Advanced search

Message boards : Graphics cards (GPUs) : BOINC 6.3.17 is out

Author Message
Wolfram1
Send message
Joined: 24 Aug 08
Posts: 45
Credit: 3,431,862
RAC: 0
Level
Ala
Scientific publications
watwatwatwatwatwat
Message 3246 - Posted: 24 Oct 2008 | 10:01:16 UTC

BOINC 6.3.17 for windows is available.

I think the funczion of bersion .17 ist like the Version .16.

Here are the changes for 6.3.16

* client: if "leave in memory" pref changes, quit suspended tasks
* manager: fix disk display when talking to pre-6.3 client
* MGR: More thorough validity checking to prevent crashes
* client: restore support for max_ncpus preference. This and max_ncpus_pct can both be specified. If both are specified, we take the min
* LIB: Fix the windows callstack dump routine for Visual Studio 2008
* WINSETUP: Don't allow parent permissions to be inhertied from the parent directory when we setup the data directories permissions
* WINSETUP: Include a fix for the ALLUSERS flag so that based on the user preference, the installation settings are honored
* MGR: Have the SkinManager parse the correct spelling of 'organization'.
* MGR: Add a new tab status to the Simple GUI for when the client has been instructed to delay starting applications via the <delay_start> option in cc_config.xml
* WINSETUP: When the shortcut to BOINC Manager remove the minimize flag, it apparently was causing the small manager window. The '/s' parameter already takes care of minimizing the window
* LIB: The max cpu preferences was named 'max_cpus' and not 'max_ncpus' back in the day
* client: the max percent needs to be based on the max cpu preference not the total number of CPUs if we are actually calculating the min of both of them

Profile Stefan Ledwina
Avatar
Send message
Joined: 16 Jul 07
Posts: 464
Credit: 135,911,881
RAC: 1,088
Level
Cys
Scientific publications
watwatwatwatwatwatwatwat
Message 3247 - Posted: 24 Oct 2008 | 10:13:03 UTC - in response to Message 3246.

Seems to run fine on Vista64.
Let's see how it does after some time and some task switches...
____________

pixelicious.at - my little photoblog

Profile Krunchin-Keith [USA]
Avatar
Send message
Joined: 17 May 07
Posts: 512
Credit: 111,288,061
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 3248 - Posted: 24 Oct 2008 | 12:00:37 UTC

6.3.17 seems ok so far, But I've had little time to observe it. I installed it last night but had other things to do and have only been awake 10 minutes today.

6.3.16 had a display bug which I found within the first hour of testing it. Hence the quick release of 6.3.17 same day as 6.3.16.

There are not many scheduler changes in these over 6.3.15 so they should remain stable, most changes were minor tweeks or cosmetic and some fine tuning on the communication between manager and client and the screen redraw to reduce flicker. That was the cause of phanton lines appearing in task display or lines appearing with wrong status of task in 6.3.15 and 6.3.16, harmless but not desired for a public release. I've done quite a few extensive combinations of tasks, projects and loads in 6.3.15 with no signs of it not keeping the proper number of tasks running.

One thing semi-fixed is the startup of manager, like after exit or a reboot, previously if you started it closed, then opened it from the icon the window size was small, not the size you left it. Now it should remain the size you left it, but the position still drifts. At least it is easier to just move the window a little than have to move and resize each time. They will continue to try to figure out why windows can;t remember the position.

It is probably safe to run this version.

You should really wait for a release memo to appear on the alpha email list before posting a version is out on public message boards. 5 of the last 8 versions were not the best and some had major flaws.

Profile [SETI.USA]Tank_Master
Avatar
Send message
Joined: 8 Jul 07
Posts: 85
Credit: 67,463,387
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwat
Message 3254 - Posted: 24 Oct 2008 | 18:32:47 UTC
Last modified: 24 Oct 2008 | 19:25:20 UTC

This is more for krunchin Keith

When I start BOINC now, I see

10/24/2008 11:27:15 AM||Unrecognized tag in cc_config.xml: <checkpoint_debug>
10/24/2008 11:27:15 AM||Unrecognized tag in cc_config.xml: <coproc_debug>
10/24/2008 11:27:15 AM||Unrecognized tag in cc_config.xml: <cpu_sched>
10/24/2008 11:27:15 AM||Unrecognized tag in cc_config.xml: <cpu_sched_debug>

Also, I keep getting the dialog box to join a project every time the manager starts up (though this build it was the account manager). I already have it attached to BAM!, so why would it be asking to reattach? I did go through and rejoin the BAM! today and the BOINC manager window auto minimized on me as soon as I clicked finish.

Running on Server 2008 x64, installed the 64bit BOINC to its own directories (d:\Datax64 and d:\progx64)

both went away after restarting BOINC. this time. strange.

Now BOINC is quiting on me a few seconds after launching. The manager will stay open with "connecting to local host". If I leave it own, about a min later, it'll reconnect for about 1 sec, then again loose connection. In the task manager, I see BOINC.exe running only for that one second. I am going back to 6.3.15

15 did the same thing, went back to 14, and it *seems* stable again

Profile Krunchin-Keith [USA]
Avatar
Send message
Joined: 17 May 07
Posts: 512
Credit: 111,288,061
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 3259 - Posted: 24 Oct 2008 | 20:24:19 UTC - in response to Message 3254.

