Chris Fallin

@jakubDoka could you summarize with a perf measurement (Sightglass or hyperfine of `wasmtime compile` is OK) the overall compile-time effect of this?

PR welcome -- I'm not really that familiar with the present allocator API situation. Adding a Cargo feature for this sounds reasonable to me. It looks like our use of...

@jakubDoka feel free to update the GitHub CI configuration too: the all-features check can probably instead become a list of `cargo check` invocations with specific feature sets, e.g. no default...

@bjorn3 and I were discussing this just now; to add a little more background: - `invoke` encapsulates both the call itself, and the (virtual) branch that is implied by the...

Neat, that's I think essentially what I describe above except two defs (exception ptr as well). Hopefully a fairly straightforward change on the producer side!

I think a reasonable design might be: - An `invoke` of a function with N return types `(T1, T2, ...)` and M exceptional returns `(Exc1, Exc2, ...)` always has a...

So to add an example, if we have ```plain fn0 = %f(i32, i32) -> (i64, i64) block0(v1: i32, v2: i32): v3, v4, v5, v6 = invoke %f(v1, v2), block1(v1), block2(v2)...

> * @cfallin > > An `invoke` of a function with N return types `(T1, T2, ...)` and M exceptional returns `(Exc1, Exc2, ...)` always has a result tuple of...

Reading the [code](https://github.com/bytecodealliance/wasmtime-py/blob/0b66b4f50fa6d45f9a981d9c6c2537e81f14c5a0/wasmtime/_ffi.py#L21-L26) that throws the error, I see ``` if machine == 'AMD64': machine = 'x86_64' if machine == 'arm64': machine = 'aarch64' if machine != 'x86_64' and machine...

I'm pretty perplexed staring at this one too -- for reference here are Godbolt links for [before](https://godbolt.org/z/oYaYYWoM4) (working) and [after](https://godbolt.org/z/ccfjPxTrv) (apparently segfaulting). The clobber-saves come in different order wrt loading...