Skip to content

detect and link against lapack libraries. use cython with monkey patch.#2

Closed
rainwoodman wants to merge 1 commit intoscikit-sparse:masterfrom
rainwoodman:master
Closed

detect and link against lapack libraries. use cython with monkey patch.#2
rainwoodman wants to merge 1 commit intoscikit-sparse:masterfrom
rainwoodman:master

Conversation

@rainwoodman
Copy link

Just a few issues to be reviewed:

I am not sure if this is a bug or not, but
I receive 'unresolved symbol' errors unless I link against all of the lapack libraries, explicitly. Everything is properly configured though. I use an old mkl.

Also the default path prefix of cholmod header is quite unflexible -- what if it is installed to /opt/SuiteSparse(eg, our machine's default installation)

I also dropped in a monkey patch to replace the .pyx handler with Cython. You may find it quite ugly.

@njsmith
Copy link
Collaborator

njsmith commented Sep 28, 2012

I have some general questions about numpy.distutils, so I asked the numpy list as well: http://mail.scipy.org/pipermail/numpy-discussion/2012-September/064058.html

If you know any of the answers feel free to answer here or there. Regarding the path for the cholmod header: yes, fair enough, but now how do we make it build? Should there be some code to look for cholmod.h and add it to the default include path? Or does everyone have to set up a site.cfg now? I would need to document how to do this, but in fact I don't know myself (I've never needed to fiddle with site.cfg so I have no idea how it works).

@rainwoodman
Copy link
Author

Sorry this has been a while. I stopped using cholmod as we went to use a different approach to the problem.

But for the path of cholmod.h, seems to me it is always easier to setup the environments (setup.py takes LDPATH and CFLAGS) with <cholmod.h> than <sparsesuite/cholmode.h>, because cholmod.h is not necessarily installed under ; for example, the (presumably) site optimized installation on our cluster.

I recall that's the biggest motivation of the patch.

The cython monkey patch originated from some standard cython recipe; I do not understand why numpy has pyrex yet no cython support.

@anntzer
Copy link
Collaborator

anntzer commented Jun 23, 2016

Seems mostly obsolete now... please reopen if a similar issue still occurs.

@anntzer anntzer closed this Jun 23, 2016
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants