From 3ef0957dac11edee7babc9746ec766dcb055d909 Mon Sep 17 00:00:00 2001 From: Jordan Justen Date: Sun, 22 May 2016 00:08:06 -0700 Subject: [PATCH] i965: Add nir based intrinsic lowering and thread ID uniform We add a lowering pass for nir intrinsics. This pass can replace nir intrinsics with driver specific nir lower code. We lower the gl_LocalInvocationIndex intrinsic based on a uniform which is loaded with a thread specific ID. We also lower the gl_LocalInvocationID based on gl_LocalInvocationIndex. v2: * Create variable during lowering pass. (Ken) v3: * Don't create a variable, but instead just insert an intrisic call to load a uniform from the allocated location. (Jason) v4: * Don't run this pass if thread_local_id_index < 0 Cc: "12.0" Signed-off-by: Jordan Justen Reviewed-by: Jason Ekstrand --- src/mesa/drivers/dri/i965/Makefile.sources | 1 + src/mesa/drivers/dri/i965/brw_fs.cpp | 1 + src/mesa/drivers/dri/i965/brw_nir.h | 2 + .../drivers/dri/i965/brw_nir_intrinsics.c | 186 ++++++++++++++++++ 4 files changed, 190 insertions(+) create mode 100644 src/mesa/drivers/dri/i965/brw_nir_intrinsics.c diff --git a/src/mesa/drivers/dri/i965/Makefile.sources b/src/mesa/drivers/dri/i965/Makefile.sources index d8711edcf39..f448551cca3 100644 --- a/src/mesa/drivers/dri/i965/Makefile.sources +++ b/src/mesa/drivers/dri/i965/Makefile.sources @@ -46,6 +46,7 @@ i965_compiler_FILES = \ brw_nir.c \ brw_nir_analyze_boolean_resolves.c \ brw_nir_attribute_workarounds.c \ + brw_nir_intrinsics.c \ brw_nir_opt_peephole_ffma.c \ brw_packed_float.c \ brw_predicated_break.cpp \ diff --git a/src/mesa/drivers/dri/i965/brw_fs.cpp b/src/mesa/drivers/dri/i965/brw_fs.cpp index 6bd28718280..975ac9eb6ba 100644 --- a/src/mesa/drivers/dri/i965/brw_fs.cpp +++ b/src/mesa/drivers/dri/i965/brw_fs.cpp @@ -6599,6 +6599,7 @@ brw_compile_cs(const struct brw_compiler *compiler, void *log_data, (unsigned)4 * (prog_data->thread_local_id_index + 1)); } + brw_nir_lower_intrinsics(shader, &prog_data->base); shader = brw_postprocess_nir(shader, compiler->devinfo, true); prog_data->local_size[0] = shader->info.cs.local_size[0]; diff --git a/src/mesa/drivers/dri/i965/brw_nir.h b/src/mesa/drivers/dri/i965/brw_nir.h index 409e49a1a07..74c354f5c82 100644 --- a/src/mesa/drivers/dri/i965/brw_nir.h +++ b/src/mesa/drivers/dri/i965/brw_nir.h @@ -91,6 +91,8 @@ void brw_nir_analyze_boolean_resolves(nir_shader *nir); nir_shader *brw_preprocess_nir(const struct brw_compiler *compiler, nir_shader *nir); +bool brw_nir_lower_intrinsics(nir_shader *nir, + struct brw_stage_prog_data *prog_data); void brw_nir_lower_vs_inputs(nir_shader *nir, const struct brw_device_info *devinfo, bool is_scalar, diff --git a/src/mesa/drivers/dri/i965/brw_nir_intrinsics.c b/src/mesa/drivers/dri/i965/brw_nir_intrinsics.c new file mode 100644 index 00000000000..972b1171730 --- /dev/null +++ b/src/mesa/drivers/dri/i965/brw_nir_intrinsics.c @@ -0,0 +1,186 @@ +/* + * Copyright (c) 2016 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. + */ + +#include "brw_nir.h" +#include "compiler/nir/nir_builder.h" + +struct lower_intrinsics_state { + nir_shader *nir; + union { + struct brw_stage_prog_data *prog_data; + struct brw_cs_prog_data *cs_prog_data; + }; + nir_function_impl *impl; + bool progress; + nir_builder builder; + bool cs_thread_id_used; +}; + +static nir_ssa_def * +read_thread_local_id(struct lower_intrinsics_state *state) +{ + assert(state->cs_prog_data->thread_local_id_index >= 0); + state->cs_thread_id_used = true; + const int id_index = state->cs_prog_data->thread_local_id_index; + + nir_builder *b = &state->builder; + nir_shader *nir = state->nir; + nir_intrinsic_instr *load = + nir_intrinsic_instr_create(nir, nir_intrinsic_load_uniform); + load->num_components = 