Revert "[AMDGPU] Handle natively unsupported types in addrspace(7) lowering" #123657
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…wering (…" This reverts commit 3805355.
|
@llvm/pr-subscribers-backend-amdgpu Author: Krzysztof Drewniak (krzysz00) ChangesReverts llvm/llvm-project#110572 Seem to have broken a buildbot, not sure why https://lab.llvm.org/buildbot/#/builders/108/builds/8346 Patch is 343.29 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/123657.diff 6 Files Affected:
diff --git a/llvm/lib/Target/AMDGPU/AMDGPULowerBufferFatPointers.cpp b/llvm/lib/Target/AMDGPU/AMDGPULowerBufferFatPointers.cpp
index 75a0c47f7c2773..657a406e9f7056 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPULowerBufferFatPointers.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPULowerBufferFatPointers.cpp
@@ -66,28 +66,6 @@
// Atomics operations on `ptr addrspace(7)` values are not suppported, as the
// hardware does not include a 160-bit atomic.
//
-// ## Buffer contents type legalization
-//
-// The underlying buffer intrinsics only support types up to 128 bits long,
-// and don't support complex types. If buffer operations were
-// standard pointer operations that could be represented as MIR-level loads,
-// this would be handled by the various legalization schemes in instruction
-// selection. However, because we have to do the conversion from `load` and
-// `store` to intrinsics at LLVM IR level, we must perform that legalization
-// ourselves.
-//
-// This involves a combination of
-// - Converting arrays to vectors where possible
-// - Otherwise, splitting loads and stores of aggregates into loads/stores of
-// each component.
-// - Zero-extending things to fill a whole number of bytes
-// - Casting values of types that don't neatly correspond to supported machine
-// value
-// (for example, an i96 or i256) into ones that would work (
-// like <3 x i32> and <8 x i32>, respectively)
-// - Splitting values that are too long (such as aforementioned <8 x i32>) into
-// multiple operations.
-//
// ## Type remapping
//
// We use a `ValueMapper` to mangle uses of [vectors of] buffer fat pointers
@@ -108,6 +86,7 @@
// This phase also records intrinsics so that they can be remangled or deleted
// later.
//
+//
// ## Splitting pointer structs
//
// The meat of this pass consists of defining semantics for operations that
@@ -239,7 +218,6 @@
#include "llvm/IR/ReplaceConstant.h"
#include "llvm/InitializePasses.h"
#include "llvm/Pass.h"
-#include "llvm/Support/Alignment.h"
#include "llvm/Support/AtomicOrdering.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/ErrorHandling.h"
@@ -573,6 +551,7 @@ bool StoreFatPtrsAsIntsVisitor::visitLoadInst(LoadInst &LI) {
auto *NLI = cast<LoadInst>(LI.clone());
NLI->mutateType(IntTy);
NLI = IRB.Insert(NLI);
+ copyMetadataForLoad(*NLI, LI);
NLI->takeName(&LI);
Value *CastBack = intsToFatPtrs(NLI, IntTy, Ty, NLI->getName());
@@ -597,540 +576,6 @@ bool StoreFatPtrsAsIntsVisitor::visitStoreInst(StoreInst &SI) {
return true;
}
-namespace {
-/// Convert loads/stores of types that the buffer intrinsics can't handle into
-/// one ore more such loads/stores that consist of legal types.
-///
-/// Do this by
-/// 1. Recursing into structs (and arrays that don't share a memory layout with
-/// vectors) since the intrinsics can't handle complex types.
-/// 2. Converting arrays of non-aggregate, byte-sized types into their
-/// corresponding vectors
-/// 3. Bitcasting unsupported types, namely overly-long scalars and byte
-/// vectors, into vectors of supported types.
-/// 4. Splitting up excessively long reads/writes into multiple operations.
-///
-/// Note that this doesn't handle complex data strucures, but, in the future,
-/// the aggregate load splitter from SROA could be refactored to allow for that
-/// case.
-class LegalizeBufferContentTypesVisitor
- : public InstVisitor<LegalizeBufferContentTypesVisitor, bool> {
- friend class InstVisitor<LegalizeBufferContentTypesVisitor, bool>;
-
- IRBuilder<> IRB;
-
- const DataLayout &DL;
-
- /// If T is [N x U], where U is a scalar type, return the vector type
- /// <N x U>, otherwise, return T.
