SPIRV-Tools/source/opt/local_single_store_elim_pass.cpp
Steven Perron 2cb589cc14 Remove uses DCEInst and call ADCE
The algorithm used in DCEInst to remove dead code is very slow.  It is
fine if you only want to remove a small number of instructions, but, if
you need to remove a large number of instructions, then the algorithm in
ADCE is much faster.

This PR removes the calls to DCEInst in the load-store removal passes
and adds a pass of ADCE afterwards.

A number of different iterations of the order of optimization, and I
believe this is the best I could find.

The results I have on 3 sets of shaders are:

Legalization:

Set 1: 5.39 -> 5.01
Set 2: 13.98 -> 8.38
Set 3: 98.00 -> 96.26

Performance passes:

Set 1: 6.90 -> 5.23
Set 2: 10.11 -> 6.62
Set 3: 253.69 -> 253.74

Size reduction passes:

Set 1: 7.16 -> 7.25
Set 2: 17.17 -> 16.81
Set 3: 112.06 -> 107.71

Note that the third set's compile time is large because of the large
number of basic blocks, not so much because of the number of
instructions.  That is why we don't see much gain there.
2018-02-27 21:06:08 -05:00

315 lines
11 KiB
C++

// Copyright (c) 2017 The Khronos Group Inc.
// Copyright (c) 2017 Valve Corporation
// Copyright (c) 2017 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 "local_single_store_elim_pass.h"
#include "cfa.h"
#include "iterator.h"
#include "latest_version_glsl_std_450_header.h"
namespace spvtools {
namespace opt {
namespace {
const uint32_t kStoreValIdInIdx = 1;
const uint32_t kVariableInitIdInIdx = 1;
} // anonymous namespace
bool LocalSingleStoreElimPass::HasOnlySupportedRefs(uint32_t ptrId) {
if (supported_ref_ptrs_.find(ptrId) != supported_ref_ptrs_.end()) return true;
if (get_def_use_mgr()->WhileEachUser(ptrId, [this](ir::Instruction* user) {
SpvOp op = user->opcode();
if (IsNonPtrAccessChain(op) || op == SpvOpCopyObject) {
if (!HasOnlySupportedRefs(user->result_id())) {
return false;
}
} else if (op != SpvOpStore && op != SpvOpLoad && op != SpvOpName &&
!IsNonTypeDecorate(op)) {
return false;
}
return true;
})) {
supported_ref_ptrs_.insert(ptrId);
return true;
}
return false;
}
void LocalSingleStoreElimPass::SingleStoreAnalyze(ir::Function* func) {
ssa_var2store_.clear();
non_ssa_vars_.clear();
store2idx_.clear();
store2blk_.clear();
for (auto bi = func->begin(); bi != func->end(); ++bi) {
uint32_t instIdx = 0;
for (auto ii = bi->begin(); ii != bi->end(); ++ii, ++instIdx) {
uint32_t varId = 0;
ir::Instruction* ptrInst = nullptr;
switch (ii->opcode()) {
case SpvOpStore: {
ptrInst = GetPtr(&*ii, &varId);
} break;
case SpvOpVariable: {
// If initializer, treat like store
if (ii->NumInOperands() > 1) {
varId = ii->result_id();
ptrInst = &*ii;
}
} break;
default:
break;
} // switch
if (varId == 0) continue;
// Verify variable is target type
if (non_ssa_vars_.find(varId) != non_ssa_vars_.end()) continue;
if (ptrInst->opcode() != SpvOpVariable) {
non_ssa_vars_.insert(varId);
ssa_var2store_.erase(varId);
continue;
}
// Verify target type and function storage class
if (!IsTargetVar(varId)) {
non_ssa_vars_.insert(varId);
continue;
}
if (!HasOnlySupportedRefs(varId)) {
non_ssa_vars_.insert(varId);
continue;
}
// Ignore variables with multiple stores
if (ssa_var2store_.find(varId) != ssa_var2store_.end()) {
non_ssa_vars_.insert(varId);
ssa_var2store_.erase(varId);
continue;
}
// Remember pointer to variable's store and it's
// ordinal position in block
ssa_var2store_[varId] = &*ii;
