[Mlir-commits] [mlir] [MLIR][Interface]: Verify index map ranks before composing loop bounds (PR #173434)

[lonely eagle]

================
@@ -165,6 +165,41 @@ func.func @generic_singular_maps(%arg0: memref<?xf32, affine_map<(i)[off]->(off
   }
 }
 
+// -----
+
+func.func @generic_index_rank0(%arg0: tensor<f32>) -> tensor<f32> {
+// expected-error @+1 {{op expected operand rank (0) to match the result rank of indexing_map #0 (1)}}
+  %0 = linalg.generic {
+    indexing_maps = [
+      affine_map<(d0) -> (d0)>,
+      affine_map<(d0) -> (d0)>
+    ],
+    iterator_types = ["parallel"]}
+      ins(%arg0 : tensor<f32>)
+     outs(%arg0 : tensor<f32>) {
+  ^bb(%0: f32, %1: f32):
+    linalg.yield %1 : f32
+  } -> tensor<f32>
+  return %0 : tensor<f32>
+}
+
+// -----
+
+func.func @generic_index_domain_error(%arg0: tensor<4xf32>) -> tensor<4xf32> {
+// expected-error @+1 {{op expected operand rank (1) to match the result rank of indexing_map #1 (2)}}
----------------
linuxlonelyeagle wrote:

"I believe the number of dimensions in the indexing_maps should be consistent first. Therefore, you should not move the code that checks the dimension count later.

https://github.com/llvm/llvm-project/pull/173434