Skip to content

Commit

Permalink
[GCAtomic] Treat single capability structs like a capability
Browse files Browse the repository at this point in the history 
This allows emitting inline atomics instead of needing a libcall and
avoids casting such structs to i128.

See https://git.morello-project.org/morello/llvm-project/-/issues/75
  • Loading branch information
@arichardson
arichardson committed Jan 9, 2025
1 parent 13f4598 commit bd82445
Show file tree
Hide file tree
Showing 2 changed files with 90 additions and 35 deletions.
23 changes: 15 additions & 8 deletions clang/lib/CodeGen/CGAtomic.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -148,7 +148,8 @@ bool isAtomicStoreOp(AtomicExpr::AtomicOp Op) {
}
UseLibcall = !C.getTargetInfo().hasBuiltinAtomic(
AtomicSizeInBits, C.toBits(lvalue.getAlignment()),
AtomicTy->isCHERICapabilityType(CGF.CGM.getContext()));
AtomicTy->isCHERICapabilityType(CGF.getContext()) ||
AtomicTy->isSingleCapabilityRecord(CGF.getContext()));
}

QualType getAtomicType() const { return AtomicTy; }
Expand Down Expand Up @@ -549,7 +550,8 @@ static void EmitAtomicOp(CodeGenFunction &CGF, AtomicExpr *E, Address Dest,
bool PostOpMinMax = false;
unsigned PostOp = 0;
QualType AtomicTy = E->getPtr()->getType()->getPointeeType();
bool IsCheriCap = AtomicTy->isCHERICapabilityType(CGF.CGM.getContext());
bool IsCheriCap = AtomicTy->isCHERICapabilityType(CGF.getContext()) ||
AtomicTy->isSingleCapabilityRecord(CGF.getContext());

switch (E->getOp()) {
case AtomicExpr::AO__c11_atomic_init:
Expand Down Expand Up @@ -820,12 +822,14 @@ static void
AddDirectArgument(CodeGenFunction &CGF, CallArgList &Args,
bool UseOptimizedLibcall, llvm::Value *Val, QualType ValTy,
SourceLocation Loc, CharUnits SizeInChars) {
bool IsCapTy = ValTy->isCHERICapabilityType(CGF.getContext()) ||
ValTy->isSingleCapabilityRecord(CGF.getContext());
if (UseOptimizedLibcall) {
// Load value and pass it to the function directly.
CharUnits Align = CGF.getContext().getTypeAlignInChars(ValTy);
int64_t SizeInBits = CGF.getContext().toBits(SizeInChars);
llvm::Type *ITy;
if (ValTy->isCHERICapabilityType(CGF.getContext())) {
if (IsCapTy) {
ValTy = CGF.getContext().getPointerType(CGF.getContext().VoidTy,
PIK_Capability);
ITy = CGF.Int8CheriCapTy;
Expand All @@ -845,7 +849,7 @@ AddDirectArgument(CodeGenFunction &CGF, CallArgList &Args,
} else {
// Non-optimized functions always take a reference.
// NB: Capabilities must be passed directly to the optimized libcall
assert(!ValTy->isCHERICapabilityType(CGF.getContext()) &&
assert(!IsCapTy &&
"Capabilities should not be passed to the generic libcall");
Args.add(RValue::get(Val), CGF.getContext().VoidPtrTy);
}
Expand Down Expand Up @@ -874,7 +878,8 @@ RValue CodeGenFunction::EmitAtomicExpr(AtomicExpr *E) {
uint64_t Size = TInfo.Width.getQuantity();
unsigned MaxInlineWidthInBits = getTarget().getMaxAtomicInlineWidth();

bool IsCheriCap = AtomicTy->isCHERICapabilityType(CGM.getContext());
bool IsCheriCap = AtomicTy->isCHERICapabilityType(CGM.getContext()) ||
AtomicTy->isSingleCapabilityRecord(CGM.getContext());
bool Oversized = (!IsCheriCap &&
getContext().toBits(TInfo.Width) > MaxInlineWidthInBits) ||
(IsCheriCap && MaxInlineWidthInBits == 0);
Expand Down Expand Up @@ -1523,14 +1528,16 @@ RValue CodeGenFunction::EmitAtomicExpr(AtomicExpr *E) {