- Type *scalarArrayTypeAsVector(Type *MaybeArrayType);
- Value *arrayToVector(Value *V, Type *TargetType, const Twine &Name);
- Value *vectorToArray(Value *V, Type *OrigType, const Twine &Name);
-
- /// Break up the loads of a struct into the loads of its components
-
- /// Convert a vector or scalar type that can't be operated on by buffer
- /// intrinsics to one that would be legal through bitcasts and/or truncation.
- /// Uses the wider of i32, i16, or i8 where possible.
- Type *legalNonAggregateFor(Type *T);
- Value *makeLegalNonAggregate(Value *V, Type *TargetType, const Twine &Name);
- Value *makeIllegalNonAggregate(Value *V, Type *OrigType, const Twine &Name);
-
- struct VecSlice {
- uint64_t Index = 0;
- uint64_t Length = 0;
- VecSlice() = delete;
- };
- /// Return the [index, length] pairs into which `T` needs to be cut to form
- /// legal buffer load or store operations. Clears `Slices`. Creates an empty
- /// `Slices` for non-vector inputs and creates one slice if no slicing will be
- /// needed.
- void getVecSlices(Type *T, SmallVectorImpl<VecSlice> &Slices);
-
- Value *extractSlice(Value *Vec, VecSlice S, const Twine &Name);
- Value *insertSlice(Value *Whole, Value *Part, VecSlice S, const Twine &Name);
-
- /// In most cases, return `LegalType`. However, when given an input that would
- /// normally be a legal type for the buffer intrinsics to return but that
- /// isn't hooked up through SelectionDAG, return a type of the same width that
- /// can be used with the relevant intrinsics. Specifically, handle the cases:
- /// - <1 x T> => T for all T
- /// - <N x i8> <=> i16, i32, 2xi32, 4xi32 (as needed)
- /// - <N x T> where T is under 32 bits and the total size is 96 bits <=> <3 x
- /// i32>
- Type *intrinsicTypeFor(Type *LegalType);
-
- bool visitLoadImpl(LoadInst &OrigLI, Type *PartType,
- SmallVectorImpl<uint32_t> &AggIdxs, uint64_t AggByteOffset,
- Value *&Result, const Twine &Name);
- /// Return value is (Changed, ModifiedInPlace)
- std::pair<bool, bool> visitStoreImpl(StoreInst &OrigSI, Type *PartType,
- SmallVectorImpl<uint32_t> &AggIdxs,
- uint64_t AggByteOffset,
- const Twine &Name);
-
- bool visitInstruction(Instruction &I) { return false; }
- bool visitLoadInst(LoadInst &LI);
- bool visitStoreInst(StoreInst &SI);
-
-public:
- LegalizeBufferContentTypesVisitor(const DataLayout &DL, LLVMContext &Ctx)
- : IRB(Ctx), DL(DL) {}
- bool processFunction(Function &F);
-};
-} // namespace
-
-Type *LegalizeBufferContentTypesVisitor::scalarArrayTypeAsVector(Type *T) {
- ArrayType *AT = dyn_cast<ArrayType>(T);
- if (!AT)
- return T;
- Type *ET = AT->getElementType();
- if (!ET->isSingleValueType() || isa<VectorType>(ET))
- report_fatal_error("loading non-scalar arrays from buffer fat pointers "
- "should have recursed");
- if (!DL.typeSizeEqualsStoreSize(AT))
- report_fatal_error(
- "loading padded arrays from buffer fat pinters should have recursed");
- return FixedVectorType::get(ET, AT->getNumElements());
-}
-
-Value *LegalizeBufferContentTypesVisitor::arrayToVector(Value *V,
- Type *TargetType,
- const Twine &Name) {
- Value *VectorRes = PoisonValue::get(TargetType);
- auto *VT = cast<FixedVectorType>(TargetType);
- unsigned EC = VT->getNumElements();
- for (auto I : iota_range<unsigned>(0, EC, /*Inclusive=*/false)) {
- Value *Elem = IRB.CreateExtractValue(V, I, Name + ".elem." + Twine(I));
- VectorRes = IRB.CreateInsertElement(VectorRes, Elem, I,
- Name + ".as.vec." + Twine(I));
- }
