**** BEGIN LOGGING AT Tue Sep 12 02:59:57 2006
Sep 12 04:36:59 <joshin> Anyone around.  I have some interesting results of comparing big and little endian network performance...
Sep 12 04:39:25 <jacques> oh?
Sep 12 04:40:38 <joshin> I have a pair of slugs.  One running the current Openslug image used for debianslug 3.10 (little endian), the other running a fresh build of openslug (big endian)
Sep 12 04:41:26 <joshin> doing a wget of an 871MB file from my server to each slug seperately, the little endian slug took 2m 22.5s and the big endian slug took 1m 41s
Sep 12 04:41:43 <joshin> little endian slug is running 2.6.16, big endian slug is running 2.6.17
Sep 12 04:41:54 <joshin> neither are running any processes other than dropbear
Sep 12 04:42:11 <joshin> That's a heck of a difference 142.5s vs 101s
Sep 12 04:42:32 <joshin> Am building a new debianslug image right now for a more apples to apples test.
Sep 12 04:43:23 <jacques> wow that's a huge difference
Sep 12 04:43:52 <EvilDevil> joshin, i knew that's an difference between LE and BE, but i didn't thought that the impact is that big
Sep 12 04:44:09 <joshin> I have a bunch more tests to do but yeah, that is a huge difference.
Sep 12 04:44:47 <joshin> About to run it again after rebooting the little endian slug.
Sep 12 04:51:55 <joshin> Round 2 (same test):  LE-138s  BE-102s
Sep 12 04:52:15 <joshin> Anyhow, bedtime for me.  More tomorrow...
Sep 12 04:53:34 <jacques> that's a pretty good argument for running in BE mode :-)
Sep 12 14:12:44 <joshin> Ok, installed a freshly made debianslug image.  Still around 140s for the LE version and 100s for the BE version.
Sep 12 14:12:58 <joshin> Both machines are now 2.6.17
Sep 12 14:17:48 <NAiL> very nice
Sep 12 14:18:22 * NAiL just switched from a LE gw to a BE gw
Sep 12 14:18:33 <NAiL> hours before you discovered the timing differences
Sep 12 14:20:54 <joshin> Well, my second slug arrived yesterday so I had a real easy way to compare.  Looking for other benchmarks to try...
Sep 12 14:21:09 <blaster8> ? gw
Sep 12 14:22:25 <joshin> gateway
Sep 12 14:22:29 <joshin> firewall
Sep 12 14:22:32 <NAiL> yeah
Sep 12 14:32:53 <joshin> NAiL, any suggestions as to tests to try?
Sep 12 14:36:08 <NAiL> hmm
Sep 12 14:36:20 <NAiL> smbclient on the slug
Sep 12 14:36:30 <NAiL> transfer files to /dev/null
Sep 12 14:37:59 <joshin> NAiL, unforturnately that would hit the network interface.  (My first test was doing scp pulls.  Boy did that tax the CPU.)
Sep 12 14:38:07 <joshin> Actually, I could try that with localhost...
Sep 12 15:10:29 <NAiL> is hitting the network interface a "Bad Thing"? Isn't network performance what you're trying to measure?
Sep 12 15:12:05 <joshin> I'm trying to compare all aspects of performance.  Some people may not care about network speed.  But yeah, network performance is a huge thing to me.
Sep 12 15:12:26 <NAiL> aha
Sep 12 17:33:32 <kitno455> re all
Sep 12 17:38:55 <kitno455> having a problem with a few week old checkout of openslug HEAD- 
Sep 12 17:39:57 <kitno455> udevsynthesize does not seem to work
Sep 12 17:40:31 <kitno455> during bootup, any flash drive that is plugged in will not run udev rules, so will not mount, etc
Sep 12 17:40:52 <kitno455> so after bootup, i have to unplug/plug the drive to get it mounted
Sep 12 17:40:55 <kitno455> any ideas?
