From ccb2b1047ea9bae05231cceac2632153d92ef076 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Sun, 23 Nov 2025 14:28:05 +0000
Subject: [PATCH] [X86] X86ISelDAGToDAG - don't let ADD/SUB(X,1) ->
 SUB/ADD(X,-1) constant fold

Extension to #168726 - ensure we peek through bitcasts to look for constants (as constant folding will)

Fixes #169205
---
 llvm/lib/Target/X86/X86ISelDAGToDAG.cpp |  3 ++-
 llvm/test/CodeGen/X86/pr169205.ll       | 23 +++++++++++++++++++++++
 2 files changed, 25 insertions(+), 1 deletion(-)
 create mode 100644 llvm/test/CodeGen/X86/pr169205.ll

diff --git a/llvm/lib/Target/X86/X86ISelDAGToDAG.cpp b/llvm/lib/Target/X86/X86ISelDAGToDAG.cpp
index 2b5a63aa66926..e7903a72d85bb 100644
--- a/llvm/lib/Target/X86/X86ISelDAGToDAG.cpp
+++ b/llvm/lib/Target/X86/X86ISelDAGToDAG.cpp
@@ -1004,7 +1004,8 @@ void X86DAGToDAGISel::PreprocessISelDAG() {
     if ((N->getOpcode() == ISD::ADD || N->getOpcode() == ISD::SUB) &&
         N->getSimpleValueType(0).isVector() && !mayPreventLoadFold()) {
       APInt SplatVal;
-      if (!ISD::isBuildVectorOfConstantSDNodes(N->getOperand(0).getNode()) &&
+      if (!ISD::isBuildVectorOfConstantSDNodes(
+              peekThroughBitcasts(N->getOperand(0)).getNode()) &&
           X86::isConstantSplat(N->getOperand(1), SplatVal) &&
           SplatVal.isOne()) {
         SDLoc DL(N);
diff --git a/llvm/test/CodeGen/X86/pr169205.ll b/llvm/test/CodeGen/X86/pr169205.ll
new file mode 100644
index 0000000000000..1416102d6ba77
--- /dev/null
+++ b/llvm/test/CodeGen/X86/pr169205.ll
@@ -0,0 +1,23 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=x86_64-- -mcpu=x86-64    | FileCheck %s --check-prefixes=SSE
+; RUN: llc < %s -mtriple=x86_64-- -mcpu=x86-64-v2 | FileCheck %s --check-prefixes=SSE
+; RUN: llc < %s -mtriple=x86_64-- -mcpu=x86-64-v3 | FileCheck %s --check-prefixes=AVX
+; RUN: llc < %s -mtriple=x86_64-- -mcpu=x86-64-v4 | FileCheck %s --check-prefixes=AVX
+
+define <4 x i16> @PR169205() {
+; SSE-LABEL: PR169205:
+; SSE:       # %bb.0:
+; SSE-NEXT:    movaps {{.*#+}} xmm0 = [1,1,1,1,u,u,u,u]
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: PR169205:
+; AVX:       # %bb.0:
+; AVX-NEXT:    vpxor %xmm0, %xmm0, %xmm0
+; AVX-NEXT:    vpaddw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; AVX-NEXT:    retq
+  %avg = tail call <16 x i8> @llvm.x86.sse2.pavg.b(<16 x i8> zeroinitializer, <16 x i8> zeroinitializer)
+  %shuffle24 = shufflevector <16 x i8> %avg, <16 x i8> zeroinitializer, <4 x i32> <i32 2, i32 4, i32 9, i32 9>
+  %conv25 = zext <4 x i8> %shuffle24 to <4 x i16>
+  %not.neg = add <4 x i16> %conv25, splat (i16 1)
+  ret <4 x i16> %not.neg
+}