Posts by catavalon21

1) Message boards : Number crunching : Computation Error (Message 80328)
Posted 4 Jul 2016 by catavalon21

That is true if that was indeed what was happening. I just think that it actually means that it never made it back. Based on the outcomes explanation screen:

Client error The task was sent to a computer and an error occurred.
Success A computer completed and reported the task successfully.

I thought it was also a mobile device issue but the majority of the projects sent to my vaio also ended in a compute error. Oh well, its not really like my individual contribution was going to make or break anything. If the work being done does not result in the the results they want I don't think it would come up as a compute error since they would have received data that they can use to modify their design or whatever. I don't necessarily care that my credit numbers aren't going up since they don't mean anything outside of the app but I do like seeing changes so that I know that my computer and phone are making contributions, as small as they seem.

Sorry, I don't know. It's weird some successes occurred too.

Hmm, maybe it would be better to shift support to other projects until its sorted out. At least that way work units don't go to waste.

I'm noticing when I go to look at the work units and their history on other systems that my lack of success is followed by or follows other's lack of success on the same work units. Perhaps the work units that can run on ARM7 systems all have a 25% success rate in general and failure is part of the plan for 75% of them??? It seems that the ones the fail for me fail for many systems across different versions of phone, OS, etc. I mean is this a possibility at Rosetta@Home that there is a built in 75% failure rate for certain types of work units and the failures and the successes are all leading to the success of the project as an whole? Like a failure proves something about the work inside the work unit that leads to conclusions about the work and directions of research? Just throwing it out there as an open question since I don't know at all and the thought occurred. Cause if that is true, then none of the work, failure or success, is wasted work, just uncredited work, which in the long term credit means nothing to the work being done, as long as it gets done.

I don't think this is entirely the reason. I have a very large percentage of Android tasks ending in compute failure, and they can't have "not made it back", as the deadline hasn't occurred yet.

It's possible the results made it back with an error, but as another post said earlier, tasks that fail appear to fail repeatedly across several users consistently.

Hopefully it'll get sorted out, but will crunch something else in the interim. F@H has an Android client, though not on BOINC.

©2022 University of Washington