commad line

Message boards : Number crunching : commad line
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile Bill G Special Project $75 donor
Avatar

Send message
Joined: 1 Jun 01
Posts: 1282
Credit: 187,688,550
RAC: 182
United States
Message 1871319 - Posted: 5 Jun 2017, 20:43:12 UTC
Last modified: 5 Jun 2017, 20:44:52 UTC

OK, probably a basic question: Where do you put the command line and with what do you put it there?
Windows system. 750ti

SETI@home classic workunits 4,019
SETI@home classic CPU time 34,348 hours
ID: 1871319 · Report as offensive
Profile Brent Norman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 1 Dec 99
Posts: 2786
Credit: 685,657,289
RAC: 835
Canada
Message 1871322 - Posted: 5 Jun 2017, 20:51:08 UTC - in response to Message 1871319.  

In your S@H data folder you will find cmdline_xxx_yyy.txt files which will be empty, they go there. There are also readme files for the rxxxx/cuda apps.

For Lunatics SoG it is mb_cmdline_win_x86_SSE3_OpenCL_NV_SoG.txt
ID: 1871322 · Report as offensive
Profile Tom M
Volunteer tester

Send message
Joined: 28 Nov 02
Posts: 5124
Credit: 276,046,078
RAC: 462
Message 1871353 - Posted: 5 Jun 2017, 22:26:36 UTC - in response to Message 1871322.  

In your S@H data folder you will find cmdline_xxx_yyy.txt files which will be empty, they go there. There are also readme files for the rxxxx/cuda apps.

For Lunatics SoG it is mb_cmdline_win_x86_SSE3_OpenCL_NV_SoG.txt


Still might be a little confusing. In stock Seti, in the hidden ProgramData/Boinc/projects/seti folder you are looking for a text files that starts with MB*SOG.txt

The minimum you should put in there is: -sbs 192 -spike_fft_thresh 2048 -tune 1 64 1 4

This all assumes you are getting SOG type gpu tasks.

If you are getting Mcuda32/40/50 tasks then you are looking for a file that ends in .cfg which is currently (almost certainly) empty. You should be able to find a file with a similar name that ends in *.sample You can delete the ; from in front of the last 3 lines, paste that text into the *.cfg file(s) and save.

The last step is either click on the "read config" files or restart Boinc.

Now if you are looking for "fancy" for the SOG command line, this may (or may not) produce some faster processing. Please note, this is for a non-dedicated machine. If you are looking to run it as a dedicated machine you probably will want some additional things.

-tt 700 -period_iterations_num 1 -sbs 1024 -spike_fft_thresh 4096 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 16 - ocfft_tune_wg 256 -oclfft_tune_ls 512 - oclfft_tune_bn 64 -oclfft_tune_cw 64

Yes, this is my current GTX 750 Ti as recommended to me by people who have researched this a lot more than me.

The other thing you will want is an "app_config.xml" file to cause one of your cpu cores to be dedicated to driving your Gpu at "full speed."

<app_config>
<app>
<name>setiathome_v8</name>
<gpu_versions>
<gpu_usage>1.0</gpu_usage>
<cpu_usage>1.0</cpu_usage>
</gpu_versions>
</app>
</app_config>

This is also how you arrange to run two tasks on the same Gpu at once. The catch is two tasks take at LEAST twice as long as one task. If they take much more than twice as long, the results are slower overall processing. The trick is to set gpu usage to "0.5" to get two tasks going. This will also tie up 2 cpu cores.

HTH,
Tom
A proud member of the OFA (Old Farts Association).
ID: 1871353 · Report as offensive
Profile Bill G Special Project $75 donor
Avatar

Send message
Joined: 1 Jun 01
Posts: 1282
Credit: 187,688,550
RAC: 182
United States
Message 1871408 - Posted: 6 Jun 2017, 3:21:18 UTC - in response to Message 1871353.  
Last modified: 6 Jun 2017, 4:07:37 UTC

Thanks, I will give it a try. The computer I am testing this on is a dedicated cruncher, at present running `1 task per GPU.

