[SciPy-User] Modify package dependencies, or issue a runtime warning, if a fast BLAS implementation not found?

David Cournapeau cournape@gmail....
Thu Sep 13 11:08:49 CDT 2012


On Thu, Sep 13, 2012 at 1:14 PM, Daπid <davidmenhur@gmail.com> wrote:
> On Wed, Sep 12, 2012 at 11:28 AM, David Cournapeau <cournape@gmail.com> wrote:
>> This is out of our hands, and mostly an issue with distributions.
>> Distributions often have different priorities than people looking for
>> best performances, so there is always a tradeoff between
>> distribution/platform support/performance.
>
> Another option, as suggested recently elsewhere, would be to write a
> nice installation tutorial with both  the "quick and dirty" way
> (plainly install the distributed version) and the optimized way
> (linking to Blas, Blas goto and ATLAS). As far as  I know, there are
> no good simple explanations on how to build numpy with these links.

The problem is that there is no such thing as a "quick and dirty" way
to build with optimized libraries, because the way to build numpy with
optimized libraries depend on many parameters.

I think the problem is that the process is too arcane/complicated, and
I strongly believe in making this easier instead of documenting all
the idiosyncraties [1]

cheers,

David

[1] For example, this WE, I added support so that we can now finally
do this for NumPy:

bentomaker configure
--with-blas-lapack-type=atlas/openblas/accelerate/mkl
--with-blas-lapack-dir=where_you_install_the_libs

with the options clearly documented in bentomaker configure --help


More information about the SciPy-User mailing list