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

Investigate and implement switching to fp16 precision as default on mobile platforms #2596

Open
raytranuk opened this issue Dec 7, 2020 · 1 comment
Labels
area: graphics Graphics related issue feature request performance Relating to load times or frame rate

Comments

@raytranuk
Copy link
Contributor

Assess workload and measure performance gains and quality.

Inspired by this excellent presentation: https://community.arm.com/cfs-file/__key/communityserver-blogs-components-weblogfiles/00-00-00-20-66/5_2D00_mmg2020_2D00_filament_2D00_romain.pdf

image

@raytranuk raytranuk added enhancement area: graphics Graphics related issue labels Dec 7, 2020
@mvaligursky mvaligursky added the performance Relating to load times or frame rate label Jan 23, 2021
@raytranuk
Copy link
Contributor Author

This was investigated - on low end devices that are GPU bound doing a global switch resulted in 20-25% increase in performance - and some severe artifacts. More experimentation showed that most the found artifacts can be dealt with by marking some values as highp. Further experimentation has been paused and after discussion, we plan to:

  1. Mark only things we know are safe to medp when a #define is specified
  2. Only perform need clamping when the #define is specified
  3. Switch the #define on by default on mobile devices
  4. Test thoroughly on a range mobile devices

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
area: graphics Graphics related issue feature request performance Relating to load times or frame rate
Projects
None yet
Development

No branches or pull requests

3 participants