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

improve OMP parallelization with scheduling #1065

Merged
merged 1 commit into from May 31, 2016

Conversation

Projects
None yet
2 participants
@sdorkenw
Contributor

sdorkenw commented May 28, 2016

Many jobs deployed by prange finish very quickly when using a mask. Since the default scheduling seems to assign jobs statically to workers by creation some worker run out of jobs quickly. I tested all possible settings with default parameters, num_threads=10 on a virtualized system on an HCP dataset (HCP/994273/T1w/Diffusion). The mask was computed with dipy and covers (=1) ~25% of the dataset.
Single runs, hence high variance, but the trend is pretty clear:

default: 1990s
guided: 1065s
dynamic: 1011s
static: 2077s

improve OMP parallelization with scheduling
Many jobs deployed by prange finish very quickly when using a mask. Since the default scheduling seems to assign jobs statically to workers by creation some worker run out of jobs quickly. I tested all possible settings with default parameters, num_threads=10 on a virtualized system on an HCP dataset (HCP/994273/T1w/Diffusion). The mask was computed with dipy and covers (=1) ~25% of the dataset.
Single runs, hence high variance, but the trend is pretty clear:

default: 1990s
guided: 1065s
dynamic: 1011s
static: 2077s
@arokem

This comment has been minimized.

Member

arokem commented May 28, 2016

Looks like an easy win to me. I am +1 for the merge. Anyone else have any comments on this (is there something I'm missing?)? If not, I will merge this in a couple of days.

@arokem

This comment has been minimized.

Member

arokem commented May 31, 2016

Thanks @sdorkenw !

@arokem arokem merged commit 3dd9e27 into nipy:master May 31, 2016

1 check passed

continuous-integration/travis-ci/pr The Travis CI build passed
Details
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment