New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
ipcluster only running on one cpu #3724
Comments
This is most commonly a result of numpy and env flags, e.g. MKL_NUM_THREADS. Do you perhaps have numpy linked against mkl? |
@minrk How do I know - I just do python setup.py install. How do I find out (sorry about the stupid questions) |
actually I don't use an intel compiler and only open source stuff. |
you installed numpy yourself, manually? |
Actually, some BLAS implementations other than MKL can set CPU affinity. See this SO Question for a possible answer. |
@minrk ( I don't remember if I installed the numpy manually). The SO question works! It would be great to put this somewhere on an FAQ. Thanks for your help. I'm being slightly naughty here (I asked this on the mailing list but no one answered): How do I shift data (a class with pandas dataframe) to each individual node engine and make it accessible in the loadbalanced view (don't worry if you don't know). |
Should be |
@minrk: I thought I tried this, but probably did something wrong. Thanks for your help and good luck with 1.0. |
I'm launching several tasks with a load_balanced_view. Despite having 8 cores (and the
ipcluster start
command launching 8 engines). All tasks run on one CPU (and thus have 12% each). This is with the newest ipython version pulled from github an hour ago. This is on ubuntu 12.04 - on my 10.8 mac this works (with only 4 cores).Thanks for the help.
The text was updated successfully, but these errors were encountered: