Is there a performance gain with GTSAM_BUILD_WITH_MARCH_NATIVE ? #1760
Unanswered
TobiasLangner-TomTom
asked this question in
Q&A
Replies: 1 comment
-
|
You definitely need to measure on your own platform, please refer to the Eigen documentation for details. |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
I'm optimizing a simple SLAM problem with pose (Pose3) and observation (Point3) variables. The factors are various priors and between factors. The problem sizes are rather large with thousands of poses and observations.
TBB is not enabled because multi-threading is managed outside of GTSAM.
According to INSTALL.md building with the option
GTSAM_BUILD_WITH_MARCH_NATIVEshould bring a performance gain of 25-30%. However, I did not observe any improvement for my use case.I would be grateful if anyone could share their experiences with GTSAM_BUILD_WITH_MARCH_NATIVE .
Does anyone know how the 25-30% were measured?
Which parts of GTSAM would benefit the most from GTSAM_BUILD_WITH_MARCH_NATIVE?
Beta Was this translation helpful? Give feedback.
All reactions