- return VectorRes;
-}
-
-Value *LegalizeBufferContentTypesVisitor::vectorToArray(Value *V,
- Type *OrigType,
- const Twine &Name) {
- Value *ArrayRes = PoisonValue::get(OrigType);
- ArrayType *AT = cast<ArrayType>(OrigType);
- unsigned EC = AT->getNumElements();
- for (auto I : iota_range<unsigned>(0, EC, /*Inclusive=*/false)) {
- Value *Elem = IRB.CreateExtractElement(V, I, Name + ".elem." + Twine(I));
- ArrayRes = IRB.CreateInsertValue(ArrayRes, Elem, I,
- Name + ".as.array." + Twine(I));
- }
- return ArrayRes;
-}
-
-Type *LegalizeBufferContentTypesVisitor::legalNonAggregateFor(Type *T) {
- TypeSize Size = DL.getTypeStoreSizeInBits(T);
- // Implicitly zero-extend to the next byte if needed
- if (!DL.typeSizeEqualsStoreSize(T))
- T = IRB.getIntNTy(Size.getFixedValue());
- Type *ElemTy = T->getScalarType();
- if (isa<PointerType, ScalableVectorType>(ElemTy)) {
- // Pointers are always big enough, and we'll let scalable vectors through to
- // fail in codegen.
- return T;
- }
- unsigned ElemSize = DL.getTypeSizeInBits(ElemTy).getFixedValue();
- if (isPowerOf2_32(ElemSize) && ElemSize >= 16 && ElemSize <= 128) {
- // [vectors of] anything that's 16/32/64/128 bits can be cast and split into
- // legal buffer operations.
- return T;
- }
- Type *BestVectorElemType = nullptr;
- if (Size.isKnownMultipleOf(32))
- BestVectorElemType = IRB.getInt32Ty();
- else if (Size.isKnownMultipleOf(16))
- BestVectorElemType = IRB.getInt16Ty();
- else
- BestVectorElemType = IRB.getInt8Ty();
- unsigned NumCastElems =
- Size.getFixedValue() / BestVectorElemType->getIntegerBitWidth();
- if (NumCastElems == 1)
- return BestVectorElemType;
- return FixedVectorType::get(BestVectorElemType, NumCastElems);
-}
-
-Value *LegalizeBufferContentTypesVisitor::makeLegalNonAggregate(
- Value *V, Type *TargetType, const Twine &Name) {
- Type *SourceType = V->getType();
- TypeSize SourceSize = DL.getTypeSizeInBits(SourceType);
- TypeSize TargetSize = DL.getTypeSizeInBits(TargetType);
- if (SourceSize != TargetSize) {
- Type *ShortScalarTy = IRB.getIntNTy(SourceSize.getFixedValue());
- Type *ByteScalarTy = IRB.getIntNTy(TargetSize.getFixedValue());
- Value *AsScalar = IRB.CreateBitCast(V, ShortScalarTy, Name + ".as.scalar");
- Value *Zext = IRB.CreateZExt(AsScalar, ByteScalarTy, Name + ".zext");
- V = Zext;
- SourceType = ByteScalarTy;
- }
- return IRB.CreateBitCast(V, TargetType, Name + ".legal");
-}
-
-Value *LegalizeBufferContentTypesVisitor::makeIllegalNonAggregate(
- Value *V, Type *OrigType, const Twine &Name) {
- Type *LegalType = V->getType();
- TypeSize LegalSize = DL.getTypeSizeInBits(LegalType);
- TypeSize OrigSize = DL.getTypeSizeInBits(OrigType);
- if (LegalSize != OrigSize) {
- Type *ShortScalarTy = IRB.getIntNTy(OrigSize.getFixedValue());
- Type *ByteScalarTy = IRB.getIntNTy(LegalSize.getFixedValue());
- Value *AsScalar = IRB.CreateBitCast(V, ByteScalarTy, Name + ".bytes.cast");
- Value *Trunc = IRB.CreateTrunc(AsScalar, ShortScalarTy, Name + ".trunc");
- return IRB.CreateBitCast(Trunc, OrigType, Name + ".orig");
- }
- return IRB.CreateBitCast(V, OrigType, Name + ".real.ty");
-}
-
-Type *LegalizeBufferContentTypesVisitor::intrinsicTypeFor(Type *LegalType) {
- auto *VT = dyn_cast<FixedVectorType>(LegalType);
- if (!VT)
- return LegalType;
- Type *ET = VT->getElementType();
- // Explicitly return the element type of 1-element vectors because the
- // underlying intrinsics don't like <1 x T> even though it's a synonym for T.
