Open annevk opened 7 years ago
OpenGL uses this notation: RGB10_A2
Converted to a string enum that would be "rgb10-a2", which seems more reasonable to me.
It was indeed surprising to see float16
, 8-8-8-8
(and, in the WhatWG Spec PR, 10-10-10-2
). I guess the idea of avoiding OpenGL's Sized Internal Format
(GL_RGBA16F
, GL_RGBA8
and GL_RGB10_A2
, respectively), is to not specify the channel endianness and using instead what the platform provides.
Alternatively, what we can do is to use OpenGL's Type
which will give us: GL_UNSIGNED_BYTE
, GL_UNSIGNED_INT_2_10_10_10_REV
and GL_HALF_FLOAT
, specifying the storage of the channels but leaving the channels undefined.
@fserb WDYT?
OpenGL's relevant Spec is glTexImage2D, see the Table 2 here.
I see now that 10-10-10-2
(and 12-12-12-12
, for the case) were removed in #22 in an effort to simplify the options offered.
"10-10-10-2" doesn't seem great. Are there no names for this kind of thing we could use?