From: Eric Anholt Date: Fri, 4 May 2012 20:08:46 +0000 (-0700) Subject: glsl: Implement the GLSL 1.30+ discard control flow rule in GLSL IR. X-Git-Url: https://git.libre-soc.org/?a=commitdiff_plain;h=3de1395fa5a563c13774ac1c38722c16cecc521d;p=mesa.git glsl: Implement the GLSL 1.30+ discard control flow rule in GLSL IR. Previously, I tried implementing this in the i965 driver, but did so in a way that violated the intent of the spec, and broke Tropics. Reviewed-by: Kenneth Graunke --- diff --git a/src/glsl/Makefile.sources b/src/glsl/Makefile.sources index b8f7e5f46eb..fba03604cc5 100644 --- a/src/glsl/Makefile.sources +++ b/src/glsl/Makefile.sources @@ -51,6 +51,7 @@ LIBGLSL_CXX_FILES := \ loop_unroll.cpp \ lower_clip_distance.cpp \ lower_discard.cpp \ + lower_discard_flow.cpp \ lower_if_to_cond_assign.cpp \ lower_instructions.cpp \ lower_jumps.cpp \ diff --git a/src/glsl/ir_optimization.h b/src/glsl/ir_optimization.h index 35678358307..cca63824d05 100644 --- a/src/glsl/ir_optimization.h +++ b/src/glsl/ir_optimization.h @@ -66,6 +66,7 @@ bool do_tree_grafting(exec_list *instructions); bool do_vec_index_to_cond_assign(exec_list *instructions); bool do_vec_index_to_swizzle(exec_list *instructions); bool lower_discard(exec_list *instructions); +bool lower_discard_flow(exec_list *instructions); bool lower_instructions(exec_list *instructions, unsigned what_to_lower); bool lower_noise(exec_list *instructions); bool lower_variable_index_to_cond_assign(exec_list *instructions, diff --git a/src/glsl/linker.cpp b/src/glsl/linker.cpp index 6ba297237c7..bdab499f046 100644 --- a/src/glsl/linker.cpp +++ b/src/glsl/linker.cpp @@ -2296,6 +2296,17 @@ link_shaders(struct gl_context *ctx, struct gl_shader_program *prog) prog->LinkStatus = true; } + /* Implement the GLSL 1.30+ rule for discard vs infinite loops Do + * it before optimization because we want most of the checks to get + * dropped thanks to constant propagation. + */ + if (max_version >= 130) { + struct gl_shader *sh = prog->_LinkedShaders[MESA_SHADER_FRAGMENT]; + if (sh) { + lower_discard_flow(sh->ir); + } + } + /* Do common optimization before assigning storage for attributes, * uniforms, and varyings. Later optimization could possibly make * some of that unused. diff --git a/src/glsl/lower_discard_flow.cpp b/src/glsl/lower_discard_flow.cpp new file mode 100644 index 00000000000..d385c143547 --- /dev/null +++ b/src/glsl/lower_discard_flow.cpp @@ -0,0 +1,144 @@ +/* + * Copyright © 2010 Intel Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ + +/** @file lower_discard_flow.cpp + * + * Implements the GLSL 1.30 revision 9 rule for fragment shader + * discard handling: + * + * "Control flow exits the shader, and subsequent implicit or + * explicit derivatives are undefined when this control flow is + * non-uniform (meaning different fragments within the primitive + * take different control paths)." + * + * There seem to be two conflicting things here. "Control flow exits + * the shader" sounds like the discarded fragments should effectively + * jump to the end of the shader, but that breaks derivatives in the + * case of uniform control flow and causes rendering failure in the + * bushes in Unigine Tropics. + * + * The question, then, is whether the intent was "loops stop at the + * point that the only active channels left are discarded pixels" or + * "discarded pixels become inactive at the point that control flow + * returns to the top of a loop". This implements the second + * interpretation. + */ + +#include "glsl_types.h" +#include "ir.h" +#include "program/hash_table.h" + +class lower_discard_flow_visitor : public ir_hierarchical_visitor { +public: + lower_discard_flow_visitor(ir_variable *discarded) + : discarded(discarded) + { + mem_ctx = ralloc_parent(discarded); + } + + ~lower_discard_flow_visitor() + { + } + + ir_visitor_status visit_enter(ir_discard *ir); + ir_visitor_status visit_enter(ir_loop_jump *ir); + ir_visitor_status visit_enter(ir_loop *ir); + ir_visitor_status visit_enter(ir_function_signature *ir); + + ir_if *generate_discard_break(); + + ir_variable *discarded; + void *mem_ctx; +}; + +ir_visitor_status +lower_discard_flow_visitor::visit_enter(ir_loop_jump *ir) +{ + if (ir->mode != ir_loop_jump::jump_continue) + return visit_continue; + + ir->insert_before(generate_discard_break()); + + return visit_continue; +} + +ir_visitor_status +lower_discard_flow_visitor::visit_enter(ir_discard *ir) +{ + ir_dereference *lhs = new(mem_ctx) ir_dereference_variable(discarded); + ir_rvalue *rhs = new(mem_ctx) ir_constant(true); + ir_assignment *assign = new(mem_ctx) ir_assignment(lhs, rhs); + ir->insert_before(assign); + + return visit_continue; +} + +ir_visitor_status +lower_discard_flow_visitor::visit_enter(ir_loop *ir) +{ + ir->body_instructions.push_tail(generate_discard_break()); + + return visit_continue; +} + +ir_visitor_status +lower_discard_flow_visitor::visit_enter(ir_function_signature *ir) +{ + if (strcmp(ir->function_name(), "main") != 0) + return visit_continue; + + ir_dereference *lhs = new(mem_ctx) ir_dereference_variable(discarded); + ir_rvalue *rhs = new(mem_ctx) ir_constant(false); + ir_assignment *assign = new(mem_ctx) ir_assignment(lhs, rhs); + ir->body.push_head(assign); + + return visit_continue; +} + +ir_if * +lower_discard_flow_visitor::generate_discard_break() +{ + ir_rvalue *if_condition = new(mem_ctx) ir_dereference_variable(discarded); + ir_if *if_inst = new(mem_ctx) ir_if(if_condition); + + ir_instruction *br = new(mem_ctx) ir_loop_jump(ir_loop_jump::jump_break); + if_inst->then_instructions.push_tail(br); + + return if_inst; +} + +void +lower_discard_flow(exec_list *ir) +{ + void *mem_ctx = ir; + + ir_variable *var = new(mem_ctx) ir_variable(glsl_type::bool_type, + "discarded", + ir_var_temporary); + + ir->push_head(var); + + lower_discard_flow_visitor v(var); + + visit_list_elements(&v, ir); +}