From: Willy TARREAU <willy@w.ods.org>
To: "David S. Miller" <davem@redhat.com>
Cc: scrosby@cs.rice.edu, alexander.riesen@synopsys.COM,
linux-kernel@vger.kernel.org, marcelo@conectiva.com.br
Subject: Re: Algoritmic Complexity Attacks and 2.4.20 the dcache code
Date: Sat, 31 May 2003 10:02:05 +0200 [thread overview]
Message-ID: <20030531080205.GA776@pcw.home.local> (raw)
In-Reply-To: <20030530.231813.59666274.davem@redhat.com>
Hi David,
On Fri, May 30, 2003 at 11:18:13PM -0700, David S. Miller wrote:
> It turns out to not be the case at all. There is too much work
> involved in the main loop just maintaining the 3-word + curbyte
> state.
>
> It needs to be optimized a bit.
With this simple change, jhash_mix is *exactly* three times faster for me
on athlon-xp, whatever gcc I use (2.95.3 or 3.2.3), on the following do_hash()
function, and about 40% faster when used on local variables. I think that
gcc is not always smart enough to avoid assignments, and consumes memory
bandwidth and perhaps prevents better optimizations.
void do_hash(unsigned *a, unsigned *b, unsigned *c) {
__jhash_mix(*a, *b, *c);
}
This function is 189 bytes long, and takes about 72 cycles to complete with the
original macro, and is now 130 bytes long for about 24 cycles, which means about
1.5 operation/cycle... not bad :-)
I've just tested on other architectures, it's even more interesting :
- On a sparc ultra2/400, 100 million hashes drop from 38.8 seconds to 8.28 s (4.68x)
- And the best win : on Alpha EV6/466, it goes from 51.5 secs to 5.75 s (8.96x) !!!
I've checked that the results are the same on every arch, before and after the
modification.
I believe it's trivial enough to go into 2.4.21, don't you think ?
Regards,
Willy
--- linux-2.4/include/linux/jhash.h Sat May 10 11:36:02 2003
+++ /tmp/jhash.h Sat May 31 09:38:17 2003
@@ -23,15 +23,15 @@
/* NOTE: Arguments are modified. */
#define __jhash_mix(a, b, c) \
{ \
- a -= b; a -= c; a ^= (c>>13); \
- b -= c; b -= a; b ^= (a<<8); \
- c -= a; c -= b; c ^= (b>>13); \
- a -= b; a -= c; a ^= (c>>12); \
- b -= c; b -= a; b ^= (a<<16); \
- c -= a; c -= b; c ^= (b>>5); \
- a -= b; a -= c; a ^= (c>>3); \
- b -= c; b -= a; b ^= (a<<10); \
- c -= a; c -= b; c ^= (b>>15); \
+ a = (a - b - c) ^ (c>>13); \
+ b = (b - c - a) ^ (a<<8); \
+ c = (c - a - b) ^ (b>>13); \
+ a = (a - b - c) ^ (c>>12); \
+ b = (b - c - a) ^ (a<<16); \
+ c = (c - a - b) ^ (b>>5); \
+ a = (a - b - c) ^ (c>>3); \
+ b = (b - c - a) ^ (a<<10); \
+ c = (c - a - b) ^ (b>>15); \
}
/* The golden ration: an arbitrary value */
next prev parent reply other threads:[~2003-05-31 7:49 UTC|newest]
Thread overview: 29+ messages / expand[flat|nested] mbox.gz Atom feed top
2003-05-29 20:42 Algoritmic Complexity Attacks and 2.4.20 the dcache code Scott A Crosby
2003-05-30 3:57 ` David S. Miller
2003-05-30 4:29 ` Ingo Molnar
2003-05-30 18:16 ` Timothy Miller
2003-05-30 18:53 ` Scott A Crosby
2003-05-30 5:04 ` Scott A Crosby
2003-05-30 6:24 ` David S. Miller
2003-05-30 6:46 ` Scott A Crosby
2003-05-30 6:56 ` David S. Miller
2003-05-30 8:59 ` Alex Riesen
2003-05-30 9:00 ` David S. Miller
2003-05-30 15:05 ` Scott A Crosby
2003-05-31 6:18 ` David S. Miller
2003-05-31 8:02 ` Willy TARREAU [this message]
2003-05-31 8:12 ` David S. Miller
2003-05-31 8:56 ` Willy Tarreau
2003-05-31 8:58 ` David S. Miller
2003-05-31 8:58 ` David Schwartz
2003-05-31 9:01 ` David S. Miller
2003-05-31 6:30 ` William Lee Irwin III
2003-05-31 6:33 ` David S. Miller
2003-05-31 6:41 ` William Lee Irwin III
2003-05-31 6:45 ` David S. Miller
2003-05-31 18:40 ` Aaron Lehmann
2003-05-30 4:02 ` Ingo Molnar
2003-05-30 4:42 ` Scott A Crosby
2003-05-30 5:01 ` David S. Miller
2003-05-30 13:48 ` Nikita Danilov
2003-06-01 1:15 ` Daniel Phillips
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20030531080205.GA776@pcw.home.local \
--to=willy@w.ods.org \
--cc=alexander.riesen@synopsys.COM \
--cc=davem@redhat.com \
--cc=linux-kernel@vger.kernel.org \
--cc=marcelo@conectiva.com.br \
--cc=scrosby@cs.rice.edu \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox