[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Pan-users] Re: Can creating task list be optimized?

From: Duncan
Subject: [Pan-users] Re: Can creating task list be optimized?
Date: Sun, 27 Aug 2006 19:11:54 +0000 (UTC)
User-agent: pan 0.109 (Beable)

"Charles Kerr" <address@hidden> posted
address@hidden, excerpted below, on  Sun, 27 Aug 2006 05:08:31

> IIRC you have a much better computer than mine, but you say klibido eats
> around half a CPU?  Would you like to run a 0.110/klibido speed test for
> me? :)

I like to be safe in my claims, so that's in the same category as the
earlier claim I was making that I expected pan 1.0 out by the end of the
year. =8^)

Standard baseline CPU is ~10% of one (of two), here.  That's with xmms
(yes, I still use xmms) streaming netradio (with level and fft display),
composite and window semi-transparency, and various per-second computer
status graphs.  Take a look at the screenshots for an idea...

IIRC actual klibido CPU use was 20-30%, IIRC, minus the baseline so
10-20%.  However, being conservative, I only claimed 50%, which still made
the point since pan was pegging 100% and judging by the lack of interactive
response would be at least doubling that if it could, to do in
(conservatively) an hour, what klibido did in seconds to minutes.

I actually haven't had klibido installed in awhile, because the existing
ebuild was stale last I looked, and while I know how to fix it, and wanted
to do so and submit the ebuild upstream (to Gentoo), it just never got
high enough on my priority list to do, for the same reason I hadn't tested
pan 0.1xx on binaries until this week -- I had other things ranking higher
on my priority list than USENET binaries, so working on a client to grab
those binaries was likewise low on my list.

I'm into work days again now, but I had a bit more time last week and if I
get it this week too, you might just get that comparison. =8^)  (As you
can tell if you check my headers, I'm still running 0.109 ATM, tho I'm
doing a system update in another window and will be merging the new pan
after that.)

BTW, my dual Opteron is only the 242, which runs 1.6 GHz, tho I'm running
two of them and have gobs of memory @ 8 gig.  Your Athlon XP 2500 looks to
be clocked at 1.83 GHz according to my quick pricewatch lookup, so it's
probably pretty close, given pan is single-threaded so confined to using
one CPU, and looked to me to be running ~70-80 MB during the time it was
pegging here, so the half-gig vs 8-gig thing shouldn't matter much either,
except under swap-storm conditions due to what else you might be running.
Of course, for disk access I'm running a 4-disk RAID-6 for the important
parts of my system, so effectively 2-way striped, which would make a
difference on disk access. 

FWIW, before I upgraded to the RAID, that or the gig of memory was the
bottleneck.  After the RAID upgrade, it was the gig memory.  After
upgrading that to 8 gig (4 would have been plenty for now but I expect
I'll be keeping this system for awhile so I went ahead with 8), its
definitely the CPUs again.  I'll be solving that with a pair of dual-core
Opti 270s or better later this year or maybe early next.  I'm hoping they
drop in price and I can do 275s or even 280s or 285s, as the 22xx series
gets stronger.  I'm figuring that'll be my last major system upgrade for 3
years, maybe more.  I'm taking a break from upgrading my system to upgrade
my eyes -- I'm looking at lasik.  I'm hoping by the next system upgrade,
therefore, to be able to go oct-core in a single socket, thereby avoiding
the $400 price-point mobo I had to do for dual Opteron.  I figure 3 years
out quad-core should be well established and oct-core coming online, so an
oct-core system at 3-5 years should be reasonable, and provide a rather
decent upgrade while putting me back in the mid-range as opposed to the
high end I'm in now, which will be low end by then.

Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman

reply via email to

[Prev in Thread] Current Thread [Next in Thread]