Author Topic: V10/11 of modified SETI MB CUDA + opt AP package for full multi-GPU+CPU use (Read 230266 times)

sunu · « **Reply #315 on:** 07 Apr 2009, 11:53:51 pm »

Quote from: codeman05 on 07 Apr 2009, 11:21:44 pm

I did resets the project to see if that would make a difference...I now have 11 WU's running at once rather than 8.

You're ok, that's how it should be, 8 CPU + 3 GPU.

codeman05 · « **Reply #316 on:** 08 Apr 2009, 12:01:23 am »

Quote from: sunu on 07 Apr 2009, 11:53:51 pm

Quote from: codeman05 on 07 Apr 2009, 11:21:44 pm
I did resets the project to see if that would make a difference...I now have 11 WU's running at once rather than 8.

You're ok, that's how it should be, 8 CPU + 3 GPU.

Thanks Sunu.

BOINC just started telling me that there aren't any AP wu's available, so I guess thats why I'm only getting MBs right now.
The MB's are flying though. About 7 mins RT for each GPU's and ~22min for each CPU core.

sunu · « **Reply #317 on:** 08 Apr 2009, 12:09:37 am »

Quote from: codeman05 on 08 Apr 2009, 12:01:23 am

BOINC just started telling me that there aren't any AP wu's available, so I guess thats why I'm only getting MBs right now.
The MB's are flying though. About 7 mins RT for each GPU's and ~22min for each CPU core.

Happy crunching then. Watch boinc to see when you get AP workunits.

Claggy · « **Reply #318 on:** 08 Apr 2009, 09:00:38 am »

Quote from: codeman05 on 07 Apr 2009, 11:21:44 pm

Hi Sunu,

Thanks for the help.

I am running 6.6.20, sorry left that out before.
Sounds good, I was thinking it was something to get with the angle, but since it happened at the exact same time I switched over. I wasn't sure.

I did resets the project to see if that would make a difference...I now have 11 WU's running at once rather than 8.

here is my AP info

Code: [Select]
- <app_info> - <app> <name>astropulse</name> </app> - <file_info> <name>ap_5.00r103_SSE3.exe</name> <executable /> </file_info> - <app_version> <app_name>astropulse</app_name> <version_num>500</version_num> - <file_ref> <file_name>ap_5.00r103_SSE3.exe</file_name> <main_program /> </file_ref> </app_version> - <app> <name>astropulse_v5</name> </app> - <file_info> <name>ap_5.03r112_SSE3.exe</name> <executable /> </file_info> - <app_version> <app_name>astropulse_v5</app_name> <version_num>503</version_num> - <file_ref> <file_name>ap_5.03r112_SSE3.exe</file_name> <main_program /> </file_ref> </app_version> - <app> <name>setiathome_enhanced</name> </app> - <file_info> <name>AK_v8b_win_x64_SSSE3x_GPU_CPU_team_V10.exe</name> <executable /> </file_info> - <file_info> <name>cudart.dll</name> <executable /> </file_info> - <file_info> <name>cufft.dll</name> <executable /> </file_info> - <file_info> <name>libfftw3f-3-1-1a_upx.dll</name> <executable /> </file_info> - <file_info> <name>MB_6.08_mod_CUDA_V10.exe</name> <executable /> </file_info> - <app_version> <app_name>setiathome_enhanced</app_name> <version_num>528</version_num> - <file_ref> <file_name>AK_v8b_win_x64_SSSE3x_GPU_CPU_team_V10.exe</file_name> <main_program /> </file_ref> - <file_ref> <file_name>cudart.dll</file_name> </file_ref> - <file_ref> <file_name>cufft.dll</file_name> </file_ref> - <file_ref> <file_name>libfftw3f-3-1-1a_upx.dll</file_name> </file_ref> - <file_ref> <file_name>MB_6.08_mod_CUDA_V10.exe</file_name> </file_ref> </app_version> - <app_version> <app_name>setiathome_enhanced</app_name> <version_num>603</version_num> - <file_ref> <file_name>AK_v8b_win_x64_SSSE3x_GPU_CPU_team_V10.exe</file_name> <main_program /> </file_ref> - <file_ref> <file_name>cudart.dll</file_name> </file_ref> - <file_ref> <file_name>cufft.dll</file_name> </file_ref> - <file_ref> <file_name>libfftw3f-3-1-1a_upx.dll</file_name> </file_ref> - <file_ref> <file_name>MB_6.08_mod_CUDA_V10.exe</file_name> </file_ref> </app_version> - <app_version> <app_name>setiathome_enhanced</app_name> <version_num>607</version_num> - <file_ref> <file_name>AK_v8b_win_x64_SSSE3x_GPU_CPU_team_V10.exe</file_name> <main_program /> </file_ref> - <file_ref> <file_name>cudart.dll</file_name> </file_ref> - <file_ref> <file_name>cufft.dll</file_name> </file_ref> - <file_ref> <file_name>libfftw3f-3-1-1a_upx.dll</file_name> </file_ref> - <file_ref> <file_name>MB_6.08_mod_CUDA_V10.exe</file_name> </file_ref> </app_version> - <app_version> <app_name>setiathome_enhanced</app_name> <version_num>608</version_num> - <file_ref> <file_name>AK_v8b_win_x64_SSSE3x_GPU_CPU_team_V10.exe</file_name> <main_program /> </file_ref> - <file_ref> <file_name>cudart.dll</file_name> </file_ref> - <file_ref> <file_name>cufft.dll</file_name> </file_ref> - <file_ref> <file_name>libfftw3f-3-1-1a_upx.dll</file_name> </file_ref> - <file_ref> <file_name>MB_6.08_mod_CUDA_V10.exe</file_name> </file_ref> </app_version> </app_info>

