spirv-opt: add pass to Spread Volatile semantics (#4667)
Add a pass to spread Volatile semantics to variables with SMIDNV,
WarpIDNV, SubgroupSize, SubgroupLocalInvocationId, SubgroupEqMask,
SubgroupGeMask, SubgroupGtMask, SubgroupLeMask, or SubgroupLtMask BuiltIn
decorations or OpLoad for them when the shader model is the ray
generation, closest hit, miss, intersection, or callable shaders. This
pass can be used for VUID-StandaloneSpirv-VulkanMemoryModel-04678 and
VUID-StandaloneSpirv-VulkanMemoryModel-04679 (See "Standalone SPIR-V
Validation" section of Vulkan spec "Appendix A: Vulkan Environment for
SPIR-V").
Handle variables used by multiple entry points:
1. Update error check to make it working regardless of the order of
entry points.
2. For a variable, if it is used by two entry points E1 and E2 and
it needs the Volatile semantics for E1 while it does not for E2
- If VulkanMemoryModel capability is enabled, which means we have to
set memory operation of load instructions for the variable, we
update load instructions in E1, but do not update the ones in E2.
- If VulkanMemoryModel capability is disabled, which means we have
to add Volatile decoration for the variable, we report an error
because E1 needs to add Volatile decoration for the variable while
E2 does not.
For the simplicity of the implementation, we assume that all functions
other than entry point functions are inlined.
2022-01-25 18:14:36 +00:00
|
|
|
// Copyright (c) 2022 Google LLC
|
|
|
|
//
|
|
|
|
// 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/spread_volatile_semantics.h"
|
|
|
|
|
|
|
|
#include "source/opt/decoration_manager.h"
|
|
|
|
#include "source/opt/ir_builder.h"
|
|
|
|
#include "source/spirv_constant.h"
|
|
|
|
|
|
|
|
namespace spvtools {
|
|
|
|
namespace opt {
|
|
|
|
namespace {
|
|
|
|
|
|
|
|
const uint32_t kOpDecorateInOperandBuiltinDecoration = 2u;
|
|
|
|
const uint32_t kOpLoadInOperandMemoryOperands = 1u;
|
|
|
|
const uint32_t kOpEntryPointInOperandEntryPoint = 1u;
|
|
|
|
const uint32_t kOpEntryPointInOperandInterface = 3u;
|
|
|
|
|
|
|
|
bool HasBuiltinDecoration(analysis::DecorationManager* decoration_manager,
|
|
|
|
uint32_t var_id, uint32_t built_in) {
|
|
|
|
return decoration_manager->FindDecoration(
|
|
|
|
var_id, SpvDecorationBuiltIn, [built_in](const Instruction& inst) {
|
|
|
|
return built_in == inst.GetSingleWordInOperand(
|
|
|
|
kOpDecorateInOperandBuiltinDecoration);
|
|
|
|
});
|
|
|
|
}
|
|
|
|
|
|
|
|
bool IsBuiltInForRayTracingVolatileSemantics(uint32_t built_in) {
|
|
|
|
switch (built_in) {
|
|
|
|
case SpvBuiltInSMIDNV:
|
|
|
|
case SpvBuiltInWarpIDNV:
|
|
|
|
case SpvBuiltInSubgroupSize:
|
|
|
|
case SpvBuiltInSubgroupLocalInvocationId:
|
|
|
|
case SpvBuiltInSubgroupEqMask:
|
|
|
|
case SpvBuiltInSubgroupGeMask:
|
|
|
|
case SpvBuiltInSubgroupGtMask:
|
|
|
|
case SpvBuiltInSubgroupLeMask:
|
|
|
|
case SpvBuiltInSubgroupLtMask:
|
|
|
|
return true;
|
|
|
|
default:
