Improved version of 2a3a and 3b2b transposes #363
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
feathern
approved these changes
Apr 11, 2022
|
There was a small conflict in changlog.md that I resolved. I'm merging this now. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR provides an option to use slightly faster versions of the 2a3a and 3b2b transposes. The original transpose routines (now the default) assumed that the m_values were stored as high/low pairs as: m = 0, m_max, 1, m_max-1, 2, m_max-2, .... Then various pairs of m values are distributed across rows/columns. For a process that holds the first 3 pairs of m values, the order would be: m=0,m_max,1.m_max-1,2,m_max-2. This means accessing the m values in order is not contiguous leading to a small performance hit.
The new 2a3a transpose routine relies on a different storage of the m values, such that the local chunk of m values is contiguous. For a process that holds the first 3 pairs of m values, the ordering will now be: m=0,1,2,m_max,m_max-1,m_max-2. This allows for contiguous access of the m values.
The new 3b2b transpose routine simply reorders some of the for loops to take advantage of the Fortran layout, i.e., column major loop ordering.