OpenCL AstroPulse on Intel GPUs - report issues here

Message boards : AstroPulse : OpenCL AstroPulse on Intel GPUs - report issues here
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 47962 - Posted: 7 Nov 2013, 13:53:03 UTC

Currently we are testing new Intel OpenCL AP buildwith improved precision and reduced CPU consumption.
Please report any found issues in this thread.
ID: 47962 · Report as offensive
Richard Haselgrove
Volunteer tester

Send message
Joined: 3 Jan 07
Posts: 1451
Credit: 3,266,428
RAC: 0
United Kingdom
Message 47971 - Posted: 7 Nov 2013, 16:37:25 UTC

I've been running an advance copy of r2058, and before that r1922, to get an idea of CPU usage - All tasks for computer 67008.

The interesting question for me is the extent to which the saved CPU time is actually usable, without slowing down either the intel_gpu app itself, or CPU apps from other projects.

The tasks which you see at the moment were all run with 75% CPU loading: the tasks issued yesterday, 6 Nov, are just starting to run at 100% loading this afternoon. I'll watch and report.
ID: 47971 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 47978 - Posted: 7 Nov 2013, 19:30:20 UTC - in response to Message 47971.  

To answer your question look at my host results. CPU runs fully loaded, 4 CPU tasks + 1 intel GPU task. No slowdown for intel GPU task, at least comparing with older app version.

ID: 47978 · Report as offensive
Richard Haselgrove
Volunteer tester

Send message
Joined: 3 Jan 07
Posts: 1451
Credit: 3,266,428
RAC: 0
United Kingdom
Message 47982 - Posted: 7 Nov 2013, 21:20:42 UTC - in response to Message 47978.  

To answer your question look at my host results. CPU runs fully loaded, 4 CPU tasks + 1 intel GPU task. No slowdown for intel GPU task, at least comparing with older app version.

Yes, I'd agree with that here - no slowdown for the intel_gpu app, and the slowdown for the co-running CPU apps - while enough (probably) to be statistically significant - is tolerable.

Which means I've now got an equally interesting question to pose to Oliver Bock.

I'm away for the next couple of days (early train in the morning), so that will probably have to wait until Monday.
ID: 47982 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 47984 - Posted: 7 Nov 2013, 21:55:27 UTC - in response to Message 47982.  

Good ride then. To monday we will have more validations I hope to look at.
ID: 47984 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 29 May 06
Posts: 1037
Credit: 8,440,339
RAC: 21
United Kingdom
Message 47985 - Posted: 7 Nov 2013, 22:10:00 UTC - in response to Message 47984.  

Got my three v6.05 tasks resent as v6.07 tasks, first one has started on the Intel HD4000,
GT650 is fully occupied with PrimeGrid Genefer Cuda task, CPU is fully occupied with two v7 MB tasks, and two Genefer CPU tasks (one AVX and one SSE3)

Claggy
ID: 47985 · Report as offensive
Werinbert
Volunteer tester

Send message
Joined: 11 May 13
Posts: 36
Credit: 1,001,497
RAC: 0
United States
Message 48000 - Posted: 8 Nov 2013, 23:42:46 UTC
Last modified: 8 Nov 2013, 23:44:18 UTC

I am having trouble getting the IntelGPU WUs. I keep getting a message that SETI Beta is not the highest priority. I am trying to run this on my i7-3770K and have set my preferences to only accept the Intel app (I am running PrimeGrid/CAS on my cpu cores, Collatz on my GPU). And I am not running any other IntelGPU apps.

Is Boinc treating it as using the cpu for scheduling or as a separate flow like the gpu apps?

Or is there something else I missed and need to check to get the IntelGPU app running.
ID: 48000 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 48001 - Posted: 9 Nov 2013, 6:51:30 UTC - in response to Message 48000.  

