Mateusz Mikuła

> > > I think we need to document about using Rustc on Alpine in Rustup's README file. > The requirement is that Rustc needs gcc or clang pre-existing. It's...

> > > Yes, I agree. While people could run rustc and cargo when having libgcc, missing C compiler prevents users from compiling libc, which is a huge loss anyway....

> My expectations for the musl toolchain binaries were for them to be entirely statically linked. After all, when you build a crate with the musl toolchain, it defaults to...

Rust's llvm already [has this fix](https://github.com/rust-lang/llvm/commit/027b97fe0ac6c6b7fcaca4f4dd09e72093918a6d#diff-9981905953a4b72b0283a2a5ca6abcca). I'll see if I can reproduce it later on today.

64bit Windows 10, latest gnu nightly and the same result as in https://github.com/rust-lang-nursery/packed_simd/issues/72#issuecomment-424458210 BTW I had to remove `deny` from `src/lib.rs` because of dozens of warnings: ``` warning: variable does...

Nice, as I said in my previous comment tests still fail (with NaN).

I tested it again recently and it still happens for debug builds (release is fine) using `packed_simd`. Failing example extracted from testsuite: ```rust use packed_simd::f32x2; fn main() { let z...

I'm pretty much beginner at this stuff but by my understanding outstanding thing here is alignment. Direct code operates on 8-byte alignment and `packed_simd` uses 4-byte alignment. Direct IR: https://gist.github.com/mati865/7ff6c9f76284d1526d947d0f9d82130e#file-direct-ll...

> @mati865 you can use `-C debuginfo=0` to "prune" all the debug information from the IR - for some reason the Direct IR appears to contain `packed_simd` stuff. I was...

Source code: ```rust use packed_simd::f32x2; macro_rules! assert_eq_ { ($a:expr, $b:expr) => { if $a != $b { panic!() } } } fn main() { let z = f32x2::splat(0.0); let o...