blob: 5553a3362f8a6eb3bfd758888819b1267bd2d75a [file] [log] [blame]
// Copyright (c) 2022 The Khronos Group Inc.
// Copyright (c) 2022 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 "source/opt/eliminate_dead_io_components_pass.h"
#include <vector>
#include "source/opt/instruction.h"
#include "source/opt/ir_context.h"
#include "source/util/bit_vector.h"
namespace spvtools {
namespace opt {
namespace {
constexpr uint32_t kAccessChainBaseInIdx = 0;
constexpr uint32_t kAccessChainIndex0InIdx = 1;
constexpr uint32_t kAccessChainIndex1InIdx = 2;
constexpr uint32_t kConstantValueInIdx = 0;
} // namespace
Pass::Status EliminateDeadIOComponentsPass::Process() {
// Only process input and output variables
if (elim_sclass_ != spv::StorageClass::Input &&
elim_sclass_ != spv::StorageClass::Output) {
if (consumer()) {
std::string message =
"EliminateDeadIOComponentsPass only valid for input and output "
"variables.";
consumer()(SPV_MSG_ERROR, 0, {0, 0, 0}, message.c_str());
}
return Status::Failure;
}
// If safe mode, only process Input variables in vertex shader
const auto stage = context()->GetStage();
if (safe_mode_ && !(stage == spv::ExecutionModel::Vertex &&
elim_sclass_ == spv::StorageClass::Input))
return Status::SuccessWithoutChange;
// Current functionality assumes shader capability.
if (!context()->get_feature_mgr()->HasCapability(spv::Capability::Shader))
return Status::SuccessWithoutChange;
// Current functionality assumes vert, frag, tesc, tese or geom shader.
// TODO(issue #4988): Add GLCompute.
if (stage != spv::ExecutionModel::Vertex &&
stage != spv::ExecutionModel::Fragment &&
stage != spv::ExecutionModel::TessellationControl &&
stage != spv::ExecutionModel::TessellationEvaluation &&
stage != spv::ExecutionModel::Geometry)
return Status::SuccessWithoutChange;
analysis::DefUseManager* def_use_mgr = context()->get_def_use_mgr();
analysis::TypeManager* type_mgr = context()->get_type_mgr();
bool modified = false;
std::vector<Instruction*> vars_to_move;
for (auto& var : context()->types_values()) {
if (var.opcode() != spv::Op::OpVariable) {
continue;
}
analysis::Type* var_type = type_mgr->GetType(var.type_id());
analysis::Pointer* ptr_type = var_type->AsPointer();
if (ptr_type == nullptr) {
continue;
}
const auto sclass = ptr_type->storage_class();
if (sclass != elim_sclass_) {
continue;
}
// For tesc, or input variables in tese or geom shaders,
// there is a outer per-vertex-array that must be ignored
// for the purposes of this analysis/optimization. Do the
// analysis on the inner type in these cases.
bool skip_first_index = false;
auto core_type = ptr_type->pointee_type();
if (stage == spv::ExecutionModel::TessellationControl ||
(sclass == spv::StorageClass::Input &&
(stage == spv::ExecutionModel::TessellationEvaluation ||
stage == spv::ExecutionModel::Geometry))) {
auto arr_type = core_type->AsArray();
if (!arr_type) continue;
core_type = arr_type->element_type();
skip_first_index = true;
}
const analysis::Array* arr_type = core_type->AsArray();
if (arr_type != nullptr) {
// Only process array if input of vertex shader, or output of
// fragment shader. Otherwise, if one shader has a runtime index and the
// other does not, interface incompatibility can occur.
if (!((sclass == spv::StorageClass::Input &&
stage == spv::ExecutionModel::Vertex) ||
(sclass == spv::StorageClass::Output &&
stage == spv::ExecutionModel::Fragment)))
continue;
unsigned arr_len_id = arr_type->LengthId();
Instruction* arr_len_inst = def_use_mgr->GetDef(arr_len_id);
if (arr_len_inst->opcode() != spv::Op::OpConstant) {
continue;
}
// SPIR-V requires array size is >= 1, so this works for signed or
// unsigned size.
unsigned original_max =
arr_len_inst->GetSingleWordInOperand(kConstantValueInIdx) - 1;
unsigned max_idx = FindMaxIndex(var, original_max);
if (max_idx != original_max) {
ChangeArrayLength(var, max_idx + 1);
vars_to_move.push_back(&var);
modified = true;
}
continue;
}
const analysis::Struct* struct_type = core_type->AsStruct();
if (struct_type == nullptr) continue;
const auto elt_types = struct_type->element_types();
unsigned original_max = static_cast<unsigned>(elt_types.size()) - 1;
unsigned max_idx = FindMaxIndex(var, original_max, skip_first_index);
if (max_idx != original_max) {
ChangeIOVarStructLength(var, max_idx + 1);
vars_to_move.push_back(&var);
modified = true;
}
}
// Move changed vars after their new type instruction to preserve backward
// referencing.
