Allow to fold vector load if there is more than one bitcast, so in the case:
%0 = load <8 x i16>* %dest
%1 = shufflevector <8 x i16> %0, <8 x i16> %in,
<8 x i32> < i32 0, i32 1, i32 2, i32 3, i32 13, i32 undef, i32 14, i32 14>
store <8 x i16> %1, <8 x i16>* %dest
We get:
vmovlpd (%eax), %xmm0, %xmm0
instead of:
vmovaps (%eax), %xmm1
vmovsd %xmm1, %xmm0, %xmm0
No extra test-case is added. I just fixed the existing one
(also it uses FileCheck now).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@166971 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index 06b8329..b65fcf1 100644
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -6438,17 +6438,17 @@
}
static bool MayFoldVectorLoad(SDValue V) {
- if (V.hasOneUse() && V.getOpcode() == ISD::BITCAST)
+ while (V.hasOneUse() && V.getOpcode() == ISD::BITCAST)
V = V.getOperand(0);
+
if (V.hasOneUse() && V.getOpcode() == ISD::SCALAR_TO_VECTOR)
V = V.getOperand(0);
if (V.hasOneUse() && V.getOpcode() == ISD::BUILD_VECTOR &&
V.getNumOperands() == 2 && V.getOperand(1).getOpcode() == ISD::UNDEF)
// BUILD_VECTOR (load), undef
V = V.getOperand(0);
- if (MayFoldLoad(V))
- return true;
- return false;
+
+ return MayFoldLoad(V);
}
// FIXME: the version above should always be used. Since there's