llvm-project/clang/lib/CodeGen/HLSLBufferLayoutBuilder.cpp
Matheus Izvekov 91cdd35008
[clang] Improve nested name specifier AST representation (#147835)
This is a major change on how we represent nested name qualifications in
the AST.

* The nested name specifier itself and how it's stored is changed. The
prefixes for types are handled within the type hierarchy, which makes
canonicalization for them super cheap, no memory allocation required.
Also translating a type into nested name specifier form becomes a no-op.
An identifier is stored as a DependentNameType. The nested name
specifier gains a lightweight handle class, to be used instead of
passing around pointers, which is similar to what is implemented for
TemplateName. There is still one free bit available, and this handle can
be used within a PointerUnion and PointerIntPair, which should keep
bit-packing aficionados happy.
* The ElaboratedType node is removed, all type nodes in which it could
previously apply to can now store the elaborated keyword and name
qualifier, tail allocating when present.
* TagTypes can now point to the exact declaration found when producing
these, as opposed to the previous situation of there only existing one
TagType per entity. This increases the amount of type sugar retained,
and can have several applications, for example in tracking module
ownership, and other tools which care about source file origins, such as
IWYU. These TagTypes are lazily allocated, in order to limit the
increase in AST size.

This patch offers a great performance benefit.

It greatly improves compilation time for
[stdexec](https://github.com/NVIDIA/stdexec). For one datapoint, for
`test_on2.cpp` in that project, which is the slowest compiling test,
this patch improves `-c` compilation time by about 7.2%, with the
`-fsyntax-only` improvement being at ~12%.

This has great results on compile-time-tracker as well:

![image](https://github.com/user-attachments/assets/700dce98-2cab-4aa8-97d1-b038c0bee831)

This patch also further enables other optimziations in the future, and
will reduce the performance impact of template specialization resugaring
when that lands.

It has some other miscelaneous drive-by fixes.

About the review: Yes the patch is huge, sorry about that. Part of the
reason is that I started by the nested name specifier part, before the
ElaboratedType part, but that had a huge performance downside, as
ElaboratedType is a big performance hog. I didn't have the steam to go
back and change the patch after the fact.

There is also a lot of internal API changes, and it made sense to remove
ElaboratedType in one go, versus removing it from one type at a time, as
that would present much more churn to the users. Also, the nested name
specifier having a different API avoids missing changes related to how
prefixes work now, which could make existing code compile but not work.

How to review: The important changes are all in
`clang/include/clang/AST` and `clang/lib/AST`, with also important
changes in `clang/lib/Sema/TreeTransform.h`.

The rest and bulk of the changes are mostly consequences of the changes
in API.

PS: TagType::getDecl is renamed to `getOriginalDecl` in this patch, just
for easier to rebasing. I plan to rename it back after this lands.

Fixes #136624
Fixes https://github.com/llvm/llvm-project/issues/43179
Fixes https://github.com/llvm/llvm-project/issues/68670
Fixes https://github.com/llvm/llvm-project/issues/92757
2025-08-09 05:06:53 -03:00

