open Beta Test for new cmsearch VM application

[ChristianB]

Please have in mind that the current cmsearch VM tasks are in fact XXL tasks that didn't finish until now because of the missing checkpointing. With the VM there is checkpointing now and the tasks should finish at the end. Also keep in mind that the reference system is not using any SSE version so you get a speedup on your system and it may not take the whole 24 weeks. I don't know how much credit this gets you as there is no valid task yet.

[Michael H.W. Weber]

I am not really sure whether CMSEARCH employs SSE at all.

Michael.

[ChristianB]

I found a bug in the newly converted tasks. They need to be recreated in order to have a working progress indicator and a reliable scientific output. Everyone with a VM task may cancel it. I will recreate the tasks later when there is a new vboxwrapper available that fixes the shutdown issue.

[JeromeC]

!!! so now you are telling I let running more than 200 hours for... nothing.

It's a pleasure helping you with the beta.

WUs cancelled, project suspended, waiting for RNA to have a stable VM application now.

[Crystal Pellet]

I found a bug in the newly converted tasks. They need to be recreated in order to have a working progress indicator and a reliable scientific output. Everyone with a VM task may cancel it. I will recreate the tasks later when there is a new vboxwrapper available that fixes the shutdown issue.

"newly converted". Does that mean that the older converted from 18th September (6330530, 6330531, 6330532, 6330533, 6330534) may run on? Already done 1005 hours.

[ChristianB]

Those five seem to be more stable but as far as I can see, they have the same bug that affects the newly created ones. This bug should not affect the scientific output but we will know for sure when the first task is finished.

[Michael H.W. Weber]

!!! so now you are telling I let running more than 200 hours for... nothing.

Well, not really. We have collected a number of very valuable (!) bug reports from those who helped with our testing. So we can build on this and make the project better.
Your help is much appreciated.

Michael.

[Crystal Pellet]

Those five seem to be more stable but as far as I can see, they have the same bug that affects the newly created ones. This bug should not affect the scientific output but we will know for sure when the first task is finished.

Hurray!

My wingman returned his task of the same workunit 6330531 last night (CET).

His machine a Xeon CPU E5520 @ 2.27GHz and mine a AMD Phenom 9350e @ 2.0GHz.

His machine needed 974 hours and I'm at 1071 hours now, so hopefully I return before X-mas too

[ChristianB]

The result file of the task is empty. I have a suspicion why but need some more time to investigate. It seems conversion is trickier than I thought.

[Crystal Pellet]

The result file of the task is empty. I have a suspicion why but need some more time to investigate. It seems conversion is trickier than I thought.

Do you think the result file (the scientific one not the BOINC stderr.txt) was available in the shared folder on the host, but not uploaded?
If yes, what is/are the filename(s)?

Anyway, I'll let my counter-task finish and try to save files before BOINC clears things.

[ChristianB]

The file is not included into the out.zip archive that is copied to the shared directory at the end of the computation. I can see a missing file message in the stdout file from the VM that get's copied to shared/

[Beorn]

Just to be sure: Does that mean, the scientific result file should have been copied into out.zip but it has not been copied for the result in question here?