LHC@home

Some folks won't entertain this unless it open source. Is it? If not why not? Have I got this wrong and missed a link to the open source web page?

Not sure how many takers they want long term but it could be difficult unless they are prepared to open up the source for people to compile themselves.

---

> Some folks won't entertain this unless it open source. Is it? If not why not?
> Have I got this wrong and missed a link to the open source web page?
>
> Not sure how many takers they want long term but it could be difficult unless
> they are prepared to open up the source for people to compile themselves.
>

It's not open source for a very good reason. Sixtrack is VERY sensitive to differences in FP calculations. The arithmetic has to match on different platforms all the way down to the eightieth bit. Early on in the life of the project it didn't and the final results varied, sometimes a little, sometimes a lot.

The LHC team have put a lot of effort into solving this. Opening the source up for users to compile their own versions will inevitably undo all the work that's been done, and render a lot of the computation worthless.

Giskard

---

Gaspode the UnDressed
http://www.littlevale.co.uk

> > Some folks won't entertain this unless it open source. Is it? If not why
> not?
> > Have I got this wrong and missed a link to the open source web page?
> >
> > Not sure how many takers they want long term but it could be difficult
> unless
> > they are prepared to open up the source for people to compile
> themselves.
> >
>
> It's not open source for a very good reason. Sixtrack is VERY sensitive to
> differences in FP calculations. The arithmetic has to match on different
> platforms all the way down to the eightieth bit. Early on in the life of the
> project it didn't and the final results varied, sometimes a little, sometimes
> a lot.
>
> The LHC team have put a lot of effort into solving this. Opening the source up
> for users to compile their own versions will inevitably undo all the work
> that's been done, and render a lot of the computation worthless.
>
> Giskard
>
>
>

good point well made. Thank you

---

> > Some folks won't entertain this unless it open source. Is it? If not why
> not?
> > Have I got this wrong and missed a link to the open source web page?
> >
> > Not sure how many takers they want long term but it could be difficult
> unless
> > they are prepared to open up the source for people to compile
> themselves.
> >
>
> It's not open source for a very good reason. Sixtrack is VERY sensitive to
> differences in FP calculations. The arithmetic has to match on different
> platforms all the way down to the eightieth bit. Early on in the life of the
> project it didn't and the final results varied, sometimes a little, sometimes
> a lot.
>
> The LHC team have put a lot of effort into solving this. Opening the source up
> for users to compile their own versions will inevitably undo all the work
> that's been done, and render a lot of the computation worthless.
>
> Giskard
>
I'm not sure if this is the real reason, but the folks at CERN have already optimised the application a fair amount. I quote....

"At the time we started this analysis, an average WU would take about one hour to compute on a 2 GHz P4 machine. This number is also highly influenced by the options the physicist specify. During our work with the application we had to change Fortran compilers many times, sometimes just versions and sometimes also the compiler manufacturer, and each time the execution time for a WU became smaller. The version we finally used for the public had almost halved the time an average WU takes."

Live long and crunch.

---

Paul
(S@H1 8888)
BOINC/SAH BETA