New AstroPulse for GPU ( ATi & NV) released (r1316)


log in

Advanced search

Message boards : Number crunching : New AstroPulse for GPU ( ATi & NV) released (r1316)

Previous · 1 · 2 · 3 · 4 · 5 . . . 11 · Next
Author Message
Profile Raistmer
Volunteer developer
Volunteer tester
Avatar
Send message
Joined: 16 Jun 01
Posts: 3649
Credit: 49,424,765
RAC: 27,024
Russia
Message 1256931 - Posted: 7 Jul 2012, 9:10:39 UTC - in response to Message 1256704.


I'm gonna try this on 2 ATI 5870 GPUs, can I still use command-line parameters like unroll 15, ffa_block 10240 & fetch 5120, 1 per device)
?

As menioned in release notes, yes, you can, listed params work as before.

Profile Raistmer
Volunteer developer
Volunteer tester
Avatar
Send message
Joined: 16 Jun 01
Posts: 3649
Credit: 49,424,765
RAC: 27,024
Russia
Message 1256932 - Posted: 7 Jul 2012, 9:14:20 UTC - in response to Message 1256916.

Completed my first Task. It ran very fast - was on track for about an hour running with 2 MB 6.10 tasks at the same time, but exited early at the 35 minute mark with "Found 30 single pulses and 30 repeating pulses, exiting."

Please keep looking at outcome of this task. 30/30 exits, especially if they come first, are always suspicious it could be just noisy task, or some app failure.

Profile Fred E.Project donor
Volunteer tester
Send message
Joined: 22 Jul 99
Posts: 768
Credit: 24,139,004
RAC: 0
United States
Message 1256935 - Posted: 7 Jul 2012, 9:27:49 UTC

Completed my first Task. It ran very fast - was on track for about an hour running with 2 MB 6.10 tasks at the same time, but exited early at the 35 minute mark with "Found 30 single pulses and 30 repeating pulses, exiting."

Please keep looking at outcome of this task. 30/30 exits, especially if they come first, are always suspicious it could be just noisy task, or some app failure.


Will do that. I started another task and will advise if anything is unusual. Glad we're starting to get some AP work. Thanks for the responses.
____________
Another Fred
Support SETI@home when you search the Web with GoodSearch or shop online with GoodShop.

Profile Fred E.Project donor
Volunteer tester
Send message
Joined: 22 Jul 99
Posts: 768
Credit: 24,139,004
RAC: 0
United States
Message 1257512 - Posted: 8 Jul 2012, 12:49:25 UTC
Last modified: 8 Jul 2012, 12:50:47 UTC

Please keep looking at outcome of this task. 30/30 exits, especially if they come first, are always suspicious it could be just noisy task, or some app failure.

That first task, workunit 10249935008 has gone inconclusive. Another 30/30 exit, workunit 1024964421 has also gone inconclusive. Both are against the stock cpu app. Edit - both found 30/30 end edit. I've had 2 other 30/30 results, one of those validated against a stock cpu app. All 4 of the 30/30 exits came from ap_01jl10_aa_B5, so maybe its just a noisy file. Anyone else seeing this?

Have 4 results from 01jl11 that ran full length and are pending validation.
____________
Another Fred
Support SETI@home when you search the Web with GoodSearch or shop online with GoodShop.

Profile MikeProject donor
Volunteer tester
Avatar
Send message
Joined: 17 Feb 01
Posts: 25223
Credit: 34,842,581
RAC: 21,542
Germany
Message 1257527 - Posted: 8 Jul 2012, 14:35:18 UTC

So long your wingman returns a 30/30 also your result should get validated at the end.

____________

Profile perryjay
Volunteer tester
Avatar
Send message
Joined: 20 Aug 02
Posts: 3377
Credit: 16,405,522
RAC: 9,692
United States
Message 1257621 - Posted: 8 Jul 2012, 18:07:22 UTC
Last modified: 8 Jul 2012, 18:13:18 UTC

Mike,
I don't know if it will hurt anything but I noticed one little problem with your NV app_info. You have this line... <name>AP6_win_x86_SSE2_OpenCL_NV_r1316.exe<app> It starts out as <name> and ends as <app> That should be changed to </name> shouldn't it?

