BRP4 1.28 OpenCL and CUDA app : feedback thread

log in

Advanced search

Message boards : Problems and Bug Reports : BRP4 1.28 OpenCL and CUDA app : feedback thread

1 · 2 · 3 · Next
Author Message
Profile Bikeman (Heinz-Bernd Eggenstein)
Volunteer moderator
Project administrator
Project developer
Avatar
Send message
Joined: 28 Aug 06
Posts: 3502
Credit: 149,228,510
RAC: 116,906
Message 118735 - Posted: 17 Aug 2012, 12:13:00 UTC
Last modified: 20 Aug 2012, 11:20:37 UTC

This thread is for discussing problems (or any other feedback if you like) related to the app BRP4 version 1.28 OpenCL for ATI/AMD graphics cards
____________

transient
Send message
Joined: 3 Jun 05
Posts: 63
Credit: 36,589,746
RAC: 23,542
Message 118745 - Posted: 18 Aug 2012, 10:00:30 UTC
Last modified: 18 Aug 2012, 10:05:03 UTC

This is a report not a problem. :)

Runtimes on a HD5770, 2 tasks at a time.

App Average of Count of Run Time(s) results Binary Radio Pulsar Search (Arecibo) v1.24 (opencl-ati) 14490.90 38 Binary Radio Pulsar Search (Arecibo) v1.28 (opencl-ati) 6105.84 7


Quite in improvement, I would say. Next step would be collecting a few more results on the 1.28 app, after that reverting to running one task at a time.
____________
Profile Alex
Send message
Joined: 1 Mar 05
Posts: 438
Credit: 47,775,980
RAC: 1,256
Message 118748 - Posted: 18 Aug 2012, 21:35:42 UTC

Today I installed latest AMD-drivers 12.8. Runtimes on my HD6950 increased from ~3200 sec's (2wu's) to ~3770 sec's (2 wu's).
Will switch back to 12.6
____________

Profile Ageless
Avatar
Send message
Joined: 26 Jan 05
Posts: 2974
Credit: 5,374,792
RAC: 0
Message 118754 - Posted: 19 Aug 2012, 0:26:39 UTC
Last modified: 19 Aug 2012, 0:26:55 UTC

App v1.28 halved both my GPU and CPU run times. From ~4,600 seconds GPU and ~750 seconds CPU to ~2,300 seconds GPU and ~370 seconds CPU. Thumbs up. AMD HD6850, 2GB.
____________
Jord

JHMarshall
Send message
Joined: 24 Jul 12
Posts: 12
Credit: 87,185,812
RAC: 340,674
Message 118757 - Posted: 19 Aug 2012, 3:41:12 UTC - in response to Message 118748.

Alex,

Saw the same problem with 12.8 -> increased run times. I run Win7 64bit. I don't know why but I changed from "Balanced" performance to "High" performance under "Power Options" and saw run times return to the 12.6 values. This was on Milky Way at Home. I haven't timed Einstein yet.

Joe
____________

Mad_Max
Send message
Joined: 2 Jan 10
Posts: 65
Credit: 20,765,404
RAC: 17,861
Message 118771 - Posted: 20 Aug 2012, 13:34:22 UTC
Last modified: 20 Aug 2012, 13:41:37 UTC

After update E@H OpenCL application from 1.24 to version 1.28 error with the lack of free memory at once disappeared (which confirms that the problem was not in amount of free RAM, but a bug - in the application or in the AMD drivers or probably both).

