Welcome to MilkyWay@home

No GPU tasks?

Message boards : Number crunching : No GPU tasks?
Message board moderation

To post messages, you must log in.

Previous · 1 · 2

AuthorMessage
Zydor
Avatar

Send message
Joined: 24 Feb 09
Posts: 620
Credit: 100,587,625
RAC: 0
Message 47097 - Posted: 7 Apr 2011, 0:16:26 UTC

Ok Try this first.

Take out any app_info file from Project directories you have and any cc_config file you have. Open up Notepad (not others - use Notepad)and paste this into it:

<cc_config>
<options>
<ignore_ati_dev>0</ignore_ati_dev>
</options>
</cc_config>

Save the file as cc_config.xml in the BOINC Data Directory (go up the directory tree from wherever you have the Project Directories, its there).

You need to have BOINC 6.10.19 or above installed on the PC to use the ATI statement in cc_config.

Unless you need an app_info file for any specific Project information, doing that should make available any of your Projects you Crunch to Device 1 (the card) as cc_config says "ignore device zero". Likely as not thats it.... all done, just switch Projects inside BOINC Client as usual.

Might need an app_info for MW depending on how the server settles down, tackle that later, for now do the cc_config, and test.

Regards
Zy
ID: 47097 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Sunny129
Avatar

Send message
Joined: 25 Jan 11
Posts: 271
Credit: 346,072,284
RAC: 0
Message 47105 - Posted: 7 Apr 2011, 1:32:44 UTC - in response to Message 47097.  

well that's pretty much how i've been operating all along. like i said a few posts back, i have a cc_config.xml file in the BOINC data directory. it looks like this when i'm crunching S@H:
<cc_config>
<options>
<ignore_ati_dev>1</ignore_ati_dev>
</options>
</cc_config>

...and i edit it to look like this when i want to crunch MW@H:
<cc_config>
<options>
<ignore_ati_dev>0</ignore_ati_dev>
</options>
</cc_config>


while i do have an app_info.xml file in my S@H project folder, i do not have one in my MW@H project folder. its probably not pertinent, but just in case i'm wrong, it looks like this:
<app_info>
<app>
<name>setiathome_enhanced</name>
</app>
<file_info>
<name>AK_v8b_win_SSE3_AMD.exe</name>
<executable/>
</file_info>
<app_version>
<app_name>setiathome_enhanced</app_name>
<version_num>603</version_num>
<file_ref>
<file_name>AK_v8b_win_SSE3_AMD.exe</file_name>
<main_program/>
</file_ref>
</app_version>
<app>
<name>astropulse_v505</name>
</app>
<file_info>
<name>ap_5.05r409_SSE.exe</name>
<executable/>
</file_info>
<app_version>
<app_name>astropulse_v505</app_name>
<version_num>505</version_num>
<file_ref>
<file_name>ap_5.05r409_SSE.exe</file_name>
<main_program/>
</file_ref>
</app_version>
<app>
<name>astropulse_v505</name>
</app>
<file_info>
<name>ap_5.06_win_x86_SSE2_OpenCL_ATI_r456.exe</name>
<executable/>
</file_info>
<file_info>
<name>AstroPulse_Kernels.cl</name>
<executable/>
</file_info>
<app_version>
<app_name>astropulse_v505</app_name>
<version_num>506</version_num>
<avg_ncpus>0.01</avg_ncpus>
<max_ncpus>0.01</max_ncpus>
<plan_class>ati13ati</plan_class>
<cmdline>-ffa_block 8192 -ffa_block_fetch 2048</cmdline>
<coproc>
<type>ATI</type>
<count>1</count>
</coproc>
<flops>15987654321</flops>
<file_ref>
<file_name>ap_5.06_win_x86_SSE2_OpenCL_ATI_r456.exe</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>AstroPulse_Kernels.cl</file_name>
<copy_file/>
</file_ref>
</app_version>
<app>
<name>setiathome_enhanced</name>
</app>
<file_info>
<name>MB_6.10_win_SSE3_ATI_HD5_r177.exe</name>
<executable/>
</file_info>
<file_info>
<name>MultiBeam_Kernels.cl</name>
<executable/>
</file_info>
<app_version>
<app_name>setiathome_enhanced</app_name>
<version_num>610</version_num>
<platform>windows_intelx86</platform>
<avg_ncpus>0.05</avg_ncpus>
<max_ncpus>0.05</max_ncpus>
<plan_class>ati13ati</plan_class>
<cmdline>-period_iterations_num 2 -instances_per_device 1</cmdline>
<flops>20987654321</flops>
<file_ref>
<file_name>MB_6.10_win_SSE3_ATI_HD5_r177.exe</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>MultiBeam_Kernels.cl</file_name>
<copy_file/>
</file_ref>
<coproc>
<type>ATI</type>
<count>1</count>
</coproc>
</app_version>
</app_info>