Raistmer, I'm still using the old 560. Would it be okay to just change the name over to 1316 and leave the version number at 601 rather than 604 in the app_info? Is it necessary to ask for 604s?

Edit: to Mike, that mistake was made in your NV app_info example, I see now you have left the version name as 601 in your example for both. Is this working for you?
____________


PROUD MEMBER OF Team Starfire World BOINC

Profile [seti.international] Dirk Sadowski
Volunteer tester
Avatar
Send message
Joined: 6 Apr 07
Posts: 7124
Credit: 61,645,641
RAC: 15,414
Germany
Message 1257623 - Posted: 8 Jul 2012, 18:32:47 UTC - in response to Message 1256374.
Last modified: 8 Jul 2012, 18:36:14 UTC

Raistmer wrote:
New AstroPulse binaries are available here: http://lunatics.kwsn.net/12-gpu-crunching/ap6-for-nv-ati-gpus-r1316-released.msg49053.html#msg49053

Please, report noticed issues in this thread.


Raistmer, a big *Thank you!* to you for your time and hard work - and the same to the alpha and beta tester!

- - - - - - - - - -

What are the default settings (values) for/of the application if nothing is in the <cmdline> in the app_info.xml file for...

-unroll
-ffa_block
-ffa_block_fetch

?

- - - - - - - - - -

How I could find/know which settings (& values) are the best for the own system for max. performance?


Thanks


* Best regards! :-) * Sutaru Tsureku, team seti.international founder. * Optimize your PC for higher RAC. * SETI@home needs your help. *
____________
BR

SETI@home Needs your Help ... $10 & U get a Star!

Team seti.international

Das Deutsche Cafe. The German Cafe.

Profile arkaynProject donor
Volunteer tester
Avatar
Send message
Joined: 14 May 99
Posts: 3748
Credit: 48,777,915
RAC: 1,076
United States
Message 1257630 - Posted: 8 Jul 2012, 19:04:39 UTC - in response to Message 1257621.

Mike,
I don't know if it will hurt anything but I noticed one little problem with your NV app_info. You have this line... <name>AP6_win_x86_SSE2_OpenCL_NV_r1316.exe<app> It starts out as <name> and ends as <app> That should be changed to </name> shouldn't it?

Raistmer, I'm still using the old 560. Would it be okay to just change the name over to 1316 and leave the version number at 601 rather than 604 in the app_info? Is it necessary to ask for 604s?

Edit: to Mike, that mistake was made in your NV app_info example, I see now you have left the version name as 601 in your example for both. Is this working for you?


I left mine the same and it is working just fine on my GTX-560.

