[Mlir-commits] [mlir] Allow 16 bit floating point operand for LLVM_AtomicRMWOp (PR #110553)

[Giuseppe Rossini]

================
@@ -420,11 +420,13 @@ func.func @atomic_store(%val : f32, %large_val : i256, %ptr : !llvm.ptr) {
 }
 
 // CHECK-LABEL: @atomicrmw
-func.func @atomicrmw(%ptr : !llvm.ptr, %val : f32) {
+func.func @atomicrmw(%ptr : !llvm.ptr, %val : f32, %f16_vec : vector<2xf16>) {
   // CHECK: llvm.atomicrmw fadd %{{.*}}, %{{.*}} monotonic : !llvm.ptr, f32
   %0 = llvm.atomicrmw fadd %ptr, %val monotonic : !llvm.ptr, f32
+  // CHECK: llvm.atomicrmw fadd %{{.*}}, %{{.*}} monotonic : !llvm.ptr, vector<2xf16>
----------------
giuseros wrote:

Same here (about adding your test as the last one)

https://github.com/llvm/llvm-project/pull/110553