Skip to content
This repository has been archived by the owner on Sep 28, 2024. It is now read-only.

support GPU #7

Merged
merged 6 commits into from
Aug 11, 2021
Merged

support GPU #7

merged 6 commits into from
Aug 11, 2021

Conversation

foldfelis
Copy link
Contributor

@foldfelis foldfelis commented Aug 11, 2021 •
edited
Loading

This is awkward...

  1. FFTW.fft and FFTW.ifft work on CUDA only if dims=1.
  2. According to documents, NNlib.batched_adjoint has same behavior with both batched_transpose, and PermutedDimsArray, but neither of them works on CUDA. See this and this.
  3. I use @tullio to construct CUDA kernels for batched transpose and einsum to do the calculation.

@foldfelis foldfelis added the enhancement New feature or request label Aug 11, 2021
@foldfelis foldfelis self-assigned this Aug 11, 2021
@foldfelis foldfelis merged commit 609e6dd into master Aug 11, 2021
@foldfelis foldfelis deleted the gpu branch August 11, 2021 18:45
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Reviewers
No reviews
Assignees

@foldfelis foldfelis

Labels
enhancement New feature or request
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
1 participant
@foldfelis