Query about the graphlab deployment on a single multicore machine

User 207 | 4/7/2014, 9:36:44 AM

I had a query about the graphlab deployment on a single multicore machine, I have configured my graphlab correctly with the -no_mpi option and when I run any standard graphlab command with the --nlcus =10 option, graphlab still only runs on a single core! Is there something I am missing, or something I can do to ensure that gaphlab uses all the cores available?

Thank you,

sincerely, Kushalveer Singh

Comments

User 6 | 4/7/2014, 2:59:19 PM

Please try --ncpus=XX and not as you wrote.


User 207 | 4/7/2014, 6:45:57 PM

Thank You Danny Bickson

In last post by mistake i write --nclus=10 but i used ncpus=10

following command i am using ./partitioning --graph=/root/snap_0.txt --format=tsv --partitions=100 --ncpus=20

where snap_0.txt file contains the graph in tsv format (edge list)

But graphlab still only runs on a single core!


User 20 | 4/8/2014, 4:46:14 AM

There are certain stages where only one core is used. Notably input file parsing is one core per input file.


User 207 | 4/10/2014, 11:56:01 AM

Thanks Yucheng