This is more for krunchin Keith

When I start BOINC now, I see

10/24/2008 11:27:15 AM||Unrecognized tag in cc_config.xml: <checkpoint_debug>
10/24/2008 11:27:15 AM||Unrecognized tag in cc_config.xml: <coproc_debug>
10/24/2008 11:27:15 AM||Unrecognized tag in cc_config.xml: <cpu_sched>
10/24/2008 11:27:15 AM||Unrecognized tag in cc_config.xml: <cpu_sched_debug>

Also, I keep getting the dialog box to join a project every time the manager starts up (though this build it was the account manager). I already have it attached to BAM!, so why would it be asking to reattach? I did go through and rejoin the BAM! today and the BOINC manager window auto minimized on me as soon as I clicked finish.

Running on Server 2008 x64, installed the 64bit BOINC to its own directories (d:\Datax64 and d:\progx64)

both went away after restarting BOINC. this time. strange.

Now BOINC is quiting on me a few seconds after launching. The manager will stay open with "connecting to local host". If I leave it own, about a min later, it'll reconnect for about 1 sec, then again loose connection. In the task manager, I see BOINC.exe running only for that one second. I am going back to 6.3.15

15 did the same thing, went back to 14, and it *seems* stable again

Hum...

First I suspect you have an unclosed or invalid tag in cc_config.xml before those, those are all valid tags and I have all those in my file. This might be causing all the problems.

As for x64 version it should be the same as x86, but since I do not have there could be some flaw I can't test for. I also do not use an account manager. This question would be better posted on the alpha email list.

But if we could check your cc_config first that might be a good starts, Could you PM me your entire cc_config.xml file.

bytzmaster
Send message
Joined: 4 Oct 08
Posts: 1
Credit: 72,993
RAC: 0
Level

Scientific publications
wat
Message 3261 - Posted: 24 Oct 2008 | 23:29:45 UTC

first, sorry 4 the bad english.
i had a problem with my dualcore-cpu. after upgrading from 6.3.15 > 6.3.17, i had only one cpu in the messagewindow of boinc.
then i have craete the cc_config file, but nothing.
so i have look into the file "global_prefs" and have change the line "<max_cpus>x</max_cpus>" (x= number of cpu´s)manual to 2 cpu´s.
now its all ok...i have my both cpu´s again 4 crunching ;D

I hope that I could help!

JKuehl2
Send message
Joined: 18 Jul 08
Posts: 33
Credit: 3,233,174
RAC: 0
Level
Ala
Scientific publications
watwatwatwatwat
Message 3264 - Posted: 25 Oct 2008 | 7:41:38 UTC - in response to Message 3261.

the underlying logic of cpu-s changed from ncpus to max_cpus in 6.3.17

this will be used for exacter scheduling in the next versions (according to boinc-dev mailinglist some are already trying with different implementations)

Profile Krunchin-Keith [USA]
Avatar
Send message
Joined: 17 May 07
Posts: 512
Credit: 111,288,061
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 3271 - Posted: 25 Oct 2008 | 12:45:02 UTC - in response to Message 3264.

the underlying logic of cpu-s changed from ncpus to max_cpus in 6.3.17

this will be used for exacter scheduling in the next versions (according to boinc-dev mailinglist some are already trying with different implementations)

Thats not exactly correct.

max_cpus was the old way before 6.X something, where in your preferences you specified # (1,2,4) of processors, to limit less than max usage. At some point it was changed to a percentage max_ncpus_pct where you specify this as a percentage 25%,50%,100% of the processors as reported by the os.

What has happened is some people run new clients on old projects only using the old preference way, not being able to set this, so the old way was added back, and combined. The client can use either way or both where it will take the minimum of the two if both are specified, always with a minimum of 1 as you cannot run with 0 cpus.

ncpus is the internal designation inside the client and also the debug flag to be used to simulate a higher number of processors for testing, so testers and developers could test the logic of having 4 cpus without actually having a quad-core.

See changeset 16281
David 22 Oct 2008
- client: restore support for max_ncpus preference.
This and max_ncpus_pct can both be specified.
If both are specified, we take the min.

Jayargh
Send message
Joined: 21 Dec 07
Posts: 47
Credit: 5,252,135
RAC: 0
Level
Ser
Scientific publications
watwatwatwatwat
Message 3273 - Posted: 25 Oct 2008 | 12:49:28 UTC

It would be really nice in these threads about BOINC clients if posters would specify whether they are talking about Windows or Linux clients. I am using Linux only and have to do a lot of scrolling around to see which platform is being discussed to see if it applies to my setups....Thanks

Profile Krunchin-Keith [USA]
Avatar
Send message
Joined: 17 May 07
Posts: 512
Credit: 111,288,061
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 3277 - Posted: 25 Oct 2008 | 14:41:08 UTC - in response to Message 3254.