276 lines
11 KiB
C++

//===- HLSLBufferLayoutBuilder.cpp ----------------------------------------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
#include "HLSLBufferLayoutBuilder.h"
#include "CGHLSLRuntime.h"
#include "CodeGenModule.h"
#include "clang/AST/Type.h"
#include <climits>
//===----------------------------------------------------------------------===//
// Implementation of constant buffer layout common between DirectX and
// SPIR/SPIR-V.
//===----------------------------------------------------------------------===//
using namespace clang;
using namespace clang::CodeGen;
using llvm::hlsl::CBufferRowSizeInBytes;
namespace {
// Creates a new array type with the same dimentions but with the new
// element type.
static llvm::Type *
createArrayWithNewElementType(CodeGenModule &CGM,
const ConstantArrayType *ArrayType,
llvm::Type *NewElemType) {
const clang::Type *ArrayElemType = ArrayType->getArrayElementTypeNoTypeQual();
if (ArrayElemType->isConstantArrayType())
NewElemType = createArrayWithNewElementType(
CGM, cast<const ConstantArrayType>(ArrayElemType), NewElemType);
return llvm::ArrayType::get(NewElemType, ArrayType->getSExtSize());
}
// Returns the size of a scalar or vector in bytes
static unsigned getScalarOrVectorSizeInBytes(llvm::Type *Ty) {
assert(Ty->isVectorTy() || Ty->isIntegerTy() || Ty->isFloatingPointTy());
if (Ty->isVectorTy()) {
llvm::FixedVectorType *FVT = cast<llvm::FixedVectorType>(Ty);
return FVT->getNumElements() *
(FVT->getElementType()->getScalarSizeInBits() / 8);
}
return Ty->getScalarSizeInBits() / 8;
}
} // namespace
namespace clang {
namespace CodeGen {
// Creates a layout type for given struct or class with HLSL constant buffer
// layout taking into account PackOffsets, if provided.
// Previously created layout types are cached by CGHLSLRuntime.
//
// The function iterates over all fields of the record type (including base
// classes) and calls layoutField to converts each field to its corresponding
// LLVM type and to calculate its HLSL constant buffer layout. Any embedded
// structs (or arrays of structs) are converted to target layout types as well.
//
// When PackOffsets are specified the elements will be placed based on the
// user-specified offsets. Not all elements must have a packoffset/register(c#)
// annotation though. For those that don't, the PackOffsets array will contain
// -1 value instead. These elements must be placed at the end of the layout
// after all of the elements with specific offset.
llvm::TargetExtType *HLSLBufferLayoutBuilder::createLayoutType(
const RecordType *RT, const llvm::SmallVector<int32_t> *PackOffsets) {
// check if we already have the layout type for this struct
if (llvm::TargetExtType *Ty =
CGM.getHLSLRuntime().getHLSLBufferLayoutType(RT))
return Ty;
SmallVector<unsigned> Layout;
SmallVector<llvm::Type *> LayoutElements;
unsigned Index = 0; // packoffset index
unsigned EndOffset = 0;
SmallVector<std::pair<const FieldDecl *, unsigned>> DelayLayoutFields;
// reserve first spot in the layout vector for buffer size
Layout.push_back(0);
// iterate over all fields of the record, including fields on base classes
llvm::SmallVector<const RecordType *> RecordTypes;
RecordTypes.push_back(RT);
while (RecordTypes.back()->getAsCXXRecordDecl()->getNumBases()) {
CXXRecordDecl *D = RecordTypes.back()->getAsCXXRecordDecl();
assert(D->getNumBases() == 1 &&
"HLSL doesn't support multiple inheritance");
RecordTypes.push_back(D->bases_begin()->getType()->getAs<RecordType>());
}
unsigned FieldOffset;
llvm::Type *FieldType;
while (!RecordTypes.empty()) {
const RecordType *RT = RecordTypes.back();
RecordTypes.pop_back();
for (const auto *FD :
RT->getOriginalDecl()->getDefinitionOrSelf()->fields()) {
assert((!PackOffsets || Index < PackOffsets->size()) &&
"number of elements in layout struct does not match number of "
"packoffset annotations");
// No PackOffset info at all, or have a valid packoffset/register(c#)
// annotations value -> layout the field.
const int PO = PackOffsets ? (*PackOffsets)[Index++] : -1;
if (!PackOffsets || PO != -1) {
if (!layoutField(FD, EndOffset, FieldOffset, FieldType, PO))
return nullptr;
Layout.push_back(FieldOffset);
LayoutElements.push_back(FieldType);
continue;
}
// Have PackOffset info, but there is no packoffset/register(cX)
// annotation on this field. Delay the layout until after all of the
// other elements with packoffsets/register(cX) are processed.
DelayLayoutFields.emplace_back(FD, LayoutElements.size());
// reserve space for this field in the layout vector and elements list
Layout.push_back(UINT_MAX);
LayoutElements.push_back(nullptr);
}
}
// process delayed layouts
for (auto I : DelayLayoutFields) {
const FieldDecl *FD = I.first;
const unsigned IndexInLayoutElements = I.second;
// the first item in layout vector is size, so we need to offset the index
// by 1
const unsigned IndexInLayout = IndexInLayoutElements + 1;
assert(Layout[IndexInLayout] == UINT_MAX &&