A couple of things:

1: Boinc 6.6.x no longer needs the <ncpus> workaround, that was just for Boinc 6.4.x, Boinc 6.6.x now knows about the GPU's fitted, and so doesn't need this entry.

2: Your app_info has been opened and saved in Internet Explorer, you should only use notepad to edit it, it now has an extra space (known as Claggy's Space)
every time there is a <main_program /> or <executable /> entry, should be <main_program/> and <executable/>, also has lots of - entries that shouldn't be there.

3: The Team mode was get round problems with Boinc 6.4.x, you should now use V10 app stand alone, i.e. without the team part, and use a normal AK_V8 app for the CPU,
See this thread for a Proper app_info for Boinc 6.6.x: app_info for AP500, AP503, MB603 and MB608

Claggy

codeman05 · « **Reply #319 on:** 08 Apr 2009, 09:08:14 am »

So basically to sum up, delete the CC_Config file since I no longer need the ncpu's flag. Install the V10 stand-alone app, and create a new app_info file using the link you provided.

In regards to the third item on your list, where can I find the V10 stand-alone app? Did a search but did not get any results back.

Thanks for the help Claggy, I'll give this a shot tonight!

Claggy · « **Reply #320 on:** 08 Apr 2009, 09:17:09 am »

Quote from: codeman05 on 08 Apr 2009, 09:08:14 am

So basically to sum up, delete the CC_Config file since I no longer need the ncpu's flag. Install the V10 stand-alone app, and create a new app_info file using the link you provided.

In regards to the third item on your list, where can I find the V10 stand-alone app? Did a search but did not get any results back.

Thanks Claggy, I'll give this a shot tonight!

If you've been running the V10 team package, you already have it, you just need to re-use the .dll's and the 'MB_6.08_mod_CUDA_V10.exe' file,
or you could use the V11 app from this post: MB_6.08_mod_CUDA_V11_def_func_FFTW_ESTIMATE_update which has the same file name for easy upgrading!

Claggy

codeman05 · « **Reply #321 on:** 08 Apr 2009, 09:48:50 pm »

Hey guys,
I think I got everything finally running correctly...atleast initially.

Did a fresh reinstall of 6.6.20. Installed the V10 packages in post one of this thread, then copying in the V11 CUDA package. Using the stock app_info that came with the V10 pkg.

Now when I started up the project after all of this, I got 3 CUDA MB wu's and 8 AP units. My 3 GPU's knocked out the MB's in about 10 mins each. So great, CPU's and GPU's running 100%...greatness!

However once the MB units were finished, I was unable to get more.
I just have all 8 AP's running.

