[Mlir-commits] [mlir] [mlir][amdgpu] Fuse adjacent `MemoryCounterWaitOp` (PR #171148)

[Jakub Kuderski]

================
@@ -244,3 +244,14 @@ func.func @scaled_mfma_ugly_shapes(%opA: vector<32xf4E2M1FN>, %opB: vector<32xf4
   %res_7 = amdgpu.scaled_mfma 16x16x128 (%sA_0_7[0] * %opA) * (%sB_6_19[0] * %opB) + %cst_0 : vector<4xf8E8M0FNU>, vector<32xf4E2M1FN>, vector<4xf8E8M0FNU>, vector<32xf4E2M1FN>, vector<4xf32>
   return %res_4, %res_5, %res_6, %res_7 : vector<4xf32>, vector<4xf32>, vector<4xf32>, vector<4xf32>
 }
+
+// -----
+
+// CHECK-LABEL fuse_memory_counter_wait
+func.func @fuse_memory_counter_wait() {
----------------
kuhar wrote:

Can you add a testcase when counters are not consecutive? (e.g., with a function call in between)

https://github.com/llvm/llvm-project/pull/171148