Linux S5R2 App 4.21 available for Beta test

Bernd Machenschalk
Bernd Machenschalk
Moderator
Administrator
Joined: 15 Oct 04
Posts: 3,915
Credit: 193,354,862
RAC: 24,157

Thanks. It looked only

Thanks. It looked only slightly slower on the machines I tested it so far. I had to use a slightly older version of gcc for this build (gcc-4.1.0 instead of gcc-4.1.2). If this App really fixes tes SIGABRT problem, this would be worth it, and it'll give me time to look into speeding up the code for the benefit of all platforms, beyond the 4.18 App.

BM

BM

Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3,515
Credit: 450,902,216
RAC: 101,364

RE: Not wanting to cry wolf

Message 63871 in response to message 63869

Quote:
Not wanting to cry wolf or sth, but atm it looks like this app is extremely slow on my box. This box used to finish 500-credit-WUs in hardly more than 100,000 seconds, which would be around 28 hours iirc (not just one large WU but at least half a dozen so far which should be halfway significant) but atm it has been crunching on a WU for more than 6 hours and it's only 10% done...

Seems like you got Kirstin's WU :-)

Seriously, is it possible that this is somehow related to the "graphics appear only once" prob? Is it possible that the graphics aren't really closed but just don't appear on screen, and are still eating CPU cycles in the backgraound? Speculation, of course. Maybe stopping boinc CC and restarting would help.

OK, I've just started the beta on a Pentium M that used to suffer from "signal 11" crashes.

http://einsteinathome.org/task/83943971

I don't have the "graphics work only once" problem, BTW. Go figure.

Estimated time to completion is in line with previous workunits.

Here's the head of the stderr output, note the warning from the "screensaver":

========================
2007-05-07 22:21:15.9710 [normal]: Built at: May 4 2007 15:43:48

2007-05-07 22:21:15.9712 [normal]: Start of BOINC application 'einstein_S5R2_4.21_i686-pc-linux-gnu'.
2007-05-07 22:21:16.7304 [debug]: Reading SFTs and setting up stacks ... done
2007-05-07 22:21:26.3764 [normal]: ERROR: Couldn't open existing checkpointing toplist file h1_0350.50_S5R2__105_S5R2c_2_0
2007-05-07 22:21:26.3766 [debug]: Couldn't open checkpoint - starting from beginning
2007-05-07 22:21:26.3766 [debug]: Total skypoints = 30191. Progress: 0, 1, 2, GLUT: Warning in screensaver: invalid glutGet parameter: 508
3, 4, 5, 6, 7, 8, 9, c
10, 11, 12, 13, 14, 15, 16, 17, 18, c
19, 20, 21, 22, 23, 24, 25, 26, 27, c
=======================

CU

BRM

Annika
Annika
Joined: 8 Aug 06
Posts: 720
Credit: 494,410
RAC: 0

Yeah, looks like Kirsten's WU

Yeah, looks like Kirsten's WU allright :-) I'll see how this turns out, but I don't think the box can still catch up and finish in a "normal" amount of time. 18% in 11 hours now, so, it's not really getting better... as for the graphics problem, I understood Bernd in a way that it was mildly annoying and harmless... i don't really think that's what's slowing me down atm. Bernd mentioned the app being a bit slower on other systems aswell, so maybe it's just my CPU or my distro or whatever causing this to affect me more badly...

Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3,515
Credit: 450,902,216
RAC: 101,364

RE: Yeah, looks like

Message 63873 in response to message 63872

Quote:
Yeah, looks like Kirsten's WU allright :-) I'll see how this turns out, but I don't think the box can still catch up and finish in a "normal" amount of time. 18% in 11 hours now, so, it's not really getting better... as for the graphics problem, I understood Bernd in a way that it was mildly annoying and harmless... i don't really think that's what's slowing me down atm. Bernd mentioned the app being a bit slower on other systems aswell, so maybe it's just my CPU or my distro or whatever causing this to affect me more badly...

Yup, but a factor of 2 and more is more than just a bit slower.

"top" is still showing only one einstein process consuming CPU significantly, I guess?

CU

BRM

Annika
Annika
Joined: 8 Aug 06
Posts: 720
Credit: 494,410
RAC: 0

As I said, maybe it affects

As I said, maybe it affects me worse... no idea why, it might be the distro, graphics, BOINC client, the CPU, maybe a combination... you know as much as I do ;-)
And yes, "top" shows exactly what one would expect (meaning one Einstein thread taking up around 98% of my CPU time).

Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3,515
Credit: 450,902,216
RAC: 101,364

RE: As I said, maybe it

Message 63875 in response to message 63874

Quote:
As I said, maybe it affects me worse... no idea why, it might be the distro, graphics, BOINC client, the CPU, maybe a combination... you know as much as I do ;-)
And yes, "top" shows exactly what one would expect (meaning one Einstein thread taking up around 98% of my CPU time).

Hmmm, so look forward to getting 1000 credits for the mythical monster WU? I have some WUs with really large predicted completion times in the pipeline, using the production app 4.18. Who knows ;-)

CU

BRM

Annika
Annika
Joined: 8 Aug 06
Posts: 720
Credit: 494,410
RAC: 0

...and a nice trophy to hang

...and a nice trophy to hang on my wall, or stick to my computer screen? :-D
Well, seriously... one of my wingmen has already finished this WU and claimed 529 credits for it, so it's a perfectly normal size compared to what I got before- and finished in much less time.

Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3,515
Credit: 450,902,216
RAC: 101,364

RE: ...and a nice trophy to

Message 63877 in response to message 63876

Quote:
...and a nice trophy to hang on my wall, or stick to my computer screen? :-D
Well, seriously... one of my wingmen has already finished this WU and claimed 529 credits for it, so it's a perfectly normal size compared to what I got before- and finished in much less time.

Hmmm...that's rather serious, performance-wise. If this went into production, we would see a lot of complaints like the one from Kirsten, don't you think? But then again, Kirsten wasn't using the beta at all, on the other hand. Very strange indeed!!!

CU

BRM

Annika
Annika
Joined: 8 Aug 06
Posts: 720
Credit: 494,410
RAC: 0

It certainly is strange, I'm

It certainly is strange, I'm surprised myself. But I can only comment on what I see here, and what I see atm is an app that is so slow we'd probably be getting much less work done even without the crashes if it affected lots of people... I would say it was a problem with my box but it never had problems before, except the two SIGABRTs, which affected more or less everyone... So, it probably is the app not liking my box. Still don't have a clue why, though, so it's of course impossible to say if it's "exotic" or sth which could potentially affect lots of users...

Melvyn Bobo Slacke
Melvyn Bobo Slacke
Joined: 22 Jan 05
Posts: 32
Credit: 1,692,164
RAC: 0

With this one 428973 I could

With this one 428973 I could trigger crashes with stopping/restarting Boinc and do Update on 4.18.
Not now though on 4.21 and suspend/run always and benchmark stoppings seems to work alright too.
4.21 seems about 8% slower

Fedora C4 with glibc 2.3.6

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.