Double FGRP5 runtime
11 Jan 2018 13:03:48 UTC
Topic 212599
(moderation:
Current lot ...19F seems to run twice as long as previous lots. That's similar to an occasion we've had by mid December, but it isn't clear how it was sorted out. So what to do with those tasks? Aborting and hope for a better download? Probably not, so for now, keep those until things clear up.
I'd like to confirm that it
)
I'd like to confirm that it is almost twice longer for 14F then for 18F:
LATeah0014F_72.0_480_-2.6e-11_1 15:19:54
LATeah0014F_72.0_264_-4.1e-11_2 15:20:33
LATeah0014F_56.0_72_-8e-11_1 14:45:22
LATeah0018F_1480.0_904689_0.0_0 08:23:32
LATeah0018F_1480.0_914166_0.0_0 08:29:38
LATeah0018F_1480.0_914004_0.0_0 08:25:58
In my experience, the longer
)
In my experience, the longer running time is the normal time. There have been quite a few examples recently of particular data files (like LATeah0018F.dat) that give 'shorter than norma'l run times. My assumption is that there is 'less than normal' work content in each task for these shorter running tasks. I don't believe there is any particular 'problem' as such. The work needs to be done. Aborting them would be pointless. Until the data file is completed, you'll just get more of the same.
Just enjoy the bonus when you get short running tasks :-).
Cheers,
Gary.
I am seeing the same, almost
)
I am seeing the same, almost twice as long runtime for some batches than others. Another observation, it really did help to create a wisdom file, I got some 40% reduction in runtime compared to earlier reports of improvements from 0% up to 25% with wisdom for fgrpsse, for an earlier set of tasks. Either I measured this wrongly or FGRP5 benefits more from wisdom. Anyway worth trying for people coming from running the O1SpotLo/O1SpotHi tasks, which has no wisdom feature.