[CodeGen][X86] Fix handling of __fp16 vectors.
This commit fixes a bug in IRGen where it generates completely broken
code for __fp16 vectors on X86. For example when the following code is
compiled:
half4 hv0, hv1, hv2; // these are vectors of __fp16.
void foo221() {
hv0 = hv1 + hv2;
}
clang generates the following IR, in which two i16 vectors are added:
@hv1 = common global <4 x i16> zeroinitializer, align 8
@hv2 = common global <4 x i16> zeroinitializer, align 8
@hv0 = common global <4 x i16> zeroinitializer, align 8
define void @foo221() {
%0 = load <4 x i16>, <4 x i16>* @hv1, align 8
%1 = load <4 x i16>, <4 x i16>* @hv2, align 8
%add = add <4 x i16> %0, %1
store <4 x i16> %add, <4 x i16>* @hv0, align 8
ret void
}
To fix the bug, this commit uses the code committed in r314056, which
modified clang to promote and truncate __fp16 vectors to and from float
vectors in the AST. It also fixes another IRGen bug where a short value
is assigned to an __fp16 variable without any integer-to-floating-point
conversion, as shown in the following example:
__fp16 a;
short b;
void foo1() {
a = b;
}
@b = common global i16 0, align 2
@a = common global i16 0, align 2
define void @foo1() #0 {
%0 = load i16, i16* @b, align 2
store i16 %0, i16* @a, align 2
ret void
}
rdar://problem/20625184
Differential Revision: https://reviews.llvm.org/D40112
llvm-svn: 320215
diff --git a/clang/lib/Basic/Targets/AArch64.h b/clang/lib/Basic/Targets/AArch64.h
index 1b8a774..33268f0 100644
--- a/clang/lib/Basic/Targets/AArch64.h
+++ b/clang/lib/Basic/Targets/AArch64.h
@@ -48,6 +48,10 @@
bool isValidCPUName(StringRef Name) const override;
bool setCPU(const std::string &Name) override;
+ bool useFP16ConversionIntrinsics() const override {
+ return false;
+ }
+
void getTargetDefinesARMV81A(const LangOptions &Opts,
MacroBuilder &Builder) const;
void getTargetDefinesARMV82A(const LangOptions &Opts,
diff --git a/clang/lib/Basic/Targets/ARM.h b/clang/lib/Basic/Targets/ARM.h
index 232b23c..fb0e7e6 100644
--- a/clang/lib/Basic/Targets/ARM.h
+++ b/clang/lib/Basic/Targets/ARM.h
@@ -126,6 +126,10 @@
bool setFPMath(StringRef Name) override;
+ bool useFP16ConversionIntrinsics() const override {
+ return false;
+ }
+
void getTargetDefinesARMV81A(const LangOptions &Opts,
MacroBuilder &Builder) const;
diff --git a/clang/lib/Basic/Targets/X86.h b/clang/lib/Basic/Targets/X86.h
index 5a4b6b5..b181159 100644
--- a/clang/lib/Basic/Targets/X86.h
+++ b/clang/lib/Basic/Targets/X86.h
@@ -193,6 +193,10 @@
return "";
}
+ bool useFP16ConversionIntrinsics() const override {
+ return false;
+ }
+
void getTargetDefines(const LangOptions &Opts,
MacroBuilder &Builder) const override;
diff --git a/clang/lib/CodeGen/CGExprConstant.cpp b/clang/lib/CodeGen/CGExprConstant.cpp
index 3e35d4c..d1b9e13 100644
--- a/clang/lib/CodeGen/CGExprConstant.cpp
+++ b/clang/lib/CodeGen/CGExprConstant.cpp
@@ -1825,7 +1825,7 @@
const llvm::APFloat &Init = Value.getFloat();
if (&Init.getSemantics() == &llvm::APFloat::IEEEhalf() &&
!CGM.getContext().getLangOpts().NativeHalfType &&
- !CGM.getContext().getLangOpts().HalfArgsAndReturns)
+ CGM.getContext().getTargetInfo().useFP16ConversionIntrinsics())
return llvm::ConstantInt::get(CGM.getLLVMContext(),
Init.bitcastToAPInt());
else
diff --git a/clang/lib/CodeGen/CGExprScalar.cpp b/clang/lib/CodeGen/CGExprScalar.cpp
index 7c11103..32e0f88 100644
--- a/clang/lib/CodeGen/CGExprScalar.cpp
+++ b/clang/lib/CodeGen/CGExprScalar.cpp
@@ -951,7 +951,7 @@
if (SrcType->isHalfType() && !CGF.getContext().getLangOpts().NativeHalfType) {
// Cast to FP using the intrinsic if the half type itself isn't supported.
if (DstTy->isFloatingPointTy()) {
- if (!CGF.getContext().getLangOpts().HalfArgsAndReturns)
+ if (CGF.getContext().getTargetInfo().useFP16ConversionIntrinsics())
return Builder.CreateCall(
CGF.CGM.getIntrinsic(llvm::Intrinsic::convert_from_fp16, DstTy),
Src);
@@ -959,7 +959,7 @@
// Cast to other types through float, using either the intrinsic or FPExt,
// depending on whether the half type itself is supported
// (as opposed to operations on half, available with NativeHalfType).
