reg_index; + // Fragment programs unsigned int uiFP_AttributeMap[FRAG_ATTRIB_MAX]; unsigned int uiFP_OutputMap[FRAG_RESULT_MAX]; @@ -330,9 +375,6 @@ typedef struct r700_AssemblerBase unsigned int FCSP; FC_LEVEL fc_stack[32]; - unsigned int branch_depth; - unsigned int max_branch_depth; - //----------------------------------------------------------------------------------- // ArgSubst used in Assemble_Source() function //----------------------------------------------------------------------------------- @@ -378,6 +420,18 @@ 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 @@ -446,6 +500,10 @@ 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 +525,21 @@ 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, GLuint unIF); + GLboolean assemble_math_function(r700_AssemblerBase* pAsm, BITS opcode); GLboolean assemble_ABS(r700_AssemblerBase *pAsm); GLboolean assemble_ADD(r700_AssemblerBase *pAsm); @@ -481,7 +554,7 @@ GLboolean assemble_EXP(r700_AssemblerBase *pAsm); GLboolean assemble_FLR(r700_AssemblerBase *pAsm); GLboolean assemble_FLR_INT(r700_AssemblerBase *pAsm); GLboolean assemble_FRC(r700_AssemblerBase *pAsm); -GLboolean assemble_KIL(r700_AssemblerBase *pAsm); +GLboolean assemble_KIL(r700_AssemblerBase *pAsm, GLuint opcode); GLboolean assemble_LG2(r700_AssemblerBase *pAsm); GLboolean assemble_LRP(r700_AssemblerBase *pAsm); GLboolean assemble_LOG(r700_AssemblerBase *pAsm); @@ -497,14 +570,32 @@ 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 +607,16 @@ 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 2b2b4d748f6..8126777bf48 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 ccafd433bfa..e9ef6c86953 100644 --- a/src/mesa/drivers/dri/r600/r700_fragprog.c +++ b/src/mesa/drivers/dri/r600/r700_fragprog.c @@ -44,12 +44,18 @@ //TODO : Validate FP input with VP output. void Map_Fragment_Program(r700_AssemblerBase *pAsm, - struct gl_fragment_program *mesa_fp) + struct gl_fragment_program *mesa_fp, + GLcontext *ctx) { unsigned int unBit; unsigned int i; GLuint ui; + /* match fp inputs with vp exports. */ + struct r700_vertex_program_cont *vpc = + (struct r700_vertex_program_cont *)ctx->VertexProgram._Current; + GLbitfield OutputsWritten = vpc->mesa_program.Base.OutputsWritten; + pAsm->number_used_registers = 0; //Input mapping : mesa_fp->Base.InputsRead set the flag, set in @@ -61,33 +67,89 @@ void Map_Fragment_Program(r700_AssemblerBase *pAsm, pAsm->uiFP_AttributeMap[FRAG_ATTRIB_WPOS] = pAsm->number_used_registers++; } - unBit = 1 << FRAG_ATTRIB_COL0; - if(mesa_fp->Base.InputsRead & unBit) + unBit = 1 << VERT_RESULT_COL0; + if(OutputsWritten & unBit) { pAsm->uiFP_AttributeMap[FRAG_ATTRIB_COL0] = pAsm->number_used_registers++; } - unBit = 1 << FRAG_ATTRIB_COL1; - if(mesa_fp->Base.InputsRead & unBit) + unBit = 1 << VERT_RESULT_COL1; + if(OutputsWritten & unBit) { pAsm->uiFP_AttributeMap[FRAG_ATTRIB_COL1] = pAsm->number_used_registers++; } - unBit = 1 << FRAG_ATTRIB_FOGC; - if(mesa_fp->Base.InputsRead & unBit) - { - pAsm->uiFP_AttributeMap[FRAG_ATTRIB_FOGC] = pAsm->number_used_registers++; - } + unBit = 1 << VERT_RESULT_FOGC; + if(OutputsWritten & unBit) + { + pAsm->uiFP_AttributeMap[FRAG_ATTRIB_FOGC] = pAsm->number_used_registers++; + } for(i=0; i<8; i++) { - unBit = 1 << (FRAG_ATTRIB_TEX0 + i); - if(mesa_fp->Base.InputsRead & unBit) + unBit = 1 << (VERT_RESULT_TEX0 + i); + if(OutputsWritten & unBit) { pAsm->uiFP_AttributeMap[FRAG_ATTRIB_TEX0 + i] = pAsm->number_used_registers++; } } +/* order has been taken care of */ +#if 1 + for(i=VERT_RESULT_VAR0; i<VERT_RESULT_MAX; i++) + { + unBit = 1 << i; + if(OutputsWritten & unBit) + { + pAsm->uiFP_AttributeMap[i-VERT_RESULT_VAR0+FRAG_ATTRIB_VAR0] = pAsm->number_used_registers++; + } + } +#else + if( (mesa_fp->Base.InputsRead >> FRAG_ATTRIB_VAR0) > 0 ) + { + struct r700_vertex_program_cont *vpc = + (struct r700_vertex_program_cont *)ctx->VertexProgram._Current; + struct gl_program_parameter_list * VsVarying = vpc->mesa_program.Base.Varying; + struct gl_program_parameter_list * PsVarying = mesa_fp->Base.Varying; + struct gl_program_parameter * pVsParam; + struct gl_program_parameter * pPsParam; + GLuint j, k; + GLuint unMaxVarying = 0; + + for(i=0; i<VsVarying->NumParameters; i++) + { + pAsm->uiFP_AttributeMap[i + FRAG_ATTRIB_VAR0] = 0; + } + + for(i=FRAG_ATTRIB_VAR0; i<FRAG_ATTRIB_MAX; i++) + { + unBit = 1 << i; + if(mesa_fp->Base.InputsRead & unBit) + { + j = i - FRAG_ATTRIB_VAR0; + pPsParam = PsVarying->Parameters + j; + + for(k=0; k<VsVarying->NumParameters; k++) + { + pVsParam = VsVarying->Parameters + k; + + if( strcmp(pPsParam->Name, pVsParam->Name) == 0) + { + pAsm->uiFP_AttributeMap[i] = pAsm->number_used_registers + k; + if(k > unMaxVarying) + { + unMaxVarying = k; + } + break; + } + } + } + } + + pAsm->number_used_registers += unMaxVarying + 1; + } +#endif + /* Map temporary registers (GPRs) */ pAsm->starting_temp_register_number = pAsm->number_used_registers; @@ -127,6 +189,8 @@ void Map_Fragment_Program(r700_AssemblerBase *pAsm, pAsm->pucOutMask[ui] = 0x0; } + pAsm->flag_reg_index = pAsm->number_used_registers++; + pAsm->uFirstHelpReg = pAsm->number_used_registers; } @@ -233,7 +297,8 @@ GLboolean Find_Instruction_Dependencies_fp(struct r700_fragment_program *fp, } GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp, - struct gl_fragment_program *mesa_fp) + struct gl_fragment_program *mesa_fp, + GLcontext *ctx) { GLuint number_of_colors_exported; GLboolean z_enabled = GL_FALSE; @@ -241,14 +306,17 @@ GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp, //Init_Program Init_r700_AssemblerBase( SPT_FP, &(fp->r700AsmCode), &(fp->r700Shader) ); - Map_Fragment_Program(&(fp->r700AsmCode), mesa_fp); + Map_Fragment_Program(&(fp->r700AsmCode), mesa_fp, ctx); if( GL_FALSE == Find_Instruction_Dependencies_fp(fp, mesa_fp) ) { return GL_FALSE; } + + InitShaderProgram(&(fp->r700AsmCode)); - if( GL_FALSE == AssembleInstr(mesa_fp->Base.NumInstructions, + if( GL_FALSE == AssembleInstr(0, + mesa_fp->Base.NumInstructions, &(mesa_fp->Base.Instructions[0]), &(fp->r700AsmCode)) ) { @@ -260,6 +328,11 @@ 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); @@ -300,7 +373,7 @@ void r700SelectFragmentShader(GLcontext *ctx) } if (GL_FALSE == fp->translated) - r700TranslateFragmentShader(fp, &(fp->mesa_program)); + r700TranslateFragmentShader(fp, &(fp->mesa_program), ctx); } void * r700GetActiveFpShaderBo(GLcontext * ctx) @@ -394,6 +467,9 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) EXPORT_MODE_shift, EXPORT_MODE_mask); // emit ps input map + struct r700_vertex_program_cont *vpc = + (struct r700_vertex_program_cont *)ctx->VertexProgram._Current; + GLbitfield OutputsWritten = vpc->mesa_program.Base.OutputsWritten; unBit = 1 << FRAG_ATTRIB_WPOS; if(mesa_fp->Base.InputsRead & unBit) { @@ -407,8 +483,8 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); } - unBit = 1 << FRAG_ATTRIB_COL0; - if(mesa_fp->Base.InputsRead & unBit) + unBit = 1 << VERT_RESULT_COL0; + if(OutputsWritten & unBit) { ui = pAsm->uiFP_AttributeMap[FRAG_ATTRIB_COL0]; SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, SEL_CENTROID_bit); @@ -420,8 +496,8 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); } - unBit = 1 << FRAG_ATTRIB_COL1; - if(mesa_fp->Base.InputsRead & unBit) + unBit = 1 << VERT_RESULT_COL1; + if(OutputsWritten & unBit) { ui = pAsm->uiFP_AttributeMap[FRAG_ATTRIB_COL1]; SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, SEL_CENTROID_bit); @@ -433,8 +509,8 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); } - unBit = 1 << FRAG_ATTRIB_FOGC; - if(mesa_fp->Base.InputsRead & unBit) + unBit = 1 << VERT_RESULT_FOGC; + if(OutputsWritten & unBit) { ui = pAsm->uiFP_AttributeMap[FRAG_ATTRIB_FOGC]; SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, SEL_CENTROID_bit); @@ -448,8 +524,8 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) for(i=0; i<8; i++) { - unBit = 1 << (FRAG_ATTRIB_TEX0 + i); - if(mesa_fp->Base.InputsRead & unBit) + unBit = 1 << (VERT_RESULT_TEX0 + i); + if(OutputsWritten & unBit) { ui = pAsm->uiFP_AttributeMap[FRAG_ATTRIB_TEX0 + i]; SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, SEL_CENTROID_bit); @@ -459,6 +535,22 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) } } + for(i=VERT_RESULT_VAR0; i<VERT_RESULT_MAX; i++) + { + unBit = 1 << i; + if(OutputsWritten & unBit) + { + ui = pAsm->uiFP_AttributeMap[i-VERT_RESULT_VAR0+FRAG_ATTRIB_VAR0]; + SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, SEL_CENTROID_bit); + SETfield(r700->SPI_PS_INPUT_CNTL[ui].u32All, ui, + SEMANTIC_shift, SEMANTIC_mask); + if (r700->SPI_INTERP_CONTROL_0.u32All & FLAT_SHADE_ENA_bit) + SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); + else + CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); + } + } + exportCount = (r700->ps.SQ_PGM_EXPORTS_PS.u32All & EXPORT_MODE_mask) / (1 << EXPORT_MODE_shift); if (r700->CB_SHADER_CONTROL.u32All != ((1 << exportCount) - 1)) { @@ -469,7 +561,8 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) /* sent out shader constants. */ paramList = fp->mesa_program.Base.Parameters; - if(NULL != paramList) { + if(NULL != paramList) + { _mesa_load_state_parameters(ctx, paramList); if (paramList->NumParameters > R700_MAX_DX9_CONSTS) @@ -482,10 +575,10 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) unNumParamData = paramList->NumParameters; for(ui=0; ui<unNumParamData; ui++) { - r700->ps.consts[ui][0].f32All = paramList->ParameterValues[ui][0]; - r700->ps.consts[ui][1].f32All = paramList->ParameterValues[ui][1]; - r700->ps.consts[ui][2].f32All = paramList->ParameterValues[ui][2]; - r700->ps.consts[ui][3].f32All = paramList->ParameterValues[ui][3]; + r700->ps.consts[ui][0].f32All = paramList->ParameterValues[ui][0]; + r700->ps.consts[ui][1].f32All = paramList->ParameterValues[ui][1]; + r700->ps.consts[ui][2].f32All = paramList->ParameterValues[ui][2]; + r700->ps.consts[ui][3].f32All = paramList->ParameterValues[ui][3]; } } else r700->ps.num_consts = 0; diff --git a/src/mesa/drivers/dri/r600/r700_fragprog.h b/src/mesa/drivers/dri/r600/r700_fragprog.h index cbb108d2128..e562bfa4789 100644 --- a/src/mesa/drivers/dri/r600/r700_fragprog.h +++ b/src/mesa/drivers/dri/r600/r700_fragprog.h @@ -49,12 +49,14 @@ struct r700_fragment_program /* Internal */ void Map_Fragment_Program(r700_AssemblerBase *pAsm, - struct gl_fragment_program *mesa_fp); + struct gl_fragment_program *mesa_fp, + GLcontext *ctx); GLboolean Find_Instruction_Dependencies_fp(struct r700_fragment_program *fp, struct gl_fragment_program *mesa_fp); GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp, - struct gl_fragment_program *mesa_vp); + struct gl_fragment_program *mesa_vp, + GLcontext *ctx); /* Interface */ extern void r700SelectFragmentShader(GLcontext *ctx); diff --git a/src/mesa/drivers/dri/r600/r700_render.c b/src/mesa/drivers/dri/r600/r700_render.c index 47f89c91f87..eab27cbd842 100644 --- a/src/mesa/drivers/dri/r600/r700_render.c +++ b/src/mesa/drivers/dri/r600/r700_render.c @@ -526,6 +526,9 @@ static void r700ConvertAttrib(GLcontext *ctx, int count, radeonAllocDmaRegion(&context->radeon, &attr->bo, &attr->bo_offset, sizeof(GLfloat) * input->Size * count, 32); + + radeon_bo_map(attr->bo, 1); + dst_ptr = (GLfloat *)ADD_POINTERS(attr->bo->ptr, attr->bo_offset); assert(src_ptr != NULL); @@ -559,6 +562,8 @@ static void r700ConvertAttrib(GLcontext *ctx, int count, break; } + radeon_bo_unmap(attr->bo); + if (mapped_named_bo) { ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER, input->BufferObj); @@ -577,6 +582,8 @@ static void r700AlignDataToDword(GLcontext *ctx, radeonAllocDmaRegion(&context->radeon, &attr->bo, &attr->bo_offset, size, 32); + radeon_bo_map(attr->bo, 1); + if (!input->BufferObj->Pointer) { ctx->Driver.MapBuffer(ctx, GL_ARRAY_BUFFER, GL_READ_ONLY_ARB, input->BufferObj); @@ -596,6 +603,7 @@ static void r700AlignDataToDword(GLcontext *ctx, } } + radeon_bo_unmap(attr->bo); if (mapped_named_bo) { ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER, input->BufferObj); @@ -664,14 +672,18 @@ static void r700SetupStreams(GLcontext *ctx, const struct gl_client_array *input radeonAllocDmaRegion(&context->radeon, &context->stream_desc[index].bo, &context->stream_desc[index].bo_offset, size, 32); + + radeon_bo_map(context->stream_desc[index].bo, 1); assert(context->stream_desc[index].bo->ptr != NULL); + + dst = (uint32_t *)ADD_POINTERS(context->stream_desc[index].bo->ptr, context->stream_desc[index].bo_offset); switch (context->stream_desc[index].dwords) { case 1: - radeonEmitVec4(dst, input[i]->Ptr, input[i]->StrideB, local_count); + radeonEmitVec4(dst, input[i]->Ptr, input[i]->StrideB, local_count); break; case 2: radeonEmitVec8(dst, input[i]->Ptr, input[i]->StrideB, local_count); @@ -686,6 +698,7 @@ static void r700SetupStreams(GLcontext *ctx, const struct gl_client_array *input assert(0); break; } + radeon_bo_unmap(context->stream_desc[index].bo); } } @@ -757,6 +770,7 @@ static void r700FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer radeonAllocDmaRegion(&context->radeon, &context->ind_buf.bo, &context->ind_buf.bo_offset, size, 4); + radeon_bo_map(context->ind_buf.bo, 1); assert(context->ind_buf.bo->ptr != NULL); out = (GLuint *)ADD_POINTERS(context->ind_buf.bo->ptr, context->ind_buf.bo_offset); @@ -770,6 +784,7 @@ static void r700FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer *out++ = in[i]; } + radeon_bo_unmap(context->ind_buf.bo); #if MESA_BIG_ENDIAN } else @@ -780,6 +795,7 @@ static void r700FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer radeonAllocDmaRegion(&context->radeon, &context->ind_buf.bo, &context->ind_buf.bo_offset, size, 4); + radeon_bo_map(context->ind_buf.bo, 1); assert(context->ind_buf.bo->ptr != NULL); out = (GLuint *)ADD_POINTERS(context->ind_buf.bo->ptr, context->ind_buf.bo_offset); @@ -792,6 +808,7 @@ static void r700FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer { *out++ = in[i]; } + radeon_bo_unmap(context->ind_buf.bo); #endif } @@ -837,11 +854,13 @@ static void r700SetupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer radeonAllocDmaRegion(&context->radeon, &context->ind_buf.bo, &context->ind_buf.bo_offset, size, 4); + radeon_bo_map(context->ind_buf.bo, 1); assert(context->ind_buf.bo->ptr != NULL); dst_ptr = ADD_POINTERS(context->ind_buf.bo->ptr, context->ind_buf.bo_offset); _mesa_memcpy(dst_ptr, src_ptr, size); + radeon_bo_unmap(context->ind_buf.bo); context->ind_buf.is_32bit = (mesa_ind_buf->type == GL_UNSIGNED_INT); context->ind_buf.count = mesa_ind_buf->count; diff --git a/src/mesa/drivers/dri/r600/r700_shader.c b/src/mesa/drivers/dri/r600/r700_shader.c index 955ea4e4e1d..2eed1acc2f5 100644 --- a/src/mesa/drivers/dri/r600/r700_shader.c +++ b/src/mesa/drivers/dri/r600/r700_shader.c @@ -159,13 +159,18 @@ void Init_R700_Shader(R700_Shader * pShader) pShader->lstVTXInstructions.uNumOfNode=0; } +void SetActiveCFlist(R700_Shader *pShader, TypedShaderList * plstCF) +{ + pShader->plstCFInstructions_active = plstCF; +} + void AddCFInstruction(R700_Shader *pShader, R700ControlFlowInstruction *pCFInst) { R700ControlFlowSXClause* pSXClause; R700ControlFlowSMXClause* pSMXClause; - pCFInst->m_uIndex = pShader->lstCFInstructions.uNumOfNode; - AddInstToList(&(pShader->lstCFInstructions), + pCFInst->m_uIndex = pShader->plstCFInstructions_active->uNumOfNode; + AddInstToList(pShader->plstCFInstructions_active, (R700ShaderInstruction*)pCFInst); pShader->uShaderBinaryDWORDSize += GetInstructionSize(pCFInst->m_ShaderInstType); diff --git a/src/mesa/drivers/dri/r600/r700_shader.h b/src/mesa/drivers/dri/r600/r700_shader.h index c6a058617ec..0599ffd901f 100644 --- a/src/mesa/drivers/dri/r600/r700_shader.h +++ b/src/mesa/drivers/dri/r600/r700_shader.h @@ -109,6 +109,7 @@ typedef struct R700_Shader GLuint uStackSize; GLuint uMaxCallDepth; + TypedShaderList * plstCFInstructions_active; TypedShaderList lstCFInstructions; TypedShaderList lstALUInstructions; TypedShaderList lstTEXInstructions; @@ -132,13 +133,13 @@ void TakeInstOutFromList(TypedShaderList * plstCFInstructions, R700ShaderInstruc void ResolveLinks(R700_Shader *pShader); void Assemble(R700_Shader *pShader); - //Interface void Init_R700_Shader(R700_Shader * pShader); void AddCFInstruction(R700_Shader *pShader, R700ControlFlowInstruction *pCFInst); void AddVTXInstruction(R700_Shader *pShader, R700VertexInstruction *pVTXInst); void AddTEXInstruction(R700_Shader *pShader, R700TextureInstruction *pTEXInst); void AddALUInstruction(R700_Shader *pShader, R700ALUInstruction *pALUInst); +void SetActiveCFlist(R700_Shader *pShader, TypedShaderList * plstCF); void LoadProgram(R700_Shader *pShader); void UpdateShaderRegisters(R700_Shader *pShader); diff --git a/src/mesa/drivers/dri/r600/r700_vertprog.c b/src/mesa/drivers/dri/r600/r700_vertprog.c index ffc6068bd88..d3d1da79592 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); @@ -612,6 +631,12 @@ GLboolean r700SetupVertexProgram(GLcontext * ctx) paramList = vp->mesa_program->Base.Parameters; if(NULL != paramList) { + /* vp->mesa_program was cloned, not updated by glsl shader api. */ + /* _mesa_reference_program has already checked glsl shProg is ok and set ctx->VertexProgem._Current */ + /* so, use ctx->VertexProgem._Current */ + struct gl_program_parameter_list *paramListOrginal = + paramListOrginal = ctx->VertexProgram._Current->Base.Parameters; + _mesa_load_state_parameters(ctx, paramList); if (paramList->NumParameters > R700_MAX_DX9_CONSTS) @@ -624,10 +649,20 @@ GLboolean r700SetupVertexProgram(GLcontext * ctx) unNumParamData = paramList->NumParameters; for(ui=0; ui<unNumParamData; ui++) { - r700->vs.consts[ui][0].f32All = paramList->ParameterValues[ui][0]; - r700->vs.consts[ui][1].f32All = paramList->ParameterValues[ui][1]; - r700->vs.consts[ui][2].f32All = paramList->ParameterValues[ui][2]; - r700->vs.consts[ui][3].f32All = paramList->ParameterValues[ui][3]; + if(paramList->Parameters[ui].Type == PROGRAM_UNIFORM) + { + r700->vs.consts[ui][0].f32All = paramListOrginal->ParameterValues[ui][0]; + r700->vs.consts[ui][1].f32All = paramListOrginal->ParameterValues[ui][1]; + r700->vs.consts[ui][2].f32All = paramListOrginal->ParameterValues[ui][2]; + r700->vs.consts[ui][3].f32All = paramListOrginal->ParameterValues[ui][3]; + } + else + { + r700->vs.consts[ui][0].f32All = paramList->ParameterValues[ui][0]; + r700->vs.consts[ui][1].f32All = paramList->ParameterValues[ui][1]; + r700->vs.consts[ui][2].f32All = paramList->ParameterValues[ui][2]; + r700->vs.consts[ui][3].f32All = paramList->ParameterValues[ui][3]; + } } } else r700->vs.num_consts = 0; diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.h b/src/mesa/drivers/dri/radeon/radeon_common_context.h index ded81fff29f..ad953ddbb5a 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common_context.h +++ b/src/mesa/drivers/dri/radeon/radeon_common_context.h @@ -328,6 +328,7 @@ struct radeon_swtcl_info { GLuint vertex_attr_count; GLuint emit_prediction; + struct radeon_bo *bo; }; #define RADEON_MAX_AOS_ARRAYS 16 diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.c b/src/mesa/drivers/dri/radeon/radeon_dma.c index c9a32c808be..b8c65f4ce62 100644 --- a/src/mesa/drivers/dri/radeon/radeon_dma.c +++ b/src/mesa/drivers/dri/radeon/radeon_dma.c @@ -151,6 +151,7 @@ void