store2idx_[&*ii] = instIdx;
store2blk_[&*ii] = &*bi;
}
}
}
LocalSingleStoreElimPass::GetBlocksFunction
LocalSingleStoreElimPass::AugmentedCFGSuccessorsFunction() const {
return [this](const ir::BasicBlock* block) {
auto asmi = augmented_successors_map_.find(block);
if (asmi != augmented_successors_map_.end()) return &(*asmi).second;
auto smi = successors_map_.find(block);
return &(*smi).second;
};
}
LocalSingleStoreElimPass::GetBlocksFunction
LocalSingleStoreElimPass::AugmentedCFGPredecessorsFunction() const {
return [this](const ir::BasicBlock* block) {
auto apmi = augmented_predecessors_map_.find(block);
if (apmi != augmented_predecessors_map_.end()) return &(*apmi).second;
auto pmi = predecessors_map_.find(block);
return &(*pmi).second;
};
}
void LocalSingleStoreElimPass::CalculateImmediateDominators(
ir::Function* func) {
// Compute CFG
vector<ir::BasicBlock*> ordered_blocks;
predecessors_map_.clear();
successors_map_.clear();
for (auto& blk : *func) {
ordered_blocks.push_back(&blk);
const auto& const_blk = blk;
const_blk.ForEachSuccessorLabel([&blk, this](const uint32_t sbid) {
successors_map_[&blk].push_back(label2block_[sbid]);
predecessors_map_[label2block_[sbid]].push_back(&blk);
});
}
// Compute Augmented CFG
augmented_successors_map_.clear();
augmented_predecessors_map_.clear();
successors_map_[cfg()->pseudo_exit_block()] = {};
predecessors_map_[cfg()->pseudo_entry_block()] = {};
auto succ_func = [this](const ir::BasicBlock* b) {
return &successors_map_[b];
};
auto pred_func = [this](const ir::BasicBlock* b) {
return &predecessors_map_[b];
};
CFA<ir::BasicBlock>::ComputeAugmentedCFG(
ordered_blocks, cfg()->pseudo_entry_block(), cfg()->pseudo_exit_block(),
&augmented_successors_map_, &augmented_predecessors_map_, succ_func,
pred_func);
// Compute Dominators
vector<const ir::BasicBlock*> postorder;
auto ignore_block = [](cbb_ptr) {};
auto ignore_edge = [](cbb_ptr, cbb_ptr) {};
spvtools::CFA<ir::BasicBlock>::DepthFirstTraversal(
ordered_blocks[0], AugmentedCFGSuccessorsFunction(), ignore_block,
[&](cbb_ptr b) { postorder.push_back(b); }, ignore_edge);
auto edges = spvtools::CFA<ir::BasicBlock>::CalculateDominators(
postorder, AugmentedCFGPredecessorsFunction());
idom_.clear();
for (auto edge : edges) idom_[edge.first] = edge.second;
}
bool LocalSingleStoreElimPass::Dominates(ir::BasicBlock* blk0, uint32_t idx0,
ir::BasicBlock* blk1, uint32_t idx1) {
if (blk0 == blk1) return idx0 <= idx1;
ir::BasicBlock* b = blk1;
while (idom_[b] != b) {
b = idom_[b];
if (b == blk0) return true;
}
return false;
}
bool LocalSingleStoreElimPass::SingleStoreProcess(ir::Function* func) {
CalculateImmediateDominators(func);
bool modified = false;
for (auto bi = func->begin(); bi != func->end(); ++bi) {
uint32_t instIdx = 0;
for (auto ii = bi->begin(); ii != bi->end(); ++ii, ++instIdx) {
if (ii->opcode() != SpvOpLoad) continue;
uint32_t varId;
ir::Instruction* ptrInst = GetPtr(&*ii, &varId);
// Skip access chain loads
if (ptrInst->opcode() != SpvOpVariable) continue;
const auto vsi = ssa_var2store_.find(varId);
if (vsi == ssa_var2store_.end()) continue;
if (non_ssa_vars_.find(varId) != non_ssa_vars_.end()) continue;
// store must dominate load
if (!Dominates(store2blk_[vsi->second], store2idx_[vsi->second], &*bi,
instIdx))
continue;
// Determine replacement id depending on OpStore or OpVariable
uint32_t replId;
if (vsi->second->opcode() == SpvOpStore)
