1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239
|
/* Copyright (c) 2024-2025 LunarG, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "descriptor_class_general_buffer_pass.h"
#include "generated/spirv_grammar_helper.h"
#include "containers/container_utils.h"
#include "state_tracker/shader_instruction.h"
#include "module.h"
#include <spirv/unified1/spirv.hpp>
#include <iostream>
#include "generated/device_features.h"
#include "generated/gpuav_offline_spirv.h"
#include "gpuav/shaders/gpuav_shaders_constants.h"
namespace gpuav {
namespace spirv {
const static OfflineModule kOfflineModule = {instrumentation_descriptor_class_general_buffer_comp,
instrumentation_descriptor_class_general_buffer_comp_size, UseErrorPayloadVariable};
const static OfflineFunction kOfflineFunction = {"inst_descriptor_class_general_buffer",
instrumentation_descriptor_class_general_buffer_comp_function_0_offset};
DescriptorClassGeneralBufferPass::DescriptorClassGeneralBufferPass(Module& module) : Pass(module, kOfflineModule) {
module.use_bda_ = true;
}
// By appending the LinkInfo, it will attempt at linking stage to add the function.
uint32_t DescriptorClassGeneralBufferPass::GetLinkFunctionId() { return GetLinkFunction(link_function_id_, kOfflineFunction); }
void DescriptorClassGeneralBufferPass::CreateFunctionCall(BasicBlock& block, InstructionIt* inst_it, const InstructionMeta& meta) {
assert(!meta.access_chain_insts.empty());
const Constant& set_constant = module_.type_manager_.GetConstantUInt32(meta.descriptor_set);
const uint32_t descriptor_index_id = CastToUint32(meta.descriptor_index_id, block, inst_it); // might be int32
const uint32_t descriptor_offset_id =
GetLastByte(*meta.descriptor_type, meta.access_chain_insts, block, inst_it); // Get Last Byte Index
BindingLayout binding_layout = module_.set_index_to_bindings_layout_lut_[meta.descriptor_set][meta.descriptor_binding];
const Constant& binding_layout_offset = module_.type_manager_.GetConstantUInt32(binding_layout.start);
const uint32_t inst_position = meta.target_instruction->GetPositionIndex();
const uint32_t inst_position_id = module_.type_manager_.CreateConstantUInt32(inst_position).Id();
const uint32_t function_result = module_.TakeNextId();
const uint32_t function_def = GetLinkFunctionId();
const uint32_t void_type = module_.type_manager_.GetTypeVoid().Id();
block.CreateInstruction(spv::OpFunctionCall,
{void_type, function_result, function_def, inst_position_id, set_constant.Id(), descriptor_index_id,
descriptor_offset_id, binding_layout_offset.Id()},
inst_it);
module_.need_log_error_ = true;
}
bool DescriptorClassGeneralBufferPass::RequiresInstrumentation(const Function& function, const Instruction& inst,
InstructionMeta& meta) {
const uint32_t opcode = inst.Opcode();
if (!IsValueIn(spv::Op(opcode), {spv::OpLoad, spv::OpStore, spv::OpAtomicStore, spv::OpAtomicLoad, spv::OpAtomicExchange})) {
return false;
}
const Instruction* next_access_chain = function.FindInstruction(inst.Operand(0));
if (!next_access_chain || !next_access_chain->IsNonPtrAccessChain()) {
return false;
}
const Variable* variable = nullptr;
// We need to walk down possibly multiple chained OpAccessChains or OpCopyObject to get the variable
while (next_access_chain && next_access_chain->IsNonPtrAccessChain()) {
meta.access_chain_insts.push_back(next_access_chain);
const uint32_t access_chain_base_id = next_access_chain->Operand(0);
variable = module_.type_manager_.FindVariableById(access_chain_base_id);
if (variable) {
break; // found
}
next_access_chain = function.FindInstruction(access_chain_base_id);
}
if (!variable) {
return false;
}
uint32_t storage_class = variable->StorageClass();
if (storage_class != spv::StorageClassUniform && storage_class != spv::StorageClassStorageBuffer) {
return false;
}
meta.descriptor_type = variable->PointerType(module_.type_manager_);
if (!meta.descriptor_type || meta.descriptor_type->spv_type_ == SpvType::kRuntimeArray) {
return false; // TODO - Currently we mark these as "bindless"
}
const bool is_descriptor_array = meta.descriptor_type->IsArray();
meta.descriptor_id = is_descriptor_array ? meta.descriptor_type->inst_.Operand(0) : meta.descriptor_type->Id();
// Check for deprecated storage block form
if (storage_class == spv::StorageClassUniform) {
assert(module_.type_manager_.FindTypeById(meta.descriptor_id)->spv_type_ == SpvType::kStruct && "unexpected block type");
const bool block_found = GetDecoration(meta.descriptor_id, spv::DecorationBlock) != nullptr;
// If block decoration not found, verify deprecated form of SSBO
if (!block_found) {
