[X86] Add negative test for D145930

[n-omer]

Summary

This patch adds a negative test for the issue discovered in D145930.

[n-omer]

Update diff.

[n-omer]

This comment was removed by n-omer.

[pengfei]

llvm/test/CodeGen/X86/atomic-rm-bit-test-64.ll

2 ↗

(On Diff #504624)

Why adding it? Doesn't look necessary.

[pengfei]

llvm/test/CodeGen/X86/atomic-rm-bit-test-64.ll

2 ↗

(On Diff #504624)

Sorry, just see the reason in D145930.

[n-omer]

llvm/test/CodeGen/X86/atomic-rm-bit-test-64.ll

2 ↗

(On Diff #504624)

I tried removing it but that eliminates the # fixup A - offset: 5, value: a-5, kind: reloc_riprel_4byte line which demonstrates the issue from my test case.

[RKSimon]

llvm/test/CodeGen/X86/atomic-rm-bit-test-64.ll

1668 ↗

(On Diff #504624)

If this is the only line that is being changed by D145930 - it'd be better to move this to its own pr61384.ll test file

[n-omer]

Address review comments.

[RKSimon]

LGTM

llvm/test/CodeGen/X86/pr61384.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=x86_64-unknown-linux-gnu -show-mc-encoding < %s | FileCheck %s
+
+@a = external dso_local global { { i64 } }
+
+define i32 @atomic_global() nounwind {
+; CHECK-LABEL: atomic_global:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    xorl %eax, %eax # encoding: [0x31,0xc0]
+; CHECK-NEXT:    lock btsq %rax, a(%rip) # encoding: [0xf0,0x48,0x0f,0xab,0x05,A,A,A,A]
+; CHECK-NEXT:    # fixup A - offset: 5, value: a-5, kind: reloc_riprel_4byte
+; CHECK-NEXT:    xorl %eax, %eax # encoding: [0x31,0xc0]
+; CHECK-NEXT:    retq # encoding: [0xc3]
+entry:
+  %shl.i = shl i64 1, 0
+  %0 = atomicrmw or ptr @a, i64 %shl.i monotonic, align 8
+  %and.i = and i64 %shl.i, %0
+  ret i32 0
+}