<app_info> <app> <name>setiathome_enhanced</name> </app> <file_info> <name>AK_v8b2_win_x64_SSE3.exe</name> <executable/> </file_info> <app_version> <app_name>setiathome_enhanced</app_name> <version_num>603</version_num> <platform>windows_intelx86</platform> <file_ref> <file_name>AK_v8b2_win_x64_SSE3.exe</file_name> <main_program/> </file_ref> </app_version> <app_version> <app_name>setiathome_enhanced</app_name> <version_num>603</version_num> <platform>windows_x86_64</platform> <file_ref> <file_name>AK_v8b2_win_x64_SSE3.exe</file_name> <main_program/> </file_ref> </app_version> <app> <name>astropulse_v6</name> </app> <file_info> <name>AP6_win_x86_SSE2_OpenCL_NV_r1316.exe</name> <executable/> </file_info> <app_version> <app_name>astropulse_v6</app_name> <version_num>601</version_num> <avg_ncpus>0.04</avg_ncpus> <max_ncpus>0.2</max_ncpus> <platform>windows_intelx86</platform> <plan_class>cuda_fermi</plan_class> <cmdline>-instances_per_device 1 -unroll 10 -ffa_block 6144 -ffa_block_fetch 1536 -sbs 256</cmdline> <coproc> <type>CUDA</type> <count>0.51</count> </coproc> <file_ref> <file_name>AP6_win_x86_SSE2_OpenCL_NV_r1316.exe</file_name> <main_program/> </file_ref> </app_version> <app_version> <app_name>astropulse_v6</app_name> <version_num>601</version_num> <avg_ncpus>0.04</avg_ncpus> <max_ncpus>0.2</max_ncpus> <platform>windows_x86_64</platform> <plan_class>cuda_fermi</plan_class> <cmdline>-instances_per_device 1 -unroll 10 -ffa_block 6144 -ffa_block_fetch 1536 -sbs 256</cmdline> <coproc> <type>CUDA</type> <count>0.51</count> </coproc> <file_ref> <file_name>AP6_win_x86_SSE2_OpenCL_NV_r1316.exe</file_name> <main_program/> </file_ref> </app_version> <app> <name>setiathome_enhanced</name> </app> <file_info> <name>Lunatics_x41y_win32_cuda50.exe</name> <executable/> </file_info> <file_info> <name>cudart32_50_7.dll</name> <executable/> </file_info> <file_info> <name>cufft32_50_7.dll</name> <executable/> </file_info> <app_version> <app_name>setiathome_enhanced</app_name> <version_num>610</version_num> <platform>windows_intelx86</platform> <plan_class>cuda_fermi</plan_class> <avg_ncpus>0.040000</avg_ncpus> <max_ncpus>0.040000</max_ncpus> <coproc> <type>CUDA</type> <count>0.49</count> </coproc> <file_ref> <file_name>Lunatics_x41y_win32_cuda50.exe</file_name> <main_program/> </file_ref> <file_ref> <file_name>cudart32_50_7.dll</file_name> </file_ref> <file_ref> <file_name>cufft32_50_7.dll</file_name> </file_ref> </app_version> <app_version> <app_name>setiathome_enhanced</app_name> <version_num>609</version_num> <platform>windows_intelx86</platform> <plan_class>cuda23</plan_class> <avg_ncpus>0.040000</avg_ncpus> <max_ncpus>0.040000</max_ncpus> <coproc> <type>CUDA</type> <count>0.49</count> </coproc> <file_ref> <file_name>Lunatics_x41y_win32_cuda50.exe</file_name> <main_program/> </file_ref> <file_ref> <file_name>cudart32_50_7.dll</file_name> </file_ref> <file_ref> <file_name>cufft32_50_7.dll</file_name> </file_ref> </app_version> <app_version> <app_name>setiathome_enhanced</app_name> <version_num>608</version_num> <platform>windows_intelx86</platform> <plan_class>cuda</plan_class> <avg_ncpus>0.040000</avg_ncpus> <max_ncpus>0.040000</max_ncpus> <coproc> <type>CUDA</type> <count>0.49</count> </coproc> <file_ref> <file_name>Lunatics_x41y_win32_cuda50.exe</file_name> <main_program/> </file_ref> <file_ref> <file_name>cudart32_50_7.dll</file_name> </file_ref> <file_ref> <file_name>cufft32_50_7.dll</file_name> </file_ref> </app_version> <app_version> <app_name>setiathome_enhanced</app_name> <version_num>610</version_num> <platform>windows_x86_64</platform> <plan_class>cuda_fermi</plan_class> <avg_ncpus>0.040000</avg_ncpus> <max_ncpus>0.040000</max_ncpus> <coproc> <type>CUDA</type> <count>0.49</count> </coproc> <file_ref> <file_name>Lunatics_x41y_win32_cuda50.exe</file_name> <main_program/> </file_ref> <file_ref> <file_name>cudart32_50_7.dll</file_name> </file_ref> <file_ref> <file_name>cufft32_50_7.dll</file_name> </file_ref> </app_version> <app_version> <app_name>setiathome_enhanced</app_name> <version_num>609</version_num> <platform>windows_x86_64</platform> <plan_class>cuda23</plan_class> <avg_ncpus>0.040000</avg_ncpus> <max_ncpus>0.040000</max_ncpus> <coproc> <type>CUDA</type> <count>0.49</count> </coproc> <file_ref> <file_name>Lunatics_x41y_win32_cuda50.exe</file_name> <main_program/> </file_ref> <file_ref> <file_name>cudart32_50_7.dll</file_name> </file_ref> <file_ref> <file_name>cufft32_50_7.dll</file_name> </file_ref> </app_version> <app_version> <app_name>setiathome_enhanced</app_name> <version_num>608</version_num> <platform>windows_x86_64</platform> <plan_class>cuda</plan_class> <avg_ncpus>0.040000</avg_ncpus> <max_ncpus>0.040000</max_ncpus> <coproc> <type>CUDA</type> <count>0.49</count> </coproc> <file_ref> <file_name>Lunatics_x41y_win32_cuda50.exe</file_name> <main_program/> </file_ref> <file_ref> <file_name>cudart32_50_7.dll</file_name> </file_ref> <file_ref> <file_name>cufft32_50_7.dll</file_name> </file_ref> </app_version> </app_info>