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
bool HasBuiltinForRayTracingVolatileSemantics(
|
|
|
|
analysis::DecorationManager* decoration_manager, uint32_t var_id) {
|
|
|
|
return decoration_manager->FindDecoration(
|
|
|
|
var_id, SpvDecorationBuiltIn, [](const Instruction& inst) {
|
|
|
|
uint32_t built_in =
|
|
|
|
inst.GetSingleWordInOperand(kOpDecorateInOperandBuiltinDecoration);
|
|
|
|
return IsBuiltInForRayTracingVolatileSemantics(built_in);
|
|
|
|
});
|
|
|
|
}
|
|
|
|
|
|
|
|
bool HasVolatileDecoration(analysis::DecorationManager* decoration_manager,
|
|
|
|
uint32_t var_id) {
|
|
|
|
return decoration_manager->HasDecoration(var_id, SpvDecorationVolatile);
|
|
|
|
}
|
|
|
|
|
|
|
|
} // namespace
|
|
|
|
|
|
|
|
Pass::Status SpreadVolatileSemantics::Process() {
|
2022-03-25 17:54:46 +00:00
|
|
|
if (HasNoExecutionModel()) {
|
|
|
|
return Status::SuccessWithoutChange;
|
|
|
|
}
|
spirv-opt: add pass to Spread Volatile semantics (#4667)
Add a pass to spread Volatile semantics to variables with SMIDNV,
WarpIDNV, SubgroupSize, SubgroupLocalInvocationId, SubgroupEqMask,
SubgroupGeMask, SubgroupGtMask, SubgroupLeMask, or SubgroupLtMask BuiltIn
decorations or OpLoad for them when the shader model is the ray
generation, closest hit, miss, intersection, or callable shaders. This
pass can be used for VUID-StandaloneSpirv-VulkanMemoryModel-04678 and
VUID-StandaloneSpirv-VulkanMemoryModel-04679 (See "Standalone SPIR-V
Validation" section of Vulkan spec "Appendix A: Vulkan Environment for
SPIR-V").
Handle variables used by multiple entry points:
1. Update error check to make it working regardless of the order of
entry points.
2. For a variable, if it is used by two entry points E1 and E2 and
it needs the Volatile semantics for E1 while it does not for E2
- If VulkanMemoryModel capability is enabled, which means we have to
set memory operation of load instructions for the variable, we
update load instructions in E1, but do not update the ones in E2.
- If VulkanMemoryModel capability is disabled, which means we have
to add Volatile decoration for the variable, we report an error
because E1 needs to add Volatile decoration for the variable while
E2 does not.
For the simplicity of the implementation, we assume that all functions
other than entry point functions are inlined.
2022-01-25 18:14:36 +00:00
|
|
|
const bool is_vk_memory_model_enabled =
|
|
|
|
context()->get_feature_mgr()->HasCapability(
|
|
|
|
SpvCapabilityVulkanMemoryModel);
|
|
|
|
CollectTargetsForVolatileSemantics(is_vk_memory_model_enabled);
|
|
|
|
|
|
|
|
// If VulkanMemoryModel capability is not enabled, we have to set Volatile
|
|
|
|
// decoration for interface variables instead of setting Volatile for load
|
|
|
|
// instructions. If an interface (or pointers to it) is used by two load
|
|
|
|
// instructions in two entry points and one must be volatile while another
|
|
|
|
// is not, we have to report an error for the conflict.
