[Trac #336] Gamma-ram workaround for data line shift only for Btest

[Zarro Boogs per Child]

#336: Gamma-ram workaround for data line shift only for Btest
-------------------------------+--------------------------------------------
 Reporter:  wmb at firmworks.com  |        Owner:  dwmw2  
     Type:  defect             |       Status:  new    
 Priority:  blocker            |    Milestone:  BTest-3
Component:  hardware           |   Resolution:         
 Keywords:                     |  
-------------------------------+--------------------------------------------
Comment (by AlbertCahalan):

 Replying to [comment:3 dcbw]:
 > We need to _really_ fix this for B3 in hardware.  The problem is that
 the Geode can only apply the gamma correction table to _one_ stream of
 [graphics, video].  That breaks non fullscreen camera usage completely.

 Uh, couldn't an updated DCON just do this nicely? It'll be a custom ASIC
 right in the middle anyway, so it can do whatever: separate mappings for
 each sRGB rendering intent and for Y'CbCr, loadable 5x5 convolution
 matrixes separate for each pixel color, dithering from 24-bit to 18-bit,
 etc. With either 16-bit 1:5:5:5 or 32-bit 8:8:8:8 data, you could spare a
 bit for a convolution bypass flag. I'm guessing you're limited by pin
 count, not silicon real estate. Heck, throw in a random number generator.

 BTW, it's creepy and probably bad to be changing things for video. Better
 would be a global setting for sRGB rendering intent. It could just be a
 color adjustment, to be chosen much like backlight brightness is chosen.
 Color mapping is a personal preference.

-- 
Ticket URL: <http://dev.laptop.org/ticket/336#comment:7>
One Laptop Per Child <http://laptop.org/>