Reiner Pope

Here's a possible way. Given this: ``` #[derive(ZeroableInOption)] #[repr(transparent)] struct Foo(Bar); ``` the proc-macro should check that `#[repr(transparent)]` is present, and if so generate the following code: ``` unsafe impl...

I'd also request the `portable_simd` types when compiling under nightly: `impl Pod for std::simd::Simd where T: Pod`. Something for `std::simd::Mask` too, but the requirements there are more subtle, and I'm...

> We should add an unsafe shuffle1_dyn_unchecked API that has undefined behavior if the index is out-of-bounds and not 0x80. > We should change shuffle1_dyn to panic! if any index...

I guess another option would be to pick an encoding for lengths that always produces valid utf8. E.g. use a length encoding that never sets the top bit of any...

For what it's worth, const fn getters (with `&self` argument) are already possible in stable Rust: https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=6d129f311dcfce598be4446a298818fe. I'd find const fn getters useful, even if const fn setters were unavailable....

Glad you like it! Here's what we use: ``` - jax==0.4.26 - jaxlib==0.4.26+cuda12.cudnn89 ``` Hope that helps?

I tried that. It works if the VCS cache is disabled, because that uses a `git clone` command that does in fact fetch tags. However, if I enable the VCS...

Thanks!

> Haha, thank you! > > I like this idea, and I think the approach is using `jax.debug.inspect_array_sharding` is probably the correct one. > > Regarding the syntax, I don't...

Given the discussion of `PartitionSpec`, I realized that `PartitionSpec` is a little more flexible than we have discussed so far. Quoting the docs: a [PartitionSpec](https://jax.readthedocs.io/en/latest/jax.sharding.html#jax.sharding.PartitionSpec) is a tuple, whose elements...