diff options
| author | Justin Holewinski <jholewinski@nvidia.com> | 2013-07-26 13:28:29 +0000 |
|---|---|---|
| committer | Justin Holewinski <jholewinski@nvidia.com> | 2013-07-26 13:28:29 +0000 |
| commit | 320185fa5f5838b3892962f6e91083e9729cd946 (patch) | |
| tree | f7f8866b408f466375b99c6430bb5d88e9b0e6f3 | |
| parent | c735c1c2aed2cbaeb61296f4269535b5d13d8b0a (diff) | |
| download | external_llvm-320185fa5f5838b3892962f6e91083e9729cd946.zip external_llvm-320185fa5f5838b3892962f6e91083e9729cd946.tar.gz external_llvm-320185fa5f5838b3892962f6e91083e9729cd946.tar.bz2 | |
Add a target legalize hook for SplitVectorOperand (again)
CustomLowerNode was not being called during SplitVectorOperand,
meaning custom legalization could not be used by targets.
This also adds a test case for NVPTX that depends on this custom
legalization.
Differential Revision: http://llvm-reviews.chandlerc.com/D1195
Attempt to fix the buildbots by making the X86 test I just added platform independent
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187202 91177308-0d34-0410-b5e6-96231b3b80d8
| -rw-r--r-- | lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp | 4 | ||||
| -rw-r--r-- | lib/Target/X86/X86ISelLowering.cpp | 2 | ||||
| -rw-r--r-- | test/CodeGen/NVPTX/vector-stores.ll | 30 | ||||
| -rw-r--r-- | test/CodeGen/X86/floor-soft-float.ll | 13 |
4 files changed, 48 insertions, 1 deletions
diff --git a/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp b/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp index 75bb609..72c16b5 100644 --- a/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp +++ b/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp @@ -1031,6 +1031,10 @@ bool DAGTypeLegalizer::SplitVectorOperand(SDNode *N, unsigned OpNo) { dbgs() << "\n"); SDValue Res = SDValue(); + // See if the target wants to custom split this node. + if (CustomLowerNode(N, N->getOperand(OpNo).getValueType(), false)) + return false; + if (Res.getNode() == 0) { switch (N->getOpcode()) { default: diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp index e75781e..ad2d308 100644 --- a/lib/Target/X86/X86ISelLowering.cpp +++ b/lib/Target/X86/X86ISelLowering.cpp @@ -996,7 +996,7 @@ void X86TargetLowering::resetOperationActions() { setLoadExtAction(ISD::EXTLOAD, MVT::v2f32, Legal); } - if (Subtarget->hasSSE41()) { + if (!TM.Options.UseSoftFloat && Subtarget->hasSSE41()) { setOperationAction(ISD::FFLOOR, MVT::f32, Legal); setOperationAction(ISD::FCEIL, MVT::f32, Legal); setOperationAction(ISD::FTRUNC, MVT::f32, Legal); diff --git a/test/CodeGen/NVPTX/vector-stores.ll b/test/CodeGen/NVPTX/vector-stores.ll new file mode 100644 index 0000000..4941812 --- /dev/null +++ b/test/CodeGen/NVPTX/vector-stores.ll @@ -0,0 +1,30 @@ +; RUN: llc < %s -march=nvptx -mcpu=sm_20 | FileCheck %s + +; CHECK: .visible .func foo1 +; CHECK: st.v2.f32 +define void @foo1(<2 x float> %val, <2 x float>* %ptr) { + store <2 x float> %val, <2 x float>* %ptr + ret void +} + +; CHECK: .visible .func foo2 +; CHECK: st.v4.f32 +define void @foo2(<4 x float> %val, <4 x float>* %ptr) { + store <4 x float> %val, <4 x float>* %ptr + ret void +} + +; CHECK: .visible .func foo3 +; CHECK: st.v2.u32 +define void @foo3(<2 x i32> %val, <2 x i32>* %ptr) { + store <2 x i32> %val, <2 x i32>* %ptr + ret void +} + +; CHECK: .visible .func foo4 +; CHECK: st.v4.u32 +define void @foo4(<4 x i32> %val, <4 x i32>* %ptr) { + store <4 x i32> %val, <4 x i32>* %ptr + ret void +} + diff --git a/test/CodeGen/X86/floor-soft-float.ll b/test/CodeGen/X86/floor-soft-float.ll new file mode 100644 index 0000000..8e7ee09 --- /dev/null +++ b/test/CodeGen/X86/floor-soft-float.ll @@ -0,0 +1,13 @@ +; RUN: llc < %s -march=x86-64 -mattr=+sse41,-avx -soft-float=0 | FileCheck %s --check-prefix=CHECK-HARD-FLOAT +; RUN: llc < %s -march=x86-64 -mattr=+sse41,-avx -soft-float=1 | FileCheck %s --check-prefix=CHECK-SOFT-FLOAT + +target triple = "x86_64-unknown-linux-gnu" + +declare float @llvm.floor.f32(float) + +; CHECK-SOFT-FLOAT: callq floorf +; CHECK-HARD-FLOAT: roundss $1, %xmm0, %xmm0 +define float @myfloor(float %a) { + %val = tail call float @llvm.floor.f32(float %a) + ret float %val +} |
