Posts by Bikeman (Heinz-Bernd Eggenstein)

WARNING: This website is obsolete! Please follow this link to get to the new Albert@Home website!

21) Message boards : Problems and Bug Reports : S6 Directed Search (CasA) Feedback thread (Message 112586) Posted 3 Jun 2013 by Bikeman (Heinz-Bernd Eggenstein) Post: First result validated, nice! Link to workunit #371351 At a first glance on my machine the new run takes about 11000s or 3 hours longer per task to run than the LineVeto search over at Einstein but got the same amount paid. I suspect that this is going to be adjusted before we go live on Einstein, right? Testing here is also used to adjust the credits for Einstein. The variations in run time of different "kinds" of tasks (frequency, etc) is much more pronounced in this run, e.g. older Core2 CPUs show a greater "run time scatter" than modern Sandy Bridge or Ivy Bridge CPUs. We need to understand these variations first and then find a compromise credit policy that is more or less fair to everyone. Cheers HB
22) Message boards : Problems and Bug Reports : S6 Directed Search (CasA) Feedback thread (Message 112585) Posted 3 Jun 2013 by Bikeman (Heinz-Bernd Eggenstein) Post: Apropos, why are the auto-format buttons like [url] so far off-screen to the right here? I'm using Chrome on Windows XP. It's the same on IE 10 on Win 7 Claggy Hi! There will be a mojor overhaul of the Web presence in the near future, so forgive us for not investing any significant time in fixing stuff in the legacy web front-end now. Cheers HB
23) Message boards : Problems and Bug Reports : BRP4U (Raspberry Pi , single DM tasks) feedback thread (Message 112584) Posted 3 Jun 2013 by Bikeman (Heinz-Bernd Eggenstein) Post: Any idea on the reason for the Validate error yet? Is it a Android Boinc Bug, an App Bug, a Server problem, or some other reason? Only one more segmentation violation since my last report, rest of Wu's all Validate error or pending. Claggy No idea yet :-(. Since this seems to affect other projects as well, I guess it's most likely not a science app problem. I also can't think of a way how a BOinc server side problem could trigger this. Strange.. HB
24) Message boards : Wish List : App for Intel HD graphics? (Message 112579) Posted 31 May 2013 by Bikeman (Heinz-Bernd Eggenstein) Post: Hi Thanks guys, that's a plan. OK, so I put an item on my TODO list, expect something next week. Feel free to nag me if I should not deliver :-) Cheers HB
25) Message boards : Wish List : App for Intel HD graphics? (Message 112575) Posted 30 May 2013 by Bikeman (Heinz-Bernd Eggenstein) Post: Any progress on this? I have successfully crunched GPU tasks for both collatz and SETI (astropulse), on my intel HD 4000. Hi! No, unfortunately no progress yet. Our BOINC server side code is not yet able to handle Intel GPUs consistently (certainly not on Einstein and not even on Albert IMHO). We (the E@H admins and devs) have quite a lot of other stuff on our desks at the moment (to name just a few: launch of S6directed on Einstein, transition of BRP4 to a CPU only run, evaluating the early BRP5 results, a complete overhaul of the Web Homepage and forum platform, postprocessing of finished GW runs, ....). Yeah, it's a shame really, especially with the next Intel CPU generation to be launched soon (Haswell) which is said to feature some relatively powerful GPUs. Sigh. What would be possible at the moment: we provide (in the timeframe of a few days) an app to download that someone could test using an app_info.xml file that would pretend the app is a CPU app. So the Boinc client would ask for work for a CPU anonymous platform task, but the task would then in fact be calculated on the Intel GPU, so one would at least get an idea about the runtime. Is anyone who is experienced in app_info.xml writing interested in such an experiment? Cheers HB
26) Message boards : Problems and Bug Reports : BRP4U (Raspberry Pi , single DM tasks) feedback thread (Message 112573) Posted 27 May 2013 by Bikeman (Heinz-Bernd Eggenstein) Post: I've attached my HTC One S to Albert@home, not had any validations yet, eithier the app gets a SIGSEGV: segmentation violation, or if it completes, it then gets a Validate error, running the Official Android 7.0.58 Boinc, one thing to Note, my Android kernel is newer than anyone else's here, being Android 3.4.10-g8d7cd09, it has some new Power Save Options (which i have switched off, but i don't trust the kernel not to do other power save things unbided) Computer 7471 I was previously running NativeBoinc on the HTC