Predictor@Home

[mrwizer]

I noticed that not many are doing this project, but of those that are... has anyone noticed that work units will go beyond 100% completion? How can that be?

I was watching as I reached 100%, expecting the WU to finish. Then it started going beyond, and the time to completion went up to 6 hours like it said in the beginning. Right now I am at 102.99%. Then, it just jumped back down to 97%. hmmm...

 

[Rattledagger]

Yes, Predictor@home-wu seems to have a small problem at the end with displaying correct %done, but atleast in my experience they've finished without any real problem in roughly 1/3rd of the expected crunch-time.

It's possible this wrong-displaying have to do with the interaction between BOINC core client and the predictor@home-application... It's not all science-applications that always correctly reports cpu-time and progress when asked to do this by the core client, and the latest core client in these instances AFAIK acts as if linear progress since last time answered... When the science-application reports the correct cpu-time and #done next time, the display can jump up or down...

Not certain this is the reason in Predictor@home, but this atleast explain why it increases and suddenly drops down again later...

 

[Doc Brown]

Originally posted by: Rattledagger
Yes, Predictor@home-wu seems to have a small problem at the end with displaying correct %done, but atleast in my experience they've finished without any real problem in roughly 1/3rd of the expected crunch-time.

It's possible this wrong-displaying have to do with the interaction between BOINC core client and the predictor@home-application... It's not all science-applications that always correctly reports cpu-time and progress when asked to do this by the core client, and the latest core client in these instances AFAIK acts as if linear progress since last time answered... When the science-application reports the correct cpu-time and #done next time, the display can jump up or down...

Not certain this is the reason in Predictor@home, but this atleast explain why it increases and suddenly drops down again later...

OK ... RD ... another question changing the subject ... I am playing with E@home and this is the 2nd time that E@Home has lost a WU on me ... is this a bonic problem or what ... version 4.19 ...

 

[mrwizer]

Originally posted by: Rattledagger
Yes, Predictor@home-wu seems to have a small problem at the end with displaying correct %done, but atleast in my experience they've finished without any real problem in roughly 1/3rd of the expected crunch-time.

It's possible this wrong-displaying have to do with the interaction between BOINC core client and the predictor@home-application... It's not all science-applications that always correctly reports cpu-time and progress when asked to do this by the core client, and the latest core client in these instances AFAIK acts as if linear progress since last time answered... When the science-application reports the correct cpu-time and #done next time, the display can jump up or down...

Not certain this is the reason in Predictor@home, but this atleast explain why it increases and suddenly drops down again later...

Yeah, I kinda figured the same, but was not sure if any others have had the same problem. Just kinda wierd that it is only predictor, but that also makes sense because each project use thier own app. And it seems to happen to most of the predictor WU's. But they do finish, so at least that is okay. And if anyone is curious, this is on BOINC 4.25. Thanks!