Welcome to MilkyWay@home

Weird one

Message boards : Number crunching : Weird one
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile Steve Dodd

Send message
Joined: 7 Nov 07
Posts: 25
Credit: 103,270,051
RAC: 17,541
Message 36278 - Posted: 5 Feb 2010, 17:03:31 UTC
Last modified: 5 Feb 2010, 17:10:10 UTC

This task took almost 10 hours to complete (GPU). Screwed up the completion time for all the rest of the tasks.

ps_s222_opt_1_v01_623806_1265035969

Is there checkpointing in the GPU application?

Work unit id - 40922146

Task ID - 43028238

Using BOINC Version 6.6.38 on Core i7, Windows 7, GTX 260 Core 216 (I know the version is old, but I'm running AQUA and there are still problems with the 'official' version of BOINC and AQUA.)
ID: 36278 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Paul D. Buck

Send message
Joined: 12 Apr 08
Posts: 621
Credit: 161,934,067
RAC: 0
Message 36280 - Posted: 5 Feb 2010, 18:28:47 UTC - in response to Message 36278.  

Double check the task, I just got one, well two actually, that were issued to the CPU even though my preferences are set to only use GPU ... and sure enough, took about 11 hours to complete...
ID: 36280 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Steve Dodd

Send message
Joined: 7 Nov 07
Posts: 25
Credit: 103,270,051
RAC: 17,541
Message 36283 - Posted: 5 Feb 2010, 22:08:09 UTC - in response to Message 36280.  

Hi Paul,
No, it ran on the GPU. Just saw the same thing happening again. The task suspended at about 71% complete, and another started. I suspended all remaining Milkway tasks to force the recently suspended task to run, and sure enough, it reset to 0% complete and started over again, except the elapsed time continued from where it left off. I haven't seen this happen before, but it sure will cut into the RAC if it continues :)
ID: 36283 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Steve Dodd

Send message
Joined: 7 Nov 07
Posts: 25
Credit: 103,270,051
RAC: 17,541
Message 36286 - Posted: 5 Feb 2010, 23:46:57 UTC - in response to Message 36283.  

Verified. Tasks are not saving their state when being swapped. Always restart from 0 if not running to completion.
ID: 36286 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Paul D. Buck

Send message
Joined: 12 Apr 08
Posts: 621
Credit: 161,934,067
RAC: 0
Message 36291 - Posted: 6 Feb 2010, 6:43:12 UTC

Ah ... :)

Ok, no checkpointing? Well, with 1 to 10 minute run times on my systems (GPUs only), I had not seen this issue. You *MAY* be seeing the prioritization bug that Richard Haslegrove, me and others have been trying to get them to look at and fix. So far there have been two fixes tried with the first one being a complete failure and the second, well, the jury is still out on that one (version 6.10.32).

ID: 36291 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote

Message boards : Number crunching : Weird one

©2024 Astroinformatics Group