[llvm-branch-commits] [llvm] [AtomicExpand] Add bitcasts when expanding load atomic vector (PR #148900)

[Matt Arsenault via llvm-branch-commits]

================
@@ -1933,6 +1933,20 @@ def atomic_load_64 :
   let MemoryVT = i64;
 }
 
+def atomic_load_128_v2i64 :
+  PatFrag<(ops node:$ptr),
+          (atomic_load node:$ptr)> {
+  let IsAtomic = true;
+  let MemoryVT = v2i64;
+}
+
+def atomic_load_128_v4i32 :
+  PatFrag<(ops node:$ptr),
+          (atomic_load node:$ptr)> {
+  let IsAtomic = true;
+  let MemoryVT = v4i32;
+}
----------------
arsenm wrote:

I don't get it. The core change here only changes the IR pass, in such a way that the later codegen pieces do not. It would make more sense if this was the patch to directly handling the casting in SelectionDAG 

https://github.com/llvm/llvm-project/pull/148900