rcommon_emit_vector(GLcontext * ctx, struct radeon_aos *aos, aos->components = size; aos->count = count; + radeon_bo_map(aos->bo, 1); out = (uint32_t*)((char*)aos->bo->ptr + aos->offset); switch (size) { case 1: radeonEmitVec4(out, data, stride, count); break; @@ -161,6 +162,7 @@ void rcommon_emit_vector(GLcontext * ctx, struct radeon_aos *aos, assert(0); break; } + radeon_bo_unmap(aos->bo); } void radeon_init_dma(radeonContextPtr rmesa) @@ -183,10 +185,6 @@ void radeonRefillCurrentDmaRegion(radeonContextPtr rmesa, int size) __FUNCTION__, size, rmesa->dma.minimum_size); - /* unmap old reserved bo */ - if (!is_empty_list(&rmesa->dma.reserved)) - radeon_bo_unmap(first_elem(&rmesa->dma.reserved)->bo); - if (is_empty_list(&rmesa->dma.free) || last_elem(&rmesa->dma.free)->bo->size < size) { dma_bo = CALLOC_STRUCT(radeon_dma_bo); @@ -224,8 +222,6 @@ again_alloc: /* Cmd buff have been flushed in radeon_revalidate_bos */ goto again_alloc; } - - radeon_bo_map(first_elem(&rmesa->dma.reserved)->bo, 1); } /* Allocates a region from rmesa->dma.current. If there isn't enough @@ -282,7 +278,6 @@ void radeonFreeDmaRegions(radeonContextPtr rmesa) foreach_s(dma_bo, temp, &rmesa->dma.reserved) { remove_from_list(dma_bo); - radeon_bo_unmap(dma_bo->bo); radeon_bo_unref(dma_bo->bo); FREE(dma_bo); } @@ -362,9 +357,6 @@ void radeonReleaseDmaRegions(radeonContextPtr rmesa) insert_at_tail(&rmesa->dma.free, dma_bo); } - /* unmap the last dma region */ - if (!is_empty_list(&rmesa->dma.reserved)) - radeon_bo_unmap(first_elem(&rmesa->dma.reserved)->bo); /* move reserved to wait list */ foreach_s(dma_bo, temp, &rmesa->dma.reserved) { /* free objects that are too small to be used because of large request */ @@ -398,11 +390,12 @@ void rcommon_flush_last_swtcl_prim( GLcontext *ctx ) radeonContextPtr rmesa = RADEON_CONTEXT(ctx); struct radeon_dma *dma = &rmesa->dma; - if (RADEON_DEBUG & RADEON_IOCTL) fprintf(stderr, "%s\n", __FUNCTION__); dma->flush = NULL; + radeon_bo_unmap(rmesa->swtcl.bo); + if (!is_empty_list(&dma->reserved)) { GLuint current_offset = dma->current_used; @@ -417,6 +410,8 @@ void rcommon_flush_last_swtcl_prim( GLcontext *ctx ) } rmesa->swtcl.numverts = 0; } + radeon_bo_unref(rmesa->swtcl.bo); + rmesa->swtcl.bo = NULL; } /* Alloc space in the current dma region. */ @@ -427,6 +422,7 @@ rcommonAllocDmaLowVerts( radeonContextPtr rmesa, int nverts, int vsize ) void *head; if (RADEON_DEBUG & RADEON_IOCTL) fprintf(stderr, "%s\n", __FUNCTION__); + if(is_empty_list(&rmesa->dma.reserved) ||rmesa->dma.current_vertexptr + bytes > first_elem(&rmesa->dma.reserved)->bo->size) { if (rmesa->dma.flush) { @@ -450,7 +446,13 @@ rcommonAllocDmaLowVerts( radeonContextPtr rmesa, int nverts, int vsize ) rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 == rmesa->dma.current_vertexptr ); - head = (first_elem(&rmesa->dma.reserved)->bo->ptr + rmesa->dma.current_vertexptr); + if (!rmesa->swtcl.bo) { + rmesa->swtcl.bo = first_elem(&rmesa->dma.reserved)->bo; + radeon_bo_ref(rmesa->swtcl.bo); + radeon_bo_map(rmesa->swtcl.bo, 1); + } + + head = (rmesa->swtcl.bo->ptr + rmesa->dma.current_vertexptr); rmesa->dma.current_vertexptr += bytes; rmesa->swtcl.numverts += nverts; return head; diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c b/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c index 08e1c5d00df..d810e6080eb 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c +++ b/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c @@ -76,12 +76,14 @@ static void emit_vecfog(GLcontext *ctx, struct radeon_aos *aos, /* Emit the data */ + radeon_bo_map(aos->bo, 1); out = (uint32_t*)((char*)aos->bo->ptr + aos->offset); for (i = 0; i < count; i++) { out[0] = radeonComputeFogBlendFactor( ctx, *(GLfloat *)data ); out++; data += stride; } + radeon_bo_unmap(aos->bo); } static void emit_s0_vec(uint32_t *out, GLvoid *data, int stride, int count) @@ -151,6 +153,7 @@ static void emit_tex_vector(GLcontext *ctx, struct radeon_aos *aos, /* Emit the data */ + radeon_bo_map(aos->bo, 1); out = (uint32_t*)((char*)aos->bo->ptr + aos->offset); switch (size) { case 1: @@ -170,6 +173,7 @@ static void emit_tex_vector(GLcontext *ctx, struct radeon_aos *aos, exit(1); break; } + radeon_bo_unmap(aos->bo); } @@ -196,12 +200,12 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (!rmesa->tcl.obj.buf) rcommon_emit_vector( ctx, &(rmesa->tcl.aos[nr]), - (char *)VB->ObjPtr->data, - VB->ObjPtr->size, - VB->ObjPtr->stride, + (char *)VB->AttribPtr[_TNL_ATTRIB_POS]->data, + VB->AttribPtr[_TNL_ATTRIB_POS]->size, + VB->AttribPtr[_TNL_ATTRIB_POS]->stride, count); - switch( VB->ObjPtr->size ) { + switch( VB->AttribPtr[_TNL_ATTRIB_POS]->size ) { case 4: vfmt |= RADEON_CP_VC_FRMT_W0; case 3: vfmt |= RADEON_CP_VC_FRMT_Z; case 2: vfmt |= RADEON_CP_VC_FRMT_XY; @@ -216,9 +220,9 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (!rmesa->tcl.norm.buf) rcommon_emit_vector( ctx, &(rmesa->tcl.aos[nr]), - (char *)VB->NormalPtr->data, + (char *)VB->AttribPtr[_TNL_ATTRIB_NORMAL]->data, 3, - VB->NormalPtr->stride, + VB->AttribPtr[_TNL_ATTRIB_NORMAL]->stride, count); vfmt |= RADEON_CP_VC_FRMT_N0; @@ -227,9 +231,9 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (inputs & VERT_BIT_COLOR0) { int emitsize; - if (VB->ColorPtr[0]->size == 4 && - (VB->ColorPtr[0]->stride != 0 || - VB->ColorPtr[0]->data[0][3] != 1.0)) { + if (VB->AttribPtr[_TNL_ATTRIB_COLOR0]->size == 4 && + (VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride != 0 || + VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data[0][3] != 1.0)) { vfmt |= RADEON_CP_VC_FRMT_FPCOLOR | RADEON_CP_VC_FRMT_FPALPHA; emitsize = 4; } @@ -242,9 +246,9 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (!rmesa->tcl.rgba.buf) rcommon_emit_vector( ctx, &(rmesa->tcl.aos[nr]), - (char *)VB->ColorPtr[0]->data, + (char *)VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data, emitsize, - VB->ColorPtr[0]->stride, + VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride, count); nr++; @@ -256,9 +260,9 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) rcommon_emit_vector( ctx, &(rmesa->tcl.aos[nr]), - (char *)VB->SecondaryColorPtr[0]->data, + (char *)VB->AttribPtr[_TNL_ATTRIB_COLOR1]->data, 3, - VB->SecondaryColorPtr[0]->stride, + VB->AttribPtr[_TNL_ATTRIB_COLOR1]->stride, count); } @@ -273,8 +277,8 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (!rmesa->tcl.fog.buf) emit_vecfog( ctx, &(rmesa->tcl.aos[nr]), - (char *)VB->FogCoordPtr->data, - VB->FogCoordPtr->stride, + (char *)VB->AttribPtr[_TNL_ATTRIB_FOG]->data, + VB->AttribPtr[_TNL_ATTRIB_FOG]->stride, count); vfmt |= RADEON_CP_VC_FRMT_FPFOG; @@ -290,24 +294,24 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (!rmesa->tcl.tex[unit].buf) emit_tex_vector( ctx, &(rmesa->tcl.aos[nr]), - (char *)VB->TexCoordPtr[unit]->data, - VB->TexCoordPtr[unit]->size, - VB->TexCoordPtr[unit]->stride, + (char *)VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->data, + VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size, + VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->stride, count ); nr++; vfmt |= RADEON_ST_BIT(unit); /* assume we need the 3rd coord if texgen is active for r/q OR at least 3 coords are submitted. This may not be 100% correct */ - if (VB->TexCoordPtr[unit]->size >= 3) { + if (VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size >= 3) { vtx |= RADEON_Q_BIT(unit); vfmt |= RADEON_Q_BIT(unit); } if ( (ctx->Texture.Unit[unit].TexGenEnabled & (R_BIT | Q_BIT)) ) vtx |= RADEON_Q_BIT(unit); - else if ((VB->TexCoordPtr[unit]->size >= 3) && + else if ((VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size >= 3) && ((ctx->Texture.Unit[unit]._ReallyEnabled & (TEXTURE_CUBE_BIT)) == 0)) { - GLuint swaptexmatcol = (VB->TexCoordPtr[unit]->size - 3); + GLuint swaptexmatcol = (VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size - 3); if (((rmesa->NeedTexMatrix >> unit) & 1) && (swaptexmatcol != ((rmesa->TexMatColSwap >> unit) & 1))) radeonUploadTexMatrix( rmesa, unit, swaptexmatcol ) ; diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h b/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h index 515783135d6..d764ccb9826 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h +++ b/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h @@ -56,18 +56,18 @@ static void TAG(emit)( GLcontext *ctx, radeon_print(RADEON_SWRENDER, RADEON_VERBOSE, "%s\n", __FUNCTION__); - coord = (GLuint (*)[4])VB->ObjPtr->data; - coord_stride = VB->ObjPtr->stride; + coord = (GLuint (*)[4])VB->AttribPtr[_TNL_ATTRIB_POS]->data; + coord_stride = VB->AttribPtr[_TNL_ATTRIB_POS]->stride; if (DO_TEX2) { - if (VB->TexCoordPtr[2]) { + if (VB->AttribPtr[_TNL_ATTRIB_TEX2]) { const GLuint t2 = GET_TEXSOURCE(2); - tc2 = (GLuint (*)[4])VB->TexCoordPtr[t2]->data; - tc2_stride = VB->TexCoordPtr[t2]->stride; - if (DO_PTEX && VB->TexCoordPtr[t2]->size < 3) { + tc2 = (GLuint (*)[4])VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->data; + tc2_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->stride; + if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->size < 3) { fill_tex |= (1<<2); } - else if (DO_PTEX && VB->TexCoordPtr[t2]->size < 4) { + else if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->size < 4) { rqcoordsnoswap |= (1<<2); } } else { @@ -77,14 +77,14 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_TEX1) { - if (VB->TexCoordPtr[1]) { + if (VB->AttribPtr[_TNL_ATTRIB_TEX1]) { const GLuint t1 = GET_TEXSOURCE(1); - tc1 = (GLuint (*)[4])VB->TexCoordPtr[t1]->data; - tc1_stride = VB->TexCoordPtr[t1]->stride; - if (DO_PTEX && VB->TexCoordPtr[t1]->size < 3) { + tc1 = (GLuint (*)[4])VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->data; + tc1_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->stride; + if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->size < 3) { fill_tex |= (1<<1); } - else if (DO_PTEX && VB->TexCoordPtr[t1]->size < 4) { + else if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->size < 4) { rqcoordsnoswap |= (1<<1); } } else { @@ -94,14 +94,14 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_TEX0) { - if (VB->TexCoordPtr[0]) { + if (VB->AttribPtr[_TNL_ATTRIB_TEX0]) { const GLuint t0 = GET_TEXSOURCE(0); - tc0_stride = VB->TexCoordPtr[t0]->stride; - tc0 = (GLuint (*)[4])VB->TexCoordPtr[t0]->data; - if (DO_PTEX && VB->TexCoordPtr[t0]->size < 3) { + tc0_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->stride; + tc0 = (GLuint (*)[4])VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->data; + if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->size < 3) { fill_tex |= (1<<0); } - else if (DO_PTEX && VB->TexCoordPtr[t0]->size < 4) { + else if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->size < 4) { rqcoordsnoswap |= (1<<0); } } else { @@ -112,9 +112,9 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_NORM) { - if (VB->NormalPtr) { - norm_stride = VB->NormalPtr->stride; - norm = (GLuint (*)[4])VB->NormalPtr->data; + if (VB->AttribPtr[_TNL_ATTRIB_NORMAL]) { + norm_stride = VB->AttribPtr[_TNL_ATTRIB_NORMAL]->stride; + norm = (GLuint (*)[4])VB->AttribPtr[_TNL_ATTRIB_NORMAL]->data; } else { norm_stride = 0; norm = (GLuint (*)[4])&ctx->Current.Attrib[VERT_ATTRIB_NORMAL]; @@ -122,9 +122,9 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_RGBA) { - if (VB->ColorPtr[0]) { - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + if (VB->AttribPtr[_TNL_ATTRIB_COLOR0]) { + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; } else { col = (GLfloat (*)[4])ctx->Current.Attrib[VERT_ATTRIB_COLOR0]; col_stride = 0; @@ -132,9 +132,9 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_SPEC_OR_FOG) { - if (VB->SecondaryColorPtr[0]) { - spec = VB->SecondaryColorPtr[0]->data; - spec_stride = VB->SecondaryColorPtr[0]->stride; + if (VB->AttribPtr[_TNL_ATTRIB_COLOR1]) { + spec = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->data; + spec_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->stride; } else { spec = (GLfloat (*)[4])ctx->Current.Attrib[VERT_ATTRIB_COLOR1]; spec_stride = 0; @@ -142,9 +142,9 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_SPEC_OR_FOG) { - if (VB->FogCoordPtr) { - fog = VB->FogCoordPtr->data; - fog_stride = VB->FogCoordPtr->stride; + if (VB->AttribPtr[_TNL_ATTRIB_FOG]) { + fog = VB->AttribPtr[_TNL_ATTRIB_FOG]->data; + fog_stride = VB->AttribPtr[_TNL_ATTRIB_FOG]->stride; } else { fog = (GLfloat (*)[4])ctx->Current.Attrib[VERT_ATTRIB_FOG]; fog_stride = 0; diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c index 78ec1193026..98f96ff2a76 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c +++ b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c @@ -326,7 +326,7 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (1) { req |= RADEON_CP_VC_FRMT_Z; - if (VB->ObjPtr->size == 4) { + if (VB->AttribPtr[_TNL_ATTRIB_POS]->size == 4) { req |= RADEON_CP_VC_FRMT_W0; } } @@ -348,15 +348,15 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) req |= RADEON_ST_BIT(unit); /* assume we need the 3rd coord if texgen is active for r/q OR at least 3 coords are submitted. This may not be 100% correct */ - if (VB->TexCoordPtr[unit]->size >= 3) { + if (VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size >= 3) { req |= RADEON_Q_BIT(unit); vtx |= RADEON_Q_BIT(unit); } if ( (ctx->Texture.Unit[unit].TexGenEnabled & (R_BIT | Q_BIT)) ) vtx |= RADEON_Q_BIT(unit); - else if ((VB->TexCoordPtr[unit]->size >= 3) && + else if ((VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size >= 3) && ((ctx->Texture.Unit[unit]._ReallyEnabled & (TEXTURE_CUBE_BIT)) == 0)) { - GLuint swaptexmatcol = (VB->TexCoordPtr[unit]->size - 3); + GLuint swaptexmatcol = (VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size - 3); if (((rmesa->NeedTexMatrix >> unit) & 1) && (swaptexmatcol != ((rmesa->TexMatColSwap >> unit) & 1))) radeonUploadTexMatrix( rmesa, unit, swaptexmatcol ) ; @@ -390,19 +390,19 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) * this, add more vertex code (for obj-2, obj-3) or preferably move * to maos. */ - if (VB->ObjPtr->size < 3 || - (VB->ObjPtr->size == 3 && + if (VB->AttribPtr[_TNL_ATTRIB_POS]->size < 3 || + (VB->AttribPtr[_TNL_ATTRIB_POS]->size == 3 && (setup_tab[i].vertex_format & RADEON_CP_VC_FRMT_W0))) { _math_trans_4f( rmesa->tcl.ObjClean.data, - VB->ObjPtr->data, - VB->ObjPtr->stride, + VB->AttribPtr[_TNL_ATTRIB_POS]->data, + VB->AttribPtr[_TNL_ATTRIB_POS]->stride, GL_FLOAT, - VB->ObjPtr->size, + VB->AttribPtr[_TNL_ATTRIB_POS]->size, 0, VB->Count ); - switch (VB->ObjPtr->size) { + switch (VB->AttribPtr[_TNL_ATTRIB_POS]->size) { case 1: _mesa_vector4f_clean_elem(&rmesa->tcl.ObjClean, VB->Count, 1); case 2: @@ -416,14 +416,14 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) break; } - VB->ObjPtr = &rmesa->tcl.ObjClean; + VB->AttribPtr[_TNL_ATTRIB_POS] = &rmesa->tcl.ObjClean; } - + radeon_bo_map(rmesa->radeon.tcl.aos[0].bo, 1); setup_tab[i].emit( ctx, 0, VB->Count, rmesa->radeon.tcl.aos[0].bo->ptr + rmesa->radeon.tcl.aos[0].offset); - + radeon_bo_unmap(rmesa->radeon.tcl.aos[0].bo); // rmesa->radeon.tcl.aos[0].size = setup_tab[i].vertex_size; rmesa->radeon.tcl.aos[0].stride = setup_tab[i].vertex_size; rmesa->tcl.vertex_format = setup_tab[i].vertex_format; diff --git a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c index a9d601a0b5f..bdbb9460bf0 100644 --- a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c +++ b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c @@ -68,6 +68,19 @@ static unsigned get_compressed_image_size( return rowStride * ((height + blockHeight - 1) / blockHeight); } +static int find_next_power_of_two(GLuint value) +{ + int i, tmp; + + i = 0; + tmp = value - 1; + while (tmp) { + tmp >>= 1; + i++; + } + return (1 << i); +} + /** * Compute sizes and fill in offset and blit information for the given * image (determined by \p face and \p level). @@ -80,25 +93,28 @@ static void compute_tex_image_offset(radeonContextPtr rmesa, radeon_mipmap_tree { radeon_mipmap_level *lvl = &mt->levels[level]; uint32_t row_align; + GLuint height; + + height = find_next_power_of_two(lvl->height); /* Find image size in bytes */ if (_mesa_is_format_compressed(mt->mesaFormat)) { lvl->rowstride = get_aligned_compressed_row_stride(mt->mesaFormat, lvl->width, rmesa->texture_compressed_row_align); - lvl->size = get_compressed_image_size(mt->mesaFormat, lvl->rowstride, lvl->height); + lvl->size = get_compressed_image_size(mt->mesaFormat, lvl->rowstride, height); } else if (mt->target == GL_TEXTURE_RECTANGLE_NV) { row_align = rmesa->texture_rect_row_align - 1; lvl->rowstride = (_mesa_format_row_stride(mt->mesaFormat, lvl->width) + row_align) & ~row_align; - lvl->size = lvl->rowstride * lvl->height; + lvl->size = lvl->rowstride * height; } else if (mt->tilebits & RADEON_TXO_MICRO_TILE) { /* tile pattern is 16 bytes x2. mipmaps stay 32 byte aligned, * though the actual offset may be different (if texture is less than * 32 bytes width) to the untiled case */ lvl->rowstride = (_mesa_format_row_stride(mt->mesaFormat, lvl->width) * 2 + 31) & ~31; - lvl->size = lvl->rowstride * ((lvl->height + 1) / 2) * lvl->depth; + lvl->size = lvl->rowstride * ((height + 1) / 2) * lvl->depth; } else { row_align = rmesa->texture_row_align - 1; lvl->rowstride = (_mesa_format_row_stride(mt->mesaFormat, lvl->width) + row_align) & ~row_align; - lvl->size = lvl->rowstride * lvl->height * lvl->depth; + lvl->size = lvl->rowstride * height * lvl->depth; } assert(lvl->size > 0); @@ -110,7 +126,7 @@ static void compute_tex_image_offset(radeonContextPtr rmesa, radeon_mipmap_tree if (RADEON_DEBUG & RADEON_TEXTURE) fprintf(stderr, "level %d, face %d: rs:%d %dx%d at %d\n", - level, face, lvl->rowstride, lvl->width, lvl->height, lvl->faces[face].offset); + level, face, lvl->rowstride, lvl->width, height, lvl->faces[face].offset); } static GLuint minify(GLuint size, GLuint levels) diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.c b/src/mesa/drivers/dri/radeon/radeon_swtcl.c index e61f59eaeaf..8bf1bfbc575 100644 --- a/src/mesa/drivers/dri/radeon/radeon_swtcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.c @@ -179,7 +179,7 @@ static void radeonSetVertexFormat( GLcontext *ctx ) for (i = 0; i < ctx->Const.MaxTextureUnits; i++) { if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX(i) )) { - GLuint sz = VB->TexCoordPtr[i]->size; + GLuint sz = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->size; switch (sz) { case 1: @@ -309,7 +309,7 @@ void r100_swtcl_flush(GLcontext *ctx, uint32_t current_offset) radeonEmitState(&rmesa->radeon); radeonEmitVertexAOS( rmesa, rmesa->radeon.swtcl.vertex_size, - first_elem(&rmesa->radeon.dma.reserved)->bo, + rmesa->radeon.swtcl.bo, current_offset); diff --git a/src/mesa/drivers/dri/savage/savagerender.c b/src/mesa/drivers/dri/savage/savagerender.c index 32c74f9467e..8221edf3872 100644 --- a/src/mesa/drivers/dri/savage/savagerender.c +++ b/src/mesa/drivers/dri/savage/savagerender.c @@ -252,13 +252,13 @@ static GLboolean run_texnorm_stage( GLcontext *ctx, const GLboolean normalizeS = (texObj->WrapS == GL_REPEAT); const GLboolean normalizeT = (reallyEnabled & TEXTURE_2D_BIT) && (texObj->WrapT == GL_REPEAT); - const GLfloat *in = (GLfloat *)VB->TexCoordPtr[i]->data; - const GLint instride = VB->TexCoordPtr[i]->stride; + const GLfloat *in = (GLfloat *)VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->data; + const GLint instride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->stride; GLfloat (*out)[4] = store->texcoord[i].data; GLint j; if (!ctx->Texture.Unit[i]._ReallyEnabled || - VB->TexCoordPtr[i]->size == 4) + VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->size == 4) /* Never try to normalize homogenous tex coords! */ continue; @@ -297,7 +297,7 @@ static GLboolean run_texnorm_stage( GLcontext *ctx, } if (normalizeS || normalizeT) - VB->AttribPtr[VERT_ATTRIB_TEX0+i] = VB->TexCoordPtr[i] = &store->texcoord[i]; + VB->AttribPtr[_TNL_ATTRIB_TEX0 + i] = &store->texcoord[i]; } } diff --git a/src/mesa/drivers/dri/savage/savagetris.c b/src/mesa/drivers/dri/savage/savagetris.c index c04763b40e1..e9529d19391 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 76d12d07b3c..3cf10007b58 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 49288022328..c200ba3255a 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_vb.