replId = vsi->second->GetSingleWordInOperand(kStoreValIdInIdx);
else
replId = vsi->second->GetSingleWordInOperand(kVariableInitIdInIdx);
// Replace all instances of the load's id with the SSA value's id
// and add load to removal list
context()->KillNamesAndDecorates(&*ii);
context()->ReplaceAllUsesWith(ii->result_id(), replId);
modified = true;
}
}
return modified;
}
bool LocalSingleStoreElimPass::LocalSingleStoreElim(ir::Function* func) {
bool modified = false;
SingleStoreAnalyze(func);
if (ssa_var2store_.empty()) return false;
modified |= SingleStoreProcess(func);
return modified;
}
void LocalSingleStoreElimPass::Initialize(ir::IRContext* irContext) {
InitializeProcessing(irContext);
// Initialize function and block maps
label2block_.clear();
for (auto& fn : *get_module()) {
for (auto& blk : fn) {
uint32_t bid = blk.id();
label2block_[bid] = &blk;
}
}
// Initialize Target Type Caches
seen_target_vars_.clear();
seen_non_target_vars_.clear();
// Initialize Supported Ref Pointer Cache
supported_ref_ptrs_.clear();
// Initialize extension whitelist
InitExtensions();
};
bool LocalSingleStoreElimPass::AllExtensionsSupported() const {
// If any extension not in whitelist, return false
for (auto& ei : get_module()->extensions()) {
const char* extName =
reinterpret_cast<const char*>(&ei.GetInOperand(0).words[0]);
if (extensions_whitelist_.find(extName) == extensions_whitelist_.end())
return false;
}
return true;
}
Pass::Status LocalSingleStoreElimPass::ProcessImpl() {
// Assumes relaxed logical addressing only (see instruction.h)
if (context()->get_feature_mgr()->HasCapability(SpvCapabilityAddresses))
return Status::SuccessWithoutChange;
// Do not process if module contains OpGroupDecorate. Additional
// support required in KillNamesAndDecorates().
// TODO(greg-lunarg): Add support for OpGroupDecorate
for (auto& ai : get_module()->annotations())
if (ai.opcode() == SpvOpGroupDecorate) return Status::SuccessWithoutChange;
// Do not process if any disallowed extensions are enabled
if (!AllExtensionsSupported()) return Status::SuccessWithoutChange;
// Process all entry point functions
ProcessFunction pfn = [this](ir::Function* fp) {
return LocalSingleStoreElim(fp);
};
bool modified = ProcessEntryPointCallTree(pfn, get_module());
return modified ? Status::SuccessWithChange : Status::SuccessWithoutChange;
}
LocalSingleStoreElimPass::LocalSingleStoreElimPass() {}
Pass::Status LocalSingleStoreElimPass::Process(ir::IRContext* irContext) {
Initialize(irContext);
return ProcessImpl();
}
void LocalSingleStoreElimPass::InitExtensions() {
extensions_whitelist_.clear();
extensions_whitelist_.insert({
"SPV_AMD_shader_explicit_vertex_parameter",
"SPV_AMD_shader_trinary_minmax",
"SPV_AMD_gcn_shader",
"SPV_KHR_shader_ballot",
"SPV_AMD_shader_ballot",
"SPV_AMD_gpu_shader_half_float",
"SPV_KHR_shader_draw_parameters",
"SPV_KHR_subgroup_vote",
"SPV_KHR_16bit_storage",
"SPV_KHR_device_group",
"SPV_KHR_multiview",
"SPV_NVX_multiview_per_view_attributes",
"SPV_NV_viewport_array2",
"SPV_NV_stereo_view_rendering",
"SPV_NV_sample_mask_override_coverage",
"SPV_NV_geometry_shader_passthrough",
"SPV_AMD_texture_gather_bias_lod",
"SPV_KHR_storage_buffer_storage_class",
// SPV_KHR_variable_pointers
// Currently do not support extended pointer expressions
"SPV_AMD_gpu_shader_int16",
"SPV_KHR_post_depth_coverage",
"SPV_KHR_shader_atomic_counter_ops",
});
}
} // namespace opt
} // namespace spvtools