1; + load->src[0] = nir_src_for_ssa(nir_imm_int(b, 0)); + nir_ssa_dest_init(&load->instr, &load->dest, 1, 32, NULL); + nir_intrinsic_set_base(load, id_index * sizeof(uint32_t)); + nir_intrinsic_set_range(load, sizeof(uint32_t)); + nir_builder_instr_insert(b, &load->instr); + return &load->dest.ssa; +} + +static bool +lower_cs_intrinsics_convert_block(struct lower_intrinsics_state *state, + nir_block *block) +{ + bool progress = false; + nir_builder *b = &state->builder; + nir_shader *nir = state->nir; + + nir_foreach_instr_safe(instr, block) { + if (instr->type != nir_instr_type_intrinsic) + continue; + + nir_intrinsic_instr *intrinsic = nir_instr_as_intrinsic(instr); + + b->cursor = nir_after_instr(&intrinsic->instr); + + nir_ssa_def *sysval; + switch (intrinsic->intrinsic) { + case nir_intrinsic_load_local_invocation_index: { + assert(nir->stage == MESA_SHADER_COMPUTE); + /* We construct the local invocation index from: + * + * gl_LocalInvocationIndex = + * cs_thread_local_id + channel_num; + */ + nir_ssa_def *thread_local_id = read_thread_local_id(state); + nir_ssa_def *channel = + nir_load_system_value(b, nir_intrinsic_load_channel_num, 0); + sysval = nir_iadd(b, channel, thread_local_id); + break; + } + + case nir_intrinsic_load_local_invocation_id: { + assert(nir->stage == MESA_SHADER_COMPUTE); + /* We lower gl_LocalInvocationID from gl_LocalInvocationIndex based + * on this formula: + * + * gl_LocalInvocationID.x = + * gl_LocalInvocationIndex % gl_WorkGroupSize.x; + * gl_LocalInvocationID.y = + * (gl_LocalInvocationIndex / gl_WorkGroupSize.x) % + * gl_WorkGroupSize.y; + * gl_LocalInvocationID.z = + * (gl_LocalInvocationIndex / + * (gl_WorkGroupSize.x * gl_WorkGroupSize.y)) % + * gl_WorkGroupSize.z; + */ + unsigned *size = nir->info.cs.local_size; + + nir_ssa_def *local_index = + nir_load_system_value(b, nir_intrinsic_load_local_invocation_index, 0); + + nir_const_value uvec3; + uvec3.u32[0] = 1; + uvec3.u32[1] = size[0]; + uvec3.u32[2] = size[0] * size[1]; + nir_ssa_def *div_val = nir_build_imm(b, 3, 32, uvec3); + uvec3.u32[0] = size[0]; + uvec3.u32[1] = size[1]; + uvec3.u32[2] = size[2]; + nir_ssa_def *mod_val = nir_build_imm(b, 3, 32, uvec3); + + sysval = nir_imod(b, nir_idiv(b, local_index, div_val), mod_val); + break; + } + + default: + continue; + } + + nir_ssa_def_rewrite_uses(&intrinsic->dest.ssa, nir_src_for_ssa(sysval)); + nir_instr_remove(&intrinsic->instr); + + state->progress = true; + } + + return progress; +} + +static void +lower_cs_intrinsics_convert_impl(struct lower_intrinsics_state *state) +{ + nir_builder_init(&state->builder, state->impl); + + nir_foreach_block(block, state->impl) { + lower_cs_intrinsics_convert_block(state, block); + } + + nir_metadata_preserve(state->impl, + nir_metadata_block_index | nir_metadata_dominance); +} + +bool +brw_nir_lower_intrinsics(nir_shader *nir, struct brw_stage_prog_data *prog_data) +{ + /* Currently we only lower intrinsics for compute shaders */ + if (nir->stage != MESA_SHADER_COMPUTE) + return false; + + bool progress = false; + struct lower_intrinsics_state state; + memset(&state, 0, sizeof(state)); + state.nir = nir; + state.prog_data = prog_data; + + /* Currently this pass only lowers intrinsics using the uniform specified + * by thread_local_id_index. + */ + if (nir->stage == MESA_SHADER_COMPUTE && + state.cs_prog_data->thread_local_id_index < 0) + return false; + + do { + state.progress = false; + nir_foreach_function(function, nir) { + if (function->impl) { + state.impl = function->impl; + lower_cs_intrinsics_convert_impl(&state); + } + } + progress |= state.progress; + } while (state.progress); + + if (nir->stage == MESA_SHADER_COMPUTE && !state.cs_thread_id_used) + state.cs_prog_data->thread_local_id_index = -1; + + return progress; +} -- 2.30.2