================
@@ -73,7 +73,7 @@ gpu.func @no_scf_i8(%arg0: memref<64x64xi8>, %arg1:
vector<8x32xi8>) -> vector<8
// CHECK: %{{.*}} = scf.for %[[K:.*]] = %{{.*}} to %{{.*}} step
%{{.*}} iter_args(%{{.*}}) -> (vector<8x16xf32>) {
// CHECK: %[[T7:.*]] = arith.shrui %[[K]], %[[C1]] : index
// CHECK-NEXT: %[[T8:.*]] = xegpu.load_nd %[[T4]][%{{.*}}, %[[T7]]]
-// CHECK-SAME: {layout_result_0 = #xegpu.layout<lane_layout = [16,
1], lane_data = [1, 1]>} :
+// CHECK-SAME: <{layout = #xegpu.layout<lane_layout = [16, 1],
lane_data = [1, 1]>}> :
----------------
Jianhui-Li wrote:
We keep layout_operand/result for non-anchor ops, and call them local layout.
Anchor op only has anchor layout, so there is no redundant layout.
https://github.com/llvm/llvm-project/pull/172125
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