One S, Both the Setiathome app and the PrimeGrid apps would also regularly get segmentation violations too, the GridGrid app would sometimes get app error: 9, i think this is when android kills the app to free up memory, also often one of the PrimeGrid Wu's would error if: a) the HTC switched over to battery power, b) If the Play Store downloaded and then updated any apps, it wasn't the downloading that caused the errors, more the updating, earlier when i ran the older kernel (I think it was around 3.0.x), i'm sure i only got the app error: 9 condition, Claggy Thanks for the report! I see one, maybe two other devices in the DB that might have similar issues. I'll have to investigate. Thanks for reporting this!! Cheers HB
27) Message boards : Problems and Bug Reports : S6 Directed Search (CasA) Feedback thread (Message 112569) Posted 24 May 2013 by Bikeman (Heinz-Bernd Eggenstein) Post: Just happened to open the Windows task manager and noticed that each of the three running S6 Directed Search tasks was using ~600 MB of RAM each, compared to the S6 LineVeto search over at Einstein where each task use ~110-115 MB of RAM. I'm now wondering if this is to be expected of the new run or is this temporary during testing? I don't mind as I have 16 GB of RAM but other volunteers will probably have less RAM available. One of the (many) differences between this and the previous run is that work units will vary a lot more, both in memory consumption and in the data volume that is required for a specific task. Among the currently generated test work units will be some of the most demanding in this respect, kind of a worst-case test. BOINC allows to configure limits both for RAM and disk usage by BOINC, so it's possible with some care to make sure that BOINC will not take more ressources that the user is prepared to contribute. Cheers HB
28) Message boards : Problems and Bug Reports : BRP4U (Raspberry Pi , single DM tasks) feedback thread (Message 112563) Posted 23 May 2013 by Bikeman (Heinz-Bernd Eggenstein) Post: I notice the Raspberry Pi app has updated to 1.06 now. Were there problems with 1.05 or some further improvements? There was a small bug fix, and the new app was compiled with a different toolchain (cross compiled on an Intel x86 Linux host like the Android (and Windows) app versions). Before that, it was a bit cumbersome to build the app on the Raspberry Pi itself, and it didn't integrate well into our automated continuous integration solution (Jenkins). I expect no significant performance change. Cheers HB
29) Message boards : Problems and Bug Reports : S6 Directed Search (CasA) Feedback thread (Message 112560) Posted 22 May 2013 by Bikeman (Heinz-Bernd Eggenstein) Post: Hi Thanks for the feedback. We are still fine tuning and fixing things, so expect some hickups. We are aware of the app selection problem, and we fixed some work generation problem. Please keep testing this, at this point in testing we want to see errors and problems (better now than later in production :-) ). Cheers HB
30) Message boards : Problems and Bug Reports : BRP4U (Raspberry Pi , single DM tasks) feedback thread (Message 112555) Posted 20 May 2013 by Bikeman (Heinz-Bernd Eggenstein) Post: Hi! I think it's too early for funeral eulogies wrt. Intel Atom. There are (more than) rumors that Samsung will use Atoms in some future tablet and possibly even smart phones. I guess they know what they are doing. We'll see. Cheers HB
31) Message boards : Problems and Bug Reports : S6 Directed Search (CasA) Feedback thread (Message 112554) Posted 20 May 2013 by Bikeman (Heinz-Bernd Eggenstein) Post: Both tasks finished and uploaded normally, waiting for validation now. Good!! Looking at performance the new tasks take almost twice as long as the gravity wave search currently running on Einstein, I assume this is by design. Indeed. The code is more or less the same, but the parameters of the tasks which tell the code 'where' to look for what kind of signals, is radically different. In fact, there has probably never been so much difference between a E@H run to the previous run in this respect, so we are now verifying and possibly tweaking the set-up of this run, and that's what the test here on Albert is all about. The final run parameters on Einstein might be different, but hopefully we got it about right in the first attempt already :-). Thanks for testing and feedback, Cheers HBE
32) Message boards : Problems and Bug Reports : FGRP application v 1.07 (OPENCL) feedback thread (Message 112545) Posted 19 May 2013 by Bikeman (Heinz-Bernd Eggenstein) Post: Nice!!! Thanks for the feedback. Cheers HB
