Going back to a month ago me, I think I agree - I was using scare quotes to indicate that this is spelled pretty badly and the spelling may be hiding some simplifications. Agreed with your point below of generalizing LinAlg (although it seems there would be some real design work needed to isolate exactly how to do that).
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| [RFC] Promoting `linalg.init_tensor` to the `bufferize` dialect | 43 | 1910 | May 22, 2022 | |
| RFC for introducing `abstract_tensor` as a `builtin` type | 25 | 1170 | January 10, 2022 | |
| Linal.generic on tensor seems to have a weird semantics | 20 | 1630 | October 11, 2021 | |
| Computing output shapes of structured ops on tensors | 9 | 1110 | April 15, 2020 | |
| Shape constraint representation | 10 | 1493 | December 15, 2020 |