feat: parallel implementation

seung-lab / euclidean-distance-transform-3d

Euclidean distance & signed distance transform for multi-label 3D anisotropic images using marching parabolas.

GNU General Public License v3.0

234 stars 37 forks source link

feat: parallel implementation #13

Closed william-silversmith closed 5 years ago

william-silversmith commented 5 years ago

Resolves #11

william-silversmith commented 5 years ago

Volume: 1024x1024x1024 uint8 single label

Red: scipy.ndimage.distance_transform_edt (memory error at end) Black: edt.edt single thread Blue: edt.edt eight threads

william-silversmith commented 5 years ago

Threading has a little overhead for single core because of the lambda functions. This eliminates some of the advantage of the boolean type. Maybe should break single core out from multicore code to not penalize them.