i wouldn't think that having an app_info file in my S@H project folder should affect how MW@H runs. nevertheless, i removed it from the S@H project folder for the time being, put the "<ignore_ati_dev>0</ignore_ati_dev>" cc_config file in the BOINC data directory, and restarted BOINC. unfortunately it still will not fetch MW@H tasks for my GPU.
ID: 47105 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Zydor
Avatar

Send message
Joined: 24 Feb 09
Posts: 620
Credit: 100,587,625
RAC: 0
Message 47106 - Posted: 7 Apr 2011, 1:37:59 UTC - in response to Message 47105.  
Last modified: 7 Apr 2011, 1:42:01 UTC

yes, for now i only plan on crunching MW@H and S@H with the 5870 GPU (device 1).....


So if device zero is only used for the screen, and nothing to do with BOINC - all thats needed is tell BOINC to ignore Device Zero. Thats done in the cc_config file, from that point BOINC will default to Device One which is your Card.

After that its just a case of putting any app_info file needed for individual Projects. Its then all automatic, no messing around.

Am I missing something here re your intentions/outcome needed ?

EDIT:
Supply of MW WUs is down for the present - Travis is working on the server to resolve issues. If the cc_config works as expected, then put the app_info back to S&H, test that. If that works, we can prepare one for MW directory - Principle being change one thing at a time so you know where something might be wrong if it falls over.

Regards
Zy
ID: 47106 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Sunny129
Avatar

Send message
Joined: 25 Jan 11
Posts: 271
Credit: 346,072,284
RAC: 0
Message 47110 - Posted: 7 Apr 2011, 2:40:47 UTC - in response to Message 47106.  
Last modified: 7 Apr 2011, 2:50:32 UTC

So if device zero is only used for the screen, and nothing to do with BOINC - all thats needed is tell BOINC to ignore Device Zero. Thats done in the cc_config file, from that point BOINC will default to Device One which is your Card.

After that its just a case of putting any app_info file needed for individual Projects. Its then all automatic, no messing around.

Am I missing something here re your intentions/outcome needed ?

EDIT:
Supply of MW WUs is down for the present - Travis is working on the server to resolve issues. If the cc_config works as expected, then put the app_info back to S&H, test that. If that works, we can prepare one for MW directory - Principle being change one thing at a time so you know where something might be wrong if it falls over.

Regards
Zy

yeah, sorry if i wasn't clear about my situation the first time. up until ~24 hours ago, i had no problem crunching either S@H or MW@H GPU tasks. all i had to do was edit the simple cc_config.xml file to ignore either device 0 or device 1, depending on which project i want to work on. i know it seems counter-intuitive that ignoring device 1 (my 5870) would allow S@H to run properly on my 5870, but that's what worked for S@H. likewise, for MW@H, i simply had to ignore device 0 in the cc_config.xml file and everything would run fine.

so now that its clear that i was indeed already using a cc_config file (and thus already following the steps you laid out for me), my only issue is that i haven't been able to get MW@H GPU tasks at all in the past 24 hours. i understand that there have been server upgrade issues over the last 24 hours. but it appears as though every around me is at least getting tasks intermittently if not regularly. i'm subscribed to your other thread in which you posted up about the intermittent supply of work the instant you noticed it. then a few minutes later you had a full queue of 36 tasks. i know you ran out again and switched projects, but at least the MW@H work was coming intermittently. i haven't seen MW@H work in over 24 hours now. i'm just worried that something isn't right with my client all of the sudden, as unlikely as that may be. why wouldn't i get work for an entire 24 hours while everyone else is at least getting intermittent work? who knows...for all i know, the instant that i get completely fed up with it will be instant that my BOINC client starts fetching MW@H GPU work again. but i'm not counting on it.


