Skip to content

[clang] [libc++] fix _Atomic c11 compare exchange does not update expected results #78707

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Open
wants to merge 6 commits into
base: main
Choose a base branch
from
Open

[clang] [libc++] fix _Atomic c11 compare exchange does not update expected results #78707

wants to merge 6 commits into from

Conversation

huixie90
Copy link
Member

@huixie90 huixie90 commented Jan 19, 2024 •
edited
Loading

fixes #30023

The issue is that for compare exchange builtin, if the type's size is not power of 2, it creates a temporary of size power of 2, then emit the compare exchange operation. And later, the results of the compare exchange operation has two components: 1. a boolean whether or not the exchange happens. 2. the old value
we are supposed to write the old value into user's "expected" value. However, in case the type is not power of 2, what we actually wrote to is the temporary that was created.

The fix is to pass the "expected" address all the way down so it can wrote to the correct address

@github-actions GitHub Actions
Copy link

github-actions bot commented Jan 19, 2024 •
edited
Loading

✅ With the latest revision this PR passed the C/C++ code formatter.

ldionne
ldionne reviewed Jan 19, 2024
View reviewed changes
@huixie90 huixie90 changed the title [libc++] fix _Atomic c11 compare exchange does not update expected results [clang] [libc++] fix _Atomic c11 compare exchange does not update expected results Jun 7, 2025
@huixie90 huixie90 marked this pull request as ready for review June 7, 2025 07:32
@huixie90 huixie90 requested a review from a team as a code owner June 7, 2025 07:32
@llvmbot llvmbot added clang Clang issues not falling into any other category libc++ libc++ C++ Standard Library. Not GNU libstdc++. Not libc++abi. clang:codegen IR generation bugs: mangling, exceptions, etc. labels Jun 7, 2025
@llvmbot
Copy link
Member

llvmbot commented Jun 7, 2025

@llvm/pr-subscribers-clang-codegen

Author: Hui (huixie90)

Changes

fixes #30023

Patch is 24.34 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/78707.diff

3 Files Affected:

  • (modified) clang/lib/CodeGen/CGAtomic.cpp (+69-23)
  • (added) clang/test/CodeGenCXX/builtin-atomic-compare_exchange.cpp (+126)
  • (added) libcxx/test/std/atomics/atomics.types.generic/30023.pass.cpp (+65)
diff --git a/clang/lib/CodeGen/CGAtomic.cpp b/clang/lib/CodeGen/CGAtomic.cpp
index 51f0799a792fd..9eee834a520e0 100644
--- a/clang/lib/CodeGen/CGAtomic.cpp
+++ b/clang/lib/CodeGen/CGAtomic.cpp
@@ -376,6 +376,7 @@ bool AtomicInfo::emitMemSetZeroIfNecessary() const {
 static void emitAtomicCmpXchg(CodeGenFunction &CGF, AtomicExpr *E, bool IsWeak,
                               Address Dest, Address Ptr,
                               Address Val1, Address Val2,
+                              Address ExpectedResult,
                               uint64_t Size,
                               llvm::AtomicOrdering SuccessOrder,
                               llvm::AtomicOrdering FailureOrder,
@@ -411,7 +412,48 @@ static void emitAtomicCmpXchg(CodeGenFunction &CGF, AtomicExpr *E, bool IsWeak,
 
   CGF.Builder.SetInsertPoint(StoreExpectedBB);
   // Update the memory at Expected with Old's value.
-  CGF.Builder.CreateStore(Old, Val1);
+llvm::Type *ExpectedType = ExpectedResult.getElementType();
+const llvm::DataLayout &DL = CGF.CGM.getDataLayout();
+uint64_t ExpectedSizeInBytes = DL.getTypeStoreSize(ExpectedType);
+
+if (ExpectedSizeInBytes == Size) {
+  // Sizes match: store directly
+  CGF.Builder.CreateStore(Old, ExpectedResult);
+
+} else {
+  // store only the first ExpectedSizeInBytes bytes of Old
+  llvm::Type *OldType = Old->getType();
+
+  llvm::Align SrcAlignLLVM = DL.getABITypeAlign(OldType);
+  llvm::Align DstAlignLLVM = DL.getABITypeAlign(ExpectedType);
+
+  clang::CharUnits SrcAlign = clang::CharUnits::fromQuantity(SrcAlignLLVM.value());
+  clang::CharUnits DstAlign = clang::CharUnits::fromQuantity(DstAlignLLVM.value());
+
+  // Allocate temporary storage for Old value
+  llvm::AllocaInst *Alloca = CGF.CreateTempAlloca(OldType, "old.tmp");
+
+  // Wrap into clang::CodeGen::Address with proper type and alignment
+  Address OldStorage(Alloca, OldType, SrcAlign);
+
+  // Store Old into this temporary
+  CGF.Builder.CreateStore(Old, OldStorage);
+
+  // Bitcast pointers to i8*
+  llvm::Type *I8PtrTy = llvm::PointerType::getUnqual(CGF.getLLVMContext());
+
+  llvm::Value *SrcPtr = CGF.Builder.CreateBitCast(OldStorage.getBasePointer(), I8PtrTy);
+  llvm::Value *DstPtr = CGF.Builder.CreateBitCast(ExpectedResult.getBasePointer(), I8PtrTy);
+
+  // Perform memcpy for first ExpectedSizeInBytes bytes
+  CGF.Builder.CreateMemCpy(
+    DstPtr, DstAlignLLVM,
+    SrcPtr, SrcAlignLLVM,
+    llvm::ConstantInt::get(CGF.IntPtrTy, ExpectedSizeInBytes),
+    /*isVolatile=*/false);
+}
+
+  
   // Finally, branch to the exit point.
   CGF.Builder.CreateBr(ContinueBB);
 
@@ -426,6 +468,7 @@ static void emitAtomicCmpXchg(CodeGenFunction &CGF, AtomicExpr *E, bool IsWeak,
 static void emitAtomicCmpXchgFailureSet(CodeGenFunction &CGF, AtomicExpr *E,
                                         bool IsWeak, Address Dest, Address Ptr,
                                         Address Val1, Address Val2,
+                                        Address ExpectedResult,
                                         llvm::Value *FailureOrderVal,
                                         uint64_t Size,
                                         llvm::AtomicOrdering SuccessOrder,
@@ -456,7 +499,7 @@ static void emitAtomicCmpXchgFailureSet(CodeGenFunction &CGF, AtomicExpr *E,
     // success argument". This condition has been lifted and the only
     // precondition is 31.7.2.18. Effectively treat this as a DR and skip
     // language version checks.
-    emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2, Size, SuccessOrder,
+    emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2, ExpectedResult, Size, SuccessOrder,
                       FailureOrder, Scope);
     return;
   }
@@ -481,17 +524,17 @@ static void emitAtomicCmpXchgFailureSet(CodeGenFunction &CGF, AtomicExpr *E,
 
   // Emit all the different atomics
   CGF.Builder.SetInsertPoint(MonotonicBB);
-  emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2,
+  emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2, ExpectedResult,
                     Size, SuccessOrder, llvm::AtomicOrdering::Monotonic, Scope);
   CGF.Builder.CreateBr(ContBB);
 
   CGF.Builder.SetInsertPoint(AcquireBB);
-  emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2, Size, SuccessOrder,
+  emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2, ExpectedResult, Size, SuccessOrder,
                     llvm::AtomicOrdering::Acquire, Scope);
   CGF.Builder.CreateBr(ContBB);
 
   CGF.Builder.SetInsertPoint(SeqCstBB);
-  emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2, Size, SuccessOrder,
+  emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2, ExpectedResult, Size, SuccessOrder,
                     llvm::AtomicOrdering::SequentiallyConsistent, Scope);
   CGF.Builder.CreateBr(ContBB);
 
@@ -524,6 +567,7 @@ static llvm::Value *EmitPostAtomicMinMax(CGBuilderTy &Builder,
 
 static void EmitAtomicOp(CodeGenFunction &CGF, AtomicExpr *E, Address Dest,
                          Address Ptr, Address Val1, Address Val2,
+                         Address ExpectedResult,
                          llvm::Value *IsWeak, llvm::Value *FailureOrder,
                          uint64_t Size, llvm::AtomicOrdering Order,
                          llvm::SyncScope::ID Scope) {
@@ -540,13 +584,13 @@ static void EmitAtomicOp(CodeGenFunction &CGF, AtomicExpr *E, Address Dest,
   case AtomicExpr::AO__hip_atomic_compare_exchange_strong:
   case AtomicExpr::AO__opencl_atomic_compare_exchange_strong:
     emitAtomicCmpXchgFailureSet(CGF, E, false, Dest, Ptr, Val1, Val2,
-                                FailureOrder, Size, Order, Scope);
+                                ExpectedResult, FailureOrder, Size, Order, Scope);
     return;
   case AtomicExpr::AO__c11_atomic_compare_exchange_weak:
   case AtomicExpr::AO__opencl_atomic_compare_exchange_weak:
   case AtomicExpr::AO__hip_atomic_compare_exchange_weak:
     emitAtomicCmpXchgFailureSet(CGF, E, true, Dest, Ptr, Val1, Val2,
-                                FailureOrder, Size, Order, Scope);
+                                ExpectedResult, FailureOrder, Size, Order, Scope);
     return;
   case AtomicExpr::AO__atomic_compare_exchange:
   case AtomicExpr::AO__atomic_compare_exchange_n:
@@ -554,7 +598,7 @@ static void EmitAtomicOp(CodeGenFunction &CGF, AtomicExpr *E, Address Dest,
   case AtomicExpr::AO__scoped_atomic_compare_exchange_n: {
     if (llvm::ConstantInt *IsWeakC = dyn_cast<llvm::ConstantInt>(IsWeak)) {
       emitAtomicCmpXchgFailureSet(CGF, E, IsWeakC->getZExtValue(), Dest, Ptr,
-                                  Val1, Val2, FailureOrder, Size, Order, Scope);
+                                  Val1, Val2, ExpectedResult, FailureOrder, Size, Order, Scope);
     } else {
       // Create all the relevant BB's
       llvm::BasicBlock *StrongBB =
@@ -568,12 +612,12 @@ static void EmitAtomicOp(CodeGenFunction &CGF, AtomicExpr *E, Address Dest,
 