But 1.28 app still not working - just another error in all Wus:
http://einstein.phys.uwm.edu/results.php?hostid=2896418
[15:51:43][3012][ERROR] Error during OpenCL kernel setup: HSFFB (error: -5)
[15:51:43][3012][ERROR] Demodulation failed (error: 2019)!
(0x7e3) - exit code 2019 (0x7e3)
or
[16:51:34][6032][ERROR] Error during OpenCL kernel setup: TSMP_T (error: -5)
[16:51:34][6032][ERROR] Demodulation failed (error: 2019)!
(0x7e3) - exit code 2019 (0x7e3)

And POEM@Home OpenCL app still work fine on same hard&soft:
http://boinc.fzk.de/poem/results.php?hostid=109617

P.S.
Errors with free RAM (in 1.24 version) described in another topic:
http://einstein.phys.uwm.edu/forum_thread.php?id=9445&nowrap=true#118714
http://einstein.phys.uwm.edu/forum_thread.php?id=9445&nowrap=true#118752

Profile Ageless
Avatar
Send message
Joined: 26 Jan 05
Posts: 2974
Credit: 5,374,792
RAC: 0
Message 118772 - Posted: 20 Aug 2012, 13:59:48 UTC - in response to Message 118771.

But 1.28 app still not working - just another error in all Wus:

Actually, it's the same error. Before you also had exit code 2019 on all your tasks.

For the developers, so far I've found that this exit code 2019 is the same as There was an error while deleting the color transform. (0x7e3) - exit code 2019 (0x7e3) that I had on Albert. Ask Oliver how he fixed that.
____________
Jord
enginerd
Avatar
Send message
Joined: 9 Feb 05
Posts: 20
Credit: 4,307,082
RAC: 0
Message 118773 - Posted: 20 Aug 2012, 19:01:41 UTC

still no dice for me on HD5670/winXP32.
I get the same "color transform" error #2019 from the earlier 1.24 app.
fyi

<core_client_version>7.0.28</core_client_version>
<![CDATA[
<message>
There was an error while deleting the color transform. (0x7e3) - exit code 2019 (0x7e3)
</message>
<stderr_txt>
Activated exception handling...
[11:53:20][3960][INFO ] Starting data processing...
[11:53:22][3960][INFO ] Using OpenCL platform provided by: Advanced Micro Devices, Inc.
[11:53:22][3960][INFO ] Using OpenCL device "Redwood" by: Advanced Micro Devices, Inc.
[11:53:22][3960][INFO ] Checkpoint file unavailable: status.cpt (No such file or directory).
------> Starting from scratch...
[11:53:22][3960][INFO ] Header contents:
------> Original WAPP file: ./p2030.20101226.G193.96-00.20.C.b4s0g0.00000_DM151.20
------> Sample time in microseconds: 65.4762
------> Observation time in seconds: 274.62705
------> Time stamp (MJD): 55556.179870079133
------> Number of samples/record: 0
------> Center freq in MHz: 1214.289551
------> Channel band in MHz: 0.33605957
------> Number of channels/record: 960
------> Nifs: 1
------> RA (J2000): 61508.1564999
------> DEC (J2000): 163727.051001
------> Galactic l: 0
------> Galactic b: 0
------> Name: G193.96-00.20.C
------> Lagformat: 0
------> Sum: 1
------> Level: 3
------> AZ at start: 0
------> ZA at start: 0
------> AST at start: 0
------> LST at start: 0
------> Project ID: --
------> Observers: --
------> File size (bytes): 0
------> Data size (bytes): 0
------> Number of samples: 4194304
------> Trial dispersion measure: 151.2 cm^-3 pc
------> Scale factor: 0.00137665
[11:53:25][3960][INFO ] Seed for random number generator is 1168041574.
[11:53:34][3960][INFO ] Derived global search parameters:
------> f_A probability = 0.08
------> single bin prob(P_noise > P_thr) = 1.32531e-008
------> thr1 = 18.139
------> thr2 = 21.241
------> thr4 = 26.2686
------> thr8 = 34.6478
------> thr16 = 48.9581
[11:53:41][3960][ERROR] Error during OpenCL kernel setup: PS_R3 (error: -5)
[11:53:41][3960][ERROR] Demodulation failed (error: 2019)!
11:53:41 (3960): called boinc_finish

</stderr_txt>
]]>
____________

Profile Alex
Send message
Joined: 1 Mar 05
Posts: 438
Credit: 47,775,980
RAC: 1,256
Message 118774 - Posted: 20 Aug 2012, 20:05:21 UTC

You're using XP, so pls check your driver-version

http://einstein.phys.uwm.edu/forum_thread.php?id=9445&nowrap=true#118752
____________

Mad_Max
Send message
Joined: 2 Jan 10
Posts: 65
Credit: 20,765,404
RAC: 17,861
Message 118792 - Posted: 22 Aug 2012, 1:41:33 UTC - in response to Message 118772.

