Closed NagabhushanSN95 closed 2 years ago
I think this is an oversight on our part. I doubt it matters, they're both valid choices, and dy and dx should always be pretty similar values.
Please let me know if performance is very sensitive to dx vs (dx+dy)/2, that would be interesting!
Okay. Thanks for the clarification. In case I test it out, I'll update here :)
The
radii
computation code is different for non-ndc and ndc spaces. In particular, without ndc,radii
computation uses onlydx
derived fromdirections
, but when ndc is enabled,(dx+dy)/2
is used, which is derived fromndc_origins
. Can you please shed some light on why it is done so?