Path: csiph.com!weretis.net!feeder9.news.weretis.net!panix!.POSTED.spitfire.i.gajendra.net!not-for-mail
From: cross@spitfire.i.gajendra.net (Dan Cross)
Newsgroups: comp.programming
Subject: Re: Optimization by programmers
Date: Fri, 9 Oct 2026 14:38:28 -0000 (UTC)
Organization: PANIX Public Access Internet and UNIX, NYC
Message-ID: <11aau94$7a3$1@reader2.panix.com>
References: <1177fdi$26ea4$1@dont-email.me> <11a6a6h$37i1n$1@dont-email.me> <11a91un$a7p$1@reader2.panix.com> <11aau1v$sbtb$1@dont-email.me>
Injection-Date: Fri, 9 Oct 2026 14:38:28 -0000 (UTC)
Injection-Info: reader2.panix.com; posting-host="spitfire.i.gajendra.net:166.84.136.80"; logging-data="7491"; mail-complaints-to="abuse@panix.com"
X-Newsreader: trn 4.0-test77 (Sep 1, 2010)
Originator: cross@spitfire.i.gajendra.net (Dan Cross)
Xref: csiph.com comp.programming:16935

In article <11aau1v$sbtb$1@dont-email.me>,
Terje Mathisen  <terje.mathisen@tmsw.no> wrote:
>Dan Cross wrote:
>> [snip]
>> Right. This lets you potentially go sub-linear looking for the
>> tokens you care about.  I wonder how it would stack up against
>> using something like `memchr::memmem`, though, since that'll get
>> you hardware acceleration using SIMD instructions and goofy
>> stuff like that.
>
>We're already well past that point!
>
>The current bunch of optimization freaks employs portable SIMD anywhere 
>it results in even small speedups.
>
>For the current setup, the main issue is that is so relatively small 
>that any kind of significant setup overhead mans that you lose out in 
>the end. The combined BM could be mostly precompiled instead of runtime 
>skip table geneation overhead.

Yeah, I'd think that if you wanted to use BM you'd be able to
pre-compute the skip tables by hand and just write them out as
constants at compile-time.

Sounds like the folks who are into this are deep into the weeds
already; if it's fun for you then rock on.  I wish you luck!

	- Dan C.