// Transform fneg(bitconvert(x)) -> bitconvert(x^sign) to avoid loading
// constant pool values.
+ // TODO: We can also optimize for vectors here, but we need to make sure
+ // that the sign mask is created properly for each vector element.
if (!TLI.isFNegFree(VT) && N0.getOpcode() == ISD::BITCAST &&
!VT.isVector() &&
N0.getNode()->hasOneUse() &&
// Transform fabs(bitconvert(x)) -> bitconvert(x&~sign) to avoid loading
// constant pool values.
+ // TODO: We can also optimize for vectors here, but we need to make sure
+ // that the sign mask is created properly for each vector element.
if (!TLI.isFAbsFree(VT) &&
N0.getOpcode() == ISD::BITCAST && N0.getNode()->hasOneUse() &&
N0.getOperand(0).getValueType().isInteger() &&
- !N0.getOperand(0).getValueType().isVector()) {
+ !VT.isVector()) {
SDValue Int = N0.getOperand(0);
EVT IntVT = Int.getValueType();
if (IntVT.isInteger() && !IntVT.isVector()) {
-; RUN: llc < %s -mtriple=x86_64-apple-darwin -march=x86 -mcpu=corei7-avx | FileCheck %s
+; RUN: llc < %s -march=x86 -mattr=+avx | FileCheck %s
define <2 x double> @fabs_v2f64(<2 x double> %p)
{
- ; CHECK: fabs_v2f64
+ ; CHECK-LABEL: fabs_v2f64
; CHECK: vandps
%t = call <2 x double> @llvm.fabs.v2f64(<2 x double> %p)
ret <2 x double> %t
define <4 x float> @fabs_v4f32(<4 x float> %p)
{
- ; CHECK: fabs_v4f32
+ ; CHECK-LABEL: fabs_v4f32
; CHECK: vandps
%t = call <4 x float> @llvm.fabs.v4f32(<4 x float> %p)
ret <4 x float> %t
define <4 x double> @fabs_v4f64(<4 x double> %p)
{
- ; CHECK: fabs_v4f64
+ ; CHECK-LABEL: fabs_v4f64
; CHECK: vandps
%t = call <4 x double> @llvm.fabs.v4f64(<4 x double> %p)
ret <4 x double> %t
define <8 x float> @fabs_v8f32(<8 x float> %p)
{
- ; CHECK: fabs_v8f32
+ ; CHECK-LABEL: fabs_v8f32
; CHECK: vandps
%t = call <8 x float> @llvm.fabs.v8f32(<8 x float> %p)
ret <8 x float> %t
}
declare <8 x float> @llvm.fabs.v8f32(<8 x float> %p)
+
+; PR20354: when generating code for a vector fabs op,
+; make sure the correct mask is used for all vector elements.
+; CHECK-LABEL: LCPI4_0
+; CHECK: .long 2147483648
+; CHECK: .long 2147483648
+; CHECK-LABEL: LCPI4_1
+; CHECK: .long 2147483647
+; CHECK: .long 2147483647
+; CHECK-LABEL: fabs_v2f32_1
+; CHECK: vmovdqa LCPI4_0, %xmm0
+; CHECK: vpand LCPI4_1, %xmm0, %xmm0
+; CHECK: vmovd %xmm0, %eax
+; CHECK: vpextrd $1, %xmm0, %edx
+define i64 @fabs_v2f32_1() {
+ %highbits = bitcast i64 9223372039002259456 to <2 x float> ; 0x8000_0000_8000_0000
+ %fabs = call <2 x float> @llvm.fabs.v2f32(<2 x float> %highbits)
+ %ret = bitcast <2 x float> %fabs to i64
+ ret i64 %ret
+}
+
+declare <2 x float> @llvm.fabs.v2f32(<2 x float> %p)