[X86,AArch64] Extend vcmp w/ unary op combine to work w/ more constants.

The transform to constant fold unary operations with an AND across a
vector comparison applies when the constant is not a splat of a scalar
as well.

llvm-svn: 213800

diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
index 9b8ff72..0247030 100644 (file)
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -6501,14 +6501,14 @@ static SDValue performVectorCompareAndMaskUnaryOpCombine(SDNode *N,
       VT.getSizeInBits() != N->getOperand(0)->getValueType(0).getSizeInBits())
     return SDValue();
 
-  // Now check that the other operand of the AND is a constant splat. We could
+  // Now check that the other operand of the AND is a constant. We could
   // make the transformation for non-constant splats as well, but it's unclear
   // that would be a benefit as it would not eliminate any operations, just
   // perform one more step in scalar code before moving to the vector unit.
   if (BuildVectorSDNode *BV =
           dyn_cast<BuildVectorSDNode>(N->getOperand(0)->getOperand(1))) {
-    // Bail out if the vector isn't a constant splat.
-    if (!BV->getConstantSplatNode())
+    // Bail out if the vector isn't a constant.
+    if (!BV->isConstant())
       return SDValue();
 
     // Everything checks out. Build up the new and improved node.

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 95666a4..81889d0 100644 (file)
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -21815,14 +21815,14 @@ static SDValue performVectorCompareAndMaskUnaryOpCombine(SDNode *N,
       VT.getSizeInBits() != N->getOperand(0)->getValueType(0).getSizeInBits())
     return SDValue();
 
-  // Now check that the other operand of the AND is a constant splat. We could
+  // Now check that the other operand of the AND is a constant. We could
   // make the transformation for non-constant splats as well, but it's unclear
   // that would be a benefit as it would not eliminate any operations, just
   // perform one more step in scalar code before moving to the vector unit.
   if (BuildVectorSDNode *BV =
           dyn_cast<BuildVectorSDNode>(N->getOperand(0)->getOperand(1))) {
-    // Bail out if the vector isn't a constant splat.
-    if (!BV->getConstantSplatNode())
+    // Bail out if the vector isn't a constant.
+    if (!BV->isConstant())
       return SDValue();
 
     // Everything checks out. Build up the new and improved node.

diff --git a/llvm/test/CodeGen/AArch64/arm64-setcc-int-to-fp-combine.ll b/llvm/test/CodeGen/AArch64/arm64-setcc-int-to-fp-combine.ll
index 045c9cd..74ca61f 100644 (file)
--- a/llvm/test/CodeGen/AArch64/arm64-setcc-int-to-fp-combine.ll
+++ b/llvm/test/CodeGen/AArch64/arm64-setcc-int-to-fp-combine.ll
@@ -25,3 +25,23 @@ define void @foo1(<4 x float> %val, <4 x float> %test, <4 x double>* %p) nounwin
   store <4 x double> %result, <4 x double>* %p
   ret void
 }
+
+; Fold explicit AND operations when the constant isn't a splat of a single
+; scalar value like what the zext creates.
+define <4 x float> @foo2(<4 x float> %val, <4 x float> %test) nounwind {
+; CHECK-LABEL: lCPI2_0:
+; CHECK-NEXT: .long 1065353216
+; CHECK-NEXT: .long 0
+; CHECK-NEXT: .long 1065353216
+; CHECK-NEXT: .long 0
+; CHECK-LABEL: foo2:
+; CHECK: adrp  x8, lCPI2_0@PAGE
+; CHECK: ldr q2, [x8, lCPI2_0@PAGEOFF]
+; CHECK-NEXT:  fcmeq.4s  v0, v0, v1
+; CHECK-NEXT:  and.16b v0, v0, v2
+  %cmp = fcmp oeq <4 x float> %val, %test
+  %ext = zext <4 x i1> %cmp to <4 x i32>
+  %and = and <4 x i32> %ext, <i32 255, i32 256, i32 257, i32 258>
+  %result = sitofp <4 x i32> %and to <4 x float>
+  ret <4 x float> %result
+}

diff --git a/llvm/test/CodeGen/X86/x86-setcc-int-to-fp-combine.ll b/llvm/test/CodeGen/X86/x86-setcc-int-to-fp-combine.ll
index f737519..4fe6c66 100644 (file)
--- a/llvm/test/CodeGen/X86/x86-setcc-int-to-fp-combine.ll
+++ b/llvm/test/CodeGen/X86/x86-setcc-int-to-fp-combine.ll
@@ -54,3 +54,21 @@ define void @foo2(<4 x float>* noalias %result) nounwind {
   store <4 x float> %val, <4 x float>* %result
   ret void
 }
+
+; Fold explicit AND operations when the constant isn't a splat of a single
+; scalar value like what the zext creates.
+define <4 x float> @foo3(<4 x float> %val, <4 x float> %test) nounwind {
+; CHECK-LABEL: LCPI3_0:
+; CHECK-NEXT: .long 1065353216              ## float 1.000000e+00
+; CHECK-NEXT: .long 0                       ## float 0.000000e+00
+; CHECK-NEXT: .long 1065353216              ## float 1.000000e+00
+; CHECK-NEXT: .long 0                       ## float 0.000000e+00
+; CHECK-LABEL: foo3:
+; CHECK: cmpeqps %xmm1, %xmm0
+; CHECK-NEXT: andps LCPI3_0(%rip), %xmm0
+  %cmp = fcmp oeq <4 x float> %val, %test
+  %ext = zext <4 x i1> %cmp to <4 x i32>
+  %and = and <4 x i32> %ext, <i32 255, i32 256, i32 257, i32 258>
+  %result = sitofp <4 x i32> %and to <4 x float>
+  ret <4 x float> %result
+}