[llvm] [X86] combineStore - remove rangedata when converting 64-bit copies to f64 load/store (PR #147904)

[via llvm-commits]

llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-backend-x86

Author: Simon Pilgrim (RKSimon)

<details>
<summary>Changes</summary>

We're changing from i64 to f64 - we can't retain any range metadata

Fixes #<!-- -->147781

---
Full diff: https://github.com/llvm/llvm-project/pull/147904.diff


2 Files Affected:

- (modified) llvm/lib/Target/X86/X86ISelLowering.cpp (+4) 
- (added) llvm/test/CodeGen/X86/pr147781.ll (+19) 


``````````diff
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 347ba1262b66b..1ad1b47a94d28 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -53574,6 +53574,10 @@ static SDValue combineStore(SDNode *N, SelectionDAG &DAG,
 
     SDLoc LdDL(Ld);
     SDLoc StDL(N);
+
+    // Remove any range metadata as we're converting to f64 load/store.
+    Ld->getMemOperand()->clearRanges();
+
     // Lower to a single movq load/store pair.
     SDValue NewLd = DAG.getLoad(MVT::f64, LdDL, Ld->getChain(),
                                 Ld->getBasePtr(), Ld->getMemOperand());
diff --git a/llvm/test/CodeGen/X86/pr147781.ll b/llvm/test/CodeGen/X86/pr147781.ll
new file mode 100644
index 0000000000000..9f65fd5984e7b
--- /dev/null
+++ b/llvm/test/CodeGen/X86/pr147781.ll
@@ -0,0 +1,19 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=i686-pc-windows-gnu -verify-machineinstrs | FileCheck %s
+
+; Ensure i64 !range data is stripped when converting to f64 load/store.
+define void @test(ptr %p, ptr %p2) #0 {
+; CHECK-LABEL: test:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; CHECK-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; CHECK-NEXT:    movsd {{.*#+}} xmm0 = mem[0],zero
+; CHECK-NEXT:    movsd %xmm0, (%eax)
+; CHECK-NEXT:    retl
+  %val = load i64, ptr %p, align 8, !range !0
+  store i64 %val, ptr %p2, align 8
+  ret void
+}
+
+attributes #0 = { "target-cpu"="pentium4" }
+!0 = !{i64 1, i64 0}

``````````

</details>


https://github.com/llvm/llvm-project/pull/147904