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.

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

Refine global - register -global transfers #55

Closed
tnowotny opened this issue Aug 19, 2015 · 4 comments
Closed

Refine global - register -global transfers #55

tnowotny opened this issue Aug 19, 2015 · 4 comments

Comments

@tnowotny
Copy link
Member

within the learn code, there can be repeated reference to dd_remapXX[iprePos]. Whenever it is used more than once, it would probably be efficient to fetch it into a local register.

@tnowotny tnowotny self-assigned this Aug 19, 2015
@tnowotny
Copy link
Member Author

tnowotny commented Dec 4, 2018

  1. load dd_remapXX[...] into a register by default.
  2. But more generally, introduce an annotation "readonly XXX YYY" e.g. readonly scalar x; for variables. Those would not be written back to global memory; the register copy could be declared const

@tnowotny tnowotny changed the title In learning kernel there seem to be inefficiencies in sparse implementation Refine global - register -global transfers Dec 4, 2018
@tapaswenipathak
Copy link

Hi folks: Is this ticket still open?

@neworderofjamie
Copy link
Contributor

Well, the read-only part has been implemented by #247 and the equivalent of dd_remap[iprepos] is now loaded into a register here. The issue is that that is done whether it is required or not as discussed in #248

@tapaswenipathak
Copy link

All PRs are there then right, can you ref the ticket in the PRs?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

3 participants