____________

Profile Paul D Harris
Volunteer tester
Send message
Joined: 1 Dec 99
Posts: 1123
Credit: 33,598,472
RAC: 0
United States
Message 1257636 - Posted: 8 Jul 2012, 19:45:19 UTC
Last modified: 8 Jul 2012, 19:53:42 UTC

How does this look it does not mention (it would be better to include libfftw3f-3.dll in app_info) and I changed <max_ncpus>0.2</max_ncpus> to <max_ncpus>0.04</max_ncpus> and
I changed <count>0.51</count> to <count>0.5</count>
Do I need to include <cmdline>-instances_per_device 1 -unroll 10 -ffa_block 6144 -ffa_block_fetch 1536 -sbs 256</cmdline>

<app> <name>astropulse_v6</name> </app> <file_info> <name>AP6_win_x86_SSE2_OpenCL_NV_r1316.exe</name> <executable/> </file_info> <app_version> <app_name>astropulse_v6</app_name> <version_num>601</version_num> <avg_ncpus>0.04</avg_ncpus> <max_ncpus>0.04</max_ncpus> <platform>windows_intelx86</platform> <plan_class>cuda_fermi</plan_class> <cmdline>-instances_per_device 1 -unroll 10 -ffa_block 6144 -ffa_block_fetch 1536 -sbs 256</cmdline> <coproc> <type>CUDA</type> <count>0.5</count> </coproc> <file_ref> <file_name>AP6_win_x86_SSE2_OpenCL_NV_r1316.exe</file_name> <main_program/> </file_ref> </app_version> <app_version> <app_name>astropulse_v6</app_name> <version_num>601</version_num> <avg_ncpus>0.04</avg_ncpus> <max_ncpus>0.04</max_ncpus> <platform>windows_x86_64</platform> <plan_class>cuda_fermi</plan_class> <cmdline>-instances_per_device 1 -unroll 10 -ffa_block 6144 -ffa_block_fetch 1536 -sbs 256</cmdline> <coproc> <type>CUDA</type> <count>0.5</count> </coproc> <file_ref> <file_name>AP6_win_x86_SSE2_OpenCL_NV_r1316.exe</file_name> <main_program/> </file_ref> </app_version>

____________

Profile perryjay
Volunteer tester
Avatar
Send message
Joined: 20 Aug 02
Posts: 3377
Credit: 16,405,522
RAC: 9,692
United States
Message 1257637 - Posted: 8 Jul 2012, 19:45:55 UTC - in response to Message 1257630.
Last modified: 8 Jul 2012, 19:51:46 UTC

Okay, changed mine too and left it at 601. Also noticed that the old 560 was sse3 and the 1316 is sse2. Changed that also. Left everything else (unroll, ffa, ect) the same as I had it. Suspended a bunch of 603s to let a 601 start. Seemed to be running okay so I opened up the 603s again. It will be awhile before it restarts the 601 but it should be alright. I also changed my instances per to 2 and all my counts back to .5. My little GTS 450 should be able to handle it.

Thanks Arkayn.

Mine now looks like this

<app>
<name>astropulse_v6</name>
</app>
<file_info>
<name>AP6_win_x86_SSE2_OpenCL_NV_r1316.exe</name>
<executable/>
</file_info>
<app_version>
<app_name>astropulse_v6</app_name>
<version_num>601</version_num>
<flops>225.24e09</flops>
<avg_ncpus>0.04</avg_ncpus>
<max_ncpus>0.2</max_ncpus>
<plan_class>CUDA</plan_class>
<cmdline>-instances_per_device 2 -unroll 10 -ffa_block 8192 -ffa_block_fetch 4096</cmdline>
<coproc>
<type>CUDA</type>
<count>.5</count>
</coproc>
<file_ref>
<file_name>AP6_win_x86_SSE2_OpenCL_NV_r1316.exe</file_name>
<main_program/>
</file_ref>
</app_version>
____________


PROUD MEMBER OF Team Starfire World BOINC

Profile Raistmer
Volunteer developer
Volunteer tester
Avatar
Send message
Joined: 16 Jun 01
Posts: 3649
Credit: 49,424,765
RAC: 27,024
Russia
Message 1257659 - Posted: 8 Jul 2012, 20:29:18 UTC - in response to Message 1257621.