This is more for krunchin Keith

When I start BOINC now, I see

10/24/2008 11:27:15 AM||Unrecognized tag in cc_config.xml: <checkpoint_debug>
10/24/2008 11:27:15 AM||Unrecognized tag in cc_config.xml: <coproc_debug>
10/24/2008 11:27:15 AM||Unrecognized tag in cc_config.xml: <cpu_sched>
10/24/2008 11:27:15 AM||Unrecognized tag in cc_config.xml: <cpu_sched_debug>

Also, I keep getting the dialog box to join a project every time the manager starts up (though this build it was the account manager). I already have it attached to BAM!, so why would it be asking to reattach? I did go through and rejoin the BAM! today and the BOINC manager window auto minimized on me as soon as I clicked finish.

Running on Server 2008 x64, installed the 64bit BOINC to its own directories (d:\Datax64 and d:\progx64)

both went away after restarting BOINC. this time. strange.

Now BOINC is quiting on me a few seconds after launching. The manager will stay open with "connecting to local host". If I leave it own, about a min later, it'll reconnect for about 1 sec, then again loose connection. In the task manager, I see BOINC.exe running only for that one second. I am going back to 6.3.15

15 did the same thing, went back to 14, and it *seems* stable again

Do you run boinc as a service ?

I have no problem, but someone else is now reporting that when running as a service on windows XP (32bit) version that the service is cutting out after 30 seconds.

Profile Krunchin-Keith [USA]
Avatar
Send message
Joined: 17 May 07
Posts: 512
Credit: 111,288,061
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 3278 - Posted: 25 Oct 2008 | 15:02:29 UTC - in response to Message 3273.

It would be really nice in these threads about BOINC clients if posters would specify whether they are talking about Windows or Linux clients. I am using Linux only and have to do a lot of scrolling around to see which platform is being discussed to see if it applies to my setups....Thanks

Well, since there is no linux 6.3.17 client yet....I think we are talking about windows. Most options apply equally to all clients though.

First post says ...for windows...

Another post says vista64

Another says Server 2008 x64 (That is Windows)

Profile [SETI.USA]Tank_Master
Avatar
Send message
Joined: 8 Jul 07
Posts: 85
Credit: 67,463,387
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwat
Message 3281 - Posted: 25 Oct 2008 | 16:55:31 UTC - in response to Message 3277.

cant run this project as a service, so no, its not.

Profile [SETI.USA]Tank_Master
Avatar
Send message
Joined: 8 Jul 07
Posts: 85
Credit: 67,463,387
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwat
Message 3282 - Posted: 25 Oct 2008 | 17:25:31 UTC

Others on my team are asking why every thing they start BOINC a window pops up to attach to a project.

Profile Stefan Ledwina
Avatar
Send message
Joined: 16 Jul 07
Posts: 464
Credit: 135,911,881
RAC: 1,088
Level
Cys
Scientific publications
watwatwatwatwatwatwatwat
Message 3284 - Posted: 25 Oct 2008 | 17:30:40 UTC - in response to Message 3282.
Last modified: 25 Oct 2008 | 17:31:28 UTC

Others on my team are asking why every thing they start BOINC a window pops up to attach to a project.


Weird, I only had that with 6.3.8, or maybe it was 6.3.10.
6.3.17 runs fine here with Vista 64 SP1. I haven't seen any problems `til now...

What exactly are they doing to get the attach to project wizard pop up?
____________

pixelicious.at - my little photoblog

Profile [SETI.USA]Tank_Master
Avatar
Send message
Joined: 8 Jul 07
Posts: 85
Credit: 67,463,387
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwat
Message 3286 - Posted: 25 Oct 2008 | 17:38:46 UTC
Last modified: 25 Oct 2008 | 17:39:16 UTC

starting the BOINC manager

Profile X-Files 27
Avatar
Send message
Joined: 11 Oct 08
Posts: 95
Credit: 68,023,693
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 3288 - Posted: 25 Oct 2008 | 17:52:21 UTC - in response to Message 3282.

Others on my team are asking why every thing they start BOINC a window pops up to attach to a project.


if they are upgrading, most likely the registry was corrupted.

My installation is always manual - extract the boinc files by /a switch
placed the extracted files in one location. Then the data files in other location. etc..etc

You must have some knowledge about MSI inorder to do manual installs. Like understading the Custom Actions, Services, etc...

Here's the registry (keyword INSTALLDIR and DATADIR):
[HKEY_LOCAL_MACHINE\SOFTWARE\Space Sciences Laboratory, U.C. Berkeley\BOINC Setup]
"INSTALLDIR"="D:\\BOINC\\"
"LAUNCHPROGRAM"=""
"ENABLELAUNCHATLOGON"=""
"ENABLESCREENSAVER"=""
"SERVICE_DOMAIN"=""
"SERVICE_USERNAME"=""
"SETUPSTATESTORED"="TRUE"
"MIGRATIONVERSION"=""
"MIGRATION"="TRUE"
"MIGRATIONDIR"="D:\\BOINC\\"
"RETURN_REBOOTREQUESTED"="0"
"REBOOTPROMPT"=""
"RETURN_VALIDATEINSTALL"="1"
"DATADIR"="D:\\BOINC Data\\"
"ENABLEPROTECTEDAPPLICATIONEXECUTION"=""
"ENABLEUSEBYALLUSERS"="1"

Profile DoctorNow
Avatar
Send message
Joined: 18 Aug 07
Posts: 83
Credit: 122,995,082
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 3289 - Posted: 25 Oct 2008 | 18:07:41 UTC - in response to Message 3284.