|
|
|
|
if (!is_vk_memory_model_enabled &&
|
|
|
|
HasInterfaceInConflictOfVolatileSemantics()) {
|
|
|
|
return Status::Failure;
|
|
|
|
}
|
|
|
|
|
|
|
|
return SpreadVolatileSemanticsToVariables(is_vk_memory_model_enabled);
|
|
|
|
}
|
|
|
|
|
|
|
|
Pass::Status SpreadVolatileSemantics::SpreadVolatileSemanticsToVariables(
|
|
|
|
const bool is_vk_memory_model_enabled) {
|
|
|
|
Status status = Status::SuccessWithoutChange;
|
|
|
|
for (Instruction& var : context()->types_values()) {
|
|
|
|
auto entry_function_ids =
|
|
|
|
EntryFunctionsToSpreadVolatileSemanticsForVar(var.result_id());
|
|
|
|
if (entry_function_ids.empty()) {
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (is_vk_memory_model_enabled) {
|
|
|
|
SetVolatileForLoadsInEntries(&var, entry_function_ids);
|
|
|
|
} else {
|
|
|
|
DecorateVarWithVolatile(&var);
|
|
|
|
}
|
|
|
|
status = Status::SuccessWithChange;
|
|
|
|
}
|
|
|
|
return status;
|
|
|
|
}
|
|
|
|
|
|
|
|
bool SpreadVolatileSemantics::IsTargetUsedByNonVolatileLoadInEntryPoint(
|
|
|
|
uint32_t var_id, Instruction* entry_point) {
|
|
|
|
uint32_t entry_function_id =
|
|
|
|
entry_point->GetSingleWordInOperand(kOpEntryPointInOperandEntryPoint);
|
2022-05-04 14:52:58 +00:00
|
|
|
std::unordered_set<uint32_t> funcs;
|
|
|
|
context()->CollectCallTreeFromRoots(entry_function_id, &funcs);
|
spirv-opt: add pass to Spread Volatile semantics (#4667)
Add a pass to spread Volatile semantics to variables with SMIDNV,
WarpIDNV, SubgroupSize, SubgroupLocalInvocationId, SubgroupEqMask,
SubgroupGeMask, SubgroupGtMask, SubgroupLeMask, or SubgroupLtMask BuiltIn
decorations or OpLoad for them when the shader model is the ray
generation, closest hit, miss, intersection, or callable shaders. This
pass can be used for VUID-StandaloneSpirv-VulkanMemoryModel-04678 and
VUID-StandaloneSpirv-VulkanMemoryModel-04679 (See "Standalone SPIR-V
Validation" section of Vulkan spec "Appendix A: Vulkan Environment for
SPIR-V").
Handle variables used by multiple entry points:
1. Update error check to make it working regardless of the order of
entry points.
2. For a variable, if it is used by two entry points E1 and E2 and
it needs the Volatile semantics for E1 while it does not for E2
- If VulkanMemoryModel capability is enabled, which means we have to
set memory operation of load instructions for the variable, we
update load instructions in E1, but do not update the ones in E2.
- If VulkanMemoryModel capability is disabled, which means we have
to add Volatile decoration for the variable, we report an error
because E1 needs to add Volatile decoration for the variable while
E2 does not.
For the simplicity of the implementation, we assume that all functions
other than entry point functions are inlined.
2022-01-25 18:14:36 +00:00
|
|
|
return !VisitLoadsOfPointersToVariableInEntries(
|
|
|
|
var_id,
|
|
|
|
[](Instruction* load) {
|
|
|
|
// If it has a load without volatile memory operand, finish traversal
|
|
|
|
// and return false.
|
|
|
|
if (load->NumInOperands() <= kOpLoadInOperandMemoryOperands) {
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
uint32_t memory_operands =
|
|
|
|
load->GetSingleWordInOperand(kOpLoadInOperandMemoryOperands);
|
|
|
|
return (memory_operands & SpvMemoryAccessVolatileMask) != 0;
|
|
|
|
},
|
2022-05-04 14:52:58 +00:00
|
|
|
funcs);
|
spirv-opt: add pass to Spread Volatile semantics (#4667)
Add a pass to spread Volatile semantics to variables with SMIDNV,
WarpIDNV, SubgroupSize, SubgroupLocalInvocationId, SubgroupEqMask,
SubgroupGeMask, SubgroupGtMask, SubgroupLeMask, or SubgroupLtMask BuiltIn
decorations or OpLoad for them when the shader model is the ray
generation, closest hit, miss, intersection, or callable shaders. This
pass can be used for VUID-StandaloneSpirv-VulkanMemoryModel-04678 and
VUID-StandaloneSpirv-VulkanMemoryModel-04679 (See "Standalone SPIR-V
Validation" section of Vulkan spec "Appendix A: Vulkan Environment for
SPIR-V").
Handle variables used by multiple entry points:
1. Update error check to make it working regardless of the order of
entry points.
2. For a variable, if it is used by two entry points E1 and E2 and
it needs the Volatile semantics for E1 while it does not for E2
- If VulkanMemoryModel capability is enabled, which means we have to
set memory operation of load instructions for the variable, we
update load instructions in E1, but do not update the ones in E2.