       CGF.Builder.SetInsertPoint(StrongBB);
       emitAtomicCmpXchgFailureSet(CGF, E, false, Dest, Ptr, Val1, Val2,
-                                  FailureOrder, Size, Order, Scope);
+                                  ExpectedResult, FailureOrder, Size, Order, Scope);
       CGF.Builder.CreateBr(ContBB);
 
       CGF.Builder.SetInsertPoint(WeakBB);
       emitAtomicCmpXchgFailureSet(CGF, E, true, Dest, Ptr, Val1, Val2,
-                                  FailureOrder, Size, Order, Scope);
+                                  ExpectedResult, FailureOrder, Size, Order, Scope);
       CGF.Builder.CreateBr(ContBB);
 
       CGF.Builder.SetInsertPoint(ContBB);
@@ -777,6 +821,7 @@ EmitValToTemp(CodeGenFunction &CGF, Expr *E) {
 
 static void EmitAtomicOp(CodeGenFunction &CGF, AtomicExpr *Expr, Address Dest,
                          Address Ptr, Address Val1, Address Val2,
+                         Address OriginalVal1,
                          llvm::Value *IsWeak, llvm::Value *FailureOrder,
                          uint64_t Size, llvm::AtomicOrdering Order,
                          llvm::Value *Scope) {
@@ -796,7 +841,7 @@ static void EmitAtomicOp(CodeGenFunction &CGF, AtomicExpr *Expr, Address Dest,
                                                    Order, CGF.getLLVMContext());
     else
       SS = llvm::SyncScope::System;
-    EmitAtomicOp(CGF, Expr, Dest, Ptr, Val1, Val2, IsWeak, FailureOrder, Size,
+    EmitAtomicOp(CGF, Expr, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, FailureOrder, Size,
                  Order, SS);
     return;
   }
@@ -806,7 +851,7 @@ static void EmitAtomicOp(CodeGenFunction &CGF, AtomicExpr *Expr, Address Dest,
     auto SCID = CGF.getTargetHooks().getLLVMSyncScopeID(
         CGF.CGM.getLangOpts(), ScopeModel->map(SC->getZExtValue()),
         Order, CGF.CGM.getLLVMContext());
-    EmitAtomicOp(CGF, Expr, Dest, Ptr, Val1, Val2, IsWeak, FailureOrder, Size,
+    EmitAtomicOp(CGF, Expr, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, FailureOrder, Size,
                  Order, SCID);
     return;
   }
@@ -832,7 +877,7 @@ static void EmitAtomicOp(CodeGenFunction &CGF, AtomicExpr *Expr, Address Dest,
       SI->addCase(Builder.getInt32(S), B);
 
     Builder.SetInsertPoint(B);
-    EmitAtomicOp(CGF, Expr, Dest, Ptr, Val1, Val2, IsWeak, FailureOrder, Size,
+    EmitAtomicOp(CGF, Expr, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, FailureOrder, Size,
                  Order,
                  CGF.getTargetHooks().getLLVMSyncScopeID(CGF.CGM.getLangOpts(),
                                                          ScopeModel->map(S),
@@ -1036,6 +1081,7 @@ RValue CodeGenFunction::EmitAtomicExpr(AtomicExpr *E) {
   LValue AtomicVal = MakeAddrLValue(Ptr, AtomicTy);
   AtomicInfo Atomics(*this, AtomicVal);
 
+  Address OriginalVal1 = Val1;
   if (ShouldCastToIntPtrTy) {
     Ptr = Atomics.castToAtomicIntPointer(Ptr);
     if (Val1.isValid())
@@ -1279,30 +1325,30 @@ RValue CodeGenFunction::EmitAtomicExpr(AtomicExpr *E) {
     if (llvm::isValidAtomicOrderingCABI(ord))
       switch ((llvm::AtomicOrderingCABI)ord) {
       case llvm::AtomicOrderingCABI::relaxed:
-        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                      llvm::AtomicOrdering::Monotonic, Scope);
         break;
       case llvm::AtomicOrderingCABI::consume:
       case llvm::AtomicOrderingCABI::acquire:
         if (IsStore)
           break; // Avoid crashing on code with undefined behavior
-        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                      llvm::AtomicOrdering::Acquire, Scope);
         break;
       case llvm::AtomicOrderingCABI::release:
         if (IsLoad)
           break; // Avoid crashing on code with undefined behavior
-        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                      llvm::AtomicOrdering::Release, Scope);
         break;
       case llvm::AtomicOrderingCABI::acq_rel:
         if (IsLoad || IsStore)
           break; // Avoid crashing on code with undefined behavior
-        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                      llvm::AtomicOrdering::AcquireRelease, Scope);
         break;
       case llvm::AtomicOrderingCABI::seq_cst:
-        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                      llvm::AtomicOrdering::SequentiallyConsistent, Scope);
         break;
       }
@@ -1338,12 +1384,12 @@ RValue CodeGenFunction::EmitAtomicExpr(AtomicExpr *E) {
 
