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

Updated and fixed the MapExpansion transformation. #1532

Merged
merged 11 commits into from
Feb 27, 2024

Conversation

philip-paul-mueller
Copy link
Collaborator

Before the transformation ignored the tiling parameter of the range, this was fixed. The transformation is now also able to limit the expansion. Thus it is possible to only expand the first k dimensions, the remaining dimensions will be kept in a multi dimensional map.

This is a feature that will be need in GT4Py.

Before the transformation ignored the tiling parameter of the range, this was fixed.
The transformation is now also able to limit the expansion.
Thus it is possible to only expand the first k dimensions, the remaining dimensions will be kept in a multi dimensional map.

This is a feature that will be need in GT4Py.
The computation of the new maps is now its own function, which allows subclassing.
In addition now the schedule of teh first map is explicitly set, before there was a slight error, that was, however, not important because the value was not used.
This commit makes it a bit more verbose, hope this helps me to debug it.
…lso seems that nothing was in the state.

Now be even more verbose and explore everything.
It seems that there is an init state on the CI which is not present on my machine.
This commit actually fixes a bug in the test.
Copy link
Contributor

@lukastruemper lukastruemper left a comment

Choose a reason for hiding this comment

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

The PR itself looks good, but the question is whether this is needed:

  • One goal is to have simple, composable transformations, e.g., MapExpansion+MapCollapse.
  • The tile size parameter should be deprecated IMO and users should use MapTiling or StripMining which creates explicit maps? The parameter is not considered during dataflow analysis

@tbennun

Copy link
Contributor

@lukastruemper lukastruemper left a comment

Choose a reason for hiding this comment

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

LGTM

@BenWeber42 BenWeber42 added this pull request to the merge queue Feb 27, 2024
Merged via the queue into spcl:master with commit ba1587e Feb 27, 2024
11 checks passed
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