[llvm] [AArch64] Fix DUP-of-extload combine to ignore chain uses (PR #201351)

[via llvm-commits]

llvmorg-github-actions[bot] wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-backend-aarch64

Author: Guy David (guy-david)

<details>
<summary>Changes</summary>

The original combine bailed when the load had more than one use, but counted chain uses too.

---
Full diff: https://github.com/llvm/llvm-project/pull/201351.diff


2 Files Affected:

- (modified) llvm/lib/Target/AArch64/AArch64ISelLowering.cpp (+1-1) 
- (modified) llvm/test/CodeGen/AArch64/dup.ll (+6-6) 


``````````diff
diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
index 89466adf7d4f3..40dd3fe11e982 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -28461,7 +28461,7 @@ static SDValue performDUPCombine(SDNode *N,
       EVT ElemVT = VT.getVectorElementType();
       if ((ExtType == ISD::EXTLOAD || ExtType == ISD::ZEXTLOAD) &&
           (MemVT == MVT::i8 || MemVT == MVT::i16 || MemVT == MVT::i32) &&
-          ElemVT != MemVT && LD->hasOneUse()) {
+          ElemVT != MemVT && Op.hasOneUse()) {
         EVT Vec128VT = EVT::getVectorVT(*DCI.DAG.getContext(), ElemVT,
                                         128 / ElemVT.getSizeInBits());
         SDValue ScalarToVec =
diff --git a/llvm/test/CodeGen/AArch64/dup.ll b/llvm/test/CodeGen/AArch64/dup.ll
index 7d5a2fad6de6b..593f30de20b1a 100644
--- a/llvm/test/CodeGen/AArch64/dup.ll
+++ b/llvm/test/CodeGen/AArch64/dup.ll
@@ -74,9 +74,9 @@ entry:
 define <2 x i8> @loaddup_str_v2i8(ptr %p) {
 ; CHECK-SD-LABEL: loaddup_str_v2i8:
 ; CHECK-SD:       // %bb.0: // %entry
-; CHECK-SD-NEXT:    ldrb w8, [x0]
+; CHECK-SD-NEXT:    ldr b0, [x0]
 ; CHECK-SD-NEXT:    strb wzr, [x0]
-; CHECK-SD-NEXT:    dup v0.2s, w8
+; CHECK-SD-NEXT:    dup v0.2s, v0.s[0]
 ; CHECK-SD-NEXT:    ret
 ;
 ; CHECK-GI-LABEL: loaddup_str_v2i8:
@@ -247,9 +247,9 @@ entry:
 define <4 x i8> @loaddup_str_v4i8(ptr %p) {
 ; CHECK-SD-LABEL: loaddup_str_v4i8:
 ; CHECK-SD:       // %bb.0: // %entry
-; CHECK-SD-NEXT:    ldrb w8, [x0]
+; CHECK-SD-NEXT:    ldr b0, [x0]
 ; CHECK-SD-NEXT:    strb wzr, [x0]
-; CHECK-SD-NEXT:    dup v0.4h, w8
+; CHECK-SD-NEXT:    dup v0.4h, v0.h[0]
 ; CHECK-SD-NEXT:    ret
 ;
 ; CHECK-GI-LABEL: loaddup_str_v4i8:
@@ -498,9 +498,9 @@ entry:
 define <2 x i16> @loaddup_str_v2i16(ptr %p) {
 ; CHECK-SD-LABEL: loaddup_str_v2i16:
 ; CHECK-SD:       // %bb.0: // %entry
-; CHECK-SD-NEXT:    ldrh w8, [x0]
+; CHECK-SD-NEXT:    ldr h0, [x0]
 ; CHECK-SD-NEXT:    strh wzr, [x0]
-; CHECK-SD-NEXT:    dup v0.2s, w8
+; CHECK-SD-NEXT:    dup v0.2s, v0.s[0]
 ; CHECK-SD-NEXT:    ret
 ;
 ; CHECK-GI-LABEL: loaddup_str_v2i16:

``````````

</details>


https://github.com/llvm/llvm-project/pull/201351