search

ucberkeley / bce

Berkeley Common Environment provides a common Linux computational environment for classwork and research.
Apache License 2.0
13 stars 6 forks source link

dot product in numpy is very slow #21

Open paciorek opened 9 years ago

paciorek commented 9 years ago

Chris ran into a possibly similar issue on AMD machines on the SCF that we addressed so he can go back to his notes and try to see if this is a related issue.

Alternatively, link numpy to openblas and see if the issue persists. At the moment reinstallation of numpy after install of openblas is NOT detecting openblas.

openblas_info: libraries openblas not found in ['/usr/local/lib', '/usr/lib', '/usr/lib/x86_64-linux-gnu'] NOT AVAILABLE

Not sure why as /usr/lib contains openblas-base and libopenblas.so.0

Also, look into whether python/numpy should be linking to system blas (/usr/lib/libblas.so{,.3}), which (after install of openblas) is pointed to openblas via the alternatives system.

paciorek commented 8 years ago

dot is now using openBLAS so I think things are fine on that front.