Skip to content

Loading…

ENH: support for detecting libraries in several directories simultaneously #229

Closed
wants to merge 4 commits into from

4 participants

@npinto

Cleaned up PR from the following discussion:
http://projects.scipy.org/numpy/ticket/993

Let me know what you think.

@rgommers
NumPy member

If you could squash the ENH and FIX commits it's even easier to review.

This looks like a welcome improvement. With enough testing, I don't see a real problem merging this.

@npinto

Awesome.

How do we go about testing this ?

@rgommers
NumPy member

Well, we've got Windows and Debian-on-SPARC buildbots (for after the merge), I can test on OS X and Wine. The X11 part needed some fixing on Ubuntu recently, so that would be a good system to test on.

We need to find one or two people who use most or all of the options (Umfpack, X11, MKL). I assume you're one of those people?

@rgommers
NumPy member

Now you also squashed in the PEP8 fixes. That achieves the opposite of what I wanted. Do you still have the original branch? I meant squash the (ENH + FIX), keep the second commit with stylistic changes separate.

@charris
NumPy member

It may be possible to use git reflog for recovery, it should be good for a month or so... undo rebase

@npinto

Thanks. git reset --hard ORIG_HEAD worked here. Let me push.

@npinto

Let me know if this is in the format you expected.

We need to find one or two people who use most or all of the options (Umfpack, X11, MKL). I assume you're one of those people?

Yes, it works for me on Gentoo with ICC 12.1, see e.g. (https://github.com/npinto/numpy-mkl-bootstrap).

@rgommers
NumPy member

These commits look good, thanks.

@rgommers
NumPy member

In the weekend I can test and review in detail.

@npinto

Fixed / rebased.

@rgommers
NumPy member

This fails on OS X:

gcc-4.0 -arch ppc -arch i386 -isysroot /Developer/SDKs/MacOSX10.4u.sdk -g -bundle -undefined dynamic_lookup build/temp.macosx-10.3-fat-2.6/numpy/core/blasdot/_dotblas.o -Lbuild/temp.macosx-10.3-fat-2.6 -o build/lib.macosx-10.3-fat-2.6/numpy/core/_dotblas.so -Wl, -framework -Wl, Accelerate
i686-apple-darwin10-gcc-4.0.1: Accelerate: No such file or directory
powerpc-apple-darwin10-gcc-4.0.1: Accelerate: No such file or directory
lipo: can't figure out the architecture type of: /var/folders/Uu/UuXfo1NLFae4yyYpsCz-XE+++TI/-Tmp-//ccWfS3yr.out
i686-apple-darwin10-gcc-4.0.1: Accelerate: No such file or directory
powerpc-apple-darwin10-gcc-4.0.1: Accelerate: No such file or directory
lipo: can't figure out the architecture type of: /var/folders/Uu/UuXfo1NLFae4yyYpsCz-XE+++TI/-Tmp-//ccWfS3yr.out
error: Command "gcc-4.0 -arch ppc -arch i386 -isysroot /Developer/SDKs/MacOSX10.4u.sdk -g -bundle -undefined dynamic_lookup build/temp.macosx-10.3-fat-2.6/numpy/core/blasdot/_dotblas.o -Lbuild/temp.macosx-10.3-fat-2.6 -o build/lib.macosx-10.3-fat-2.6/numpy/core/_dotblas.so -Wl, -framework -Wl, Accelerate" failed with exit status 1

That's with python setup.py install, Python 2.6, OS X 10.6. Looks like a simple mistake somewhere, but I don't have time to debug right now.

@npinto

Fixed. It should work now (at least it does on my OSX 10.6 w/ Python 2.7.2).

@rgommers
NumPy member

That works now. I've tested on OS X 10.6 with Python 2.4 and 3.2, plus with numscons on 2.6. Also on Windows with Python 2.5. I don't use a site.cfg file though.

Could you perhaps ask on the mailing list if there's anyone who'd want to help test this? Preferably users of MKL / Umfpack / ...

@npinto

We can test MKL (again ;-) and UMFPACK here with @poilvert. Anything else you would like to test?

@rgommers
NumPy member

No, that should be enough.

@cournape: you weren't sure about this change. It looks fine to me, and not all that invasive of a change. A number of people find this useful. Please comment if you still have some reservations.

@npinto

I tried Intel MKL (from Composer XE 12.1) again on Gentoo with Python 2.7.2 and it works.

@npinto

I also confirm that it works on Python 3.2.2 (Gentoo).

@cournape
NumPy member
@npinto

@rgommers, what tests should I run to insure that umfpack is working correctly? Do I need to install/test scipy as well?

@npinto

Update: umfpack support compiles correctly on 2.7.2

@rgommers
NumPy member

OK, merged in commit a4dbfc1. Thanks Nicolas!

@rgommers rgommers closed this
@npinto

Awesome. You're welcome. I'll contribute more, if time allows ;-)

@rgommers
NumPy member

More contributions would be great.

@npinto

No problem, I'm working on Ticket #2073. I'll get you a PR to review soon.

@teoliphant teoliphant referenced this pull request
Commit has since been removed from the repository and is no longer available.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
This page is out of date. Refresh to see the latest.
Showing with 488 additions and 406 deletions.
  1. +488 −406 numpy/distutils/system_info.py
View
894 numpy/distutils/system_info.py
488 additions, 406 deletions not shown because the diff is too large. Please use a local Git client to view these changes.
Something went wrong with that request. Please try again.