Address AtomicInfo::castToAtomicIntPointer(Address addr) const {
llvm::Type *ty;
if (AtomicTy->isCHERICapabilityType(CGF.getContext())) {
if (AtomicTy->isCHERICapabilityType(CGF.getContext()) ||
AtomicTy->isSingleCapabilityRecord(CGF.getContext())) {
// If capability atomics are natively supported the instruction expects
// a capability type. We also pass capabilities directly to the atomic
// libcalls (i.e. always use optimized ones) since this is required to
// support the RMW operations and special-casing the load/store/xchg to
// use the generic libcalls (with mutex+memcpy) adds unncessary complexity.
if (!UseLibcall) {
// If we aren't using a libcall there is no need to cast to i8*
if (!UseLibcall && !AtomicTy->isSingleCapabilityRecord(CGF.getContext())) {
// If we aren't using a libcall and aren't using a single-capability
// struct a there is no need to cast to i8*
return addr.withElementType(getAtomicAddress().getElementType());
}
ty = CGF.CGM.Int8CheriCapTy;
Expand Down
102 changes: 75 additions & 27 deletions clang/test/CodeGen/cheri/c11-atomic-caps-struct.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -4,6 +4,7 @@
// RUN: | opt -S --passes=mem2reg | FileCheck --check-prefix=HYBRID %s
// RUN: %riscv64_cheri_purecap_cc1 -target-feature +a -std=c11 -o - -emit-llvm -disable-O0-optnone %s -verify \
// RUN: | opt -S --passes=mem2reg | FileCheck --check-prefix=PURECAP %s
// expected-no-diagnostics

typedef struct capstruct {
unsigned __intcap value;
Expand Down Expand Up @@ -36,24 +37,25 @@ void test_init(_Atomic(capstruct) *f, capstruct value) {
// HYBRID-NEXT: entry:
// HYBRID-NEXT: [[RETVAL:%.*]] = alloca [[STRUCT_CAPSTRUCT:%.*]], align 16
// HYBRID-NEXT: [[ATOMIC_TEMP:%.*]] = alloca [[STRUCT_CAPSTRUCT]], align 16
// HYBRID-NEXT: call void @__atomic_load(i64 noundef 16, ptr noundef [[F]], ptr noundef [[ATOMIC_TEMP]], i32 noundef signext 5)
// HYBRID-NEXT: [[TMP0:%.*]] = load atomic ptr addrspace(200), ptr [[F]] seq_cst, align 16
// HYBRID-NEXT: store ptr addrspace(200) [[TMP0]], ptr [[ATOMIC_TEMP]], align 16
// HYBRID-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 16 [[RETVAL]], ptr align 16 [[ATOMIC_TEMP]], i64 16, i1 false)
// HYBRID-NEXT: [[TMP0:%.*]] = load [[STRUCT_CAPSTRUCT]], ptr [[RETVAL]], align 16
// HYBRID-NEXT: ret [[STRUCT_CAPSTRUCT]] [[TMP0]]
// HYBRID-NEXT: [[TMP1:%.*]] = load [[STRUCT_CAPSTRUCT]], ptr [[RETVAL]], align 16
// HYBRID-NEXT: ret [[STRUCT_CAPSTRUCT]] [[TMP1]]
//
// PURECAP-LABEL: define {{[^@]+}}@test_load
// PURECAP-SAME: (ptr addrspace(200) noundef [[F:%.*]]) addrspace(200) #[[ATTR0]] {
// PURECAP-NEXT: entry:
// PURECAP-NEXT: [[RETVAL:%.*]] = alloca [[STRUCT_CAPSTRUCT:%.*]], align 16, addrspace(200)
// PURECAP-NEXT: [[ATOMIC_TEMP:%.*]] = alloca [[STRUCT_CAPSTRUCT]], align 16, addrspace(200)
// PURECAP-NEXT: call void @__atomic_load(i64 noundef 16, ptr addrspace(200) noundef [[F]], ptr addrspace(200) noundef [[ATOMIC_TEMP]], i32 noundef signext 5)
// PURECAP-NEXT: [[TMP0:%.*]] = load atomic ptr addrspace(200), ptr addrspace(200) [[F]] seq_cst, align 16
// PURECAP-NEXT: store ptr addrspace(200) [[TMP0]], ptr addrspace(200) [[ATOMIC_TEMP]], align 16
// PURECAP-NEXT: call void @llvm.memcpy.p200.p200.i64(ptr addrspace(200) align 16 [[RETVAL]], ptr addrspace(200) align 16 [[ATOMIC_TEMP]], i64 16, i1 false)
// PURECAP-NEXT: [[TMP0:%.*]] = load [[STRUCT_CAPSTRUCT]], ptr addrspace(200) [[RETVAL]], align 16
// PURECAP-NEXT: ret [[STRUCT_CAPSTRUCT]] [[TMP0]]
// PURECAP-NEXT: [[TMP1:%.*]] = load [[STRUCT_CAPSTRUCT]], ptr addrspace(200) [[RETVAL]], align 16
// PURECAP-NEXT: ret [[STRUCT_CAPSTRUCT]] [[TMP1]]
//
capstruct test_load(_Atomic(capstruct) *f) {
return __c11_atomic_load(f, __ATOMIC_SEQ_CST);
// expected-warning@-1{{large atomic operation may incur significant performance penalty}}
}