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_vb.c @@ -69,11 +69,11 @@ static void interp_extras( GLcontext *ctx, /*fprintf(stderr, "%s\n", __FUNCTION__);*/ - if (VB->ColorPtr[1]) { + if (VB->BackfaceColorPtr) { INTERP_4F( t, - GET_COLOR(VB->ColorPtr[1], dst), - GET_COLOR(VB->ColorPtr[1], out), - GET_COLOR(VB->ColorPtr[1], in) ); + GET_COLOR(VB->BackfaceColorPtr, dst), + GET_COLOR(VB->BackfaceColorPtr, out), + GET_COLOR(VB->BackfaceColorPtr, in) ); } if (VB->EdgeFlag) { @@ -88,9 +88,9 @@ static void copy_pv_extras( GLcontext *ctx, GLuint dst, GLuint src ) { struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; - if (VB->ColorPtr[1]) { - COPY_4FV( GET_COLOR(VB->ColorPtr[1], dst), - GET_COLOR(VB->ColorPtr[1], src) ); + if (VB->BackfaceColorPtr) { + COPY_4FV( GET_COLOR(VB->BackfaceColorPtr, dst), + GET_COLOR(VB->BackfaceColorPtr, src) ); } setup_tab[TDFX_CONTEXT(ctx)->SetupIndex].copy_pv(ctx, dst, src); diff --git a/src/mesa/drivers/dri/tdfx/tdfx_vbtmp.h b/src/mesa/drivers/dri/tdfx/tdfx_vbtmp.h index 9b780761f42..19baf7d0d25 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_vbtmp.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_vbtmp.h @@ -58,32 +58,32 @@ static void TAG(emit)( GLcontext *ctx, /* fprintf(stderr, "%s\n", __FUNCTION__); */ if (IND & TDFX_TEX0_BIT) { - tc0_stride = VB->TexCoordPtr[tmu0_source]->stride; - tc0 = VB->TexCoordPtr[tmu0_source]->data; + tc0_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu0_source]->stride; + tc0 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu0_source]->data; u0scale = fxMesa->sScale0; v0scale = fxMesa->tScale0; if (IND & TDFX_PTEX_BIT) - tc0_size = VB->TexCoordPtr[tmu0_source]->size; + tc0_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu0_source]->size; } if (IND & TDFX_TEX1_BIT) { - tc1 = VB->TexCoordPtr[tmu1_source]->data; - tc1_stride = VB->TexCoordPtr[tmu1_source]->stride; + tc1 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu1_source]->data; + tc1_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu1_source]->stride; u1scale = fxMesa->sScale1; v1scale = fxMesa->tScale1; if (IND & TDFX_PTEX_BIT) - tc1_size = VB->TexCoordPtr[tmu1_source]->size; + tc1_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu1_source]->size; } if (IND & TDFX_RGBA_BIT) { - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; - col_size = VB->ColorPtr[0]->size; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; + col_size = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->size; } if (IND & TDFX_FOGC_BIT) { - fog = VB->FogCoordPtr->data; - fog_stride = VB->FogCoordPtr->stride; + fog = VB->AttribPtr[_TNL_ATTRIB_FOG]->data; + fog_stride = VB->AttribPtr[_TNL_ATTRIB_FOG]->stride; } { @@ -168,14 +168,14 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; if (IND & TDFX_TEX1_BIT) { - if (VB->TexCoordPtr[0] == 0) - VB->TexCoordPtr[0] = VB->TexCoordPtr[1]; + if (VB->AttribPtr[_TNL_ATTRIB_TEX0] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0] = VB->AttribPtr[_TNL_ATTRIB_TEX1]; - if (VB->TexCoordPtr[1]->size == 4) + if (VB->AttribPtr[_TNL_ATTRIB_TEX1]->size == 4) return GL_FALSE; } - if (VB->TexCoordPtr[0]->size == 4) + if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size == 4) return GL_FALSE; } diff --git a/src/mesa/drivers/dri/unichrome/via_tris.c b/src/mesa/drivers/dri/unichrome/via_tris.c index 79e67620c9e..ab457d41dc8 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); diff --git a/src/mesa/drivers/glide/fxvb.c b/src/mesa/drivers/glide/fxvb.c index 1dc5f9891a1..cc9ad0e8b83 100644 --- a/src/mesa/drivers/glide/fxvb.c +++ b/src/mesa/drivers/glide/fxvb.c @@ -104,24 +104,24 @@ static void interp_extras( GLcontext *ctx, { struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; - if (VB->ColorPtr[1]) { - /* If stride is zero, ColorPtr[1] is constant across the VB, so + if (VB->BackfaceColorPtr) { + /* If stride is zero, BackfaceColorPtr is constant across the VB, so * there is no point interpolating between two values as they will * be identical. This case is handled in t_dd_tritmp.h */ - if (VB->ColorPtr[1]->stride) { - assert(VB->ColorPtr[1]->stride == 4 * sizeof(GLfloat)); + if (VB->BackfaceColorPtr->stride) { + assert(VB->BackfaceColorPtr->stride == 4 * sizeof(GLfloat)); INTERP_4F( t, - GET_COLOR(VB->ColorPtr[1], dst), - GET_COLOR(VB->ColorPtr[1], out), - GET_COLOR(VB->ColorPtr[1], in) ); + GET_COLOR(VB->BackfaceColorPtr, dst), + GET_COLOR(VB->BackfaceColorPtr, out), + GET_COLOR(VB->BackfaceColorPtr, in) ); } - if (VB->SecondaryColorPtr[1]) { + if (VB->BackfaceSecondaryColorPtr) { INTERP_3F( t, - GET_COLOR(VB->SecondaryColorPtr[1], dst), - GET_COLOR(VB->SecondaryColorPtr[1], out), - GET_COLOR(VB->SecondaryColorPtr[1], in) ); + GET_COLOR(VB->BackfaceSecondaryColorPtr, dst), + GET_COLOR(VB->BackfaceSecondaryColorPtr, out), + GET_COLOR(VB->BackfaceSecondaryColorPtr, in) ); } } @@ -137,13 +137,13 @@ static void copy_pv_extras( GLcontext *ctx, GLuint dst, GLuint src ) { struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; - if (VB->ColorPtr[1]) { - COPY_4FV( GET_COLOR(VB->ColorPtr[1], dst), - GET_COLOR(VB->ColorPtr[1], src) ); + if (VB->BackfaceColorPtr) { + COPY_4FV( GET_COLOR(VB->BackfaceColorPtr, dst), + GET_COLOR(VB->BackfaceColorPtr, src) ); - if (VB->SecondaryColorPtr[1]) { - COPY_3FV( GET_COLOR(VB->SecondaryColorPtr[1], dst), - GET_COLOR(VB->SecondaryColorPtr[1], src) ); + if (VB->BackfaceSecondaryColorPtr) { + COPY_3FV( GET_COLOR(VB->BackfaceSecondaryColorPtr, dst), + GET_COLOR(VB->BackfaceSecondaryColorPtr, src) ); } } diff --git a/src/mesa/drivers/glide/fxvbtmp.h b/src/mesa/drivers/glide/fxvbtmp.h index f7970c78e29..f7893c1573f 100644 --- a/src/mesa/drivers/glide/fxvbtmp.h +++ b/src/mesa/drivers/glide/fxvbtmp.h @@ -62,37 +62,37 @@ static void TAG(emit)( GLcontext *ctx, } if (IND & SETUP_TMU0) { - tc0 = VB->TexCoordPtr[tmu0_source]->data; - tc0_stride = VB->TexCoordPtr[tmu0_source]->stride; + tc0 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu0_source]->data; + tc0_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu0_source]->stride; u0scale = fxMesa->s0scale; v0scale = fxMesa->t0scale; if (IND & SETUP_PTEX) - tc0_size = VB->TexCoordPtr[tmu0_source]->size; + tc0_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu0_source]->size; } if (IND & SETUP_TMU1) { - tc1 = VB->TexCoordPtr[tmu1_source]->data; - tc1_stride = VB->TexCoordPtr[tmu1_source]->stride; + tc1 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu1_source]->data; + tc1_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu1_source]->stride; u1scale = fxMesa->s1scale; /* wrong if tmu1_source == 0, possible? */ v1scale = fxMesa->t1scale; if (IND & SETUP_PTEX) - tc1_size = VB->TexCoordPtr[tmu1_source]->size; + tc1_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu1_source]->size; } if (IND & SETUP_RGBA) { - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; - col_size = VB->ColorPtr[0]->size; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; + col_size = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->size; } if (IND & SETUP_SPEC) { - spec = VB->SecondaryColorPtr[0]->data; - spec_stride = VB->SecondaryColorPtr[0]->stride; + spec = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->data; + spec_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->stride; } if (IND & SETUP_FOGC) { - fog = VB->FogCoordPtr->data; - fog_stride = VB->FogCoordPtr->stride; + fog = VB->AttribPtr[_TNL_ATTRIB_FOG]->data; + fog_stride = VB->AttribPtr[_TNL_ATTRIB_FOG]->stride; } if (start) { @@ -220,14 +220,15 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; if (IND & SETUP_TMU1) { - if (VB->TexCoordPtr[0] == 0) - VB->TexCoordPtr[0] = VB->TexCoordPtr[1]; + if (VB->AttribPtr[_TNL_ATTRIB_TEX0] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0] = VB->AttribPtr[_TNL_ATTRIB_TEX1]; - if (VB->TexCoordPtr[1]->size == 4) + if (VB->AttribPtr[_TNL_ATTRIB_TEX1]->size == 4) return GL_FALSE; } - if (VB->TexCoordPtr[0] && VB->TexCoordPtr[0]->size == 4) + if (VB->AttribPtr[_TNL_ATTRIB_TEX0] && + VB->AttribPtr[_TNL_ATTRIB_TEX0]->size == 4) return GL_FALSE; } diff --git a/src/mesa/drivers/windows/gldirect/dx7/gld_primitive_dx7.c b/src/mesa/drivers/windows/gldirect/dx7/gld_primitive_dx7.c index c99ba0bba55..0b373814fee 100644 --- a/src/mesa/drivers/windows/gldirect/dx7/gld_primitive_dx7.c +++ b/src/mesa/drivers/windows/gldirect/dx7/gld_primitive_dx7.c @@ -189,9 +189,9 @@ GLfloat ex,ey,fx,fy,cc; \ /* Get vars for later */ \ VB = &TNL_CONTEXT(ctx)->vb; \ - vbcolor = (GLchan (*)[4])VB->ColorPtr[1]->data; \ - if (VB->SecondaryColorPtr[1]) { \ - vbspec = (GLchan (*)[4])VB->SecondaryColorPtr[1]->data; \ + vbcolor = (GLchan (*)[4])VB->BackfaceColorPtr->data; \ + if (VB->BackfaceSecondaryColorPtr) { \ + vbspec = (GLchan (*)[4])VB->BackfaceSecondaryColorPtr->data; \ } else { \ vbspec = NULL; \ } \ @@ -241,33 +241,33 @@ DWORD dwColor; #define GLD_SETUP_3D_VERTEX(v) \ - p4f = VB->ObjPtr->data; \ + p4f = VB->AttribPtr[_TNL_ATTRIB_POS]->data; \ pV->Position.x = p4f[##v][0]; \ pV->Position.y = p4f[##v][1]; \ pV->Position.z = p4f[##v][2]; #define GLD_SETUP_SMOOTH_COLOUR_3D(v) \ - p4f = (GLfloat (*)[4])VB->ColorPtr[0]->data; \ + p4f = (GLfloat (*)[4])VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; \ pV->Diffuse = D3DCOLOR_COLORVALUE(p4f[##v][0], p4f[##v][1], p4f[##v][2], p4f[##v][3]); #define GLD_SETUP_GET_FLAT_COLOUR_3D(v) \ - p4f = (GLfloat (*)[4])VB->ColorPtr[0]->data; \ + p4f = (GLfloat (*)[4])VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; \ dwColor = D3DCOLOR_COLORVALUE(p4f[##v][0], p4f[##v][1], p4f[##v][2], p4f[##v][3]); #define GLD_SETUP_USE_FLAT_COLOUR_3D \ pV->Diffuse = dwColor; #define GLD_SETUP_TEX0_3D(v) \ - if (VB->TexCoordPtr[0]) { \ - tc = VB->TexCoordPtr[0]->data; \ + if (VB->AttribPtr[_TNL_ATTRIB_TEX0]) { \ + tc = VB->AttribPtr[_TNL_ATTRIB_TEX0]->data; \ pV->TexUnit0.x = tc[##v][0]; \ pV->TexUnit0.y = tc[##v][1]; \ } #define GLD_SETUP_TEX1_3D(v) \ - if (VB->TexCoordPtr[1]) { \ - tc = VB->TexCoordPtr[1]->data; \ + if (VB->AttribPtr[_TNL_ATTRIB_TEX1]) { \ + tc = VB->AttribPtr[_TNL_ATTRIB_TEX1]->data; \ pV->TexUnit1.x = tc[##v][0]; \ pV->TexUnit1.y = tc[##v][1]; \ } diff --git a/src/mesa/drivers/windows/gldirect/dx7/gld_vb_d3d_render_dx7.c b/src/mesa/drivers/windows/gldirect/dx7/gld_vb_d3d_render_dx7.c index a85620dde8d..c39775cad32 100644 --- a/src/mesa/drivers/windows/gldirect/dx7/gld_vb_d3d_render_dx7.c +++ b/src/mesa/drivers/windows/gldirect/dx7/gld_vb_d3d_render_dx7.c @@ -151,7 +151,7 @@ static GLboolean gld_d3d_render_stage_run( #if 0 // For debugging: Useful to see if an app passes colour data in // an unusual format. - switch (VB->ColorPtr[0]->Type) { + switch (VB->AttribPtr[_TNL_ATTRIB_COLOR0]->Type) { case GL_FLOAT: ddlogMessage(GLDLOG_SYSTEM, "ColorPtr: GL_FLOAT\n"); break; diff --git a/src/mesa/drivers/windows/gldirect/dx8/gld_primitive_dx8.c b/src/mesa/drivers/windows/gldirect/dx8/gld_primitive_dx8.c index a5b5462f039..990922580aa 100644 --- a/src/mesa/drivers/windows/gldirect/dx8/gld_primitive_dx8.c +++ b/src/mesa/drivers/windows/gldirect/dx8/gld_primitive_dx8.c @@ -189,9 +189,9 @@ GLfloat ex,ey,fx,fy,cc; \ /* Get vars for later */ \ VB = &TNL_CONTEXT(ctx)->vb; \ - vbcolor = (GLchan (*)[4])VB->ColorPtr[1]->data; \ - if (VB->SecondaryColorPtr[1]) { \ - vbspec = (GLchan (*)[4])VB->SecondaryColorPtr[1]->data; \ + vbcolor = (GLchan (*)[4])VB->BackfaceColorPtr->data; \ + if (VB->BackfaceSecondaryColorPtr) { \ + vbspec = (GLchan (*)[4])VB->BackfaceSecondaryColorPtr->data; \ } else { \ vbspec = NULL; \ } \ @@ -241,33 +241,33 @@ DWORD dwColor; #define GLD_SETUP_3D_VERTEX(v) \ - p4f = VB->ObjPtr->data; \ + p4f = VB->AttribPtr[_TNL_ATTRIB_POS]->data; \ pV->Position.x = p4f[##v][0]; \ pV->Position.y = p4f[##v][1]; \ pV->Position.z = p4f[##v][2]; #define GLD_SETUP_SMOOTH_COLOUR_3D(v) \ - p4f = (GLfloat (*)[4])VB->ColorPtr[0]->data; \ + p4f = (GLfloat (*)[4])VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; \ pV->Diffuse = D3DCOLOR_COLORVALUE(p4f[##v][0], p4f[##v][1], p4f[##v][2], p4f[##v][3]); #define GLD_SETUP_GET_FLAT_COLOUR_3D(v) \ - p4f = (GLfloat (*)[4])VB->ColorPtr[0]->data; \ + p4f = (GLfloat (*)[4])VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; \ dwColor = D3DCOLOR_COLORVALUE(p4f[##v][0], p4f[##v][1], p4f[##v][2], p4f[##v][3]); #define GLD_SETUP_USE_FLAT_COLOUR_3D \ pV->Diffuse = dwColor; #define GLD_SETUP_TEX0_3D(v) \ - if (VB->TexCoordPtr[0]) { \ - tc = VB->TexCoordPtr[0]->data; \ + if (VB->AttribPtr[_TNL_ATTRIB_TEX0]) { \ + tc = VB->TnlAttribPtr[_TNL_ATTRIB_TEX0]->data; \ pV->TexUnit0.x = tc[##v][0]; \ pV->TexUnit0.y = tc[##v][1]; \ } #define GLD_SETUP_TEX1_3D(v) \ - if (VB->TexCoordPtr[1]) { \ - tc = VB->TexCoordPtr[1]->data; \ + if (VB->TnlAttribPtr[_TNL_ATTRIB_TEX1]) { \ + tc = VB->TnlAttribPtr[_TNL_ATTRIB_TEX1]->data; \ pV->TexUnit1.x = tc[##v][0]; \ pV->TexUnit1.y = tc[##v][1]; \ } diff --git a/src/mesa/drivers/windows/gldirect/dx8/gld_vb_d3d_render_dx8.c b/src/mesa/drivers/windows/gldirect/dx8/gld_vb_d3d_render_dx8.c index cafbf4f5c50..265c81fb4a4 100644 --- a/src/mesa/drivers/windows/gldirect/dx8/gld_vb_d3d_render_dx8.c +++ b/src/mesa/drivers/windows/gldirect/dx8/gld_vb_d3d_render_dx8.c @@ -149,7 +149,7 @@ static GLboolean gld_d3d_render_stage_run( #if 0 // For debugging: Useful to see if an app passes colour data in // an unusual format. - switch (VB->ColorPtr[0]->Type) { + switch (VB->AttribPtr[_TNL_ATTRIB_COLOR0]->Type) { case GL_FLOAT: ddlogMessage(GLDLOG_SYSTEM, "ColorPtr: GL_FLOAT\n"); break; diff --git a/src/mesa/drivers/windows/gldirect/dx9/gld_primitive_dx9.c b/src/mesa/drivers/windows/gldirect/dx9/gld_primitive_dx9.c index 403a9d5f86e..fd4dd4ed751 100644 --- a/src/mesa/drivers/windows/gldirect/dx9/gld_primitive_dx9.c +++ b/src/mesa/drivers/windows/gldirect/dx9/gld_primitive_dx9.c @@ -189,9 +189,9 @@ GLfloat ex,ey,fx,fy,cc; \ /* Get vars for later */ \ VB = &TNL_CONTEXT(ctx)->vb; \ - vbcolor = (GLchan (*)[4])VB->ColorPtr[1]->data; \ - if (VB->SecondaryColorPtr[1]) { \ - vbspec = (GLchan (*)[4])VB->SecondaryColorPtr[1]->data; \ + vbcolor = (GLchan (*)[4])VB->BackfaceColorPtr->data; \ + if (VB->BackfaceSecondaryColorPtr) { \ + vbspec = (GLchan (*)[4])VB->BackfaceSecondaryColorPtr->data; \ } else { \ vbspec = NULL; \ } \ @@ -241,33 +241,33 @@ DWORD dwColor; #define GLD_SETUP_3D_VERTEX(v) \ - p4f = VB->ObjPtr->data; \ + p4f = VB->AttribPtr[_TNL_ATTRIB_POS]->data; \ pV->Position.x = p4f[##v][0]; \ pV->Position.y = p4f[##v][1]; \ pV->Position.z = p4f[##v][2]; #define GLD_SETUP_SMOOTH_COLOUR_3D(v) \ - p4f = (GLfloat (*)[4])VB->ColorPtr[0]->data; \ + p4f = (GLfloat (*)[4])VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; \ pV->Diffuse = D3DCOLOR_COLORVALUE(p4f[##v][0], p4f[##v][1], p4f[##v][2], p4f[##v][3]); #define GLD_SETUP_GET_FLAT_COLOUR_3D(v) \ - p4f = (GLfloat (*)[4])VB->ColorPtr[0]->data; \ + p4f = (GLfloat (*)[4])VB->AttribPtr[_TNL_ATTRIB_COLOR00]->data; \ dwColor = D3DCOLOR_COLORVALUE(p4f[##v][0], p4f[##v][1], p4f[##v][2], p4f[##v][3]); #define GLD_SETUP_USE_FLAT_COLOUR_3D \ pV->Diffuse = dwColor; #define GLD_SETUP_TEX0_3D(v) \ - if (VB->TexCoordPtr[0]) { \ - tc = VB->TexCoordPtr[0]->data; \ + if (VB->AttribPtr[_TNL_ATTRIB_TEX0]) { \ + tc = VB->AttribPtr[_TNL_ATTRIB_TEX0]->data; \ pV->TexUnit0.x = tc[##v][0]; \ pV->TexUnit0.y = tc[##v][1]; \ } #define GLD_SETUP_TEX1_3D(v) \ - if (VB->TexCoordPtr[1]) { \ - tc = VB->TexCoordPtr[1]->data; \ + if (VB->AttribPtr[_TNL_ATTRIB_TEX1]) { \ + tc = VB->AttribPtr[_TNL_ATTRIB_TEX1]->data; \ pV->TexUnit1.x = tc[##v][0]; \ pV->TexUnit1.y = tc[##v][1]; \ } diff --git a/src/mesa/drivers/windows/gldirect/dx9/gld_vb_d3d_render_dx9.c b/src/mesa/drivers/windows/gldirect/dx9/gld_vb_d3d_render_dx9.c index 4fa6bcaf1ab..91a68b3f2d9 100644 --- a/src/mesa/drivers/windows/gldirect/dx9/gld_vb_d3d_render_dx9.c +++ b/src/mesa/drivers/windows/gldirect/dx9/gld_vb_d3d_render_dx9.c @@ -149,7 +149,7 @@ static GLboolean gld_d3d_render_stage_run( #if 0 // For debugging: Useful to see if an app passes colour data in // an unusual format. - switch (VB->ColorPtr[0]->Type) { + switch (VB->AttribPtr[_TNL_ATTRIB_COLOR0]->Type) { case GL_FLOAT: ddlogMessage(GLDLOG_SYSTEM, "ColorPtr: GL_FLOAT\n"); break; diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index b5bf46718f7..03fc57e6654 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -575,6 +575,7 @@ _mesa_init_constants(GLcontext *ctx) #if FEATURE_ARB_vertex_shader ctx->Const.MaxVertexTextureImageUnits = MAX_VERTEX_TEXTURE_IMAGE_UNITS; + ctx->Const.MaxCombinedTextureImageUnits = MAX_COMBINED_TEXTURE_IMAGE_UNITS; ctx->Const.MaxVarying = MAX_VARYING; #endif diff --git a/src/mesa/main/ffvertex_prog.c b/src/mesa/main/ffvertex_prog.c index fe2416d8946..5cfa898031e 100644 --- a/src/mesa/main/ffvertex_prog.c +++ b/src/mesa/main/ffvertex_prog.c @@ -523,7 +523,6 @@ static void emit_dst( struct prog_dst_register *dst, dst->CondMask = COND_TR; /* always pass cond test */ dst->CondSwizzle = SWIZZLE_NOOP; dst->CondSrc = 0; - dst->pad = 0; /* Check that bitfield sizes aren't exceeded */ ASSERT(dst->Index == reg.idx); } diff --git a/src/mesa/main/get.c b/src/mesa/main/get.c index 6c5ce029135..e8932f83b6a 100644 --- a/src/mesa/main/get.c +++ b/src/mesa/main/get.c @@ -1876,7 +1876,7 @@ _mesa_GetBooleanv( GLenum pname, GLboolean *params ) break; case GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS_ARB: CHECK_EXT1(ARB_vertex_shader, "GetBooleanv"); - params[0] = INT_TO_BOOLEAN(MAX_COMBINED_TEXTURE_IMAGE_UNITS); + params[0] = INT_TO_BOOLEAN(ctx->Const.MaxCombinedTextureImageUnits); break; case GL_CURRENT_PROGRAM: CHECK_EXT1(ARB_shader_objects, "GetBooleanv"); @@ -3711,7 +3711,7 @@ _mesa_GetFloatv( GLenum pname, GLfloat *params ) break; case GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS_ARB: CHECK_EXT1(ARB_vertex_shader, "GetFloatv"); - params[0] = (GLfloat)(MAX_COMBINED_TEXTURE_IMAGE_UNITS); + params[0] = (GLfloat)(ctx->Const.MaxCombinedTextureImageUnits); break; case GL_CURRENT_PROGRAM: CHECK_EXT1(ARB_shader_objects, "GetFloatv"); @@ -5546,7 +5546,7 @@ _mesa_GetIntegerv( GLenum pname, GLint *params ) break; case GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS_ARB: CHECK_EXT1(ARB_vertex_shader, "GetIntegerv"); - params[0] = MAX_COMBINED_TEXTURE_IMAGE_UNITS; + params[0] = ctx->Const.MaxCombinedTextureImageUnits; break; case GL_CURRENT_PROGRAM: CHECK_EXT1(ARB_shader_objects, "GetIntegerv"); @@ -7382,7 +7382,7 @@ _mesa_GetInteger64v( GLenum pname, GLint64 *params ) break; case GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS_ARB: CHECK_EXT1(ARB_vertex_shader, "GetInteger64v"); - params[0] = (GLint64)(MAX_COMBINED_TEXTURE_IMAGE_UNITS); + params[0] = (GLint64)(ctx->Const.MaxCombinedTextureImageUnits); break; case GL_CURRENT_PROGRAM: CHECK_EXT1(ARB_shader_objects, "GetInteger64v"); diff --git a/src/mesa/main/get_gen.py b/src/mesa/main/get_gen.py index 930c3362fae..a29962d3348 100644 --- a/src/mesa/main/get_gen.py +++ b/src/mesa/main/get_gen.py @@ -1006,7 +1006,7 @@ StateVars = [ ( "GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS_ARB", GLint, ["ctx->Const.MaxVertexTextureImageUnits"], "", ["ARB_vertex_shader"] ), ( "GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS_ARB", GLint, - ["MAX_COMBINED_TEXTURE_IMAGE_UNITS"], "", ["ARB_vertex_shader"] ), + ["ctx->Const.MaxCombinedTextureImageUnits"], "", ["ARB_vertex_shader"] ), # GL_ARB_shader_objects # Actually, this token isn't part of GL_ARB_shader_objects, but is diff --git a/src/mesa/main/imports.c b/src/mesa/main/imports.c index 46ffb929b6c..c9e00cf7528 100644 --- a/src/mesa/main/imports.c +++ b/src/mesa/main/imports.c @@ -629,11 +629,15 @@ _mesa_ffsll(int64_t val) unsigned int _mesa_bitcount(unsigned int n) { +#if defined(__GNUC__) + return __builtin_popcount(n); +#else unsigned int bits; for (bits = 0; n > 0; n = n >> 1) { bits += (n & 1); } return bits; +#endif } diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h index 881d233ca3d..5f012448275 100644 --- a/src/mesa/main/mtypes.h +++ b/src/mesa/main/mtypes.h @@ -2319,6 +2319,7 @@ struct gl_constants GLuint MaxTextureCoordUnits; GLuint MaxTextureImageUnits; GLuint MaxVertexTextureImageUnits; + GLuint MaxCombinedTextureImageUnits; GLuint MaxTextureUnits; /**< = MIN(CoordUnits, ImageUnits) */ GLfloat MaxTextureMaxAnisotropy; /**< GL_EXT_texture_filter_anisotropic */ GLfloat MaxTextureLodBias; /**< GL_EXT_texture_lod_bias */ diff --git a/src/mesa/main/texgetimage.c b/src/mesa/main/texgetimage.c index 23765d27530..bd7cc8d2780 100644 --- a/src/mesa/main/texgetimage.c +++ b/src/mesa/main/texgetimage.c @@ -555,7 +555,8 @@ _mesa_get_compressed_teximage(GLcontext *ctx, GLenum target, GLint level, _mesa_get_format_block_size(texImage->TexFormat, &bw, &bh); for (i = 0; i < (texImage->Height + bh - 1) / bh; i++) { memcpy((GLubyte *)img + i * row_stride, - (GLubyte *)texImage->Data + i * row_stride_stored, row_stride); + (GLubyte *)texImage->Data + i * row_stride_stored, + row_stride); } } diff --git a/src/mesa/main/version.h b/src/mesa/main/version.h index fddb9a851f4..dc55cb7ccc9 100644 --- a/src/mesa/main/version.h +++ b/src/mesa/main/version.h @@ -1,6 +1,6 @@ /* * Mesa 3-D graphics library - * Version: 7.7 + * Version: 7.8 * * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. * Copyright (C) 2009 VMware, Inc. All Rights Reserved. @@ -30,9 +30,9 @@ /* Mesa version */ #define MESA_MAJOR 7 -#define MESA_MINOR 7 +#define MESA_MINOR 8 #define MESA_PATCH 0 -#define MESA_VERSION_STRING "7.7-rc1" +#define MESA_VERSION_STRING "7.8-devel" /* To make version comparison easy */ #define MESA_VERSION(a,b,c) (((a) << 16) + ((b) << 8) + (c)) diff --git a/src/mesa/shader/prog_instruction.h b/src/mesa/shader/prog_instruction.h index 1c687bc16cf..224350caac6 100644 --- a/src/mesa/shader/prog_instruction.h +++ b/src/mesa/shader/prog_instruction.h @@ -312,7 +312,6 @@ struct prog_dst_register */ GLuint CondSrc:1; /*@}*/ - GLuint pad:28; }; diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index c0e4b27aa5f..99eb254cee0 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -81,8 +81,8 @@ new_subroutine(slang_emit_info *emitInfo, GLuint *id) emitInfo->Subroutines = (struct gl_program **) _mesa_realloc(emitInfo->Subroutines, - n * sizeof(struct gl_program), - (n + 1) * sizeof(struct gl_program)); + n * sizeof(struct gl_program *), + (n + 1) * sizeof(struct gl_program *)); emitInfo->Subroutines[n] = ctx->Driver.NewProgram(ctx, emitInfo->prog->Target, 0); emitInfo->Subroutines[n]->Parameters = emitInfo->prog->Parameters; emitInfo->NumSubroutines++; @@ -551,6 +551,9 @@ emit_instruction(slang_emit_info *emitInfo, &srcRelAddr, NULL, NULL); + if (!inst) { + return NULL; + } src[i] = &newSrc[i]; } @@ -948,6 +951,9 @@ emit_compare(slang_emit_info *emitInfo, slang_ir_node *n) n->Children[0]->Store, n->Children[1]->Store, NULL); + if (!inst) { + return NULL; + } inst_comment(inst, "Compare values"); /* Compute val = DOT(temp, temp) (reduction) */ @@ -957,6 +963,9 @@ emit_compare(slang_emit_info *emitInfo, slang_ir_node *n) &tempStore, &tempStore, NULL); + if (!inst) { + return NULL; + } inst->SrcReg[0].Swizzle = inst->SrcReg[1].Swizzle = swizzle; /*override*/ inst_comment(inst, "Reduce vec to bool"); @@ -972,6 +981,9 @@ emit_compare(slang_emit_info *emitInfo, slang_ir_node *n) n->Store, &zero, NULL); + if (!inst) { + return NULL; + } inst_comment(inst, "Invert true/false"); } } @@ -1001,6 +1013,9 @@ emit_compare(slang_emit_info *emitInfo, slang_ir_node *n) &srcStore0, &srcStore1, NULL); + if (!inst) { + return NULL; + } inst_comment(inst, "Begin struct/array comparison"); } else { @@ -1010,12 +1025,18 @@ emit_compare(slang_emit_info *emitInfo, slang_ir_node *n) &srcStore0, &srcStore1, NULL); + if (!inst) { + return NULL; + } /* ADD accTemp, accTemp, sneTemp; # like logical-OR */ inst = emit_instruction(emitInfo, OPCODE_ADD, &accTemp, /* dest */ &accTemp, &sneTemp, NULL); + if (!inst) { + return NULL; + } } } @@ -1025,6 +1046,9 @@ emit_compare(slang_emit_info *emitInfo, slang_ir_node *n) &accTemp, &accTemp, NULL); + if (!inst) { + return NULL; + } inst_comment(inst, "End struct/array comparison"); if (n->Opcode == IR_EQUAL) { @@ -1036,6 +1060,9 @@ emit_compare(slang_emit_info *emitInfo, slang_ir_node *n) n->Store, &zero, NULL); + if (!inst) { + return NULL; + } inst_comment(inst, "Invert true/false"); } @@ -1119,6 +1146,9 @@ emit_clamp(slang_emit_info *emitInfo, slang_ir_node *n) n->Children[0]->Store, n->Children[1]->Store, NULL); + if (!inst) { + return NULL; + } /* n->dest = min(tmp, ch[2]) */ inst = emit_instruction(emitInfo, OPCODE_MIN, @@ -1153,7 +1183,9 @@ emit_negation(slang_emit_info *emitInfo, slang_ir_node *n) n->Children[0]->Store, NULL, NULL); - inst->SrcReg[0].Negate = NEGATE_XYZW; + if (inst) { + inst->SrcReg[0].Negate = NEGATE_XYZW; + } return inst; } @@ -1356,6 +1388,9 @@ emit_tex(slang_emit_info *emitInfo, slang_ir_node *n) n->Children[1]->Store, NULL, NULL); + if (!inst) { + return NULL; + } inst->TexShadow = shadow; @@ -1458,6 +1493,9 @@ emit_copy(slang_emit_info *emitInfo, slang_ir_node *n) &srcStore, NULL, NULL); + if (!inst) { + return NULL; + } inst_comment(inst, "IR_COPY block"); srcStore.Index++; dstStore.Index++; @@ -1473,6 +1511,9 @@ emit_copy(slang_emit_info *emitInfo, slang_ir_node *n) n->Children[1]->Store, NULL, NULL); + if (!inst) { + return NULL; + } dstAnnot = storage_annotation(n->Children[0], emitInfo->prog); srcAnnot = storage_annotation(n->Children[1], emitInfo->prog); inst->Comment = instruction_annotation(inst->Opcode, dstAnnot, @@ -1534,6 +1575,9 @@ emit_cond(slang_emit_info *emitInfo, slang_ir_node *n) n->Children[0]->Store, NULL, NULL); + if (!inst) { + return NULL; + } inst->CondUpdate = GL_TRUE; inst_comment(inst, "COND expr"); _slang_free_temp(emitInfo->vt, n->Store); @@ -1596,6 +1640,9 @@ emit_not(slang_emit_info *emitInfo, slang_ir_node *n) n->Children[0]->Store, &zero, NULL); + if (!inst) { + return NULL; + } inst_comment(inst, "NOT"); free_node_storage(emitInfo->vt, n->Children[0]); @@ -1646,12 +1693,17 @@ emit_if(slang_emit_info *emitInfo, slang_ir_node *n) ifInst->DstReg.CondSwizzle = writemask_to_swizzle(condWritemask); } else { + struct prog_instruction *inst; + /* IF src[0] THEN ... */ - emit_instruction(emitInfo, OPCODE_IF, - NULL, /* dst */ - n->Children[0]->Store, /* op0 */ - NULL, - NULL); + inst = emit_instruction(emitInfo, OPCODE_IF, + NULL, /* dst */ + n->Children[0]->Store, /* op0 */ + NULL, + NULL); + if (!inst) { + return NULL; + } } } else { @@ -1875,6 +1927,9 @@ emit_cont_break_if_true(slang_emit_info *emitInfo, slang_ir_node *n) n->Children[0]->Store, NULL, NULL); + if (!inst) { + return NULL; + } n->InstLocation = emitInfo->prog->NumInstructions; inst = new_instruction(emitInfo, opcode); @@ -2045,6 +2100,9 @@ emit_array_element(slang_emit_info *emitInfo, slang_ir_node *n) indexStore, /* the index */ &elemSizeStore, NULL); + if (!inst) { + return NULL; + } indexStore = indexTemp; } @@ -2071,6 +2129,9 @@ emit_array_element(slang_emit_info *emitInfo, slang_ir_node *n) indexStore, /* the index */ &indirectArray, /* indirect array base */ NULL); + if (!inst) { + return NULL; + } indexStore = indexTemp; } diff --git a/src/mesa/shader/slang/slang_link.c b/src/mesa/shader/slang/slang_link.c index 0a2bc497803..ed27821a951 100644 --- a/src/mesa/shader/slang/slang_link.c +++ b/src/mesa/shader/slang/slang_link.c @@ -590,11 +590,16 @@ concat_shaders(struct gl_shader_program *shProg, GLenum shaderType) { struct gl_shader *newShader; const struct gl_shader *firstShader = NULL; - GLuint shaderLengths[100]; + GLuint *shaderLengths; GLchar *source; GLuint totalLen = 0, len = 0; GLuint i; + shaderLengths = (GLuint *)_mesa_malloc(shProg->NumShaders * sizeof(GLuint)); + if (!shaderLengths) { + return NULL; + } + /* compute total size of new shader source code */ for (i = 0; i < shProg->NumShaders; i++) { const struct gl_shader *shader = shProg->Shaders[i]; @@ -606,12 +611,16 @@ concat_shaders(struct gl_shader_program *shProg, GLenum shaderType) } } - if (totalLen == 0) + if (totalLen == 0) { + _mesa_free(shaderLengths); return NULL; + } source = (GLchar *) _mesa_malloc(totalLen + 1); - if (!source) + if (!source) { + _mesa_free(shaderLengths); return NULL; + } /* concatenate shaders */ for (i = 0; i < shProg->NumShaders; i++) { @@ -626,9 +635,16 @@ concat_shaders(struct gl_shader_program *shProg, GLenum shaderType) _mesa_printf("---NEW CONCATENATED SHADER---:\n%s\n------------\n", source); */ + _mesa_free(shaderLengths); + remove_extra_version_directives(source); newShader = CALLOC_STRUCT(gl_shader); + if (!newShader) { + _mesa_free(source); + return NULL; + } + newShader->Type = shaderType; newShader->Source = source; newShader->Pragmas = firstShader->Pragmas; diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index 0e89a624c4e..73df44d1988 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -46,7 +46,8 @@ static const struct st_tracked_state *atoms[] = &st_update_clip, &st_finalize_textures, - &st_update_shader, + &st_update_fp, + &st_update_vp, &st_update_rasterizer, &st_update_polygon_stipple, diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h index c7cffd85c8a..f34b49203b2 100644 --- a/src/mesa/state_tracker/st_atom.h +++ b/src/mesa/state_tracker/st_atom.h @@ -47,7 +47,8 @@ void st_validate_state( struct st_context *st ); extern const struct st_tracked_state st_update_framebuffer; extern const struct st_tracked_state st_update_clip; extern const struct st_tracked_state st_update_depth_stencil_alpha; -extern const struct st_tracked_state st_update_shader; +extern const struct st_tracked_state st_update_fp; +extern const struct st_tracked_state st_update_vp; extern const struct st_tracked_state st_update_rasterizer; extern const struct st_tracked_state st_update_polygon_stipple; extern const struct st_tracked_state st_update_viewport; diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c index e18c0f6e0a1..8ca4335e330 100644 --- a/src/mesa/state_tracker/st_atom_framebuffer.c +++ b/src/mesa/state_tracker/st_atom_framebuffer.c @@ -40,6 +40,7 @@ #include "pipe/p_inlines.h" #include "cso_cache/cso_context.h" #include "util/u_rect.h" +#include "util/u_math.h" @@ -64,8 +65,8 @@ update_renderbuffer_surface(struct st_context *st, GLuint level; /* find matching mipmap level size */ for (level = 0; level <= texture->last_level; level++) { - if (texture->width[level] == rtt_width && - texture->height[level] == rtt_height) { + if (u_minify(texture->width0, level) == rtt_width && + u_minify(texture->height0, level) == rtt_height) { pipe_surface_reference(&strb->surface, NULL); diff --git a/src/mesa/state_tracker/st_atom_pixeltransfer.c b/src/mesa/state_tracker/st_atom_pixeltransfer.c index babfcc87b4a..4b35f59cc2f 100644 --- a/src/mesa/state_tracker/st_atom_pixeltransfer.c +++ b/src/mesa/state_tracker/st_atom_pixeltransfer.c @@ -145,7 +145,7 @@ load_color_map_texture(GLcontext *ctx, struct pipe_texture *pt) const GLuint gSize = ctx->PixelMaps.GtoG.Size; const GLuint bSize = ctx->PixelMaps.BtoB.Size; const GLuint aSize = ctx->PixelMaps.AtoA.Size; - const uint texSize = pt->width[0]; + const uint texSize = pt->width0; uint *dest; uint i, j; diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index 6611956ae83..d6e3a3e5610 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -229,14 +229,23 @@ update_samplers(struct st_context *st) /*printf("%s su=%u non-null\n", __FUNCTION__, su);*/ cso_single_sampler(st->cso_context, su, sampler); + if (su < st->ctx->Const.MaxVertexTextureImageUnits) { + cso_single_vertex_sampler(st->cso_context, su, sampler); + } } else { /*printf("%s su=%u null\n", __FUNCTION__, su);*/ cso_single_sampler(st->cso_context, su, NULL); + if (su < st->ctx->Const.MaxVertexTextureImageUnits) { + cso_single_vertex_sampler(st->cso_context, su, NULL); + } } } cso_single_sampler_done(st->cso_context); + if (st->ctx->Const.MaxVertexTextureImageUnits > 0) { + cso_single_vertex_sampler_done(st->cso_context); + } } diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index 6e311e537ef..09baff875bc 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -56,82 +56,18 @@ #include "st_mesa_to_tgsi.h" -/** - * This represents a vertex program, especially translated to match - * the inputs of a particular fragment shader. - */ -struct translated_vertex_program -{ - struct st_vertex_program *master; - - /** The fragment shader "signature" this vertex shader is meant for: */ - GLbitfield frag_inputs; - /** Compared against master vertex program's serialNo: */ - GLuint serialNo; - /** Maps VERT_RESULT_x to slot */ - GLuint output_to_slot[VERT_RESULT_MAX]; - ubyte output_to_semantic_name[VERT_RESULT_MAX]; - ubyte output_to_semantic_index[VERT_RESULT_MAX]; - - /** Pointer to the translated vertex program */ - struct st_vertex_program *vp; - - struct translated_vertex_program *next; /**< next in linked list */ -}; - - -/** - * Given a vertex program output attribute, return the corresponding - * fragment program input attribute. - * \return -1 for vertex outputs that have no corresponding fragment input +/* + * Translate fragment program if needed. */ -static GLint -vp_out_to_fp_in(GLuint vertResult) -{ - if (vertResult >= VERT_RESULT_TEX0 && - vertResult < VERT_RESULT_TEX0 + MAX_TEXTURE_COORD_UNITS) - return FRAG_ATTRIB_TEX0 + (vertResult - VERT_RESULT_TEX0); - - if (vertResult >= VERT_RESULT_VAR0 && - vertResult < VERT_RESULT_VAR0 + MAX_VARYING) - return FRAG_ATTRIB_VAR0 + (vertResult - VERT_RESULT_VAR0); - - switch (vertResult) { - case VERT_RESULT_HPOS: - return FRAG_ATTRIB_WPOS; - case VERT_RESULT_COL0: - return FRAG_ATTRIB_COL0; - case VERT_RESULT_COL1: - return FRAG_ATTRIB_COL1; - case VERT_RESULT_FOGC: - return FRAG_ATTRIB_FOGC; - default: - /* Back-face colors, edge flags, etc */ - return -1; - } -} - - -/** - * Find a translated vertex program that corresponds to stvp and - * has outputs matched to stfp's inputs. - * This performs vertex and fragment translation (to TGSI) when needed. - */ -static struct translated_vertex_program * -find_translated_vp(struct st_context *st, - struct st_vertex_program *stvp, - struct st_fragment_program *stfp) +static void +translate_fp(struct st_context *st, + struct st_fragment_program *stfp) { - static const GLuint UNUSED = ~0; - struct translated_vertex_program *xvp; const GLbitfield fragInputsRead = stfp->Base.Base.InputsRead; - /* - * Translate fragment program if needed. - */ if (!stfp->state.tokens) { GLuint inAttr, numIn = 0; @@ -141,7 +77,7 @@ find_translated_vp(struct st_context *st, numIn++; } else { - stfp->input_to_slot[inAttr] = UNUSED; + stfp->input_to_slot[inAttr] = -1; } } @@ -151,170 +87,63 @@ find_translated_vp(struct st_context *st, st_translate_fragment_program(st, stfp, stfp->input_to_slot); } +} - /* See if we've got a translated vertex program whose outputs match - * the fragment program's inputs. - * XXX This could be a hash lookup, using InputsRead as the key. - */ - for (xvp = stfp->vertex_programs; xvp; xvp = xvp->next) { - if (xvp->master == stvp && xvp->frag_inputs == fragInputsRead) { - break; - } - } - /* No? Allocate translated vp object now */ - if (!xvp) { - xvp = ST_CALLOC_STRUCT(translated_vertex_program); - xvp->frag_inputs = fragInputsRead; - xvp->master = stvp; +/** + * Find a translated vertex program that corresponds to stvp and + * has outputs matched to stfp's inputs. + * This performs vertex and fragment translation (to TGSI) when needed. + */ +static struct st_vp_varient * +find_translated_vp(struct st_context *st, + struct st_vertex_program *stvp ) +{ + struct st_vp_varient *vpv; + struct st_vp_varient_key key; - xvp->next = stfp->vertex_programs; - stfp->vertex_programs = xvp; - } + /* Nothing in our key yet. This will change: + */ + memset(&key, 0, sizeof key); + key.dummy = 0; - /* See if we need to translate vertex program to TGSI form */ - if (xvp->serialNo != stvp->serialNo) { - GLuint outAttr; - const GLbitfield64 outputsWritten = stvp->Base.Base.OutputsWritten; - GLuint numVpOuts = 0; - GLboolean emitPntSize = GL_FALSE, emitBFC0 = GL_FALSE, emitBFC1 = GL_FALSE; - GLbitfield usedGenerics = 0x0; - GLbitfield usedOutputSlots = 0x0; - - /* Compute mapping of vertex program outputs to slots, which depends - * on the fragment program's input->slot mapping. + /* Do we need to throw away old translations after a change in the + * GL program string? + */ + if (stvp->serialNo != stvp->lastSerialNo) { + /* These may have changed if the program string changed. */ - for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) { - /* set defaults: */ - xvp->output_to_slot[outAttr] = UNUSED; - xvp->output_to_semantic_name[outAttr] = TGSI_SEMANTIC_COUNT; - xvp->output_to_semantic_index[outAttr] = 99; - - if (outAttr == VERT_RESULT_HPOS) { - /* always put xformed position into slot zero */ - GLuint slot = 0; - xvp->output_to_slot[VERT_RESULT_HPOS] = slot; - xvp->output_to_semantic_name[outAttr] = TGSI_SEMANTIC_POSITION; - xvp->output_to_semantic_index[outAttr] = 0; - numVpOuts++; - usedOutputSlots |= (1 << slot); - } - else if (outputsWritten & (1 << outAttr)) { - /* see if the frag prog wants this vert output */ - GLint fpInAttrib = vp_out_to_fp_in(outAttr); - if (fpInAttrib >= 0) { - GLuint fpInSlot = stfp->input_to_slot[fpInAttrib]; - if (fpInSlot != ~0) { - /* match this vp output to the fp input */ - GLuint vpOutSlot = stfp->input_map[fpInSlot]; - xvp->output_to_slot[outAttr] = vpOutSlot; - xvp->output_to_semantic_name[outAttr] = stfp->input_semantic_name[fpInSlot]; - xvp->output_to_semantic_index[outAttr] = stfp->input_semantic_index[fpInSlot]; - numVpOuts++; - usedOutputSlots |= (1 << vpOutSlot); - } - else { -#if 0 /*debug*/ - printf("VP output %d not used by FP\n", outAttr); -#endif - } - } - else if (outAttr == VERT_RESULT_PSIZ) - emitPntSize = GL_TRUE; - else if (outAttr == VERT_RESULT_BFC0) - emitBFC0 = GL_TRUE; - else if (outAttr == VERT_RESULT_BFC1) - emitBFC1 = GL_TRUE; - } -#if 0 /*debug*/ - printf("assign vp output_to_slot[%d] = %d\n", outAttr, - xvp->output_to_slot[outAttr]); -#endif - } - - /* must do these last */ - if (emitPntSize) { - GLuint slot = numVpOuts++; - xvp->output_to_slot[VERT_RESULT_PSIZ] = slot; - xvp->output_to_semantic_name[VERT_RESULT_PSIZ] = TGSI_SEMANTIC_PSIZE; - xvp->output_to_semantic_index[VERT_RESULT_PSIZ] = 0; - usedOutputSlots |= (1 << slot); - } - if (emitBFC0) { - GLuint slot = numVpOuts++; - xvp->output_to_slot[VERT_RESULT_BFC0] = slot; - xvp->output_to_semantic_name[VERT_RESULT_BFC0] = TGSI_SEMANTIC_COLOR; - xvp->output_to_semantic_index[VERT_RESULT_BFC0] = 0; - usedOutputSlots |= (1 << slot); - } - if (emitBFC1) { - GLuint slot = numVpOuts++; - xvp->output_to_slot[VERT_RESULT_BFC1] = slot; - xvp->output_to_semantic_name[VERT_RESULT_BFC1] = TGSI_SEMANTIC_COLOR; - xvp->output_to_semantic_index[VERT_RESULT_BFC1] = 1; - usedOutputSlots |= (1 << slot); - } - - /* build usedGenerics mask */ - usedGenerics = 0x0; - for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) { - if (xvp->output_to_semantic_name[outAttr] == TGSI_SEMANTIC_GENERIC) { - usedGenerics |= (1 << xvp->output_to_semantic_index[outAttr]); - } - } + st_prepare_vertex_program( st, stvp ); - /* For each vertex program output that doesn't match up to a fragment - * program input, map the vertex program output to a free slot and - * free generic attribute. + /* We are now up-to-date: */ - for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) { - if (outputsWritten & (1 << outAttr)) { - if (xvp->output_to_slot[outAttr] == UNUSED) { - GLint freeGeneric = _mesa_ffs(~usedGenerics) - 1; - GLint freeSlot = _mesa_ffs(~usedOutputSlots) - 1; - usedGenerics |= (1 << freeGeneric); - usedOutputSlots |= (1 << freeSlot); - xvp->output_to_slot[outAttr] = freeSlot; - xvp->output_to_semantic_name[outAttr] = TGSI_SEMANTIC_GENERIC; - xvp->output_to_semantic_index[outAttr] = freeGeneric; - } - } - -#if 0 /*debug*/ - printf("vp output_to_slot[%d] = %d\n", outAttr, - xvp->output_to_slot[outAttr]); -#endif + stvp->lastSerialNo = stvp->serialNo; + } + + /* See if we've got a translated vertex program whose outputs match + * the fragment program's inputs. + */ + for (vpv = stvp->varients; vpv; vpv = vpv->next) { + if (memcmp(&vpv->key, &key, sizeof key) == 0) { + break; } + } - assert(stvp->Base.Base.NumInstructions > 1); - - st_translate_vertex_program(st, stvp, xvp->output_to_slot, - xvp->output_to_semantic_name, - xvp->output_to_semantic_index); - - xvp->vp = stvp; - - /* translated VP is up to date now */ - xvp->serialNo = stvp->serialNo; + /* No? Perform new translation here. */ + if (!vpv) { + vpv = st_translate_vertex_program(st, stvp, &key); + if (!vpv) + return NULL; + + vpv->next = stvp->varients; + stvp->varients = vpv; } - return xvp; + return vpv; } -void -st_free_translated_vertex_programs(struct st_context *st, - struct translated_vertex_program *xvp) -{ - struct translated_vertex_program *next; - - while (xvp) { - next = xvp->next; - _mesa_free(xvp); - xvp = next; - } -} static void * @@ -328,32 +157,19 @@ get_passthrough_fs(struct st_context *st) return st->passthrough_fs; } - static void -update_linkage( struct st_context *st ) +update_fp( struct st_context *st ) { - struct st_vertex_program *stvp; struct st_fragment_program *stfp; - struct translated_vertex_program *xvp; - - /* find active shader and params -- Should be covered by - * ST_NEW_VERTEX_PROGRAM - */ - assert(st->ctx->VertexProgram._Current); - stvp = st_vertex_program(st->ctx->VertexProgram._Current); - assert(stvp->Base.Base.Target == GL_VERTEX_PROGRAM_ARB); assert(st->ctx->FragmentProgram._Current); stfp = st_fragment_program(st->ctx->FragmentProgram._Current); assert(stfp->Base.Base.Target == GL_FRAGMENT_PROGRAM_ARB); - xvp = find_translated_vp(st, stvp, stfp); + translate_fp(st, stfp); - st_reference_vertprog(st, &st->vp, stvp); st_reference_fragprog(st, &st->fp, stfp); - cso_set_vertex_shader_handle(st->cso_context, stvp->driver_shader); - if (st->missing_textures) { /* use a pass-through frag shader that uses no textures */ void *fs = get_passthrough_fs(st); @@ -362,16 +178,48 @@ update_linkage( struct st_context *st ) else { cso_set_fragment_shader_handle(st->cso_context, stfp->driver_shader); } +} + +const struct st_tracked_state st_update_fp = { + "st_update_fp", /* name */ + { /* dirty */ + 0, /* mesa */ + ST_NEW_FRAGMENT_PROGRAM /* st */ + }, + update_fp /* update */ +}; + + + + +static void +update_vp( struct st_context *st ) +{ + struct st_vertex_program *stvp; + + /* find active shader and params -- Should be covered by + * ST_NEW_VERTEX_PROGRAM + */ + assert(st->ctx->VertexProgram._Current); + stvp = st_vertex_program(st->ctx->VertexProgram._Current); + assert(stvp->Base.Base.Target == GL_VERTEX_PROGRAM_ARB); + + st->vp_varient = find_translated_vp(st, stvp); + + st_reference_vertprog(st, &st->vp, stvp); + + cso_set_vertex_shader_handle(st->cso_context, + st->vp_varient->driver_shader); - st->vertex_result_to_slot = xvp->output_to_slot; + st->vertex_result_to_slot = stvp->result_to_output; } -const struct st_tracked_state st_update_shader = { - "st_update_shader", /* name */ +const struct st_tracked_state st_update_vp = { + "st_update_vp", /* name */ { /* dirty */ 0, /* mesa */ - ST_NEW_VERTEX_PROGRAM | ST_NEW_FRAGMENT_PROGRAM /* st */ + ST_NEW_VERTEX_PROGRAM /* st */ }, - update_linkage /* update */ + update_vp /* update */ }; diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c index 4d4f97da7ed..0b68447d212 100644 --- a/src/mesa/state_tracker/st_atom_texture.c +++ b/src/mesa/state_tracker/st_atom_texture.c @@ -32,6 +32,8 @@ */ +#include "main/macros.h" + #include "st_context.h" #include "st_atom.h" #include "st_texture.h" @@ -99,6 +101,12 @@ update_textures(struct st_context *st) cso_set_sampler_textures(st->cso_context, st->state.num_textures, st->state.sampler_texture); + if (st->ctx->Const.MaxVertexTextureImageUnits > 0) { + cso_set_vertex_sampler_textures(st->cso_context, + MIN2(st->state.num_textures, + st->ctx->Const.MaxVertexTextureImageUnits), + st->state.sampler_texture); + } } diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 1960d171bf6..1bdeaccda3e 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -169,11 +169,6 @@ make_bitmap_fragment_program(GLcontext *ctx, GLuint samplerIndex) stfp = (struct st_fragment_program *) p; stfp->Base.UsesKill = GL_TRUE; - /* No need to send this incomplete program down to hardware: - * - * st_translate_fragment_program(ctx->st, stfp, NULL); - */ - return stfp; } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 1d33e81c2c9..03617b7a939 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -60,6 +60,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" +#include "tgsi/tgsi_ureg.h" #include "util/u_tile.h" #include "util/u_draw_quad.h" #include "util/u_math.h" @@ -236,78 +237,41 @@ make_fragment_shader_z(struct st_context *st) * Create a simple vertex shader that just passes through the * vertex position and texcoord (and optionally, color). */ -static struct st_vertex_program * -st_make_passthrough_vertex_shader(struct st_context *st, GLboolean passColor) +static void * +st_make_passthrough_vertex_shader(struct st_context *st, + GLboolean passColor) { - GLcontext *ctx = st->ctx; - struct st_vertex_program *stvp; - struct gl_program *p; - GLuint ic = 0; - - if (st->drawpix.vert_shaders[passColor]) - return st->drawpix.vert_shaders[passColor]; - - /* - * Create shader now - */ - p = ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0); - if (!p) - return NULL; - - if (passColor) - p->NumInstructions = 4; - else - p->NumInstructions = 3; - - p->Instructions = _mesa_alloc_instructions(p->NumInstructions); - if (!p->Instructions) { - ctx->Driver.DeleteProgram(ctx, p); - return NULL; - } - _mesa_init_instructions(p->Instructions, p->NumInstructions); - /* MOV result.pos, vertex.pos; */ - p->Instructions[0].Opcode = OPCODE_MOV; - p->Instructions[0].DstReg.File = PROGRAM_OUTPUT; - p->Instructions[0].DstReg.Index = VERT_RESULT_HPOS; - p->Instructions[0].SrcReg[0].File = PROGRAM_INPUT; - p->Instructions[0].SrcReg[0].Index = VERT_ATTRIB_POS; - /* MOV result.texcoord0, vertex.texcoord0; */ - p->Instructions[1].Opcode = OPCODE_MOV; - p->Instructions[1].DstReg.File = PROGRAM_OUTPUT; - p->Instructions[1].DstReg.Index = VERT_RESULT_TEX0; - p->Instructions[1].SrcReg[0].File = PROGRAM_INPUT; - p->Instructions[1].SrcReg[0].Index = VERT_ATTRIB_TEX0; - ic = 2; - if (passColor) { - /* MOV result.color0, vertex.color0; */ - p->Instructions[ic].Opcode = OPCODE_MOV; - p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT; - p->Instructions[ic].DstReg.Index = VERT_RESULT_COL0; - p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT; - p->Instructions[ic].SrcReg[0].Index = VERT_ATTRIB_COLOR0; - ic++; - } - - /* END; */ - p->Instructions[ic].Opcode = OPCODE_END; - ic++; - - assert(ic == p->NumInstructions); + if (!st->drawpix.vert_shaders[passColor]) { + struct ureg_program *ureg = + ureg_create( TGSI_PROCESSOR_VERTEX ); + + if (ureg == NULL) + return NULL; + + /* MOV result.pos, vertex.pos; */ + ureg_MOV(ureg, + ureg_DECL_output( ureg, TGSI_SEMANTIC_POSITION, 0 ), + ureg_DECL_vs_input( ureg, 0 )); + + /* MOV result.texcoord0, vertex.texcoord0; */ + ureg_MOV(ureg, + ureg_DECL_output( ureg, TGSI_SEMANTIC_GENERIC, 0 ), + ureg_DECL_vs_input( ureg, 1 )); + + if (passColor) { + /* MOV result.color0, vertex.color0; */ + ureg_MOV(ureg, + ureg_DECL_output( ureg, TGSI_SEMANTIC_COLOR, 0 ), + ureg_DECL_vs_input( ureg, 2 )); + } - p->InputsRead = VERT_BIT_POS | VERT_BIT_TEX0; - p->OutputsWritten = ((1 << VERT_RESULT_TEX0) | - (1 << VERT_RESULT_HPOS)); - if (passColor) { - p->InputsRead |= VERT_BIT_COLOR0; - p->OutputsWritten |= (1 << VERT_RESULT_COL0); + ureg_END( ureg ); + + st->drawpix.vert_shaders[passColor] = + ureg_create_shader_and_destroy( ureg, st->pipe ); } - stvp = (struct st_vertex_program *) p; - st_translate_vertex_program(st, stvp, NULL, NULL, NULL); - - st->drawpix.vert_shaders[passColor] = stvp; - - return stvp; + return st->drawpix.vert_shaders[passColor]; } @@ -539,8 +503,8 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, GLsizei width, GLsizei height, GLfloat zoomX, GLfloat zoomY, struct pipe_texture *pt, - struct st_vertex_program *stvp, - struct st_fragment_program *stfp, + void *driver_vp, + void *driver_fp, const GLfloat *color, GLboolean invertTex) { @@ -575,10 +539,10 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, } /* fragment shader state: TEX lookup program */ - cso_set_fragment_shader_handle(cso, stfp->driver_shader); + cso_set_fragment_shader_handle(cso, driver_fp); /* vertex shader state: position + texcoord pass-through */ - cso_set_vertex_shader_handle(cso, stvp->driver_shader); + cso_set_vertex_shader_handle(cso, driver_vp); /* texture sampling state: */ @@ -621,10 +585,10 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, struct pipe_texture *textures[2]; textures[0] = pt; textures[1] = st->pixel_xfer.pixelmap_texture; - pipe->set_sampler_textures(pipe, 2, textures); + pipe->set_fragment_sampler_textures(pipe, 2, textures); } else { - pipe->set_sampler_textures(pipe, 1, &pt); + pipe->set_fragment_sampler_textures(pipe, 1, &pt); } /* Compute window coords (y=0=bottom) with pixel zoom. @@ -637,8 +601,8 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, y1 = y + height * ctx->Pixel.ZoomY; draw_quad(ctx, x0, y0, z, x1, y1, color, invertTex, - (GLfloat) width / pt->width[0], - (GLfloat) height / pt->height[0]); + (GLfloat) width / pt->width0, + (GLfloat) height / pt->height0); /* restore state */ cso_restore_rasterizer(cso); @@ -806,7 +770,7 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, const struct gl_pixelstore_attrib *unpack, const GLvoid *pixels) { struct st_fragment_program *stfp; - struct st_vertex_program *stvp; + void *driver_vp; struct st_context *st = st_context(ctx); struct pipe_surface *ps; const GLfloat *color; @@ -826,13 +790,13 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, if (format == GL_DEPTH_COMPONENT) { ps = st->state.framebuffer.zsbuf; stfp = make_fragment_shader_z(st); - stvp = st_make_passthrough_vertex_shader(st, GL_TRUE); + driver_vp = st_make_passthrough_vertex_shader(st, GL_TRUE); color = ctx->Current.RasterColor; } else { ps = st->state.framebuffer.cbufs[0]; stfp = combined_drawpix_fragment_program(ctx); - stvp = st_make_passthrough_vertex_shader(st, GL_FALSE); + driver_vp = st_make_passthrough_vertex_shader(st, GL_FALSE); color = NULL; } @@ -843,7 +807,10 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, if (pt) { draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2], width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY, - pt, stvp, stfp, color, GL_FALSE); + pt, + driver_vp, + stfp->driver_shader, + color, GL_FALSE); pipe_texture_reference(&pt, NULL); } } @@ -960,7 +927,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, struct pipe_context *pipe = st->pipe; struct pipe_screen *screen = pipe->screen; struct st_renderbuffer *rbRead; - struct st_vertex_program *stvp; + void *driver_vp; struct st_fragment_program *stfp; struct pipe_texture *pt; GLfloat *color; @@ -1009,14 +976,14 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, rbRead = st_get_color_read_renderbuffer(ctx); color = NULL; stfp = combined_drawpix_fragment_program(ctx); - stvp = st_make_passthrough_vertex_shader(st, GL_FALSE); + driver_vp = st_make_passthrough_vertex_shader(st, GL_FALSE); } else { assert(type == GL_DEPTH); rbRead = st_renderbuffer(ctx->ReadBuffer->_DepthBuffer); color = ctx->Current.Attrib[VERT_ATTRIB_COLOR0]; stfp = make_fragment_shader_z(st); - stvp = st_make_passthrough_vertex_shader(st, GL_TRUE); + driver_vp = st_make_passthrough_vertex_shader(st, GL_TRUE); } srcFormat = rbRead->texture->format; @@ -1148,7 +1115,10 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, /* draw textured quad */ draw_textured_quad(ctx, dstx, dsty, ctx->Current.RasterPos[2], width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY, - pt, stvp, stfp, color, GL_TRUE); + pt, + driver_vp, + stfp->driver_shader, + color, GL_TRUE); pipe_texture_reference(&pt, NULL); } diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 65ce12ccd4f..659a6c91938 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -128,9 +128,9 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, template.target = PIPE_TEXTURE_2D; template.format = format; pf_get_block(format, &template.block); - template.width[0] = width; - template.height[0] = height; - template.depth[0] = 1; + template.width0 = width; + template.height0 = height; + template.depth0 = 1; template.last_level = 0; template.nr_samples = rb->NumSamples; if (pf_is_depth_stencil(format)) { @@ -376,7 +376,7 @@ st_render_texture(GLcontext *ctx, rb->_BaseFormat = texImage->_BaseFormat; /*printf("***** render to texture level %d: %d x %d\n", att->TextureLevel, rb->Width, rb->Height);*/ - /*printf("***** pipe texture %d x %d\n", pt->width[0], pt->height[0]);*/ + /*printf("***** pipe texture %d x %d\n", pt->width0, pt->height0);*/ pipe_texture_reference( &strb->texture, pt ); diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index b2d5c39a3a0..8c276f8128e 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -138,24 +138,7 @@ st_delete_program(GLcontext *ctx, struct gl_program *prog) case GL_VERTEX_PROGRAM_ARB: { struct st_vertex_program *stvp = (struct st_vertex_program *) prog; - - if (stvp->driver_shader) { - cso_delete_vertex_shader(st->cso_context, stvp->driver_shader); - stvp->driver_shader = NULL; - } - - if (stvp->draw_shader) { -#if FEATURE_feedback || FEATURE_drawpix - /* this would only have been allocated for the RasterPos path */ - draw_delete_vertex_shader(st->draw, stvp->draw_shader); - stvp->draw_shader = NULL; -#endif - } - - if (stvp->state.tokens) { - st_free_tokens(stvp->state.tokens); - stvp->state.tokens = NULL; - } + st_vp_release_varients( st, stvp ); } break; case GL_FRAGMENT_PROGRAM_ARB: @@ -177,8 +160,6 @@ st_delete_program(GLcontext *ctx, struct gl_program *prog) _mesa_reference_program(ctx, &prg, NULL); stfp->bitmap_program = NULL; } - - st_free_translated_vertex_programs(st, stfp->vertex_programs); } break; default: @@ -219,8 +200,6 @@ static void st_program_string_notify( GLcontext *ctx, stfp->state.tokens = NULL; } - stfp->param_state = stfp->Base.Base.Parameters->StateFlags; - if (st->fp == stfp) st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM; } @@ -229,25 +208,7 @@ static void st_program_string_notify( GLcontext *ctx, stvp->serialNo++; - if (stvp->driver_shader) { - cso_delete_vertex_shader(st->cso_context, stvp->driver_shader); - stvp->driver_shader = NULL; - } - - if (stvp->draw_shader) { -#if FEATURE_feedback || FEATURE_drawpix - /* this would only have been allocated for the RasterPos path */ - draw_delete_vertex_shader(st->draw, stvp->draw_shader); - stvp->draw_shader = NULL; -#endif - } - - if (stvp->state.tokens) { - st_free_tokens(stvp->state.tokens); - stvp->state.tokens = NULL; - } - - stvp->param_state = stvp->Base.Base.Parameters->StateFlags; + st_vp_release_varients( st, stvp ); if (st->vp == stvp) st->dirty.st |= ST_NEW_VERTEX_PROGRAM; diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index 772bb3bb69e..103861d6f9c 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -243,7 +243,7 @@ st_fast_readpixels(GLcontext *ctx, struct st_renderbuffer *strb, GLint row, col, dy, dstStride; if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) { - y = strb->texture->height[0] - y - height; + y = strb->texture->height0 - y - height; } trans = st_cond_flush_get_tex_transfer(st_context(ctx), strb->texture, diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index d4630a514f8..3a2337802fa 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -406,9 +406,9 @@ compress_with_blit(GLcontext * ctx, templ.target = PIPE_TEXTURE_2D; templ.format = st_mesa_format_to_pipe_format(mesa_format); pf_get_block(templ.format, &templ.block); - templ.width[0] = width; - templ.height[0] = height; - templ.depth[0] = 1; + templ.width0 = width; + templ.height0 = height; + templ.depth0 = 1; templ.last_level = 0; templ.