feat: use a very large coil dimension for 2D stacked NUFFT #39
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
paquiteau
changed the title
paquiteau
force-pushed
the
faster-stacked
branch
from
eadde29 to
0f96c5f
Compare
paquiteau
force-pushed
the
faster-stacked
branch
from
e9b9885 to
8380064
Compare
|
Alright some simple timing (H2D2H) reports the followings on a QUADRO P5000 with the data from the
Overall that's a 2x speed up on the Forward + Adjoint step (= a data consistency), it can be made faster by avoiding the roundtrip in the middle, and even more when cufinufft support the async copy / computations The stacked 2D Nufft on GPU also has a lower memory footprint (the oversampling grid is only 4 times bigger , not 8 times) making it possible to potentially have multiple operators running together (think fMRI) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR improve the stacked nufft, by leveraging the coil dimension of the 2D operator.
For gpu backend which does asynchronous copy of coil data (e.g. only gpuNUFFT for now, cufinufft needs a few PR being merged upstream) this is great news, and potentially leads to increase in speed.
I still need to setup some rudimentary benchmark to compare between the two.