[Mlir-commits] [mlir] [mlir][AMDGPU] Plumb address space 7 through MLIR, add address_space attr. (PR #125594)

[Jakub Kuderski]

================
@@ -46,7 +159,9 @@ func.func @gpu_gcn_raw_buffer_load_i32_strided(%buf: memref<16x16xi32, strided<[
     // CHECK: %[[ext_j:.*]] = llvm.mul %[[sz_j]], %[[stride_j]] : i64
     // CHECK: %[[num_records:.*]] = llvm.intr.umax(%[[ext_i]], %[[ext_j]]) : (i64, i64) -> i64
     // CHECK: %[[num_rec_i32:.*]] = llvm.trunc %[[num_records]] : i64 to i32
-    // CHECK: %[[num_rec_bytes_i32:.*]] = llvm.mul %[[num_rec_i32]], %[[elem_size]] : i32
+    // CHECK: %[[elem_size_2:.*]] = llvm.mlir.constant(4 : i32) : i32
+    // CHECK: %[[num_rec_bytes_i32:.*]] = llvm.mul %[[num_rec_i32]], %[[elem_size_2]] : i32
+    // CHECK: %[[stride:.*]] = llvm.mlir.constant(0 : i16) : i16
----------------
kuhar wrote:

Should we use `CHECK-DAG` here to avoid random changes in the future?

https://github.com/llvm/llvm-project/pull/125594