[SciPy-dev] Building newcore with the intel compilers on Itanium2...

Arnd Baecker arnd.baecker at web.de
Thu Nov 3 04:11:06 CST 2005


Hi Fernando,

On Wed, 2 Nov 2005, Fernando Perez wrote:

> For now, a more viable solution (for me) seems to be just to use gcc and lose
> some of the benefits of icc on this architecture.  But I think it would be
> great to be able to offer users the ability to use the C compiler of their
> choice for scipy.

Just out of curiosity: how big are the benefits of icc?
(The computing center here in Dresden is just
installing the first parts of 1500 ItaniumR2
and so we will very much benifit from your pioneering work on
Itanium ...).
Do you also have the Intel Math Kernel Library
http://www.intel.com/cd/software/products/asmo-na/eng/perflib/mkl/index.htm
?
It would be very interesting to see how much performance
increase one can get in practice when using their
Linear Algebra, FFT, vector math, ...
compared to ATLAS, fftw, etc.).
For example, for fft it seems to be really good:
http://www.intel.com/cd/software/products/asmo-na/eng/perflib/mkl/219662.htm
And also for atlas (especially multiprocesssor support):
http://www.intel.com/cd/software/products/asmo-na/eng/perflib/mkl/219823.htm

((No - I don't have stock options with them ;-))

> I can get scipy newcore to pass all tests on this Itanium2 box:
>
> In [4]: scipy.test(10,10)
> ...
> ----------------------------------------------------------------------
> Ran 140 tests in 2.025s
>
> OK
>
> One more 64-bit architecture where things look good, thanks to Travis and
> Arnd's relentless work!

Excellent!

((Remark: even if all the unit tests pass, it does
not mean that code which used Numeric directly works with newcore
- we just had such a case yesterday, where it should have worked ...
so far we have not yet managed to isolate the problem.))

Best, Arnd




More information about the Scipy-dev mailing list