Introduce __builtin_nontemporal_store and __builtin_nontemporal_load.
Summary:
Currently clang provides no general way to generate nontemporal loads/stores.
There are some architecture specific builtins for doing so (e.g. in x86), but
there is no way to generate non-temporal store on, e.g. AArch64. This patch adds
generic builtins which are expanded to a simple store with '!nontemporal'
attribute in IR.
Differential Revision: http://reviews.llvm.org/D12313
llvm-svn: 247104
diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index c35f25a..729c0a1 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -111,6 +111,28 @@
return EmitFromInt(CGF, Result, T, ValueType);
}
+static Value *EmitNontemporalStore(CodeGenFunction &CGF, const CallExpr *E) {
+ Value *Val = CGF.EmitScalarExpr(E->getArg(0));
+ Value *Address = CGF.EmitScalarExpr(E->getArg(1));
+
+ // Convert the type of the pointer to a pointer to the stored type.
+ Val = CGF.EmitToMemory(Val, E->getArg(0)->getType());
+ Value *BC = CGF.Builder.CreateBitCast(
+ Address, llvm::PointerType::getUnqual(Val->getType()), "cast");
+ LValue LV = CGF.MakeNaturalAlignAddrLValue(BC, E->getArg(0)->getType());
+ LV.setNontemporal(true);
+ CGF.EmitStoreOfScalar(Val, LV, false);
+ return nullptr;
+}
+
+static Value *EmitNontemporalLoad(CodeGenFunction &CGF, const CallExpr *E) {
+ Value *Address = CGF.EmitScalarExpr(E->getArg(0));
+
+ LValue LV = CGF.MakeNaturalAlignAddrLValue(Address, E->getType());
+ LV.setNontemporal(true);
+ return CGF.EmitLoadOfScalar(LV, E->getExprLoc());
+}
+
static RValue EmitBinaryAtomic(CodeGenFunction &CGF,
llvm::AtomicRMWInst::BinOp Kind,
const CallExpr *E) {
@@ -1143,6 +1165,10 @@
return RValue::get(nullptr);
}
+ case Builtin::BI__builtin_nontemporal_load:
+ return RValue::get(EmitNontemporalLoad(*this, E));
+ case Builtin::BI__builtin_nontemporal_store:
+ return RValue::get(EmitNontemporalStore(*this, E));
case Builtin::BI__c11_atomic_is_lock_free:
case Builtin::BI__atomic_is_lock_free: {
// Call "bool __atomic_is_lock_free(size_t size, void *ptr)". For the
diff --git a/clang/lib/CodeGen/CGExpr.cpp b/clang/lib/CodeGen/CGExpr.cpp
index 6635e57..b7d1ef1 100644
--- a/clang/lib/CodeGen/CGExpr.cpp
+++ b/clang/lib/CodeGen/CGExpr.cpp
@@ -1160,7 +1160,8 @@
return EmitLoadOfScalar(lvalue.getAddress(), lvalue.isVolatile(),
lvalue.getType(), Loc, lvalue.getAlignmentSource(),
lvalue.getTBAAInfo(),
- lvalue.getTBAABaseType(), lvalue.getTBAAOffset());
+ lvalue.getTBAABaseType(), lvalue.getTBAAOffset(),
+ lvalue.isNontemporal());
}
static bool hasBooleanRepresentation(QualType Ty) {
@@ -1226,7 +1227,8 @@
AlignmentSource AlignSource,
llvm::MDNode *TBAAInfo,
QualType TBAABaseType,
- uint64_t TBAAOffset) {
+ uint64_t TBAAOffset,
+ bool isNontemporal) {
// For better performance, handle vector loads differently.
if (Ty->isVectorType()) {
const llvm::Type *EltTy = Addr.getElementType();
@@ -1258,6 +1260,11 @@
}
llvm::LoadInst *Load = Builder.CreateLoad(Addr, Volatile);
+ if (isNontemporal) {
+ llvm::MDNode *Node = llvm::MDNode::get(
+ Load->getContext(), llvm::ConstantAsMetadata::get(Builder.getInt32(1)));
+ Load->setMetadata(CGM.getModule().getMDKindID("nontemporal"), Node);
+ }
if (TBAAInfo) {
llvm::MDNode *TBAAPath = CGM.getTBAAStructTagInfo(TBAABaseType, TBAAInfo,
TBAAOffset);
@@ -1330,7 +1337,8 @@
AlignmentSource AlignSource,
llvm::MDNode *TBAAInfo,
bool isInit, QualType TBAABaseType,
- uint64_t TBAAOffset) {
+ uint64_t TBAAOffset,
+ bool isNontemporal) {
// Handle vectors differently to get better performance.