I also have another dedicated cruncher running 2 tasks per GPU. The computers are very similar and I have never been able to actually figure out which is faster (more APR) as it seems to be so close.

SETI@home classic workunits 4,019
SETI@home classic CPU time 34,348 hours
ID: 1871408 · Report as offensive
Profile Brent Norman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 1 Dec 99
Posts: 2786
Credit: 685,657,289
RAC: 835
Canada
Message 1871409 - Posted: 6 Jun 2017, 3:30:09 UTC - in response to Message 1871408.  

With my 750Ti's I found 2 tasks had a very marginal gain, BUT losing 1 more core for that small gain was definitely not worth it.
I ran 1 task with -hp flag so it got priority over the CPU.
ID: 1871409 · Report as offensive
Profile Bill G Special Project $75 donor
Avatar

Send message
Joined: 1 Jun 01
Posts: 1282
Credit: 187,688,550
RAC: 182
United States
Message 1871417 - Posted: 6 Jun 2017, 4:03:44 UTC - in response to Message 1871409.  
Last modified: 6 Jun 2017, 4:05:19 UTC

With my 750Ti's I found 2 tasks had a very marginal gain, BUT losing 1 more core for that small gain was definitely not worth it.
I ran 1 task with -hp flag so it got priority over the CPU.

Is that additional Brent?

I am getting an error in the start of the operation :
WARNING: can't open binary kernel file for oclFFT plan: C:\ProgramData\BOINC/projects/setiathome.berkeley.edu\MB_clFFTplan_GeForceGTX750Ti_524288_gr256_lr16_wg256_tw0_ls512_bn16_cw64_r3557.bin_36510, continue with recompile...
But the WU completed and validated with a copy of the previously listed command line.

More to check in the morning.......eyes starting to glaze over.

I am running 1 WU per GPU with 2 GPUs on the computer I am testing this on.

SETI@home classic workunits 4,019
SETI@home classic CPU time 34,348 hours
ID: 1871417 · Report as offensive
Profile Brent Norman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 1 Dec 99
Posts: 2786
Credit: 685,657,289
RAC: 835
Canada
Message 1871425 - Posted: 6 Jun 2017, 5:13:44 UTC - in response to Message 1871417.  

.... Is that additional Brent?

Yes, -hp added anywhere in the command line just puts the app at a higher priority than the CPU tasks to make sure the GPU doesn't get robbed of access time.
ID: 1871425 · Report as offensive
Profile Bill G Special Project $75 donor
Avatar

Send message
Joined: 1 Jun 01
Posts: 1282
Credit: 187,688,550
RAC: 182
United States
Message 1874194 - Posted: 21 Jun 2017, 3:35:15 UTC - in response to Message 1871353.  

In your S@H data folder you will find cmdline_xxx_yyy.txt files which will be empty, they go there. There are also readme files for the rxxxx/cuda apps.

For Lunatics SoG it is mb_cmdline_win_x86_SSE3_OpenCL_NV_SoG.txt


Still might be a little confusing. In stock Seti, in the hidden ProgramData/Boinc/projects/seti folder you are looking for a text files that starts with MB*SOG.txt

The minimum you should put in there is: -sbs 192 -spike_fft_thresh 2048 -tune 1 64 1 4

This all assumes you are getting SOG type gpu tasks.

If you are getting Mcuda32/40/50 tasks then you are looking for a file that ends in .cfg which is currently (almost certainly) empty. You should be able to find a file with a similar name that ends in *.sample You can delete the ; from in front of the last 3 lines, paste that text into the *.cfg file(s) and save.

The last step is either click on the "read config" files or restart Boinc.

Now if you are looking for "fancy" for the SOG command line, this may (or may not) produce some faster processing. Please note, this is for a non-dedicated machine. If you are looking to run it as a dedicated machine you probably will want some additional things.
-tt 700 -period_iterations_num 1 -sbs 1024 -spike_fft_thresh 4096 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 16 - ocfft_tune_wg 256 -oclfft_tune_ls 512 - oclfft_tune_bn 64 -oclfft_tune_cw 64
Yes, this is my current GTX 750 Ti as recommended to me by people who have researched this a lot more than me.
The other thing you will want is an "app_config.xml" file to cause one of your cpu cores to be dedicated to driving your Gpu at "full speed."
<app_config>
<app>
<name>setiathome_v8</name>
<gpu_versions>
<gpu_usage>1.0</gpu_usage>
<cpu_usage>1.0</cpu_usage>
</gpu_versions>
</app>
</app_config>
This is also how you arrange to run two tasks on the same Gpu at once. The catch is two tasks take at LEAST twice as long as one task. If they take much more than twice as long, the results are slower overall processing. The trick is to set gpu usage to "0.5" to get two tasks going. This will also tie up 2 cpu cores.
HTH,
Tom

I had been running this cmd line for a few weeks when we started to get APs.....the APs all failed with -9 overflows so I removed the cmd line and now the APs are validating.
(after proving that all was working I upgraded the computer to 1050 ti's from the 750 ti's that were in it. The computer is running much faster, or should I say WUs are processing much faster now.)
Do you think the same cmd line should be tried on the 1050's?

SETI@home classic workunits 4,019
SETI@home classic CPU time 34,348 hours
ID: 1874194 · Report as offensive
Profile Zalster Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 27 May 99
Posts: 5517
Credit: 528,817,460
RAC: 242
United States
Message 1874198 - Posted: 21 Jun 2017, 4:08:38 UTC - in response to Message 1874194.  

Command lines for SoG are not the same commandlines used for Astropulses. You have to have different commandlines for the 2 different types of work units
ID: 1874198 · Report as offensive
Mark Stevenson Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 8 Sep 11
Posts: 1736
Credit: 174,899,165
RAC: 91
United Kingdom
Message 1874224 - Posted: 21 Jun 2017, 6:40:05 UTC
Last modified: 21 Jun 2017, 6:41:07 UTC

-sbs 1024 -spike_fft_thresh 2048 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 16 -oclfft_tune_wg 256 -oclfft_tune_ls 512 -oclfft_tune_bn 32 -oclfft_tune_cw 32-sbs 768 -hp -high_perf -high_perc_timer -period_iterations_num 30

That's the Com line i use with my 1050ti's running 2 wu's at a time for SOG , Mike helped me with it when i got the cards and after a few days tweeking this is what iv'e been running on computer ID: 7420805
Ain't got a Com line for AP's but run times are ok on the 1050's so ain't to worried bout it
5822591291 2581573223 20 Jun 2017, 10:12:52 UTC 21 Jun 2017, 2:50:17 UTC Completed and validated 3,782.24 3,758.50 522.67 AstroPulse v7
Anonymous platform (NVIDIA GPU)

copy/paste of a AP's runtimes all are very similar
Life is what you make of it :-)

When i'm good i'm very good , but when i'm bad i'm shi#eloads better ;-) In't I " buttercups " p.m.s.l at authoritie !!;-)
ID: 1874224 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 13720
Credit: 208,696,464
RAC: 304
Australia
Message 1874225 - Posted: 21 Jun 2017, 6:48:35 UTC - in response to Message 1874224.  

-high_perc_timer

I think that should be -high_prec_timer
Grant
Darwin NT
ID: 1874225 · Report as offensive
Profile Bill G Special Project $75 donor
Avatar

Send message
Joined: 1 Jun 01
Posts: 1282
Credit: 187,688,550
RAC: 182
United States
Message 1874269 - Posted: 21 Jun 2017, 13:26:50 UTC - in response to Message 1874198.  

