Skip to content

Revert "Prepare drop packunpack (#9994)"#10037

Merged
fjetter merged 1 commit intodask:mainfrom
fjetter:revert_hlg_foo
Mar 8, 2023
Merged

Revert "Prepare drop packunpack (#9994)"#10037
fjetter merged 1 commit intodask:mainfrom
fjetter:revert_hlg_foo

Conversation

@fjetter
Copy link
Copy Markdown
Member

@fjetter fjetter commented Mar 8, 2023

This reverts commit 4574dc5.

Apparently there was a significant performance regression with #9994 so I suggest reverting it until we figured this out.

cc @hendrikmakait

Copy link
Copy Markdown
Member

@jrbourbeau jrbourbeau left a comment

Choose a reason for hiding this comment

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

This is fine by me. Is there somewhere with more info about the regression?

Copy link
Copy Markdown
Member

@hendrikmakait hendrikmakait left a comment

Choose a reason for hiding this comment

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

LGTM assuming CI will be green.

@hendrikmakait
Copy link
Copy Markdown
Member

coiled/benchmarks#706 is the mentioned regression

Comment on lines +412 to +424
self.annotations = self.annotations or {}
if "priority" not in self.annotations:
self.annotations["priority"] = {}
self.annotations["priority"]["__expanded_annotations__"] = None
self.annotations["priority"].update({_key: 1 for _key in self.get_split_keys()})

def get_split_keys(self):
if self._split_keys is None:
self._split_keys = [
stringify((self.split_name, part_out, part_in))
for part_in in range(self.npartitions_input)
for part_out in self.parts_out
]
return self._split_keys
# Return SimpleShuffleLayer "split" keys
return [
stringify((self.split_name, part_out, part_in))
for part_in in range(self.npartitions_input)
for part_out in self.parts_out
]
Copy link
Copy Markdown
Member Author

Choose a reason for hiding this comment

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

I suspect this is causing the regression. What I need for the refactoring is to drop the __expanded_annotations__

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

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants