Skip to content

Apply __new__ approach to disabling __init__#484

Merged
leofang merged 3 commits intoNVIDIA:mainfrom
rwgk:uniform_init_disables
Mar 3, 2025
Merged

Apply __new__ approach to disabling __init__#484
leofang merged 3 commits intoNVIDIA:mainfrom
rwgk:uniform_init_disables

Conversation

@rwgk
Copy link
Collaborator

@rwgk rwgk commented Mar 3, 2025 •
edited
Loading

Contributes to #198

Make disabling __init__ uniform across cuda.core:

  • Use @classmethod as the only constructor: Best for strict enforcement.

  • Uniform raise RuntimeError() with a standardized message form.

The starting point for this change was extracted from PR #458:

@rwgk
Apply __new__ approach to disabling __init__
2e6201f 
The starting point for this change was extracted from PR NVIDIA#458:

* NVIDIA#458 (comment)

* NVIDIA#458 (comment)
@copy-pr-bot
Copy link
Contributor

copy-pr-bot bot commented Mar 3, 2025

Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.

@rwgk
Copy link
Collaborator Author

rwgk commented Mar 3, 2025

/ok to test

@github-actions

This comment has been minimized.

Sign in to view
@leofang
Copy link
Member

leofang commented Mar 3, 2025

Looks fine to me! Merge?

@rwgk
Copy link
Collaborator Author

rwgk commented Mar 3, 2025

I forgot that I wanted to add a couple more small tests. Doing that now.

@rwgk
Copy link
Collaborator Author

rwgk commented Mar 3, 2025

@leofang Test coverage is now systematically complete, covering "front doors" and "back doors"; see comments added with commit eda324f.

Testing that the back doors are locked is done through APIs that are unavoidably (for all practical purposes) visible through public APIs.

Recently I experienced pushback when testing private functionality under the CCCL repo, so I asked ChatGPT about it:

The lists of pros and cons are complex. I'm convinced it would be unproductive to be to pure one way or another, case-by-case judgement is important. That's how I decided to add tests that ensure the back doors are locked, and remain so.

@rwgk
Copy link
Collaborator Author

rwgk commented Mar 3, 2025

/ok to test

leofang
leofang previously approved these changes Mar 3, 2025
View reviewed changes
Copy link
Member

@leofang leofang left a comment

Choose a reason for hiding this comment

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

LGTM other than fixing one error message.

cuda_core/cuda/core/experimental/_stream.py
Comment on lines +88 to +89
self._device_id = None # delayed
self._ctx_handle = None # delayed
Copy link
Member

@leofang leofang Mar 3, 2025 •
edited
Loading

Choose a reason for hiding this comment

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

IIUC this is actually a bug fix too! I wonder why it wasn't caught before...

cuda_core/cuda/core/experimental/_stream.py
Comment on lines +98 to +99
self._device_id = None # delayed
self._ctx_handle = None # delayed
Copy link
Member

@leofang leofang Mar 3, 2025

Choose a reason for hiding this comment

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

ditto

@rwgk
Remove wait from list of suggested Stream APIs:
201d3fc 
NVIDIA#484 (comment)

Right now, if one wants to create a stream from a protocol-supported object, the intended way is to do `s = Device().create_stream(object_with_dunder_cuda_stream)`.
@leofang
Copy link
Member

leofang commented Mar 3, 2025

Since the previous run was green and the most recent change is comment-only, I suggest we admin-merge this without another CI run. @rwgk ready?

@rwgk
Copy link
Collaborator Author

rwgk commented Mar 3, 2025

Since the previous run was green and the most recent change is comment-only, I suggest we admin-merge this without another CI run. @rwgk ready?

Yes, please! :-)

@leofang leofang marked this pull request as ready for review March 3, 2025 17:49
@copy-pr-bot
Copy link
Contributor

copy-pr-bot bot commented Mar 3, 2025

Auto-sync is disabled for ready for review pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.

@leofang leofang merged commit 043669b into NVIDIA:main Mar 3, 2025
1 check passed
@leofang leofang added bug Something isn't working enhancement Any code-related improvements P0 High priority - Must do! cuda.core Everything related to the cuda.core module labels Mar 3, 2025
@leofang leofang added this to the cuda.core beta 3 milestone Mar 3, 2025
@rwgk rwgk deleted the uniform_init_disables branch March 3, 2025 17:51
@github-actions
Copy link

github-actions bot commented Mar 3, 2025

Doc Preview CI
Preview removed because the pull request was closed or merged.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@leofang leofang leofang approved these changes

Assignees

@rwgk rwgk

Labels

bug Something isn't working cuda.core Everything related to the cuda.core module enhancement Any code-related improvements P0 High priority - Must do!

Projects

None yet

Milestone

cuda.core beta 3

Development

Successfully merging this pull request may close these issues.

2 participants

@rwgk @leofang