alexcrichton commented on issue #5976:
Mind re-reviewing the latest commit when you get a chance @cfallin? Turns out I was only running Cranelift's test suite and the wasm test suite exercised a case where a right-shift happened at the full bit-width of the type triggering assertions in the backend for emission of these instructions with a 0-shift. I've special-cased the ISLE lowerings to not emit an instruction with 0-shifts but I'm not sure if the ISLE is correct since it relies on rule priorities to be correct, which I don't think is how we'd like to rely on ISLE, right?
Last updated: Dec 23 2024 at 13:07 UTC