Issue #16487 has been updated by alanwu (Alan Wu).


I think we can sidestep some maintainability problems by being liberal abou=
t ripping out SIMD code if we need to make changes.
Some methods basically never change and I think we might be able to write o=
nce and forget for those.
Having multiple implementations leaves room for bugs, though.

We also need people with enough SIMD chop to review and merge the initial i=
mplementation. It seems like there are some footguns
when it comes to SIMD. For examples, a long time ago Linus Torvolds said us=
ing SIMD for `memcpy` might make multi thread programs slower
due to FPU use [1] and using AVX-512 can cause the CPU to throttle[2].



[1]: https://yarchive.net/comp/linux/page_zeroing_strategy.html
[2]: https://blog.cloudflare.com/on-the-dangers-of-intels-frequency-scaling/

----------------------------------------
Misc #16487: Potential for SIMD usage in ruby-core
https://bugs.ruby-lang.org/issues/16487#change-83895

* Author: byroot (Jean Boussier)
* Status: Open
* Priority: Normal
* Assignee: =

----------------------------------------
### Context

There are several ruby core methods that could be optimized with the use of=
 SIMD instructions.

I experimented a bit on `coderange_scan` https://github.com/Shopify/ruby/pu=
ll/2, and Pavel Rosick=FD experimented on `String#strip` https://github.com=
/ruby/ruby/pull/2815.

### Problem

The downside of SIMD instructions is that they are not universally availabl=
e.
So it means maintaining several versions of the same code, and switching th=
em either statically or dynamically.

And since most Ruby users use precompiled binaries from repositories and su=
ch, it would need to be dynamic if we want most users to benefit from it.

So it's not exactly "free speed", as it means a complexified codebase.

### Question

So the question is to know wether ruby-core is open to patches using SIMD i=
nstructions ? And if so under which conditions.

cc @shyouhei





-- =

https://bugs.ruby-lang.org/

Unsubscribe: <mailto:ruby-core-request / ruby-lang.org?subject=3Dunsubscribe>
<http://lists.ruby-lang.org/cgi-bin/mailman/options/ruby-core>