<t>QuoteOriginally posted by: outrunI need to do a lot of linear algebra operations on vectors/matrices. I was thinking about C(++) with Lapack, via ATLAS (atlas optimizes performance by using special instructions, looking at cache hits etc)How does numpy perform in this respect? Numpy does rely on ...

Numpy for .NET should be relased soon and will not require anymore using Ironclad. There is a talk this week in London about the project (cf. the London Financial Python User Group). I'll keep you posted when the official release gets out.

You also could have a look at Python with Numpy and Scipy. Numpy allows efficient handling of big arrays and natively support memory maps.

<t>Now it does ;-)Output is :In [35]: run sum_generation.py 10+13+20+40 = 83 13+20+40 = 73 10+20+40 = 70 20+40 = 60 10+13+40 = 63 13+40 = 53 10+40 = 50 40 = 40 10+13+20 = 43 13+20 = 33 10+20 = 30 20 = 20 10+13 = 23 13 = 13 10 = 10 </t>

Some python code that should to the trick, no ?Output is :In [23]: run sum_generation.py 10+13 = 23 10+13+20 = 43 10+13+20+40 = 83 13+20 = 33 13+20+40 = 73 20+40 = 60

<t>It seems the Financial Derivatives Toolbox of Matlab should provide you with some of the needed stuff. If you look at potential integration of your tool with powerful visualisation, deployment or inside of a large app, you should definitely keep using an open environment like Python & numpy/s...

<r>For those interested, I have posted a rough benchmark comparing the usage of the MKL against the old BLAS/LAPACK with numpy. The blog post is here : <URL url="http://dpinte.wordpress.com/2010/01/15/numpy-performance-improvement-with-the-mkl/"><LINK_TEXT text="http://dpinte.wordpress.com/2010/01/1...