Welcome to MilkyWay@home

Locking Up Moving From Milkyway to Collatz


Advanced search

Message boards : Number crunching : Locking Up Moving From Milkyway to Collatz
Message board moderation

To post messages, you must log in.

AuthorMessage
Zydor
Avatar

Send message
Joined: 24 Feb 09
Posts: 620
Credit: 100,587,625
RAC: 0
100 million credit badge12 year member badgeextraordinary contributions badge
Message 37621 - Posted: 21 Mar 2010, 15:27:50 UTC
Last modified: 21 Mar 2010, 15:31:06 UTC

I dont want to get into an "other enditis" saga on this, I am trying to track down the reality.

I recently got a new toy - a 5970 - its working very well, except when I suspend Milkyway and startup Collatz. On that specific sequence of actions, the GPU use on Collatz plummets to circa 40%, and the machine gets very stuttery on the mouse or any screen related activity, verging on unworkable.

If I then suspend Collatz, nothing changes, even though both projects are by now suspended ...

If I do this in the reverse ie move from Collatz to Milkyway, all is well, no problems. It only happens when I move from Milkyway (having suspended it) to Collatz.

The only way to clear this is to reboot. Then all is well, until the next time I move from Milkyway to Collatz...

The card and machine seem fine, main memory is rock solid, all the tests under the Sun seem to indicate no hardware errors. At present my deduction is that on suspending milkyway, the application stops, but does not clear the GPU memory for some reason, and when I startup Collatz the GPU useage reported is low (circa 40%) because it cant get at a large chunk of GPU memory still occupied by the old Milkyway stuff.

BOINC is not set to keep apps in memory when suspended. I have also noticed in Task Manager that the two milky way apps are still running, despite being suspended. If I force a thread closure in Task Manager, milkyway opens up another thread with another instance of the app, despite being suspended in BOINC....

All weired and very frustrating.....

All this only happens when moving from Milkyway to Collatz, it does not occur when moving from Collatz to Milkyway - in the latter case all is fine.

I have the initial impression that milkyway is not exiting the GPU memory gracefully, and leaving bits behind in GPU memory locking out Collatz from those bits still occupied by milkyway parts left behind, and Collatz uses all thats left leaving nothing else there for anything hence the stuttering screen and mouse - not casting dispertions, it could be something silly I'm doing - just trying to make sense of this.

Anyone else getting the same behaviour or any other suggestions I can try/test to sort this.

Regards
Zy
ID: 37621 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profilekashi

Send message
Joined: 30 Dec 07
Posts: 311
Credit: 149,490,184
RAC: 0
100 million credit badge13 year member badge
Message 37623 - Posted: 21 Mar 2010, 15:53:09 UTC - in response to Message 37621.  

I always complete MilkyWay tasks before suspending the MilkyWay project. To do this I suspend all MilkyWay tasks in the cache except for the 2 that are running and wait for those 2 to complete. Only takes a few minutes to do it this way. When one project has completed any tasks it has running I change the GPU memory speed to suit and start up the other project.
ID: 37623 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
BarryAZ

Send message
Joined: 1 Sep 08
Posts: 519
Credit: 289,414,784
RAC: 24,063
200 million credit badge13 year member badgeextraordinary contributions badge
Message 37624 - Posted: 21 Mar 2010, 16:00:52 UTC

Milkyway GPU tasks have to complete before I can suspend them, elsewise they reset to 0%. Collatz tasks don't have this bug.
ID: 37624 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Zydor
Avatar

Send message
Joined: 24 Feb 09
Posts: 620
Credit: 100,587,625
RAC: 0
100 million credit badge12 year member badgeextraordinary contributions badge
Message 37625 - Posted: 21 Mar 2010, 16:07:28 UTC

Thanks guys, appreciated - I'll give the individual suspension of cache tasks and completing those running before switching, a whirl, next time I switch milkyway to collatz.

I'll post the result.

Regards
Zy
ID: 37625 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Zydor
Avatar

Send message
Joined: 24 Feb 09
Posts: 620
Credit: 100,587,625
RAC: 0
100 million credit badge12 year member badgeextraordinary contributions badge
Message 37627 - Posted: 21 Mar 2010, 20:15:39 UTC

Yup - that did it, left the "suspension" alone, and allowed the switching to go ahead without my fingers and thumbs (!). Works fine, as was pointed out above just need to watch what its running for the GPU settings.

Thanks guys, appreciated :)

Regards
Zy
ID: 37627 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
ProfileSlicker [TopGun]

Send message
Joined: 20 Mar 08
Posts: 46
Credit: 69,382,802
RAC: 0
50 million credit badge13 year member badge
Message 37628 - Posted: 21 Mar 2010, 20:19:48 UTC - in response to Message 37624.  

Milkyway GPU tasks have to complete before I can suspend them, elsewise they reset to 0%. Collatz tasks don't have this bug.


I'm not sure it is so much of a bug as a design decision. The MW tasks don't run long enough on fast GPUs to hit a checkpoint which only occurs every 3 minutes (or is it 5, I don't recall). Back when MW WUs took under a minute, there really wasn't any reason for checkpointing. Also, when the apps were first created, there were no other projects running ATI.

As far as the slow response goes, that can be controlled via the command line parameters via the app_info. The options are all in the readme that CP includes with the opt apps.
ID: 37628 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
John Clark

Send message
Joined: 4 Oct 08
Posts: 1734
Credit: 64,228,409
RAC: 0
50 million credit badge13 year member badge
Message 37630 - Posted: 21 Mar 2010, 20:38:55 UTC

When moving from one project to another (I use an HD5850) I NNT the project to be moved from and run it until all caches are cleared.

This is easy going from Collatz to Milkyway, as Collatz runs in preference to Milkyway on my GPU when both projects have work in their caches.

Swapping from Collatz to Milkyway means running down the Collatz cache. Going the other way is to NNT Milkyway, which finishes fairly quickly prior to starting Collatz and allowing work to download.
Go away, I was asleep


ID: 37630 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Zydor
Avatar

Send message
Joined: 24 Feb 09
Posts: 620
Credit: 100,587,625
RAC: 0
100 million credit badge12 year member badgeextraordinary contributions badge
Message 37631 - Posted: 21 Mar 2010, 20:43:31 UTC

Okie Doke - Thanks John :)

Regards
Zy
ID: 37631 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
ProfileGalaxyIce
Avatar

Send message
Joined: 6 Apr 08
Posts: 2018
Credit: 100,142,856
RAC: 0
100 million credit badge13 year member badge
Message 37696 - Posted: 23 Mar 2010, 20:43:25 UTC

I don't know how you have the patience to sit there waiting for caches to run dry before switiching to the other, especially if you have multiple boxes.

I have the problem that in some cases Milkyway can dominate the system so that mouse etc. run extremely slow, as Zydor describes, but as soon as I switch to Collatz everything whizzes again. So long as one or the other is running I don't really worry about it.



ID: 37696 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote

Message boards : Number crunching : Locking Up Moving From Milkyway to Collatz

©2021 Astroinformatics Group