Min/Max/Abs/Relu recognition (starter project?)

[rengolin]

@aartbik

In practice, however, many of the current paths into MLIR lower some of the very relevant ops too early, and having some reasonable matcher to correct such cases would go a long way

Another path is to make projects like torch-mlir a canonicalizer of ingress. It can already understand HLO, Torch and ONNX inputs, if it can lower them all to the same Linalg representation (high level named ops), then the relevance of such matcher would be only for special cases.

[RFC] Torch-mlir lowering to named ops

Looking into torch-mlir, the lowering to linalg is as generic but convoluted as I expected. Lots of corner cases, but also lots of generic builders and code reuse. Trying to create an ingress in linalg named ops there may not be simple enough, especially if we want to keep the choice (generic vs. named ops) for where there’s an option. But given that we have a generalization pass, for the recently added, we could still produce the same result after lowering + generalization. However, as discus…

This is important because the matchers take a considerable effort to build correctly and to run at compile time.