Changes between Version 1 and Version 4 of Ticket #26118


Ignore:
Timestamp:
08/24/18 03:01:26 (4 years ago)
Author:
saraedum
Comment:

The hangs do not happen anymore with OPENBLAS_NUM_THREADS=1 and sage -tp --all finishes in 2:29 minutes (which is the time it takes to run the tests in sage.manifolds.differentiable.tensorfield).

Legend:

Unmodified
Added
Removed
Modified
  • Ticket #26118

    • Property Cc roed added
  • Ticket #26118 – Description

    v1 v4  
    11Strangely `SAGE_NUM_THREADS=160 sage -tp --long --all` produces lots of timeouts on a 160 core machine.
    22
    3 It turns out that only a few cores are actually used (three or four most of the time) which seems to be related to the set CPU affinity. I've read some reports online that numpy sets that affinity.
     3It turns out that only a few cores are actually used (three or four most of the time) which seems to be related to the set CPU affinity.
    44
    55This ugly workaround fixes it:
     
    1919             # Block SIGCHLD and SIGINT except during the pselect() call
    2020}}}
     21
     22A better workaround, turned out to be setting `OPENBLAS_NUM_THREADS=1`.