Sep 12 17:46:17 <NAiL> if it's a couple of weeks old, I think there's been udev updates since then
Sep 12 17:46:54 <koen> NAiL: did you see joshin's net benchmarks? 
Sep 12 17:47:10 <NAiL> more or less yes
Sep 12 17:47:11 <koen> NAiL: I wonder how those will look with the new ipx-eth driver
Sep 12 17:47:13 <joshin> koen, he did.  We were talking about it earlier.
Sep 12 17:47:52 <NAiL> koen: yeah, that's gonna be interesting, seeing the lack of quality in Intel's code
Sep 12 17:49:34 <joshin> Ooh, when might that be available for testing?  :) 
Sep 12 17:50:13 <NAiL> by the looks of it, within a couple of days
Sep 12 17:50:31 <koen> isn't it in svn already?
Sep 12 17:50:34 <NAiL> blaster8 is working on it 
Sep 12 17:50:45 <blaster8> allo
Sep 12 17:50:46 <NAiL> yeah, but I dunno if it'll compile in OE yet
Sep 12 17:50:51 <blaster8> it's compiling
Sep 12 17:51:06 <blaster8> I managed to compile it with a stale defconfig, but that's fixed now
Sep 12 17:51:12 <koen> I still don't get where when to use svn with *slug*
Sep 12 17:51:24 <blaster8> ?
Sep 12 17:51:36 <koen> A while back I assumed development happened monotone, but then it didn't
Sep 12 17:51:46 <NAiL> development still happens in monotone
Sep 12 17:51:49 <blaster8> slugos releases are dropped into svn
Sep 12 17:52:07 <blaster8> then minor fixes go on there
Sep 12 17:52:11 <NAiL> the kernel development was moved to svn for some reason I can't recall
Sep 12 17:52:15 <koen> blaster8: a while back a lot of bugfixes were in svn but not in mtn
Sep 12 17:52:34 <blaster8> mm
Sep 12 17:52:40 <blaster8> I think that's been sorted
Sep 12 17:52:41 <koen> haven't built it since
Sep 12 17:53:01 <koen> I should revive my loft soon
Sep 12 17:53:13 <blaster8> [g2] still has to refresh his patches
Sep 12 17:53:34 <blaster8> http://trac.nslu2-linux.org/kernel/query?status=new&status=assigned&status=reopened&group=owner&milestone=2.6.18&order=priority
Sep 12 17:54:40 <NAiL> koen: if there's any fixes in svn that should be in mtn, I'm the one who should be to blame :P
Sep 12 17:55:00 <NAiL> most of the fixes go into mtn as well
Sep 12 17:55:14 <NAiL> but some are specific to a release and won't work in mtn
Sep 12 17:57:59 <NAiL> blaster8: mind CC'ing me on the zd1211rw discussion?
Sep 12 17:58:08 <blaster8> ok
Sep 12 18:01:00 <NAiL> thanks
Sep 12 18:01:15 <blaster8> he's thought of something :)
Sep 12 18:01:26 <NAiL> I saw that
Sep 12 18:01:40 <NAiL> As you might've noticed, he CC'd me on the mail you replied to
Sep 12 18:02:01 <blaster8> oh, I didn't spot that
Sep 12 18:02:09 <blaster8> sorry, I'll cc from now on
Sep 12 18:02:27 <NAiL> nothing to apologize for ;)
Sep 12 18:08:19 <blaster8> can I publish a preliminary openslug image somewhere?
Sep 12 18:09:37 <blaster8> testers need to have a sense of humour and a serial connection
Sep 12 18:09:38 <kitno455> guys- does redboot have a copy of the NPE firmware in it?
Sep 12 18:09:42 <blaster8> yes
Sep 12 18:09:48 <blaster8> but it is as old as the hills
Sep 12 18:09:53 <blaster8> roughly
Sep 12 18:10:02 <NAiL> blaster8: publish, no. It requires the click-through licence
Sep 12 18:10:03 <kitno455> same size?
