diff options
author | Zack Rusin <[email protected]> | 2013-04-16 20:31:22 -0700 |
---|---|---|
committer | Zack Rusin <[email protected]> | 2013-04-16 23:38:47 -0700 |
commit | f01f754ca13373d62f5f4ba5ff76d83aa4eac62b (patch) | |
tree | af3c40e5cebfeb05e40efd447b6af18a035653fb /src/gallium/auxiliary/draw | |
parent | be497ac9d3b5c50a4cd126578081bd54b68f16a9 (diff) |
draw/gs: make sure geometry shaders don't overflow
The specification says that the geometry shader should exit if the
number of emitted vertices is bigger or equal to max_output_vertices and
we can't do that because we're running in the SoA mode, which means that
our storing routines will keep getting called on channels that have
overflown (even though they will be masked out, but we just can't skip
them).
So we need some scratch area where we can keep writing the overflown
vertices without overwriting anything important or crashing.
Signed-off-by: Zack Rusin <[email protected]>
Reviewed-by: Jose Fonseca <[email protected]>
Diffstat (limited to 'src/gallium/auxiliary/draw')
-rw-r--r-- | src/gallium/auxiliary/draw/draw_gs.c | 50 | ||||
-rw-r--r-- | src/gallium/auxiliary/draw/draw_gs.h | 1 | ||||
-rw-r--r-- | src/gallium/auxiliary/draw/draw_llvm.c | 6 |
3 files changed, 47 insertions, 10 deletions
diff --git a/src/gallium/auxiliary/draw/draw_gs.c b/src/gallium/auxiliary/draw/draw_gs.c index 987c01a7919..2f94eaeda4f 100644 --- a/src/gallium/auxiliary/draw/draw_gs.c +++ b/src/gallium/auxiliary/draw/draw_gs.c @@ -280,6 +280,7 @@ llvm_fetch_gs_outputs(struct draw_geometry_shader *shader, int max_prims_per_invocation = 0; char *output_ptr = (char*)shader->gs_output; int i, j, prim_idx; + unsigned next_prim_boundary = shader->primitive_boundary; for (i = 0; i < shader->vector_length; ++i) { int prims = shader->llvm_emitted_primitives[i]; @@ -290,19 +291,42 @@ llvm_fetch_gs_outputs(struct draw_geometry_shader *shader, total_verts += shader->llvm_emitted_vertices[i]; } - output_ptr += shader->emitted_vertices * shader->vertex_size; for (i = 0; i < shader->vector_length - 1; ++i) { int current_verts = shader->llvm_emitted_vertices[i]; - - if (current_verts != shader->max_output_vertices) { - memcpy(output_ptr + (vertex_count + current_verts) * shader->vertex_size, - output_ptr + (vertex_count + shader->max_output_vertices) * shader->vertex_size, - shader->vertex_size * (total_verts - vertex_count)); + int next_verts = shader->llvm_emitted_vertices[i + 1]; +#if 0 + int j; + for (j = 0; j < current_verts; ++j) { + struct vertex_header *vh = (struct vertex_header *) + (output_ptr + shader->vertex_size * (i * next_prim_boundary + j)); + debug_printf("--- %d) [%f, %f, %f, %f]\n", j + vertex_count, + vh->data[0][0], vh->data[0][1], vh->data[0][2], vh->data[0][3]); + + } +#endif + debug_assert(current_verts <= shader->max_output_vertices); + debug_assert(next_verts <= shader->max_output_vertices); + if (next_verts) { + memmove(output_ptr + (vertex_count + current_verts) * shader->vertex_size, + output_ptr + ((i + 1) * next_prim_boundary) * shader->vertex_size, + shader->vertex_size * next_verts); } vertex_count += current_verts; } +#if 0 + { + int i; + for (i = 0; i < total_verts; ++i) { + struct vertex_header *vh = (struct vertex_header *)(output_ptr + shader->vertex_size * i); + debug_printf("%d) [%f, %f, %f, %f]\n", i, + vh->data[0][0], vh->data[0][1], vh->data[0][2], vh->data[0][3]); + + } + } +#endif + prim_idx = 0; for (i = 0; i < shader->vector_length; ++i) { int num_prims = shader->llvm_emitted_primitives[i]; @@ -513,10 +537,12 @@ int draw_geometry_shader_run(struct draw_geometry_shader *shader, output_verts->vertex_size = vertex_size; output_verts->stride = output_verts->vertex_size; + /* we allocate exactly one extra vertex per primitive to allow the GS to emit + * overflown vertices into some area where they won't harm anyone */ output_verts->verts = (struct vertex_header *)MALLOC(output_verts->vertex_size * max_out_prims * - shader->max_output_vertices); + shader->primitive_boundary); #if 0 debug_printf("%s count = %d (in prims # = %d)\n", @@ -724,6 +750,16 @@ draw_create_geometry_shader(struct draw_context *draw, TGSI_PROPERTY_GS_MAX_OUTPUT_VERTICES) gs->max_output_vertices = gs->info.properties[i].data[0]; } + /* Primitive boundary is bigger than max_output_vertices by one, because + * the specification says that the geometry shader should exit if the + * number of emitted vertices is bigger or equal to max_output_vertices and + * we can't do that because we're running in the SoA mode, which means that + * our storing routines will keep getting called on channels that have + * overflown. + * So we need some scratch area where we can keep writing the overflown + * vertices without overwriting anything important or crashing. + */ + gs->primitive_boundary = gs->max_output_vertices + 1; for (i = 0; i < gs->info.num_outputs; i++) { if (gs->info.output_semantic_name[i] == TGSI_SEMANTIC_POSITION && diff --git a/src/gallium/auxiliary/draw/draw_gs.h b/src/gallium/auxiliary/draw/draw_gs.h index 7c841396aa2..ca744cebfc9 100644 --- a/src/gallium/auxiliary/draw/draw_gs.h +++ b/src/gallium/auxiliary/draw/draw_gs.h @@ -68,6 +68,7 @@ struct draw_geometry_shader { unsigned position_output; unsigned max_output_vertices; + unsigned primitive_boundary; unsigned input_primitive; unsigned output_primitive; diff --git a/src/gallium/auxiliary/draw/draw_llvm.c b/src/gallium/auxiliary/draw/draw_llvm.c index 0e349783198..33fe40d5e7f 100644 --- a/src/gallium/auxiliary/draw/draw_llvm.c +++ b/src/gallium/auxiliary/draw/draw_llvm.c @@ -1287,8 +1287,8 @@ draw_gs_llvm_emit_vertex(const struct lp_build_tgsi_gs_iface *gs_base, LLVMValueRef clipmask = lp_build_const_int_vec(gallivm, lp_int_type(gs_type), 0); LLVMValueRef indices[LP_MAX_VECTOR_LENGTH]; - LLVMValueRef max_output_vertices = - lp_build_const_int32(gallivm, variant->shader->base.max_output_vertices); + LLVMValueRef next_prim_offset = + lp_build_const_int32(gallivm, variant->shader->base.primitive_boundary); LLVMValueRef io = variant->io_ptr; unsigned i; const struct tgsi_shader_info *gs_info = &variant->shader->base.info; @@ -1297,7 +1297,7 @@ draw_gs_llvm_emit_vertex(const struct lp_build_tgsi_gs_iface *gs_base, LLVMValueRef ind = lp_build_const_int32(gallivm, i); LLVMValueRef currently_emitted = LLVMBuildExtractElement(builder, emitted_vertices_vec, ind, ""); - indices[i] = LLVMBuildMul(builder, ind, max_output_vertices, ""); + indices[i] = LLVMBuildMul(builder, ind, next_prim_offset, ""); indices[i] = LLVMBuildAdd(builder, indices[i], currently_emitted, ""); } |