UT theoretical chemistry code forum

Howdy folks,

I have a problem with eOn... it seems every time I allow new tasks, they blast in with really tight deadlines and thus end up running high priority killing everything else running.

It appears all the tasks are given about 30 hours to complete. Most other projects allow a lot more time than that.

In any case, doesn't matter if I set the preferences to switch every 60 minutes or 9999 minutes; eOn loads and takes over all cores.

Since I run about 24 different CPU projects, this mean I have to let eOn load, then set "no new tasks" after I get some WU's in the que... that's a pain.

Can ya'll maybe double the time or something? This is the only CPU project I have that does this... like takes over my computer and allows me no control at all. If I let it load up, eventually everything is running high priority until the new WU's arrive, then its only & all eOn again...

Tex1954

I will second this and add that the benchmarking is also a little off, to the point where running eOn on a modern i5 or six-year old P4 gives the same estimate of 300+ hours til completion, where it really takes one to two hours.

But yep, definitely a deadline thing. Since they don't actually take that long to complete, why not raise it? People shouldn't be babysitting BOINC (that is, manually reporting) anyway, and plus anything from 6.12.xx onward automatically reports finished tasks - so no issue there.

I posted current benchmarks in the Low Points thread... interesting...

:D

Tex1954

Well, somebody reads these threads! Now what is happening is that only ONE task is being sent per host at a time.

I can live with that until they extend the times and let us que up some more stuff and limit the internet activity.

Thanks to whomever!! A fast quick fix I can live with for now!

Please extend the times in the future batches!

Much appreciated!

:D

Tex1954

LOL! Well, I take back that last post... it was running well, one task per host... then suddenly started loading me up again with one task per core...

Sigh...

We have to fix this someday!

LOL!

:D

I have taken to opening work fetch, allowing one to download, then setting NNT again.

Yup, I have had to do that with 3.10 eOn also... first it sent some 1.5-2.5 minute tasks, then sent some really long ones with deadline too short so they hog my systems again. They always run high priority... take every core... These one day and two day deadlines suck big time... It doesn't let anything else get done ever if I let it go. So now, they get about 10% or less of my compute power because of short deadlines. I would be happy to let it use ONE core constantly, but not 4 or 6 or 8 depending on which machine I let go.

Sigh...

Tex1954