src/amd/compiler/aco_interface.cpp

   1 /*
   2  * Copyright © 2018 Google
   3  *
   4  * Permission is hereby granted, free of charge, to any person obtaining a
   5  * copy of this software and associated documentation files (the "Software"),
   6  * to deal in the Software without restriction, including without limitation
   7  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
   8  * and/or sell copies of the Software, and to permit persons to whom the
   9  * Software is furnished to do so, subject to the following conditions:
  10  *
  11  * The above copyright notice and this permission notice (including the next
  12  * paragraph) shall be included in all copies or substantial portions of the
  13  * Software.
  14  *
  15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  16  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  17  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
  18  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  19  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
  20  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
  21  * IN THE SOFTWARE.
  22  */
  23
  24 #include "aco_interface.h"
  25 #include "aco_ir.h"
  26 #include "vulkan/radv_shader.h"
  27 #include "vulkan/radv_shader_args.h"
  28
  29 #include <iostream>
  30 #include <sstream>
  31
  32 static radv_compiler_statistic_info statistic_infos[] = {
  33    [aco::statistic_hash] = {"Hash", "CRC32 hash of code and constant data"},
  34    [aco::statistic_instructions] = {"Instructions", "Instruction count"},
  35    [aco::statistic_copies] = {"Copies", "Copy instructions created for pseudo-instructions"},
  36    [aco::statistic_branches] = {"Branches", "Branch instructions"},
  37    [aco::statistic_cycles] = {"Busy Cycles", "Estimate of busy cycles"},
  38    [aco::statistic_vmem_clauses] = {"VMEM Clause", "Number of VMEM clauses (includes 1-sized clauses)"},
  39    [aco::statistic_smem_clauses] = {"SMEM Clause", "Number of SMEM clauses (includes 1-sized clauses)"},
  40    [aco::statistic_vmem_score] = {"VMEM Score", "Average VMEM def-use distances"},
  41    [aco::statistic_smem_score] = {"SMEM Score", "Average SMEM def-use distances"},
  42    [aco::statistic_sgpr_presched] = {"Pre-Sched SGPRs", "SGPR usage before scheduling"},
  43    [aco::statistic_vgpr_presched] = {"Pre-Sched VGPRs", "VGPR usage before scheduling"},
  44 };
  45
  46 static void validate(aco::Program *program)
  47 {
  48    if (!(aco::debug_flags & aco::DEBUG_VALIDATE))
  49       return;
  50
  51    bool is_valid = aco::validate(program, stderr);
  52    assert(is_valid);
  53 }
  54
  55 void aco_compile_shader(unsigned shader_count,
  56                         struct nir_shader *const *shaders,
  57                         struct radv_shader_binary **binary,
  58                         struct radv_shader_args *args)
  59 {
  60    aco::init();
  61
  62    ac_shader_config config = {0};
  63    std::unique_ptr<aco::Program> program{new aco::Program};
  64
  65    program->collect_statistics = args->options->record_stats;
  66    if (program->collect_statistics)
  67       memset(program->statistics, 0, sizeof(program->statistics));
  68
  69    /* Instruction Selection */
  70    if (args->is_gs_copy_shader)
  71       aco::select_gs_copy_shader(program.get(), shaders[0], &config, args);
  72    else
  73       aco::select_program(program.get(), shader_count, shaders, &config, args);
  74    if (args->options->dump_preoptir) {
  75       std::cerr << "After Instruction Selection:\n";
  76       aco_print_program(program.get(), stderr);
  77    }
  78
  79    /* Phi lowering */
  80    aco::lower_phis(program.get());
  81    aco::dominator_tree(program.get());
  82    validate(program.get());
  83
  84    /* Optimization */
  85    aco::value_numbering(program.get());
  86    aco::optimize(program.get());
  87
  88    /* cleanup and exec mask handling */
  89    aco::setup_reduce_temp(program.get());
  90    aco::insert_exec_mask(program.get());
  91    validate(program.get());
  92
  93    /* spilling and scheduling */
  94    aco::live live_vars = aco::live_var_analysis(program.get(), args->options);
  95    aco::spill(program.get(), live_vars, args->options);
  96