Sep 12 18:10:20 <blaster8> hmm
Sep 12 18:10:21 <NAiL> kitno455: not same format I think
Sep 12 18:10:33 <NAiL> there's two formats used for the NPE firmware
Sep 12 18:10:38 <kitno455> the hardware has not changed :)
Sep 12 18:10:41 <kitno455> weird
Sep 12 18:10:43 <NAiL> redboot uses the old one
Sep 12 18:11:00 <NAiL> the open driver uses the new (but it was discussed that it could be adapted to use both)
Sep 12 18:11:08 <NAiL> gtg
Sep 12 18:11:10 <NAiL> later all
Sep 12 18:11:18 <kitno455> i am glad to see developement of the open driver
Sep 12 18:11:26 <blaster8> good news, the experimental kernel compiles in OE :)
Sep 12 18:11:31 <kitno455> access lib is monstrous
Sep 12 18:11:45 <kitno455> blaster8- is this in HEAD now?
Sep 12 18:11:46 <blaster8> indeed
Sep 12 18:11:50 <blaster8> not yet
Sep 12 18:11:58 <blaster8> you'll need a serial port for now
Sep 12 18:12:10 <kitno455> dude- you must be new around here :)
Sep 12 18:12:31 <kitno455> i've got serial :)
Sep 12 18:12:38 <blaster8> well, I was around last year
Sep 12 18:12:43 <blaster8> then Uni term got in the way
Sep 12 18:12:49 <blaster8> good good
Sep 12 18:12:50 <kitno455> thanks for working on this. 
Sep 12 18:12:54 <blaster8> you keen to test?
Sep 12 18:13:06 <kitno455> yeah- 
Sep 12 18:13:09 <kitno455> i can help
Sep 12 18:13:13 <blaster8> now?
Sep 12 18:13:18 <kitno455> tomorrow?
Sep 12 18:13:21 <blaster8> better
Sep 12 18:13:33 <blaster8> I need to do some more work here, preferably
Sep 12 18:13:50 <kitno455> i am actually USING slugos for real work, so i cant drop everything and move forward all that often
Sep 12 18:14:25 <blaster8> aah
Sep 12 18:14:30 <kitno455> but i could get a second tree and come up to your revision there
Sep 12 18:14:40 <kitno455> might take a few hours :)
Sep 12 18:14:47 <blaster8> This is very very experimental
Sep 12 18:15:02 <kitno455> no problem. i was the only guy running ucslugc for quite awhile
Sep 12 18:15:10 <blaster8> hehe
Sep 12 18:15:10 <kitno455> not bothered by exp.
Sep 12 18:15:14 <blaster8> me too :)
Sep 12 18:15:18 <blaster8> do you run ucslugc now?
Sep 12 18:15:34 <kitno455> no, dropped it a few weeks back when we jumped to gcc 411
Sep 12 18:15:39 <blaster8> ok
Sep 12 18:15:57 <kitno455> but supposedly it works again
Sep 12 18:16:03 <kitno455> i need to try it.
Sep 12 18:16:04 <blaster8> I'm probably going to do some work on it in the next week or so
Sep 12 18:16:42 <kitno455> i am not much of a kernel hacker, just a basic level C guy, so once the compiler needs work, i give up :)
Sep 12 18:17:11 <kitno455> seems like big endian arm is the bastard stepchild of gcc
Sep 12 18:17:33 <blaster8> hmm
Sep 12 18:18:23 <blaster8> add in thumb instructions and ...
Sep 12 18:18:46 <kitno455> ?
Sep 12 18:18:50 <blaster8> hell
Sep 12 18:19:03 <blaster8> but well with low RAM usage :)
Sep 12 18:19:33 <kitno455> actually, uclibc is most of the savings
Sep 12 18:19:47 <kitno455> i find that the average binary does not shink all that much
Sep 12 18:19:58 <koen> I'm trying to get uclibc/eabi to work
Sep 12 18:20:12 <koen> getting that to work on armeb shouldn't be too hard after that
Sep 12 18:20:13 <blaster8> thanks
Sep 12 18:20:23 <kitno455> uclibc is braindead in enough places that i got anoyed
Sep 12 18:20:49 <blaster8> I think they have suffered from fork issues
Sep 12 18:21:07 <kitno455> who hasn't (ask my wife :)
Sep 12 18:21:21 <kitno455> oh- FORK, sorry...
