Sign in to follow this  
leoptimus

The Best Numerics Library

Recommended Posts

Hi. I'm currently searching a numerical methods library, specially linear algebra, maybe for physics simulations, some one that makes large vector operations faster, takes advantage of the modern CPU/GPU and does paralele processing/computing. There is a rumor claiming that c++ for numerics is not as fast as the fortran routines (G77). Some numerics libraries comes with Fortran code, like ATLAS or BLAS. I want to know where I can get a good numerics library (for c++) efficient enough for industrial uses (including games )? At this time, I've found these: GSL (GNU Scientific Library). http://graphics.stanford.edu/projects/brookgpu/lang.html - Links to fortran BLAS and manages sparce matrices. [Edited by - leoptimus on September 6, 2006 7:40:01 PM]

Share this post


Link to post
Share on other sites
Fortran indeed has a slight performance edge in certain situations. It isn't so much a question of capabilities as it is Fortran's handling of arrays in a way which makes efficient programming clearer and easier (in this area, anyway; it's a horrible piece of garbage in many other respects, even compared to C++).

C++ is no slouch, though. uBLAS is robust and efficient, and is definitely ready for prime time. And if uBLAS isn't efficient enough for you, Fortran coding probably isn't either... and you'll be more interested in directly coding vectorized SSE routines.

EDIT: Oh, you mentioned leveraging the GPU. Personally, I wouldn't bother... but if you want to, you'll want to use Brook or Sh. GPGPU.org will have more resources in that area.

Share this post


Link to post
Share on other sites
There is quite a lot of work going on, as we write, in the area of exploiting GPU's and emerging multicore processors to accelerate linear algebra and related numerical methods. I attended the "EDGE" workshop @ the University of North Carolina at Chapel Hill a few months ago, and the . I was very impressed with what people are doing in this area.

Workshop on Edge Computing Using New Commodity Architectures (EDGE)

Note that the slides from the workshop are available for download.

Share this post


Link to post
Share on other sites
Aha, well....
and what about the existing packages avaliables for today's programmers? is there any ready to use library for mathematics? tested?
For example i got Blitz++, but, is there any better?

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

Sign in to follow this