Comment by keithwinstein
1 year ago
Yes, wasm2c implements the Wasm tail-call feature with trampolines, exactly this way. (https://github.com/WebAssembly/wabt/blob/main/test/wasm2c/ta... has an example.)
Doing it with a trampoline is probably slower than if C really had tail calls. On the other hand, adding "real" tail calls to C would probably require changing the ABI (e.g. to "tailcc" or "fastcc -tailcallopt"), and I think there's some reason to think this would probably impose a penalty everywhere (https://llvm.org/docs/CodeGenerator.html#tail-call-optimizat...).
> On the other hand, adding "real" tail calls to C would probably require changing the ABI (e.g. to "tailcc" or "fastcc -tailcallopt")
But [[musttail]] does exactly this while respecting existing calling conventions: https://clang.llvm.org/docs/AttributeReference.html#musttail
No -- as discussed upthread, clang's musttail attribute requires the target function to have the same number of arguments as the caller and for each argument to be similar to the corresponding caller argument. That's stricter than the underlying LLVM musttail marker (when targeting the tailcc/swifttailcc calling conventions) and is too restrictive to implement Wasm's tail-call feature (and probably Scheme's, etc.), at least if arguments are getting passed to functions natively.
It would be nice if the more relaxed rules of the LLVM musttail marker with tailcc could be exposed in clang (and gcc). I think that's basically what "return goto" would do.