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER; src_tex = screen->texture_create(screen, &templ); @@ -1761,9 +1761,9 @@ st_finalize_texture(GLcontext *ctx, if (stObj->pt->target != gl_target_to_pipe(stObj->base.Target) || stObj->pt->format != fmt || stObj->pt->last_level < stObj->lastLevel || - stObj->pt->width[0] != firstImage->base.Width2 || - stObj->pt->height[0] != firstImage->base.Height2 || - stObj->pt->depth[0] != firstImage->base.Depth2 || + stObj->pt->width0 != firstImage->base.Width2 || + stObj->pt->height0 != firstImage->base.Height2 || + stObj->pt->depth0 != firstImage->base.Depth2 || stObj->pt->block.size != blockSize) { pipe_texture_reference(&stObj->pt, NULL); diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 18adb35e872..b7607286583 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -127,6 +127,8 @@ struct st_context struct st_vertex_program *vp; /**< Currently bound vertex program */ struct st_fragment_program *fp; /**< Currently bound fragment program */ + struct st_vp_varient *vp_varient; + struct gl_texture_object *default_texture; struct { diff --git a/src/mesa/state_tracker/st_debug.c b/src/mesa/state_tracker/st_debug.c index 3009cde9d51..6e699ca5522 100644 --- a/src/mesa/state_tracker/st_debug.c +++ b/src/mesa/state_tracker/st_debug.c @@ -86,7 +86,8 @@ st_print_current(void) } #endif - tgsi_dump( st->vp->state.tokens, 0 ); + if (st->vp->varients) + tgsi_dump( st->vp->varients[0].state.tokens, 0 ); if (st->vp->Base.Base.Parameters) _mesa_print_parameter_list(st->vp->Base.Base.Parameters); diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 68bc76b5723..e13ae57a0ec 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -573,7 +573,7 @@ st_draw_vbo(GLcontext *ctx, /* must get these after state validation! */ vp = ctx->st->vp; - vs = &ctx->st->vp->state; + vs = &ctx->st->vp_varient->state; #if 0 if (MESA_VERBOSE & VERBOSE_GLSL) { diff --git a/src/mesa/state_tracker/st_draw_feedback.c b/src/mesa/state_tracker/st_draw_feedback.c index b2d682ef640..d793f820bc3 100644 --- a/src/mesa/state_tracker/st_draw_feedback.c +++ b/src/mesa/state_tracker/st_draw_feedback.c @@ -120,10 +120,10 @@ st_feedback_draw_vbo(GLcontext *ctx, /* must get these after state validation! */ vp = ctx->st->vp; - vs = &st->vp->state; + vs = &st->vp_varient->state; - if (!st->vp->draw_shader) { - st->vp->draw_shader = draw_create_vertex_shader(draw, vs); + if (!st->vp_varient->draw_shader) { + st->vp_varient->draw_shader = draw_create_vertex_shader(draw, vs); } /* @@ -136,7 +136,7 @@ st_feedback_draw_vbo(GLcontext *ctx, draw_set_viewport_state(draw, &st->state.viewport); draw_set_clip_state(draw, &st->state.clip); draw_set_rasterizer_state(draw, &st->state.rasterizer); - draw_bind_vertex_shader(draw, st->vp->draw_shader); + draw_bind_vertex_shader(draw, st->vp_varient->draw_shader); set_feedback_vertex_format(ctx); /* loop over TGSI shader inputs to determine vertex buffer diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c index 57fe72d76a7..ef3cbc53ee0 100644 --- a/src/mesa/state_tracker/st_extensions.c +++ b/src/mesa/state_tracker/st_extensions.c @@ -92,6 +92,10 @@ void st_init_limits(struct st_context *st) = _min(screen->get_param(screen, PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS), MAX_VERTEX_TEXTURE_IMAGE_UNITS); + c->MaxCombinedTextureImageUnits + = _min(screen->get_param(screen, PIPE_CAP_MAX_COMBINED_SAMPLERS), + MAX_COMBINED_TEXTURE_IMAGE_UNITS); + c->MaxTextureCoordUnits = _min(c->MaxTextureImageUnits, MAX_TEXTURE_COORD_UNITS); diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index 16ca2771b0f..f8068fa12be 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -38,6 +38,7 @@ #include "pipe/p_defines.h" #include "pipe/p_inlines.h" #include "util/u_gen_mipmap.h" +#include "util/u_math.h" #include "cso_cache/cso_cache.h" #include "cso_cache/cso_context.h" @@ -133,14 +134,14 @@ fallback_generate_mipmap(GLcontext *ctx, GLenum target, srcTrans = st_cond_flush_get_tex_transfer(st_context(ctx), pt, face, srcLevel, zslice, PIPE_TRANSFER_READ, 0, 0, - pt->width[srcLevel], - pt->height[srcLevel]); + u_minify(pt->width0, srcLevel), + u_minify(pt->height0, srcLevel)); dstTrans = st_cond_flush_get_tex_transfer(st_context(ctx), pt, face, dstLevel, zslice, PIPE_TRANSFER_WRITE, 0, 0, - pt->width[dstLevel], - pt->height[dstLevel]); + u_minify(pt->width0, dstLevel), + u_minify(pt->height0, dstLevel)); srcData = (ubyte *) screen->transfer_map(screen, srcTrans); dstData = (ubyte *) screen->transfer_map(screen, dstTrans); @@ -149,13 +150,17 @@ fallback_generate_mipmap(GLcontext *ctx, GLenum target, dstStride = dstTrans->stride / dstTrans->block.size; _mesa_generate_mipmap_level(target, datatype, comps, - 0 /*border*/, - pt->width[srcLevel], pt->height[srcLevel], pt->depth[srcLevel], - srcData, - srcStride, /* stride in texels */ - pt->width[dstLevel], pt->height[dstLevel], pt->depth[dstLevel], - dstData, - dstStride); /* stride in texels */ + 0 /*border*/, + u_minify(pt->width0, srcLevel), + u_minify(pt->height0, srcLevel), + u_minify(pt->depth0, srcLevel), + srcData, + srcStride, /* stride in texels */ + u_minify(pt->width0, dstLevel), + u_minify(pt->height0, dstLevel), + u_minify(pt->depth0, dstLevel), + dstData, + dstStride); /* stride in texels */ screen->transfer_unmap(screen, srcTrans); screen->transfer_unmap(screen, dstTrans); @@ -232,9 +237,9 @@ st_generate_mipmap(GLcontext *ctx, GLenum target, oldTex->target, oldTex->format, lastLevel, - oldTex->width[0], - oldTex->height[0], - oldTex->depth[0], + oldTex->width0, + oldTex->height0, + oldTex->depth0, oldTex->tex_usage); /* The texture isn't in a "complete" state yet so set the expected @@ -269,9 +274,9 @@ st_generate_mipmap(GLcontext *ctx, GLenum target, = _mesa_get_tex_image(ctx, texObj, target, srcLevel); struct gl_texture_image *dstImage; struct st_texture_image *stImage; - uint dstWidth = pt->width[dstLevel]; - uint dstHeight = pt->height[dstLevel]; - uint dstDepth = pt->depth[dstLevel]; + uint dstWidth = u_minify(pt->width0, dstLevel); + uint dstHeight = u_minify(pt->height0, dstLevel); + uint dstDepth = u_minify(pt->depth0, dstLevel); uint border = srcImage->Border; dstImage = _mesa_get_tex_image(ctx, texObj, target, dstLevel); diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 3d6c2158191..5e8e61135e6 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -676,6 +676,31 @@ emit_inverted_wpos( struct st_translate *t, /** + * OpenGL's fragment gl_FrontFace input is 1 for front-facing, 0 for back. + * TGSI uses +1 for front, -1 for back. + * This function converts the TGSI value to the GL value. Simply clamping/ + * saturating the value to [0,1] does the job. + */ +static void +emit_face_var( struct st_translate *t, + const struct gl_program *program ) +{ + struct ureg_program *ureg = t->ureg; + struct ureg_dst face_temp = ureg_DECL_temporary( ureg ); + struct ureg_src face_input = t->inputs[t->inputMapping[FRAG_ATTRIB_FACE]]; + + /* MOV_SAT face_temp, input[face] + */ + face_temp = ureg_saturate( face_temp ); + ureg_MOV( ureg, face_temp, face_input ); + + /* Use face_temp as face input from here on: + */ + t->inputs[t->inputMapping[FRAG_ATTRIB_FACE]] = ureg_src(face_temp); +} + + +/** * Translate Mesa program to TGSI format. * \param program the program to translate * \param numInputs number of input registers used @@ -704,12 +729,10 @@ st_translate_mesa_program( const ubyte inputSemanticName[], const ubyte inputSemanticIndex[], const GLuint interpMode[], - const GLbitfield inputFlags[], GLuint numOutputs, const GLuint outputMapping[], const ubyte outputSemanticName[], - const ubyte outputSemanticIndex[], - const GLbitfield outputFlags[] ) + const ubyte outputSemanticIndex[] ) { struct st_translate translate, *t; struct ureg_program *ureg; @@ -748,6 +771,10 @@ st_translate_mesa_program( emit_inverted_wpos( t, program ); } + if (program->InputsRead & FRAG_BIT_FACE) { + emit_face_var( t, program ); + } + /* * Declare output attributes. */ diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.h b/src/mesa/state_tracker/st_mesa_to_tgsi.h index c0d1ff59e1f..dc0362fe790 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.h +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.h @@ -49,12 +49,10 @@ st_translate_mesa_program( const ubyte inputSemanticName[], const ubyte inputSemanticIndex[], const GLuint interpMode[], - const GLbitfield inputFlags[], GLuint numOutputs, const GLuint outputMapping[], const ubyte outputSemanticName[], - const ubyte outputSemanticIndex[], - const GLbitfield outputFlags[] ); + const ubyte outputSemanticIndex[] ); void st_free_tokens(const struct tgsi_token *tokens); diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 190b6a5526d..d66f45d13ea 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -49,6 +49,36 @@ #include "st_mesa_to_tgsi.h" #include "cso_cache/cso_context.h" + /* Clean out any old compilations: + */ +void +st_vp_release_varients( struct st_context *st, + struct st_vertex_program *stvp ) +{ + struct st_vp_varient *vpv; + + for (vpv = stvp->varients; vpv; ) { + struct st_vp_varient *next = vpv->next; + + if (vpv->driver_shader) + cso_delete_vertex_shader(st->cso_context, vpv->driver_shader); + + if (vpv->draw_shader) + draw_delete_vertex_shader( st->draw, vpv->draw_shader ); + + if (vpv->state.tokens) + st_free_tokens(vpv->state.tokens); + + FREE( vpv ); + + vpv = next; + } + + stvp->varients = NULL; +} + + + /** * Translate a Mesa vertex shader into a TGSI shader. @@ -58,31 +88,13 @@ * \return pointer to cached pipe_shader object. */ void -st_translate_vertex_program(struct st_context *st, - struct st_vertex_program *stvp, - const GLuint outputMapping[], - const ubyte *outputSemanticName, - const ubyte *outputSemanticIndex) +st_prepare_vertex_program(struct st_context *st, + struct st_vertex_program *stvp) { - struct pipe_context *pipe = st->pipe; - GLuint defaultOutputMapping[VERT_RESULT_MAX]; - GLuint attr, i; - GLuint num_generic = 0; - - ubyte vs_input_semantic_name[PIPE_MAX_SHADER_INPUTS]; - ubyte vs_input_semantic_index[PIPE_MAX_SHADER_INPUTS]; - uint vs_num_inputs = 0; - - ubyte vs_output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; - ubyte vs_output_semantic_index[PIPE_MAX_SHADER_OUTPUTS]; - uint vs_num_outputs = 0; - - GLbitfield input_flags[MAX_PROGRAM_INPUTS]; - GLbitfield output_flags[MAX_PROGRAM_OUTPUTS]; + GLuint attr; - /*memset(&vs, 0, sizeof(vs));*/ - memset(input_flags, 0, sizeof(input_flags)); - memset(output_flags, 0, sizeof(output_flags)); + stvp->num_inputs = 0; + stvp->num_outputs = 0; if (stvp->Base.IsPositionInvariant) _mesa_insert_mvp_code(st->ctx, &stvp->Base); @@ -93,162 +105,56 @@ st_translate_vertex_program(struct st_context *st, */ for (attr = 0; attr < VERT_ATTRIB_MAX; attr++) { if (stvp->Base.Base.InputsRead & (1 << attr)) { - const GLuint slot = vs_num_inputs; - - vs_num_inputs++; - - stvp->input_to_index[attr] = slot; - stvp->index_to_input[slot] = attr; - - switch (attr) { - case VERT_ATTRIB_POS: - vs_input_semantic_name[slot] = TGSI_SEMANTIC_POSITION; - vs_input_semantic_index[slot] = 0; - break; - case VERT_ATTRIB_WEIGHT: - /* fall-through */ - case VERT_ATTRIB_NORMAL: - /* just label as a generic */ - vs_input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - vs_input_semantic_index[slot] = 0; - break; - case VERT_ATTRIB_COLOR0: - vs_input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; - vs_input_semantic_index[slot] = 0; - break; - case VERT_ATTRIB_COLOR1: - vs_input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; - vs_input_semantic_index[slot] = 1; - break; - case VERT_ATTRIB_FOG: - vs_input_semantic_name[slot] = TGSI_SEMANTIC_FOG; - vs_input_semantic_index[slot] = 0; - break; - case VERT_ATTRIB_POINT_SIZE: - vs_input_semantic_name[slot] = TGSI_SEMANTIC_PSIZE; - vs_input_semantic_index[slot] = 0; - break; - case VERT_ATTRIB_TEX0: - case VERT_ATTRIB_TEX1: - case VERT_ATTRIB_TEX2: - case VERT_ATTRIB_TEX3: - case VERT_ATTRIB_TEX4: - case VERT_ATTRIB_TEX5: - case VERT_ATTRIB_TEX6: - case VERT_ATTRIB_TEX7: - assert(slot < Elements(vs_input_semantic_name)); - vs_input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - vs_input_semantic_index[slot] = num_generic++; - break; - case VERT_ATTRIB_GENERIC0: - case VERT_ATTRIB_GENERIC1: - case VERT_ATTRIB_GENERIC2: - case VERT_ATTRIB_GENERIC3: - case VERT_ATTRIB_GENERIC4: - case VERT_ATTRIB_GENERIC5: - case VERT_ATTRIB_GENERIC6: - case VERT_ATTRIB_GENERIC7: - case VERT_ATTRIB_GENERIC8: - case VERT_ATTRIB_GENERIC9: - case VERT_ATTRIB_GENERIC10: - case VERT_ATTRIB_GENERIC11: - case VERT_ATTRIB_GENERIC12: - case VERT_ATTRIB_GENERIC13: - case VERT_ATTRIB_GENERIC14: - case VERT_ATTRIB_GENERIC15: - assert(attr < VERT_ATTRIB_MAX); - assert(slot < Elements(vs_input_semantic_name)); - vs_input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - vs_input_semantic_index[slot] = num_generic++; - break; - default: - assert(0); - } - - input_flags[slot] = stvp->Base.Base.InputFlags[attr]; + stvp->input_to_index[attr] = stvp->num_inputs; + stvp->index_to_input[stvp->num_inputs] = attr; + stvp->num_inputs++; } } -#if 0 - if (outputMapping && outputSemanticName) { - printf("VERT_RESULT written out_slot semantic_name semantic_index\n"); - for (attr = 0; attr < VERT_RESULT_MAX; attr++) { - printf(" %-2d %c %3d %2d %2d\n", - attr, - ((stvp->Base.Base.OutputsWritten & (1 << attr)) ? 'Y' : ' '), - outputMapping[attr], - outputSemanticName[attr], - outputSemanticIndex[attr]); - } - } -#endif - - /* initialize output semantics to defaults */ - for (i = 0; i < PIPE_MAX_SHADER_OUTPUTS; i++) { - assert(i < Elements(vs_output_semantic_name)); - vs_output_semantic_name[i] = TGSI_SEMANTIC_GENERIC; - vs_output_semantic_index[i] = 0; - output_flags[i] = 0x0; - } - - num_generic = 0; - /* - * Determine number of outputs, the (default) output register - * mapping and the semantic information for each output. + /* Compute mapping of vertex program outputs to slots. */ for (attr = 0; attr < VERT_RESULT_MAX; attr++) { - if (stvp->Base.Base.OutputsWritten & (1 << attr)) { - GLuint slot; - - /* XXX - * Pass in the fragment program's input's semantic info. - * Use the generic semantic indexes from there, instead of - * guessing below. - */ - - if (outputMapping) { - slot = outputMapping[attr]; - assert(slot != ~0); - } - else { - slot = vs_num_outputs; - vs_num_outputs++; - defaultOutputMapping[attr] = slot; - } + if ((stvp->Base.Base.OutputsWritten & (1 << attr)) == 0) { + stvp->result_to_output[attr] = ~0; + } + else { + unsigned slot = stvp->num_outputs++; + + stvp->result_to_output[attr] = slot; switch (attr) { case VERT_RESULT_HPOS: - assert(slot == 0); - vs_output_semantic_name[slot] = TGSI_SEMANTIC_POSITION; - vs_output_semantic_index[slot] = 0; + stvp->output_semantic_name[slot] = TGSI_SEMANTIC_POSITION; + stvp->output_semantic_index[slot] = 0; break; case VERT_RESULT_COL0: - vs_output_semantic_name[slot] = TGSI_SEMANTIC_COLOR; - vs_output_semantic_index[slot] = 0; + stvp->output_semantic_name[slot] = TGSI_SEMANTIC_COLOR; + stvp->output_semantic_index[slot] = 0; break; case VERT_RESULT_COL1: - vs_output_semantic_name[slot] = TGSI_SEMANTIC_COLOR; - vs_output_semantic_index[slot] = 1; + stvp->output_semantic_name[slot] = TGSI_SEMANTIC_COLOR; + stvp->output_semantic_index[slot] = 1; break; case VERT_RESULT_BFC0: - vs_output_semantic_name[slot] = TGSI_SEMANTIC_BCOLOR; - vs_output_semantic_index[slot] = 0; + stvp->output_semantic_name[slot] = TGSI_SEMANTIC_BCOLOR; + stvp->output_semantic_index[slot] = 0; break; case VERT_RESULT_BFC1: - vs_output_semantic_name[slot] = TGSI_SEMANTIC_BCOLOR; - vs_output_semantic_index[slot] = 1; + stvp->output_semantic_name[slot] = TGSI_SEMANTIC_BCOLOR; + stvp->output_semantic_index[slot] = 1; break; case VERT_RESULT_FOGC: - vs_output_semantic_name[slot] = TGSI_SEMANTIC_FOG; - vs_output_semantic_index[slot] = 0; + stvp->output_semantic_name[slot] = TGSI_SEMANTIC_FOG; + stvp->output_semantic_index[slot] = 0; break; case VERT_RESULT_PSIZ: - vs_output_semantic_name[slot] = TGSI_SEMANTIC_PSIZE; - vs_output_semantic_index[slot] = 0; + stvp->output_semantic_name[slot] = TGSI_SEMANTIC_PSIZE; + stvp->output_semantic_index[slot] = 0; break; case VERT_RESULT_EDGE: assert(0); break; + case VERT_RESULT_TEX0: case VERT_RESULT_TEX1: case VERT_RESULT_TEX2: @@ -257,92 +163,50 @@ st_translate_vertex_program(struct st_context *st, case VERT_RESULT_TEX5: case VERT_RESULT_TEX6: case VERT_RESULT_TEX7: - /* fall-through */ + stvp->output_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + stvp->output_semantic_index[slot] = attr - VERT_RESULT_TEX0; + break; + case VERT_RESULT_VAR0: - /* fall-through */ default: - assert(slot < Elements(vs_output_semantic_name)); - if (outputSemanticName) { - /* use provided semantic into */ - assert(outputSemanticName[attr] != TGSI_SEMANTIC_COUNT); - vs_output_semantic_name[slot] = outputSemanticName[attr]; - vs_output_semantic_index[slot] = outputSemanticIndex[attr]; - } - else { - /* use default semantic info */ - vs_output_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - vs_output_semantic_index[slot] = num_generic++; - } + assert(attr < VERT_RESULT_MAX); + stvp->output_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + stvp->output_semantic_index[slot] = (FRAG_ATTRIB_VAR0 - + FRAG_ATTRIB_TEX0 + + attr - + VERT_RESULT_VAR0); + break; } - - assert(slot < Elements(output_flags)); - output_flags[slot] = stvp->Base.Base.OutputFlags[attr]; } } +} - if (outputMapping) { - /* find max output slot referenced to compute vs_num_outputs */ - GLuint maxSlot = 0; - for (attr = 0; attr < VERT_RESULT_MAX; attr++) { - if (outputMapping[attr] != ~0 && outputMapping[attr] > maxSlot) - maxSlot = outputMapping[attr]; - } - vs_num_outputs = maxSlot + 1; - } - else { - outputMapping = defaultOutputMapping; - } - -#if 0 /* debug */ - { - GLuint i; - printf("outputMapping? %d\n", outputMapping ? 