// HYBRID-LABEL: define {{[^@]+}}@test_store
Expand All @@ -64,7 +66,8 @@ capstruct test_load(_Atomic(capstruct) *f) {
// HYBRID-NEXT: [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_CAPSTRUCT]], ptr [[VALUE]], i32 0, i32 0
// HYBRID-NEXT: store ptr addrspace(200) [[VALUE_COERCE]], ptr [[COERCE_DIVE]], align 16
// HYBRID-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 16 [[DOTATOMICTMP]], ptr align 16 [[VALUE]], i64 16, i1 false)
// HYBRID-NEXT: call void @__atomic_store(i64 noundef 16, ptr noundef [[F]], ptr noundef [[DOTATOMICTMP]], i32 noundef signext 5)
// HYBRID-NEXT: [[TMP0:%.*]] = load ptr addrspace(200), ptr [[DOTATOMICTMP]], align 16
// HYBRID-NEXT: store atomic ptr addrspace(200) [[TMP0]], ptr [[F]] seq_cst, align 16
// HYBRID-NEXT: ret void
//
// PURECAP-LABEL: define {{[^@]+}}@test_store
Expand All @@ -75,12 +78,12 @@ capstruct test_load(_Atomic(capstruct) *f) {
// PURECAP-NEXT: [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_CAPSTRUCT]], ptr addrspace(200) [[VALUE]], i32 0, i32 0
// PURECAP-NEXT: store ptr addrspace(200) [[VALUE_COERCE]], ptr addrspace(200) [[COERCE_DIVE]], align 16
// PURECAP-NEXT: call void @llvm.memcpy.p200.p200.i64(ptr addrspace(200) align 16 [[DOTATOMICTMP]], ptr addrspace(200) align 16 [[VALUE]], i64 16, i1 false)
// PURECAP-NEXT: call void @__atomic_store(i64 noundef 16, ptr addrspace(200) noundef [[F]], ptr addrspace(200) noundef [[DOTATOMICTMP]], i32 noundef signext 5)
// PURECAP-NEXT: [[TMP0:%.*]] = load ptr addrspace(200), ptr addrspace(200) [[DOTATOMICTMP]], align 16
// PURECAP-NEXT: store atomic ptr addrspace(200) [[TMP0]], ptr addrspace(200) [[F]] seq_cst, align 16
// PURECAP-NEXT: ret void
//
void test_store(_Atomic(capstruct) *f, capstruct value) {
__c11_atomic_store(f, value, __ATOMIC_SEQ_CST);
// expected-warning@-1{{large atomic operation may incur significant performance penalty}}
}

