Nicolas Vasilache

The back and forth integration has now landed. I added a new github action testing IREE integration on the iree/sandbox branch that is meant to be non-blocking for pushes. The...

Nice! apply/apply_dynamic seem analogous to call/call_indirect, right? I think I see recursion being useful in cases where we write super generic stuff that targets an abstract machine model. I also...

Hi @vikasbz Basically here: https://github.com/google/iree-llvm-sandbox/blob/main/python/core/transforms.py#L15 We would need to print the `self.pipeline` if some TBD `os.getenv` is set. I am unsure about ergonomics atm (e.g. print and exit, or print...

Thanks for your Q, this is a missing feature that is being pushed back into our stack of shorter term things thanks to recent developments. I'd recommend you come ask...

> Do you think that a) is the wrong way to go (or it is harder to do than b) )? I don't think a) is wrong in itself but...

@Mogball thanks for looking into this and starting the discussion. On our end we have been so far iterating on a top-down breakdown of "the transformations + state via attribute...

> The discussion around how passes and transformations should handle attributes does not seem to have been resolved. The general discussion sounds a bit hard to reconcile with all possible...

> Any ideas as to how such rollbacks could be implemented? I would imagine that it might be doable since the dialect conversion framework does maintain two parallel copies of...

I don't have a good definition for it but basically if you have a `set1{ops}` that get transformed to `set2{ops} ... setn{ops}` and at every step the only modifications to...

also @michel-steuwer and @Bastacyclop once they accept the invite