atan2 is implemented as atan(y/x) in gpu backend #647
Comments
|
https://tc39.es/ecma262/#sec-math.atan2 The spec for atan2 is pretty sophisticated and long... |
|
IIUC, GLSL atan is overloaded to take both 1 argument (like Math.atan) and 2 arguments (like Math.atan2). AFAIK the second implementation is identical to C/javascript. https://www.khronos.org/registry/OpenGL-Refpages/gl4/html/atan.xhtml |
|
Can confirm that pull request fixes supplied repro case.
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
What is wrong?
atan2 should return values from -pi to pi, but instead returns values from -pi/2 to pi/2 in gpu mode. It also returns undefined results for x=0.
Where does it happen?
In Chrome browser.
How do we replicate the issue?
I'm attaching a repro case - correct behavior can be seen in mode:cpu, incorrect behavior in mode:gpu.
atan2-bug.zip
How important is this (1-5)?
3 - I can work around, but also should be an easy fix.
Expected behavior (i.e. solution)
The backend GLSL implementations of atan2 use the single-argument version of atan(y_over_x), instead of the two-argument version atan(y, x).
The text was updated successfully, but these errors were encountered: