[llvm-branch-commits] [clang] clang/AMDGPU: Emit atomicrmw for global/flat fadd v2bf16 builtins (PR #96875)

[Matt Arsenault via llvm-branch-commits]

================
@@ -48,7 +48,7 @@ void test_local_add_2f16_noret(__local half2 *addr, half2 x) {
 }
 
 // CHECK-LABEL: test_flat_add_2f16
-// CHECK: [[RMW:%.+]] = atomicrmw fadd ptr %{{.+}}, <2 x half> %{{.+}} syncscope("agent") seq_cst, align 4, !amdgpu.no.fine.grained.memory !{{[0-9]+$}}
+// CHECK: [[RMW:%.+]] = atomicrmw fadd ptr %{{.+}}, <2 x half> %{{.+}} syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory !{{[0-9]+$}}
----------------
arsenm wrote:

I'm not sure. My guess is this changed in the downstream patch change to seq_cst to monotonic, and this diff wasn't updated? Before this patch the atomicrmw wouldn't have been produced at all 

https://github.com/llvm/llvm-project/pull/96875