Should be separate like other GPUs.
But with BOINC's scheduler no one can say definitely IMO.
Try to enable debug options, post logs with debug info and hope that someone could understand what is wrong then :)
ID: 48001 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 29 May 06
Posts: 1037
Credit: 8,440,339
RAC: 21
United Kingdom
Message 48004 - Posted: 9 Nov 2013, 20:22:14 UTC - in response to Message 48000.  
Last modified: 9 Nov 2013, 20:27:10 UTC

I am having trouble getting the IntelGPU WUs. I keep getting a message that SETI Beta is not the highest priority. I am trying to run this on my i7-3770K and have set my preferences to only accept the Intel app (I am running PrimeGrid/CAS on my cpu cores, Collatz on my GPU). And I am not running any other IntelGPU apps.

Try upgrading to Boinc 7.2.28 (It became recommended yesterday), there have been scheduling fixes since Boinc 7.0.64

Also, check your Boinc startup, there are no intel GPUs showing on your computers at present.

Claggy
ID: 48004 · Report as offensive
Werinbert
Volunteer tester

Send message
Joined: 11 May 13
Posts: 36
Credit: 1,001,497
RAC: 0
United States
Message 48006 - Posted: 10 Nov 2013, 3:07:03 UTC

I tried the recommended: debug, upgraded to 7.2.28, and checked the startup. Still not working...

I moved my problem to the BOINC forums as it seemed more appropriate.
http://boinc.berkeley.edu/dev/forum_thread.php?id=8716
ID: 48006 · Report as offensive
Richard Haselgrove
Volunteer tester

Send message
Joined: 3 Jan 07
Posts: 1451
Credit: 3,266,428
RAC: 0
United Kingdom
Message 48008 - Posted: 10 Nov 2013, 9:33:32 UTC

As Werinbert has worked out for himself over at BOINC, Intel GPU crunching requires that there is a monitor loading (monitor, dummy plug, or secondary video cable) connected to the Intel video output.

How quickly we all forget these things!
ID: 48008 · Report as offensive
Profile Tetsuji Maverick Rai
Project developer
Volunteer developer
Avatar

Send message
Joined: 15 Jun 05
Posts: 399
Credit: 16,571,350
RAC: 0
Japan
Message 48480 - Posted: 14 Dec 2013, 20:37:13 UTC
Last modified: 14 Dec 2013, 20:42:32 UTC

Hi,
I'm not a newbie about s@h/ap, but a newbie about opencl/cuda. 8 years ago, I was optimizing s@h_enhanced, but do you remember me?

I installed nvidia driver 331.20 downloaded from nvidia site on my Fedora 19, and ran the boinc-client 7.0.65 (as provided by Fedora), downloaded AP v6 6.07 (opencl_nvidia_100) and some wu's. However none of them is processed. They even don't start. What's needed? Where is "howto about opencl"?

Thanks in advance.

btw, cuda apps work fine on Windows xp on the same machine. So the hardware has no problem.
Luckiest in the world. WMD = Weapon of Mass Distraction
ID: 48480 · Report as offensive
Profile Tetsuji Maverick Rai
Project developer
Volunteer developer
Avatar

Send message
Joined: 15 Jun 05
Posts: 399
Credit: 16,571,350
RAC: 0
Japan
Message 48483 - Posted: 15 Dec 2013, 10:11:18 UTC

I'm using boinc-7.0.65. This may be the problem. I use "always use GPU" option to run the app.
Luckiest in the world. WMD = Weapon of Mass Distraction
ID: 48483 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 48488 - Posted: 15 Dec 2013, 21:39:37 UTC - in response to Message 48483.  
Last modified: 15 Dec 2013, 21:39:57 UTC

Well, I'm not sure that your troubles have anything with OpenCL app for Intel GPUs ;)
And yes, we remember you, of course :).

