[Mlir-commits] [mlir] [MLIR] Add `InParallelOpInterface` for parallel combining operations (PR #157736)

[Quinn Dawkins]

================
@@ -628,11 +628,9 @@ func.func @invalid_insert_dest(%in: tensor<100xf32>, %out: tensor<100xf32>) {
 
   %result = scf.forall (%thread_idx) in (%num_threads) shared_outs(%o = %out) -> (tensor<100xf32>) {
       %1 = tensor.extract_slice %in[%thread_idx][1][1] : tensor<100xf32> to tensor<1xf32>
-      scf.forall.in_parallel {
-        // expected-error @+1 {{may only insert into an output block argument}}
-        tensor.parallel_insert_slice %1 into %out[%thread_idx][1][1] :
-          tensor<1xf32> into tensor<100xf32>
-      }
+      // expected-error @+1 {{in_parallel interface op must have an iterating parent}}
+      tensor.parallel_insert_slice %1 into %out[%thread_idx][1][1] :
+        tensor<1xf32> into tensor<100xf32>
----------------
qedawkins wrote:

This test needs to go back to what it was. A lot of the changes here were part of a draft towards a larger change that we're not going forward with (allowing `parallel_insert_slice` outside the terminator).

https://github.com/llvm/llvm-project/pull/157736