Summary: | ebuild for sci-libs/scipy-1.2.0_rc1 - add USE=mkl | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Joel Berendzen <joel> |
Component: | Current packages | Assignee: | Gentoo Science Related Packages <sci> |
Status: | RESOLVED FIXED | ||
Severity: | normal | Keywords: | EBUILD |
Priority: | Normal | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Attachments: | ebuild for scipy 1.2.0_rc1 with mkl use flags |
The bug has been closed via the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=2f5e92ccad73395c0aae63cbad530c64e4d7eb9a commit 2f5e92ccad73395c0aae63cbad530c64e4d7eb9a Author: Mo Zhou <cdluminate@gmail.com> AuthorDate: 2019-07-09 12:08:35 +0000 Commit: Benda Xu <heroxbd@gentoo.org> CommitDate: 2019-07-09 15:44:15 +0000 sci-libs/mkl-rt: Intel MKL runtime libraries with eselect-ldso mechanism. Closes: https://bugs.gentoo.org/672466 Closes: https://github.com/gentoo/gentoo/pull/12405 Signed-off-by: Mo Zhou <cdluminate@gmail.com> Signed-off-by: Benda Xu <heroxbd@gentoo.org> sci-libs/mkl-rt/Manifest | 1 + sci-libs/mkl-rt/metadata.xml | 25 ++++++++++ sci-libs/mkl-rt/mkl-rt-2019.4.243.ebuild | 81 ++++++++++++++++++++++++++++++++ 3 files changed, 107 insertions(+) |
Created attachment 557028 [details] ebuild for scipy 1.2.0_rc1 with mkl use flags scipy doesn't work with MKL BLAS from science overlay with errors at build time or run time. I got it to work with this: 1. Install numpy with mkl USE flag enabled. 2. Download scipy-1.2.0_rc1.tar.gz to /usr/portage/distfiles. 3. Put something into /usr/portage/distfiles/scipy-1.2.0_rc1-html.zip and /usr/portage/distfiles/scipy-1.2.0_rc1-ref.pdf. 4. Install the attached ebuild. 5. Set the mkl USE flag for scipy. 6. Emerge scipy. 7. Run benchmarks. Known problems with this ebuild: 1. Only works if mkl USE flag is enabled; breaks other BLAS implementations. 2. MKL library location is hard-coded in for now. 3. May break some compiled dependencies such as scikits_learn.