// HYBRID-LABEL: define {{[^@]+}}@test_xchg
Expand All @@ -93,10 +96,12 @@ void test_store(_Atomic(capstruct) *f, capstruct value) {
// HYBRID-NEXT: [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_CAPSTRUCT]], ptr [[VALUE]], i32 0, i32 0
// HYBRID-NEXT: store ptr addrspace(200) [[VALUE_COERCE]], ptr [[COERCE_DIVE]], align 16
// HYBRID-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 16 [[DOTATOMICTMP]], ptr align 16 [[VALUE]], i64 16, i1 false)
// HYBRID-NEXT: call void @__atomic_exchange(i64 noundef 16, ptr noundef [[F]], ptr noundef [[DOTATOMICTMP]], ptr noundef [[ATOMIC_TEMP]], i32 noundef signext 5)
// HYBRID-NEXT: [[TMP0:%.*]] = load ptr addrspace(200), ptr [[DOTATOMICTMP]], align 16
// HYBRID-NEXT: [[TMP1:%.*]] = atomicrmw xchg ptr [[F]], ptr addrspace(200) [[TMP0]] seq_cst, align 16
// HYBRID-NEXT: store ptr addrspace(200) [[TMP1]], ptr [[ATOMIC_TEMP]], align 16
// HYBRID-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 16 [[RETVAL]], ptr align 16 [[ATOMIC_TEMP]], i64 16, i1 false)
// HYBRID-NEXT: [[TMP0:%.*]] = load [[STRUCT_CAPSTRUCT]], ptr [[RETVAL]], align 16
// HYBRID-NEXT: ret [[STRUCT_CAPSTRUCT]] [[TMP0]]
// HYBRID-NEXT: [[TMP2:%.*]] = load [[STRUCT_CAPSTRUCT]], ptr [[RETVAL]], align 16
// HYBRID-NEXT: ret [[STRUCT_CAPSTRUCT]] [[TMP2]]
//
// PURECAP-LABEL: define {{[^@]+}}@test_xchg
// PURECAP-SAME: (ptr addrspace(200) noundef [[F:%.*]], ptr addrspace(200) [[VALUE_COERCE:%.*]]) addrspace(200) #[[ATTR0]] {
Expand All @@ -108,14 +113,15 @@ void test_store(_Atomic(capstruct) *f, capstruct value) {
// PURECAP-NEXT: [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_CAPSTRUCT]], ptr addrspace(200) [[VALUE]], i32 0, i32 0
// PURECAP-NEXT: store ptr addrspace(200) [[VALUE_COERCE]], ptr addrspace(200) [[COERCE_DIVE]], align 16
// PURECAP-NEXT: call void @llvm.memcpy.p200.p200.i64(ptr addrspace(200) align 16 [[DOTATOMICTMP]], ptr addrspace(200) align 16 [[VALUE]], i64 16, i1 false)
// PURECAP-NEXT: call void @__atomic_exchange(i64 noundef 16, ptr addrspace(200) noundef [[F]], ptr addrspace(200) noundef [[DOTATOMICTMP]], ptr addrspace(200) noundef [[ATOMIC_TEMP]], i32 noundef signext 5)
// PURECAP-NEXT: [[TMP0:%.*]] = load ptr addrspace(200), ptr addrspace(200) [[DOTATOMICTMP]], align 16
// PURECAP-NEXT: [[TMP1:%.*]] = atomicrmw xchg ptr addrspace(200) [[F]], ptr addrspace(200) [[TMP0]] seq_cst, align 16
// PURECAP-NEXT: store ptr addrspace(200) [[TMP1]], ptr addrspace(200) [[ATOMIC_TEMP]], align 16
// PURECAP-NEXT: call void @llvm.memcpy.p200.p200.i64(ptr addrspace(200) align 16 [[RETVAL]], ptr addrspace(200) align 16 [[ATOMIC_TEMP]], i64 16, i1 false)
// PURECAP-NEXT: [[TMP0:%.*]] = load [[STRUCT_CAPSTRUCT]], ptr addrspace(200) [[RETVAL]], align 16
// PURECAP-NEXT: ret [[STRUCT_CAPSTRUCT]] [[TMP0]]
// PURECAP-NEXT: [[TMP2:%.*]] = load [[STRUCT_CAPSTRUCT]], ptr addrspace(200) [[RETVAL]], align 16
// PURECAP-NEXT: ret [[STRUCT_CAPSTRUCT]] [[TMP2]]
//
capstruct test_xchg(_Atomic(capstruct) *f, capstruct value) {
return __c11_atomic_exchange(f, value, __ATOMIC_SEQ_CST);
// expected-warning@-1{{large atomic operation may incur significant performance penalty}}
}