This is what I see in the BOINC log:

Quote

4/8/2009 8:43:29 PM   SETI@home   [sched_op_debug] Starting scheduler request
4/8/2009 8:43:29 PM   SETI@home   Sending scheduler request: To fetch work.
4/8/2009 8:43:29 PM   SETI@home   Requesting new tasks
4/8/2009 8:43:29 PM   SETI@home   [sched_op_debug] CPU work request: 0.00 seconds; 0 idle CPUs
4/8/2009 8:43:29 PM   SETI@home   [sched_op_debug] CUDA work request: 864000.86 seconds; 3 idle GPUs
4/8/2009 8:43:34 PM   SETI@home   Scheduler request completed: got 0 new tasks
4/8/2009 8:43:34 PM   SETI@home   [sched_op_debug] Server version 607
4/8/2009 8:43:34 PM   SETI@home   Message from server: No work sent
4/8/2009 8:43:34 PM   SETI@home   Message from server: No work is available for Astropulse
4/8/2009 8:43:34 PM   SETI@home   Project requested delay of 11 seconds
4/8/2009 8:43:34 PM   SETI@home   [sched_op_debug] Deferring communication for 11 sec
4/8/2009 8:43:34 PM   SETI@home   [sched_op_debug] Reason: requested by project
4/8/2009 8:45:50 PM   SETI@home   [sched_op_debug] Starting scheduler request
4/8/2009 8:45:50 PM   SETI@home   Sending scheduler request: To fetch work.
4/8/2009 8:45:50 PM   SETI@home   Requesting new tasks
4/8/2009 8:45:50 PM   SETI@home   [sched_op_debug] CPU work request: 0.00 seconds; 0 idle CPUs
4/8/2009 8:45:50 PM   SETI@home   [sched_op_debug] CUDA work request: 864000.86 seconds; 3 idle GPUs
4/8/2009 8:45:55 PM   SETI@home   Scheduler request completed: got 0 new tasks
4/8/2009 8:45:55 PM   SETI@home   [sched_op_debug] Server version 607
4/8/2009 8:45:55 PM   SETI@home   Message from server: No work sent
4/8/2009 8:45:55 PM   SETI@home   Message from server: No work is available for Astropulse

Do I just need to give it more time, or could I have something off in the install?

AP wu's also look to be going at a 13hr-16 hour pace. I think it was about 40 hours before, so that's a nice improvment.

Raistmer · « **Reply #322 on:** 11 Apr 2009, 03:46:06 am »

non-teamed AK_v8b for AMD SSE3-level CPUs (for both x86 and x64 usage).

[attachment deleted by admin]

KarVi · « **Reply #323 on:** 11 Apr 2009, 04:39:08 am »

Quote from: Raistmer on 11 Apr 2009, 03:46:06 am

non-teamed AK_v8b for AMD SSE3-level CPUs (for both x86 and x64 usage).

I have been trying to find info earlier in thread, but found nothing.

What are the advantages of this build?

From the name I gather its normal MB, not CUDA? Is it faster than previously released ak_v8b?

Raistmer · « **Reply #324 on:** 11 Apr 2009, 05:21:41 am »

Quote from: KarVi on 11 Apr 2009, 04:39:08 am

Quote from: Raistmer on 11 Apr 2009, 03:46:06 am
non-teamed AK_v8b for AMD SSE3-level CPUs (for both x86 and x64 usage).

I have been trying to find info earlier in thread, but found nothing.

What are the advantages of this build?

From the name I gather its normal MB, not CUDA? Is it faster than previously released ak_v8b?

As I can remember there was no v8b for AMD available in non-teamed flavor. Only pre-release ones in closed section. This one should have same performance as pre-release (but it's fresh rebuild - small speed deviation in both sides possible). It can be used as "standalone" CPU MB opt app for AMD CPUs. Posted in this thread as replacement to "teamed" version for those who will do V10 "team" ->BOINC 6.6.20 transition.

KarVi · « **Reply #325 on:** 11 Apr 2009, 05:37:24 am »

OK.

So if I'm allready running 6.6.20 with the official ak_v8b SSE3_AMD (which I believe is publicly available somewhere), there should be no difference?

I'll stick to that then :-)

Raistmer · « **Reply #326 on:** 11 Apr 2009, 06:09:41 am »

Quote from: KarVi on 11 Apr 2009, 05:37:24 am

So if I'm allready running 6.6.20 with the official ak_v8b SSE3_AMD (which I believe is publicly available somewhere), there should be no difference?