- If VulkanMemoryModel capability is disabled, which means we have
to add Volatile decoration for the variable, we report an error
because E1 needs to add Volatile decoration for the variable while
E2 does not.
For the simplicity of the implementation, we assume that all functions
other than entry point functions are inlined.
2022-01-25 18:14:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
bool SpreadVolatileSemantics::HasInterfaceInConflictOfVolatileSemantics() {
|
|
|
|
for (Instruction& entry_point : get_module()->entry_points()) {
|
|
|
|
SpvExecutionModel execution_model =
|
|
|
|
static_cast<SpvExecutionModel>(entry_point.GetSingleWordInOperand(0));
|
|
|
|
for (uint32_t operand_index = kOpEntryPointInOperandInterface;
|
|
|
|
operand_index < entry_point.NumInOperands(); ++operand_index) {
|
|
|
|
uint32_t var_id = entry_point.GetSingleWordInOperand(operand_index);
|
|
|
|
if (!EntryFunctionsToSpreadVolatileSemanticsForVar(var_id).empty() &&
|
|
|
|
!IsTargetForVolatileSemantics(var_id, execution_model) &&
|
|
|
|
IsTargetUsedByNonVolatileLoadInEntryPoint(var_id, &entry_point)) {
|
|
|
|
Instruction* inst = context()->get_def_use_mgr()->GetDef(var_id);
|
|
|
|
context()->EmitErrorMessage(
|
|
|
|
"Variable is a target for Volatile semantics for an entry point, "
|
|
|
|
"but it is not for another entry point",
|
|
|
|
inst);
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
void SpreadVolatileSemantics::MarkVolatileSemanticsForVariable(
|
|
|
|
uint32_t var_id, Instruction* entry_point) {
|
|
|
|
uint32_t entry_function_id =
|
|
|
|
entry_point->GetSingleWordInOperand(kOpEntryPointInOperandEntryPoint);
|
|
|
|
auto itr = var_ids_to_entry_fn_for_volatile_semantics_.find(var_id);
|
|
|
|
if (itr == var_ids_to_entry_fn_for_volatile_semantics_.end()) {
|
|
|
|
var_ids_to_entry_fn_for_volatile_semantics_[var_id] = {entry_function_id};
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
itr->second.insert(entry_function_id);
|
|
|
|
}
|
|
|
|
|
|
|
|
void SpreadVolatileSemantics::CollectTargetsForVolatileSemantics(
|
|
|
|
const bool is_vk_memory_model_enabled) {
|
|
|
|
for (Instruction& entry_point : get_module()->entry_points()) {
|
|
|
|
SpvExecutionModel execution_model =
|
|
|
|
static_cast<SpvExecutionModel>(entry_point.GetSingleWordInOperand(0));
|
|
|
|
for (uint32_t operand_index = kOpEntryPointInOperandInterface;
|
|
|
|
operand_index < entry_point.NumInOperands(); ++operand_index) {
|
|
|
|
uint32_t var_id = entry_point.GetSingleWordInOperand(operand_index);
|
|
|
|
if (!IsTargetForVolatileSemantics(var_id, execution_model)) {
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
if (is_vk_memory_model_enabled ||
|
|
|
|
IsTargetUsedByNonVolatileLoadInEntryPoint(var_id, &entry_point)) {
|
|
|
|
MarkVolatileSemanticsForVariable(var_id, &entry_point);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void SpreadVolatileSemantics::DecorateVarWithVolatile(Instruction* var) {
|
|
|
|
analysis::DecorationManager* decoration_manager =
|
|
|
|
context()->get_decoration_mgr();
|
|
|
|
uint32_t var_id = var->result_id();
|
|
|
|
if (HasVolatileDecoration(decoration_manager, var_id)) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
get_decoration_mgr()->AddDecoration(
|
|
|
|
SpvOpDecorate, {{spv_operand_type_t::SPV_OPERAND_TYPE_ID, {var_id}},
|
|
|
|
{spv_operand_type_t::SPV_OPERAND_TYPE_LITERAL_INTEGER,
|
|
|
|
{SpvDecorationVolatile}}});
|
|
|
|
}
|
|
|
|
|
|
|
|
bool SpreadVolatileSemantics::VisitLoadsOfPointersToVariableInEntries(
|
|
|
|
uint32_t var_id, const std::function<bool(Instruction*)>& handle_load,
|
2022-05-04 14:52:58 +00:00
|
|
|
const std::unordered_set<uint32_t>& function_ids) {
|
spirv-opt: add pass to Spread Volatile semantics (#4667)
Add a pass to spread Volatile semantics to variables with SMIDNV,
WarpIDNV, SubgroupSize, SubgroupLocalInvocationId, SubgroupEqMask,
SubgroupGeMask, SubgroupGtMask, SubgroupLeMask, or SubgroupLtMask BuiltIn
decorations or OpLoad for them when the shader model is the ray
generation, closest hit, miss, intersection, or callable shaders. This
pass can be used for VUID-StandaloneSpirv-VulkanMemoryModel-04678 and
VUID-StandaloneSpirv-VulkanMemoryModel-04679 (See "Standalone SPIR-V
Validation" section of Vulkan spec "Appendix A: Vulkan Environment for
SPIR-V").