if (Ty->isVectorType()) {
@@ -1365,6 +1373,12 @@
}
llvm::StoreInst *Store = Builder.CreateStore(Value, Addr, Volatile);
+ if (isNontemporal) {
+ llvm::MDNode *Node =
+ llvm::MDNode::get(Store->getContext(),
+ llvm::ConstantAsMetadata::get(Builder.getInt32(1)));
+ Store->setMetadata(CGM.getModule().getMDKindID("nontemporal"), Node);
+ }
if (TBAAInfo) {
llvm::MDNode *TBAAPath = CGM.getTBAAStructTagInfo(TBAABaseType, TBAAInfo,
TBAAOffset);
@@ -1378,7 +1392,7 @@
EmitStoreOfScalar(value, lvalue.getAddress(), lvalue.isVolatile(),
lvalue.getType(), lvalue.getAlignmentSource(),
lvalue.getTBAAInfo(), isInit, lvalue.getTBAABaseType(),
- lvalue.getTBAAOffset());
+ lvalue.getTBAAOffset(), lvalue.isNontemporal());
}
/// EmitLoadOfLValue - Given an expression that represents a value lvalue, this
diff --git a/clang/lib/CodeGen/CGValue.h b/clang/lib/CodeGen/CGValue.h
index 195571b..3ccc4cd 100644
--- a/clang/lib/CodeGen/CGValue.h
+++ b/clang/lib/CodeGen/CGValue.h
@@ -202,6 +202,10 @@
unsigned AlignSource : 2;
+ // This flag shows if a nontemporal load/stores should be used when accessing
+ // this lvalue.
+ bool Nontemporal : 1;
+
Expr *BaseIvarExp;
/// Used by struct-path-aware TBAA.
@@ -228,6 +232,7 @@
// Initialize Objective-C flags.
this->Ivar = this->ObjIsArray = this->NonGC = this->GlobalObjCRef = false;
this->ImpreciseLifetime = false;
+ this->Nontemporal = false;
this->ThreadLocalRef = false;
this->BaseIvarExp = nullptr;
@@ -277,6 +282,8 @@
void setARCPreciseLifetime(ARCPreciseLifetime_t value) {
ImpreciseLifetime = (value == ARCImpreciseLifetime);
}
+ bool isNontemporal() const { return Nontemporal; }
+ void setNontemporal(bool Value) { Nontemporal = Value; }
bool isObjCWeak() const {
return Quals.getObjCGCAttr() == Qualifiers::Weak;
diff --git a/clang/lib/CodeGen/CodeGenFunction.h b/clang/lib/CodeGen/CodeGenFunction.h
index 86154e9..3925932 100644
--- a/clang/lib/CodeGen/CodeGenFunction.h
+++ b/clang/lib/CodeGen/CodeGenFunction.h
@@ -2449,7 +2449,8 @@
AlignmentSource::Type,
llvm::MDNode *TBAAInfo = nullptr,
QualType TBAABaseTy = QualType(),
- uint64_t TBAAOffset = 0);
+ uint64_t TBAAOffset = 0,
+ bool isNontemporal = false);
/// EmitLoadOfScalar - Load a scalar value from an address, taking
/// care to appropriately convert from the memory representation to
@@ -2465,7 +2466,7 @@
AlignmentSource AlignSource = AlignmentSource::Type,
llvm::MDNode *TBAAInfo = nullptr, bool isInit = false,
QualType TBAABaseTy = QualType(),
- uint64_t TBAAOffset = 0);
+ uint64_t TBAAOffset = 0, bool isNontemporal = false);
/// EmitStoreOfScalar - Store a scalar value to an address, taking
/// care to appropriately convert from the memory representation to
diff --git a/clang/lib/Sema/SemaChecking.cpp b/clang/lib/Sema/SemaChecking.cpp
index 92c9456..c6a00b1 100644
--- a/clang/lib/Sema/SemaChecking.cpp
+++ b/clang/lib/Sema/SemaChecking.cpp
@@ -441,6 +441,9 @@
case Builtin::BI__sync_swap_8:
case Builtin::BI__sync_swap_16:
return SemaBuiltinAtomicOverloaded(TheCallResult);
+ case Builtin::BI__builtin_nontemporal_load:
+ case Builtin::BI__builtin_nontemporal_store:
+ return SemaBuiltinNontemporalOverloaded(TheCallResult);
#define BUILTIN(ID, TYPE, ATTRS)
#define ATOMIC_BUILTIN(ID, TYPE, ATTRS) \
case Builtin::BI##ID: \
@@ -2210,6 +2213,78 @@
return TheCallResult;
}
+/// SemaBuiltinNontemporalOverloaded - We have a call to
+/// __builtin_nontemporal_store or __builtin_nontemporal_load, which is an
+/// overloaded function based on the pointer type of its last argument.
+///
+/// This function goes through and does final semantic checking for these
+/// builtins.
+ExprResult Sema::SemaBuiltinNontemporalOverloaded(ExprResult TheCallResult) {
+ CallExpr *TheCall = (CallExpr *)TheCallResult.get();
+ DeclRefExpr *DRE =
+ cast<DeclRefExpr>(TheCall->getCallee()->IgnoreParenCasts());
+ FunctionDecl *FDecl = cast<FunctionDecl>(DRE->getDecl());
+ unsigned BuiltinID = FDecl->getBuiltinID();
+ assert((BuiltinID == Builtin::BI__builtin_nontemporal_store ||
+ BuiltinID == Builtin::BI__builtin_nontemporal_load) &&
+ "Unexpected nontemporal load/store builtin!");
+ bool isStore = BuiltinID == Builtin::BI__builtin_nontemporal_store;
+ unsigned numArgs = isStore ? 2 : 1;
+
+ // Ensure that we have the proper number of arguments.
+ if (checkArgCount(*this, TheCall, numArgs))
+ return ExprError();
+
+ // Inspect the last argument of the nontemporal builtin. This should always
+ // be a pointer type, from which we imply the type of the memory access.
+ // Because it is a pointer type, we don't have to worry about any implicit
+ // casts here.
+ Expr *PointerArg = TheCall->getArg(numArgs - 1);
+ ExprResult PointerArgResult =
+ DefaultFunctionArrayLvalueConversion(PointerArg);
+
+ if (PointerArgResult.isInvalid())
+ return ExprError();
+ PointerArg = PointerArgResult.get();
+ TheCall->setArg(numArgs - 1, PointerArg);
+
+ const PointerType *pointerType = PointerArg->getType()->getAs<PointerType>();
+ if (!pointerType) {
+ Diag(DRE->getLocStart(), diag::err_nontemporal_builtin_must_be_pointer)
+ << PointerArg->getType() << PointerArg->getSourceRange();
+ return ExprError();
+ }
+
+ QualType ValType = pointerType->getPointeeType();
+
+ // Strip any qualifiers off ValType.
+ ValType = ValType.getUnqualifiedType();
+ if (!ValType->isIntegerType() && !ValType->isAnyPointerType() &&
+ !ValType->isBlockPointerType() && !ValType->isFloatingType() &&
+ !ValType->isVectorType()) {
+ Diag(DRE->getLocStart(),
+ diag::err_nontemporal_builtin_must_be_pointer_intfltptr_or_vector)
+ << PointerArg->getType() << PointerArg->getSourceRange();
+ return ExprError();
+ }
+
+ if (!isStore) {
+ TheCall->setType(ValType);
+ return TheCallResult;
+ }
+
+ ExprResult ValArg = TheCall->getArg(0);
+ InitializedEntity Entity = InitializedEntity::InitializeParameter(
+ Context, ValType, /*consume*/ false);
+ ValArg = PerformCopyInitialization(Entity, SourceLocation(), ValArg);
+ if (ValArg.isInvalid())
+ return ExprError();
+
+ TheCall->setArg(0, ValArg.get());
+ TheCall->setType(Context.VoidTy);
+ return TheCallResult;
+}
+
/// CheckObjCString - Checks that the argument to the builtin
/// CFString constructor is correct
/// Note: It might also make sense to do the UTF-16 conversion here (would