Skip to content

WIP: Implement __setitem__ for DeviceBuffer #351

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
Closed
wants to merge 1 commit into from
Closed

WIP: Implement __setitem__ for DeviceBuffer #351

wants to merge 1 commit into from

Conversation

jakirkham
Copy link
Member

This is a rough first pass at implementing __setitem__ for DeviceBuffer. It works based on a quick check, but the error handling could be improved.

@jakirkham jakirkham mentioned this pull request Apr 21, 2020
Open
@GPUtester
Copy link
Contributor

Please update the changelog in order to start CI tests.

View the gpuCI docs here.

@jrhemstad
Copy link
Contributor

What is this doing?

@jakirkham
Copy link
Member Author

Initially PR ( dask/distributed#3732 ) was using Numba to view and write into DeviceBuffers, but the overhead was a bit high as @kkraus14 suspected and later testing bore out. This is an attempt to add __setitem__ to bypass needing Numba. That said, it appears using CuPy works well. We may still want this, but don't seem to need it currently.

@harrism
Copy link
Member

harrism commented Apr 22, 2020

Can you explain what __setitem__ is for us Python noobs?

@kkraus14
Copy link
Contributor

__setitem__ is a Python special method that allows you to do object[key] = some_other_object. For us here, we want to handle where key is a slice which under the hood we use to translate to a pointer and a size to run a memcopy.

kkraus14
kkraus14 reviewed Apr 22, 2020
View reviewed changes
python/rmm/_lib/device_buffer.pyx
Comment on lines +120 to +122
rng = range(self.size)[key]
if not isinstance(rng, range):
rng = range(rng, rng + 1)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We should probably handle negative slices / scalars here

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Negative indices are normalized by slicing the range object.

In [1]: range(5)[-1]
Out[1]: 4

Currently we are raising an error for negative indices or any other step size that is not 1 since this could be a bit tricky to do in practice and it's unclear if we have use cases for those yet.

@harrism harrism added Python Related to RMM Python API 2 - In Progress Currently a work in progress labels May 28, 2020
@harrism harrism changed the base branch from branch-0.14 to branch-0.15 May 28, 2020 00:34
@harrism
Copy link
Member

harrism commented Aug 7, 2020

@jakirkham @kkraus14 what are the plans for this? Moving to 0.16.

@harrism harrism changed the base branch from branch-0.15 to branch-0.16 August 7, 2020 03:30
@jakirkham
Copy link
Member Author

I think we still want this. Has slipped lately though due to other priorities. Will take a look after the release.

@github-actions
Copy link

This PR has been marked rotten due to no recent activity in the past 90d. Please close this PR if it is no longer required. Otherwise, please respond with a comment indicating any updates.

@github-actions
Copy link

This PR has been marked stale due to no recent activity in the past 30d. Please close this PR if it is no longer required. Otherwise, please respond with a comment indicating any updates. This PR will be marked rotten if there is no activity in the next 60d.

@harrism
Copy link
Member

harrism commented Feb 16, 2021

@jakirkham @kkraus14 can we close this?

@kkraus14
Copy link
Contributor

Yes, closing this for now. Feel free to reopen if desired.

@kkraus14 kkraus14 closed this Feb 17, 2021
@jakirkham
Copy link
Member Author

Yeah sounds good. I think people can probably just use CuPy or Numba for this today

@jakirkham jakirkham deleted the add_devbuf_setitem branch February 17, 2021 03:14
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@kkraus14 kkraus14 kkraus14 left review comments

Assignees
No one assigned
Labels
2 - In Progress Currently a work in progress Python Related to RMM Python API
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
5 participants
@jakirkham @GPUtester @jrhemstad @harrism @kkraus14