Why TextureFormat::R16Unorm missing?

[aloucks]

Description
TextureFormat::R16Unorm does not exist. I thought this this previously existed in WGPU.

Expected vs observed behavior
I would expect there to be a 16 bit unsigned normalized type that has a float sample type.
https://gpuweb.github.io/gpuweb/#texture-format-caps
r16unorm, float

Extra materials
I want to sample from 16bit height data. The current work around is to convert it to 8bit which loses detail (and looks awful).

Platform
wgpu 0.11.0 (via bevy pipelined-rendering branch)

[CatCode79]

Hello @aloucks, I found this issue which I think is related:
#999

[aloucks]

Although he mentions R16Unorm in the title, if you look in the details, he's actually trying to sample unsigned integer data. I'm trying to sample floating point data that's stored as an unsigned 16bit value.

I'm looking for the 16bit version of R8unorm, for higher precision.

[kvark]

The short answer is - because 16-bit normalized formats are not portable, so they were excluded from WebGPU baseline.
Both us and WebGPU upstream would welcome an optional feature to expose them.
Closing as a dupe of #1934