It is currently not possible to rewrite side-effecting instructions within the acyclic egraphs framework if I'm not mistaken, or has anything changed since the RFC regarding this?
That's still the case, yeah
Thanks for the confirmation, Chris! In my work I allow unpinning statements from the skeleton when a rewrite is applied, I imagine this is something that could also be useful in cranelift? (ignoring potential footguns this hands to rule writers)
yes, absolutely, we've talked about ways we could handle this but nothing concrete -- would be interesting to see how you address this!
Great to hear, will keep you updated!
Last updated: Nov 22 2024 at 17:03 UTC