Others on my team are asking why every thing they start BOINC a window pops up to attach to a project.


Weird, I only had that with 6.3.8, or maybe it was 6.3.10.

Yes, I had that with the 6.3.10, too, every time I restarted Windows and BOINC the Attach-Window did come up. I first thought it could be a missed setting during the installation, but it wasn't.
After changing to the 6.3.17 all is fine now.
____________
Member of BOINC@Heidelberg and ATA!

fractal
Send message
Joined: 16 Aug 08
Posts: 87
Credit: 1,248,879,715
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 3291 - Posted: 25 Oct 2008 | 18:44:40 UTC - in response to Message 3289.

Others on my team are asking why every thing they start BOINC a window pops up to attach to a project.


Weird, I only had that with 6.3.8, or maybe it was 6.3.10.

Yes, I had that with the 6.3.10, too, every time I restarted Windows and BOINC the Attach-Window did come up. I first thought it could be a missed setting during the installation, but it wasn't.
After changing to the 6.3.17 all is fine now.

Same here. I had 6.3.10 on this machine that I use to check the rest of my grove. Every time I selected a new machine, it would pop up a window reminding me to use BAM to select a project for most of my machines. Some would not do that weirdness, I guess since they were to old. Anyway, 6.3.17 doesn't do this any more.

Profile [SETI.USA]Tank_Master
Avatar
Send message
Joined: 8 Jul 07
Posts: 85
Credit: 67,463,387
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwat
Message 3294 - Posted: 25 Oct 2008 | 19:18:04 UTC

I had 6.3.17 ask to attach to BAM! every time it restarted... got old real quick. Id see the windows "communicating with BAM!" then it would ask for my account info. I finally resubmitted the account info (even though in the message logs I would see it had already connected)and the message never came back.

Profile Krunchin-Keith [USA]
Avatar
Send message
Joined: 17 May 07
Posts: 512
Credit: 111,288,061
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 3297 - Posted: 25 Oct 2008 | 19:24:12 UTC - in response to Message 3281.
Last modified: 25 Oct 2008 | 19:27:30 UTC

cant run this project as a service, so no, its not.

I thought that applied only to Windows Vista.
I had 6.3.17 ask to attach to BAM! every time it restarted... got old real quick. Id see the windows "communicating with BAM!" then it would ask for my account info. I finally resubmitted the account info (even though in the message logs I would see it had already connected)and the message never came back.

So is 6.3.17 still crashing or is it running now ?

Profile Krunchin-Keith [USA]
Avatar
Send message
Joined: 17 May 07
Posts: 512
Credit: 111,288,061
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 3298 - Posted: 25 Oct 2008 | 19:31:22 UTC - in response to Message 3284.

Others on my team are asking why every thing they start BOINC a window pops up to attach to a project.


Weird, I only had that with 6.3.8, or maybe it was 6.3.10.
6.3.17 runs fine here with Vista 64 SP1. I haven't seen any problems `til now...

What exactly are they doing to get the attach to project wizard pop up?

Same for me, I only had that in 6.3.10. When up grading to 6.3.14, 6.3.15, 6.3.16 and 6.3.17 I never saw it again on 3 windows computers.

When I did have it all I did was hit cancel and the window would go away and client would then load all existing projects and go on like it never asked that.

Profile X-Files 27
Avatar
Send message
Joined: 11 Oct 08
Posts: 95
Credit: 68,023,693
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 3332 - Posted: 25 Oct 2008 | 23:36:23 UTC - in response to Message 3298.
Last modified: 25 Oct 2008 | 23:56:44 UTC

One of my rig with v6.3.17 doesnt have any work..

178.13
Vistax64

Also I got this messages (lots of them):
25-Oct-2008 18:17:16 [---] Internal error: expected process to be executing

After reboot so far the message didnt came back...

edit: changed my resource share to 100 from .2 - now im getting task.
So the resource share plays some role? I thought no matter what your settings, GPU will always run a task.

Profile Krunchin-Keith [USA]
Avatar
Send message
Joined: 17 May 07
Posts: 512
Credit: 111,288,061
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 3338 - Posted: 26 Oct 2008 | 2:08:09 UTC - in response to Message 3332.

One of my rig with v6.3.17 doesnt have any work..

178.13
Vistax64

Also I got this messages (lots of them):
25-Oct-2008 18:17:16 [---] Internal error: expected process to be executing

After reboot so far the message didnt came back...

edit: changed my resource share to 100 from .2 - now im getting task.
So the resource share plays some role? I thought no matter what your settings, GPU will always run a task.

Yes, GPU should always run tasks, but tasks are fetched by a separate process, work fetch. The scheduler can't run tasks if work fetch does not get any for it to run. I'm not sure exactly how resource shares play a role in that.

As for the internal error, I saw that too today, for the first time in any client. it came after I had tried to test the graphics button, not sure if that was the cause, but after a restart of the client I have not seen it again, have not tried graphics since then either.

