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_IR))
  49       return;
  50
  51    bool is_valid = aco::validate_ir(program);
  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    program->debug.func = args->options->debug.func;
  70    program->debug.private_data = args->options->debug.private_data;
  71
  72    /* Instruction Selection */
  73    if (args->is_gs_copy_shader)
  74       aco::select_gs_copy_shader(program.get(), shaders[0], &config, args);
  75    else
  76       aco::select_program(program.get(), shader_count, shaders, &config, args);
  77    if (args->options->dump_preoptir) {
  78       std::cerr << "After Instruction Selection:\n";
  79       aco_print_program(program.get(), stderr);
  80    }
  81
  82    aco::live live_vars;
  83    if (!args->is_trap_handler_shader) {
  84       /* Phi lowering */
  85       aco::lower_phis(program.get());
  86       aco::dominator_tree(program.get());
  87       validate(program.get());
  88
  89       /* Optimization */
  90       aco::value_numbering(program.get());
  91       aco::optimize(program.get());
  92
  93       /* cleanup and exec mask handling */
  94       aco::setup_reduce_temp(program.get());
  95       aco::insert_exec_mask(program.get());
  96       validate(program.get());
  97
  98       /* spilling and scheduling */
  99       live_vars = aco::live_var_analysis(program.get(), args->options);
 100       aco::spill(program.get(), live_vars, args->options);
 101    }
 102
 103    std::string llvm_ir;
 104    if (args->options->record_ir) {
 105       char *data = NULL;
 106       size_t size = 0;
 107       FILE *f = open_memstream(&data, &size);
 108       if (f) {
 109          aco_print_program(program.get(), f);
 110          fputc(0, f);
 111          fclose(f);
 112       }
 113
 114       llvm_ir = std::string(data, data + size);
 115       free(data);
 116    }
 117
 118    if (program->collect_statistics)
 119       aco::collect_presched_stats(program.get());
 120
 121    if (!args->is_trap_handler_shader) {
 122       aco::schedule_program(program.get(), live_vars);
 123       validate(program.get());
 124
 125       /* Register Allocation */
 126       aco::register_allocation(program.get(), live_vars.live_out);
 127       if (args->options->dump_shader) {
 128          std::cerr << "After RA:\n";
 129          aco_print_program(program.get(), stderr);
 130       }
 131
 132       if (aco::validate_ra(program.get(), args->options)) {
 133          std::cerr << "Program after RA validation failure:\n";
 134          aco_print_program(program.get(), stderr);
 135          abort();
 136       }
 137
 138       validate(program.get());
 139
 140       aco::ssa_elimination(program.get());
 141    }
 142
 143    /* Lower to HW Instructions */
 144    aco::lower_to_hw_instr(program.get());
 145
 146    /* Insert Waitcnt */
 147    aco::insert_wait_states(program.get());
 148    aco::insert_NOPs(program.get());
 149
 150    if (program->collect_statistics)
 151       aco::collect_preasm_stats(program.get());
 152
 153    /* Assembly */
 154    std::vector<uint32_t> code;
 155    unsigned exec_size = aco::emit_program(program.get(), code);
 156
 157    if (program->collect_statistics)
 158       aco::collect_postasm_stats(program.get(), code);
 159
 160    bool get_disasm = args->options->dump_shader || args->options->record_ir;
 161
 162    size_t size = llvm_ir.size();
 163
 164    std::string disasm;
 165    if (get_disasm) {
 166       std::ostringstream stream;
 167       aco::print_asm(program.get(), code, exec_size / 4u, stream);
 168       stream << '\0';
 169       disasm = stream.str();
 170       size += disasm.size();
 171    }
 172
 173    size_t stats_size = 0;
 174    if (program->collect_statistics)
 175       stats_size = sizeof(radv_compiler_statistics) + aco::num_statistics * sizeof(uint32_t);
 176    size += stats_size;
 177
 178    size += code.size() * sizeof(uint32_t) + sizeof(radv_shader_binary_legacy);
 179    /* We need to calloc to prevent unintialized data because this will be used
 180     * directly for the disk cache. Uninitialized data can appear because of
 181     * padding in the struct or because legacy_binary->data can be at an offset
 182     * from the start less than sizeof(radv_shader_binary_legacy). */
 183    radv_shader_binary_legacy* legacy_binary = (radv_shader_binary_legacy*) calloc(size, 1);
 184
 185    legacy_binary->base.type = RADV_BINARY_TYPE_LEGACY;
 186    legacy_binary->base.stage = shaders[shader_count-1]->info.stage;
 187    legacy_binary->base.is_gs_copy_shader = args->is_gs_copy_shader;
 188    legacy_binary->base.total_size = size;
 189
 190    if (program->collect_statistics) {
 191       radv_compiler_statistics *statistics = (radv_compiler_statistics *)legacy_binary->data;
 192       statistics->count = aco::num_statistics;
 193       statistics->infos = statistic_infos;
 194       memcpy(statistics->values, program->statistics, aco::num_statistics * sizeof(uint32_t));
 195    }
 196    legacy_binary->stats_size = stats_size;
 197
 198    memcpy(legacy_binary->data + legacy_binary->stats_size, code.data(), code.size() * sizeof(uint32_t));
 199    legacy_binary->exec_size = exec_size;
 200    legacy_binary->code_size = code.size() * sizeof(uint32_t);
 201
 202    legacy_binary->config = config;
 203    legacy_binary->disasm_size = 0;
 204    legacy_binary->ir_size = llvm_ir.size();
 205
 206    llvm_ir.copy((char*) legacy_binary->data + legacy_binary->stats_size + legacy_binary->code_size, llvm_ir.size());
 207
 208    if (get_disasm) {
 209       disasm.copy((char*) legacy_binary->data + legacy_binary->stats_size + legacy_binary->code_size + llvm_ir.size(), disasm.size());
 210       legacy_binary->disasm_size = disasm.size();
 211    }
 212
 213    *binary = (radv_shader_binary*) legacy_binary;
 214 }