XLL or not XLL? This is the question.

[Lem]

So far I've crunched a few of those new cmsearch XLL workunits... that are quite smaller than the regular cmsearch S ones, indeed. LOL.

They cannot saturate a core: cpu idle time rises, cpu temperature falls... And in fact the spread between the run time and the cpu time is much higher than usual. I've seen that this spread lowers when the workunits are crunched by "virtual cores" of a hyperthreading capable cpu... which may have sense.
For example: http://www.rnaworld.de/rnaworld/workuni ... id=5091189

I've finally decided to abort most of them, thinking they're perfectly suitable even for a bit slower crunching boxes not always on, and to keep on going with the regular S ones.

Bye.

[yoyo]

Currently the XXL app is in testing, therefore I use short WUs to check if the XXL app, the validator and the whole backend processing is working as it should. Later the XXL app should run only the large Wus, so the user can decide what he wants.
yoyo

[Michael H.W. Weber]

So far I've crunched a few of those new cmsearch XLL workunits... that are quite smaller than the regular cmsearch S ones, indeed. LOL.

They cannot saturate a core: cpu idle time rises, cpu temperature falls... And in fact the spread between the run time and the cpu time is much higher than usual. I've seen that this spread lowers when the workunits are crunched by "virtual cores" of a hyperthreading capable cpu... which may have sense.
For example: http://www.rnaworld.de/rnaworld/workuni ... id=5091189

I've finally decided to abort most of them, thinking they're perfectly suitable even for a bit slower crunching boxes not always on, and to keep on going with the regular S ones.

A few points: First of all, as said by Yoyo above, these are test WUs to check the performance of the XXL application. BUT: test WU does not mean, these go to waste! By contrast, these test WUs are a very important new project taking on viruses and how they corrupt human cell regulatory mechanisms by bringing in small RNAs. And finally, the reason for why your core appears not fully occupied is that the virus genomes are extremely small, i.e. at the order of 10 kB per genome as opposed by 4 MB per bacterial genome or even upt to 500 MB per large eukaryotic chromosomes. So, the computation is almost over before it starts. For this reason, these genomes are bundled in a single WU pack that aims at completing several hundred of these computations successively. And there are many viruses out there (in fact, thousands).

Michael.

[Lem]

A few points: First of all, as said by Yoyo above, these are test WUs to check the performance of the XXL application. BUT: test WU does not mean, these go to waste! By contrast, these test WUs are a very important new project taking on viruses and how they corrupt human cell regulatory mechanisms by bringing in small RNAs.

Of course, of course, I agree.

The only reason why I decided to abort most of the XLL I still had in queue was that I thought they would have been a light work, suitable for everyone, whereas I like better to go with heavy WUs if I'm fast enough.

And finally, the reason for why your core appears not fully occupied is that the virus genomes are extremely small, i.e. at the order of 10 kB per genome as opposed by 4 MB per bacterial genome or even upt to 500 MB per large eukaryotic chromosomes. So, the computation is almost over before it starts. For this reason, these genomes are bundled in a single WU pack that aims at completing several hundred of these computations successively. And there are many viruses out there (in fact, thousands).

Thanks to both of you for your clarification.

Bye.

[Michael H.W. Weber]

Michael.