Profile Krunchin-Keith [USA]
Avatar
Send message
Joined: 17 May 07
Posts: 512
Credit: 111,288,061
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 3339 - Posted: 26 Oct 2008 | 2:09:50 UTC

Suddenly too I am seeing bad behavior in 6.3.17 which was not in the 6.3.15 client, I think a new bug has crept in. A moment ago I brought up manager to find the first PS3GRID CPU<1 running (has been for 5 hours), all looked fine, had two other cpu tasks running too (my max). Now one of those two CPU has quit without me touching anything, another did not start, which is not correct. Also too the CPU<1 is running at low priority and in earlier tests all CPU<1 ran at normal priority. If I now suspend PS3GRID CUDA tasks, my CUDA tasks from test project won't start which didn't happen before. Suddenly I got lots of internal errors logged. something is out of whack in this client. After a restart the CUDA runs at normal priority. I see some debugging needs to be done, but that will have to wait until tomorrow. I've been up for 16 hours already today. I'll also have to inquire about tweeks made in 6.3.17 over 6.3.15 to see if we can narrow this down.

Profile Stefan Ledwina
Avatar
Send message
Joined: 16 Jul 07
Posts: 464
Credit: 135,911,881
RAC: 1,088
Level
Cys
Scientific publications
watwatwatwatwatwatwatwat
Message 3345 - Posted: 26 Oct 2008 | 7:45:19 UTC - in response to Message 3339.
Last modified: 26 Oct 2008 | 7:48:03 UTC

Hi Keith,

I just have set up a cc_config.xml, but BOINC won't recognize it. It only shows missing start tag in cc_config.xml.

The cc_config looks like this one -

<cc_config>
<log_flags>
<checkpoint_debug>1</checkpoint_debug>
<coproc_debug>1</coproc_debug>
<cpu_sched>1</cpu_sched>
<cpu_sched_debug>1</cpu_sched_debug>
</log_flags>
</cc_config>

and is placed in the BOINC Data dir.

Any ideas what's wrong?

[edit] Nevermind... I saved it in UTF-8, now as ANSI it works...
____________

pixelicious.at - my little photoblog

Profile K1atOdessa
Send message
Joined: 25 Feb 08
Posts: 249
Credit: 370,320,941
RAC: 0
Level
Asp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 3347 - Posted: 26 Oct 2008 | 8:19:23 UTC - in response to Message 3339.

Suddenly too I am seeing bad behavior in 6.3.17 which was not in the 6.3.15 client, I think a new bug has crept in. A moment ago I brought up manager to find the first PS3GRID CPU<1 running (has been for 5 hours), all looked fine, had two other cpu tasks running too (my max). Now one of those two CPU has quit without me touching anything, another did not start, which is not correct.


Similar problem with the 6.3.17 (reported in the <1 CPU thread). Went back to 6.3.14, 3 CPU tasks + 2 GPU tasks running for 10 minutes so far (1 CPU went to "Waiting to run" within seconds on 6.3.17). I'll watch this for a while in 6.3.14 to see if it ever goes back again for me (to rule out <1 CPU issue versus 6.3.17 issue).

http://www.gpugrid.net/forum_thread.php?id=453&nowrap=true#3342

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 3356 - Posted: 26 Oct 2008 | 10:20:45 UTC

Something strange is going on. Made the switch from 6.3.10 to 6.3.17 yesterday and it's running 4 CPU + 1 GPU task, as expected. But my log is spammed with entries like this:


26/10/2008 09:46:55|QMC@HOME|Resuming task three_114_peptidsm-ecp2.13554_0 using Amolqc-preRC1 version 501
26/10/2008 09:48:02|QMC@HOME|Resuming task three_114_peptidsm-ecp2.13554_0 using Amolqc-preRC1 version 501
26/10/2008 09:49:09|QMC@HOME|Resuming task three_114_peptidsm-ecp2.13554_0 using Amolqc-preRC1 version 501
26/10/2008 09:50:14|QMC@HOME|Resuming task three_114_peptidsm-ecp2.13554_0 using Amolqc-preRC1 version 501
26/10/2008 09:51:19|QMC@HOME|Resuming task three_114_peptidsm-ecp2.13554_0 using Amolqc-preRC1 version 501


It's not doing this all the time, but really a lot. I have no idea what this means.

MrS
____________
Scanning for our furry friends since Jan 2002

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 3361 - Posted: 26 Oct 2008 | 11:10:59 UTC

Stefan@home wrote:
From the logs it looked like it has to do with the preemption of tasks, but I'm not a dev, I only report what I see... ;)


Yes, but BOINC didn't stop the tasks. Or if it did it's not reporting it.

MrS
____________
Scanning for our furry friends since Jan 2002

Profile Stefan Ledwina
Avatar
Send message
Joined: 16 Jul 07
Posts: 464
Credit: 135,911,881
RAC: 1,088
Level
Cys
Scientific publications
watwatwatwatwatwatwatwat
Message 3362 - Posted: 26 Oct 2008 | 11:37:58 UTC - in response to Message 3361.
Last modified: 26 Oct 2008 | 11:38:25 UTC

Stefan@home wrote:
From the logs it looked like it has to do with the preemption of tasks, but I'm not a dev, I only report what I see... ;)


Yes, but BOINC didn't stop the tasks. Or if it did it's not reporting it.