Raistmer, I'm still using the old 560. Would it be okay to just change the name over to 1316 and leave the version number at 601 rather than 604 in the app_info? Is it necessary to ask for 604s?

It's not nessasary to ask for 6.04 though it's real version of that binary.
Also, you need to remove </copy> flag from app_info regarding to CL file. Or even better - remove mention of CL file at all.

Profile Raistmer
Volunteer developer
Volunteer tester
Avatar
Send message
Joined: 16 Jun 01
Posts: 3649
Credit: 49,424,765
RAC: 27,024
Russia
Message 1257663 - Posted: 8 Jul 2012, 20:34:24 UTC - in response to Message 1257623.
Last modified: 8 Jul 2012, 20:44:29 UTC


What are the default settings (values) for/of the application if nothing is in the <cmdline> in the app_info.xml file for...

-unroll
-ffa_block
-ffa_block_fetch

?

- - - - - - - - - -

How I could find/know which settings (& values) are the best for the own system for max. performance?


Thanks

unroll=2, ffa_block=1024, ffa_block_fetch=512 are defaults.

Usual method can be used - to try and to see.
There are some heuristics though I and others used before. They still work for this new build.
Can't mention all of those here but understanding what these params change should give insights how to use them.
So:
-unroll - changes number of data chnks processed at once in main loop. To loow number will leave some of GPU's compute units idle, too big can cause issues with stride of memory access (depends of GPU memory controller and number of channels it uses).
-ffa_block - how many different periods calculated on GPU at once. Each such period corresponds to OpenCL workitem. So again, if this value too low this can result in part of GPU being idle. Too big values have other issue - more and more unneeded work will be done cause data arrays for different periods have different sizes but GPU will process rectangular area corresponding to longest of arrays in block. Also, r1316 includes FFA pre-compute. It means that whole block will be redone if GPU found some signals inside (signals are rare so in most cases this result in faster operation than checking if there was signal or not after each small step). Too big block will reduce performance in case of signal (and usually each task has at least 2-3 signals).

Profile perryjay
Volunteer tester
Avatar
Send message
Joined: 20 Aug 02
Posts: 3377
Credit: 16,405,522
RAC: 9,692
United States
Message 1257670 - Posted: 8 Jul 2012, 20:50:38 UTC - in response to Message 1257659.

</copy> flag? I don't see it. Are you talking to me? Only OpenCL reference I see is in the name.
____________


PROUD MEMBER OF Team Starfire World BOINC

Profile Paul D Harris
Volunteer tester
Send message
Joined: 1 Dec 99
Posts: 1123
Credit: 33,598,472
RAC: 0
United States
Message 1257672 - Posted: 8 Jul 2012, 20:51:56 UTC

None of the app_info mention libfftw3f-3.dll in app_info
Should it be included?
____________

Profile MikeProject donor
Volunteer tester
Avatar
Send message
Joined: 17 Feb 01
Posts: 25223
Credit: 34,842,581
RAC: 21,542
Germany
Message 1257685 - Posted: 8 Jul 2012, 21:09:24 UTC - in response to Message 1257672.
Last modified: 8 Jul 2012, 21:10:05 UTC

None of the app_info mention libfftw3f-3.dll in app_info
Should it be included?


Because we are using it for long time already.

For those running these apps for the fisrt time it avoids that boinc deletes it.
____________

Profile Paul D Harris
Volunteer tester
Send message
Joined: 1 Dec 99
Posts: 1123
Credit: 33,598,472
RAC: 0
United States
Message 1257705 - Posted: 8 Jul 2012, 21:43:45 UTC

So I added llibfftw3f-3.dll 3 times. Is this OK also I added the default <cmdline>unroll=2, ffa_block=1024, ffa_block_fetch=512</cmdline> what else I need to add?

