[llvm] [X86] Use fence(seq_cst) in IdempotentRMWIntoFencedLoad (PR #126521)

[Valentin Churavy via llvm-commits]

================
@@ -31807,21 +31807,10 @@ X86TargetLowering::lowerIdempotentRMWIntoFencedLoad(AtomicRMWInst *AI) const {
   // otherwise, we might be able to be more aggressive on relaxed idempotent
   // rmw. In practice, they do not look useful, so we don't try to be
   // especially clever.
-  if (SSID == SyncScope::SingleThread)
----------------
vchuravy wrote:

In  https://github.com/llvm/llvm-project/pull/126521/commits/02f6abd67801b008aa43891b8915a9b6b6d5c71c I added a test

and https://github.com/llvm/llvm-project/pull/126521/commits/a01668d4e65115e4e68bd1efe4920cdcc74724e9 shows the changes with this PR


https://github.com/llvm/llvm-project/pull/126521