Hello! I have made a PR implementing LLVM's ghcc in Cranelift: https://github.com/bytecodealliance/wasmtime/pull/13913. Would like to have a review and would like to know if theres any change of merging this into the main tree? My motivation was simple: Tail callconv currently still builds frames, and basically pins only one register so for FP compiles for languages like scheme/sml&nj you can't pin registers for runtime state, nor can you do control flow via jmps instead of call properly.
I am going to test this callconv with my Scheme compiler in a few days to give a proper numbers as in code size reduction and performance increase
Last updated: Jul 29 2026 at 05:03 UTC