Command lines for SoG are not the same commandlines used for Astropulses. You have to have different commandlines for the 2 different types of work units

I did not think of that. But something happened to me.

SETI@home classic workunits 4,019
SETI@home classic CPU time 34,348 hours
ID: 1874269 · Report as offensive
Profile Bill G Special Project $75 donor
Avatar

Send message
Joined: 1 Jun 01
Posts: 1282
Credit: 187,688,550
RAC: 182
United States
Message 1874278 - Posted: 21 Jun 2017, 14:42:22 UTC - in response to Message 1874224.  

-sbs 1024 -spike_fft_thresh 2048 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 16 -oclfft_tune_wg 256 -oclfft_tune_ls 512 -oclfft_tune_bn 32 -oclfft_tune_cw 32-sbs 768 -hp -high_perf -high_perc_timer -period_iterations_num 30

That's the Com line i use with my 1050ti's running 2 wu's at a time for SOG , Mike helped me with it when i got the cards and after a few days tweeking this is what iv'e been running on computer ID: 7420805
Ain't got a Com line for AP's but run times are ok on the 1050's so ain't to worried bout it
5822591291 2581573223 20 Jun 2017, 10:12:52 UTC 21 Jun 2017, 2:50:17 UTC Completed and validated 3,782.24 3,758.50 522.67 AstroPulse v7
Anonymous platform (NVIDIA GPU)

copy/paste of a AP's runtimes all are very similar

Mark, I assume you are dedicating a CPU core to each WU?

SETI@home classic workunits 4,019
SETI@home classic CPU time 34,348 hours
ID: 1874278 · Report as offensive
Mark Stevenson Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 8 Sep 11
Posts: 1736
Credit: 174,899,165
RAC: 91
United Kingdom
Message 1874301 - Posted: 21 Jun 2017, 17:33:31 UTC - in response to Message 1874278.  
Last modified: 21 Jun 2017, 17:33:52 UTC

Mark, I assume you are dedicating a CPU core to each WU?


Yea for SOG i do with every machine I've got since I've been running the app , coz of the temps here I'm just running GPU so got plenty of " free " cores avalable

And thanks Grant for spotting the mistake in the com line , I'd totaly missed that but sorted it now ( might even run slightly better now hehe ;-) )
Life is what you make of it :-)

When i'm good i'm very good , but when i'm bad i'm shi#eloads better ;-) In't I " buttercups " p.m.s.l at authoritie !!;-)
ID: 1874301 · Report as offensive
Profile Bill G Special Project $75 donor
Avatar

Send message
Joined: 1 Jun 01
Posts: 1282
Credit: 187,688,550
RAC: 182
United States
Message 1874304 - Posted: 21 Jun 2017, 17:47:16 UTC - in response to Message 1874301.  

Thanks
and I did change the spelling although I am now getting this in the Stdrr: High-performance path selected. If GUI lags occur consider to remove -high_perf option from tuning line
and there is severe lag but it is only a cruncher so it does not matter.

SETI@home classic workunits 4,019
SETI@home classic CPU time 34,348 hours
ID: 1874304 · Report as offensive
Profile Bill G Special Project $75 donor
Avatar

Send message
Joined: 1 Jun 01
Posts: 1282
Credit: 187,688,550
RAC: 182
United States
Message 1874694 - Posted: 23 Jun 2017, 18:05:19 UTC - in response to Message 1874304.  

Thanks
and I did change the spelling although I am now getting this in the Stdrr: High-performance path selected. If GUI lags occur consider to remove -high_perf option from tuning line
and there is severe lag but it is only a cruncher so it does not matter.

I had to remove the - high_perf option as the system got so laggy that I was unable to do anything with my remote control (TeamViewer).

SETI@home classic workunits 4,019
SETI@home classic CPU time 34,348 hours
ID: 1874694 · Report as offensive

Message boards : Number crunching : commad line


 
©2024 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.