Posts by Jeroen

WARNING: This website is obsolete!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

1) Message boards : Problems and Bug Reports : S6 Directed Search (CasA) Feedback thread (Message 112588) Posted 5 Jun 2013 by Jeroen Post: Hello, I have had four of the new GW search tasks validate so far. Runtime ranges from 7908 seconds average on one host to 8886 seconds average on another host. Currently memory usage per task in Linux is approximately 100 MB. The memory usage was at approximately 130 MB yesterday evening when I first started running the new tasks. Jeroen
2) Message boards : Problems and Bug Reports : FGRP application v 1.07 (OPENCL) feedback thread (Message 112534) Posted 17 May 2013 by Jeroen Post: I ran 81 tasks in Linux via an AMD 7970. 37 tasks have validated so far and the remaining are pending validation. Runtime: ~610 seconds GPU Load: 58-60% Via the same system, these tasks are running 4-5 times faster on the GPU and CPU combined compared to the CPU alone.
3) Message boards : Problems and Bug Reports : FGRP application v 1.07 (OPENCL) feedback thread (Message 112531) Posted 16 May 2013 by Jeroen Post: Hi Thanks for the feedback. This is surprising, we didn't change anything that should have a significant effect on performance. We made the logging less verbose which might help performance a tiny bit, but not more. Some volunteers might see performance increases or even decreases because the tasks are now actually running on the GPU intended by BOINC for it and not the one picked erroneously by the app before (which might be busy with other tasks already), but I understand your host here at Albert has only one GPU installed, so this should not happen in your case. Hello, I am not too sure on the difference. A single task used to have a runtime of around 1,100 seconds when I tried previously and now has a runtime of around 800 seconds. I will take a look at my system configuration to see if anything may have changed there since when I ran the previous version. Note however that setting the nr of concurrent GPU jobs in the profile currently has NO effect on the FGRP app, we didn't enable this feature for the Fermi search, yet. So one would need a app_info.xml or app_config.xml file for the FGRP app. I'm wondering whether BOINC is running BRP4 jobs and FGRP jobs in parallel in a "mixed" configuration (e.g. FGRP needs 1 GPU, BRP4 0.333 GPUs. Will BOINC let all BRP4 tasks finish and hold of running new ones before letting FGRP crunch?) Cheers HB I had an app_config.xml file setup for a short while as I wanted to see how well the new application would scale on a single GPU. However, since then I removed the app_config.xml file. Thanks for the work done on porting the application to GPU.
4) Message boards : Problems and Bug Reports : FGRP application v 1.07 (OPENCL) feedback thread (Message 112510) Posted 6 May 2013 by Jeroen Post: I have been running the new version for the past two days. I am actually seeing significantly better performance compared to the previous version I ran a while back. In Linux via a GTX 680 and quad core processor - HT disabled, I have seen the following runtimes: 1-task - ~796 seconds per task 3-tasks - 885-963 seconds per task 36 tasks have validated so far. Unfortunately, I am not able to retrieve GPU load in Linux. I suspect the load is fairly low as the GPU temperature is around 38-41C with 80% fan speed. If I had more CPU cores and cache memory, I think I would be able to run additional tasks at once via a single GPU without significant performance drop per task. The tasks that errored out happened because I accidentally overwrote my 64-bit BOINC with 32-bit BOINC and did not have the 32-bit OpenCL libraries installed.
5) Message boards : Problems and Bug Reports : FGRP application v 1.05 (OPENCL) feedback thread (Message 112406) Posted 5 Apr 2013 by Jeroen Post: I have had 1617 tasks validated and 207 tasks pending in the last week or so via one of my Linux systems with a NVIDIA card. There was one task that failed to validate yesterday but otherwise, the new OpenCL application appears to be running stable overall with Linux/NVIDIA.
6) Message boards : Problems and Bug Reports : FGRP application v 1.05 (OPENCL) feedback thread (Message 112397) Posted 27 Mar 2013 by Jeroen Post: I am very glad to see that FGRP search has been ported to GPUs. Thanks for developing the GPU applications for FGRP search. Here are some results I have seen on different hardware and OS. Tasks running per GPU: 1 OS, Driver, Card, CPU Frequency, GPU Usage, Runtime, PCI-E Linux, 12.11 Beta 11, 7970, 3930K 4.4 GHz, 55%, 955 sec, x16 3.0 Linux, 295.33, GTX 580, 920 4.2 GHz, NA, 1180 sec, x16 2.0 Linux, 295.33, GTX 580, 920 4.2 GHz, NA, 1280 sec, x8 2.0 XP 64, 314.07, GTX 680, 920 4.2 GHz, 0-30%, 1065-1157 sec, x16 2.0 FGRP search CPU only (via Einstein): 3930K 4.4 GHz No HT, 2850 sec So far there have been no failed tasks due to the new application. I did have one task error out but that was due to something missing in my Linux image. The FGRP tasks take around 35% longer than the BRP4 tasks to complete but complete much quicker than FGRP search on CPUs alone. I did not have GPU usage via Linux NVIDIA as nvidia-smi does not report usage for Geforce cards.
7) Message boards : Problems and Bug Reports : BRP application v 1.33 feedback thread (Message 112321) Posted 11 Jan 2013 by Jeroen Post: I have 1.33 running on one host so far. So far 9 tasks have completed and 3 tasks have validated. The other 6 are pending validation. The file size reduction is very significant from 2MB to 475K per file. Thank you.
8) Message boards : Problems and Bug Reports : [New release] BRP app v1.30 (new 64bit versions) (Message 112257) Posted 27 Oct 2012 by Jeroen Post: One of my Linux systems has completed 20 tasks last night with BRP4 version 1.30 CUDA 64-bit. All tasks completed successfully with a few tasks that have validated so far. Run time is very similar to the CUDA 32-bit application. I plan to try out the CPU version as well. Thanks for the updates!
9) Message boards : Problems and Bug Reports : [New release] BRP app v1.28 feedback thread (Message 112211) Posted 31 Aug 2012 by Jeroen Post: The older cards are also running well with the new version. 8800GT G92 512 MB - x16 slot @ 5.0 GT/s 1.28: 2940 seconds 1.24: ~3600 seconds
10) Message boards : Problems and Bug Reports : [New release] BRP app v1.28 feedback thread (Message 112204) Posted 28 Aug 2012 by Jeroen Post: Here are some preliminary numbers for the GTX 680. One task per GPU System #1 - Single GPU x16 3.0 - 721 seconds System #2 - Multi GPU x16 3.0 - 785 seconds x8 3.0 - 901 seconds Overall, the performance looks great so far. I want to do some more testing with multiple tasks running at once, different PCI-E configurations, and with the CPU dedicated for BRP4 GPU only. The above tests were done with ~50% CPU load from running other CPU tasks at the same time.
11) Message boards : Problems and Bug Reports : [New release] BRP app v1.28 feedback thread (Message 112200) Posted 26 Aug 2012 by Jeroen Post: I ran the new CUDA 1.28 app via one of my Windows systems today. I have not been able to get much work today but the two tasks that ran via my GTX 580, completed at 834 seconds each. This is with one task running at a time. GPU load was at approximately 90-91% while running one task. If memory serves me right, the previous application ran at around 1360 seconds per task with the 1.25 app via this system. This is a very decent improvement in performance. Thanks for the work put into optimizing the BRP4 applications.
12) Message boards : Problems and Bug Reports : [New release] BRP app v1.28 feedback thread (Message 112194) Posted 24 Aug 2012 by Jeroen Post: I am looking forward to testing out BRP4 v1.28 for Linux. Thanks for the updates!

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