for (auto var : vars_to_move) {
auto type_id = var->type_id();
auto type_inst = def_use_mgr->GetDef(type_id);
var->RemoveFromList();
var->InsertAfter(type_inst);
}
return modified ? Status::SuccessWithChange : Status::SuccessWithoutChange;
}
unsigned EliminateDeadIOComponentsPass::FindMaxIndex(
const Instruction& var, const unsigned original_max,
const bool skip_first_index) {
unsigned max = 0;
bool seen_non_const_ac = false;
assert(var.opcode() == spv::Op::OpVariable && "must be variable");
context()->get_def_use_mgr()->WhileEachUser(
var.result_id(), [&max, &seen_non_const_ac, var, skip_first_index,
this](Instruction* use) {
auto use_opcode = use->opcode();
if (use_opcode == spv::Op::OpLoad || use_opcode == spv::Op::OpStore ||
use_opcode == spv::Op::OpCopyMemory ||
use_opcode == spv::Op::OpCopyMemorySized ||
use_opcode == spv::Op::OpCopyObject) {
seen_non_const_ac = true;
return false;
}
if (use->opcode() != spv::Op::OpAccessChain &&
use->opcode() != spv::Op::OpInBoundsAccessChain) {
return true;
}
// OpAccessChain with no indices currently not optimized
if (use->NumInOperands() == 1 ||
(skip_first_index && use->NumInOperands() == 2)) {
seen_non_const_ac = true;
return false;
}
const unsigned base_id =
use->GetSingleWordInOperand(kAccessChainBaseInIdx);
USE_ASSERT(base_id == var.result_id() && "unexpected base");
const unsigned in_idx = skip_first_index ? kAccessChainIndex1InIdx
: kAccessChainIndex0InIdx;
const unsigned idx_id = use->GetSingleWordInOperand(in_idx);
Instruction* idx_inst = context()->get_def_use_mgr()->GetDef(idx_id);
if (idx_inst->opcode() != spv::Op::OpConstant) {
seen_non_const_ac = true;
return false;
}
unsigned value = idx_inst->GetSingleWordInOperand(kConstantValueInIdx);
if (value > max) max = value;
return true;
});
return seen_non_const_ac ? original_max : max;
}
void EliminateDeadIOComponentsPass::ChangeArrayLength(Instruction& arr_var,
unsigned length) {
analysis::TypeManager* type_mgr = context()->get_type_mgr();
analysis::ConstantManager* const_mgr = context()->get_constant_mgr();
analysis::DefUseManager* def_use_mgr = context()->get_def_use_mgr();
analysis::Pointer* ptr_type =
type_mgr->GetType(arr_var.type_id())->AsPointer();
const analysis::Array* arr_ty = ptr_type->pointee_type()->AsArray();
assert(arr_ty && "expecting array type");
uint32_t length_id = const_mgr->GetUIntConstId(length);
analysis::Array new_arr_ty(arr_ty->element_type(),
arr_ty->GetConstantLengthInfo(length_id, length));
analysis::Type* reg_new_arr_ty = type_mgr->GetRegisteredType(&new_arr_ty);
analysis::Pointer new_ptr_ty(reg_new_arr_ty, ptr_type->storage_class());
analysis::Type* reg_new_ptr_ty = type_mgr->GetRegisteredType(&new_ptr_ty);
uint32_t new_ptr_ty_id = type_mgr->GetTypeInstruction(reg_new_ptr_ty);
arr_var.SetResultType(new_ptr_ty_id);
def_use_mgr->AnalyzeInstUse(&arr_var);
}
void EliminateDeadIOComponentsPass::ChangeIOVarStructLength(Instruction& io_var,
unsigned length) {
analysis::TypeManager* type_mgr = context()->get_type_mgr();
analysis::Pointer* ptr_type =
type_mgr->GetType(io_var.type_id())->AsPointer();
auto core_type = ptr_type->pointee_type();
// Check for per-vertex-array of struct from tesc, tese and geom and grab
// embedded struct type.
const auto arr_type = core_type->AsArray();
if (arr_type) core_type = arr_type->element_type();
const analysis::Struct* struct_ty = core_type->AsStruct();
assert(struct_ty && "expecting struct type");
const auto orig_elt_types = struct_ty->element_types();
std::vector<const analysis::Type*> new_elt_types;
for (unsigned u = 0; u < length; ++u)
new_elt_types.push_back(orig_elt_types[u]);
analysis::Struct new_struct_ty(new_elt_types);
uint32_t old_struct_ty_id = type_mgr->GetTypeInstruction(struct_ty);
std::vector<Instruction*> decorations =
context()->get_decoration_mgr()->GetDecorationsFor(old_struct_ty_id,
true);
for (auto dec : decorations) {
if (dec->opcode() == spv::Op::OpMemberDecorate) {
uint32_t midx = dec->GetSingleWordInOperand(1);
if (midx >= length) continue;
}
type_mgr->AttachDecoration(*dec, &new_struct_ty);
}
// Clone name instructions for new struct type
analysis::Type* reg_new_str_ty = type_mgr->GetRegisteredType(&new_struct_ty);
uint32_t new_struct_ty_id = type_mgr->GetTypeInstruction(reg_new_str_ty);
context()->CloneNames(old_struct_ty_id, new_struct_ty_id, length);
// Attach new type to var
analysis::Type* reg_new_var_ty = reg_new_str_ty;
if (arr_type) {
analysis::Array new_arr_ty(reg_new_var_ty, arr_type->length_info());
reg_new_var_ty = type_mgr->GetRegisteredType(&new_arr_ty);
}
analysis::Pointer new_ptr_ty(reg_new_var_ty, elim_sclass_);
analysis::Type* reg_new_ptr_ty = type_mgr->GetRegisteredType(&new_ptr_ty);
uint32_t new_ptr_ty_id = type_mgr->GetTypeInstruction(reg_new_ptr_ty);
io_var.SetResultType(new_ptr_ty_id);
analysis::DefUseManager* def_use_mgr = context()->get_def_use_mgr();
def_use_mgr->AnalyzeInstUse(&io_var);
}
} // namespace opt
} // namespace spvtools