// HYBRID-LABEL: define {{[^@]+}}@test_cmpxchg_weak
Expand All @@ -126,8 +132,19 @@ capstruct test_xchg(_Atomic(capstruct) *f, capstruct value) {
// HYBRID-NEXT: [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_CAPSTRUCT]], ptr [[NEW]], i32 0, i32 0
// HYBRID-NEXT: store ptr addrspace(200) [[NEW_COERCE]], ptr [[COERCE_DIVE]], align 16
// HYBRID-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 16 [[DOTATOMICTMP]], ptr align 16 [[NEW]], i64 16, i1 false)
// HYBRID-NEXT: [[CALL:%.*]] = call zeroext i1 @__atomic_compare_exchange(i64 noundef 16, ptr noundef [[F]], ptr noundef [[EXP]], ptr noundef [[DOTATOMICTMP]], i32 noundef signext 0, i32 noundef signext 0)
// HYBRID-NEXT: ret i1 [[CALL]]
// HYBRID-NEXT: [[TMP0:%.*]] = load ptr addrspace(200), ptr [[EXP]], align 16
// HYBRID-NEXT: [[TMP1:%.*]] = load ptr addrspace(200), ptr [[DOTATOMICTMP]], align 16
// HYBRID-NEXT: [[TMP2:%.*]] = cmpxchg weak ptr [[F]], ptr addrspace(200) [[TMP0]], ptr addrspace(200) [[TMP1]] monotonic monotonic, align 16
// HYBRID-NEXT: [[TMP3:%.*]] = extractvalue { ptr addrspace(200), i1 } [[TMP2]], 0
// HYBRID-NEXT: [[TMP4:%.*]] = extractvalue { ptr addrspace(200), i1 } [[TMP2]], 1
// HYBRID-NEXT: br i1 [[TMP4]], label [[CMPXCHG_CONTINUE:%.*]], label [[CMPXCHG_STORE_EXPECTED:%.*]]
// HYBRID: cmpxchg.store_expected:
// HYBRID-NEXT: store ptr addrspace(200) [[TMP3]], ptr [[EXP]], align 16
// HYBRID-NEXT: br label [[CMPXCHG_CONTINUE]]
// HYBRID: cmpxchg.continue:
// HYBRID-NEXT: [[FROMBOOL:%.*]] = zext i1 [[TMP4]] to i8
// HYBRID-NEXT: [[TOBOOL:%.*]] = trunc i8 [[FROMBOOL]] to i1
// HYBRID-NEXT: ret i1 [[TOBOOL]]
//
// PURECAP-LABEL: define {{[^@]+}}@test_cmpxchg_weak
// PURECAP-SAME: (ptr addrspace(200) noundef [[F:%.*]], ptr addrspace(200) noundef [[EXP:%.*]], ptr addrspace(200) [[NEW_COERCE:%.*]]) addrspace(200) #[[ATTR0]] {
Expand All @@ -137,12 +154,22 @@ capstruct test_xchg(_Atomic(capstruct) *f, capstruct value) {
// PURECAP-NEXT: [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_CAPSTRUCT]], ptr addrspace(200) [[NEW]], i32 0, i32 0
// PURECAP-NEXT: store ptr addrspace(200) [[NEW_COERCE]], ptr addrspace(200) [[COERCE_DIVE]], align 16
// PURECAP-NEXT: call void @llvm.memcpy.p200.p200.i64(ptr addrspace(200) align 16 [[DOTATOMICTMP]], ptr addrspace(200) align 16 [[NEW]], i64 16, i1 false)
// PURECAP-NEXT: [[CALL:%.*]] = call zeroext i1 @__atomic_compare_exchange(i64 noundef 16, ptr addrspace(200) noundef [[F]], ptr addrspace(200) noundef [[EXP]], ptr addrspace(200) noundef [[DOTATOMICTMP]], i32 noundef signext 0, i32 noundef signext 0)
// PURECAP-NEXT: ret i1 [[CALL]]
// PURECAP-NEXT: [[TMP0:%.*]] = load ptr addrspace(200), ptr addrspace(200) [[EXP]], align 16
// PURECAP-NEXT: [[TMP1:%.*]] = load ptr addrspace(200), ptr addrspace(200) [[DOTATOMICTMP]], align 16
// PURECAP-NEXT: [[TMP2:%.*]] = cmpxchg weak ptr addrspace(200) [[F]], ptr addrspace(200) [[TMP0]], ptr addrspace(200) [[TMP1]] monotonic monotonic, align 16
// PURECAP-NEXT: [[TMP3:%.*]] = extractvalue { ptr addrspace(200), i1 } [[TMP2]], 0
// PURECAP-NEXT: [[TMP4:%.*]] = extractvalue { ptr addrspace(200), i1 } [[TMP2]], 1
// PURECAP-NEXT: br i1 [[TMP4]], label [[CMPXCHG_CONTINUE:%.*]], label [[CMPXCHG_STORE_EXPECTED:%.*]]
// PURECAP: cmpxchg.store_expected:
// PURECAP-NEXT: store ptr addrspace(200) [[TMP3]], ptr addrspace(200) [[EXP]], align 16
// PURECAP-NEXT: br label [[CMPXCHG_CONTINUE]]
// PURECAP: cmpxchg.continue:
// PURECAP-NEXT: [[FROMBOOL:%.*]] = zext i1 [[TMP4]] to i8
// PURECAP-NEXT: [[TOBOOL:%.*]] = trunc i8 [[FROMBOOL]] to i1
// PURECAP-NEXT: ret i1 [[TOBOOL]]
//
_Bool test_cmpxchg_weak(_Atomic(capstruct) *f, capstruct *exp, capstruct new) {
return __c11_atomic_compare_exchange_weak(f, exp, new, __ATOMIC_RELAXED, __ATOMIC_RELAXED);
// expected-warning@-1{{large atomic operation may incur significant performance penalty}}
}

