[Mlir-commits] [mlir] [memref] Support non-scalar copies in `reinterpret_cast` elision (PR #203873)

Andrzej WarzyƄski llvmlistbot at llvm.org
Mon Jun 29 01:02:40 PDT 2026


https://github.com/banach-space commented:

Hey @ioghiban , thank you for updating the tests!

There are quite a few tests and test-categories here and I am struggling to navigate that. I don't think that the original split into "positive" and "negative" works anymore. I have made a few suggestions inline, but basically I am recommending this:

```mlir
//===----------------------------------------------------------------------===//
// 1. The destination is effectively a 0D array within a MemRef with rank == 0 
//===----------------------------------------------------------------------===//

// Positive + negative tests
(...)

//===----------------------------------------------------------------------===//
// 2. The destination is effectively a 1D array within a MemRef with rank >= 1 
//===----------------------------------------------------------------------===//

// Positive + negative tests
(...)

//===----------------------------------------------------------------------===//
// 3. The destination is effectively a 2D array within a MemRef with rank >= 2 
//===----------------------------------------------------------------------===//

// Positive + negative tests
(...)
```

Please also make sure to
* Be consistent about test naming between and within categories (for both positive and negative tests).
* Only change one thing at a time (e.g. don't change src/dst dimensions when switching from static to dynamic offset).

More guidelines are available here: 
* https://mlir.llvm.org/getting_started/TestingGuide/#test-formatting-best-practices
* https://mlir.llvm.org/getting_started/TestingGuide/#test-documentation-best-practices

Thank you!

https://github.com/llvm/llvm-project/pull/203873


More information about the Mlir-commits mailing list