assert(GetDecoration(meta.descriptor_id, spv::DecorationBufferBlock) != nullptr && "block decoration not found");
storage_class = spv::StorageClassStorageBuffer;
}
}
// Grab front() as it will be the "final" type we access
const Type* value_type = module_.type_manager_.FindValueTypeById(meta.access_chain_insts.front()->TypeId());
if (!value_type) return false;
if (is_descriptor_array) {
// Because you can't have 2D array of descriptors, the first index of the last accessChain is the descriptor index
meta.descriptor_index_id = meta.access_chain_insts.back()->Operand(1);
} else {
// There is no array of this descriptor, so we essentially have an array of 1
meta.descriptor_index_id = module_.type_manager_.GetConstantZeroUint32().Id();
}
for (const auto& annotation : module_.annotations_) {
if (annotation->Opcode() == spv::OpDecorate && annotation->Word(1) == variable->Id()) {
if (annotation->Word(2) == spv::DecorationDescriptorSet) {
meta.descriptor_set = annotation->Word(3);
} else if (annotation->Word(2) == spv::DecorationBinding) {
meta.descriptor_binding = annotation->Word(3);
}
}
}
if (meta.descriptor_set >= glsl::kDebugInputBindlessMaxDescSets) {
module_.InternalWarning(Name(), "Tried to use a descriptor slot over the current max limit");
return false;
}
if (!module_.settings_.safe_mode) {
meta.access_offset = FindOffsetInStruct(meta.descriptor_id, is_descriptor_array, meta.access_chain_insts);
}
// Save information to be used to make the Function
meta.target_instruction = &inst;
return true;
}
void DescriptorClassGeneralBufferPass::PrintDebugInfo() const {
std::cout << "DescriptorClassGeneralBufferPass instrumentation count: " << instrumentations_count_ << '\n';
}
// Created own Instrument() because need to control finding the largest offset in a given block
bool DescriptorClassGeneralBufferPass::Instrument() {
if (module_.set_index_to_bindings_layout_lut_.empty()) {
return false; // If there is no bindings, nothing to instrument
}
// Can safely loop function list as there is no injecting of new Functions until linking time
for (const auto& function : module_.functions_) {
if (function->instrumentation_added_) continue;
for (auto block_it = function->blocks_.begin(); block_it != function->blocks_.end(); ++block_it) {
BasicBlock& current_block = **block_it;
cf_.Update(current_block);
if (debug_disable_loops_ && cf_.in_loop) continue;
auto& block_instructions = current_block.instructions_;
// < Descriptor SSA ID, Highest offset byte that will be accessed >
vvl::unordered_map<uint32_t, uint32_t> block_highest_offset_map;
if (!module_.settings_.safe_mode) {
// Pre-pass loop the Block to get the highest offset accessed (statically known)
// Do here before we inject instructions into the block list below
for (auto inst_it = block_instructions.begin(); inst_it != block_instructions.end(); ++inst_it) {
InstructionMeta meta;
// Every instruction is analyzed by the specific pass and lets us know if we need to inject a function or not
if (!RequiresInstrumentation(*function, *(inst_it->get()), meta)) continue;
if (meta.access_offset != 0) {
// set offset for the first loop of the block
auto map_it = block_highest_offset_map.find(meta.descriptor_id);
if (map_it == block_highest_offset_map.end()) {
block_highest_offset_map[meta.descriptor_id] = meta.access_offset;
} else {
map_it->second = std::max(map_it->second, meta.access_offset);
}
}
}
}
for (auto inst_it = block_instructions.begin(); inst_it != block_instructions.end(); ++inst_it) {
InstructionMeta meta;
// Every instruction is analyzed by the specific pass and lets us know if we need to inject a function or not
if (!RequiresInstrumentation(*function, *(inst_it->get()), meta)) continue;
if (!module_.settings_.safe_mode && meta.access_offset != 0) {
const uint32_t block_highest_offset = block_highest_offset_map[meta.descriptor_id];
if (meta.access_offset < block_highest_offset) {
continue; // skipping because other instruction in block will be a higher offset
}
}
if (IsMaxInstrumentationsCount()) continue;
instrumentations_count_++;
// inst_it is updated to the instruction after the new function call, it will not add/remove any Blocks
CreateFunctionCall(current_block, &inst_it, meta);
}
}
}
if (instrumentations_count_ > 75) {
module_.InternalWarning(
"GPUAV-Compile-time-general-buffer",
"This shader will be very slow to compile and runtime performance may also be slow. This is due to the number of OOB "
"checks for storage/uniform "
"buffers. Turn on the |gpuav_force_on_robustness| setting to skip these checks and improve GPU-AV performance.");
}
return instrumentations_count_ != 0;
}
} // namespace spirv
} // namespace gpuav
|