Handle variables used by multiple entry points:
1. Update error check to make it working regardless of the order of
entry points.
2. For a variable, if it is used by two entry points E1 and E2 and
it needs the Volatile semantics for E1 while it does not for E2
- If VulkanMemoryModel capability is enabled, which means we have to
set memory operation of load instructions for the variable, we
update load instructions in E1, but do not update the ones in E2.
- If VulkanMemoryModel capability is disabled, which means we have
to add Volatile decoration for the variable, we report an error
because E1 needs to add Volatile decoration for the variable while
E2 does not.
For the simplicity of the implementation, we assume that all functions
other than entry point functions are inlined.
2022-01-25 18:14:36 +00:00
|
|
|
std::vector<uint32_t> worklist({var_id});
|
|
|
|
auto* def_use_mgr = context()->get_def_use_mgr();
|
|
|
|
while (!worklist.empty()) {
|
|
|
|
uint32_t ptr_id = worklist.back();
|
|
|
|
worklist.pop_back();
|
|
|
|
bool finish_traversal = !def_use_mgr->WhileEachUser(
|
|
|
|
ptr_id, [this, &worklist, &ptr_id, handle_load,
|
2022-05-04 14:52:58 +00:00
|
|
|
&function_ids](Instruction* user) {
|
spirv-opt: add pass to Spread Volatile semantics (#4667)
Add a pass to spread Volatile semantics to variables with SMIDNV,
WarpIDNV, SubgroupSize, SubgroupLocalInvocationId, SubgroupEqMask,
SubgroupGeMask, SubgroupGtMask, SubgroupLeMask, or SubgroupLtMask BuiltIn
decorations or OpLoad for them when the shader model is the ray
generation, closest hit, miss, intersection, or callable shaders. This
pass can be used for VUID-StandaloneSpirv-VulkanMemoryModel-04678 and
VUID-StandaloneSpirv-VulkanMemoryModel-04679 (See "Standalone SPIR-V
Validation" section of Vulkan spec "Appendix A: Vulkan Environment for
SPIR-V").
Handle variables used by multiple entry points:
1. Update error check to make it working regardless of the order of
entry points.
2. For a variable, if it is used by two entry points E1 and E2 and
it needs the Volatile semantics for E1 while it does not for E2
- If VulkanMemoryModel capability is enabled, which means we have to
set memory operation of load instructions for the variable, we
update load instructions in E1, but do not update the ones in E2.
- If VulkanMemoryModel capability is disabled, which means we have
to add Volatile decoration for the variable, we report an error
because E1 needs to add Volatile decoration for the variable while
E2 does not.
For the simplicity of the implementation, we assume that all functions
other than entry point functions are inlined.