1 : 0); - if (outputMapping) { - printf("attr -> slot\n"); - for (i = 0; i < 16; i++) { - printf(" %2d %3d\n", i, outputMapping[i]); - } - } - printf("slot sem_name sem_index\n"); - for (i = 0; i < vs_num_outputs; i++) { - printf(" %2d %d %d\n", - i, - vs_output_semantic_name[i], - vs_output_semantic_index[i]); - } - } -#endif - /* free old shader state, if any */ - if (stvp->state.tokens) { - st_free_tokens(stvp->state.tokens); - stvp->state.tokens = NULL; - } - if (stvp->driver_shader) { - cso_delete_vertex_shader(st->cso_context, stvp->driver_shader); - stvp->driver_shader = NULL; - } +struct st_vp_varient * +st_translate_vertex_program(struct st_context *st, + struct st_vertex_program *stvp, + const struct st_vp_varient_key *key) +{ + struct st_vp_varient *vpv = CALLOC_STRUCT(st_vp_varient); + struct pipe_context *pipe = st->pipe; - stvp->state.tokens = + vpv->state.tokens = st_translate_mesa_program(st->ctx, TGSI_PROCESSOR_VERTEX, &stvp->Base.Base, /* inputs */ - vs_num_inputs, + stvp->num_inputs, stvp->input_to_index, - vs_input_semantic_name, - vs_input_semantic_index, + NULL, /* input semantic name */ + NULL, /* input semantic index */ NULL, - input_flags, /* outputs */ - vs_num_outputs, - outputMapping, - vs_output_semantic_name, - vs_output_semantic_index, - output_flags ); + stvp->num_outputs, + stvp->result_to_output, + stvp->output_semantic_name, + stvp->output_semantic_index ); - stvp->num_inputs = vs_num_inputs; - stvp->driver_shader = pipe->create_vs_state(pipe, &stvp->state); + vpv->driver_shader = pipe->create_vs_state(pipe, &vpv->state); if ((ST_DEBUG & DEBUG_TGSI) && (ST_DEBUG & DEBUG_MESA)) { _mesa_print_program(&stvp->Base.Base); @@ -350,9 +214,11 @@ st_translate_vertex_program(struct st_context *st, } if (ST_DEBUG & DEBUG_TGSI) { - tgsi_dump( stvp->state.tokens, 0 ); + tgsi_dump( vpv->state.tokens, 0 ); debug_printf("\n"); } + + return vpv; } @@ -375,7 +241,6 @@ st_translate_fragment_program(struct st_context *st, GLuint attr; const GLbitfield inputsRead = stfp->Base.Base.InputsRead; GLuint vslot = 0; - GLuint num_generic = 0; uint fs_num_inputs = 0; @@ -383,13 +248,6 @@ st_translate_fragment_program(struct st_context *st, ubyte fs_output_semantic_index[PIPE_MAX_SHADER_OUTPUTS]; uint fs_num_outputs = 0; - GLbitfield input_flags[MAX_PROGRAM_INPUTS]; - GLbitfield output_flags[MAX_PROGRAM_OUTPUTS]; - - /*memset(&fs, 0, sizeof(fs));*/ - memset(input_flags, 0, sizeof(input_flags)); - memset(output_flags, 0, sizeof(output_flags)); - /* which vertex output goes to the first fragment input: */ if (inputsRead & FRAG_BIT_WPOS) vslot = 0; @@ -432,14 +290,25 @@ st_translate_fragment_program(struct st_context *st, break; case FRAG_ATTRIB_FACE: stfp->input_semantic_name[slot] = TGSI_SEMANTIC_FACE; - stfp->input_semantic_index[slot] = num_generic++; + stfp->input_semantic_index[slot] = 0; interpMode[slot] = TGSI_INTERPOLATE_CONSTANT; break; - case FRAG_ATTRIB_PNTC: - stfp->input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - stfp->input_semantic_index[slot] = num_generic++; - interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; - break; + + /* In most cases, there is nothing special about these + * inputs, so adopt a convention to use the generic + * semantic name and the mesa FRAG_ATTRIB_ number as the + * index. + * + * All that is required is that the vertex shader labels + * its own outputs similarly, and that the vertex shader + * generates at least every output required by the + * fragment shader plus fixed-function hardware (such as + * BFC). + * + * There is no requirement that semantic indexes start at + * zero or be restricted to a particular range -- nobody + * should be building tables based on semantic index. + */ case FRAG_ATTRIB_TEX0: case FRAG_ATTRIB_TEX1: case FRAG_ATTRIB_TEX2: @@ -448,19 +317,18 @@ st_translate_fragment_program(struct st_context *st, case FRAG_ATTRIB_TEX5: case FRAG_ATTRIB_TEX6: case FRAG_ATTRIB_TEX7: - stfp->input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - stfp->input_semantic_index[slot] = num_generic++; - interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; - break; + case FRAG_ATTRIB_PNTC: case FRAG_ATTRIB_VAR0: - /* fall-through */ default: + /* Actually, let's try and zero-base this just for + * readability of the generated TGSI. + */ + assert(attr >= FRAG_ATTRIB_TEX0); + stfp->input_semantic_index[slot] = (attr - FRAG_ATTRIB_TEX0); stfp->input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - stfp->input_semantic_index[slot] = num_generic++; interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; + break; } - - input_flags[slot] = stfp->Base.Base.InputFlags[attr]; } } @@ -498,8 +366,6 @@ st_translate_fragment_program(struct st_context *st, break; } - output_flags[fs_num_outputs] = stfp->Base.Base.OutputFlags[attr]; - fs_num_outputs++; } } @@ -518,13 +384,11 @@ st_translate_fragment_program(struct st_context *st, stfp->input_semantic_name, stfp->input_semantic_index, interpMode, - input_flags, /* outputs */ fs_num_outputs, outputMapping, fs_output_semantic_name, - fs_output_semantic_index, - output_flags ); + fs_output_semantic_index ); stfp->driver_shader = pipe->create_fs_state(pipe, &stfp->state); diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index e2e5eddef22..88aadbd7510 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -64,41 +64,70 @@ struct st_fragment_program struct pipe_shader_state state; void *driver_shader; - GLuint param_state; - - /** List of vertex programs which have been translated such that their - * outputs match this fragment program's inputs. - */ - struct translated_vertex_program *vertex_programs; - /** Program prefixed with glBitmap prologue */ struct st_fragment_program *bitmap_program; uint bitmap_sampler; }; + +struct st_vp_varient_key +{ + char dummy; /* currently unused */ +}; + + +/** + * This represents a vertex program, especially translated to match + * the inputs of a particular fragment shader. + */ +struct st_vp_varient +{ + /* Parameters which generated this translated version of a vertex + * shader: + */ + struct st_vp_varient_key key; + + /** TGSI tokens -- why? + */ + struct pipe_shader_state state; + + /** Driver's compiled shader */ + void *driver_shader; + + /** For using our private draw module (glRasterPos) */ + struct draw_vertex_shader *draw_shader; + + /** Next in linked list */ + struct st_vp_varient *next; +}; + + + + /** * Derived from Mesa gl_fragment_program: */ struct st_vertex_program { struct gl_vertex_program Base; /**< The Mesa vertex program */ - GLuint serialNo; + GLuint serialNo, lastSerialNo; /** maps a Mesa VERT_ATTRIB_x to a packed TGSI input index */ GLuint input_to_index[VERT_ATTRIB_MAX]; /** maps a TGSI input index back to a Mesa VERT_ATTRIB_x */ GLuint index_to_input[PIPE_MAX_SHADER_INPUTS]; - GLuint num_inputs; - struct pipe_shader_state state; - void *driver_shader; + /** Maps VERT_RESULT_x to slot */ + GLuint result_to_output[VERT_RESULT_MAX]; + ubyte output_semantic_name[VERT_RESULT_MAX]; + ubyte output_semantic_index[VERT_RESULT_MAX]; + GLuint num_outputs; - /** For using our private draw module (glRasterPos) */ - struct draw_vertex_shader *draw_shader; - - GLuint param_state; + /** List of translated varients of this vertex program. + */ + struct st_vp_varient *varients; }; @@ -143,13 +172,21 @@ st_translate_fragment_program(struct st_context *st, const GLuint inputMapping[]); +/* Called after program string change, discard all previous + * compilation results. + */ extern void +st_prepare_vertex_program(struct st_context *st, + struct st_vertex_program *stvp); + +extern struct st_vp_varient * st_translate_vertex_program(struct st_context *st, - struct st_vertex_program *vp, - const GLuint vert_output_to_slot[], - const ubyte *fs_input_semantic_name, - const ubyte *fs_input_semantic_index); + struct st_vertex_program *stvp, + const struct st_vp_varient_key *key); +void +st_vp_release_varients( struct st_context *st, + struct st_vertex_program *stvp ); extern void st_print_shaders(GLcontext *ctx); diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c index 10f13512830..dbccee86c1c 100644 --- a/src/mesa/state_tracker/st_texture.c +++ b/src/mesa/state_tracker/st_texture.c @@ -44,6 +44,7 @@ #include "pipe/p_defines.h" #include "pipe/p_inlines.h" #include "util/u_rect.h" +#include "util/u_math.h" #define DBG if(0) printf @@ -100,9 +101,9 @@ st_texture_create(struct st_context *st, pt.target = target; pt.format = format; pt.last_level = last_level; - pt.width[0] = width0; - pt.height[0] = height0; - pt.depth[0] = depth0; + pt.width0 = width0; + pt.height0 = height0; + pt.depth0 = depth0; pf_get_block(format, &pt.block); pt.tex_usage = usage; @@ -135,9 +136,9 @@ st_texture_match_image(const struct pipe_texture *pt, /* Test if this image's size matches what's expected in the * established texture. */ - if (image->Width != pt->width[level] || - image->Height != pt->height[level] || - image->Depth != pt->depth[level]) + if (image->Width != u_minify(pt->width0, level) || + image->Height != u_minify(pt->height0, level) || + image->Depth != u_minify(pt->depth0, level)) return GL_FALSE; return GL_TRUE; @@ -265,7 +266,7 @@ st_texture_image_data(struct st_context *st, { struct pipe_context *pipe = st->pipe; struct pipe_screen *screen = pipe->screen; - GLuint depth = dst->depth[level]; + GLuint depth = u_minify(dst->depth0, level); GLuint i; const GLubyte *srcUB = src; struct pipe_transfer *dst_transfer; @@ -275,15 +276,16 @@ st_texture_image_data(struct st_context *st, for (i = 0; i < depth; i++) { dst_transfer = st_no_flush_get_tex_transfer(st, dst, face, level, i, PIPE_TRANSFER_WRITE, 0, 0, - dst->width[level], - dst->height[level]); + u_minify(dst->width0, level), + u_minify(dst->height0, level)); st_surface_data(pipe, dst_transfer, 0, 0, /* dstx, dsty */ srcUB, src_row_stride, 0, 0, /* source x, y */ - dst->width[level], dst->height[level]); /* width, height */ + u_minify(dst->width0, level), + u_minify(dst->height0, level)); /* width, height */ screen->tex_transfer_destroy(dst_transfer); @@ -301,9 +303,9 @@ st_texture_image_copy(struct pipe_context *pipe, GLuint face) { struct pipe_screen *screen = pipe->screen; - GLuint width = dst->width[dstLevel]; - GLuint height = dst->height[dstLevel]; - GLuint depth = dst->depth[dstLevel]; + GLuint width = u_minify(dst->width0, dstLevel); + GLuint height = u_minify(dst->height0, dstLevel); + GLuint depth = u_minify(dst->depth0, dstLevel); struct pipe_surface *src_surface; struct pipe_surface *dst_surface; GLuint i; @@ -313,13 +315,13 @@ st_texture_image_copy(struct pipe_context *pipe, /* find src texture level of needed size */ for (srcLevel = 0; srcLevel <= src->last_level; srcLevel++) { - if (src->width[srcLevel] == width && - src->height[srcLevel] == height) { + if (u_minify(src->width0, srcLevel) == width && + u_minify(src->height0, srcLevel) == height) { break; } } - assert(src->width[srcLevel] == width); - assert(src->height[srcLevel] == height); + assert(u_minify(src->width0, srcLevel) == width); + assert(u_minify(src->height0, srcLevel) == height); #if 0 { diff --git a/src/mesa/swrast_setup/ss_tritmp.h b/src/mesa/swrast_setup/ss_tritmp.h index 724b5e94fa5..17f38639563 100644 --- a/src/mesa/swrast_setup/ss_tritmp.h +++ b/src/mesa/swrast_setup/ss_tritmp.h @@ -67,8 +67,8 @@ static void TAG(triangle)(GLcontext *ctx, GLuint e0, GLuint e1, GLuint e2 ) if (facing == 1) { if (IND & SS_TWOSIDE_BIT) { if (IND & SS_RGBA_BIT) { - if (VB->ColorPtr[1]) { - GLfloat (*vbcolor)[4] = VB->ColorPtr[1]->data; + if (VB->BackfaceColorPtr) { + GLfloat (*vbcolor)[4] = VB->BackfaceColorPtr->data; if (swsetup->intColors) { COPY_CHAN4(saved_color[0], v[0]->color); @@ -81,7 +81,7 @@ static void TAG(triangle)(GLcontext *ctx, GLuint e0, GLuint e1, GLuint e2 ) COPY_4V(saved_col0[2], v[2]->attrib[FRAG_ATTRIB_COL0]); } - if (VB->ColorPtr[1]->stride) { + if (VB->BackfaceColorPtr->stride) { if (swsetup->intColors) { SS_COLOR(v[0]->color, vbcolor[e0]); SS_COLOR(v[1]->color, vbcolor[e1]); @@ -108,14 +108,14 @@ static void TAG(triangle)(GLcontext *ctx, GLuint e0, GLuint e1, GLuint e2 ) } } - if (VB->SecondaryColorPtr[1]) { - GLfloat (*vbspec)[4] = VB->SecondaryColorPtr[1]->data; + if (VB->BackfaceSecondaryColorPtr) { + GLfloat (*vbspec)[4] = VB->BackfaceSecondaryColorPtr->data; COPY_4V(saved_spec[0], v[0]->attrib[FRAG_ATTRIB_COL1]); COPY_4V(saved_spec[1], v[1]->attrib[FRAG_ATTRIB_COL1]); COPY_4V(saved_spec[2], v[2]->attrib[FRAG_ATTRIB_COL1]); - if (VB->SecondaryColorPtr[1]->stride) { + if (VB->BackfaceSecondaryColorPtr->stride) { SS_SPEC(v[0]->attrib[FRAG_ATTRIB_COL1], vbspec[e0]); SS_SPEC(v[1]->attrib[FRAG_ATTRIB_COL1], vbspec[e1]); SS_SPEC(v[2]->attrib[FRAG_ATTRIB_COL1], vbspec[e2]); @@ -127,7 +127,7 @@ static void TAG(triangle)(GLcontext *ctx, GLuint e0, GLuint e1, GLuint e2 ) } } } else { - GLfloat *vbindex = (GLfloat *)VB->IndexPtr[1]->data; + GLfloat *vbindex = (GLfloat *)VB->BackfaceIndexPtr->data; saved_index[0] = v[0]->attrib[FRAG_ATTRIB_CI][0]; saved_index[1] = v[1]->attrib[FRAG_ATTRIB_CI][0]; saved_index[2] = v[2]->attrib[FRAG_ATTRIB_CI][0]; @@ -200,7 +200,7 @@ static void TAG(triangle)(GLcontext *ctx, GLuint e0, GLuint e1, GLuint e2 ) if (IND & SS_TWOSIDE_BIT) { if (facing == 1) { if (IND & SS_RGBA_BIT) { - if (VB->ColorPtr[1]) { + if (VB->BackfaceColorPtr) { if (swsetup->intColors) { COPY_CHAN4(v[0]->color, saved_color[0]); COPY_CHAN4(v[1]->color, saved_color[1]); @@ -213,7 +213,7 @@ static void TAG(triangle)(GLcontext *ctx, GLuint e0, GLuint e1, GLuint e2 ) } } - if (VB->SecondaryColorPtr[1]) { + if (VB->BackfaceSecondaryColorPtr) { COPY_4V(v[0]->attrib[FRAG_ATTRIB_COL1], saved_spec[0]); COPY_4V(v[1]->attrib[FRAG_ATTRIB_COL1], saved_spec[1]); COPY_4V(v[2]->attrib[FRAG_ATTRIB_COL1], saved_spec[2]); diff --git a/src/mesa/tnl/t_context.h b/src/mesa/tnl/t_context.h index 6137c2d2fe5..ebaae6335b9 100644 --- a/src/mesa/tnl/t_context.h +++ b/src/mesa/tnl/t_context.h @@ -198,26 +198,23 @@ struct vertex_buffer */ GLuint Count; /**< Number of vertices currently in buffer */ - /* Pointers to current data. - * XXX some of these fields alias AttribPtr below and should be removed - * such as NormalPtr, TexCoordPtr, FogCoordPtr, etc. + /* Pointers to current data. Most of the data is in AttribPtr -- all of + * it that is one of VERT_ATTRIB_X. For things only produced by TNL, + * such as backface color or eye-space coordinates, they are stored + * here. */ GLuint *Elts; - GLvector4f *ObjPtr; /* _TNL_BIT_POS */ GLvector4f *EyePtr; /* _TNL_BIT_POS */ GLvector4f *ClipPtr; /* _TNL_BIT_POS */ GLvector4f *NdcPtr; /* _TNL_BIT_POS */ GLubyte ClipOrMask; /* _TNL_BIT_POS */ GLubyte ClipAndMask; /* _TNL_BIT_POS */ GLubyte *ClipMask; /* _TNL_BIT_POS */ - GLvector4f *NormalPtr; /* _TNL_BIT_NORMAL */ GLfloat *NormalLengthPtr; /* _TNL_BIT_NORMAL */ GLboolean *EdgeFlag; /* _TNL_BIT_EDGEFLAG */ - GLvector4f *TexCoordPtr[MAX_TEXTURE_COORD_UNITS]; /* VERT_TEX_0..n */ - GLvector4f *IndexPtr[2]; /* _TNL_BIT_INDEX */ - GLvector4f *ColorPtr[2]; /* _TNL_BIT_COLOR0 */ - GLvector4f *SecondaryColorPtr[2]; /* _TNL_BIT_COLOR1 */ - GLvector4f *FogCoordPtr; /* _TNL_BIT_FOG */ + GLvector4f *BackfaceIndexPtr; + GLvector4f *BackfaceColorPtr; + GLvector4f *BackfaceSecondaryColorPtr; const struct _mesa_prim *Primitive; GLuint PrimitiveCount; @@ -402,11 +399,6 @@ struct tnl_device_driver /* Alert tnl-aware drivers of changes to material. */ - void (*NotifyInputChanges)(GLcontext *ctx, GLuint bitmask); - /* Alert tnl-aware drivers of changes to size and stride of input - * arrays. - */ - /*** *** Rendering -- These functions called only from t_vb_render.c ***/ diff --git a/src/mesa/tnl/t_draw.c b/src/mesa/tnl/t_draw.c index 04fa1063003..1c7c733883c 100644 --- a/src/mesa/tnl/t_draw.c +++ b/src/mesa/tnl/t_draw.c @@ -251,22 +251,10 @@ static void bind_inputs( GLcontext *ctx, */ VB->Count = count; - - /* Legacy pointers -- remove one day. - */ - VB->ObjPtr = VB->AttribPtr[_TNL_ATTRIB_POS]; - VB->NormalPtr = VB->AttribPtr[_TNL_ATTRIB_NORMAL]; - VB->ColorPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR0]; - VB->ColorPtr[1] = NULL; - VB->IndexPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR_INDEX]; - VB->IndexPtr[1] = NULL; - VB->SecondaryColorPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR1]; - VB->SecondaryColorPtr[1] = NULL; - VB->FogCoordPtr = VB->AttribPtr[_TNL_ATTRIB_FOG]; - - for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) { - VB->TexCoordPtr[i] = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]; - } + /* These should perhaps be part of _TNL_ATTRIB_* */ + VB->BackfaceColorPtr = NULL; + VB->BackfaceIndexPtr = NULL; + VB->BackfaceSecondaryColorPtr = NULL; /* Clipping and drawing code still requires this to be a packed * array of ubytes which can be written into. TODO: Fix and diff --git a/src/mesa/tnl/t_pipeline.c b/src/mesa/tnl/t_pipeline.c index 357ef1e24b5..01b30babb48 100644 --- a/src/mesa/tnl/t_pipeline.c +++ b/src/mesa/tnl/t_pipeline.c @@ -86,10 +86,6 @@ static GLuint check_input_changes( GLcontext *ctx ) } } - if (tnl->pipeline.input_changes && - tnl->Driver.NotifyInputChanges) - tnl->Driver.NotifyInputChanges( ctx, tnl->pipeline.input_changes ); - return tnl->pipeline.input_changes; } diff --git a/src/mesa/tnl/t_vb_fog.c b/src/mesa/tnl/t_vb_fog.c index f3a7bd49f40..4a0e6ad4f99 100644 --- a/src/mesa/tnl/t_vb_fog.c +++ b/src/mesa/tnl/t_vb_fog.c @@ -156,7 +156,7 @@ run_fog_stage(GLcontext *ctx, struct tnl_pipeline_stage *stage) GLuint i; GLfloat *coord; /* Fog is computed from vertex or fragment Z values */ - /* source = VB->ObjPtr or VB->EyePtr coords */ + /* source = VB->AttribPtr[_TNL_ATTRIB_POS] or VB->EyePtr coords */ /* dest = VB->AttribPtr[_TNL_ATTRIB_FOG] = fog stage private storage */ VB->AttribPtr[_TNL_ATTRIB_FOG] = &store->fogcoord; @@ -176,11 +176,12 @@ run_fog_stage(GLcontext *ctx, struct tnl_pipeline_stage *stage) /* Full eye coords weren't required, just calculate the * eye Z values. */ - _mesa_dotprod_tab[VB->ObjPtr->size]( (GLfloat *) input->data, - 4 * sizeof(GLfloat), - VB->ObjPtr, plane ); + _mesa_dotprod_tab[VB->AttribPtr[_TNL_ATTRIB_POS]->size] + ( (GLfloat *) input->data, + 4 * sizeof(GLfloat), + VB->AttribPtr[_TNL_ATTRIB_POS], plane ); - input->count = VB->ObjPtr->count; + input->count = VB->AttribPtr[_TNL_ATTRIB_POS]->count; /* make sure coords are really positive NOTE should avoid going through array twice */ @@ -213,7 +214,7 @@ run_fog_stage(GLcontext *ctx, struct tnl_pipeline_stage *stage) /* input->count may be one if glFogCoord was only called once * before glBegin. But we need to compute fog for all vertices. */ - input->count = VB->ObjPtr->count; + input->count = VB->AttribPtr[_TNL_ATTRIB_POS]->count; VB->AttribPtr[_TNL_ATTRIB_FOG] = &store->fogcoord; /* dest data */ } @@ -227,7 +228,6 @@ run_fog_stage(GLcontext *ctx, struct tnl_pipeline_stage *stage) VB->AttribPtr[_TNL_ATTRIB_FOG] = input; } - VB->FogCoordPtr = VB->AttribPtr[_TNL_ATTRIB_FOG]; return GL_TRUE; } diff --git a/src/mesa/tnl/t_vb_light.c b/src/mesa/tnl/t_vb_light.c index f47f99397c9..8a0fe63fd8b 100644 --- a/src/mesa/tnl/t_vb_light.c +++ b/src/mesa/tnl/t_vb_light.c @@ -127,7 +127,7 @@ prepare_materials(GLcontext *ctx, const GLuint bitmask = ctx->Light.ColorMaterialBitmask; for (i = 0 ; i < MAT_ATTRIB_MAX ; i++) if (bitmask & (1<<i)) - VB->AttribPtr[_TNL_ATTRIB_MAT_FRONT_AMBIENT + i] = VB->ColorPtr[0]; + VB->AttribPtr[_TNL_ATTRIB_MAT_FRONT_AMBIENT + i] = VB->AttribPtr[_TNL_ATTRIB_COLOR0]; } /* Now, for each material attribute that's tracking vertex color, save @@ -200,7 +200,7 @@ static GLboolean run_lighting( GLcontext *ctx, struct light_stage_data *store = LIGHT_STAGE_DATA(stage); TNLcontext *tnl = TNL_CONTEXT(ctx); struct vertex_buffer *VB = &tnl->vb; - GLvector4f *input = ctx->_NeedEyeCoords ? VB->EyePtr : VB->ObjPtr; + GLvector4f *input = ctx->_NeedEyeCoords ? VB->EyePtr : VB->AttribPtr[_TNL_ATTRIB_POS]; GLuint idx; if (!ctx->Light.Enabled || ctx->VertexProgram._Current) @@ -208,13 +208,13 @@ static GLboolean run_lighting( GLcontext *ctx, /* Make sure we can talk about position x,y and z: */ - if (input->size <= 2 && input == VB->ObjPtr) { + if (input->size <= 2 && input == VB->AttribPtr[_TNL_ATTRIB_POS]) { _math_trans_4f( store->Input.