[Mlir-commits] [mlir] [mlir] Drop outermost dims in slice rank reduction inference (PR #95020)

[llvmlistbot at llvm.org]

================
@@ -282,23 +282,6 @@ func.func @insert_slice_of_insert_slice(%t: tensor<f32>, %r0: tensor<1x1xf32>, %
 
 // -----
 
-// CHECK-LABEL: func @insert_slice_of_insert_slice(
----------------
MaheshRavishankar wrote:

Thanks for the explanation! The composition is load bearing. I think the previous test was wrong. IIUC the semantics of rank-reducing extract and insert is that the outer dimensions are dropped. Maybe fix the offsets and keep the test. It's good to test the composition 

https://github.com/llvm/llvm-project/pull/95020