   // Emit all the different atomics
   Builder.SetInsertPoint(MonotonicBB);
-  EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+  EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                llvm::AtomicOrdering::Monotonic, Scope);
   Builder.CreateBr(ContBB);
   if (!IsStore) {
     Builder.SetInsertPoint(AcquireBB);
-    EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+    EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                  llvm::AtomicOrdering::Acquire, Scope);
     Builder.CreateBr(ContBB);
     SI->addCase(Builder.getInt32((int)llvm::AtomicOrderingCABI::consume),
@@ -1353,7 +1399,7 @@ RValue CodeGenFunction::EmitAtomicExpr(AtomicExpr *E) {
   }
   if (!IsLoad) {
     Builder.SetInsertPoint(ReleaseBB);
-    EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+    EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                  llvm::AtomicOrdering::Release, Scope);
     Builder.CreateBr(ContBB);
     SI->addCase(Builder.getInt32((int)llvm::AtomicOrderingCABI::release),
@@ -1361,14 +1407,14 @@ RValue CodeGenFunction::EmitAtomicExpr(AtomicExpr *E) {
   }
   if (!IsLoad && !IsStore) {
     Builder.SetInsertPoint(AcqRelBB);
-    EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+    EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                  llvm::AtomicOrdering::AcquireRelease, Scope);
     Builder.CreateBr(ContBB);
     SI->addCase(Builder.getInt32((int)llvm::AtomicOrderingCABI::acq_rel),
                 AcqRelBB);
   }
   Builder.SetInsertPoint(SeqCstBB);
-  EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+  EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                llvm::AtomicOrdering::SequentiallyConsistent, Scope);
   Builder.CreateBr(ContBB);
   SI->addCase(Builder.getInt32((int)llvm::AtomicOrderingCABI::seq_cst),
diff --git a/clang/test/CodeGenCXX/builtin-atomic-compare_exchange.cpp b/clang/test/CodeGenCXX/builtin-atomic-compare_exchange.cpp
new file mode 100644
index 0000000000000..4e2830e7a66cd
--- /dev/null
+++ b/clang/test/CodeGenCXX/builtin-atomic-compare_exchange.cpp
@@ -0,0 +1,126 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 5
+// RUN: %clang_cc1 -std=c++20 -triple=x86_64-linux-gnu -emit-llvm -o - %s | FileCheck %s
+
+
+template <unsigned Size>
+struct S {
+    char data[Size];
+};
+
+// CHECK-LABEL: define dso_local noundef zeroext i1 @_Z21test_compare_exchangePU7_Atomic1SILj3EEPS0_S0_(
+// CHECK-SAME: ptr noundef [[A:%.*]], ptr noundef [[EXPECTED:%.*]], i24 [[DESIRED_COERCE:%.*]]) #[[ATTR0:[0-9]+]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[DESIRED:%.*]] = alloca [[STRUCT_S:%.*]], align 1
+// CHECK-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT:    [[EXPECTED_ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT:    [[DOTATOMICTMP:%.*]] = alloca [[STRUCT_S]], align 1
+// CHECK-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca { [[STRUCT_S]], [1 x i8] }, align 4
+// CHECK-NEXT:    [[ATOMIC_TEMP1:%.*]] = alloca { [[STRUCT_S]], [1 x i8] }, align 4
+// CHECK-NEXT:    [[CMPXCHG_BOOL:%.*]] = alloca i8, align 1
+// CHECK-NEXT:    [[OLD_TMP:%.*]] = alloca i32, align 4
+// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds nuw [[STRUCT_S]], ptr [[DESIRED]], i32 0, i32 0
+// CHECK-NEXT:    store i24 [[DESIRED_COERCE]], ptr [[COERCE_DIVE]], align 1
+// CHECK-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// CHECK-NEXT:    store ptr [[EXPECTED]], ptr [[EXPECTED_ADDR]], align 8
+// CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// CHECK-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[EXPECTED_ADDR]], align 8
+// CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 1 [[DOTATOMICTMP]], ptr align 1 [[DESIRED]], i64 3, i1 false)
+// CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ATOMIC_TEMP]], ptr align 1 [[TMP1]], i64 3, i1 false)
+// CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ATOMIC_TEMP1]], ptr align 1 [[DOTATOMICTMP]], i64 3, i1 false)
+// CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[ATOMIC_TEMP]], align 4
+// CHECK-NEXT:    [[TMP3:%.*]] = load i32, ptr [[ATOMIC_TEMP1]], align 4
+// CHECK-NEXT:    [[TMP4:%.*]] = cmpxchg ptr [[TMP0]], i32 [[TMP2]], i32 [[TMP3]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP5:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
+// CHECK-NEXT:    br i1 [[TMP6]], label %[[CMPXCHG_CONTINUE:.*]], label %[[CMPXCHG_STORE_EXPECTED:.*]]
+// CHECK:       [[CMPXCHG_STORE_EXPECTED]]:
+// CHECK-NEXT:    store i32 [[TMP5]], ptr [[OLD_TMP]], align 4
+// CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 1 [[TMP1]], ptr align 4 [[OLD_TMP]], i64 3, i1 false)
+// CHECK-NEXT:    br label %[[CMPXCHG_CONTINUE]]
+// CHECK:       [[CMPXCHG_CONTINUE]]:
+// CHECK-NEXT:    [[STOREDV:%.*]] = zext i1 [[TMP6]] to i8
+// CHECK-NEXT:    store i8 [[STOREDV]], ptr [[CMPXCHG_BOOL]], align 1
+// CHECK-NEXT:    [[TMP7:%.*]] = load i8, ptr [[CMPXCHG_BOOL]], align 1
+// CHECK-NEXT:    [[LOADEDV:%.*]] = trunc i8 [[TMP7]] to i1
+// CHECK-NEXT:    ret i1 [[LOADEDV]]
+//
+bool test_compare_exchange(_Atomic(S<3>)* a, S<3>* expected, S<3> desired) {
+    return __c11_atomic_compare_exchange_strong(a, expected, desired, 0, 0);
+}
+
+
+// CHECK-LABEL: define dso_local noundef zeroext i1 @_Z21test_compare_exchangePU7_Atomic1SILj4EEPS0_S0_(
+// CHECK-SAME: ptr noundef [[A:%.*]], ptr noundef [[EXPECTED:%.*]], i32 [[DESIRED_COERCE:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[DESIRED:%.*]] = alloca [[STRUCT_S_0:%.*]], align 1
+// CHECK-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT:    [[EXPECTED_ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT:    [[DOTATOMICTMP:%.*]] = alloca [[STRUCT_S_0]], align 1
+// CHECK-NEXT:    [[CMPXCHG_BOOL:%.*]] = alloca i8, align 1
+// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds nuw [[STRUCT_S_0]], ptr [[DESIRED]], i32 0, i32 0
+// CHECK-NEXT:    store i32 [[DESIRED_COERCE]], ptr [[COERCE_DIVE]], align 1
+// CHECK-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// CHECK-NEXT:    store ptr [[EXPECTED]], ptr [[EXPECTED_ADDR]], align 8
+// CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// CHECK-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[EXPECTED_ADDR]], align 8
+// CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 1 [[DOTATOMICTMP]], ptr align 1 [[DESIRED]], i64 4, i1 false)
+// CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 1
+// CHECK-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTATOMICTMP]], align 1
+// CHECK-NEXT:    [[TMP4:%.*]] = cmpxchg ptr [[TMP0]], i32 [[TMP2]], i32 [[TMP3]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP5:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
+// CHECK-NEXT:    br i1 [[TMP6]], label %[[CMPXCHG_CONTINUE:.*]], label %[[CMPXCHG_STORE_EXPECTED:.*]]
+// CHECK:       [[CMPXCHG_STORE_EXPECTED]]:
+// CHECK-NEXT:    store i32 [[TMP5]], ptr [[TMP1]], align 1
+// CHECK-NEXT:    br label %[[CMPXCHG_CONTINUE]]
+// CHECK:       [[CMPXCHG_CONTINUE]]:
+// CHECK-NEXT:    [[STOREDV:%.*]] = zext i1 [[TMP6]] to i8
+// CHECK-NEXT:    store i8 [[STOREDV]], ptr [[CMPXCHG_BOOL]], align 1
+// CHECK-NEXT:    [[TMP7:%.*]] = load i8, ptr [[CMPXCHG_BOOL]], align 1
+// CHECK-NEXT:    [[LOADEDV:%.*]] = trunc i8 [[TMP7]] to i1
+// CHECK-NEXT:    ret i1 [[LOADEDV]]
+//
+bool test_compare_exchange(_Atomic(S<4>)* a, S<4>* expected, S<4> desired) {
+    return __c11_atomic_compare_exchange_strong(a, expected, desired, 0, 0);
+}
+
+// CHECK-LABEL: define dso_local noundef zeroext i1 @_Z21test_compare_exchangePU7_Atomic1SILj6EEPS0_S0_(
+// CHECK-SAME: ptr noundef [[A:%.*]], ptr noundef [[EXPECTED:%.*]], i48 [[DESIRED_COERCE:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[DESIRED:%.*]] = alloca [[STRUCT_S_1:%.*]], align 1
+// CHECK-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT:    [[EXPECTED_ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT:    [[DOTATOMICTMP:%.*]] = alloca [[STRUCT_S_1]], align 1
+// CHECK-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca { [[STRUCT_S_1]], [2 x i8] }, align 8
+// CHECK-NEXT:    [[ATOMIC_TEMP1:%.*]] = alloca { [[STRUCT_S_1]], [2 x i8...
[truncated]

@llvmbot
Copy link
Member

llvmbot commented Jun 7, 2025

@llvm/pr-subscribers-libcxx

Author: Hui (huixie90)

Changes

fixes #30023

Patch is 24.34 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/78707.diff

3 Files Affected:

  • (modified) clang/lib/CodeGen/CGAtomic.cpp (+69-23)
  • (added) clang/test/CodeGenCXX/builtin-atomic-compare_exchange.cpp (+126)
  • (added) libcxx/test/std/atomics/atomics.types.generic/30023.pass.cpp (+65)
diff --git a/clang/lib/CodeGen/CGAtomic.cpp b/clang/lib/CodeGen/CGAtomic.cpp
index 51f0799a792fd..9eee834a520e0 100644
--- a/clang/lib/CodeGen/CGAtomic.cpp
+++ b/clang/lib/CodeGen/CGAtomic.cpp
@@ -376,6 +376,7 @@ bool AtomicInfo::emitMemSetZeroIfNecessary() const {
 static void emitAtomicCmpXchg(CodeGenFunction &CGF, AtomicExpr *E, bool IsWeak,
                               Address Dest, Address Ptr,
                               Address Val1, Address Val2,
+                              Address ExpectedResult,
                               uint64_t Size,
                               llvm::AtomicOrdering SuccessOrder,
                               llvm::AtomicOrdering FailureOrder,
@@ -411,7 +412,48 @@ static void emitAtomicCmpXchg(CodeGenFunction &CGF, AtomicExpr *E, bool IsWeak,
 
