LHC@home

Kinda funny.
http://lhcathome.cern.ch/workunit.php?wuid=27871

maybe my machine sucks.


______________________________________________________________
Did your tech wear a static strap? No? Well, there ya go! :p

If you look at the result record for your result you will se 'Validate state: Invalid'

This means that your result most likely had an error in it and didn't agree with the other 2. Are you overclocking at all? Could also be a random error... who knows really... I do remember one of the admins saying that they require bit-level precision for this project which makes it very sensitive to errors.


--------------------------------------
A member of The Knights Who Say Ni!
My BOINC stats site

Well, the majority of WU's validate ok.

No overclocking here. And I wear my static strap when working on my pc.

______________________________________________________________
Did your tech wear a static strap? No? Well, there ya go! :p

> maybe my machine sucks.

I don't think so, Sixtrack is very sensitive to numerical differences, which unfortunately is very difficult to avoid with different machines. We did install a more forgiving validator, but some results still fail validation.

/Jakob Pedersen
LHC@Home Admin.

A question for the developers:

Presumably your compiler is using the floating point processor of the host system to perform its calculation, and this is where the numerical differences are accruing.

Would it be possible or even practical to use a software library that performs FP maths without using the FP processor? I realise this might slow the computation, but the random differences might be significantly reduced. The net result might actually be faster validation...

...or it might not ;-)




Giskard - the first telepathic robot.