Skip to contentSkip to conversation

Benchmarks for performance patches (-ck) for 2.4.19

DISCUSSION Con Kolivas 2 participants 2 messages Sep 2, 2002 – Sep 4, 2002

lore
Share & tools
lore
#0 lore raw

Quoting J Sloan <[email protected]>:

Quoted 10 lines

The ck5 kernel comes in more than one flavor

rmap, aa and low latency are offered -

I'm running 2.4.19-ck5-ll and I'm happy to report that it also plays well with freeswan ipsec - I'm running freeswan 1.98b and it's all good - not to mention the nice smooth feel in multimedia and 3D fps shooters ;-)

Quoted 8 lines

Thanks to conman for the tedious patching work!

Joe

Quoted 2 lines · Paul

ps. ck4 uses the aa vm, but if there was a version with rmap, I would test it

Actually Paul was correct; the low latency branch was just -aa but in my never ending descent into madness I have decided to add another branch as requested, ck5-rl which has lowest latency with -rmap instead of -aa

Get it at the usual place:
http://kernel.kolivas.net

Con.

P.S. You're welcome Joe.

#1 lore raw
Quoted 11 lines · Con Kolivas <[email protected]>, on Mon Sep 02, 2002 [07:12:07 PM] said:

Actually Paul was correct; the low latency branch was just -aa but in my never ending descent into madness I have decided to add another branch as requested, ck5-rl which has lowest latency with -rmap instead of -aa

Get it at the usual place:
http://kernel.kolivas.net

Con.

P.S. You're welcome Joe.

	Hi;

	pII 400x2 512M SMP
	Running ck5-rl now. Benno's latencytest shows that

schedualing latency is perhaps better than the patchset with the aa vm. My first run yeilded a max 2.7ms latency-- oddly during the /proc loading test, instead of disk write which seems most problematic here. Every other stress test yeilded 1.6ms max latency. A second run managed to pull off 1.5ms max latency for most, and a 1.7ms latency for disk write.
Under extreme stress (make -j4 bzImage) while running
the tests, both aa and rmap performed about the same. (3.5ms max lat.)
Under more general benchmarks, like lmbench, bonnie,
hdparm, and a kernel build, rmap came out the same or just a little worse than aa. (in the context of these patchsets)
Andrew Morton's low latency patch seems to be the
most signifigant aspect in acheiving these low schedualing latencies-- turning it off in /proc/sys/kernel/lowlatency results in readings, which are better than 2.4.19 virgin, but still unworkable for applications which require consistant low schedualing latency.
My focus on these tests is related to my (diletante)
interest in audio processing and recording. I think the interactive 'feel' of the system is improved, but I didnt really have trouble with vanilla-- on this system, IO is my real limiter. I am curious how these patchsets affect games, but I dont have much gaming experience...
If anyone wants me to run any benchmarks, just
ask. Thanks, Con. (and of course, the authors of the patchs)

Paul
[email protected]