   CGF.Builder.SetInsertPoint(StoreExpectedBB);
   // Update the memory at Expected with Old's value.
-  CGF.Builder.CreateStore(Old, Val1);
+llvm::Type *ExpectedType = ExpectedResult.getElementType();
+const llvm::DataLayout &DL = CGF.CGM.getDataLayout();
+uint64_t ExpectedSizeInBytes = DL.getTypeStoreSize(ExpectedType);
+
+if (ExpectedSizeInBytes == Size) {
+  // Sizes match: store directly
+  CGF.Builder.CreateStore(Old, ExpectedResult);
+
+} else {
+  // store only the first ExpectedSizeInBytes bytes of Old
+  llvm::Type *OldType = Old->getType();
+
+  llvm::Align SrcAlignLLVM = DL.getABITypeAlign(OldType);
+  llvm::Align DstAlignLLVM = DL.getABITypeAlign(ExpectedType);
+
+  clang::CharUnits SrcAlign = clang::CharUnits::fromQuantity(SrcAlignLLVM.value());
+  clang::CharUnits DstAlign = clang::CharUnits::fromQuantity(DstAlignLLVM.value());
+
+  // Allocate temporary storage for Old value
+  llvm::AllocaInst *Alloca = CGF.CreateTempAlloca(OldType, "old.tmp");
+
+  // Wrap into clang::CodeGen::Address with proper type and alignment
+  Address OldStorage(Alloca, OldType, SrcAlign);
+
+  // Store Old into this temporary
+  CGF.Builder.CreateStore(Old, OldStorage);
+
+  // Bitcast pointers to i8*
+  llvm::Type *I8PtrTy = llvm::PointerType::getUnqual(CGF.getLLVMContext());
+
+  llvm::Value *SrcPtr = CGF.Builder.CreateBitCast(OldStorage.getBasePointer(), I8PtrTy);
+  llvm::Value *DstPtr = CGF.Builder.CreateBitCast(ExpectedResult.getBasePointer(), I8PtrTy);
+
+  // Perform memcpy for first ExpectedSizeInBytes bytes
+  CGF.Builder.CreateMemCpy(
+    DstPtr, DstAlignLLVM,
+    SrcPtr, SrcAlignLLVM,
+    llvm::ConstantInt::get(CGF.IntPtrTy, ExpectedSizeInBytes),
+    /*isVolatile=*/false);
+}
+
+  
   // Finally, branch to the exit point.
   CGF.Builder.CreateBr(ContinueBB);
 
@@ -426,6 +468,7 @@ static void emitAtomicCmpXchg(CodeGenFunction &CGF, AtomicExpr *E, bool IsWeak,
 static void emitAtomicCmpXchgFailureSet(CodeGenFunction &CGF, AtomicExpr *E,
                                         bool IsWeak, Address Dest, Address Ptr,
                                         Address Val1, Address Val2,
+                                        Address ExpectedResult,
                                         llvm::Value *FailureOrderVal,
                                         uint64_t Size,
                                         llvm::AtomicOrdering SuccessOrder,
@@ -456,7 +499,7 @@ static void emitAtomicCmpXchgFailureSet(CodeGenFunction &CGF, AtomicExpr *E,
     // success argument". This condition has been lifted and the only
     // precondition is 31.7.2.18. Effectively treat this as a DR and skip
     // language version checks.
-    emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2, Size, SuccessOrder,
+    emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2, ExpectedResult, Size, SuccessOrder,
                       FailureOrder, Scope);
     return;
   }
@@ -481,17 +524,17 @@ static void emitAtomicCmpXchgFailureSet(CodeGenFunction &CGF, AtomicExpr *E,
 
   // Emit all the different atomics
   CGF.Builder.SetInsertPoint(MonotonicBB);
-  emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2,
+  emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2, ExpectedResult,
                     Size, SuccessOrder, llvm::AtomicOrdering::Monotonic, Scope);
   CGF.Builder.CreateBr(ContBB);
 
   CGF.Builder.SetInsertPoint(AcquireBB);
-  emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2, Size, SuccessOrder,
+  emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2, ExpectedResult, Size, SuccessOrder,
                     llvm::AtomicOrdering::Acquire, Scope);
   CGF.Builder.CreateBr(ContBB);
 
   CGF.Builder.SetInsertPoint(SeqCstBB);
-  emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2, Size, SuccessOrder,
+  emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2, ExpectedResult, Size, SuccessOrder,
                     llvm::AtomicOrdering::SequentiallyConsistent, Scope);
   CGF.Builder.CreateBr(ContBB);
 
@@ -524,6 +567,7 @@ static llvm::Value *EmitPostAtomicMinMax(CGBuilderTy &Builder,
 
 static void EmitAtomicOp(CodeGenFunction &CGF, AtomicExpr *E, Address Dest,
                          Address Ptr, Address Val1, Address Val2,
+                         Address ExpectedResult,
                          llvm::Value *IsWeak, llvm::Value *FailureOrder,
                          uint64_t Size, llvm::AtomicOrdering Order,
                          llvm::SyncScope::ID Scope) {
@@ -540,13 +584,13 @@ static void EmitAtomicOp(CodeGenFunction &CGF, AtomicExpr *E, Address Dest,
   case AtomicExpr::AO__hip_atomic_compare_exchange_strong:
   case AtomicExpr::AO__opencl_atomic_compare_exchange_strong:
     emitAtomicCmpXchgFailureSet(CGF, E, false, Dest, Ptr, Val1, Val2,
-                                FailureOrder, Size, Order, Scope);
+                                ExpectedResult, FailureOrder, Size, Order, Scope);
     return;
   case AtomicExpr::AO__c11_atomic_compare_exchange_weak:
   case AtomicExpr::AO__opencl_atomic_compare_exchange_weak:
   case AtomicExpr::AO__hip_atomic_compare_exchange_weak:
     emitAtomicCmpXchgFailureSet(CGF, E, true, Dest, Ptr, Val1, Val2,
-                                FailureOrder, Size, Order, Scope);
+                                ExpectedResult, FailureOrder, Size, Order, Scope);
     return;
   case AtomicExpr::AO__atomic_compare_exchange:
   case AtomicExpr::AO__atomic_compare_exchange_n:
@@ -554,7 +598,7 @@ static void EmitAtomicOp(CodeGenFunction &CGF, AtomicExpr *E, Address Dest,
   case AtomicExpr::AO__scoped_atomic_compare_exchange_n: {
     if (llvm::ConstantInt *IsWeakC = dyn_cast<llvm::ConstantInt>(IsWeak)) {
       emitAtomicCmpXchgFailureSet(CGF, E, IsWeakC->getZExtValue(), Dest, Ptr,
-                                  Val1, Val2, FailureOrder, Size, Order, Scope);
+                                  Val1, Val2, ExpectedResult, FailureOrder, Size, Order, Scope);
     } else {
       // Create all the relevant BB's
       llvm::BasicBlock *StrongBB =
@@ -568,12 +612,12 @@ static void EmitAtomicOp(CodeGenFunction &CGF, AtomicExpr *E, Address Dest,
 
       CGF.Builder.SetInsertPoint(StrongBB);
       emitAtomicCmpXchgFailureSet(CGF, E, false, Dest, Ptr, Val1, Val2,
-                                  FailureOrder, Size, Order, Scope);
+                                  ExpectedResult, FailureOrder, Size, Order, Scope);
       CGF.Builder.CreateBr(ContBB);
 
       CGF.Builder.SetInsertPoint(WeakBB);
       emitAtomicCmpXchgFailureSet(CGF, E, true, Dest, Ptr, Val1, Val2,
-                                  FailureOrder, Size, Order, Scope);
+                                  ExpectedResult, FailureOrder, Size, Order, Scope);
       CGF.Builder.CreateBr(ContBB);
 
       CGF.Builder.SetInsertPoint(ContBB);
@@ -777,6 +821,7 @@ EmitValToTemp(CodeGenFunction &CGF, Expr *E) {
 
 static void EmitAtomicOp(CodeGenFunction &CGF, AtomicExpr *Expr, Address Dest,
                          Address Ptr, Address Val1, Address Val2,
+                         Address OriginalVal1,
                          llvm::Value *IsWeak, llvm::Value *FailureOrder,
                          uint64_t Size, llvm::AtomicOrdering Order,
                          llvm::Value *Scope) {
@@ -796,7 +841,7 @@ static void EmitAtomicOp(CodeGenFunction &CGF, AtomicExpr *Expr, Address Dest,
                                                    Order, CGF.getLLVMContext());
     else
       SS = llvm::SyncScope::System;
-    EmitAtomicOp(CGF, Expr, Dest, Ptr, Val1, Val2, IsWeak, FailureOrder, Size,
+    EmitAtomicOp(CGF, Expr, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, FailureOrder, Size,
                  Order, SS);
     return;
   }
@@ -806,7 +851,7 @@ static void EmitAtomicOp(CodeGenFunction &CGF, AtomicExpr *Expr, Address Dest,
     auto SCID = CGF.getTargetHooks().getLLVMSyncScopeID(
         CGF.CGM.getLangOpts(), ScopeModel->map(SC->getZExtValue()),
         Order, CGF.CGM.getLLVMContext());
-    EmitAtomicOp(CGF, Expr, Dest, Ptr, Val1, Val2, IsWeak, FailureOrder, Size,
+    EmitAtomicOp(CGF, Expr, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, FailureOrder, Size,
                  Order, SCID);
     return;
   }
@@ -832,7 +877,7 @@ static void EmitAtomicOp(CodeGenFunction &CGF, AtomicExpr *Expr, Address Dest,
       SI->addCase(Builder.getInt32(S), B);
 