*EDIT* - i just tried again to get some MW@H GPU work, but no luck:
4/6/2011 10:43:59 PM | Milkyway@home | resumed by user
4/6/2011 10:44:00 PM | Milkyway@home | work fetch resumed by user
4/6/2011 10:44:04 PM | Milkyway@home | Sending scheduler request: To fetch work.
4/6/2011 10:44:04 PM | Milkyway@home | Requesting new tasks for ATI GPU
4/6/2011 10:44:05 PM | Milkyway@home | Scheduler request completed: got 0 new tasks
4/6/2011 10:44:05 PM | Milkyway@home | No work available

the project status page shows all servers as "running," however there are only 2 tasks ready to send...so i understand why i didn't get work with this particular server query. i'll keep an eye on the project status page and try again when there's more work to go around.
ID: 47110 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Zydor
Avatar

Send message
Joined: 24 Feb 09
Posts: 620
Credit: 100,587,625
RAC: 0
Message 47111 - Posted: 7 Apr 2011, 3:04:03 UTC - in response to Message 47110.  
Last modified: 7 Apr 2011, 3:12:06 UTC

Ok - I understand now.

A pole to revolve around - well two in fact.

First, the way to control BOINC access to the card in slot one is that statement in cc_config. End of, as they say - you then use BOINC Client to switch projects in the normal way, no need to build in complexity. When the app_info files go into the Project Directories, then you will have told the Projects what you need. Thats it then - you are good to go - once the supply situation is resolved. Dont mix up the latter saga with your config, it will be fine when supply restarts.

Second, Re MW supply. Travis is working in the server, nothing will speed that up, it takes what it takes. He will be just as keen to get it going for us all. The on off supply you saw before was just the symptom of earlier work in progress, not an indication of a permanent fix.

Meanwhile we sit it out - there are no back doors, especially as Travis has rightly turned off supply in order to fix the server, cant do both at the same time.

I know its frustrating, but all we can do is wait it out - there's nothing you can do with your setup that will gain MW WUs any earlier. Once its fixed, away we go - until then its alternate Project time, hit the sack, or copious pots of coffee and a good DVD Movie :)

EDIT
I cheated, did two - DVD Movie and switched crunching for now :)

Regards
Zy
ID: 47111 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile arkayn
Avatar

Send message
Joined: 14 Feb 09
Posts: 999
Credit: 74,932,619
RAC: 0
Message 47116 - Posted: 7 Apr 2011, 3:42:00 UTC - in response to Message 47110.  

So if device zero is only used for the screen, and nothing to do with BOINC - all thats needed is tell BOINC to ignore Device Zero. Thats done in the cc_config file, from that point BOINC will default to Device One which is your Card.

After that its just a case of putting any app_info file needed for individual Projects. Its then all automatic, no messing around.

Am I missing something here re your intentions/outcome needed ?

EDIT:
Supply of MW WUs is down for the present - Travis is working on the server to resolve issues. If the cc_config works as expected, then put the app_info back to S&H, test that. If that works, we can prepare one for MW directory - Principle being change one thing at a time so you know where something might be wrong if it falls over.

