[DAG] visitADDLike - update "(x - y) + -1 -> add (xor y, -1), x" fold…

… to accept UNDEF in a splat vector of -1 Make sure we use getNOT instead of reusing the allones (with undefs) vector
llvm · Apr 19, 2024 · 2e68ba9 · 2e68ba9
1 parent 7fbdadb
commit 2e68ba9
Show file tree

Hide file tree

Showing 3 changed files with 13 additions and 15 deletions.
diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -2831,9 +2831,9 @@ SDValue DAGCombiner::visitADDLike(SDNode *N) {
 
   // (x - y) + -1  ->  add (xor y, -1), x
   if (N0.getOpcode() == ISD::SUB && N0.hasOneUse() &&
-      isAllOnesOrAllOnesSplat(N1)) {
-    SDValue Xor = DAG.getNode(ISD::XOR, DL, VT, N0.getOperand(1), N1);
-    return DAG.getNode(ISD::ADD, DL, VT, Xor, N0.getOperand(0));
+      isAllOnesOrAllOnesSplat(N1, /*AllowUndefs=*/true)) {
+    SDValue Not = DAG.getNOT(DL, N0.getOperand(1), VT);
+    return DAG.getNode(ISD::ADD, DL, VT, Not, N0.getOperand(0));
   }
 
   if (SDValue Combined = visitADDLikeCommutative(N0, N1, N))

diff --git a/llvm/test/CodeGen/AArch64/xor.ll b/llvm/test/CodeGen/AArch64/xor.ll
@@ -61,9 +61,8 @@ define <4 x i32> @vec_add_of_not_decrement(<4 x i32> %x, <4 x i32> %y) {
 define <4 x i32> @vec_add_of_not_with_undef(<4 x i32> %x, <4 x i32> %y) {
 ; CHECK-LABEL: vec_add_of_not_with_undef:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    movi v2.2d, #0xffffffffffffffff
-; CHECK-NEXT:    sub v0.4s, v0.4s, v1.4s
-; CHECK-NEXT:    add v0.4s, v0.4s, v2.4s
+; CHECK-NEXT:    mvn v1.16b, v1.16b
+; CHECK-NEXT:    add v0.4s, v1.4s, v0.4s
 ; CHECK-NEXT:    ret
   %t0 = sub <4 x i32> %x, %y
   %r = add <4 x i32> %t0, <i32 -1, i32 undef, i32 -1, i32 -1>

diff --git a/llvm/test/CodeGen/X86/xor.ll b/llvm/test/CodeGen/X86/xor.ll
@@ -627,24 +627,23 @@ define <4 x i32> @vec_add_of_not_decrement(<4 x i32> %x, <4 x i32> %y) {
 define <4 x i32> @vec_add_of_not_with_undef(<4 x i32> %x, <4 x i32> %y) {
 ; X86-LABEL: vec_add_of_not_with_undef:
 ; X86:       # %bb.0:
-; X86-NEXT:    psubd %xmm1, %xmm0
-; X86-NEXT:    pcmpeqd %xmm1, %xmm1
-; X86-NEXT:    paddd %xmm1, %xmm0
+; X86-NEXT:    pcmpeqd %xmm2, %xmm2
+; X86-NEXT:    pxor %xmm1, %xmm2
+; X86-NEXT:    paddd %xmm2, %xmm0
 ; X86-NEXT:    retl
 ;
 ; X64-LIN-LABEL: vec_add_of_not_with_undef:
 ; X64-LIN:       # %bb.0:
-; X64-LIN-NEXT:    psubd %xmm1, %xmm0
-; X64-LIN-NEXT:    pcmpeqd %xmm1, %xmm1
-; X64-LIN-NEXT:    paddd %xmm1, %xmm0
+; X64-LIN-NEXT:    pcmpeqd %xmm2, %xmm2
+; X64-LIN-NEXT:    pxor %xmm1, %xmm2
+; X64-LIN-NEXT:    paddd %xmm2, %xmm0
 ; X64-LIN-NEXT:    retq
 ;
 ; X64-WIN-LABEL: vec_add_of_not_with_undef:
 ; X64-WIN:       # %bb.0:
-; X64-WIN-NEXT:    movdqa (%rcx), %xmm1
-; X64-WIN-NEXT:    psubd (%rdx), %xmm1
 ; X64-WIN-NEXT:    pcmpeqd %xmm0, %xmm0
-; X64-WIN-NEXT:    paddd %xmm1, %xmm0
+; X64-WIN-NEXT:    pxor (%rdx), %xmm0
+; X64-WIN-NEXT:    paddd (%rcx), %xmm0
 ; X64-WIN-NEXT:    retq
   %t0 = sub <4 x i32> %x, %y
   %r = add <4 x i32> %t0, <i32 -1, i32 undef, i32 -1, i32 -1>