     Builder.SetInsertPoint(B);
-    EmitAtomicOp(CGF, Expr, Dest, Ptr, Val1, Val2, IsWeak, FailureOrder, Size,
+    EmitAtomicOp(CGF, Expr, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, FailureOrder, Size,
                  Order,
                  CGF.getTargetHooks().getLLVMSyncScopeID(CGF.CGM.getLangOpts(),
                                                          ScopeModel->map(S),
@@ -1036,6 +1081,7 @@ RValue CodeGenFunction::EmitAtomicExpr(AtomicExpr *E) {
   LValue AtomicVal = MakeAddrLValue(Ptr, AtomicTy);
   AtomicInfo Atomics(*this, AtomicVal);
 
+  Address OriginalVal1 = Val1;
   if (ShouldCastToIntPtrTy) {
     Ptr = Atomics.castToAtomicIntPointer(Ptr);
     if (Val1.isValid())
@@ -1279,30 +1325,30 @@ RValue CodeGenFunction::EmitAtomicExpr(AtomicExpr *E) {
     if (llvm::isValidAtomicOrderingCABI(ord))
       switch ((llvm::AtomicOrderingCABI)ord) {
       case llvm::AtomicOrderingCABI::relaxed:
-        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                      llvm::AtomicOrdering::Monotonic, Scope);
         break;
       case llvm::AtomicOrderingCABI::consume:
       case llvm::AtomicOrderingCABI::acquire:
         if (IsStore)
           break; // Avoid crashing on code with undefined behavior
-        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                      llvm::AtomicOrdering::Acquire, Scope);
         break;
       case llvm::AtomicOrderingCABI::release:
         if (IsLoad)
           break; // Avoid crashing on code with undefined behavior
-        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                      llvm::AtomicOrdering::Release, Scope);
         break;
       case llvm::AtomicOrderingCABI::acq_rel:
         if (IsLoad || IsStore)
           break; // Avoid crashing on code with undefined behavior
-        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                      llvm::AtomicOrdering::AcquireRelease, Scope);
         break;
       case llvm::AtomicOrderingCABI::seq_cst:
-        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                      llvm::AtomicOrdering::SequentiallyConsistent, Scope);
         break;
       }
@@ -1338,12 +1384,12 @@ RValue CodeGenFunction::EmitAtomicExpr(AtomicExpr *E) {
 
   // Emit all the different atomics
   Builder.SetInsertPoint(MonotonicBB);
-  EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+  EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                llvm::AtomicOrdering::Monotonic, Scope);
   Builder.CreateBr(ContBB);
   if (!IsStore) {
     Builder.SetInsertPoint(AcquireBB);
-    EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+    EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                  llvm::AtomicOrdering::Acquire, Scope);
     Builder.CreateBr(ContBB);
     SI->addCase(Builder.getInt32((int)llvm::AtomicOrderingCABI::consume),
@@ -1353,7 +1399,7 @@ RValue CodeGenFunction::EmitAtomicExpr(AtomicExpr *E) {
   }
   if (!IsLoad) {
     Builder.SetInsertPoint(ReleaseBB);
-    EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+    EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                  llvm::AtomicOrdering::Release, Scope);
     Builder.CreateBr(ContBB);
     SI->addCase(Builder.getInt32((int)llvm::AtomicOrderingCABI::release),
@@ -1361,14 +1407,14 @@ RValue CodeGenFunction::EmitAtomicExpr(AtomicExpr *E) {
   }
   if (!IsLoad && !IsStore) {
     Builder.SetInsertPoint(AcqRelBB);
-    EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+    EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                  llvm::AtomicOrdering::AcquireRelease, Scope);
     Builder.CreateBr(ContBB);
     SI->addCase(Builder.getInt32((int)llvm::AtomicOrderingCABI::acq_rel),
                 AcqRelBB);
   }
   Builder.SetInsertPoint(SeqCstBB);
-  EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+  EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                llvm::AtomicOrdering::SequentiallyConsistent, Scope);
   Builder.CreateBr(ContBB);
   SI->addCase(Builder.getInt32((int)llvm::AtomicOrderingCABI::seq_cst),
diff --git a/clang/test/CodeGenCXX/builtin-atomic-compare_exchange.cpp b/clang/test/CodeGenCXX/builtin-atomic-compare_exchange.cpp
new file mode 100644
index 0000000000000..4e2830e7a66cd
--- /dev/null
+++ b/clang/test/CodeGenCXX/builtin-atomic-compare_exchange.cpp
@@ -0,0 +1,126 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 5
+// RUN: %clang_cc1 -std=c++20 -triple=x86_64-linux-gnu -emit-llvm -o - %s | FileCheck %s
+
+
+template <unsigned Size>
+struct S {
+    char data[Size];
+};
+
+// CHECK-LABEL: define dso_local noundef zeroext i1 @_Z21test_compare_exchangePU7_Atomic1SILj3EEPS0_S0_(
+// CHECK-SAME: ptr noundef [[A:%.*]], ptr noundef [[EXPECTED:%.*]], i24 [[DESIRED_COERCE:%.*]]) #[[ATTR0:[0-9]+]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[DESIRED:%.*]] = alloca [[STRUCT_S:%.*]], align 1
+// CHECK-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT:    [[EXPECTED_ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT:    [[DOTATOMICTMP:%.*]] = alloca [[STRUCT_S]], align 1
+// CHECK-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca { [[STRUCT_S]], [1 x i8] }, align 4
+// CHECK-NEXT:    [[ATOMIC_TEMP1:%.*]] = alloca { [[STRUCT_S]], [1 x i8] }, align 4
+// CHECK-NEXT:    [[CMPXCHG_BOOL:%.*]] = alloca i8, align 1
+// CHECK-NEXT:    [[OLD_TMP:%.*]] = alloca i32, align 4
+// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds nuw [[STRUCT_S]], ptr [[DESIRED]], i32 0, i32 0
+// CHECK-NEXT:    store i24 [[DESIRED_COERCE]], ptr [[COERCE_DIVE]], align 1
+// CHECK-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// CHECK-NEXT:    store ptr [[EXPECTED]], ptr [[EXPECTED_ADDR]], align 8
+// CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// CHECK-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[EXPECTED_ADDR]], align 8
+// CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 1 [[DOTATOMICTMP]], ptr align 1 [[DESIRED]], i64 3, i1 false)
+// CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ATOMIC_TEMP]], ptr align 1 [[TMP1]], i64 3, i1 false)
+// CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ATOMIC_TEMP1]], ptr align 1 [[DOTATOMICTMP]], i64 3, i1 false)
+// CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[ATOMIC_TEMP]], align 4
+// CHECK-NEXT:    [[TMP3:%.*]] = load i32, ptr [[ATOMIC_TEMP1]], align 4
+// CHECK-NEXT:    [[TMP4:%.*]] = cmpxchg ptr [[TMP0]], i32 [[TMP2]], i32 [[TMP3]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP5:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
+// CHECK-NEXT:    br i1 [[TMP6]], label %[[CMPXCHG_CONTINUE:.*]], label %[[CMPXCHG_STORE_EXPECTED:.*]]
+// CHECK:       [[CMPXCHG_STORE_EXPECTED]]:
+// CHECK-NEXT:    store i32 [[TMP5]], ptr [[OLD_TMP]], align 4
+// CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 1 [[TMP1]], ptr align 4 [[OLD_TMP]], i64 3, i1 false)
+// CHECK-NEXT:    br label %[[CMPXCHG_CONTINUE]]
+// CHECK:       [[CMPXCHG_CONTINUE]]:
+// CHECK-NEXT:    [[STOREDV:%.*]] = zext i1 [[TMP6]] to i8
+// CHECK-NEXT:    store i8 [[STOREDV]], ptr [[CMPXCHG_BOOL]], align 1
+// CHECK-NEXT:    [[TMP7:%.*]] = load i8, ptr [[CMPXCHG_BOOL]], align 1
+// CHECK-NEXT:    [[LOADEDV:%.*]] = trunc i8 [[TMP7]] to i1
+// CHECK-NEXT:    ret i1 [[LOADEDV]]
+//
+bool test_compare_exchange(_Atomic(S<3>)* a, S<3>* expected, S<3> desired) {
+    return __c11_atomic_compare_exchange_strong(a, expected, desired, 0, 0);
+}
+
+
+// CHECK-LABEL: define dso_local noundef zeroext i1 @_Z21test_compare_exchangePU7_Atomic1SILj4EEPS0_S0_(
+// CHECK-SAME: ptr noundef [[A:%.*]], ptr noundef [[EXPECTED:%.*]], i32 [[DESIRED_COERCE:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[DESIRED:%.*]] = alloca [[STRUCT_S_0:%.*]], align 1
+// CHECK-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT:    [[EXPECTED_ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT:    [[DOTATOMICTMP:%.*]] = alloca [[STRUCT_S_0]], align 1
+// CHECK-NEXT:    [[CMPXCHG_BOOL:%.*]] = alloca i8, align 1
+// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds nuw [[STRUCT_S_0]], ptr [[DESIRED]], i32 0, i32 0
+// CHECK-NEXT:    store i32 [[DESIRED_COERCE]], ptr [[COERCE_DIVE]], align 1
+// CHECK-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// CHECK-NEXT:    store ptr [[EXPECTED]], ptr [[EXPECTED_ADDR]], align 8
+// CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// CHECK-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[EXPECTED_ADDR]], align 8
+// CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 1 [[DOTATOMICTMP]], ptr align 1 [[DESIRED]], i64 4, i1 false)
+// CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 1
+// CHECK-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTATOMICTMP]], align 1
+// CHECK-NEXT:    [[TMP4:%.*]] = cmpxchg ptr [[TMP0]], i32 [[TMP2]], i32 [[TMP3]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP5:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
+// CHECK-NEXT:    br i1 [[TMP6]], label %[[CMPXCHG_CONTINUE:.*]], label %[[CMPXCHG_STORE_EXPECTED:.*]]
+// CHECK:       [[CMPXCHG_STORE_EXPECTED]]:
+// CHECK-NEXT:    store i32 [[TMP5]], ptr [[TMP1]], align 1
+// CHECK-NEXT:    br label %[[CMPXCHG_CONTINUE]]
+// CHECK:       [[CMPXCHG_CONTINUE]]:
+// CHECK-NEXT:    [[STOREDV:%.*]] = zext i1 [[TMP6]] to i8
+// CHECK-NEXT:    store i8 [[STOREDV]], ptr [[CMPXCHG_BOOL]], align 1
+// CHECK-NEXT:    [[TMP7:%.*]] = load i8, ptr [[CMPXCHG_BOOL]], align 1
+// CHECK-NEXT:    [[LOADEDV:%.*]] = trunc i8 [[TMP7]] to i1
+// CHECK-NEXT:    ret i1 [[LOADEDV]]
+//
+bool test_compare_exchange(_Atomic(S<4>)* a, S<4>* expected, S<4> desired) {
+    return __c11_atomic_compare_exchange_strong(a, expected, desired, 0, 0);
+}
+
+// CHECK-LABEL: define dso_local noundef zeroext i1 @_Z21test_compare_exchangePU7_Atomic1SILj6EEPS0_S0_(
+// CHECK-SAME: ptr noundef [[A:%.*]], ptr noundef [[EXPECTED:%.*]], i48 [[DESIRED_COERCE:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[DESIRED:%.*]] = alloca [[STRUCT_S_1:%.*]], align 1
+// CHECK-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT:    [[EXPECTED_ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT:    [[DOTATOMICTMP:%.*]] = alloca [[STRUCT_S_1]], align 1
+// CHECK-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca { [[STRUCT_S_1]], [2 x i8] }, align 8
+// CHECK-NEXT:    [[ATOMIC_TEMP1:%.*]] = alloca { [[STRUCT_S_1]], [2 x i8...
[truncated]

