[Intel-gfx] [PATCH] sna/uxa: Fix colormap handling at screen depth 30.

Ville Syrjälä ville.syrjala at linux.intel.com
Fri Mar 16 10:24:29 UTC 2018


On Fri, Mar 16, 2018 at 12:19:22AM +0100, Mario Kleiner wrote:
> I made a proof of concept patchset against i915-kms to use the 1024
> slot, 10 bit wide hw lut's starting with Ironlake, and to upsample
> from the legacy 256 slots to those 1024 slot luts. Also some hacks to
> enable dithering from 10 bit -> 8 bit panels via a module parameter,
> because i currently don't have a true 10 bit panel to test with.
> That's what i used for testing with the Mesa patches, but the code
> wasn't quite ready for upstreaming, and i spent all the time since
> then with the userspace bits.
> 
> Have you thought about also exposing the 512 slot, 12 bit (or was it
> 14 bits?) wide hw luts, with degamma and CTM disabled? It wouldn't
> provide a slot for each of the 1024 framebuffer output values, but
> would give some extra precision for gamma correction on the output
> side for HDMI/DisplayPort deep color with 12 bit or 16 bit panels, or
> even on 10 bit panels + dithering?

Yes, my current plan is to expose all the modes the hw has. The 10bpc
LUT stuff isn't sufficient for HDR at least. I'm also playing around
with fp16 scanout at the moment so that I can at least test the higher
precision gamma without having the lower fb precision work against me.
And I think eventually we'll want fp16 for HDR stuff as well.

-- 
Ville Syrjälä
Intel OTC


More information about the Intel-gfx mailing list