diff options
| author | Nadav Rotem <nadav.rotem@intel.com> | 2012-05-10 12:39:13 +0000 |
|---|---|---|
| committer | Nadav Rotem <nadav.rotem@intel.com> | 2012-05-10 12:39:13 +0000 |
| commit | 15946e50c1d6565676dad5dc825f9daff59920cc (patch) | |
| tree | b115199cc3fdd7d9b48cca67801d0839b3d94174 /llvm | |
| parent | b86a3fb8d058205d35141527b161d1eee9bd25d1 (diff) | |
| download | bcm5719-llvm-15946e50c1d6565676dad5dc825f9daff59920cc.tar.gz bcm5719-llvm-15946e50c1d6565676dad5dc825f9daff59920cc.zip | |
AVX2: Add an additional broadcast idiom.
llvm-svn: 156540
Diffstat (limited to 'llvm')
| -rw-r--r-- | llvm/lib/Target/X86/X86ISelLowering.cpp | 7 | ||||
| -rw-r--r-- | llvm/test/CodeGen/X86/avx-vbroadcast.ll | 12 |
2 files changed, 17 insertions, 2 deletions
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp index 4a312f4eb56..a7e72e05f7a 100644 --- a/llvm/lib/Target/X86/X86ISelLowering.cpp +++ b/llvm/lib/Target/X86/X86ISelLowering.cpp @@ -4980,8 +4980,11 @@ X86TargetLowering::LowerVectorBroadcast(SDValue &Op, SelectionDAG &DAG) const { return SDValue(); SDValue Sc = Op.getOperand(0); - if (Sc.getOpcode() != ISD::SCALAR_TO_VECTOR) - return SDValue(); + if (Sc.getOpcode() == ISD::SCALAR_TO_VECTOR) + Ld = Sc.getOperand(0); + else if (Sc.getOpcode() == ISD::BUILD_VECTOR) + Ld = Sc.getOperand(0); + else return SDValue(); Ld = Sc.getOperand(0); ConstSplatVal = (Ld.getOpcode() == ISD::Constant || diff --git a/llvm/test/CodeGen/X86/avx-vbroadcast.ll b/llvm/test/CodeGen/X86/avx-vbroadcast.ll index 26ee1d33a17..0d403d4bb12 100644 --- a/llvm/test/CodeGen/X86/avx-vbroadcast.ll +++ b/llvm/test/CodeGen/X86/avx-vbroadcast.ll @@ -129,3 +129,15 @@ entry: ret <4 x float> %vecinit6.i } + +; CHECK: _RR2 +; CHECK: vbroadcastss (% +; CHECK: ret +define <4 x float> @_RR2(float* %ptr, i32* %k) nounwind uwtable readnone ssp { +entry: + %q = load float* %ptr, align 4 + %v = insertelement <4 x float> undef, float %q, i32 0 + %t = shufflevector <4 x float> %v, <4 x float> undef, <4 x i32> zeroinitializer + ret <4 x float> %t +} + |
