Sieve Application Testing
log in

Advanced search

Message boards : News : Sieve Application Testing

Previous · 1 · 2
Author Message
RFGuy_KCCO
Send message
Joined: 8 Oct 13
Posts: 21
Credit: 1,096,742,075
RAC: 403
Message 20570 - Posted: 13 Jun 2015, 13:40:43 UTC
Last modified: 13 Jun 2015, 13:42:07 UTC

Credit for these WU's will definitely need to be adjusted. I am only receiving ~30K points for each of these, yet they run for ~12 hours. I usually get ~240K points per Large WU and they only run for ~6 hours.

When do you plan to adjust the credits granted per WU, Slicker? It isn't currently worth my time to crunch the Sieve WU's at twice the crunching time and only 1/8th the points.
____________

Profile Peciak
Avatar
Send message
Joined: 4 Aug 09
Posts: 13
Credit: 669,955,177
RAC: 931
Message 20571 - Posted: 13 Jun 2015, 15:44:07 UTC

all in good time

Rymorea
Send message
Joined: 14 Oct 14
Posts: 100
Credit: 200,411,819
RAC: 4
Message 20575 - Posted: 13 Jun 2015, 17:22:07 UTC

2 of slieve wu done and validated. But screen gets too laggy both my amd and NV so I stop slieve till the next test version .


for AMD slieve http://boinc.thesonntags.com/collatz/workunit.php?wuid=15773208

for NV slieve http://boinc.thesonntags.com/collatz/workunit.php?wuid=15773194
____________
Seti@home Classic account User ID 955 member since 8 Sep 1999 classic CPU time 539,770 hours

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 20599 - Posted: 16 Jun 2015, 15:01:35 UTC - in response to Message 20570.
Last modified: 16 Jun 2015, 15:22:34 UTC

Credit for these WU's will definitely need to be adjusted. I am only receiving ~30K points for each of these, yet they run for ~12 hours. I usually get ~240K points per Large WU and they only run for ~6 hours.

When do you plan to adjust the credits granted per WU, Slicker? It isn't currently worth my time to crunch the Sieve WU's at twice the crunching time and only 1/8th the points.


I adjusted it but it will likely need more tweaking. Now I need to write an app to parse the code from the ones done thus far and either grant (if failed in validation) or update (if validated but at the lower amount).

FYI, after the initial 400 sieve WUs are done, the next WUs are much smaller which should help with testing. I hate to see someone waste hours crunching only to have an issue when it is almost done.

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 20606 - Posted: 16 Jun 2015, 18:39:25 UTC

I also updated the total_credit for all hosts and users that submitted sieve WUs prior to the credit adjustment earlier today.

Profile Peciak
Avatar
Send message
Joined: 4 Aug 09
Posts: 13
Credit: 669,955,177
RAC: 931
Message 20625 - Posted: 18 Jun 2015, 20:02:20 UTC

whether there are different lengths WU?
http://boinc.thesonntags.com/collatz/result.php?resultid=18139352
http://boinc.thesonntags.com/collatz/result.php?resultid=18190315
as you can choose a specific type of WU?

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 20627 - Posted: 18 Jun 2015, 20:31:21 UTC

I could use some help finding good values for the default settings where the video performance is reasonable and yet GPU utilization is quite high. Could those who tried the sieve but didn't like the video lag try the following in the collatz_sieve_<version_platform_plan_class>.config file:

verbose=1
kernels_per_reduction=2
threads=6
sleep=1
lut_size=15
reduceCPU=0


It gives pretty good performance on my laptops GPU but I'd like to hear from both the nVidia and AMD users with slower GPUs on whether it works for them.

(and to answer Peciak's question)
Because I also need slower GPUs to test out the sieve app, I reduced the sieve WU size to about that of the solo WUs so they should finish much faster than the first sieve WUs.

Jon Fox
Send message
Joined: 6 Sep 09
Posts: 36
Credit: 351,495,953
RAC: 260,267
Message 20633 - Posted: 19 Jun 2015, 13:40:44 UTC - in response to Message 20627.

I've updated my config file to reflect the settings you itemized and have reenabled accepting the sieve/test work units. I'll report back as soon as I have performance data.

For the record, here's my device configuration details:

6/1/2015 3:44:29 PM | | Starting BOINC client version 7.6.2 for windows_x86_64
6/1/2015 3:44:29 PM | | log flags: file_xfer, sched_ops, task, task_debug, unparsed_xml
6/1/2015 3:44:29 PM | | Libraries: libcurl/7.39.0 OpenSSL/1.0.2a zlib/1.2.8
6/1/2015 3:44:29 PM | | Data directory: C:\ProgramData\BOINC
6/1/2015 3:44:29 PM | | CUDA: NVIDIA GPU 0: NVS 4200M (driver version 341.05, CUDA version 6.5, compute capability 2.1, 1024MB, 971MB available, 190 GFLOPS peak)
6/1/2015 3:44:29 PM | | OpenCL: NVIDIA GPU 0: NVS 4200M (driver version 341.05, device version OpenCL 1.1 CUDA, 1024MB, 971MB available, 190 GFLOPS peak)
6/1/2015 3:44:29 PM | | OpenCL CPU: Intel(R) Core(TM) i5-2520M CPU @ 2.50GHz (OpenCL driver vendor: Intel(R) Corporation, driver version 3.0.1.15216, device version OpenCL 1.2 (Build 80752))
6/1/2015 3:44:29 PM | | Processor: 4 GenuineIntel Intel(R) Core(TM) i5-2520M CPU @ 2.50GHz [Family 6 Model 42 Stepping 7]
6/1/2015 3:44:29 PM | | Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss htt tm pni ssse3 cx16 sse4_1 sse4_2 popcnt aes syscall nx lm avx vmx smx tm2 pbe
6/1/2015 3:44:29 PM | | OS: Microsoft Windows 7: Professional x64 Edition, Service Pack 1, (06.01.7601.00)
6/1/2015 3:44:29 PM | | Memory: 7.89 GB physical, 15.77 GB virtual
6/1/2015 3:44:29 PM | | Disk: 448.96 GB total, 239.73 GB free


--
Jon

Dr Who Fan
Avatar
Send message
Joined: 27 May 14
Posts: 21
Credit: 4,562,054
RAC: 0
Message 20636 - Posted: 19 Jun 2015, 18:19:51 UTC
Last modified: 19 Jun 2015, 18:21:40 UTC

Finally got one to complete using the settings Slicker posted.
STILL HAD GPU DRIVER CRASH/RESTART when the task first started and after a task suspend/restart.

At times the CPU usage was up to about 90% but never went below about 80%. The task was sent out with the following parameters > 0.581CPU + 1NV.

Did NOT notice any screen lag, I had to modify my local settings while it ran reserve a FULL CORE to run it to avoid whole system from becoming almost unresponsive.

Interesting side note is that GPU-Z claimed 0% (NO) GPU USE while it was running.

Run Information from BoincTasks:

Collatz Conjecture collatz_sieve_2397308560254556962816_844424930131968_0 1.05 Collatz Sieve (opencl_nvidia_gpu) 12:26:59 (09:59:59) Reported: OK + 0.581C + 1NV 80.321 1158.73 MB 817.50 MB


TASK# 18139359

NOTE FOR Slicker: The way Nvidia implements open_cl you NEED TO RESERVE A FULL CORE for high GPU use. Any way you can modify the Settings on the server so when the task is sent out it RESERVES A FULL CORE on Nvidia GPU's?
____________

RFGuy_KCCO
Send message
Joined: 8 Oct 13
Posts: 21
Credit: 1,096,742,075
RAC: 403
Message 20641 - Posted: 20 Jun 2015, 17:56:27 UTC - in response to Message 20636.

NOTE FOR Slicker: The way Nvidia implements open_cl you NEED TO RESERVE A FULL CORE for high GPU use. Any way you can modify the Settings on the server so when the task is sent out it RESERVES A FULL CORE on Nvidia GPU's?


Just use an app_config.xml file to specify 1 CPU core per WU. That is the best way to accomplish what you want on your machine.

Here is what I am currently using on all of my boxes:


<app_config>
<app>
<name>solo_collatz</name>
<gpu_versions>
<gpu_usage>1.0</gpu_usage>
<cpu_usage>1.0</cpu_usage>
</gpu_versions>
</app>
<app>
<name>large_collatz</name>
<gpu_versions>
<gpu_usage>1.0</gpu_usage>
<cpu_usage>1.0</cpu_usage>
</gpu_versions>
</app>
<app>
<name>mini_collatz</name>
<gpu_versions>
<gpu_usage>1.0</gpu_usage>
<cpu_usage>1.0</cpu_usage>
</gpu_versions>
</app>
<app>
<name>micro_collatz</name>
<gpu_versions>
<gpu_usage>1.0</gpu_usage>
<cpu_usage>1.0</cpu_usage>
</gpu_versions>
</app>
<app>
<name>collatz_sieve</name>
<gpu_versions>
<gpu_usage>1.0</gpu_usage>
<cpu_usage>1.0</cpu_usage>
</gpu_versions>
</app>
</app_config>

____________

Jon Fox
Send message
Joined: 6 Sep 09
Posts: 36
Credit: 351,495,953
RAC: 260,267
Message 20649 - Posted: 22 Jun 2015, 13:22:19 UTC - in response to Message 20633.

Have been running for several days with the recommended "test" configuration settings. Twelve (12) WUs processed, none successfully completed. I've included the WU links below.