@llvmbot
Copy link
Member

llvmbot commented Jun 7, 2025

@llvm/pr-subscribers-clang

Author: Hui (huixie90)

Changes

fixes #30023

Patch is 24.34 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/78707.diff

3 Files Affected:

  • (modified) clang/lib/CodeGen/CGAtomic.cpp (+69-23)
  • (added) clang/test/CodeGenCXX/builtin-atomic-compare_exchange.cpp (+126)
  • (added) libcxx/test/std/atomics/atomics.types.generic/30023.pass.cpp (+65)
diff --git a/clang/lib/CodeGen/CGAtomic.cpp b/clang/lib/CodeGen/CGAtomic.cpp
index 51f0799a792fd..9eee834a520e0 100644
--- a/clang/lib/CodeGen/CGAtomic.cpp
+++ b/clang/lib/CodeGen/CGAtomic.cpp
@@ -376,6 +376,7 @@ bool AtomicInfo::emitMemSetZeroIfNecessary() const {
 static void emitAtomicCmpXchg(CodeGenFunction &CGF, AtomicExpr *E, bool IsWeak,
                               Address Dest, Address Ptr,
                               Address Val1, Address Val2,
+                              Address ExpectedResult,
                               uint64_t Size,
                               llvm::AtomicOrdering SuccessOrder,
                               llvm::AtomicOrdering FailureOrder,
@@ -411,7 +412,48 @@ static void emitAtomicCmpXchg(CodeGenFunction &CGF, AtomicExpr *E, bool IsWeak,
 
   CGF.Builder.SetInsertPoint(StoreExpectedBB);
   // Update the memory at Expected with Old's value.
-  CGF.Builder.CreateStore(Old, Val1);
+llvm::Type *ExpectedType = ExpectedResult.getElementType();
+const llvm::DataLayout &DL = CGF.CGM.getDataLayout();
+uint64_t ExpectedSizeInBytes = DL.getTypeStoreSize(ExpectedType);
+
+if (ExpectedSizeInBytes == Size) {
+  // Sizes match: store directly
+  CGF.Builder.CreateStore(Old, ExpectedResult);
+
+} else {
+  // store only the first ExpectedSizeInBytes bytes of Old
+  llvm::Type *OldType = Old->getType();
+
+  llvm::Align SrcAlignLLVM = DL.getABITypeAlign(OldType);
+  llvm::Align DstAlignLLVM = DL.getABITypeAlign(ExpectedType);
+
+  clang::CharUnits SrcAlign = clang::CharUnits::fromQuantity(SrcAlignLLVM.value());
+  clang::CharUnits DstAlign = clang::CharUnits::fromQuantity(DstAlignLLVM.value());
+
+  // Allocate temporary storage for Old value
+  llvm::AllocaInst *Alloca = CGF.CreateTempAlloca(OldType, "old.tmp");
+
+  // Wrap into clang::CodeGen::Address with proper type and alignment
+  Address OldStorage(Alloca, OldType, SrcAlign);
+
+  // Store Old into this temporary
+  CGF.Builder.CreateStore(Old, OldStorage);
+
+  // Bitcast pointers to i8*
+  llvm::Type *I8PtrTy = llvm::PointerType::getUnqual(CGF.getLLVMContext());
+
+  llvm::Value *SrcPtr = CGF.Builder.CreateBitCast(OldStorage.getBasePointer(), I8PtrTy);
+  llvm::Value *DstPtr = CGF.Builder.CreateBitCast(ExpectedResult.getBasePointer(), I8PtrTy);
+
+  // Perform memcpy for first ExpectedSizeInBytes bytes
+  CGF.Builder.CreateMemCpy(
+    DstPtr, DstAlignLLVM,
+    SrcPtr, SrcAlignLLVM,
+    llvm::ConstantInt::get(CGF.IntPtrTy, ExpectedSizeInBytes),
+    /*isVolatile=*/false);
+}
+
+  
   // Finally, branch to the exit point.
   CGF.Builder.CreateBr(ContinueBB);
 
@@ -426,6 +468,7 @@ static void emitAtomicCmpXchg(CodeGenFunction &CGF, AtomicExpr *E, bool IsWeak,
 static void emitAtomicCmpXchgFailureSet(CodeGenFunction &CGF, AtomicExpr *E,
                                         bool IsWeak, Address Dest, Address Ptr,
                                         Address Val1, Address Val2,
+                                        Address ExpectedResult,
                                         llvm::Value *FailureOrderVal,
                                         uint64_t Size,
                                         llvm::AtomicOrdering SuccessOrder,
@@ -456,7 +499,7 @@ static void emitAtomicCmpXchgFailureSet(CodeGenFunction &CGF, AtomicExpr *E,
     // success argument". This condition has been lifted and the only
     // precondition is 31.7.2.18. Effectively treat this as a DR and skip
     // language version checks.
-    emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2, Size, SuccessOrder,
+    emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2, ExpectedResult, Size, SuccessOrder,
                       FailureOrder, Scope);
     return;
   }
@@ -481,17 +524,17 @@ static void emitAtomicCmpXchgFailureSet(CodeGenFunction &CGF, AtomicExpr *E,
 
   // Emit all the different atomics
   CGF.Builder.SetInsertPoint(MonotonicBB);
-  emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2,
+  emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2, ExpectedResult,
                     Size, SuccessOrder, llvm::AtomicOrdering::Monotonic, Scope);
   CGF.Builder.CreateBr(ContBB);
 
   CGF.Builder.SetInsertPoint(AcquireBB);
-  emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2, Size, SuccessOrder,
+  emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2, ExpectedResult, Size, SuccessOrder,
                     llvm::AtomicOrdering::Acquire, Scope);
   CGF.Builder.CreateBr(ContBB);
 
   CGF.Builder.SetInsertPoint(SeqCstBB);
-  emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2, Size, SuccessOrder,
+  emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2, ExpectedResult, Size, SuccessOrder,
                     llvm::AtomicOrdering::SequentiallyConsistent, Scope);
   CGF.Builder.CreateBr(ContBB);
 