"Didn't start" - so they failed or what? Could you be more specific?
ID: 48488 · Report as offensive
Profile jay_e
Volunteer tester

Send message
Joined: 8 Apr 13
Posts: 15
Credit: 240,305
RAC: 0
United States
Message 48558 - Posted: 19 Dec 2013, 12:26:52 UTC

Subject: Spaces needed in app_config.xml for GPU

Greetings!

I got this 'error' in the event log:
9-Dec-2013 06:21:53 [SETI@home Beta Test] Found app_config.xml
19-Dec-2013 06:21:53 [SETI@home Beta Test] app setiathome_v7 not found in app_config.xml
19-Dec-2013 06:21:53 [SETI@home Beta Test] app beta_17 not found in app_config.xml

hmmm.

This was odd, because my file looked like:

<app_config>
<app>
<name>setiathome_v7</name>
<max_concurrent>4</max_concurrent>
<gpu_versions>
<gpu_usage>1.0</gpu_usage>
<cpu_usage>1.0</cpu_usage>
</gpu_versions>
</app>
<app>
<name>astropulse_v6</name>
<max_concurrent>4</max_concurrent>
<gpu_versions>
<gpu_usage>1.0</gpu_usage>
<cpu_usage>1.0</cpu_usage>
</gpu_versions>
</app>
<app>
<name>beta_17</name>
<max_concurrent>4</max_concurrent>
<gpu_versions>
<gpu_usage>1.0</gpu_usage>
<cpu_usage>1.0</cpu_usage>
</gpu_versions>
</app>
</app_config>


I admit; I don't always know what I'm doing.
I tried putting in spaces to pretty it up.
For some reason, seti-beta *then* read the values - and did not give the error.

So. Are the spaces supposed to be required?
Or is this a small bug?


Here is the environment stuff (after the fix) from the beginning of the event log..
19-Dec-2013 06:44:39 [---] Starting BOINC client version 7.0.65 for x86_64-pc-linux-gnu
19-Dec-2013 06:44:39 [---] log flags: file_xfer, sched_ops, task
19-Dec-2013 06:44:39 [---] Libraries: libcurl/7.29.0 OpenSSL/1.0.1c zlib/1.2.7 libidn/1.25 librtmp/2.3
19-Dec-2013 06:44:39 [---] Data directory: /var/lib/boinc-client
19-Dec-2013 06:44:39 [---] Processor: 8 AuthenticAMD AMD FX(tm)-8150 Eight-Core Processor [Family 21 Model 1 Stepping 2]
19-Dec-2013 06:44:39 [---] Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx m
mxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good nopl nonstop_tsc extd_apicid aperfmperf pni pclmulqdq monitor ssse3 cx16 sse4_1 sse4_2 popcnt aes xsave a
vx lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs xop skinit wdt lwp fma4 nodeid_msr topoext perfctr_core arat cpb hw_ps
tate npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold
19-Dec-2013 06:44:39 [---] OS: Linux: 3.8.0-34-lowlatency
19-Dec-2013 06:44:39 [---] Memory: 7.70 GB physical, 9.31 GB virtual
19-Dec-2013 06:44:39 [---] Disk: 9.61 GB total, 7.36 GB free
19-Dec-2013 06:44:39 [---] Local time is UTC -5 hours
19-Dec-2013 06:44:39 [---] CAL: ATI GPU 0: AMD Radeon HD 7700 series (Cape Verde) (CAL version 1.4.1741, 2048MB, 1889MB available, 2048 GFLOPS peak)
19-Dec-2013 06:44:39 [---] OpenCL: AMD/ATI GPU 0: AMD Radeon HD 7700 series (Cape Verde) (driver version 1084.4 (VM), device version OpenCL 1.2 AMD-APP (1084.4)
, 2048MB, 1889MB available, 2048 GFLOPS peak)
19-Dec-2013 06:44:39 [Einstein@Home] Found app_config.xml
19-Dec-2013 06:44:39 [SETI@home Beta Test] Found app_config.xml
19-Dec-2013 06:44:39 [---] Config: GUI RPCs allowed from:
19-Dec-2013 06:44:39 [World Community Grid] URL http://www.worldcommunitygrid.org/; Computer ID 2635735; resource share 20
19-Dec-2013 06:44:39 [Einstein@Home] URL http://einstein.phys.uwm.edu/; Computer ID 9846906; resource share 20
19-Dec-2013 06:44:39 [malariacontrol.net] URL http://www.malariacontrol.net/; Computer ID 731405; resource share 20
19-Dec-2013 06:44:39 [LHC@home 1.0] URL http://lhcathomeclassic.cern.ch/sixtrack/; Computer ID 10310841; resource share 20
19-Dec-2013 06:44:39 [SETI@home Beta Test] URL http://setiweb.ssl.berkeley.edu/beta/; Computer ID 69470; resource share 20
19-Dec-2013 06:44:39 [World Community Grid] General prefs: from World Community Grid (last modified 14-Dec-2013 06:13:34)
19-Dec-2013 06:44:39 [World Community Grid] Host location: none
...