But 1.28 app still not working - just another error in all Wus:

Actually, it's the same error. Before you also had exit code 2019 on all your tasks.

For the developers, so far I've found that this exit code 2019 is the same as There was an error while deleting the color transform. (0x7e3) - exit code 2019 (0x7e3) that I had on Albert. Ask Oliver how he fixed that.

Hmm, you right.
More exactly - at 1.24 app was 2 errors:
-Demodulation failed (error: 2013)! - this related to bug with lack of free RAM (when lot of free RAM actually)
-Demodulation failed (error: 2019)!

At 1.28 app first error (2013) gone, but 2019 still persist.
paul
Send message
Joined: 22 Jun 08
Posts: 1
Credit: 1,179,693
RAC: 0
Message 118794 - Posted: 22 Aug 2012, 9:30:01 UTC

I run einstein in the background & pause it when I need to do processor intensive work. I recently started getting WU's for the gpu but these won't pause when I need to do graphics intensive work. The Boinc manager tasks status is listed as suspended, but the task keeps on using the gpu. I had to abort the task to get on with work. Do you have any suggestions so that I won't have to continue to abort opencl-ati-lion tasks?

Profile Bikeman (Heinz-Bernd Eggenstein)
Volunteer moderator
Project administrator
Project developer
Avatar
Send message
Joined: 28 Aug 06
Posts: 3502
Credit: 149,228,510
RAC: 116,906
Message 118798 - Posted: 22 Aug 2012, 11:57:17 UTC - in response to Message 118794.

Interesting, thanks for the report. The one result that was reported shows that the app crashed somewhere in the drivers at it seems, maybe at the moment it was suspended??

It would be interesting to know whether other Mac users are seeing this as well or whether it was a fluke.

Cheers
HB

____________

Profile DuFutur
Avatar
Send message
Joined: 8 Mar 06
Posts: 8
Credit: 504,835
RAC: 0
Message 118808 - Posted: 23 Aug 2012, 18:10:12 UTC

Hello all.
Since the update I have a AMD 12.8 boinc 7.0.28 and incompatibility with my graphics card HD 5450. In fact as soon as I run Boinc Manager, I immediately blue screen and shutdown. This was not the case with version 12.6.
Someone you have that problem ..??
I have a board version 12.6 ..??

For now and until the problem is solved, I have to work in safe mode with networking support.
With this solution, of course, I can not use the GPU.

I wonder if it takes one of modiffication boinc or wait for a change of AMD, or returned to the version 12.6 (I said with the old version, the problem of error in the color change appeared to be resolved, since my last GPU WU been validated as successful.)
____________

Profile Alex
Send message
Joined: 1 Mar 05
Posts: 438
Credit: 47,775,980
RAC: 1,256
Message 118809 - Posted: 23 Aug 2012, 19:34:46 UTC - in response to Message 118808.

Hello all.
Since the update I have a AMD 12.8 boinc 7.0.28 and incompatibility with my graphics card HD 5450. In fact as soon as I run Boinc Manager, I immediately blue screen and shutdown. This was not the case with version 12.6.
Someone you have that problem ..??
I have a board version 12.6 ..??

For now and until the problem is solved, I have to work in safe mode with networking support.
With this solution, of course, I can not use the GPU.

I wonder if it takes one of modiffication boinc or wait for a change of AMD, or returned to the version 12.6 (I said with the old version, the problem of error in the color change appeared to be resolved, since my last GPU WU been validated as successful.)


You are running XP on both pc's.

As posted in an other thread, the latest amd driver version for XP that supports open-cl is 12.1
Another reason to avoid 12.8 is that crunching time increases signisficant.
____________
Ron Voss
Send message
Joined: 20 Dec 05
Posts: 2
Credit: 3,151,882
RAC: 1,776
Message 118810 - Posted: 24 Aug 2012, 1:32:29 UTC - in response to Message 118794.