@@ -524,6 +567,7 @@ static llvm::Value *EmitPostAtomicMinMax(CGBuilderTy &Builder,
 
 static void EmitAtomicOp(CodeGenFunction &CGF, AtomicExpr *E, Address Dest,
                          Address Ptr, Address Val1, Address Val2,
+                         Address ExpectedResult,
                          llvm::Value *IsWeak, llvm::Value *FailureOrder,
                          uint64_t Size, llvm::AtomicOrdering Order,
                          llvm::SyncScope::ID Scope) {
@@ -540,13 +584,13 @@ static void EmitAtomicOp(CodeGenFunction &CGF, AtomicExpr *E, Address Dest,
   case AtomicExpr::AO__hip_atomic_compare_exchange_strong:
   case AtomicExpr::AO__opencl_atomic_compare_exchange_strong:
     emitAtomicCmpXchgFailureSet(CGF, E, false, Dest, Ptr, Val1, Val2,
-                                FailureOrder, Size, Order, Scope);
+                                ExpectedResult, FailureOrder, Size, Order, Scope);
     return;
   case AtomicExpr::AO__c11_atomic_compare_exchange_weak:
   case AtomicExpr::AO__opencl_atomic_compare_exchange_weak:
   case AtomicExpr::AO__hip_atomic_compare_exchange_weak:
     emitAtomicCmpXchgFailureSet(CGF, E, true, Dest, Ptr, Val1, Val2,
-                                FailureOrder, Size, Order, Scope);
+                                ExpectedResult, FailureOrder, Size, Order, Scope);
     return;
   case AtomicExpr::AO__atomic_compare_exchange:
   case AtomicExpr::AO__atomic_compare_exchange_n:
@@ -554,7 +598,7 @@ static void EmitAtomicOp(CodeGenFunction &CGF, AtomicExpr *E, Address Dest,
   case AtomicExpr::AO__scoped_atomic_compare_exchange_n: {
     if (llvm::ConstantInt *IsWeakC = dyn_cast<llvm::ConstantInt>(IsWeak)) {
       emitAtomicCmpXchgFailureSet(CGF, E, IsWeakC->getZExtValue(), Dest, Ptr,
-                                  Val1, Val2, FailureOrder, Size, Order, Scope);
+                                  Val1, Val2, ExpectedResult, FailureOrder, Size, Order, Scope);
     } else {
       // Create all the relevant BB's
       llvm::BasicBlock *StrongBB =
@@ -568,12 +612,12 @@ static void EmitAtomicOp(CodeGenFunction &CGF, AtomicExpr *E, Address Dest,
 
       CGF.Builder.SetInsertPoint(StrongBB);
       emitAtomicCmpXchgFailureSet(CGF, E, false, Dest, Ptr, Val1, Val2,
-                                  FailureOrder, Size, Order, Scope);
+                                  ExpectedResult, FailureOrder, Size, Order, Scope);
       CGF.Builder.CreateBr(ContBB);
 
       CGF.Builder.SetInsertPoint(WeakBB);
       emitAtomicCmpXchgFailureSet(CGF, E, true, Dest, Ptr, Val1, Val2,
-                                  FailureOrder, Size, Order, Scope);
+                                  ExpectedResult, FailureOrder, Size, Order, Scope);
       CGF.Builder.CreateBr(ContBB);
 
       CGF.Builder.SetInsertPoint(ContBB);
@@ -777,6 +821,7 @@ EmitValToTemp(CodeGenFunction &CGF, Expr *E) {
 
 static void EmitAtomicOp(CodeGenFunction &CGF, AtomicExpr *Expr, Address Dest,
                          Address Ptr, Address Val1, Address Val2,
+                         Address OriginalVal1,
                          llvm::Value *IsWeak, llvm::Value *FailureOrder,
                          uint64_t Size, llvm::AtomicOrdering Order,
                          llvm::Value *Scope) {
@@ -796,7 +841,7 @@ static void EmitAtomicOp(CodeGenFunction &CGF, AtomicExpr *Expr, Address Dest,
                                                    Order, CGF.getLLVMContext());
     else
       SS = llvm::SyncScope::System;
-    EmitAtomicOp(CGF, Expr, Dest, Ptr, Val1, Val2, IsWeak, FailureOrder, Size,
+    EmitAtomicOp(CGF, Expr, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, FailureOrder, Size,
                  Order, SS);
     return;
   }
@@ -806,7 +851,7 @@ static void EmitAtomicOp(CodeGenFunction &CGF, AtomicExpr *Expr, Address Dest,
     auto SCID = CGF.getTargetHooks().getLLVMSyncScopeID(
         CGF.CGM.getLangOpts(), ScopeModel->map(SC->getZExtValue()),
         Order, CGF.CGM.getLLVMContext());
-    EmitAtomicOp(CGF, Expr, Dest, Ptr, Val1, Val2, IsWeak, FailureOrder, Size,
+    EmitAtomicOp(CGF, Expr, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, FailureOrder, Size,
                  Order, SCID);
     return;
   }
@@ -832,7 +877,7 @@ static void EmitAtomicOp(CodeGenFunction &CGF, AtomicExpr *Expr, Address Dest,
       SI->addCase(Builder.getInt32(S), B);
 
     Builder.SetInsertPoint(B);
-    EmitAtomicOp(CGF, Expr, Dest, Ptr, Val1, Val2, IsWeak, FailureOrder, Size,
+    EmitAtomicOp(CGF, Expr, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, FailureOrder, Size,
                  Order,
                  CGF.getTargetHooks().getLLVMSyncScopeID(CGF.CGM.getLangOpts(),
                                                          ScopeModel->map(S),
@@ -1036,6 +1081,7 @@ RValue CodeGenFunction::EmitAtomicExpr(AtomicExpr *E) {
   LValue AtomicVal = MakeAddrLValue(Ptr, AtomicTy);
   AtomicInfo Atomics(*this, AtomicVal);
 
+  Address OriginalVal1 = Val1;
   if (ShouldCastToIntPtrTy) {
     Ptr = Atomics.castToAtomicIntPointer(Ptr);
     if (Val1.isValid())
@@ -1279,30 +1325,30 @@ RValue CodeGenFunction::EmitAtomicExpr(AtomicExpr *E) {
     if (llvm::isValidAtomicOrderingCABI(ord))
       switch ((llvm::AtomicOrderingCABI)ord) {
       case llvm::AtomicOrderingCABI::relaxed:
-        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                      llvm::AtomicOrdering::Monotonic, Scope);
         break;
       case llvm::AtomicOrderingCABI::consume:
       case llvm::AtomicOrderingCABI::acquire:
         if (IsStore)
           break; // Avoid crashing on code with undefined behavior
-        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                      llvm::AtomicOrdering::Acquire, Scope);
         break;
       case llvm::AtomicOrderingCABI::release:
         if (IsLoad)
           break; // Avoid crashing on code with undefined behavior
-        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                      llvm::AtomicOrdering::Release, Scope);
         break;
       case llvm::AtomicOrderingCABI::acq_rel:
         if (IsLoad || IsStore)
           break; // Avoid crashing on code with undefined behavior
-        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                      llvm::AtomicOrdering::AcquireRelease, Scope);
         break;
       case llvm::AtomicOrderingCABI::seq_cst:
-        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                      llvm::AtomicOrdering::SequentiallyConsistent, Scope);
         break;
       }
@@ -1338,12 +1384,12 @@ RValue CodeGenFunction::EmitAtomicExpr(AtomicExpr *E) {
 
