[Mlir-commits] [mlir] [MLIR][XeGPU] Use updated dpas_mx op print format. (PR #202700)

[llvmlistbot at llvm.org]

llvmorg-github-actions[bot] wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-mlir

Author: Sang Ik Lee (silee2)

<details>
<summary>Changes</summary>

Old assembly format causes parse error.

---
Full diff: https://github.com/llvm/llvm-project/pull/202700.diff


2 Files Affected:

- (modified) mlir/test/Integration/Dialect/XeGPU/LANE/xegpu_dpas_mx_prepacked_bf8.mlir (+1-1) 
- (modified) mlir/test/Integration/Dialect/XeGPU/LANE/xegpu_dpas_mx_prepacked_e2m1.mlir (+1-1) 


``````````diff
diff --git a/mlir/test/Integration/Dialect/XeGPU/LANE/xegpu_dpas_mx_prepacked_bf8.mlir b/mlir/test/Integration/Dialect/XeGPU/LANE/xegpu_dpas_mx_prepacked_bf8.mlir
index 7d6e806b16e72..d1a19556a1232 100644
--- a/mlir/test/Integration/Dialect/XeGPU/LANE/xegpu_dpas_mx_prepacked_bf8.mlir
+++ b/mlir/test/Integration/Dialect/XeGPU/LANE/xegpu_dpas_mx_prepacked_bf8.mlir
@@ -32,7 +32,7 @@ module @gemm attributes {gpu.container_module} {
       %scale_b_elem = memref.load %scale_b[%c0, %id_x] : memref<1x16xf8E8M0FNU>
       %scale_b_loaded = vector.insert %scale_b_elem, %scale_b_undef[%c0] : f8E8M0FNU into vector<1xf8E8M0FNU>
 
-      %c_result = xegpu.dpas_mx %a_trunc, %b_trunc, %c_loaded scale_a = %scale_a_loaded scale_b = %scale_b_loaded : vector<16xf8E5M2>, vector<32xf8E5M2>, vector<8xf32>, vector<1xf8E8M0FNU>, vector<1xf8E8M0FNU> -> vector<8xf32>
+      %c_result = xegpu.dpas_mx %a_trunc, %b_trunc, %c_loaded scale_a = %scale_a_loaded scale_b = %scale_b_loaded : (vector<16xf8E5M2>, vector<32xf8E5M2>, vector<8xf32>, vector<1xf8E8M0FNU>, vector<1xf8E8M0FNU>) -> vector<8xf32>
 
       xegpu.store_nd %c_result, %tdesc_c[0, 0] : vector<8xf32>, !xegpu.tensor_desc<8x16xf32>
       gpu.return
diff --git a/mlir/test/Integration/Dialect/XeGPU/LANE/xegpu_dpas_mx_prepacked_e2m1.mlir b/mlir/test/Integration/Dialect/XeGPU/LANE/xegpu_dpas_mx_prepacked_e2m1.mlir
index 67fa43d00a4f2..669c49053abde 100644
--- a/mlir/test/Integration/Dialect/XeGPU/LANE/xegpu_dpas_mx_prepacked_e2m1.mlir
+++ b/mlir/test/Integration/Dialect/XeGPU/LANE/xegpu_dpas_mx_prepacked_e2m1.mlir
@@ -49,7 +49,7 @@ module @gemm attributes {gpu.container_module} {
       %scale_b_tmp = vector.insert %first_b, %scale_b_undef[%c0] : f8E8M0FNU into vector<2xf8E8M0FNU>
       %scale_b_loaded = vector.insert %second_b, %scale_b_tmp[%c1] : f8E8M0FNU into vector<2xf8E8M0FNU>
 
-      %c_result = xegpu.dpas_mx %a_trunc, %b_trunc, %c_loaded scale_a = %scale_a_loaded scale_b = %scale_b_loaded : vector<32xf4E2M1FN>, vector<64xf4E2M1FN>, vector<8xf32>, vector<2xf8E8M0FNU>, vector<2xf8E8M0FNU> -> vector<8xf32>
+      %c_result = xegpu.dpas_mx %a_trunc, %b_trunc, %c_loaded scale_a = %scale_a_loaded scale_b = %scale_b_loaded : (vector<32xf4E2M1FN>, vector<64xf4E2M1FN>, vector<8xf32>, vector<2xf8E8M0FNU>, vector<2xf8E8M0FNU>) -> vector<8xf32>
 
       xegpu.store_nd %c_result, %tdesc_c[0, 0] : vector<8xf32>, !xegpu.tensor_desc<8x16xf32>
       gpu.return

``````````

</details>


https://github.com/llvm/llvm-project/pull/202700