Advanced search

Message boards : Number crunching : HIVPR_n1 stopps and need to be manually restartet again

Author Message
abakus
Send message
Joined: 4 Mar 09
Posts: 6
Credit: 324,547
RAC: 0
Level

Scientific publications
watwatwatwat
Message 9847 - Posted: 16 May 2009 | 12:06:10 UTC
Last modified: 16 May 2009 | 12:09:12 UTC

Hello all,
this is my first post, so please move it if I failed to choose the appropriate category. Thx.

Here is an "annoyance description":
Above mentioned HIVPR_n1 WU's stopp whenever another WU (tested with climateprediction, Einstein, Spinhenge and SHA1) is finished, uploaded and the consecutive one starts.
To restart, I need to stop all WU's, start HIV..., and the start the rest again!
It also does not tolerate another second WU on the CPU (dual core), but this has been described before.
Used: BOINC -WCG 6.4.7)on XP Pro.

To be plain and simple: I don't like this behaviour and are going to abort those WU's in the future, as long as nobody posts a solution.
Sorry - but BOINC is something that is supposed to work in the background. If it starts to force my attention: ...

Let's discuss please
abakus

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 9850 - Posted: 16 May 2009 | 12:39:58 UTC - in response to Message 9847.

I think you're reporting something new: I can't remember any post, which said that the troublesome WUs hang when BOINC switches apps (or if one finished). That may give us a clue as to why it happens and where to start searching for a solution.

It also does not tolerate another second WU on the CPU (dual core), but this has been described before.


Not sure what you mean.

I don't like this behaviour and are going to abort those WU's in the future, as long as nobody posts a solution.


Of course, noone likes this and it's not intional behaviour. Since you're running 6.4.7 don't hold your breath for a quick solution :/

MrS
____________
Scanning for our furry friends since Jan 2002

abakus
Send message
Joined: 4 Mar 09
Posts: 6
Credit: 324,547
RAC: 0
Level

Scientific publications
watwatwatwat
Message 9864 - Posted: 16 May 2009 | 15:14:01 UTC - in response to Message 9847.

Glad if it turns out my post is of any use...

not sure what you mean.

Ok, here is a more precise try:
Assume I am running an IBUCH job: simulataneously I can also crunch any one or two of the below mentioned projects (climateprediction, Einstein, Spinhenge and SHA1) or any combination of them. In total I am running two CPU and 1 GPU.

THis does not work with the HIVPR on my machine.
If I try to do this, always one CPU job gets pushed aside and is on hold.
I can restart it by manually stop and restart it again.
But then: HIVPR is unpredictable "on strike".

There has been one or the other thread down this catgegory which I understood this way.
Ahh, btw I forgot to mention: 8800GTS512 (62.92.16.00.39 - driver 6.14.11.8120)

happy crunching !
abakus

Profile Dieter Matuschek
Avatar
Send message
Joined: 28 Dec 08
Posts: 58
Credit: 231,884,297
RAC: 0
Level
Leu
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 9868 - Posted: 16 May 2009 | 17:33:32 UTC - in response to Message 9864.

@abakus

To run 2 CPU tasks + 1 GPU task simultaneously you should simply upgrade to BOINC 6.6.28. But before you do this, please update the GPU preferences with the option "Suspend GPU work while computer is in use?" deactivated.

Perhaps then the other problem will be resolved also? (Please report when it occurs again.)
____________

abakus
Send message
Joined: 4 Mar 09
Posts: 6
Credit: 324,547
RAC: 0
Level

Scientific publications
watwatwatwat
Message 9870 - Posted: 16 May 2009 | 18:28:12 UTC - in response to Message 9868.

As I wrote, the "problem" only occurs when crunching HIV...
GPUGrid works perfectly with the mentioned client, crunching two CPU plus GPU.
To be honest, I can't see a reason to upgrade right now, as none of my other projects or WU's has any problem at all. I will eventually do so later.
The option is and was deactivated - set to "no".
Thx anyway!

Profile Paul D. Buck
Send message
Joined: 9 Jun 08
Posts: 1050
Credit: 37,321,185
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 9873 - Posted: 16 May 2009 | 20:13:15 UTC - in response to Message 9870.

As I wrote, the "problem" only occurs when crunching HIV...
GPUGrid works perfectly with the mentioned client, crunching two CPU plus GPU.
To be honest, I can't see a reason to upgrade right now, as none of my other projects or WU's has any problem at all. I will eventually do so later.
The option is and was deactivated - set to "no".
Thx anyway!

If you are running into an old bug that already has been addressed, then we can be of no help at all. There were problems with older versions and work starting and like you describe. Sadly, whatever research you could do for us on this is negated by the fact that significant changes have been made to this area of code. We think we caught most of this issue, but, if you won't upgrade BOINC to find out if that helps ... well, then you have to live with fiddling with BOINC all the time.

6.4.x had a number of issues with CUDA as these were the first versions issued to enable the capabiity. Our collective experience here at GPU Grid says that the first "good" version of BOINC that used the GPU capabilities was 6.5.0 ... now, the newly minted 6.6.28 as recommended by UCB is also recommended by some of us as having addressed a number of issues with running the GPU and CPU and keeping they loaded with work ...

That does not mean that this version is bug free ... there are still significant issues that remain ... and some of us are nagging at UCB to get them addressed ... please reconsider upgrading, and if it does not help your problem we can look into it more for you ...

abakus
Send message
Joined: 4 Mar 09
Posts: 6
Credit: 324,547
RAC: 0
Level

Scientific publications
watwatwatwat
Message 9876 - Posted: 16 May 2009 | 21:02:39 UTC - in response to Message 9873.

Paul, Dieter,
ok - I will upgrade soon.
But before I can run a HIV-WU again it might take a while.
In any case I will post my experiences here (negative or postive).

Just out of curiosity:
Is there a way or setting, besides aborting other, to deliberatly crunch only those HIV-WU's?

Profile Paul D. Buck
Send message
Joined: 9 Jun 08
Posts: 1050
Credit: 37,321,185
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 9877 - Posted: 16 May 2009 | 21:07:13 UTC - in response to Message 9876.

Paul, Dieter,
ok - I will upgrade soon.
But before I can run a HIV-WU again it might take a while.
In any case I will post my experiences here (negative or postive).

Just out of curiosity:
Is there a way or setting, besides aborting other, to deliberatly crunch only those HIV-WU's?

No.

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 9882 - Posted: 16 May 2009 | 21:28:28 UTC

You could dodge the problems of 6.6.28 (not sure how serious, I think long term debts are still out of whack) by testing 6.5.0 first. It doesn't have the cpu scheduling issues 6.4.x have.

MrS
____________
Scanning for our furry friends since Jan 2002

Post to thread

Message boards : Number crunching : HIVPR_n1 stopps and need to be manually restartet again

//