This is the mail archive of the gsl-discuss@sourceware.org mailing list for the GSL project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Adding OpenMP support for some of the GSL functions


Hi Rhys,
I will have a deeper look at vectorization of GSL, but in my understanding, vectorizing can only be done with simple operations, while algorithms like RKF45 involve about 10 operations per loop iterations.


You are correct, I meant one problem with a million degrees of freedom.

Maxime

Le 2012-12-12 16:41, Rhys Ulerich a écrit :
The more intensive function is within rkf45_apply in my case. I simply added
a few pragmas to the loops, and it speed it up quite a lot.
Having looked at where you've placed the #pragma omp parallels, have
you tried enabling vectorization to see if the time spent in those
axpy-like operations could be improved?  A good SSE-ready optimizer
should nail those.

I may have misunderstood your "millions of differential equations"
statement.  Are you rather solving one problem with a million degrees
of freedom?

- Rhys



Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]