Re: Name hashing function causing a perf regression

From: Linus Torvalds <torvalds@linux-foundation.org>
To: Josef Bacik <jbacik@fb.com>
Cc: Andi Kleen <andi@firstfloor.org>,
	linux-fsdevel <linux-fsdevel@vger.kernel.org>,
	Al Viro <viro@zeniv.linux.org.uk>,
	Christoph Hellwig <hch@infradead.org>, Chris Mason <clm@fb.com>
Subject: Re: Name hashing function causing a perf regression
Date: Sat, 13 Sep 2014 11:58:08 -0700	[thread overview]
Message-ID: <CA+55aFwrYRQRHnqHsTJZYX4tZiCywT5JHji23iv2q26x-VAVoQ@mail.gmail.com> (raw)
In-Reply-To: <CA+55aFwTGwstMnLCqDz7G9+RoNCGDb0HAyErFuTTMenvwXWoFA@mail.gmail.com>

On Fri, Sep 12, 2014 at 3:25 PM, Linus Torvalds
<torvalds@linux-foundation.org> wrote:
>
> So with that, the kernel patch would look something like the attached.
> Totally untested.

Tested here, and it looks ok. The hash_64() implementation meant that
even on x86-64, where multiplies are generally fast, we'd do manual
shift-and-adds. Which annoyed me, although it didn't show up as a huge
problem in profiles. So I cleaned some of that up too.

I'm not going to wait for the next merge window, since we still have a
couple of weeks to go for 3.17. But I also didn't necessarily mark it
for stable, since I want to hear about how this affects performance at
fb first.

So if this all helps you guys, and after sufficient testing, just the
current top commit should be self-sufficient (with two other commits
before *if* that multiply-vs-shift makes a difference)

    99d263d4c5b2 (HEAD, master) vfs: fix bad hashing of dentries

and if it *doesn't* help, and you still see problems, please holler.

            Linus