This bugzilla service is closed. All entries have been migrated to https://gitlab.com/libeigen/eigen
Bug 1712 - Sub-Optimal code for mixed complex*real products
Summary: Sub-Optimal code for mixed complex*real products
Status: CONFIRMED
Alias: None
Product: Eigen
Classification: Unclassified
Component: Core - matrix products (show other bugs)
Version: 3.4 (development)
Hardware: All All
: Normal Performance Problem
Assignee: Nobody
URL:
Whiteboard:
Keywords:
Depends on: 1019
Blocks: 3.x
  Show dependency treegraph
 
Reported: 2019-05-10 16:54 UTC by Christoph Hertzberg
Modified: 2019-12-04 18:38 UTC (History)
2 users (show)



Attachments

Description Christoph Hertzberg 2019-05-10 16:54:25 UTC
For column-major matrices A*B, with A complex and B real would be equivalent to multiplying two real matrices where A has twice the height. Currently, Eigen is not able to vectorize this properly. Actually, clang is able to auto-vectorize this (for full FMA usage, sometimes this requires -ffast-math): https://godbolt.org/z/ILd8tL

I guess complex*real could easially be solved by Bug 1019.
But real*complex will require some new logic (essentially, the real matrix would need to have each row duplicated).
Comment 1 Nobody 2019-12-04 18:38:11 UTC
-- GitLab Migration Automatic Message --

This bug has been migrated to gitlab.com's GitLab instance and has been closed from further activity.

You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.com/libeigen/eigen/issues/1712.

Note You need to log in before you can comment on or make changes to this bug.