Package Details: atlas-lapack 3.10.3-4

Git Clone URL: (read-only, click to copy)
Package Base: atlas-lapack
Description: Automatically Tuned Linear Algebra Software
Upstream URL:
Licenses: BSD, custom:lapack
Conflicts: atlas-lapack-base, blas, cblas, lapack, openblas, openblas-lapack
Provides: atlas-lapack-base, blas, cblas, lapack
Submitter: ilpianista
Maintainer: a.kudelin
Last Packager: a.kudelin
Votes: 92
Popularity: 0.031945
First Submitted: 2008-04-24 01:36
Last Updated: 2019-11-15 18:37

Required by (318)

Sources (3)

Latest Comments

1 2 3 4 5 6 ... Next › Last »

a.kudelin commented on 2020-04-10 10:30

@ajong, yes, it is. "Tests" are the auto-tuning procedures which the efficiency of the ATLAS package is based on.

ajong commented on 2020-04-08 02:27

The install is running tests for what seems to be like forever. It's run for three hours so far, is this normal? Is there a way to skip the tests?

ajong commented on 2020-04-08 00:05

Got compile error about CPU throttling, followed phcerdan's advice and it worked!

a.kudelin commented on 2019-11-15 18:39

cblas link was fixed.

a.kudelin commented on 2019-10-16 12:27


einalex commented on 2019-10-14 11:30

can you change "mkdir ATLAS/build" to "mkdir -p ATLAS/build" so the package doesn't throw an error if you try to build it with the directory already existing from a previous build?

Emmatipate commented on 2019-06-12 11:51

I solved "undefined reference to 'main'" by adding unset LDFLAGS to build() in python-numpy's PKGBUILD.

sain_87 commented on 2018-08-04 08:47

Again, broken numpy with /usr/lib/ missing.

Recompiling numpy from source with "yaourt -Sb python-numpy" after installing ATLAS fails for me when compiling python_xerbla.c with "undefined reference to `main'"

Is there any solution for this?

deltaecho commented on 2018-04-26 22:40

The best solution is to recompile numpy after replacing BLAS (no hacking required): yaourt -Sb python-numpy # be sure to add python-numpy to the IgnorePkg attribute of your /etc/pacman.conf to avoid breaking updates; you will need to compile it from source from here-on-out, which is easy to do

greyltc commented on 2018-03-03 22:35

Since this package provides and conflicts with cblas, I think it should at least install all the shared objects that the cblas packages does. /usr/lib/ and /usr/lib/ are missing here.

Notably, the missing /usr/lib/ breaks python's numpy module!