MrS


Actually I was talking about my logs in the other thread. I'm sorry that I haven't stated that... ;)

Here's what I saw when BOINC switched from 4CPU plus 1 GPU to 3 CPU plus 1 GPU tasks:

26.10.2008 08:55:09||[cpu_sched_debug] CPU efficiency old 0.805795 new 0.805822 wall 40.809601 CPU 35.211000 w 0.999528 e 0.862812
26.10.2008 08:55:09|PS3GRID|Master file download succeeded
26.10.2008 08:55:10|Docking@Home|Finished download of 1m0b_mod0013sc_69336_87792.inp
26.10.2008 08:55:10||[cpu_sched_debug] CPU efficiency old 0.805822 new 0.805823 wall 4.367996 CPU 3.628000 w 0.999949 e 0.830587
26.10.2008 08:55:11||[cpu_sched_debug] Request CPU reschedule: files downloaded
26.10.2008 08:55:11||[cpu_sched_debug] schedule_cpus(): start
26.10.2008 08:55:11||[cpu_sched_debug] CPU efficiency old 0.805823 new 0.805826 wall 4.243202 CPU 3.650000 w 0.999951 e 0.860199
26.10.2008 08:55:11|PS3GRID|[cpu_sched_debug] highest debt: 86400.000000 fH25969-GPUTEST4-9-10-acemd_0
26.10.2008 08:55:11|PS3GRID|[cpu_sched_debug] already reserved coprocessors for fH25969-GPUTEST4-9-10-acemd_0
26.10.2008 08:55:11|PS3GRID|[cpu_sched_debug] scheduling fH25969-GPUTEST4-9-10-acemd_0
26.10.2008 08:55:11|PS3GRID|[cpu_sched_debug] highest debt: 84685.714286 Ms17952-GPUTEST4-7-10-acemd_0
26.10.2008 08:55:11||[cpu_sched_debug] rr_sim: insufficient coproc CUDA (1 + 1 > 1)
26.10.2008 08:55:11|PS3GRID|[cpu_sched_debug] insufficient coprocessors for Ms17952-GPUTEST4-7-10-acemd_0
26.10.2008 08:55:11|PS3GRID|[cpu_sched_debug] highest debt: 84685.714286 md16645-GPUTEST4-2-10-acemd_0
26.10.2008 08:55:11||[cpu_sched_debug] rr_sim: insufficient coproc CUDA (1 + 1 > 1)
26.10.2008 08:55:11|PS3GRID|[cpu_sched_debug] insufficient coprocessors for md16645-GPUTEST4-2-10-acemd_0
26.10.2008 08:55:11|PS3GRID|[cpu_sched_debug] highest debt: 84685.714286 Rzx1100-GPUTEST4-8-10-acemd_0
26.10.2008 08:55:11||[cpu_sched_debug] rr_sim: insufficient coproc CUDA (1 + 1 > 1)
26.10.2008 08:55:11|PS3GRID|[cpu_sched_debug] insufficient coprocessors for Rzx1100-GPUTEST4-8-10-acemd_0
26.10.2008 08:55:11|Docking@Home|[cpu_sched_debug] highest debt: 30.228069 1m0b_mod0013sc_70074_87792_0
26.10.2008 08:55:11|Docking@Home|[cpu_sched_debug] scheduling 1m0b_mod0013sc_70074_87792_0
26.10.2008 08:55:11|Docking@Home|[cpu_sched_debug] highest debt: -312.629074 1m0b_mod0013sc_70553_87792_0
26.10.2008 08:55:11|Docking@Home|[cpu_sched_debug] scheduling 1m0b_mod0013sc_70553_87792_0
26.10.2008 08:55:11|Docking@Home|[cpu_sched_debug] highest debt: -655.486217 1m0b_mod0013sc_67283_87792_0
26.10.2008 08:55:11|Docking@Home|[cpu_sched_debug] scheduling 1m0b_mod0013sc_67283_87792_0
26.10.2008 08:55:11|Docking@Home|[cpu_sched_debug] highest debt: -998.343360 1m0b_mod0013sc_69336_87792_0
26.10.2008 08:55:11|Docking@Home|[cpu_sched_debug] scheduling 1m0b_mod0013sc_69336_87792_0
26.10.2008 08:55:11||[cpu_sched_debug] Request enforce CPU schedule: schedule_cpus
26.10.2008 08:55:11||[cpu_sched_debug] enforce_schedule(): start
26.10.2008 08:55:11|PS3GRID|[cpu_sched_debug] want to run: fH25969-GPUTEST4-9-10-acemd_0
26.10.2008 08:55:11|Docking@Home|[cpu_sched_debug] want to run: 1m0b_mod0013sc_70074_87792_0
26.10.2008 08:55:11|Docking@Home|[cpu_sched_debug] want to run: 1m0b_mod0013sc_70553_87792_0
26.10.2008 08:55:11|Docking@Home|[cpu_sched_debug] want to run: 1m0b_mod0013sc_67283_87792_0
26.10.2008 08:55:11|Docking@Home|[cpu_sched_debug] want to run: 1m0b_mod0013sc_69336_87792_0
26.10.2008 08:55:11|PS3GRID|[cpu_sched_debug] processing fH25969-GPUTEST4-9-10-acemd_0
26.10.2008 08:55:11|Docking@Home|[cpu_sched_debug] processing 1m0b_mod0013sc_70074_87792_0
26.10.2008 08:55:11|Docking@Home|[cpu_sched_debug] didn't preempt wu_102208_154906_2_2_0: tr 4962.414900 tsc 649.491398
26.10.2008 08:55:11|Docking@Home|[cpu_sched_debug] processing 1m0b_mod0013sc_70553_87792_0
26.10.2008 08:55:11|Docking@Home|[cpu_sched_debug] didn't preempt wu_102208_154906_2_2_0: tr 4962.414900 tsc 649.491398
26.10.2008 08:55:11|Docking@Home|[cpu_sched_debug] processing 1m0b_mod0013sc_67283_87792_0
26.10.2008 08:55:11|Docking@Home|[cpu_sched_debug] didn't preempt wu_102208_154906_2_2_0: tr 4962.414900 tsc 649.491398
26.10.2008 08:55:11|Docking@Home|[cpu_sched_debug] processing 1m0b_mod0013sc_69336_87792_0
26.10.2008 08:55:11|Docking@Home|[cpu_sched_debug] didn't preempt wu_102208_154906_2_2_0: tr 4962.414900 tsc 649.491398
26.10.2008 08:55:11||[cpu_sched_debug] finished preempt loop, ncpus_used 3.900000
26.10.2008 08:55:11||[cpu_sched_debug] using 3.900000 out of 4 CPUs
26.10.2008 08:55:11|Einstein@Home|[cpu_sched_debug] h1_0600.55_S5R4__224_S5R4a_1 sched state 2 next 2 task state 1
26.10.2008 08:55:11|PS3GRID|[cpu_sched_debug] fH25969-GPUTEST4-9-10-acemd_0 sched state 2 next 2 task state 1
26.10.2008 08:55:11|Cosmology@Home|[cpu_sched_debug] wu_102208_154906_2_2_0 sched state 2 next 2 task state 1
26.10.2008 08:55:11|Einstein@Home|[cpu_sched_debug] h1_0600.55_S5R4__222_S5R4a_1 sched state 2 next 2 task state 1
26.10.2008 08:55:11|Cosmology@Home|[cpu_sched_debug] wu_102208_154924_0_1_0 sched state 1 next 1 task state 9
26.10.2008 08:55:11||[cpu_sched_debug] enforce_schedule: end
____________