I run einstein in the background & pause it when I need to do processor intensive work. I recently started getting WU's for the gpu but these won't pause when I need to do graphics intensive work. The Boinc manager tasks status is listed as suspended, but the task keeps on using the gpu. I had to abort the task to get on with work. Do you have any suggestions so that I won't have to continue to abort opencl-ati-lion tasks?


Same here, MacOS X 10.7 AMD Radeon HD 6750M 512 MB.

I'm sorry to have checked Einstein's "Won't get new tasks" due to sluggish GPU. :-(
____________
enginerd
Avatar
Send message
Joined: 9 Feb 05
Posts: 20
Credit: 4,307,082
RAC: 0
Message 118811 - Posted: 24 Aug 2012, 7:54:06 UTC

>You're using XP, so pls check your driver-version
>
>http://einstein.phys.uwm.edu/forum_thread.php?id=9445&nowrap=true#118752

using 12.1 driver still, i'm just a very special case:

http://einstein.phys.uwm.edu/forum_thread.php?id=9502
____________

Robert
Send message
Joined: 3 Feb 07
Posts: 2
Credit: 3,643,396
RAC: 0
Message 118870 - Posted: 28 Aug 2012, 1:58:00 UTC

This report is a little vague.
I've got an Imac with a Radeeon HD 6770M running Lion 10.7.4

I've had to turn off GPU processing because I noticed that playing flash video (you tube) while doing openCL processing had a tendency to hang the GPU.
The machine hangs hard (mouse pointer still moves but everything else is locked up) and has to be powered off.

Its not completely reproducable. It tends to happen after video has been playing for a couple of minutes. But has happened multiple times.

I assume its the openCL that causing it since it only started happening once the opencl work unit appeared. And its stopped since I unsubscribed from the openCL workunits. But its not a huge sample set.

Anyone else noticed issues?

Robert
Send message
Joined: 3 Feb 07
Posts: 2
Credit: 3,643,396
RAC: 0
Message 118872 - Posted: 28 Aug 2012, 2:01:44 UTC - in response to Message 118810.

You could try the option to not "use GPU while machine is in use".
That way the the GPU tasks should only run when your machine is idle eg overnight.

Profile Bikeman (Heinz-Bernd Eggenstein)
Volunteer moderator
Project administrator
Project developer
Avatar
Send message
Joined: 28 Aug 06
Posts: 3502
Credit: 149,228,510
RAC: 116,906
Message 118875 - Posted: 28 Aug 2012, 11:22:24 UTC

Hi!

Especially for notebooks, the GPU load can be quite nagging: slow down of the user interface, heat and fan noise. I really like that BOINC now has the option to selectively suspend the GPU in the "Activity" menu. So you can, for example, leave GPU suspended except when you really don't care (say at night). You will still get a lot more science done when using the GPU app during the night only compared to all-day CPU-only operation.

Cheers
HB

____________

Sumtec
Send message
Joined: 29 Aug 12
Posts: 2
Credit: 96,529
RAC: 0
Message 118886 - Posted: 29 Aug 2012, 13:36:39 UTC

Hi! The app "Binary Radio Pulsar Search (Arecibo) v1.28 (opencl-ati)" crash on my laptop every time it started. I don't know how to get further detail (e.g. the log file of the app) in order to get this fixed. The OS is Windows 7 Home Basic, ATI driver version: 8.951.0.0 on ATI HD 7730M. And it seems that this laptop can choose which video card to use. There is another integrated Intel HD Graphics 4000. I'm to quite sure about that. What should I do?

1 · 2 · 3 · Next

Message boards : Problems and Bug Reports : BRP4 1.28 OpenCL and CUDA app : feedback thread


Home · Your account · Message boards

This material is based upon work supported by the National Science Foundation (NSF) under Grants PHY-1104902, PHY-1104617 and PHY-1105572 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 © 2016 Bruce Allen