// HYBRID-LABEL: define {{[^@]+}}@test_cmpxchg_strong
Expand All @@ -153,8 +180,19 @@ _Bool test_cmpxchg_weak(_Atomic(capstruct) *f, capstruct *exp, capstruct new) {
// HYBRID-NEXT: [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_CAPSTRUCT]], ptr [[NEW]], i32 0, i32 0
// HYBRID-NEXT: store ptr addrspace(200) [[NEW_COERCE]], ptr [[COERCE_DIVE]], align 16
// HYBRID-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 16 [[DOTATOMICTMP]], ptr align 16 [[NEW]], i64 16, i1 false)
// HYBRID-NEXT: [[CALL:%.*]] = call zeroext i1 @__atomic_compare_exchange(i64 noundef 16, ptr noundef [[F]], ptr noundef [[EXP]], ptr noundef [[DOTATOMICTMP]], i32 noundef signext 0, i32 noundef signext 0)
// HYBRID-NEXT: ret i1 [[CALL]]
// HYBRID-NEXT: [[TMP0:%.*]] = load ptr addrspace(200), ptr [[EXP]], align 16
// HYBRID-NEXT: [[TMP1:%.*]] = load ptr addrspace(200), ptr [[DOTATOMICTMP]], align 16
// HYBRID-NEXT: [[TMP2:%.*]] = cmpxchg ptr [[F]], ptr addrspace(200) [[TMP0]], ptr addrspace(200) [[TMP1]] monotonic monotonic, align 16
// HYBRID-NEXT: [[TMP3:%.*]] = extractvalue { ptr addrspace(200), i1 } [[TMP2]], 0
// HYBRID-NEXT: [[TMP4:%.*]] = extractvalue { ptr addrspace(200), i1 } [[TMP2]], 1
// HYBRID-NEXT: br i1 [[TMP4]], label [[CMPXCHG_CONTINUE:%.*]], label [[CMPXCHG_STORE_EXPECTED:%.*]]
// HYBRID: cmpxchg.store_expected:
// HYBRID-NEXT: store ptr addrspace(200) [[TMP3]], ptr [[EXP]], align 16
// HYBRID-NEXT: br label [[CMPXCHG_CONTINUE]]
// HYBRID: cmpxchg.continue:
// HYBRID-NEXT: [[FROMBOOL:%.*]] = zext i1 [[TMP4]] to i8
// HYBRID-NEXT: [[TOBOOL:%.*]] = trunc i8 [[FROMBOOL]] to i1
// HYBRID-NEXT: ret i1 [[TOBOOL]]
//
// PURECAP-LABEL: define {{[^@]+}}@test_cmpxchg_strong
// PURECAP-SAME: (ptr addrspace(200) noundef [[F:%.*]], ptr addrspace(200) noundef [[EXP:%.*]], ptr addrspace(200) [[NEW_COERCE:%.*]]) addrspace(200) #[[ATTR0]] {
Expand All @@ -164,10 +202,20 @@ _Bool test_cmpxchg_weak(_Atomic(capstruct) *f, capstruct *exp, capstruct new) {