LayoutElements[IndexInLayoutElements] == nullptr);
if (!layoutField(FD, EndOffset, FieldOffset, FieldType))
return nullptr;
Layout[IndexInLayout] = FieldOffset;
LayoutElements[IndexInLayoutElements] = FieldType;
}
// set the size of the buffer
Layout[0] = EndOffset;
// create the layout struct type; anonymous struct have empty name but
// non-empty qualified name
const CXXRecordDecl *Decl = RT->getAsCXXRecordDecl();
std::string Name =
Decl->getName().empty() ? "anon" : Decl->getQualifiedNameAsString();
llvm::StructType *StructTy =
llvm::StructType::create(LayoutElements, Name, true);
// create target layout type
llvm::TargetExtType *NewLayoutTy = llvm::TargetExtType::get(
CGM.getLLVMContext(), LayoutTypeName, {StructTy}, Layout);
if (NewLayoutTy)
CGM.getHLSLRuntime().addHLSLBufferLayoutType(RT, NewLayoutTy);
return NewLayoutTy;
}
// The function converts a single field of HLSL Buffer to its corresponding
// LLVM type and calculates it's layout. Any embedded structs (or
// arrays of structs) are converted to target layout types as well.
// The converted type is set to the FieldType parameter, the element
// offset is set to the FieldOffset parameter. The EndOffset (=size of the
// buffer) is also updated accordingly to the offset just after the placed
// element, unless the incoming EndOffset already larger (may happen in case
// of unsorted packoffset annotations).
// Returns true if the conversion was successful.
// The packoffset parameter contains the field's layout offset provided by the
// user or -1 if there was no packoffset (or register(cX)) annotation.
bool HLSLBufferLayoutBuilder::layoutField(const FieldDecl *FD,
unsigned &EndOffset,
unsigned &FieldOffset,
llvm::Type *&FieldType,
int Packoffset) {
// Size of element; for arrays this is a size of a single element in the
// array. Total array size of calculated as (ArrayCount-1) * ArrayStride +
// ElemSize.
unsigned ElemSize = 0;
unsigned ElemOffset = 0;
unsigned ArrayCount = 1;
unsigned ArrayStride = 0;
unsigned NextRowOffset = llvm::alignTo(EndOffset, CBufferRowSizeInBytes);
llvm::Type *ElemLayoutTy = nullptr;
QualType FieldTy = FD->getType();
if (FieldTy->isConstantArrayType()) {
// Unwrap array to find the element type and get combined array size.
QualType Ty = FieldTy;
while (Ty->isConstantArrayType()) {
auto *ArrayTy = CGM.getContext().getAsConstantArrayType(Ty);
ArrayCount *= ArrayTy->getSExtSize();
Ty = ArrayTy->getElementType();
}
// For array of structures, create a new array with a layout type
// instead of the structure type.
if (Ty->isStructureOrClassType()) {
llvm::Type *NewTy =
cast<llvm::TargetExtType>(createLayoutType(Ty->getAs<RecordType>()));
if (!NewTy)
return false;
assert(isa<llvm::TargetExtType>(NewTy) && "expected target type");
ElemSize = cast<llvm::TargetExtType>(NewTy)->getIntParameter(0);
ElemLayoutTy = createArrayWithNewElementType(
CGM, cast<ConstantArrayType>(FieldTy.getTypePtr()), NewTy);
} else {
// Array of vectors or scalars
ElemSize =
getScalarOrVectorSizeInBytes(CGM.getTypes().ConvertTypeForMem(Ty));
ElemLayoutTy = CGM.getTypes().ConvertTypeForMem(FieldTy);
}
ArrayStride = llvm::alignTo(ElemSize, CBufferRowSizeInBytes);
ElemOffset = (Packoffset != -1) ? Packoffset : NextRowOffset;
} else if (FieldTy->isStructureOrClassType()) {
// Create a layout type for the structure
ElemLayoutTy =
createLayoutType(cast<RecordType>(FieldTy->getAs<RecordType>()));
if (!ElemLayoutTy)
return false;
assert(isa<llvm::TargetExtType>(ElemLayoutTy) && "expected target type");
ElemSize = cast<llvm::TargetExtType>(ElemLayoutTy)->getIntParameter(0);
ElemOffset = (Packoffset != -1) ? Packoffset : NextRowOffset;
} else {
// scalar or vector - find element size and alignment
unsigned Align = 0;
ElemLayoutTy = CGM.getTypes().ConvertTypeForMem(FieldTy);
if (ElemLayoutTy->isVectorTy()) {
// align vectors by sub element size
const llvm::FixedVectorType *FVT =
cast<llvm::FixedVectorType>(ElemLayoutTy);
unsigned SubElemSize = FVT->getElementType()->getScalarSizeInBits() / 8;
ElemSize = FVT->getNumElements() * SubElemSize;
Align = SubElemSize;
} else {
assert(ElemLayoutTy->isIntegerTy() || ElemLayoutTy->isFloatingPointTy());
ElemSize = ElemLayoutTy->getScalarSizeInBits() / 8;
Align = ElemSize;
}
// calculate or get element offset for the vector or scalar
if (Packoffset != -1) {
ElemOffset = Packoffset;
} else {
ElemOffset = llvm::alignTo(EndOffset, Align);
// if the element does not fit, move it to the next row
if (ElemOffset + ElemSize > NextRowOffset)
ElemOffset = NextRowOffset;
}
}
// Update end offset of the layout; do not update it if the EndOffset
// is already bigger than the new value (which may happen with unordered
// packoffset annotations)
unsigned NewEndOffset =
ElemOffset + (ArrayCount - 1) * ArrayStride + ElemSize;
EndOffset = std::max<unsigned>(EndOffset, NewEndOffset);
// add the layout element and offset to the lists
FieldOffset = ElemOffset;
FieldType = ElemLayoutTy;
return true;
}
} // namespace CodeGen
} // namespace clang