Regards
Zy

yeah, sorry if i wasn't clear about my situation the first time. up until ~24 hours ago, i had no problem crunching either S@H or MW@H GPU tasks. all i had to do was edit the simple cc_config.xml file to ignore either device 0 or device 1, depending on which project i want to work on. i know it seems counter-intuitive that ignoring device 1 (my 5870) would allow S@H to run properly on my 5870, but that's what worked for S@H. likewise, for MW@H, i simply had to ignore device 0 in the cc_config.xml file and everything would run fine.

so now that its clear that i was indeed already using a cc_config file (and thus already following the steps you laid out for me), my only issue is that i haven't been able to get MW@H GPU tasks at all in the past 24 hours. i understand that there have been server upgrade issues over the last 24 hours. but it appears as though every around me is at least getting tasks intermittently if not regularly. i'm subscribed to your other thread in which you posted up about the intermittent supply of work the instant you noticed it. then a few minutes later you had a full queue of 36 tasks. i know you ran out again and switched projects, but at least the MW@H work was coming intermittently. i haven't seen MW@H work in over 24 hours now. i'm just worried that something isn't right with my client all of the sudden, as unlikely as that may be. why wouldn't i get work for an entire 24 hours while everyone else is at least getting intermittent work? who knows...for all i know, the instant that i get completely fed up with it will be instant that my BOINC client starts fetching MW@H GPU work again. but i'm not counting on it.


*EDIT* - i just tried again to get some MW@H GPU work, but no luck:
4/6/2011 10:43:59 PM | Milkyway@home | resumed by user
4/6/2011 10:44:00 PM | Milkyway@home | work fetch resumed by user
4/6/2011 10:44:04 PM | Milkyway@home | Sending scheduler request: To fetch work.
4/6/2011 10:44:04 PM | Milkyway@home | Requesting new tasks for ATI GPU
4/6/2011 10:44:05 PM | Milkyway@home | Scheduler request completed: got 0 new tasks
4/6/2011 10:44:05 PM | Milkyway@home | No work available

the project status page shows all servers as "running," however there are only 2 tasks ready to send...so i understand why i didn't get work with this particular server query. i'll keep an eye on the project status page and try again when there's more work to go around.


Sunny is running Brook on Milkyway, but OpenCL on SETI. Both see the GPU's in different ways and cause problems unless he changes which GPU to ignore.

ID: 47116 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Sunny129
Avatar

Send message
Joined: 25 Jan 11
Posts: 271
Credit: 346,072,284
RAC: 0
Message 47117 - Posted: 7 Apr 2011, 3:51:12 UTC

ok, i'll just leave BOINC configured the way it is b/c that's what's been working for me thus far. and thanks for the reassurance that it is simply a server issue. sorry if it seemed like i was growing impatient - that wasn't my intention, and i more than appreciate all who give their time and effort to keep this project running. honestly, i was more worried that something might be wrong on the client-end than i was worried about the server issues, simply b/c i wasn't even getting intermittent work, or so it seemed. but now i know that what i was seeing was actually "a symptom of earlier work in progress" as you explained above. so i'm sure there have been others besides me who have been out of work for the last 24 hours straight and just haven't had the time or cared to post about it. that at least makes me feel better about my situation while i wait this thing out.

i'm off to bed then...thanks for everything.

Eric
ID: 47117 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Sunny129
Avatar

Send message
Joined: 25 Jan 11
Posts: 271
Credit: 346,072,284
RAC: 0
Message 47118 - Posted: 7 Apr 2011, 3:59:58 UTC - in response to Message 47116.  
Last modified: 7 Apr 2011, 4:01:46 UTC

Sunny is running Brook on Milkyway, but OpenCL on SETI. Both see the GPU's in different ways and cause problems unless he changes which GPU to ignore.

yes, that is correct. and it was a pain in the ass figuring out how to configure the cc_config file by trial and error lol...

i also need an app_info file in my S@H project directory b/c the GPU apps for S@H MB & AP run on anonymous platforms. plus it needs to be there for tweaking such as running multiple simultaneous tasks and what not.
ID: 47118 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile arkayn
Avatar

Send message
Joined: 14 Feb 09
Posts: 999
Credit: 74,932,619
RAC: 0
Message 47119 - Posted: 7 Apr 2011, 4:15:20 UTC - in response to Message 47118.  

I did follow that thread over at SETI so I remember how much "fun" you had
ID: 47119 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Previous · 1 · 2

Message boards : Number crunching : No GPU tasks?

©2024 Astroinformatics Group