// PURECAP-NEXT: [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_CAPSTRUCT]], ptr addrspace(200) [[NEW]], i32 0, i32 0
// PURECAP-NEXT: store ptr addrspace(200) [[NEW_COERCE]], ptr addrspace(200) [[COERCE_DIVE]], align 16
// PURECAP-NEXT: call void @llvm.memcpy.p200.p200.i64(ptr addrspace(200) align 16 [[DOTATOMICTMP]], ptr addrspace(200) align 16 [[NEW]], i64 16, i1 false)
// PURECAP-NEXT: [[CALL:%.*]] = call zeroext i1 @__atomic_compare_exchange(i64 noundef 16, ptr addrspace(200) noundef [[F]], ptr addrspace(200) noundef [[EXP]], ptr addrspace(200) noundef [[DOTATOMICTMP]], i32 noundef signext 0, i32 noundef signext 0)
// PURECAP-NEXT: ret i1 [[CALL]]
// PURECAP-NEXT: [[TMP0:%.*]] = load ptr addrspace(200), ptr addrspace(200) [[EXP]], align 16
// PURECAP-NEXT: [[TMP1:%.*]] = load ptr addrspace(200), ptr addrspace(200) [[DOTATOMICTMP]], align 16
// PURECAP-NEXT: [[TMP2:%.*]] = cmpxchg ptr addrspace(200) [[F]], ptr addrspace(200) [[TMP0]], ptr addrspace(200) [[TMP1]] monotonic monotonic, align 16
// PURECAP-NEXT: [[TMP3:%.*]] = extractvalue { ptr addrspace(200), i1 } [[TMP2]], 0
// PURECAP-NEXT: [[TMP4:%.*]] = extractvalue { ptr addrspace(200), i1 } [[TMP2]], 1
// PURECAP-NEXT: br i1 [[TMP4]], label [[CMPXCHG_CONTINUE:%.*]], label [[CMPXCHG_STORE_EXPECTED:%.*]]
// PURECAP: cmpxchg.store_expected:
// PURECAP-NEXT: store ptr addrspace(200) [[TMP3]], ptr addrspace(200) [[EXP]], align 16
// PURECAP-NEXT: br label [[CMPXCHG_CONTINUE]]
// PURECAP: cmpxchg.continue:
// PURECAP-NEXT: [[FROMBOOL:%.*]] = zext i1 [[TMP4]] to i8
// PURECAP-NEXT: [[TOBOOL:%.*]] = trunc i8 [[FROMBOOL]] to i1
// PURECAP-NEXT: ret i1 [[TOBOOL]]
//
_Bool test_cmpxchg_strong(_Atomic(capstruct) *f, capstruct *exp, capstruct new) {
return __c11_atomic_compare_exchange_strong(f, exp, new, __ATOMIC_RELAXED, __ATOMIC_RELAXED);
// expected-warning@-1{{large atomic operation may incur significant performance penalty}}
}

0 comments on commit bd82445

Please sign in to comment.