2022-01-25 18:14:36 +00:00
|
|
|
BasicBlock* block = context()->get_instr_block(user);
|
|
|
|
if (block == nullptr ||
|
2022-05-04 14:52:58 +00:00
|
|
|
function_ids.find(block->GetParent()->result_id()) ==
|
|
|
|
function_ids.end()) {
|
spirv-opt: add pass to Spread Volatile semantics (#4667)
Add a pass to spread Volatile semantics to variables with SMIDNV,
WarpIDNV, SubgroupSize, SubgroupLocalInvocationId, SubgroupEqMask,
SubgroupGeMask, SubgroupGtMask, SubgroupLeMask, or SubgroupLtMask BuiltIn
decorations or OpLoad for them when the shader model is the ray
generation, closest hit, miss, intersection, or callable shaders. This
pass can be used for VUID-StandaloneSpirv-VulkanMemoryModel-04678 and
VUID-StandaloneSpirv-VulkanMemoryModel-04679 (See "Standalone SPIR-V
Validation" section of Vulkan spec "Appendix A: Vulkan Environment for
SPIR-V").
Handle variables used by multiple entry points:
1. Update error check to make it working regardless of the order of
entry points.
2. For a variable, if it is used by two entry points E1 and E2 and
it needs the Volatile semantics for E1 while it does not for E2
- If VulkanMemoryModel capability is enabled, which means we have to
set memory operation of load instructions for the variable, we
update load instructions in E1, but do not update the ones in E2.
- If VulkanMemoryModel capability is disabled, which means we have
to add Volatile decoration for the variable, we report an error
because E1 needs to add Volatile decoration for the variable while
E2 does not.
For the simplicity of the implementation, we assume that all functions
other than entry point functions are inlined.
2022-01-25 18:14:36 +00:00
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (user->opcode() == SpvOpAccessChain ||
|
|
|
|
user->opcode() == SpvOpInBoundsAccessChain ||
|
|
|
|
user->opcode() == SpvOpPtrAccessChain ||
|
|
|
|
user->opcode() == SpvOpInBoundsPtrAccessChain ||
|
|
|
|
user->opcode() == SpvOpCopyObject) {
|
|
|
|
if (ptr_id == user->GetSingleWordInOperand(0))
|
|
|
|
worklist.push_back(user->result_id());
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (user->opcode() != SpvOpLoad) {
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
return handle_load(user);
|
|
|
|
});
|
|
|
|
if (finish_traversal) return false;
|
|
|
|
}
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
void SpreadVolatileSemantics::SetVolatileForLoadsInEntries(
|
|
|
|
Instruction* var, const std::unordered_set<uint32_t>& entry_function_ids) {
|
|
|
|
// Set Volatile memory operand for all load instructions if they do not have
|
|
|
|
// it.
|
2022-05-04 14:52:58 +00:00
|
|
|
for (auto entry_id : entry_function_ids) {
|
|
|
|
std::unordered_set<uint32_t> funcs;
|
|
|
|
context()->CollectCallTreeFromRoots(entry_id, &funcs);
|
|
|
|
VisitLoadsOfPointersToVariableInEntries(
|
|
|
|
var->result_id(),
|
|
|
|
[](Instruction* load) {
|
|
|
|
if (load->NumInOperands() <= kOpLoadInOperandMemoryOperands) {
|
|
|
|
load->AddOperand({SPV_OPERAND_TYPE_MEMORY_ACCESS,
|
|
|
|
{SpvMemoryAccessVolatileMask}});
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
uint32_t memory_operands =
|
|
|
|
load->GetSingleWordInOperand(kOpLoadInOperandMemoryOperands);
|
|
|
|
memory_operands |= SpvMemoryAccessVolatileMask;
|
|
|
|
load->SetInOperand(kOpLoadInOperandMemoryOperands, {memory_operands});
|
spirv-opt: add pass to Spread Volatile semantics (#4667)
Add a pass to spread Volatile semantics to variables with SMIDNV,
WarpIDNV, SubgroupSize, SubgroupLocalInvocationId, SubgroupEqMask,
SubgroupGeMask, SubgroupGtMask, SubgroupLeMask, or SubgroupLtMask BuiltIn
decorations or OpLoad for them when the shader model is the ray
generation, closest hit, miss, intersection, or callable shaders. This
pass can be used for VUID-StandaloneSpirv-VulkanMemoryModel-04678 and
VUID-StandaloneSpirv-VulkanMemoryModel-04679 (See "Standalone SPIR-V
Validation" section of Vulkan spec "Appendix A: Vulkan Environment for
SPIR-V").
