Hadrien G.

That... was a different bug! :face_with_head_bandage: Will teach me to write descriptions in my tickets... The bug which I am actually observing is that when running "make/ninja xbenchmark" on my...

Since you know the length at compile time, shouldn't you take an std::array instead of an initializer_list? Seems to me that it would avoid brace initialization confusion with the constructor...

# Constructors to be tested ## batch_bool - Default constructor - Explicit broadcasting constructor from a bool - Constructor from a list of bools - Converting to inner type +...

After some experimenting, I realized that this would also require a sizeable fraction of batch_bool to be constexpr, which is not straightforward to do in the current design (which e.g....

Here is a version annotated with my reading notes: ```c++ // === GENERAL IMPLEMENTATION W/ RUN-TIME MASK === // Generates a batch_bool bitmask for the run-time version of select() template...

That looks pretty nice! One question which I would have is, why do Intel propose multiple versions of the blend intrinsic for different data types (_ps, _pd, _epi32)? Is it...

Another question, considering the construction of immediates. The current design is quite specialized for blend/select use cases. Do we expect to have other use for boolean mask immediates in the...

Since the `preferred_simd_width` part requires compiler support, I brought it up here: https://internals.rust-lang.org/t/querying-the-preferred-simd-width/15829 .

This does looks cleaner and more extensible in the future, at the cost of possibly delaying the time by which this is usable on stable (since this API requires ability...

I tried to use it in various ways, but did not quite get it to work. Unfortunately, I did not take notes at the time, and the GH Actions environment...