[Trac #1017] swizzle and other DCON-related display problems

[Zarro Boogs per Child]

#1017: swizzle and other DCON-related display problems
---------------------------+------------------------------------------------
 Reporter:  AlbertCahalan  |        Owner:  blizzard
     Type:  defect         |       Status:  reopened
 Priority:  normal         |    Milestone:  BTest-3 
Component:  hardware       |   Resolution:          
 Keywords:                 |  
---------------------------+------------------------------------------------
Comment (by vmb):

 Yes, I know that filtering and luminance-extraction oughtn't be done on
 nonlinear data (I've worked on the design of chips in commercial products
 that do this thing the right way) but because of limited hardware
 resources Mark had available in the DCON we had to do a simple hack --
 Albert's non-rotationally-symmetric filter probably would have been a
 better idea.  In a later generation of the machine where we get a decent
 display driver circuit integrated with the graphics subsystem we can do
 the totally correct thing; if we ever re-do the DCON before then we can
 try to fit in a better filter even if we can't put it between gamma
 tables.

 So everyone knows what the DCON does:

 The filter kernel is (left-to-right, top-to-bottom) (0 1 0, 1 4 1, 0 1
 0)/8.  The Y channel is computed as
 (R >> 2) + (R >> 4) + (G >> 1) + (G >> 4) + (B >> 3)

 Note also that the current DCON has only 16-bit color line buffers (cost
 and space, again), so there isn't going to be a lot of visual finesse
 available in any event.

-- 
Ticket URL: <http://dev.laptop.org/ticket/1017#comment:5>
One Laptop Per Child <http://laptop.org/>