cuda_core forward compatibility changes.

[rwgk]

Description

Forward compatibility changes for cuda_core.

This helps us staying organized.

[copy-pr-bot]

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]

/ok to test

[rwgk]

/ok to test

[rwgk]

/ok to test

[copy-pr-bot]

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]

Q: If we're packing anyway, maybe we pack them together?

Suggested change

	status, _, graph, *_, _ = handle_return(driver.cuStreamGetCaptureInfo(self._mnff.stream.handle))
	status, _, graph, *_ = handle_return(driver.cuStreamGetCaptureInfo(self._mnff.stream.handle))

[leofang]

@rwgk let's merge #709, resolve the conflicts, rerun the CI and merge.

[vzhurba01]

I took a look at the graphs side and while I'm ok with it, it does feels over-engineered.
The "tricks" used with deps_info_update feel bad because it's scattered through out the code (just 2 places but still... what if we end up doing more tricks to handle other forward compatibility APIs).

Thinking out loud, what if we move these weird APIs to a compatibility file/layer? At least this way we're keeping these tricks contained in a single place. Oh and by tricks I mean logic like [None] * (len(deps_info) - 1), and the packing * which is placed for some APIs at the end while in the middle for others...

[leofang]

@rwgk let's merge #709, resolve the conflicts, rerun the CI and merge.

FYI, #709 is merged

[rwgk]

/ok to test

[rwgk]

I took a look at the graphs side and while I'm ok with it, it does feels over-engineered.

I was very careful to make the unpacking consistent between all the uses, even where I could have taken a shortcut (*_, _), to reduce the cognitive burden. I find that far easier to read than duplicate code (what the LLMs suggested initially). The unpacking feature seems very intuitive and complete to me.

[None] * (len(deps_info) - 1)

Yeah, it's a bit over the top, but there are only two, and the circumstances are special.

[leofang]

Thanks, Ralf!

[github-actions]

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