pixelicious.at - my little photoblog

Profile Krunchin-Keith [USA]
Avatar
Send message
Joined: 17 May 07
Posts: 512
Credit: 111,288,061
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 3364 - Posted: 26 Oct 2008 | 13:56:55 UTC - in response to Message 3347.

Suddenly too I am seeing bad behavior in 6.3.17 which was not in the 6.3.15 client, I think a new bug has crept in. A moment ago I brought up manager to find the first PS3GRID CPU<1 running (has been for 5 hours), all looked fine, had two other cpu tasks running too (my max). Now one of those two CPU has quit without me touching anything, another did not start, which is not correct.


Similar problem with the 6.3.17 (reported in the <1 CPU thread). Went back to 6.3.14, 3 CPU tasks + 2 GPU tasks running for 10 minutes so far (1 CPU went to "Waiting to run" within seconds on 6.3.17). I'll watch this for a while in 6.3.14 to see if it ever goes back again for me (to rule out <1 CPU issue versus 6.3.17 issue).

http://www.gpugrid.net/forum_thread.php?id=453&nowrap=true#3342

I know 6.3.14 has problems, it will not work correctly, sooner or later. You can run 6.3.14 if you want to, but it is not perfect.

6.3.15 was next to perfect, when I tested it, as far as scheduling, it has some screen redraw flaws. There is a Windows only version.

At this time I recommend users to use 6.3.10 for linux and 6.3.10 or 6.3.15 for Windows.

If you have not upgraded to 6.3.17, do not.

Some new bug has been introduced into 6.3.17 when running CPU<1. It is not harmful, only leaves you with wrong amount of tasks runing and shows some kind of internal error. once this happens you can only quit and restart, but it will then happen again.

Profile K1atOdessa
Send message
Joined: 25 Feb 08
Posts: 249
Credit: 370,320,941
RAC: 0
Level
Asp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 3366 - Posted: 26 Oct 2008 | 14:32:47 UTC - in response to Message 3364.
Last modified: 26 Oct 2008 | 15:14:00 UTC


I know 6.3.14 has problems, it will not work correctly, sooner or later. You can run 6.3.14 if you want to, but it is not perfect.

6.3.15 was next to perfect, when I tested it, as far as scheduling, it has some screen redraw flaws. There is a Windows only version.

At this time I recommend users to use 6.3.10 for linux and 6.3.10 or 6.3.15 for Windows.


I'll try 6.3.15.

*UPDATE* 6.3.15 went back to 4 tasks (2 CPU, 2 GPU) shortly after install. I will try 6.3.10, before again trying 6.3.14 again (which ran fine with 3 CPU and 2 GPU tasks for 12+ hours).

Profile Krunchin-Keith [USA]
Avatar
Send message
Joined: 17 May 07
Posts: 512
Credit: 111,288,061
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 3393 - Posted: 26 Oct 2008 | 22:11:53 UTC

I was going to do some testing today, but I'm so burned out, I gave up early this morning.

It was not running correct number of tasks this morning.

I quit boinc and re-started.

Funny thing is, since then, 8 hours ago, it has worked perfectly. At least every time I walk by and looked. Checking message log does not show that internal error. I'm afraid to touch anything else, all I have done since then is switch tabs a few times, messages and tasks. I have not touched anything else in it.