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 ); @@ -246,10 +246,6 @@ static GLboolean run_lighting( GLcontext *ctx, */ store->light_func_tab[idx]( ctx, VB, stage, input ); - VB->AttribPtr[_TNL_ATTRIB_COLOR0] = VB->ColorPtr[0]; - VB->AttribPtr[_TNL_ATTRIB_COLOR1] = VB->SecondaryColorPtr[0]; - VB->AttribPtr[_TNL_ATTRIB_COLOR_INDEX] = VB->IndexPtr[0]; - return GL_TRUE; } diff --git a/src/mesa/tnl/t_vb_lighttmp.h b/src/mesa/tnl/t_vb_lighttmp.h index 124ca3c74fe..4ebef2356f5 100644 --- a/src/mesa/tnl/t_vb_lighttmp.h +++ b/src/mesa/tnl/t_vb_lighttmp.h @@ -72,13 +72,13 @@ static void TAG(light_rgba_spec)( GLcontext *ctx, fprintf(stderr, "%s\n", __FUNCTION__ ); #endif - VB->ColorPtr[0] = &store->LitColor[0]; - VB->SecondaryColorPtr[0] = &store->LitSecondary[0]; + VB->AttribPtr[_TNL_ATTRIB_COLOR0] = &store->LitColor[0]; + VB->AttribPtr[_TNL_ATTRIB_COLOR1] = &store->LitSecondary[0]; sumA[0] = ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_DIFFUSE][3]; #if IDX & LIGHT_TWOSIDE - VB->ColorPtr[1] = &store->LitColor[1]; - VB->SecondaryColorPtr[1] = &store->LitSecondary[1]; + VB->BackfaceColorPtr = &store->LitColor[1]; + VB->BackfaceSecondaryColorPtr = &store->LitSecondary[1]; sumA[1] = ctx->Light.Material.Attrib[MAT_ATTRIB_BACK_DIFFUSE][3]; #endif @@ -259,11 +259,11 @@ static void TAG(light_rgba)( GLcontext *ctx, fprintf(stderr, "%s\n", __FUNCTION__ ); #endif - VB->ColorPtr[0] = &store->LitColor[0]; + VB->AttribPtr[_TNL_ATTRIB_COLOR0] = &store->LitColor[0]; sumA[0] = ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_DIFFUSE][3]; #if IDX & LIGHT_TWOSIDE - VB->ColorPtr[1] = &store->LitColor[1]; + VB->BackfaceColorPtr = &store->LitColor[1]; sumA[1] = ctx->Light.Material.Attrib[MAT_ATTRIB_BACK_DIFFUSE][3]; #endif @@ -449,9 +449,9 @@ static void TAG(light_fast_rgba_single)( GLcontext *ctx, (void) input; /* doesn't refer to Eye or Obj */ - VB->ColorPtr[0] = &store->LitColor[0]; + VB->AttribPtr[_TNL_ATTRIB_COLOR0] = &store->LitColor[0]; #if IDX & LIGHT_TWOSIDE - VB->ColorPtr[1] = &store->LitColor[1]; + VB->BackfaceColorPtr = &store->LitColor[1]; #endif if (nr > 1) { @@ -559,9 +559,9 @@ static void TAG(light_fast_rgba)( GLcontext *ctx, sumA[0] = ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_DIFFUSE][3]; sumA[1] = ctx->Light.Material.Attrib[MAT_ATTRIB_BACK_DIFFUSE][3]; - VB->ColorPtr[0] = &store->LitColor[0]; + VB->AttribPtr[_TNL_ATTRIB_COLOR0] = &store->LitColor[0]; #if IDX & LIGHT_TWOSIDE - VB->ColorPtr[1] = &store->LitColor[1]; + VB->BackfaceColorPtr = &store->LitColor[1]; #endif if (nr > 1) { @@ -665,14 +665,14 @@ static void TAG(light_ci)( GLcontext *ctx, fprintf(stderr, "%s\n", __FUNCTION__ ); #endif - VB->IndexPtr[0] = &store->LitIndex[0]; + VB->AttribPtr[_TNL_ATTRIB_COLOR_INDEX] = &store->LitIndex[0]; #if IDX & LIGHT_TWOSIDE - VB->IndexPtr[1] = &store->LitIndex[1]; + VB->BackfaceIndexPtr = &store->LitIndex[1]; #endif - indexResult[0] = (GLfloat *)VB->IndexPtr[0]->data; + indexResult[0] = (GLfloat *)VB->AttribPtr[_TNL_ATTRIB_COLOR_INDEX]->data; #if IDX & LIGHT_TWOSIDE - indexResult[1] = (GLfloat *)VB->IndexPtr[1]->data; + indexResult[1] = (GLfloat *)VB->BackfaceIndexPtr->data; #endif /* loop over vertices */ diff --git a/src/mesa/tnl/t_vb_normals.c b/src/mesa/tnl/t_vb_normals.c index a4821cc1ccc..693d3dc118d 100644 --- a/src/mesa/tnl/t_vb_normals.c +++ b/src/mesa/tnl/t_vb_normals.c @@ -79,7 +79,6 @@ run_normal_stage(GLcontext *ctx, struct tnl_pipeline_stage *stage) } VB->AttribPtr[_TNL_ATTRIB_NORMAL] = &store->normal; - VB->NormalPtr = &store->normal; VB->NormalLengthPtr = NULL; /* no longer valid */ return GL_TRUE; diff --git a/src/mesa/tnl/t_vb_program.c b/src/mesa/tnl/t_vb_program.c index e69f7d57664..c289cdfbaac 100644 --- a/src/mesa/tnl/t_vb_program.c +++ b/src/mesa/tnl/t_vb_program.c @@ -454,19 +454,14 @@ run_vp( GLcontext *ctx, struct tnl_pipeline_stage *stage ) VB->ClipPtr->count = VB->Count; } - VB->ColorPtr[0] = &store->results[VERT_RESULT_COL0]; - VB->ColorPtr[1] = &store->results[VERT_RESULT_BFC0]; - VB->SecondaryColorPtr[0] = &store->results[VERT_RESULT_COL1]; - VB->SecondaryColorPtr[1] = &store->results[VERT_RESULT_BFC1]; - VB->FogCoordPtr = &store->results[VERT_RESULT_FOGC]; - VB->AttribPtr[VERT_ATTRIB_COLOR0] = &store->results[VERT_RESULT_COL0]; VB->AttribPtr[VERT_ATTRIB_COLOR1] = &store->results[VERT_RESULT_COL1]; VB->AttribPtr[VERT_ATTRIB_FOG] = &store->results[VERT_RESULT_FOGC]; VB->AttribPtr[_TNL_ATTRIB_POINTSIZE] = &store->results[VERT_RESULT_PSIZ]; + VB->BackfaceColorPtr = &store->results[VERT_RESULT_BFC0]; + VB->BackfaceSecondaryColorPtr = &store->results[VERT_RESULT_BFC1]; for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) { - VB->TexCoordPtr[i] = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i] = &store->results[VERT_RESULT_TEX0 + i]; } diff --git a/src/mesa/tnl/t_vb_texgen.c b/src/mesa/tnl/t_vb_texgen.c index 7c1819b223c..9ef13bc96d8 100644 --- a/src/mesa/tnl/t_vb_texgen.c +++ b/src/mesa/tnl/t_vb_texgen.c @@ -341,7 +341,7 @@ static void texgen( GLcontext *ctx, GLvector4f *in = VB->AttribPtr[VERT_ATTRIB_TEX0 + unit]; GLvector4f *out = &store->texcoord[unit]; const struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; - const GLvector4f *obj = VB->ObjPtr; + const GLvector4f *obj = VB->AttribPtr[_TNL_ATTRIB_POS]; const GLvector4f *eye = VB->EyePtr; const GLvector4f *normal = VB->AttribPtr[_TNL_ATTRIB_NORMAL]; const GLfloat *m = store->tmp_m; @@ -498,7 +498,6 @@ static GLboolean run_texgen_stage( GLcontext *ctx, store->TexgenFunc[i]( ctx, store, i ); - VB->TexCoordPtr[i] = VB->AttribPtr[VERT_ATTRIB_TEX0 + i] = &store->texcoord[i]; } } diff --git a/src/mesa/tnl/t_vb_texmat.c b/src/mesa/tnl/t_vb_texmat.c index 0abe8cc35dd..83688290e59 100644 --- a/src/mesa/tnl/t_vb_texmat.c +++ b/src/mesa/tnl/t_vb_texmat.c @@ -73,7 +73,6 @@ static GLboolean run_texmat_stage( GLcontext *ctx, ctx->TextureMatrixStack[i].Top, VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]); - VB->TexCoordPtr[i] = VB->AttribPtr[VERT_ATTRIB_TEX0+i] = &store->texcoord[i]; } } diff --git a/src/mesa/tnl/t_vb_vertex.c b/src/mesa/tnl/t_vb_vertex.c index 4734754ea40..bc7e0951ec8 100644 --- a/src/mesa/tnl/t_vb_vertex.c +++ b/src/mesa/tnl/t_vb_vertex.c @@ -152,16 +152,16 @@ static GLboolean run_vertex_stage( GLcontext *ctx, * Use combined ModelProject to avoid some depth artifacts */ if (ctx->ModelviewMatrixStack.Top->type == MATRIX_IDENTITY) - VB->EyePtr = VB->ObjPtr; + VB->EyePtr = VB->AttribPtr[_TNL_ATTRIB_POS]; else VB->EyePtr = TransformRaw( &store->eye, ctx->ModelviewMatrixStack.Top, - VB->ObjPtr); + VB->AttribPtr[_TNL_ATTRIB_POS]); } VB->ClipPtr = TransformRaw( &store->clip, &ctx->_ModelProjectMatrix, - VB->ObjPtr ); + VB->AttribPtr[_TNL_ATTRIB_POS] ); /* Drivers expect this to be clean to element 4... */ diff --git a/src/mesa/tnl/t_vertex_generic.c b/src/mesa/tnl/t_vertex_generic.c index 9812f8c8081..fa34d11d7bd 100644 --- a/src/mesa/tnl/t_vertex_generic.c +++ b/src/mesa/tnl/t_vertex_generic.c @@ -1092,33 +1092,33 @@ void _tnl_generic_interp_extras( GLcontext *ctx, { struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; - /* If stride is zero, ColorPtr[1] is constant across the VB, so + /* If stride is zero, BackfaceColorPtr is constant across the VB, so * there is no point interpolating between two values as they will * be identical. In all other cases, this value is generated by * t_vb_lighttmp.h and has a stride of 4 dwords. */ - if (VB->ColorPtr[1] && VB->ColorPtr[1]->stride) { - assert(VB->ColorPtr[1]->stride == 4 * sizeof(GLfloat)); + if (VB->BackfaceColorPtr && VB->BackfaceColorPtr->stride) { + assert(VB->BackfaceColorPtr->stride == 4 * sizeof(GLfloat)); INTERP_4F( t, - VB->ColorPtr[1]->data[dst], - VB->ColorPtr[1]->data[out], - VB->ColorPtr[1]->data[in] ); + VB->BackfaceColorPtr->data[dst], + VB->BackfaceColorPtr->data[out], + VB->BackfaceColorPtr->data[in] ); } - if (VB->SecondaryColorPtr[1]) { - assert(VB->SecondaryColorPtr[1]->stride == 4 * sizeof(GLfloat)); + if (VB->BackfaceSecondaryColorPtr) { + assert(VB->BackfaceSecondaryColorPtr->stride == 4 * sizeof(GLfloat)); INTERP_3F( t, - VB->SecondaryColorPtr[1]->data[dst], - VB->SecondaryColorPtr[1]->data[out], - VB->SecondaryColorPtr[1]->data[in] ); + VB->BackfaceSecondaryColorPtr->data[dst], + VB->BackfaceSecondaryColorPtr->data[out], + VB->BackfaceSecondaryColorPtr->data[in] ); } - if (VB->IndexPtr[1]) { - VB->IndexPtr[1]->data[dst][0] = LINTERP( t, - VB->IndexPtr[1]->data[out][0], - VB->IndexPtr[1]->data[in][0] ); + if (VB->BackfaceIndexPtr) { + VB->BackfaceIndexPtr->data[dst][0] = LINTERP( t, + VB->BackfaceIndexPtr->data[out][0], + VB->BackfaceIndexPtr->data[in][0] ); } if (VB->EdgeFlag) { @@ -1135,18 +1135,18 @@ void _tnl_generic_copy_pv_extras( GLcontext *ctx, /* See above comment: */ - if (VB->ColorPtr[1] && VB->ColorPtr[1]->stride) { - COPY_4FV( VB->ColorPtr[1]->data[dst], - VB->ColorPtr[1]->data[src] ); + if (VB->BackfaceColorPtr && VB->BackfaceColorPtr->stride) { + COPY_4FV( VB->BackfaceColorPtr->data[dst], + VB->BackfaceColorPtr->data[src] ); } - if (VB->SecondaryColorPtr[1]) { - COPY_4FV( VB->SecondaryColorPtr[1]->data[dst], - VB->SecondaryColorPtr[1]->data[src] ); + if (VB->BackfaceSecondaryColorPtr) { + COPY_4FV( VB->BackfaceSecondaryColorPtr->data[dst], + VB->BackfaceSecondaryColorPtr->data[src] ); } - if (VB->IndexPtr[1]) { - VB->IndexPtr[1]->data[dst][0] = VB->IndexPtr[1]->data[src][0]; + if (VB->BackfaceIndexPtr) { + VB->BackfaceIndexPtr->data[dst][0] = VB->BackfaceIndexPtr->data[src][0]; } _tnl_generic_copy_pv(ctx, dst, src); diff --git a/src/mesa/tnl_dd/t_dd_dmatmp.h b/src/mesa/tnl_dd/t_dd_dmatmp.h index e4b535fb680..e5885782c7a 100644 --- a/src/mesa/tnl_dd/t_dd_dmatmp.h +++ b/src/mesa/tnl_dd/t_dd_dmatmp.h @@ -443,7 +443,7 @@ static void TAG(render_quad_strip_verts)( GLcontext *ctx, } else if (HAVE_TRI_STRIPS && ctx->Light.ShadeModel == GL_FLAT && - TNL_CONTEXT(ctx)->vb.ColorPtr[0]->stride) { + TNL_CONTEXT(ctx)->vb.AttribPtr[_TNL_ATTRIB_COLOR0]->stride) { if (HAVE_ELTS) { LOCAL_VARS; int dmasz = GET_SUBSEQUENT_VB_MAX_ELTS(); @@ -1221,7 +1221,7 @@ static GLboolean TAG(validate_render)( GLcontext *ctx, ok = GL_TRUE; } else if (HAVE_TRI_STRIPS && ctx->Light.ShadeModel == GL_FLAT && - VB->ColorPtr[0]->stride != 0) { + VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride != 0) { if (HAVE_ELTS) { ok = (GLint) count < GET_SUBSEQUENT_VB_MAX_ELTS(); } diff --git a/src/mesa/tnl_dd/t_dd_tritmp.h b/src/mesa/tnl_dd/t_dd_tritmp.h index 1ae70f4059f..8574fe618b5 100644 --- a/src/mesa/tnl_dd/t_dd_tritmp.h +++ b/src/mesa/tnl_dd/t_dd_tritmp.h @@ -195,7 +195,7 @@ static void TAG(triangle)( GLcontext *ctx, GLuint e0, GLuint e1, GLuint e2 ) } } else { - GLfloat (*vbcolor)[4] = VB->ColorPtr[1]->data; + GLfloat (*vbcolor)[4] = VB->BackfaceColorPtr->data; (void) vbcolor; if (!DO_FLAT) { @@ -204,8 +204,8 @@ static void TAG(triangle)( GLcontext *ctx, GLuint e0, GLuint e1, GLuint e2 ) } VERT_SAVE_RGBA( 2 ); - if (VB->ColorPtr[1]->stride) { - ASSERT(VB->ColorPtr[1]->stride == 4*sizeof(GLfloat)); + if (VB->BackfaceColorPtr->stride) { + ASSERT(VB->BackfaceColorPtr->stride == 4*sizeof(GLfloat)); if (!DO_FLAT) { VERT_SET_RGBA( v[0], vbcolor[e0] ); @@ -221,9 +221,9 @@ static void TAG(triangle)( GLcontext *ctx, GLuint e0, GLuint e1, GLuint e2 ) VERT_SET_RGBA( v[2], vbcolor[0] ); } - if (HAVE_SPEC && VB->SecondaryColorPtr[1]) { - GLfloat (*vbspec)[4] = VB->SecondaryColorPtr[1]->data; - ASSERT(VB->SecondaryColorPtr[1]->stride == 4*sizeof(GLfloat)); + if (HAVE_SPEC && VB->BackfaceSecondaryColorPtr) { + GLfloat (*vbspec)[4] = VB->BackfaceSecondaryColorPtr->data; + ASSERT(VB->BackfaceSecondaryColorPtr->stride == 4*sizeof(GLfloat)); if (!DO_FLAT) { VERT_SAVE_SPEC( 0 ); @@ -237,7 +237,7 @@ static void TAG(triangle)( GLcontext *ctx, GLuint e0, GLuint e1, GLuint e2 ) } } else { - GLfloat (*vbindex) = (GLfloat *)VB->IndexPtr[1]->data; + GLfloat (*vbindex) = (GLfloat *)VB->BackfaceIndexPtr->data; if (!DO_FLAT) { VERT_SAVE_IND( 0 ); VERT_SAVE_IND( 1 ); @@ -279,7 +279,7 @@ static void TAG(triangle)( GLcontext *ctx, GLuint e0, GLuint e1, GLuint e2 ) VERT_SAVE_RGBA( 1 ); VERT_COPY_RGBA( v[0], v[2] ); VERT_COPY_RGBA( v[1], v[2] ); - if (HAVE_SPEC && VB->SecondaryColorPtr[0]) { + if (HAVE_SPEC && VB->AttribPtr[_TNL_ATTRIB_COLOR1]) { VERT_SAVE_SPEC( 0 ); VERT_SAVE_SPEC( 1 ); VERT_COPY_SPEC( v[0], v[2] ); @@ -374,7 +374,7 @@ static void TAG(triangle)( GLcontext *ctx, GLuint e0, GLuint e1, GLuint e2 ) if (HAVE_RGBA) { VERT_RESTORE_RGBA( 0 ); VERT_RESTORE_RGBA( 1 ); - if (HAVE_SPEC && VB->SecondaryColorPtr[0]) { + if (HAVE_SPEC && VB->AttribPtr[_TNL_ATTRIB_COLOR1]) { VERT_RESTORE_SPEC( 0 ); VERT_RESTORE_SPEC( 1 ); } @@ -436,7 +436,7 @@ static void TAG(quadr)( GLcontext *ctx, if (DO_TWOSIDE && facing == 1) { if (HAVE_RGBA) { - GLfloat (*vbcolor)[4] = VB->ColorPtr[1]->data; + GLfloat (*vbcolor)[4] = VB->BackfaceColorPtr->data; (void)vbcolor; if (HAVE_BACK_COLORS) { @@ -471,7 +471,7 @@ static void TAG(quadr)( GLcontext *ctx, } VERT_SAVE_RGBA( 3 ); - if (VB->ColorPtr[1]->stride) { + if (VB->BackfaceColorPtr->stride) { if (!DO_FLAT) { VERT_SET_RGBA( v[0], vbcolor[e0] ); VERT_SET_RGBA( v[1], vbcolor[e1] ); @@ -488,9 +488,9 @@ static void TAG(quadr)( GLcontext *ctx, VERT_SET_RGBA( v[3], vbcolor[0] ); } - if (HAVE_SPEC && VB->SecondaryColorPtr[1]) { - GLfloat (*vbspec)[4] = VB->SecondaryColorPtr[1]->data; - ASSERT(VB->SecondaryColorPtr[1]->stride==4*sizeof(GLfloat)); + if (HAVE_SPEC && VB->BackfaceSecondaryColorPtr) { + GLfloat (*vbspec)[4] = VB->BackfaceSecondaryColorPtr->data; + ASSERT(VB->BackfaceSecondaryColorPtr->stride==4*sizeof(GLfloat)); if (!DO_FLAT) { VERT_SAVE_SPEC( 0 ); @@ -506,7 +506,7 @@ static void TAG(quadr)( GLcontext *ctx, } } else { - GLfloat *vbindex = (GLfloat *)VB->IndexPtr[1]->data; + GLfloat *vbindex = (GLfloat *)VB->BackfaceIndexPtr->data; if (!DO_FLAT) { VERT_SAVE_IND( 0 ); VERT_SAVE_IND( 1 ); @@ -553,7 +553,7 @@ static void TAG(quadr)( GLcontext *ctx, VERT_COPY_RGBA( v[0], v[3] ); VERT_COPY_RGBA( v[1], v[3] ); VERT_COPY_RGBA( v[2], v[3] ); - if (HAVE_SPEC && VB->SecondaryColorPtr[0]) { + if (HAVE_SPEC && VB->AttribPtr[_TNL_ATTRIB_COLOR1]) { VERT_SAVE_SPEC( 0 ); VERT_SAVE_SPEC( 1 ); VERT_SAVE_SPEC( 2 ); @@ -659,7 +659,7 @@ static void TAG(quadr)( GLcontext *ctx, VERT_RESTORE_RGBA( 0 ); VERT_RESTORE_RGBA( 1 ); VERT_RESTORE_RGBA( 2 ); - if (HAVE_SPEC && VB->SecondaryColorPtr[0]) { + if (HAVE_SPEC && VB->AttribPtr[_TNL_ATTRIB_COLOR1]) { VERT_RESTORE_SPEC( 0 ); VERT_RESTORE_SPEC( 1 ); VERT_RESTORE_SPEC( 2 ); @@ -708,7 +708,7 @@ static void TAG(line)( GLcontext *ctx, GLuint e0, GLuint e1 ) if (HAVE_RGBA) { VERT_SAVE_RGBA( 0 ); VERT_COPY_RGBA( v[0], v[1] ); - if (HAVE_SPEC && VB->SecondaryColorPtr[0]) { + if (HAVE_SPEC && VB->AttribPtr[_TNL_ATTRIB_COLOR1]) { VERT_SAVE_SPEC( 0 ); VERT_COPY_SPEC( v[0], v[1] ); } @@ -725,7 +725,7 @@ static void TAG(line)( GLcontext *ctx, GLuint e0, GLuint e1 ) if (HAVE_RGBA) { VERT_RESTORE_RGBA( 0 ); - if (HAVE_SPEC && VB->SecondaryColorPtr[0]) { + if (HAVE_SPEC && VB->AttribPtr[_TNL_ATTRIB_COLOR1]) { VERT_RESTORE_SPEC( 0 ); } } diff --git a/src/mesa/tnl_dd/t_dd_vb.c b/src/mesa/tnl_dd/t_dd_vb.c index b3937c29a00..a8a0a69768d 100644 --- a/src/mesa/tnl_dd/t_dd_vb.c +++ b/src/mesa/tnl_dd/t_dd_vb.c @@ -297,19 +297,19 @@ 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) ); } } @@ -326,13 +326,13 @@ 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/tnl_dd/t_dd_vbtmp.h b/src/mesa/tnl_dd/t_dd_vbtmp.h index 92dd8931c33..85101b9ceb8 100644 --- a/src/mesa/tnl_dd/t_dd_vbtmp.h +++ b/src/mesa/tnl_dd/t_dd_vbtmp.h @@ -153,46 +153,46 @@ 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_stride = VB->ColorPtr[0]->stride; - col = VB->ColorPtr[0]->data; - col_size = VB->ColorPtr[0]->size; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_size = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->size; } if (DO_SPEC) { - if (VB->SecondaryColorPtr[0]) { - spec_stride = VB->SecondaryColorPtr[0]->stride; - spec = VB->SecondaryColorPtr[0]->data; + if (VB->AttribPtr[_TNL_ATTRIB_COLOR1]) { + spec_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->stride; + spec = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->data; } else { spec = (GLfloat (*)[4])ctx->Current.Attrib[VERT_ATTRIB_COLOR1]; spec_stride = 0; @@ -200,9 +200,9 @@ static void TAG(emit)( GLcontext *ctx, } 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}; @@ -356,9 +356,9 @@ 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_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; /* fprintf(stderr, "%s(small) importable %x\n", */ /* __FUNCTION__, VB->importable_data); */ @@ -410,22 +410,22 @@ 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_TEX0 + 2] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0 + 2] = VB->AttribPtr[_TNL_ATTRIB_TEX0 + 3]; - if (DO_TEX2 && VB->TexCoordPtr[1] == 0) - VB->TexCoordPtr[1] = VB->TexCoordPtr[2]; + if (DO_TEX2 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + 1] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0 + 1] = VB->AttribPtr[_TNL_ATTRIB_TEX0 + 2]; - if (DO_TEX1 && VB->TexCoordPtr[0] == 0) - VB->TexCoordPtr[0] = VB->TexCoordPtr[1]; + if (DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + 0] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0 + 0] = VB->AttribPtr[_TNL_ATTRIB_TEX0 + 1]; 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; @@ -438,14 +438,14 @@ 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_TEX0 + 2] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0 + 2] = VB->AttribPtr[_TNL_ATTRIB_TEX0 + 3]; - if (DO_TEX2 && VB->TexCoordPtr[1] == 0) - VB->TexCoordPtr[1] = VB->TexCoordPtr[2]; + if (DO_TEX2 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + 1] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0 + 1] = VB->AttribPtr[_TNL_ATTRIB_TEX0 + 2]; - if (DO_TEX1 && VB->TexCoordPtr[0] == 0) - VB->TexCoordPtr[0] = VB->TexCoordPtr[1]; + if (DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + 0] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0 + 0] = VB->AttribPtr[_TNL_ATTRIB_TEX0 + 1]; if (DO_PTEX) return GL_TRUE; @@ -453,14 +453,14 @@ 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/x86/gen_matypes.c b/src/mesa/x86/gen_matypes.c index d56b701aa8c..0d7e0f1f98b 100644 --- a/src/mesa/x86/gen_matypes.c +++ b/src/mesa/x86/gen_matypes.c @@ -120,22 +120,22 @@ int main( int argc, char **argv ) OFFSET( "VB_COUNT ", struct vertex_buffer, Count ); printf( "\n" ); OFFSET( "VB_ELTS ", struct vertex_buffer, Elts ); - OFFSET( "VB_OBJ_PTR ", struct vertex_buffer, ObjPtr ); + OFFSET( "VB_OBJ_PTR ", struct vertex_buffer, AttribPtr[_TNL_ATTRIB_POS] ); OFFSET( "VB_EYE_PTR ", struct vertex_buffer, EyePtr ); OFFSET( "VB_CLIP_PTR ", struct vertex_buffer, ClipPtr ); OFFSET( "VB_PROJ_CLIP_PTR ", struct vertex_buffer, NdcPtr ); OFFSET( "VB_CLIP_OR_MASK ", struct vertex_buffer, ClipOrMask ); OFFSET( "VB_CLIP_MASK ", struct vertex_buffer, ClipMask ); - OFFSET( "VB_NORMAL_PTR ", struct vertex_buffer, NormalPtr ); + OFFSET( "VB_NORMAL_PTR ", struct vertex_buffer, AttribPtr[_TNL_ATTRIB_NORMAL] ); OFFSET( "VB_EDGE_FLAG ", struct vertex_buffer, EdgeFlag ); - OFFSET( "VB_TEX0_COORD_PTR ", struct vertex_buffer, TexCoordPtr[0] ); - OFFSET( "VB_TEX1_COORD_PTR ", struct vertex_buffer, TexCoordPtr[1] ); - OFFSET( "VB_TEX2_COORD_PTR ", struct vertex_buffer, TexCoordPtr[2] ); - OFFSET( "VB_TEX3_COORD_PTR ", struct vertex_buffer, TexCoordPtr[3] ); - OFFSET( "VB_INDEX_PTR ", struct vertex_buffer, IndexPtr ); - OFFSET( "VB_COLOR_PTR ", struct vertex_buffer, ColorPtr ); - OFFSET( "VB_SECONDARY_COLOR_PTR ", struct vertex_buffer, SecondaryColorPtr ); - OFFSET( "VB_FOG_COORD_PTR ", struct vertex_buffer, FogCoordPtr ); + OFFSET( "VB_TEX0_COORD_PTR ", struct vertex_buffer, AttribPtr[_TNL_ATTRIB_TEX0] ); + OFFSET( "VB_TEX1_COORD_PTR ", struct vertex_buffer, AttribPtr[_TNL_ATTRIB_TEX1] ); + OFFSET( "VB_TEX2_COORD_PTR ", struct vertex_buffer, AttribPtr[_TNL_ATTRIB_TEX2] ); + OFFSET( "VB_TEX3_COORD_PTR ", struct vertex_buffer, AttribPtr[_TNL_ATTRIB_TEX3] ); + OFFSET( "VB_INDEX_PTR ", struct vertex_buffer, AttribPtr[_TNL_ATTRIB_COLOR_INDEX] ); + OFFSET( "VB_COLOR_PTR ", struct vertex_buffer, AttribPtr[_TNL_ATTRIB_COLOR0] ); + OFFSET( "VB_SECONDARY_COLOR_PTR ", struct vertex_buffer, AttribPtr[_TNL_ATTRIB_COLOR1] ); + OFFSET( "VB_FOG_COORD_PTR ", struct vertex_buffer, AttribPtr[_TNL_ATTRIB_FOG] ); OFFSET( "VB_PRIMITIVE ", struct vertex_buffer, Primitive ); printf( "\n" ); |