Faster dti odf

[arokem]

Speeds up DTI odf calculation, by calculating it only over regions of the volume that have non-zero eigen-values.

[arokem]

This saves a little bit of time for the case in scratch/profile_dti.py:

this branch: 48.5043 s
master: 57.4576 s

But should save substantially more time when a mask is used.

[arokem]

This one is also an easy review: <40 lines of changed code for a 12% speedup for something that some users probably do 20 times a day. Any one want to take a look?

[MarcCote]

LGTM. It avoids divisions by zero and simplifies the code.

[arokem]

Thanks for taking a look. Now also rebased for ease of merging.

[MarcCote]

@arokem : it seems the rebase went wrong?

[arokem]

Looks like Travis was having a hard time connecting to Github. I just repushed that last commit.

[codecov-io]

Current coverage is 84.61% (diff: 100%)

Merging #1064 into master will increase coverage by 3.74%

@@             master      #1064   diff @@
==========================================
  Files           217        219      +2   
  Lines         24593      24879    +286   
  Methods           0          0           
  Messages          0          0           
  Branches       2491       2515     +24   
==========================================
+ Hits          19888      21051   +1163   
+ Misses         4194       3199    -995   
- Partials        511        629    +118   

Powered by Codecov. Last update da31980...58532e6

[coveralls]

Coverage decreased (-0.01%) to 82.925% when pulling 67b349f on arokem:faster-dti-odf into da31980 on nipy:master.

[coveralls]

Coverage decreased (-0.01%) to 82.925% when pulling 67b349f on arokem:faster-dti-odf into da31980 on nipy:master.

[MarcCote]

@arokem: Should I wait for you to add a small test that actually uses the mask (see codecov)?

[arokem]

oh yeah - thats no good. please wait. ill get to it in the weekend, i hope

On Oct 26, 2016 2:19 PM, "Marc-Alexandre Côté" notifications@github.com
wrote:

@arokem https://github.com/arokem: Should I wait for you to add a small
test that actually uses the mask (see codecov)?

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
#1064 (comment), or mute
the thread
https://github.com/notifications/unsubscribe-auth/AAHPNhnpdK5Ml7mFkjmr2woxUPyGJwiXks5q35mngaJpZM4Io3Kv
.

[coveralls]

Coverage increased (+4.2%) to 87.121% when pulling 03ebcd5 on arokem:faster-dti-odf into da31980 on nipy:master.

[arokem]

How about this? On my machine, the dti module is currently at 99% coverage.

[coveralls]

Coverage increased (+4.2%) to 87.142% when pulling 615283e on arokem:faster-dti-odf into da31980 on nipy:master.

[coveralls]

Coverage increased (+4.2%) to 87.142% when pulling 58532e6 on arokem:faster-dti-odf into da31980 on nipy:master.

[MarcCote]

Excellent!