NEW: Use CuPy fuse to merge some reduction kernels - Githubissues

AdvancedPhotonSource / tike

Repository for ptychography software

http://tike.readthedocs.io

Other

29 stars 15 forks source link

NEW: Use CuPy fuse to merge some reduction kernels #290

Closed carterbox closed 1 year ago

carterbox commented 1 year ago

Purpose

Use CuPy fuse to merge more function calls into single kernels which reduces the use of intermediate memory arrays and the number of kernel launches.

Approach

On a benchmarking dataset, the memory profile starts with 15.29GB of GPU memory acquired. After this PR, only 13.88GB of memory acquired.

Pre-Merge Checklists

Submitter

[ ] Write a helpfully descriptive pull request title.
[ ] Organize changes into logically grouped commits with descriptive commit messages.
[ ] Document all new functions.
[ ] Click 'details' on the readthedocs check to view the updated docs.
[ ] Write tests for new functions or explain why they are not needed.
[ ] Address any complaints from pep8speaks.

Reviewer

[ ] Actually read all of the code.
[ ] Run the new code yourself; the included tests should make this easy.
[ ] Write a summary of the changes as you understand them.
[ ] Thank the submitter.