  97    std::string llvm_ir;
  98    if (args->options->record_ir) {
  99       char *data = NULL;
 100       size_t size = 0;
 101       FILE *f = open_memstream(&data, &size);
 102       if (f) {
 103          aco_print_program(program.get(), f);
 104          fputc(0, f);
 105          fclose(f);
 106       }
 107
 108       llvm_ir = std::string(data, data + size);
 109       free(data);
 110    }
 111
 112    if (program->collect_statistics)
 113       aco::collect_presched_stats(program.get());
 114    aco::schedule_program(program.get(), live_vars);
 115    validate(program.get());
 116
 117    /* Register Allocation */
 118    aco::register_allocation(program.get(), live_vars.live_out);
 119    if (args->options->dump_shader) {
 120       std::cerr << "After RA:\n";
 121       aco_print_program(program.get(), stderr);
 122    }
 123
 124    if (aco::validate_ra(program.get(), args->options, stderr)) {
 125       std::cerr << "Program after RA validation failure:\n";
 126       aco_print_program(program.get(), stderr);
 127       abort();
 128    }
 129
 130    validate(program.get());
 131
 132    /* Lower to HW Instructions */
 133    aco::ssa_elimination(program.get());
 134    aco::lower_to_hw_instr(program.get());
 135
 136    /* Insert Waitcnt */
 137    aco::insert_wait_states(program.get());
 138    aco::insert_NOPs(program.get());
 139
 140    if (program->collect_statistics)
 141       aco::collect_preasm_stats(program.get());
 142
 143    /* Assembly */
 144    std::vector<uint32_t> code;
 145    unsigned exec_size = aco::emit_program(program.get(), code);
 146
 147    if (program->collect_statistics)
 148       aco::collect_postasm_stats(program.get(), code);
 149
 150    bool get_disasm = args->options->dump_shader || args->options->record_ir;
 151
 152    size_t size = llvm_ir.size();
 153
 154    std::string disasm;
 155    if (get_disasm) {
 156       std::ostringstream stream;
 157       aco::print_asm(program.get(), code, exec_size / 4u, stream);
 158       stream << '\0';
 159       disasm = stream.str();
 160       size += disasm.size();
 161    }
 162
 163    size_t stats_size = 0;
 164    if (program->collect_statistics)
 165       stats_size = sizeof(radv_compiler_statistics) + aco::num_statistics * sizeof(uint32_t);
 166    size += stats_size;
 167
 168    size += code.size() * sizeof(uint32_t) + sizeof(radv_shader_binary_legacy);
 169    /* We need to calloc to prevent unintialized data because this will be used
 170     * directly for the disk cache. Uninitialized data can appear because of
 171     * padding in the struct or because legacy_binary->data can be at an offset
 172     * from the start less than sizeof(radv_shader_binary_legacy). */
 173    radv_shader_binary_legacy* legacy_binary = (radv_shader_binary_legacy*) calloc(size, 1);
 174
 175    legacy_binary->base.type = RADV_BINARY_TYPE_LEGACY;
 176    legacy_binary->base.stage = shaders[shader_count-1]->info.stage;
 177    legacy_binary->base.is_gs_copy_shader = args->is_gs_copy_shader;
 178    legacy_binary->base.total_size = size;
 179
 180    if (program->collect_statistics) {
 181       radv_compiler_statistics *statistics = (radv_compiler_statistics *)legacy_binary->data;
 182       statistics->count = aco::num_statistics;
 183       statistics->infos = statistic_infos;
 184       memcpy(statistics->values, program->statistics, aco::num_statistics * sizeof(uint32_t));
 185    }
 186    legacy_binary->stats_size = stats_size;
 187
 188    memcpy(legacy_binary->data + legacy_binary->stats_size, code.data(), code.size() * sizeof(uint32_t));
 189    legacy_binary->exec_size = exec_size;
 190    legacy_binary->code_size = code.size() * sizeof(uint32_t);
 191
 192    legacy_binary->config = config;
 193    legacy_binary->disasm_size = 0;
 194    legacy_binary->ir_size = llvm_ir.size();
 195
 196    llvm_ir.copy((char*) legacy_binary->data + legacy_binary->stats_size + legacy_binary->code_size, llvm_ir.size());
 197
 198    if (get_disasm) {
 199       disasm.copy((char*) legacy_binary->data + legacy_binary->stats_size + legacy_binary->code_size + llvm_ir.size(), disasm.size());
 200       legacy_binary->disasm_size = disasm.size();
 201    }
 202
 203    *binary = (radv_shader_binary*) legacy_binary;
 204 }