Thank you for yor patience.

Happy Holidays!

Jay
ID: 48558 · Report as offensive
Profile jay_e
Volunteer tester

Send message
Joined: 8 Apr 13
Posts: 15
Credit: 240,305
RAC: 0
United States
Message 48559 - Posted: 19 Dec 2013, 12:35:52 UTC - in response to Message 48488.  

about the start ups..

I still have to restart the boinc-client after every reboot.
sudo /etc/init.d/boinc-client restart

I did a system recovery this week and the linux- UbuntuStudio that I use picked up BOINC 7.0.65..

Jay
ID: 48559 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 29 May 06
Posts: 1037
Credit: 8,440,339
RAC: 21
United Kingdom
Message 48565 - Posted: 19 Dec 2013, 16:19:25 UTC - in response to Message 48558.  

Subject: Spaces needed in app_config.xml for GPU

Greetings!

I got this 'error' in the event log:
9-Dec-2013 06:21:53 [SETI@home Beta Test] Found app_config.xml
19-Dec-2013 06:21:53 [SETI@home Beta Test] app setiathome_v7 not found in app_config.xml
19-Dec-2013 06:21:53 [SETI@home Beta Test] app beta_17 not found in app_config.xml

19-Dec-2013 06:44:39 [SETI@home Beta Test] URL http://setiweb.ssl.berkeley.edu/beta/; Computer ID 69470; resource share 20


The wording is wrong, what it really means are that those two apps aren't in the client_state.XML

That host hasn't done any setiathome_v7 work, so that's the reason for that error,
Seti Beta doesn't have an app called beta_17, so that's the reason for that error.

Claggy
ID: 48565 · Report as offensive
Profile jay_e
Volunteer tester

Send message
Joined: 8 Apr 13
Posts: 15
Credit: 240,305
RAC: 0
United States
Message 48602 - Posted: 20 Dec 2013, 14:58:53 UTC - in response to Message 48565.  

Thank you, Claggy.

I fixed my app_config file.
I owe you another espresso (or beer.)

Happy Holidays!

Jay
ID: 48602 · Report as offensive
tOM (Ottawa)
Volunteer tester
Avatar

Send message
Joined: 27 Jan 14
Posts: 3
Credit: 0
RAC: 0
Pitcairn Islands
Message 48980 - Posted: 27 Jan 2014, 3:04:27 UTC

So how do I get astropulse to run on my intel GPU?
ID: 48980 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 29 May 06
Posts: 1037
Credit: 8,440,339
RAC: 21
United Kingdom
Message 48982 - Posted: 27 Jan 2014, 3:47:46 UTC - in response to Message 48980.  

So how do I get astropulse to run on my intel GPU?

See my post in the other thread.

Claggy
ID: 48982 · Report as offensive

Message boards : AstroPulse : OpenCL AstroPulse on Intel GPUs - report issues here


 
©2019 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.