Hanna Kruppe

For the record, I did not advocate a "runtime" way for user code to behave differently in const eval, I am probably as uncomfortable with it as you are.

I am not aware of any precedent for u8 being special in Rust, and in fact all the discussion of the invariants of integer types I'm aware of treats all...

> Option (a) is kind of bad (maybe?) for alternative backends like cranelift as they would have to support / implement these LLVM intrinsics to be on parity with the...

However, note that (to the best of my knowledge) the vast majority of those intrinsics are lowered to libcalls rather than single instructions on most or all architectures. They are...

While it's not properly documented (nothing about these intrinsics is), they do in fact [lower to IR that requires the pointers to be aligned](https://godbolt.org/z/WtKT1c). This is also consistent with now...

The hard question is what happens if you want to cast `Box` to a trait object of a subset of {A, B, C, D, ...} rathern than a single one....

Took a quick look, some notes: - Experimentally, `__builtin_ia32_selectd_*` seem to just generate ordinary `select` instructions in LLVM IR, I'm not up to date with stdsimd internally but I'm sure...

Sounds good to me.

Heads up, I've been working on this for a while (https://github.com/jbush001/NyuziToolchain/issues/44 was part of that). It'll be some more time before I have results to show for, but I intend...

FWIW, the code is now public at https://github.com/rkruppe/spmd-vectorize. It's very much research-quality code at the moment, and there's a long list of known limitations that somewhat restrict the usefulness (most...