[llvm] [AMDGPU] Remove `nosync` from image atomic intrinsics. (PR #76814)

[Matt Arsenault via llvm-commits]

================
@@ -0,0 +1,356 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-attributes --check-globals all --version 4
+; RUN: opt -S -mtriple=amdgcn-unknown-unknown < %s | FileCheck -check-prefixes=CHECK %s
----------------
arsenm wrote:

These tests don't show any transforms, and I assume you're just going for the direct test of the intrinsic attributes. If you're not going to show the effect on a transform of the attribute, this probably belongs in test/Assembler rather than test/CodeGen

https://github.com/llvm/llvm-project/pull/76814