Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Allow seperate workspaces for convolution/pooling #27

Open
hobofan opened this issue Feb 21, 2016 · 0 comments
Open

Allow seperate workspaces for convolution/pooling #27

hobofan opened this issue Feb 21, 2016 · 0 comments
Labels
breaking F-CUDA Perf-Memory refactor

Comments

@hobofan
Copy link
Member

hobofan commented Feb 21, 2016

#18 reduces the memory usage for use cases where you need the workspace for both forward and backward (training).

I haven't done any testing but it could be possible the forward workspace is smaller than the backward one, leading to to higher memory usage than necessary in pure forward(/inference) use cases.

This would also help seperate the automatic convolution algorithm detection for those uses cases, leading to quicker startup time (this should also be possible right now by using an Algo different than Auto, but would clearer).
hobofan added a commit that referenced this issue Mar 2, 2016
@hobofan
feat/convolution: change convolution functions to require workspace
f9d4013 
BREAKING CHANGE: All convolution functions now require
a SharedTensor<u8> workspace to be passed.
This allows for reuse of the workspace between different
convolution operations and a global shared workspace.

REFERENCE #27
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
breaking F-CUDA Perf-Memory refactor
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
1 participant
@hobofan