IVRCompositor::SetGamma is missing and setting linear gamma is not working

[TheWhiteAmbit]

The described compositor setting is completely missing in the sources, attempt to set linear gamma by hand via

vr::ColorSpace_Linear;

will not render at all (at least with DirectX). Currently using workaround with vr::ColorSpace_Gamma and shader.

[aleiby]

What texture format are you using? Are you not seeing any visual difference in the headset when changing the color space, or are you getting an error?

[TheWhiteAmbit]

The "SetGamma" method is missing completely:
https://github.com/ValveSoftware/openvr/wiki/IVRCompositor::SetGamma

Beside this, the picture is remaining completely black when I set color space from ColorSpace_Gamma to ColorSpace_Linear. My Texture-Format is the following, but I have tried different options (including float and non-sRGB). Right now I am seeing it on a Rift DK2, but it was the same on the Vive Pre.

tdesc.SampleDesc.Count = 1;
tdesc.SampleDesc.Quality = 0;
tdesc.Usage = D3D11_USAGE_DEFAULT;
tdesc.Format = DXGI_FORMAT_R8G8B8A8_UNORM_SRGB;
tdesc.BindFlags = D3D11_BIND_SHADER_RESOURCE | D3D11_BIND_RENDER_TARGET;

[ekmett]

I'm also getting this issue. If I submit in Gamma space what I see on the display matches what I get in the screen, yet if I submit in Linear space what I see on the vive display matches what i get on the screen in the compositor. I can't see a difference for toggling this.

As a result both the compositor and and screen look too dark when I say 'linear' and try to get OpenVR to apply the gamma curve for me, leaving me to apply the gamma curve as part of my tonemapping pass.

I was looking to use glEnable(GL_FRAMEBUFFER_SRGB) at the end in my companion window and switch to linear for eye submission to avoid an admittedly miniscule amount of work.

[ekmett]

In the above case I was submitting a "linear" GL_RGBA8 texture as vr::ColorSpace_Linear.

Possibly related is that the documentation mentions that ColorSpace_Auto will pick based on if the resolve texture is 8 bits or floating point, but if I try to switch to submitting a GL_RGBA16F texture under ColorSpace_Auto it doesn't work and I just get a black compositor window.

Is perhaps the OpenGL side of the code path bit rotted or limited in the formats it will accept for each colorspace?

[TheWhiteAmbit]

This is propably not OpenGL related, since it also happens in Direct3D. The internal handling on Windows should be all Direct3D, so this seems to be also present with not OpenGL involved. Direc3D has explicit linar and sRGB texture formats, but also supports Typeless. It worked with none as expected.

[TheWhiteAmbit]

Just checked it with release 1.0.16 - still not resolved :(