   // Emit all the different atomics
   Builder.SetInsertPoint(MonotonicBB);
-  EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+  EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                llvm::AtomicOrdering::Monotonic, Scope);
   Builder.CreateBr(ContBB);
   if (!IsStore) {
     Builder.SetInsertPoint(AcquireBB);
-    EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+    EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                  llvm::AtomicOrdering::Acquire, Scope);
     Builder.CreateBr(ContBB);
     SI->addCase(Builder.getInt32((int)llvm::AtomicOrderingCABI::consume),
@@ -1353,7 +1399,7 @@ RValue CodeGenFunction::EmitAtomicExpr(AtomicExpr *E) {
   }
   if (!IsLoad) {
     Builder.SetInsertPoint(ReleaseBB);
-    EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+    EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                  llvm::AtomicOrdering::Release, Scope);
     Builder.CreateBr(ContBB);
     SI->addCase(Builder.getInt32((int)llvm::AtomicOrderingCABI::release),
@@ -1361,14 +1407,14 @@ RValue CodeGenFunction::EmitAtomicExpr(AtomicExpr *E) {
   }
   if (!IsLoad && !IsStore) {
     Builder.SetInsertPoint(AcqRelBB);
-    EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+    EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                  llvm::AtomicOrdering::AcquireRelease, Scope);
     Builder.CreateBr(ContBB);
     SI->addCase(Builder.getInt32((int)llvm::AtomicOrderingCABI::acq_rel),
                 AcqRelBB);
   }
   Builder.SetInsertPoint(SeqCstBB);
-  EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+  EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, OriginalVal1, IsWeak, OrderFail, Size,
                llvm::AtomicOrdering::SequentiallyConsistent, Scope);
   Builder.CreateBr(ContBB);
   SI->addCase(Builder.getInt32((int)llvm::AtomicOrderingCABI::seq_cst),
diff --git a/clang/test/CodeGenCXX/builtin-atomic-compare_exchange.cpp b/clang/test/CodeGenCXX/builtin-atomic-compare_exchange.cpp
new file mode 100644
index 0000000000000..4e2830e7a66cd
--- /dev/null
+++ b/clang/test/CodeGenCXX/builtin-atomic-compare_exchange.cpp
@@ -0,0 +1,126 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 5
+// RUN: %clang_cc1 -std=c++20 -triple=x86_64-linux-gnu -emit-llvm -o - %s | FileCheck %s
+
+
+template <unsigned Size>
+struct S {
+    char data[Size];
+};
+
+// CHECK-LABEL: define dso_local noundef zeroext i1 @_Z21test_compare_exchangePU7_Atomic1SILj3EEPS0_S0_(
+// CHECK-SAME: ptr noundef [[A:%.*]], ptr noundef [[EXPECTED:%.*]], i24 [[DESIRED_COERCE:%.*]]) #[[ATTR0:[0-9]+]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[DESIRED:%.*]] = alloca [[STRUCT_S:%.*]], align 1
+// CHECK-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT:    [[EXPECTED_ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT:    [[DOTATOMICTMP:%.*]] = alloca [[STRUCT_S]], align 1
+// CHECK-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca { [[STRUCT_S]], [1 x i8] }, align 4
+// CHECK-NEXT:    [[ATOMIC_TEMP1:%.*]] = alloca { [[STRUCT_S]], [1 x i8] }, align 4
+// CHECK-NEXT:    [[CMPXCHG_BOOL:%.*]] = alloca i8, align 1
+// CHECK-NEXT:    [[OLD_TMP:%.*]] = alloca i32, align 4
+// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds nuw [[STRUCT_S]], ptr [[DESIRED]], i32 0, i32 0
+// CHECK-NEXT:    store i24 [[DESIRED_COERCE]], ptr [[COERCE_DIVE]], align 1
+// CHECK-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// CHECK-NEXT:    store ptr [[EXPECTED]], ptr [[EXPECTED_ADDR]], align 8
+// CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// CHECK-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[EXPECTED_ADDR]], align 8
+// CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 1 [[DOTATOMICTMP]], ptr align 1 [[DESIRED]], i64 3, i1 false)
+// CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ATOMIC_TEMP]], ptr align 1 [[TMP1]], i64 3, i1 false)
+// CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ATOMIC_TEMP1]], ptr align 1 [[DOTATOMICTMP]], i64 3, i1 false)
+// CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[ATOMIC_TEMP]], align 4
+// CHECK-NEXT:    [[TMP3:%.*]] = load i32, ptr [[ATOMIC_TEMP1]], align 4
+// CHECK-NEXT:    [[TMP4:%.*]] = cmpxchg ptr [[TMP0]], i32 [[TMP2]], i32 [[TMP3]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP5:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
+// CHECK-NEXT:    br i1 [[TMP6]], label %[[CMPXCHG_CONTINUE:.*]], label %[[CMPXCHG_STORE_EXPECTED:.*]]
+// CHECK:       [[CMPXCHG_STORE_EXPECTED]]:
+// CHECK-NEXT:    store i32 [[TMP5]], ptr [[OLD_TMP]], align 4
+// CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 1 [[TMP1]], ptr align 4 [[OLD_TMP]], i64 3, i1 false)
+// CHECK-NEXT:    br label %[[CMPXCHG_CONTINUE]]
+// CHECK:       [[CMPXCHG_CONTINUE]]:
+// CHECK-NEXT:    [[STOREDV:%.*]] = zext i1 [[TMP6]] to i8
+// CHECK-NEXT:    store i8 [[STOREDV]], ptr [[CMPXCHG_BOOL]], align 1
+// CHECK-NEXT:    [[TMP7:%.*]] = load i8, ptr [[CMPXCHG_BOOL]], align 1
+// CHECK-NEXT:    [[LOADEDV:%.*]] = trunc i8 [[TMP7]] to i1
+// CHECK-NEXT:    ret i1 [[LOADEDV]]
+//
+bool test_compare_exchange(_Atomic(S<3>)* a, S<3>* expected, S<3> desired) {
+    return __c11_atomic_compare_exchange_strong(a, expected, desired, 0, 0);
+}
+
+
+// CHECK-LABEL: define dso_local noundef zeroext i1 @_Z21test_compare_exchangePU7_Atomic1SILj4EEPS0_S0_(
+// CHECK-SAME: ptr noundef [[A:%.*]], ptr noundef [[EXPECTED:%.*]], i32 [[DESIRED_COERCE:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[DESIRED:%.*]] = alloca [[STRUCT_S_0:%.*]], align 1
+// CHECK-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT:    [[EXPECTED_ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT:    [[DOTATOMICTMP:%.*]] = alloca [[STRUCT_S_0]], align 1
+// CHECK-NEXT:    [[CMPXCHG_BOOL:%.*]] = alloca i8, align 1
+// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds nuw [[STRUCT_S_0]], ptr [[DESIRED]], i32 0, i32 0
+// CHECK-NEXT:    store i32 [[DESIRED_COERCE]], ptr [[COERCE_DIVE]], align 1
+// CHECK-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// CHECK-NEXT:    store ptr [[EXPECTED]], ptr [[EXPECTED_ADDR]], align 8
+// CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// CHECK-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[EXPECTED_ADDR]], align 8
+// CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 1 [[DOTATOMICTMP]], ptr align 1 [[DESIRED]], i64 4, i1 false)
+// CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 1
+// CHECK-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTATOMICTMP]], align 1
+// CHECK-NEXT:    [[TMP4:%.*]] = cmpxchg ptr [[TMP0]], i32 [[TMP2]], i32 [[TMP3]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP5:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
+// CHECK-NEXT:    br i1 [[TMP6]], label %[[CMPXCHG_CONTINUE:.*]], label %[[CMPXCHG_STORE_EXPECTED:.*]]
+// CHECK:       [[CMPXCHG_STORE_EXPECTED]]:
+// CHECK-NEXT:    store i32 [[TMP5]], ptr [[TMP1]], align 1
+// CHECK-NEXT:    br label %[[CMPXCHG_CONTINUE]]
+// CHECK:       [[CMPXCHG_CONTINUE]]:
+// CHECK-NEXT:    [[STOREDV:%.*]] = zext i1 [[TMP6]] to i8
+// CHECK-NEXT:    store i8 [[STOREDV]], ptr [[CMPXCHG_BOOL]], align 1
+// CHECK-NEXT:    [[TMP7:%.*]] = load i8, ptr [[CMPXCHG_BOOL]], align 1
+// CHECK-NEXT:    [[LOADEDV:%.*]] = trunc i8 [[TMP7]] to i1
+// CHECK-NEXT:    ret i1 [[LOADEDV]]
+//
+bool test_compare_exchange(_Atomic(S<4>)* a, S<4>* expected, S<4> desired) {
+    return __c11_atomic_compare_exchange_strong(a, expected, desired, 0, 0);
+}
+
+// CHECK-LABEL: define dso_local noundef zeroext i1 @_Z21test_compare_exchangePU7_Atomic1SILj6EEPS0_S0_(
+// CHECK-SAME: ptr noundef [[A:%.*]], ptr noundef [[EXPECTED:%.*]], i48 [[DESIRED_COERCE:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[DESIRED:%.*]] = alloca [[STRUCT_S_1:%.*]], align 1
+// CHECK-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT:    [[EXPECTED_ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT:    [[DOTATOMICTMP:%.*]] = alloca [[STRUCT_S_1]], align 1
+// CHECK-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca { [[STRUCT_S_1]], [2 x i8] }, align 8
+// CHECK-NEXT:    [[ATOMIC_TEMP1:%.*]] = alloca { [[STRUCT_S_1]], [2 x i8...
[truncated]

efriedma-quic
efriedma-quic reviewed Jun 9, 2025
View reviewed changes
ldionne
ldionne reviewed Jul 4, 2025
View reviewed changes
Copy link
Member

@ldionne ldionne left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This looks quite reasonable to me, and it seems we're fixing a pretty bad bug. However, someone else with more Clang/Codegen experience needs to look at this. Thanks for working on this!

CC @efriedma-quic

@ldionne
Copy link
Member

ldionne commented Jul 4, 2025

CC @philnik777 in case you want to take a look as well

@huixie90
Copy link
Member Author

ping @efriedma-quic does the current status address your concerns?

@efriedma-quic
Copy link
Collaborator

I think the current state addresses my concerns, yes. I'd like to make one final pass over it, though; please git-clang-format and rebase the patch.

@huixie90
Copy link
Member Author

huixie90 commented Oct 5, 2025

I think the current state addresses my concerns, yes. I'd like to make one final pass over it, though; please git-clang-format and rebase the patch.

Thanks I rebased and clang-formatted the changed lines. The CI fails on the new test I added in libc++, because the libc++ is not using the just-built clang for testing for majority of the targets and they fail as expected. i will check with @ldionne how to skip the previous versions of clangs for that test

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@ldionne ldionne ldionne left review comments

@efriedma-quic efriedma-quic efriedma-quic left review comments

Assignees
No one assigned
Labels
clang:codegen IR generation bugs: mangling, exceptions, etc. clang Clang issues not falling into any other category libc++ libc++ C++ Standard Library. Not GNU libstdc++. Not libc++abi.
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

atomic::compare_exchange returns wrong value
4 participants
@huixie90 @llvmbot @ldionne @efriedma-quic