- if (VT->getNumElements() == 1)
- return ET;
- if (DL.getTypeSizeInBits(LegalType) == 96 && DL.getTypeSizeInBits(ET) < 32)
- return FixedVectorType::get(IRB.getInt32Ty(), 3);
- if (ET->isIntegerTy(8)) {
- switch (VT->getNumElements()) {
- default:
- return LegalType; // Let it crash later
- case 1:
- return IRB.getInt8Ty();
- case 2:
- return IRB.getInt16Ty();
- case 4:
- return IRB.getInt32Ty();
- case 8:
- return FixedVectorType::get(IRB.getInt32Ty(), 2);
- case 16:
- return FixedVectorType::get(IRB.getInt32Ty(), 4);
- }
- }
- return LegalType;
-}
-
-void LegalizeBufferContentTypesVisitor::getVecSlices(
- Type *T, SmallVectorImpl<VecSlice> &Slices) {
- Slices.clear();
- auto *VT = dyn_cast<FixedVectorType>(T);
- if (!VT)
- return;
-
- uint64_t ElemBitWidth =
- DL.getTypeSizeInBits(VT->getElementType()).getFixedValue();
-
- uint64_t ElemsPer4Words = 128 / ElemBitWidth;
- uint64_t ElemsPer2Words = ElemsPer4Words / 2;
- uint64_t ElemsPerWord = ElemsPer2Words / 2;
- uint64_t ElemsPerShort = ElemsPerWord / 2;
- uint64_t ElemsPerByte = ElemsPerShort / 2;
- // If the elements evenly pack into 32-bit words, we can use 3-word stores,
- // such as for <6 x bfloat> or <3 x i32>, but we can't dot his for, for
- // example, <3 x i64>, since that's not slicing.
- uint64_t ElemsPer3Words = ElemsPerWord * 3;
-
- uint64_t TotalElems = VT->getNumElements();
- uint64_t Index = 0;
- auto TrySlice = [&](unsigned MaybeLen) {
- if (MaybeLen > 0 && Index + MaybeLen <= TotalElems) {
- VecSlice Slice{/*Index=*/Index, /*Length=*/MaybeLen};
- Slices.push_back(Slice);
- Index += MaybeLen;
- return true;
- }
- return false;
- };
- while (Index < TotalElems) {
- TrySlice(ElemsPer4Words) || TrySlice(ElemsPer3Words) ||
- TrySlice(ElemsPer2Words) || TrySlice(ElemsPerWord) ||
- TrySlice(ElemsPerShort) || TrySlice(ElemsPerByte);
- }
-}
-
-Value *LegalizeBufferContentTypesVisitor::extractSlice(Value *Vec, VecSlice S,
- const Twine &Name) {
- auto *VecVT = dyn_cast<FixedVectorType>(Vec->getType());
- if (!VecVT)
- return Vec;
- if (S.Length == VecVT->getNumElements() && S.Index == 0)
- return Vec;
- if (S.Length == 1)
- return IRB.CreateExtractElement(Vec, S.Index,
- Name + ".slice." + Twine(S.Index));
- SmallVector<int> Mask = llvm::to_vector(
- llvm::iota_range<int>(S.Index, S.Index + S.Length, /*Inclusive=*/false));
- return IRB.CreateShuffleVector(Vec, Mask, Name + ".slice." + Twine(S.Index));
-}
-
-Value *LegalizeBufferContentTypesVisitor::insertSlice(Value *Whole, Value *Part,
- VecSlice S,
- const Twine &Name) {
- auto *WholeVT = dyn_cast<FixedVectorType>(Whole->getType());
- if (!WholeVT)
- return Part;
- if (S.Length == WholeVT->getNumElements() && S.Index == 0)
- return Part;
- if (S.Length == 1) {
- return IRB.CreateInsertElement(Whole, Part, S.Index,
- Name + ".slice." + Twine(S.Index));
- }
- int NumElems = cast<FixedVectorType>(Whole->getType())->getNumElements();
-
- // Extend the slice with poisons to make the main shufflevector happy.