33) Message boards : Problems and Bug Reports : S6 Directed Search (CasA) Feedback thread (Message 112543) Posted 19 May 2013 by Bikeman (Heinz-Bernd Eggenstein) Post: Year, sorry, there seems to be a general problem. I'm not sure we'll be able to fix this before Tuesday (Monday is a public holiday here), so for now I have "deprecated" the new S6Directed apps. Feel free to suspend or cancel any that you received alraedy. Cheers HBE
34) Message boards : Problems and Bug Reports : FGRP application v 1.07 (OPENCL) feedback thread (Message 112542) Posted 19 May 2013 by Bikeman (Heinz-Bernd Eggenstein) Post: since actual GPU use is so low (~20%) can you allow us to change the GPU utilization factor in settings? I guess eventually we'll do that, probably with a separate setting as this app consumes more RAM on the graphics card than the BRP4 (or future BRP5) tasks. So those who are able to execute (say) 4 tasks of BRP4 in parallel might only be able to run 3 of the FGRP2 tasks in parallel, hence separate settings. Cheers HB
35) Message boards : Problems and Bug Reports : FGRP application v 1.07 (OPENCL) feedback thread (Message 112538) Posted 18 May 2013 by Bikeman (Heinz-Bernd Eggenstein) Post: Interesting, thanks for additional data. We'll have to look deeper into this. Cheers HB
36) Message boards : Problems and Bug Reports : FGRP application v 1.07 (OPENCL) feedback thread (Message 112535) Posted 17 May 2013 by Bikeman (Heinz-Bernd Eggenstein) Post: I have a SIV picure that shows how it looks. I saw in post in previous version that similiar thing was found. Hmm.. that stuff looks odd. The card with the highest load according to the diagram over a timeline has a low temperature, while another GPU with only intermittent load has a very high temperature, and the second GPU doesn't have a temperature reading at all, even tho all GPUs are very similar (Cypress GPU). So I'm not quite sure which part of the data from this tool can be trusted, because there are obvious inconsistencies. Cheers HB
37) Message boards : Problems and Bug Reports : BRP4U (Raspberry Pi , single DM tasks) feedback thread (Message 112524) Posted 12 May 2013 by Bikeman (Heinz-Bernd Eggenstein) Post: It still hasn't quite finished the 1st one yet (its on 96% after 30 hours). I will provide links once it completes. Now its finished. 31 hours and 45 minutes. Link to wu is here Cool, validated against an Android (NEON) task that took a bit less than half as long to finish. Cheers HB
38) Message boards : Problems and Bug Reports : BRP4U (Raspberry Pi , single DM tasks) feedback thread (Message 112522) Posted 12 May 2013 by Bikeman (Heinz-Bernd Eggenstein) Post: This one is looking a lot faster. Normally they take 47 hours, currently they are looking around 31-32 hours. A substantial improvement. It still hasn't quite finished the 1st one yet (its on 96% after 30 hours). I will provide links once it completes. Excellent, thanks! The obvious question is: will it be possible to improve the Android app versions in a similar way? That remains to be seen(I'll need to make a wrapper app first that produces and dumps the FFTW3 'wisdom' on an Android device. Stay tuned. Cheers HBE
39) Message boards : Problems and Bug Reports : BRP4U (Raspberry Pi , single DM tasks) feedback thread (Message 112520) Posted 10 May 2013 by Bikeman (Heinz-Bernd Eggenstein) Post: OK, next attempt .... I just released version 1.05 for Raspberry Pi . If all goes well this should be 20 - 25 % faster than the previous version. Feel free to cancel task that are/will be running under version 1.04, this will help to get 1.05 results in soon. Thanks for testing!! Cheers HB
40) Message boards : Problems and Bug Reports : BRP4U (Raspberry Pi , single DM tasks) feedback thread (Message 112519) Posted 10 May 2013 by Bikeman (Heinz-Bernd Eggenstein) Post: Hi! yeah, I think you are right. Doesn't seem faster at all. Maybe I didn't turn off the FFT wisdom file when I tested this one....(including a wisdom file is another thing I'm working on and this WOULD speed up things indeed). So stay tuned. Thanks for testing Cheers HB

Previous 20 · Next 20

This material is based upon work supported by the National Science Foundation (NSF) under Grant PHY-0555655 and by the Max Planck Gesellschaft (MPG). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the investigators and do not necessarily reflect the views of the NSF or the MPG.

Copyright © 2024 Bruce Allen for the LIGO Scientific Collaboration