What I noticed is that in all but one WU I received a 0x5 exit code "Access Denied". In the single case (WU 18139314), where this error did not appear, the WU ran to completion but failed validation. Are there any general troubleshooting guidelines for the 0x5 exit code?

http://boinc.thesonntags.com/collatz/result.php?resultid=18327372

http://boinc.thesonntags.com/collatz/result.php?resultid=18315948

http://boinc.thesonntags.com/collatz/result.php?resultid=18293814

http://boinc.thesonntags.com/collatz/result.php?resultid=18293341

http://boinc.thesonntags.com/collatz/result.php?resultid=18293237

http://boinc.thesonntags.com/collatz/result.php?resultid=18245596

http://boinc.thesonntags.com/collatz/result.php?resultid=18245493

http://boinc.thesonntags.com/collatz/result.php?resultid=18245126

http://boinc.thesonntags.com/collatz/result.php?resultid=18244018

http://boinc.thesonntags.com/collatz/result.php?resultid=18220324

http://boinc.thesonntags.com/collatz/result.php?resultid=18219013

http://boinc.thesonntags.com/collatz/result.php?resultid=18139293

--
Jon

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 20650 - Posted: 22 Jun 2015, 13:50:05 UTC - in response to Message 20636.


NOTE FOR Slicker: The way Nvidia implements open_cl you NEED TO RESERVE A FULL CORE for high GPU use. Any way you can modify the Settings on the server so when the task is sent out it RESERVES A FULL CORE on Nvidia GPU's?


The app, at least for Windows, runs at below normal priority rather than idle priority which is the norm for BOINC apps. That should mean that it is able to use the CPU whenever it needs it but that any end-user apps will take priority as they run at normal priority.

Given the issue with driver crashes and out of resource errors on nVidia GPUs, or at least the mobile versions, I'm going to start working on a CUDA version of the sieve app since nVidia doesn't give OpenCL the same attention as CUDA.

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 20651 - Posted: 22 Jun 2015, 14:24:09 UTC - in response to Message 20649.

Have been running for several days with the recommended "test" configuration settings. Twelve (12) WUs processed, none successfully completed. I've included the WU links below.

What I noticed is that in all but one WU I received a 0x5 exit code "Access Denied". In the single case (WU 18139314), where this error did not appear, the WU ran to completion but failed validation. Are there any general troubleshooting guidelines for the 0x5 exit code?


On the bright side, the error occurred when it tried reading the results for the first time so at least it didn't run for hours before crashing.

The error is -5 "out of resources" and is coming from the nVidia driver occurring when it calls clEnqueueMapBuffer to read the results. I'll have to switch the code to use clEnqueueReadBuffer and see if that works for your GPU.

Jon Fox
Send message
Joined: 6 Sep 09
Posts: 36
Credit: 351,495,953
RAC: 260,267
Message 20652 - Posted: 22 Jun 2015, 16:13:16 UTC - in response to Message 20651.

Understood. Thanks, Slicker.

Just let me know when and how I can continue to help with the testing.

--
Jon

Profile DoctorNow
Avatar
Send message
Joined: 12 Jul 09
Posts: 30
Credit: 102,805,175
RAC: 0
Message 20853 - Posted: 19 Jul 2015, 14:40:52 UTC
Last modified: 19 Jul 2015, 15:04:13 UTC

Well, just wanted to try the first Sieve WUs - activated the app and the "run test app"-options in my account, but the project keeps telling me this:

19.07.2015 16:17:57 | Collatz Conjecture | Scheduler request completed: got 0 new tasks

Even though the server status says there are wus for the sieve app...
Am I missing something? What driver requirement does the Sieve need? I have 340.52 installed.

Edit
Never mind, found the problem. I overlooked that the app is only for OpenCL, now I got work. ;-)
____________
Life is Science, and Science rules. To the universe and beyond
Proud member of BOINC@Heidelberg
My BOINC-Stats

Profile DoctorNow
Avatar
Send message
Joined: 12 Jul 09
Posts: 30
Credit: 102,805,175
RAC: 0
Message 20854 - Posted: 19 Jul 2015, 16:42:50 UTC
Last modified: 19 Jul 2015, 16:45:02 UTC

Hm, I've tested now several values in the config to get the enormous lags out which both my video cards have (GTX 560 and GTX 660), but even with very low values my machines are still lagging heavily.
That's quite unusual. And since there isn't much difference every time I try something new I'm getting the feeling that the app doesn't seem to read the config file (I do restart BOINC every time of course).
Sounds impossible since some of you obviously have no problems with it but I actually don't know what else could be the problem...
____________
Life is Science, and Science rules. To the universe and beyond
Proud member of BOINC@Heidelberg
My BOINC-Stats

Profile JohnMD
Avatar
Send message
Joined: 25 Mar 14
Posts: 29
Credit: 15,481,538
RAC: 0
Message 20927 - Posted: 29 Jul 2015, 21:28:02 UTC - in response to Message 20627.


verbose=1
kernels_per_reduction=2
threads=6
sleep=1
lut_size=15
reduceCPU=0

I tried these on my Ion after getting failures with other recommendations.
I still get -
Error: (-4)Memory object allocation failure at 1275 of collatzOpenCL::RunSteps
Error: GPU steps do not match CPU steps. Workunit processing aborted.
Hope this means something to you
:-)

Previous · 1 · 2
Post to thread

Message boards : News : Sieve Application Testing


Main page · Your account · Message boards


Copyright © 2018 Jon Sonntag; All rights reserved.