Welcome to MilkyWay@home

Separation updated to 0.80

Message boards : News : Separation updated to 0.80
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3

AuthorMessage
robertmiles

Send message
Joined: 30 Sep 09
Posts: 211
Credit: 36,977,315
RAC: 0
Message 49317 - Posted: 15 Jun 2011, 1:25:30 UTC - in response to Message 49271.  

Switched one machine over to .82. This is what i get:

<core_client_version>6.12.28</core_client_version>
<![CDATA[
<message>
Maximum elapsed time exceeded
</message>

[snip]


I've seen a few messages on other BOINC projects saying that at least some of the 6.12.* versions of BOINC have problems calculating time limits correctly, and a few more messages that suggest that some of the 6.10.* versions have a similar problem but probably from a different cause. Therefore, anyone reporting such a problem should make sure to mention which version of BOINC the problem happened under.
ID: 49317 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Link
Avatar

Send message
Joined: 19 Jul 10
Posts: 621
Credit: 19,254,980
RAC: 2
Message 49326 - Posted: 15 Jun 2011, 11:09:05 UTC - in response to Message 49206.  

- Checkpointing for GPUs for both OpenCL/Nvidia and ATI/AMD for CAL. It will checkpoint no more frequently than after at least 10% progress, so it might not checkpoint as frequently as your settings specify if you have a particularly slow GPU. GPU checkpointing is a bit slow when it does happen, so if you don't want it, you can disable it with the flag --gpu-disable-checkpointing.

May I suggest for future releases an additional command line option to specify the min. % progress, i.e. if someone wants that checkpointing won't happen more often as after every 20% or something like that?
ID: 49326 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Beyond
Avatar

Send message
Joined: 15 Jul 08
Posts: 383
Credit: 729,293,740
RAC: 0
Message 49331 - Posted: 15 Jun 2011, 14:13:38 UTC - in response to Message 49317.  

Maximum elapsed time exceeded

I've seen a few messages on other BOINC projects saying that at least some of the 6.12.* versions of BOINC have problems calculating time limits correctly, and a few more messages that suggest that some of the 6.10.* versions have a similar problem but probably from a different cause. Therefore, anyone reporting such a problem should make sure to mention which version of BOINC the problem happened under.

The cause was listed above. It was a run of bad WUs and has nothing to do with the BOINC version. It was pointed out before in the other thread.
ID: 49331 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
ExtraTerrestrial Apes
Avatar

Send message
Joined: 1 Sep 08
Posts: 204
Credit: 219,354,537
RAC: 0
Message 49343 - Posted: 15 Jun 2011, 20:29:59 UTC

@Link: just set the "write to disk at most ...%" in your profile or local settings.

MrS
Scanning for our furry friends since Jan 2002
ID: 49343 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Link
Avatar

Send message
Joined: 19 Jul 10
Posts: 621
Credit: 19,254,980
RAC: 2
Message 49380 - Posted: 16 Jun 2011, 21:11:43 UTC - in response to Message 49343.  
Last modified: 16 Jun 2011, 21:13:01 UTC

I have there only "at most every x seconds", but that would work as well. I know this way, however I thought that if we can disable checkpointing for the GPU apps, why not have a possibility to change the hardcoded 10% limit. That way we could use the our default values for other apps, while avoiding too many CPU intensive checkpoints from the GPU app.
ID: 49380 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Matt Arsenault
Volunteer moderator
Project developer
Project tester
Project scientist

Send message
Joined: 8 May 10
Posts: 576
Credit: 15,979,383
RAC: 0
Message 49381 - Posted: 16 Jun 2011, 21:16:54 UTC - in response to Message 49380.  

I have there only "at most every x seconds", but that would work as well. I know this way, however I thought that if we can disable checkpointing for the CAL apps, why not have a possibility to change the hardcoded 10% limit. That way we could use the our default values for other apps, while avoiding too many CPU intensive checkpoints from the GPU app.
It's not 10%. It's the default time based with an additional condition that at least 10% be done. It only serves to potentially decrease the frequency of checkpoints with slow GPUs. Unless there were going to be enormous checkpoints (on the order of 100MB), there's some error introduced with each checkpoint. If it were to checkpoint at every possible opportunity, the introduced error becomes unacceptable. I decided the arbitrary limit of at least 10% kept the errors acceptable.
ID: 49381 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Previous · 1 · 2 · 3

Message boards : News : Separation updated to 0.80

©2024 Astroinformatics Group