- SmallVector<int> ExtPartMask(NumElems, -1);
- for (auto [I, E] : llvm::enumerate(
- MutableArrayRef<int>(ExtPartMask).take_front(S.Length))) {
- E = I;
- }
- Value *ExtPart = IRB.CreateShuffleVector(Part, ExtPartMask,
- Name + ".ext." + Twine(S.Index));
-
- SmallVector<int> Mask =
- llvm::to_vector(llvm::iota_range<int>(0, NumElems, /*Inclusive=*/false));
- for (auto [I, E] :
- llvm::enumerate(MutableArrayRef<int>(Mask).slice(S.Index, S.Length)))
- E = I + NumElems;
- return IRB.CreateShuffleVector(Whole, ExtPart, Mask,
- Name + ".parts." + Twine(S.Index));
-}
-
-bool LegalizeBufferContentTypesVisitor::visitLoadImpl(
- LoadInst &OrigLI, Type *PartType, SmallVectorImpl<uint32_t> &AggIdxs,
- uint64_t AggByteOff, Value *&Result, const Twine &Name) {
- if (auto *ST = dyn_cast<StructType>(PartType)) {
- const StructLayout *Layout = DL.getStructLayout(ST);
- bool Changed = false;
- for (auto [I, ElemTy, Offset] :
- llvm::enumerate(ST->elements(), Layout->getMemberOffsets())) {
- AggIdxs.push_back(I);
- Changed |= visitLoadImpl(OrigLI, ElemTy, AggIdxs,
- AggByteOff + Offset.getFixedValue(), Result,
- Name + "." + Twine(I));
- AggIdxs.pop_back();
- }
- return Changed;
- }
- if (auto *AT = dyn_cast<ArrayType>(PartType)) {
- Type *ElemTy = AT->getElementType();
- if (!ElemTy->isSingleValueType() || !DL.typeSizeEqualsStoreSize(ElemTy) ||
- ElemTy->isVectorTy()) {
- TypeSize ElemStoreSize = DL.getTypeStoreSize(ElemTy);
- bool Changed = false;
- for (auto I : llvm::iota_range<uint32_t>(0, AT->getNumElements(),
- /*Inclusive=*/false)) {
- AggIdxs.push_back(I);
- Changed |= visitLoadImpl(OrigLI, ElemTy, AggIdxs,
- AggByteOff + I * ElemStoreSize.getFixedValue(),
- Result, Name + Twine(I));
- AggIdxs.pop_back();
- }
- return Changed;
- }
- }
-
- // Typical case
-
- Type *ArrayAsVecType = scalarArrayTypeAsVector(PartType);
- Type *LegalType = legalNonAggregateFor(ArrayAsVecType);
-
- SmallVector<VecSlice> Slices;
- getVecSlices(LegalType, Slices);
- bool HasSlices = Slices.size() > 1;
- bool IsAggPart = !AggIdxs.empty();
- Value *LoadsRes;
- if (!HasSlices && !IsAggPart) {
- Type *LoadableType = intrinsicTypeFor(LegalType);
- if (LoadableType == PartType)
- return false;
-
- IRB.SetInsertPoint(&OrigLI);
- auto *NLI = cast<LoadInst>(OrigLI.clone());
- NLI->mutateType(LoadableType);
- NLI = IRB.Insert(NLI);
- NLI->setName(Name + ".loadable");
-
- LoadsRes = IRB.CreateBitCast(NLI, LegalType, Name + ".from.loadable");
- } else {
- IRB.SetInsertPoint(&OrigLI);
- LoadsRes = PoisonValue::get(LegalType);
- Value *OrigPtr = OrigLI.getPointerOperand();
- // If we're needing to spill something into more than one load, its legal
- // type will be a vector (ex. an i256 load will have LegalType = <8 x i32>).
- // But if we're already a scalar (which can happen if we're splitting up a
- // struct), the element type will be the legal type itself.
