GNU/Linux S5R3 App 4.14 available for Beta test

M. Schmitt
M. Schmitt
Joined: 27 Jun 05
Posts: 478
Credit: 15,872,262
RAC: 0

RE: You might know, right

Message 74683 in response to message 74679

Quote:

You might know, right after I'd said that I'd never had trouble with the 4.14 app and ReiserFS, two 4.14 workunits simultaneously fail on my Compaq dual-processor P-III 733 machine.

For anyone who needs to look at this, its ID number is 764145.

For the record, it's running Feisty Fawn Ubuntu from a SCSI hard drive that's formatted with ReiserFS. I don't know that it is Reiser-related, but wanted to let everyone know just-in-case.


If you google for "process got signal 11" you will find out that many projects have problems with their Linux app. I don't know why, but I already got so many crashed WUs with this error or that I wonder if the boinc client can cause such errors even so I don't believe this is possible. The client is the only major thing that changed over here. I might try out a 64-Bit boinc client I compiled myself more than a year ago.

cu,
Michael

Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3,516
Credit: 471,303,446
RAC: 52,838

Hi! That's most

Hi!

That's most interesting!

Maybe it's not the (core) client of BOINC but the boinc client API lib (whatever is the correct term) that gets compiled into the science apps.

CU
Bikeman

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

That would explain a few

That would explain a few things. I lost two WUs on my Kubuntu 7.10 laptop with the 4.12 app, next day had another one that crashed running 4.14, and a few SHA-1 WUs error out as well.
The last days everything has been running okay so I don't think it's a hardware problem (this laptop is only half a year old, and I have no reason to believe it has heat-related, hardware, or similar probs). But I was surprised, as I've never had so many client errors in a row. I was running Windows and Kubuntu 7.04 on this machine for a while and neither had problems.
I would have thought the crashed came from the beta apps, but that's not very likely if other projects react the same (I don't do beta-testing for SHA-1). A problem with the core client (or sth related, as Bikeman stated) would be a very logical explanation indeed.

tullio
tullio
Joined: 22 Jan 05
Posts: 2,081
Credit: 46,862,407
RAC: 11,497

Since updating the BOINC

Since updating the BOINC client from 5.8.17 to 5.10.21 my boincprog is stuck to a single app, whichever starts first of SETI, Einstein and QMC, so maybe the problem is in the core client.
Tullio

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

That may well be the case.

That may well be the case. Switching between apps works okay for me (I'm using 5.10.8 simply because that's what Canonical put into the repositories) but as I said, it does tend to produce odd client errors... Btw, I think I remember someone else complaining about the same problem you mentioned, Tullio, but I'm not quite sure any more who and where... might be a bit more common, though.

KSMarksPsych
KSMarksPsych
Moderator
Joined: 15 Oct 05
Posts: 2,702
Credit: 4,090,227
RAC: 0

RE: Since updating the

Message 74688 in response to message 74686

Quote:
Since updating the BOINC client from 5.8.17 to 5.10.21 my boincprog is stuck to a single app, whichever starts first of SETI, Einstein and QMC, so maybe the problem is in the core client.
Tullio

5.10.21 is switching ok for me. I suppose I could go through the change log to see what changes were made that might account for the behavior you're seeing, but my brain is too sluggish for that right now. Sorry.

Kathryn :o)

Einstein@Home Moderator

KSMarksPsych
KSMarksPsych
Moderator
Joined: 15 Oct 05
Posts: 2,702
Credit: 4,090,227
RAC: 0

Error being reported in this

Error being reported in this thread.

Result
Host

[pre]
Outcome Client error
Client state Compute error
Exit status 11 (0xb)[/pre]

Kathryn :o)

Einstein@Home Moderator

tullio
tullio
Joined: 22 Jan 05
Posts: 2,081
Credit: 46,862,407
RAC: 11,497

Not that I care much about

Not that I care much about credits since I am using a slow CPU, but my latest Einstein result, done with 4.12 app, gave me only about 2 credits/hour, while earlier apps gave me about 2.6, in agreement with SETI and QMC.
Tullio

Bernd Machenschalk
Bernd Machenschalk
Moderator
Administrator
Joined: 15 Oct 04
Posts: 4,031
Credit: 218,460,337
RAC: 44,455

RE: Not that I care much

Message 74691 in response to message 74690

Quote:
Not that I care much about credits since I am using a slow CPU, but my latest Einstein result, done with 4.12 app, gave me only about 2 credits/hour, while earlier apps gave me about 2.6, in agreement with SETI and QMC.
Tullio


We are still working on adapting the credit granting to the large variation in runtime in S5R3. Averaging over about a dozen resuts the credit/h should be similar to other projects (actually a bit more with Linux Apps since version 4.10), but for randomly picked individual results it may differ a lot from the average value.

BM

BM

Conan
Conan
Joined: 19 Jun 05
Posts: 162
Credit: 5,937,946
RAC: 0

RE: From Bernds post i take

Message 74692 in response to message 74657

Quote:

From Bernds post i take it that the code thats important for speed is the same in 4.14 as in 4.12, just a bug correction and then this new checkpointing that didnt work and was removed in 4.14 v2? So 4.14 and 4.12 performance should be the same, and since 4.12 had so short run we can continue the performance analysis for 4.12 here using 4.14.

Conan, you have some interesting results on [url=http://einsteinathome.org/host/579944/tasks]this host (Opteron 285 Fedora3), the 2 results you started with 4.09 and finished with 4.12 are quite fast. Your results arent very "lined up" because of small work cache so it will be hard to find a trend from your computers, and in addition you have too many other projecs running to get a good performance baseline for Einstein. Still, i think you had a speedup.


G'Day th3,
After my initial disappointment with the speed of 4.12 over 4.09, I kept it going and have now found that with 4.14 it seems to be getting faster with each WU.
As I am nearing the end of my current frequency run my times have dropped below 9 hours for a WU when before they were up over 12. So I am a lot happier than I was before.

Einstein team,
Application 4.14 on Linux seems to be getting better the longer it goes, in the middle of the frequency runs the run times do seem to be all over the place but steady down near the end.
My average cr/h has stopped going down and may even be slowly starting to go back up now to where it was before the new app was released (when on 4.02), so that is good news.

Comment viewing options

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