- if (!CGF.getContext().getLangOpts().HalfArgsAndReturns) {
+ if (CGF.getContext().getTargetInfo().useFP16ConversionIntrinsics()) {
Src = Builder.CreateCall(
CGF.CGM.getIntrinsic(llvm::Intrinsic::convert_from_fp16,
CGF.CGM.FloatTy),
@@ -1068,7 +1068,7 @@
if (SrcTy->isFloatingPointTy()) {
// Use the intrinsic if the half type itself isn't supported
// (as opposed to operations on half, available with NativeHalfType).
- if (!CGF.getContext().getLangOpts().HalfArgsAndReturns)
+ if (CGF.getContext().getTargetInfo().useFP16ConversionIntrinsics())
return Builder.CreateCall(
CGF.CGM.getIntrinsic(llvm::Intrinsic::convert_to_fp16, SrcTy), Src);
// If the half type is supported, just use an fptrunc.
@@ -1104,7 +1104,7 @@
}
if (DstTy != ResTy) {
- if (!CGF.getContext().getLangOpts().HalfArgsAndReturns) {
+ if (CGF.getContext().getTargetInfo().useFP16ConversionIntrinsics()) {
assert(ResTy->isIntegerTy(16) && "Only half FP requires extra conversion");
Res = Builder.CreateCall(
CGF.CGM.getIntrinsic(llvm::Intrinsic::convert_to_fp16, CGF.CGM.FloatTy),
@@ -2028,7 +2028,7 @@
if (type->isHalfType() && !CGF.getContext().getLangOpts().NativeHalfType) {
// Another special case: half FP increment should be done via float
- if (!CGF.getContext().getLangOpts().HalfArgsAndReturns) {
+ if (CGF.getContext().getTargetInfo().useFP16ConversionIntrinsics()) {
value = Builder.CreateCall(
CGF.CGM.getIntrinsic(llvm::Intrinsic::convert_from_fp16,
CGF.CGM.FloatTy),
@@ -2063,7 +2063,7 @@
value = Builder.CreateFAdd(value, amt, isInc ? "inc" : "dec");
if (type->isHalfType() && !CGF.getContext().getLangOpts().NativeHalfType) {
- if (!CGF.getContext().getLangOpts().HalfArgsAndReturns) {
+ if (CGF.getContext().getTargetInfo().useFP16ConversionIntrinsics()) {
value = Builder.CreateCall(
CGF.CGM.getIntrinsic(llvm::Intrinsic::convert_to_fp16,
CGF.CGM.FloatTy),
diff --git a/clang/lib/CodeGen/CodeGenTypes.cpp b/clang/lib/CodeGen/CodeGenTypes.cpp
index 80f2a31..529a13b 100644
--- a/clang/lib/CodeGen/CodeGenTypes.cpp
+++ b/clang/lib/CodeGen/CodeGenTypes.cpp
@@ -451,10 +451,10 @@
case BuiltinType::Half:
// Half FP can either be storage-only (lowered to i16) or native.
- ResultType =
- getTypeForFormat(getLLVMContext(), Context.getFloatTypeSemantics(T),
- Context.getLangOpts().NativeHalfType ||
- Context.getLangOpts().HalfArgsAndReturns);
+ ResultType = getTypeForFormat(
+ getLLVMContext(), Context.getFloatTypeSemantics(T),
+ Context.getLangOpts().NativeHalfType ||
+ !Context.getTargetInfo().useFP16ConversionIntrinsics());
break;
case BuiltinType::Float:
case BuiltinType::Double:
diff --git a/clang/lib/Sema/SemaExpr.cpp b/clang/lib/Sema/SemaExpr.cpp
index 3bd88c7..8c6d11b 100644
--- a/clang/lib/Sema/SemaExpr.cpp
+++ b/clang/lib/Sema/SemaExpr.cpp
@@ -11565,7 +11565,8 @@
static bool needsConversionOfHalfVec(bool OpRequiresConversion, ASTContext &Ctx,
QualType SrcType) {
return OpRequiresConversion && !Ctx.getLangOpts().NativeHalfType &&
- Ctx.getLangOpts().HalfArgsAndReturns && isVector(SrcType, Ctx.HalfTy);
+ !Ctx.getTargetInfo().useFP16ConversionIntrinsics() &&
+ isVector(SrcType, Ctx.HalfTy);
}
/// CreateBuiltinBinOp - Creates a new built-in binary operation with