Simplify code; NFC.

Also, moved test cases from CodeGen/X86/fold-buildvector-bug.ll into
CodeGen/X86/buildvec-insertvec.ll and regenerated CHECK lines using
update_llc_test_checks.py.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@239142 91177308-0d34-0410-b5e6-96231b3b80d8

Author: adibiagio

lib/CodeGen/SelectionDAG/DAGCombiner.cpp

@@ -619,7 +619,7 @@ static SDValue GetNegatedExpression(SDValue Op, SelectionDAG &DAG,
 
     // fold (fneg (fsub 0, B)) -> B
     if (ConstantFPSDNode *N0CFP = dyn_cast<ConstantFPSDNode>(Op.getOperand(0)))
-      if (N0CFP->getValueAPF().isZero())
+      if (N0CFP->isZero())
         return Op.getOperand(1);
 
     // fold (fneg (fsub A, B)) -> (fsub B, A)
@@ -7864,7 +7864,7 @@ SDValue DAGCombiner::visitFADD(SDNode *N) {
     bool AllowNewConst = (Level < AfterLegalizeDAG);
 
     // fold (fadd A, 0) -> A
-    if (N1CFP && N1CFP->getValueAPF().isZero())
+    if (N1CFP && N1CFP->isZero())
       return N0;
 
     // fold (fadd (fadd x, c1), c2) -> (fadd x, (fadd c1, c2))
@@ -7995,11 +7995,11 @@ SDValue DAGCombiner::visitFSUB(SDNode *N) {
   // If 'unsafe math' is enabled, fold lots of things.
   if (Options.UnsafeFPMath) {
     // (fsub A, 0) -> A
-    if (N1CFP && N1CFP->getValueAPF().isZero())
+    if (N1CFP && N1CFP->isZero())
       return N0;
 
     // (fsub 0, B) -> -B
-    if (N0CFP && N0CFP->getValueAPF().isZero()) {
+    if (N0CFP && N0CFP->isZero()) {
       if (isNegatibleForFree(N1, LegalOperations, TLI, &Options))
         return GetNegatedExpression(N1, DAG, LegalOperations);
       if (!LegalOperations || TLI.isOperationLegal(ISD::FNEG, VT))
@@ -8065,7 +8065,7 @@ SDValue DAGCombiner::visitFMUL(SDNode *N) {
 
   if (Options.UnsafeFPMath) {
     // fold (fmul A, 0) -> 0
-    if (N1CFP && N1CFP->getValueAPF().isZero())
+    if (N1CFP && N1CFP->isZero())
       return N1;
 
     // fold (fmul (fmul x, c1), c2) -> (fmul x, (fmul c1, c2))
@@ -12995,7 +12995,7 @@ SDValue DAGCombiner::SimplifyVBinOp(SDNode *N) {
       if (N->getOpcode() == ISD::SDIV || N->getOpcode() == ISD::UDIV ||
           N->getOpcode() == ISD::FDIV) {
         if (isNullConstant(RHSOp) || (RHSOp.getOpcode() == ISD::ConstantFP &&
-             cast<ConstantFPSDNode>(RHSOp.getNode())->getValueAPF().isZero()))
+             cast<ConstantFPSDNode>(RHSOp.getNode())->isZero()))
           break;
       }
 
@@ -13259,7 +13259,7 @@ SDValue DAGCombiner::SimplifySelectCC(SDLoc DL, SDValue N0, SDValue N1,
   // Check to see if we can simplify the select into an fabs node
   if (ConstantFPSDNode *CFP = dyn_cast<ConstantFPSDNode>(N1)) {
     // Allow either -0.0 or 0.0
-    if (CFP->getValueAPF().isZero()) {
+    if (CFP->isZero()) {
       // select (setg[te] X, +/-0.0), X, fneg(X) -> fabs
       if ((CC == ISD::SETGE || CC == ISD::SETGT) &&
           N0 == N2 && N3.getOpcode() == ISD::FNEG &&

test/CodeGen/X86/buildvec-insertvec.ll

@@ -1,15 +1,56 @@
-; RUN: llc < %s -mcpu=corei7 -mtriple=x86_64-unknown-linux-gnu | FileCheck %s
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s
 
 define void @foo(<3 x float> %in, <4 x i8>* nocapture %out) nounwind {
+; CHECK-LABEL: foo:
+; CHECK:       # BB#0:
+; CHECK-NEXT:    cvttps2dq %xmm0, %xmm0
+; CHECK-NEXT:    movl $255, %eax
+; CHECK-NEXT:    pinsrd $3, %eax, %xmm0
+; CHECK-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; CHECK-NEXT:    movd %xmm0, (%rdi)
+; CHECK-NEXT:    retq
   %t0 = fptoui <3 x float> %in to <3 x i8>
   %t1 = shufflevector <3 x i8> %t0, <3 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 undef>
   %t2 = insertelement <4 x i8> %t1, i8 -1, i32 3
   store <4 x i8> %t2, <4 x i8>* %out, align 4
   ret void
-; CHECK: foo
-; CHECK: cvttps2dq
-; CHECK-NOT: pextrd
-; CHECK: pinsrd
-; CHECK-NEXT: pshufb
-; CHECK: ret
+}
+
+; Verify that the DAGCombiner doesn't wrongly fold a build_vector into a
+; blend with a zero vector if the build_vector contains negative zero.
+;
+; TODO: the codegen for function 'test_negative_zero_1' is sub-optimal.
+; Ideally, we should generate a single shuffle blend operation.
+
+define <4 x float> @test_negative_zero_1(<4 x float> %A) {
+; CHECK-LABEL: test_negative_zero_1:
+; CHECK:       # BB#0: # %entry
+; CHECK-NEXT:    movapd %xmm0, %xmm1
+; CHECK-NEXT:    shufpd {{.*#+}} xmm1 = xmm1[1,0]
+; CHECK-NEXT:    xorps %xmm2, %xmm2
+; CHECK-NEXT:    blendps {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
+; CHECK-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; CHECK-NEXT:    unpcklpd {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; CHECK-NEXT:    retq
+entry:
+  %0 = extractelement <4 x float> %A, i32 0
+  %1 = insertelement <4 x float> undef, float %0, i32 0
+  %2 = insertelement <4 x float> %1, float -0.0, i32 1
+  %3 = extractelement <4 x float> %A, i32 2
+  %4 = insertelement <4 x float> %2, float %3, i32 2
+  %5 = insertelement <4 x float> %4, float 0.0, i32 3
+  ret <4 x float> %5
+}
+
+define <2 x double> @test_negative_zero_2(<2 x double> %A) {
+; CHECK-LABEL: test_negative_zero_2:
+; CHECK:       # BB#0: # %entry
+; CHECK-NEXT:    movhpd {{.*}}(%rip), %xmm0
+; CHECK-NEXT:    retq
+entry:
+  %0 = extractelement <2 x double> %A, i32 0
+  %1 = insertelement <2 x double> undef, double %0, i32 0
+  %2 = insertelement <2 x double> %1, double -0.0, i32 1
+  ret <2 x double> %2
 }