[llvm] BPF: Generate locked insn for __sync_fetch_and_add() with cpu v1/v2 (PR #106494)

[via llvm-commits]

================
@@ -1,19 +1,22 @@
-; RUN: llc < %s -march=bpfel -verify-machineinstrs -show-mc-encoding | FileCheck --check-prefixes=CHECK,CHECK-V2 %s
-; RUN: llc < %s -march=bpfel -verify-machineinstrs -show-mc-encoding -mcpu=v3 | FileCheck --check-prefixes=CHECK,CHECK-V3 %s
+; RUN: llc < %s -march=bpfel -verify-machineinstrs -show-mc-encoding | FileCheck %s
----------------
4ast wrote:

should we add -mcpu=v1 here and maybe flip a default to -mcpu=v3 ?
So that by default memory ordering is preserved and users that want to see xadd insn for (void)fetch_and_add
have to explicitly request it with -mcpu=v1.

https://github.com/llvm/llvm-project/pull/106494