Handle variables used by multiple entry points:
1. Update error check to make it working regardless of the order of
entry points.
2. For a variable, if it is used by two entry points E1 and E2 and
it needs the Volatile semantics for E1 while it does not for E2
- If VulkanMemoryModel capability is enabled, which means we have to
set memory operation of load instructions for the variable, we
update load instructions in E1, but do not update the ones in E2.
- If VulkanMemoryModel capability is disabled, which means we have
to add Volatile decoration for the variable, we report an error
because E1 needs to add Volatile decoration for the variable while
E2 does not.
For the simplicity of the implementation, we assume that all functions
other than entry point functions are inlined.
2022-01-25 18:14:36 +00:00
|
|
|
return true;
|
2022-05-04 14:52:58 +00:00
|
|
|
},
|
|
|
|
funcs);
|
|
|
|
}
|
spirv-opt: add pass to Spread Volatile semantics (#4667)
Add a pass to spread Volatile semantics to variables with SMIDNV,
WarpIDNV, SubgroupSize, SubgroupLocalInvocationId, SubgroupEqMask,
SubgroupGeMask, SubgroupGtMask, SubgroupLeMask, or SubgroupLtMask BuiltIn
decorations or OpLoad for them when the shader model is the ray
generation, closest hit, miss, intersection, or callable shaders. This
pass can be used for VUID-StandaloneSpirv-VulkanMemoryModel-04678 and
VUID-StandaloneSpirv-VulkanMemoryModel-04679 (See "Standalone SPIR-V
Validation" section of Vulkan spec "Appendix A: Vulkan Environment for
SPIR-V").
Handle variables used by multiple entry points:
1. Update error check to make it working regardless of the order of
entry points.
2. For a variable, if it is used by two entry points E1 and E2 and
it needs the Volatile semantics for E1 while it does not for E2
- If VulkanMemoryModel capability is enabled, which means we have to
set memory operation of load instructions for the variable, we
update load instructions in E1, but do not update the ones in E2.
- If VulkanMemoryModel capability is disabled, which means we have
to add Volatile decoration for the variable, we report an error
because E1 needs to add Volatile decoration for the variable while
E2 does not.
For the simplicity of the implementation, we assume that all functions
other than entry point functions are inlined.
2022-01-25 18:14:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
bool SpreadVolatileSemantics::IsTargetForVolatileSemantics(
|
|
|
|
uint32_t var_id, SpvExecutionModel execution_model) {
|
|
|
|
analysis::DecorationManager* decoration_manager =
|
|
|
|
context()->get_decoration_mgr();
|
|
|
|
if (execution_model == SpvExecutionModelFragment) {
|
|
|
|
return get_module()->version() >= SPV_SPIRV_VERSION_WORD(1, 6) &&
|
|
|
|
HasBuiltinDecoration(decoration_manager, var_id,
|
|
|
|
SpvBuiltInHelperInvocation);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (execution_model == SpvExecutionModelIntersectionKHR ||
|
|
|
|
execution_model == SpvExecutionModelIntersectionNV) {
|
|
|
|
if (HasBuiltinDecoration(decoration_manager, var_id,
|
|
|
|
SpvBuiltInRayTmaxKHR)) {
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
switch (execution_model) {
|
|
|
|
case SpvExecutionModelRayGenerationKHR:
|
|
|
|
case SpvExecutionModelClosestHitKHR:
|
|
|
|
case SpvExecutionModelMissKHR:
|
|
|
|
case SpvExecutionModelCallableKHR:
|
|
|
|
case SpvExecutionModelIntersectionKHR:
|
|
|
|
return HasBuiltinForRayTracingVolatileSemantics(decoration_manager,
|
|
|
|
var_id);
|
|
|
|
default:
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
} // namespace opt
|
|
|
|
} // namespace spvtools
|