<app> <name>astropulse_v6</name> </app> <file_info> <name>AP6_win_x86_SSE2_OpenCL_NV_r1316.exe</name> <executable/> </file_info> <file_info> <name>llibfftw3f-3.dll</name> <executable/> </file_info> <app_version> <app_name>astropulse_v6</app_name> <version_num>601</version_num> <avg_ncpus>0.04</avg_ncpus> <max_ncpus>0.04</max_ncpus> <platform>windows_intelx86</platform> <plan_class>cuda_fermi</plan_class> <cmdline>unroll=2, ffa_block=1024, ffa_block_fetch=512</cmdline> <coproc> <type>CUDA</type> <count>0.5</count> </coproc> <file_ref> <file_name>AP6_win_x86_SSE2_OpenCL_NV_r1316.exe</file_name> <main_program/> </file_ref> <file_info> <name>llibfftw3f-3.dll</name> <executable/> </file_info> </app_version> <app_version> <app_name>astropulse_v6</app_name> <version_num>601</version_num> <avg_ncpus>0.04</avg_ncpus> <max_ncpus>0.04</max_ncpus> <platform>windows_x86_64</platform> <plan_class>cuda_fermi</plan_class> <cmdline>-instances_per_device 1 -unroll 10 -ffa_block 6144 -ffa_block_fetch 1536 -sbs 256</cmdline> <coproc> <type>CUDA</type> <count>0.5</count> </coproc> <file_ref> <file_name>AP6_win_x86_SSE2_OpenCL_NV_r1316.exe</file_name> <main_program/> </file_ref> <file_info> <name>llibfftw3f-3.dll</name> <executable/> </file_info> </app_version>

____________

Profile perryjay
Volunteer tester
Avatar
Send message
Joined: 20 Aug 02
Posts: 3377
Credit: 16,405,522
RAC: 9,692
United States
Message 1257707 - Posted: 8 Jul 2012, 21:53:13 UTC - in response to Message 1257705.

Paul, the dll file goes in the SETI at Home folder. It doesn't need to be mentioned in the app_info.
____________


PROUD MEMBER OF Team Starfire World BOINC

Profile MikeProject donor
Volunteer tester
Avatar
Send message
Joined: 17 Feb 01
Posts: 25223
Credit: 34,842,581
RAC: 21,542
Germany
Message 1257710 - Posted: 8 Jul 2012, 21:57:56 UTC - in response to Message 1257707.

Paul, the dll file goes in the SETI at Home folder. It doesn't need to be mentioned in the app_info.


Wrong Perryjay.

Regarding what to mention in app_info file: for stock users upgrading to this release it would be better to include libfftw3f-3.dll in app_info.
That's because stock version could mention it already. If you have this file in project dir then you need to mention it to save it from deletion. It's BOINC's paranoid "kill all I know" approach - quite useless cause nothing prevents to accumulate not known to BOINC trash in that folders.

____________

Profile perryjay
Volunteer tester
Avatar
Send message
Joined: 20 Aug 02
Posts: 3377
Credit: 16,405,522
RAC: 9,692
United States
Message 1257714 - Posted: 8 Jul 2012, 22:04:40 UTC - in response to Message 1257710.

Oops, sorry, I missed that since I was coming from the old 560.

Well, I finished the first one. It validated but got a lot of warnings in the stddr. Not sure if I did something wrong or not. http://setiathome.berkeley.edu/result.php?resultid=2515639344
____________


PROUD MEMBER OF Team Starfire World BOINC

Profile Raistmer
Volunteer developer
Volunteer tester
Avatar
Send message
Joined: 16 Jun 01
Posts: 3649
Credit: 49,424,765
RAC: 27,024
Russia
Message 1257715 - Posted: 8 Jul 2012, 22:04:49 UTC - in response to Message 1257670.

</copy> flag? I don't see it. Are you talking to me? Only OpenCL reference I see is in the name.

Then you already removed it before, no additional changes required.

Previous · 1 · 2 · 3 · 4 · 5 . . . 11 · Next

Message boards : Number crunching : New AstroPulse for GPU ( ATi & NV) released (r1316)

Copyright © 2014 University of California