[Mlir-commits] [mlir] [MLIR] Fixing the memref linearization size computation (PR #138922)

[Zhuoran Yin]

jerryyin wrote:

Re: column major and non-packed test case. I tried again and find those marked illegal func ops. 

```mlir
func.func @memref_alloca_load_i4_row(%mem: memref<4x8xi4, strided<[8, 1]>>, %arg0: index, %arg1: index) {
  return
}
```
This is legal row major memref, but doesn't add any additional test coverage.

```mlir
func.func @memref_alloca_load_i4_column(%mem: memref<8x4xi4, strided<[1, 8]>>, %arg0: index, %arg1: index) {
  return
}
```
This is illegal but already tracked per https://github.com/llvm/llvm-project/blob/bbafa5214e8d5d5daf7cf428780500b13a7d6cbb/mlir/test/Dialect/MemRef/emulate-narrow-type.mlir#L553-L556

So it looks to me that non-conventional memref aren't supported seem to be a known limitation.

https://github.com/llvm/llvm-project/pull/138922