Sep 12 18:21:23 <kitno455>  :)
Sep 12 18:21:25 <blaster8> yes
Sep 12 18:21:27 <blaster8> indeed
Sep 12 18:21:29 <blaster8> ;)
Sep 12 18:21:29 <kitno455> kidding
Sep 12 18:21:34 <blaster8> :p
Sep 12 18:21:57 <kitno455> i think arm instructionset with uclibc is the sweet spot, if...
Sep 12 18:22:02 <blaster8> it works?
Sep 12 18:22:11 <kitno455> not last time i checked
Sep 12 18:22:16 <blaster8> indeed
Sep 12 18:22:44 <kitno455> you get a noticable flash size savings, but there are enough differences with glibc that some programs act funny
Sep 12 18:23:31 <blaster8> ok, I have a kernel and modules from OE
Sep 12 18:23:37 <blaster8> now to fix userspace...
Sep 12 18:33:38 <blaster8> ok, now we have a rootfs
Sep 12 18:52:56 <kitno455> blaster8, status?
Sep 12 18:53:16 <blaster8> I need someone to make a bitbake package
Sep 12 18:53:24 <kitno455> for?
Sep 12 18:53:47 <blaster8> https://trac.nslu2-linux.org/slugos/ticket/20
Sep 12 18:54:25 <kitno455> trac does not work for me- get a weird message
Sep 12 18:54:49 <blaster8> http://trac.nslu2-linux.org/slugos/ticket/20
Sep 12 18:56:07 <blaster8> ok, enough
Sep 12 18:56:27 <kitno455> enough?
Sep 12 18:56:32 <blaster8> going off for tonight - NAiL: if you see the ticket can you try and get that done before tomorrow?
Sep 12 18:56:54 <kitno455> adioa
Sep 12 18:56:56 <kitno455> adios
Sep 12 18:57:01 <kitno455> be too
Sep 12 18:57:02 <blaster8> I've got 2.6.18-rc6 building with both GCC 3.4 and 4.1, which is quite a big step
Sep 12 18:57:04 <blaster8> cya
Sep 12 18:57:09 <kitno455> cool- thanks
Sep 12 22:18:42 <joshin> Wow, ucslugc actually built...  :)
Sep 12 22:19:14 <NAiL> yeah
Sep 12 22:19:22 <joshin> Thanks!  Time to ditch my LE firmware and install it instead.
Sep 12 22:19:25 <NAiL> IIRC mwester fixed missing stuff
Sep 12 22:19:37 <NAiL> lots of stuff going on lately ;)
Sep 12 22:19:44 <joshin> I've noticed!  
Sep 12 22:23:29 <NAiL> which is very good :)
Sep 12 22:52:42 <NAiL> joshin: btw, regarding network performance:
Sep 12 22:52:42 <NAiL> http://www.openembedded.org/viewmtn/revision.psp?id=83d0e48aed0ad37c789536151c11fe1ee96e68c6
Sep 12 22:53:05 <NAiL> mickeyl added "bing" to the OE repo earlier today
Sep 12 22:56:29 <joshin> Neat.  I'll check it out.  I'm still poking at libtorrent and rtorrent (when the kids are behaving).
Sep 12 23:01:29 <joshin> Looks like ucslugc is hard coded at gcc 3.4.4.  I'm not surprised that 4.1.x is causing issues...
Sep 12 23:10:06 <NAiL> I've got no idea actually. uCslugC is basically uncharted territory for me
**** ENDING LOGGING AT Wed Sep 13 02:59:56 2006