I think what I will do is wait until tomorrow. When I go to work I will check my two systems there running this version. If they are fowled up and show errors, i'll downgrade one to 6.3.16 and one to 6.3.15. Then see how long they run and if I can reproduce the error in those versions. I'm pretty sure its not in 6.3.15. 6.3.16 I only ran about 4 hours or so before 6.3.17 came out. This should help determine when the bug crawled in and may help find out what it is.

AngleDroit
Send message
Joined: 2 Sep 08
Posts: 3
Credit: 358,759
RAC: 0
Level

Scientific publications
wat
Message 3414 - Posted: 27 Oct 2008 | 14:23:41 UTC
Last modified: 27 Oct 2008 | 14:24:47 UTC

Hi,

I'm running 6.3.17 on Vista 64 with a E6750 and 2 GTX280 in SLI mode.

I only run two projects: SETI and GPUGRID
In task manager, i see the three tasks as planned, 2 SETI at low priority and 1 GPU at normal priority.
The CPU usage, tho, is completely abnormal.

When SETI is enabled, i have 90% of CPU split between the two SETI tasks, the GPUGRID is using 3 to 5% tops and the rest is used by the system. The GPUGRID tasks take forever to finish.

When SETI is suspended, the GPUGRID task suddenly takes 20 to 40% CPU, and progresses much faster.

Is there something i should try to make that better ?
Or just stick to fedora ? ;-)

Profile [SETI.USA]Tank_Master
Avatar
Send message
Joined: 8 Jul 07
Posts: 85
Credit: 67,463,387
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwat
Message 3424 - Posted: 27 Oct 2008 | 22:24:40 UTC

disable SLI to be able to use both video cards to crunch.

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 3427 - Posted: 27 Oct 2008 | 22:34:13 UTC

However, this doesn't solve the "starving for CPU time" issue.

You could try the setting "on multiprocessor systems use at most xx cpus" and set it to 1 locally. That should give you 1 SETI and 1 GPU-Grid, which should be fine for your dual core. Otherwise I'd certainly disable SLI to crunch 2 WUs if I were you.. but then I wouldn't have bought such hardware anyway ;)

MrS
____________
Scanning for our furry friends since Jan 2002

AngleDroit
Send message
Joined: 2 Sep 08
Posts: 3
Credit: 358,759
RAC: 0
Level

Scientific publications
wat
Message 3430 - Posted: 27 Oct 2008 | 23:13:48 UTC - in response to Message 3427.

ok thx

i have this hardware to be able to play @ 2560x1600 with full details ;-)

i tried to changed from SLI to none and then back and i lost my projects >:

i'll try a fresh install and see how things are

anyway, GDF said in another topic that things should evolve in a few days, so i guess another version is coming :)
____________

Rabinovitch
Avatar
Send message
Joined: 25 Aug 08
Posts: 143
Credit: 64,937,578
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwat
Message 3433 - Posted: 28 Oct 2008 | 4:27:06 UTC

Seems that умкертп is more or less OK, excluding CPU consumpting of 6.48 GPU application (on Windows).
But today I found the following phenomena:

28.10.2008 8:29:54|PS3GRID|Sending scheduler request: To fetch work. Requesting 35166 seconds of work, reporting 0 completed tasks
28.10.2008 8:30:00|PS3GRID|Scheduler request completed: got 0 new tasks
28.10.2008 8:30:00|PS3GRID|Message from server: No work sent
28.10.2008 8:30:00|PS3GRID|Message from server: Full-atom molecular dynamics for Cell processor is not available for your type of computer.
28.10.2008 8:30:00|PS3GRID|Message from server: Full-atom molecular dynamics on Cell processor is not available for your type of computer.

Hmm....

Pepo
Avatar
Send message
Joined: 25 Dec 07
Posts: 19
Credit: 53,126
RAC: 0
Level

Scientific publications
watwat
Message 3434 - Posted: 28 Oct 2008 | 9:40:03 UTC

Heads-up for anyone who's Windows 6.3.17 or Linux 6.3.18 client is crashing (upon requesting work from any project): check whether cc_config's sched_op_debug is on.

Fixed in changeset [16326].

Peter

Profile [SETI.USA]Tank_Master
Avatar
Send message
Joined: 8 Jul 07
Posts: 85
Credit: 67,463,387
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwat
Message 3442 - Posted: 28 Oct 2008 | 15:38:31 UTC

thx. Will setting it to 0 worf for the time being? or should I just remove the line?

Pepo
Avatar
Send message
Joined: 25 Dec 07
Posts: 19
Credit: 53,126
RAC: 0
Level

Scientific publications
watwat
Message 3444 - Posted: 28 Oct 2008 | 16:27:11 UTC - in response to Message 3442.

Will setting it to 0 worf for the time being?

Yes, it should. Just the non-zero value is important for triggering the crash.
or should I just remove the line?

If you do not intend to type it in later...

(Now I do not know anymore, why I've not changed it to 0, but renamed to sched____op_debug - possibly to notice it at a later time? Weird idea :-)

Peter

Post to thread

Message boards : Graphics cards (GPUs) : BOINC 6.3.17 is out

//