| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: GPU Support Forum
Thread: [RESOLVED] Pls help debug this app_info for HD6850 on 1090T |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 49
|
|
| Author |
|
|
twilyth
Master Cruncher US Joined: Mar 30, 2007 Post Count: 2130 Status: Offline Project Badges: 
|
This may not be representative since this is an APU chip where the gpu is part of the cpu, but I just set up my HTPC with the gpu only version and it's only running one thread.
----------------------------------------It running Win 7 Home Premium 64 bit. The chip is a Llano A8-3850 which has a Radeon 6550D gpu core. This reports in device manager as a Sumo 9640. I set the count parameter to .25 and the avg. param to 1 since it's a 4 core chip with no hyper threading (finally realized those have to match up). edit: I don't really care too much about this one. I'll report in tomorrow on one of the 2600k's with a 7950 - assuming I got everything right finally.   [Edit 1 times, last edit by twilyth at Nov 14, 2012 12:11:49 AM] |
||
|
|
twilyth
Master Cruncher US Joined: Mar 30, 2007 Post Count: 2130 Status: Offline Project Badges:
|
ok, I still can't get this to work. Here is the code. It is running in a 2600k with win7 ultimate. I have avg parameter at .5 since according the instructions that is supposed to be cores (4) over desired threads (8). I have the count parameter set at .125 since that is 1/8. What am I doing wrong?
----------------------------------------<app_info> <app> <name>hcc1</name> <user_friendly_name>Help Conquer Cancer</user_friendly_name> </app> <file_info> <name>wcg_hcc1_img_7.05_windows_intelx86__ati_hcc1</name> <executable/> </file_info> <file_info> <name>hcckernel.cl.7.05</name> <executable/> </file_info> <app_version> <app_name>hcc1</app_name> <version_num>705</version_num> <platform>windows_intelx86</platform> <plan_class>ati_hcc1</plan_class> <avg_ncpus>.5</avg_ncpus> <max_ncpus>1.0</max_ncpus> <coproc> <type>ATI</type> <count>.125</count> </coproc> <file_ref> <file_name>wcg_hcc1_img_7.05_windows_intelx86__ati_hcc1</file_name> <main_program/> </file_ref> <file_ref> <file_name>hcckernel.cl.7.05</file_name> <open_name>hcckernel.cl</open_name> </file_ref> </app_version> </app_info> edit - for the record, I let the queue run to empty, I closed BOINC which also seems to shutdown the service, restarted BOINC and only then had it download new tasks. [Edit 1 times, last edit by twilyth at Nov 15, 2012 9:00:48 AM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Closing the BOINC Manager also closing down the core client or for short client [boinc.exe] ** without asking is when you set the BM to remember a prior decision. You can restore being asked in the Options screen of the BM Tools menu. The BM is only a GUI interface to interact with the client.
** [which for GPU crunching cant be running as a service on Windows, except XP], |
||
|
|
twilyth
Master Cruncher US Joined: Mar 30, 2007 Post Count: 2130 Status: Offline Project Badges:
|
Yes. Thank you but the first time I selected the 2 check boxes. One for not to be asked again and the other to unload and close all boinc processes.
----------------------------------------edit - however I just double checked and although it does close all of the running apps and the service, it does leave the boinctray.exe app running. Would that be enough to cause a problem? [Edit 1 times, last edit by twilyth at Nov 15, 2012 9:50:46 AM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
boinctray is an entirely different thing. Only for Windows, it is a little app that detects user input and signals this to the core client, to then operate per prefs / activity menu setting. It always runs, in fact it's launched before anything else BOINC to do that very detection when the client comes on.
Just a personal thing, but exiting the BM whilst having one or more CEP2 job running, hours past last checkpoint, is crunchtime-costly. So I like to be asked first, specially in distracted moments. Anyway not using the BM, when it all can be done much smarter with BOINCTasks. The BM has BTW an option on the Advanced menu to shut down the connected client, that is the one local or remote. Silly though, there you sit with a BM, and no client running, without an option to restart the local client... you'd have to exit the BM and start it again to get the client to launch. BOINC idiosyncrasies. But to the root of the discussion, gathering that after the restart you've now got 8 concurrently running on the 7950? |
||
|
|
twilyth
Master Cruncher US Joined: Mar 30, 2007 Post Count: 2130 Status: Offline Project Badges:
|
No. I'm sorry if I wasn't clear about that.
----------------------------------------I don't know what BOINC tasks is. I do a plain vanilla BOINC install taking all of the defaults. The first time I shut down BM it asked if I wanted to be asked in the future. I told it I did not. It asked if I wanted to also in the future completely stop all BOINC processes. I said that I did. When I restarted BM, I still had just one GPU task. I have since changed my profile to load other projects on the remaining cores but only a quarter day's worth of work so that I won't have to terminate many wu's when it is time to retry this. |
||
|
|
wplachy
Senior Cruncher Joined: Sep 4, 2007 Post Count: 423 Status: Offline |
I have, with one param exception, the same app_info.xml file controlling 2 BOINC clients and they are each running 8 GPU WUs at a time on a single card each.
----------------------------------------The only difference between the one you posted and mine is the <avg_ncpus> values. In mine I have <avg_ncpus>1.0</avg_ncpus> but I don't believe that would account for the single task problem. Have you verified that the app_info.xml file is recognized by the client? If it is you should have a line in the Event Log (stdoutdae.txt) about 15 lines after client start that looks like this: [World Community Grid] Found app_info.xml; using anonymous platform Also, when you are using the app_info.xml file you should be getting a message like this each time the server sends you new work: [World Community Grid] App version returned from anonymous platform project; ignoring Are you seeing these messages?
Bill P
 |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
There's weirdness [inconsistency] where sometimes there's need of a leading zero when specifying a fraction for a parm in app_info, so maybe it is 0.5 rather than .5. (seeing at least max_ncpus which has a leading zero)
|
||
|
|
twilyth
Master Cruncher US Joined: Mar 30, 2007 Post Count: 2130 Status: Offline Project Badges:
|
wplachy: No on both counts and this time I reset the project, cleaned out the queue, set avg to 1 and even rebooted.
----------------------------------------I give up. [Edit 1 times, last edit by twilyth at Nov 15, 2012 8:27:31 PM] |
||
|
|
twilyth
Master Cruncher US Joined: Mar 30, 2007 Post Count: 2130 Status: Offline Project Badges:
|
I should have mentioned that I can find no stdoutdae.txt anywhere on my C drive and that the only other BOINC directory besides the one in Program Files is one called Boinc_Master in the Users directory.
----------------------------------------But I put the data\projects\www.worldcommunitygrid.org\app_info.xml file there too and restarted BM with no success there either. [Edit 1 times, last edit by twilyth at Nov 15, 2012 9:04:17 PM] |
||
|
|
|