memset_benchmark VS nbdkit

mem* do not need to be called via ifunc. That is a toolchain decision. See e.g. https://github.com/nadavrot/memset_benchmark for recent data about the cost of PLT indirection for small copies.

Probably poorly. It is a violation to cast an unaligned pointer to an aligned type. And the code looks like it does just that right here: https://github.com/nadavrot/memset_benchmark/blob/main/src/l...

This is undefined behavior under C99 §6.3.2.3 Paragraph 7.

I was impressed by the notion until I saw the code...

This is correct, but doesn't quite explain why. It's because when you accept contributions from a variety of authors, without using a CLA, then your code base ends up with a patchwork of copyright, making relicensing practically impossible as you have to get buy-in from every author or else determine that author's contributions and remove/rewrite them.

GPL/LGPL are excellent licenses, but this patchwork of copyright can apply for any license you use. For a small project we wrote which was under BSD, we recently had to make a small (non-functional) change to the license, and we got buy-in from all the authors to do this which took quite a long time: https://gitlab.com/nbdkit/nbdkit/-/commit/952ffe0fc7685ea775...

That second link is wrong, should be: https://gitlab.com/nbdkit/nbdkit/-/commit/a956e2e75d6c88eeef...

I think an exception might be made for a plain "C-like" struct that doesn't initialize members or contain anything except basic types. In the specific example[0] the code is actually surrounded by extern "C" { ... } so I suppose that the compiler "knows" this is a plain C struct?

[0] https://gitlab.com/nbdkit/nbdkit/-/blob/cd761c9bf770b23f678f...

We relicensed[1] a project which had 10 contributors, and we got every single one of them to do an Acked-by (by email) which took some weeks. That was the advice from our lawyers. Can't imagine the impossible hassle of doing the same for something like Linux.

[1] https://gitlab.com/nbdkit/nbdkit/-/commit/952ffe0fc7685ea775...