On Friday, 24 August 2018 at 00:32:59 UTC, Guillaume Piolat wrote:
For example: why implement AVX in DMD backend? Who are the users that will be delighted by that? Those interested in performance already use some other back-end, it's imo a completely useless development since _no one_ use D_SIMD seriously apart from compiler tests (https://github.com/search?l=D&p=4&q=D_SIMD&type=Code)
But I need it to implement `memcpy` and `memcmp` in D, so we can remove the dependency on the D standard library :-)
https://github.com/JinShil/memcpyD I know. I'm weird.
Phobos shouldn't even be a thing, we always read "hopefully this will be moved into Phobos" which is entirely wrong. People should be pushed to use the community to their advantage. SemVer is where it's at.
Totally agree. It seems, from someone without much historical perspective, that Phobos was intended to be something like the .Net Framework for D. Perhaps there are a few fundamentals (std.algorithm, std.allocator, etc.) to keep, but for the others... move 'em to Dub and let the "free market" sort it out.
Mike