Fine if available. There should be no difference in speed. Only CPUID features were changed recently.

n7rfa · « **Reply #327 on:** 11 Apr 2009, 11:19:22 am »

I seem to have a problem. I am running 6.6.20 and the v10 applications. I only see 2 tasks running in BOINC and when I check the Task Manager, I see 1 MB task running at 50%, 1 CUDA task running 0-1%, and 1 MB task with nothing loaded at 0%. (I am not running AP.)

What have I done wrong?

Thanks,

Here is my app_info.xml:

<app_info>
<app>
<name>setiathome_enhanced</name>
</app>
<file_info>
<name>AK_v8b_win_SSE3_GPU_CPU_team_V10.exe</name>
<executable/>
</file_info>
<file_info>
<name>cudart.dll</name>
<executable/>
</file_info>
<file_info>
<name>cufft.dll</name>
<executable/>
</file_info>
<file_info>
<name>libfftw3f-3-1-1a_upx.dll</name>
<executable/>
</file_info>
<file_info>
<name>MB_6.08_mod_CUDA_V10.exe</name>
<executable/>
</file_info>

<app_version>
<app_name>setiathome_enhanced</app_name>
<version_num>528</version_num>
<file_ref>
<file_name>AK_v8b_win_SSE3_GPU_CPU_team_V10.exe</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>cudart.dll</file_name>
</file_ref>
<file_ref>
<file_name>cufft.dll</file_name>
</file_ref>
<file_ref>
<file_name>libfftw3f-3-1-1a_upx.dll</file_name>
</file_ref>
<file_ref>
<file_name>MB_6.08_mod_CUDA_V10.exe</file_name>
</file_ref>
</app_version>

<app_version>
<app_name>setiathome_enhanced</app_name>
<version_num>603</version_num>
<file_ref>
<file_name>AK_v8b_win_SSE3_GPU_CPU_team_V10.exe</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>cudart.dll</file_name>
</file_ref>
<file_ref>
<file_name>cufft.dll</file_name>
</file_ref>
<file_ref>
<file_name>libfftw3f-3-1-1a_upx.dll</file_name>
</file_ref>
<file_ref>
<file_name>MB_6.08_mod_CUDA_V10.exe</file_name>
</file_ref>
</app_version>

<app_version>
<app_name>setiathome_enhanced</app_name>
<version_num>607</version_num>
<file_ref>
<file_name>AK_v8b_win_SSE3_GPU_CPU_team_V10.exe</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>cudart.dll</file_name>
</file_ref>
<file_ref>
<file_name>cufft.dll</file_name>
</file_ref>
<file_ref>
<file_name>libfftw3f-3-1-1a_upx.dll</file_name>
</file_ref>
<file_ref>
<file_name>MB_6.08_mod_CUDA_V10.exe</file_name>
</file_ref>
</app_version>

<app_version>
<app_name>setiathome_enhanced</app_name>
<version_num>608</version_num>
<file_ref>
<file_name>AK_v8b_win_SSE3_GPU_CPU_team_V10.exe</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>cudart.dll</file_name>
</file_ref>
<file_ref>
<file_name>cufft.dll</file_name>
</file_ref>
<file_ref>
<file_name>libfftw3f-3-1-1a_upx.dll</file_name>
</file_ref>
<file_ref>
<file_name>MB_6.08_mod_CUDA_V10.exe</file_name>
</file_ref>
</app_version>

</app_info>

Here is my local_prefs_override.xml:

<global_preferences>
<run_on_batteries>1</run_on_batteries>
<run_if_user_active>1</run_if_user_active>
<run_gpu_if_user_active>1</run_gpu_if_user_active>
<start_hour>0.000000</start_hour>
<end_hour>0.000000</end_hour>
<net_start_hour>0.000000</net_start_hour>
<net_end_hour>0.000000</net_end_hour>
<leave_apps_in_memory>0</leave_apps_in_memory>
<confirm_before_connecting>0</confirm_before_connecting>
<hangup_if_dialed>0</hangup_if_dialed>
<dont_verify_images>0</dont_verify_images>
<work_buf_min_days>0.100000</work_buf_min_days>
<work_buf_additional_days>0.500000</work_buf_additional_days>
<max_ncpus_pct>100.000000</max_ncpus_pct>
<cpu_scheduling_period_minutes>60.000000</cpu_scheduling_period_minutes>
<disk_interval>60.000000</disk_interval>
<disk_max_used_gb>100.000000</disk_max_used_gb>
<disk_max_used_pct>50.000000</disk_max_used_pct>
<disk_min_free_gb>0.100000</disk_min_free_gb>
<vm_max_used_pct>75.000000</vm_max_used_pct>
<ram_max_used_busy_pct>50.000000</ram_max_used_busy_pct>
<ram_max_used_idle_pct>90.000000</ram_max_used_idle_pct>
<max_bytes_sec_up>0.000000</max_bytes_sec_up>
<max_bytes_sec_down>0.000000</max_bytes_sec_down>
<cpu_usage_limit>100.000000</cpu_usage_limit>
</global_preferences>

and here is my cc_config.xml:

<cc_config>
<log_flags>
<task>1</task>
<file_xfer>1</file_xfer>
<sched_ops>1</sched_ops>
<cpu_sched_debug>0</cpu_sched_debug>
<work_fetch_debug>0</work_fetch_debug>
</log_flags>
<options>
<save_stats_days>30</save_stats_days>
<report_results_immediately>1</report_results_immediately>
</options>
</cc_config>

cyclejon · « **Reply #328 on:** 11 Apr 2009, 12:18:04 pm »

Quote from: n7rfa on 11 Apr 2009, 11:19:22 am

I seem to have a problem. I am running 6.6.20 and the v10 applications. I only see 2 tasks running in BOINC and when I check the Task Manager, I see 1 MB task running at 50%, 1 CUDA task running 0-1%, and 1 MB task with nothing loaded at 0%. (I am not running AP.)

What have I done wrong?

Your app_info is from the team mod, it doesn't work with 6.6.* There are other posts in this thread about which files you need. As well as this thread on the seti forums http://setiathome.berkeley.edu/forum_thread.php?id=52589 for the correct app_info.

Raistmer · « **Reply #329 on:** 11 Apr 2009, 01:56:25 pm »

Quote from: cyclejon on 11 Apr 2009, 12:18:04 pm

Your app_info is from the team mod, it doesn't work with 6.6.*

V10 pack does work with any 6.6.x BOINC including 6.6.20. Actually I running it exactly with BOINC 6.6.20 right now.
But starting from BOINC 6.6.14 as was mentioned many times in this thread already there is possibility to use BOINC own scheduling mechanism for CUDA applications to run both CPU and GPU SETI MB versions simultaneously. That's why team pack can be disabled now.
But to do this you need replace app_info.xml _and_ SETI CPU MB "teamed" app ("teamed" AK_v8). Only SETI CUDA MB and SETI AP should be leaved untouched while upgrading from V10 pack to BOINC 6.6.20.
You need to delete cc_config.xml also (or just edit it to remove ncpus section if you use cc_config for some other purposes too).

Author Topic: V10/11 of modified SETI MB CUDA + opt AP package for full multi-GPU+CPU use (Read 230266 times)

sunu

Re: V10 of modified SETI MB CUDA + opt AP package for full multi-GPU+CPU use

codeman05

Re: V10 of modified SETI MB CUDA + opt AP package for full multi-GPU+CPU use

sunu

Re: V10 of modified SETI MB CUDA + opt AP package for full multi-GPU+CPU use

Claggy

Re: V10 of modified SETI MB CUDA + opt AP package for full multi-GPU+CPU use

codeman05

Re: V10 of modified SETI MB CUDA + opt AP package for full multi-GPU+CPU use

Claggy

Re: V10 of modified SETI MB CUDA + opt AP package for full multi-GPU+CPU use

codeman05

Re: V10 of modified SETI MB CUDA + opt AP package for full multi-GPU+CPU use

Raistmer

Re: V10 of modified SETI MB CUDA + opt AP package for full multi-GPU+CPU use

KarVi

Re: V10 of modified SETI MB CUDA + opt AP package for full multi-GPU+CPU use

Raistmer

Re: V10 of modified SETI MB CUDA + opt AP package for full multi-GPU+CPU use

KarVi

Re: V10 of modified SETI MB CUDA + opt AP package for full multi-GPU+CPU use

Raistmer

Re: V10 of modified SETI MB CUDA + opt AP package for full multi-GPU+CPU use

n7rfa

Re: V10 of modified SETI MB CUDA + opt AP package for full multi-GPU+CPU use

cyclejon

Re: V10 of modified SETI MB CUDA + opt AP package for full multi-GPU+CPU use

Raistmer

Re: V10 of modified SETI MB CUDA + opt AP package for full multi-GPU+CPU use