[SelectionDAG] Optional handling of UNDEF elements in matchBinaryPredicate (part...

Now that SimplifyDemandedBits/SimplifyDemandedVectorElts is simplifying vector elements, we're seeing more constant BUILD_VECTOR containing undefs.

This patch provides opt-in support for UNDEF elements in matchBinaryPredicate, passing NULL instead of the result ConstantSDNode* argument.

I've updated the (or (and X, c1), c2) -> (and (or X, c2), c1|c2) fold to demonstrate its use, which I believe is safe for undef cases.

Differential Revision: https://reviews.llvm.org/D55822

llvm-svn: 349629

diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index 10cce78..c4acd09 100644 (file)
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -5175,12 +5175,12 @@ SDValue DAGCombiner::visitOR(SDNode *N) {
     return ROR;
 
   // Canonicalize (or (and X, c1), c2) -> (and (or X, c2), c1|c2)
-  // iff (c1 & c2) != 0.
-  auto MatchIntersect = [](ConstantSDNode *LHS, ConstantSDNode *RHS) {
-    return LHS->getAPIntValue().intersects(RHS->getAPIntValue());
+  // iff (c1 & c2) != 0 or c1/c2 are undef.
+  auto MatchIntersect = [](ConstantSDNode *C1, ConstantSDNode *C2) {
+    return !C1 || !C2 || C1->getAPIntValue().intersects(C2->getAPIntValue());
   };
   if (N0.getOpcode() == ISD::AND && N0.getNode()->hasOneUse() &&
-      ISD::matchBinaryPredicate(N0.getOperand(1), N1, MatchIntersect)) {
+      ISD::matchBinaryPredicate(N0.getOperand(1), N1, MatchIntersect, true)) {
     if (SDValue COR = DAG.FoldConstantArithmetic(
             ISD::OR, SDLoc(N1), VT, N1.getNode(), N0.getOperand(1).getNode())) {
       SDValue IOR = DAG.getNode(ISD::OR, SDLoc(N0), VT, N0.getOperand(0), N1);

diff --git a/llvm/test/CodeGen/X86/known-bits-vector.ll b/llvm/test/CodeGen/X86/known-bits-vector.ll
index d3e4843..cedc740 100644 (file)
--- a/llvm/test/CodeGen/X86/known-bits-vector.ll
+++ b/llvm/test/CodeGen/X86/known-bits-vector.ll
@@ -162,18 +162,12 @@ define <4 x float> @knownbits_mask_shuffle_uitofp(<4 x i32> %a0) nounwind {
 define <4 x float> @knownbits_mask_or_shuffle_uitofp(<4 x i32> %a0) nounwind {
 ; X32-LABEL: knownbits_mask_or_shuffle_uitofp:
 ; X32:       # %bb.0:
-; X32-NEXT:    vandps {{\.LCPI.*}}, %xmm0, %xmm0
-; X32-NEXT:    vorps {{\.LCPI.*}}, %xmm0, %xmm0
-; X32-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
-; X32-NEXT:    vcvtdq2ps %xmm0, %xmm0
+; X32-NEXT:    vmovaps {{.*#+}} xmm0 = [6.5535E+4,6.5535E+4,6.5535E+4,6.5535E+4]
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: knownbits_mask_or_shuffle_uitofp:
 ; X64:       # %bb.0:
-; X64-NEXT:    vandps {{.*}}(%rip), %xmm0, %xmm0
-; X64-NEXT:    vorps {{.*}}(%rip), %xmm0, %xmm0
-; X64-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
-; X64-NEXT:    vcvtdq2ps %xmm0, %xmm0
+; X64-NEXT:    vmovaps {{.*#+}} xmm0 = [6.5535E+4,6.5535E+4,6.5535E+4,6.5535E+4]
 ; X64-NEXT:    retq
   %1 = and <4 x i32> %a0, <i32 -1, i32 -1, i32 255, i32 4085>
   %2 = or <4 x i32> %1, <i32 65535, i32 65535, i32 65535, i32 65535>