- Type *ElemType = LegalType->getScalarType();
- unsigned ElemBytes = DL.getTypeStoreSize(ElemType);
- AAMDNodes AANodes = OrigLI.getAAMetadata();
- if (IsAggPart && Slices.empty())
- Slices.push_back(VecSlice{/*Index=*/0, /*Length=*/1});
- for (VecSlice S : Slices) {
- Type *SliceType =
- S.Length != 1 ? FixedVectorType::get(ElemType, S.Length) : ElemType;
- int64_t ByteOffset = AggByteOff + S.Index * ElemBytes;
- // You can't reasonably expect loads to wrap around the edge of memory.
- Value *NewPtr = IRB.CreateGEP(
- IRB.getInt8Ty(), OrigLI.getPointerOperand(), IRB.getInt32(ByteOffset),
- OrigPtr->getName() + ".off.ptr." + Twine(ByteOffset),
- GEPNoWrapFlags::noUnsignedWrap());
- Type *LoadableType = intrinsicTypeFor(SliceType);
- LoadInst *NewLI = IRB.CreateAlignedLoad(
- LoadableType, NewPtr, commonAlignment(OrigLI.getAlign(), ByteOffset),
- Name + ".off." + Twine(ByteOffset));
- copyMetadataForLoad(*NewLI, OrigLI);
- NewLI->setAAMetadata(
- AANodes.adjustForAccess(ByteOffset, LoadableType, DL));
- NewLI->setAtomic(OrigLI.getOrdering(), OrigLI.getSyncScopeID());
- NewLI->setVolatile(OrigLI.isVolatile());
- Value *Loaded = IRB.CreateBitCast(NewLI, SliceType,
- NewLI->getName() + ".from.loadable");
- LoadsRes = insertSlice(LoadsRes, Loaded, S, Name);
- }
- }
- if (LegalType != ArrayAsVecType)
- LoadsRes = makeIllegalNonAggregate(LoadsRes, ArrayAsVecType, Name);
- if (ArrayAsVecType != PartType)
- LoadsRes = vectorToArray(LoadsRes, PartType, Name);
-
- if (IsAggPart)
- Result = IRB.CreateInsertValue(Result, LoadsRes, AggIdxs, Name);
- else
- Result = LoadsRes;
- return true;
-}
-
-bool LegalizeBufferContentTypesVisitor::visitLoadInst(LoadInst &LI) {
- if (LI.getPointerAddressSpace() != AMDGPUAS::BUFFER_FAT_POINTER)
- return false;
-
- SmallVector<uint32_t> AggIdxs;
- Type *OrigType = LI.getType();
- Value *Result = PoisonValue::get(OrigType);
- bool Changed = visitLoadImpl(LI, OrigType, AggIdxs, 0, Result, LI.getName());
- if (!Changed)
- return false;
- Result->takeName(&LI);
- LI.replaceAllUsesWith(Result);
- LI.eraseFromParent();
- return Changed;
-}
-
-std::pair<bool, bool> LegalizeBufferContentTypesVisitor::visitStoreImpl(
- StoreInst &OrigSI, Type *PartType, SmallVectorImpl<uint32_t> &AggIdxs,
- uint64_t AggByteOff, const Twine &Name) {...
[truncated]
|
shiltian
approved these changes
Jan 20, 2025
|
technically we can directly revert a commit w/o a PR. |
|
... Dang, I had an updating PR up but we might as well get the revert in |
krzysz00
added a commit
to krzysz00/llvm-project
that referenced
this pull request
Jan 20, 2025
…owering" (llvm#123657) This reverts commit 64749fb. Adds a constructor to VecSlice to address the failure
searlmc1
pushed a commit
to ROCm/llvm-project
that referenced
this pull request
Mar 24, 2025
…wering" (llvm#123657) Reverts llvm#110572 Seem to have broken a buildbot, not sure why https://lab.llvm.org/buildbot/#/builders/108/builds/8346
searlmc1
pushed a commit
to ROCm/llvm-project
that referenced
this pull request
Mar 24, 2025
…owering" (llvm#123660) (llvm#123657) This reverts commit 64749fb. Adds a constructor to VecSlice to address the failure
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Reverts #110572
Seem to have broken a buildbot, not sure why https://lab.llvm.org/buildbot/#/builders/108/builds/8346