Subject: Re: CG14 code
To: Jeff Wyman <wysoft@c481444-a.bremtn1.wa.home.com>
From: Erik E. Fair <fair@clock.org>
List: port-sparc
Date: 08/27/2001 17:47:01
In the archives of this list is a message from me from when I 
attempted to use the 24-bit cgeight emulation mode of the cgfourteen 
driver on an SS20 with an 8MB VSIMM in it, attached to a Sun 
GDM-20E10 20" multisync monitor, set at 1280x1024x76m (76 Hz).

It worked OK in console mode, but any attempt to fire up X didn't 
work in the following way: I got what appeared to be four tiny copies 
of the console screen (possibly with X text in 'em; was hard to 
read/interpret) across the top 1/4th of the screen, and then X would 
quit with error messages that I don't recall, but that I'm sure I 
documented in my last message on this subject.

I think our best bet to get this working (and working well) is to pry 
the driver source code loose from Sun; the SX and the SS20 are way 
past EOL, and any proprietary value that code had has long since gone 
to zero, since it has been eclipsed by the PC graphics chip market. I 
believe it's mostly a matter of prodding Sun sharply and insistently 
enough that they release it. Sun has no programming manuals for their 
graphics devices - my reading is that they never expected that 
information to be useful outside their company, so never documented 
them properly. Thus, the source code to their X/whatever display 
server is the only documentation, aside from the ASIC/circuit 
diagrams themselves (and good luck reading that if you're not an EE 
with the right training - I'm definitely not!).

I suggest our goal should be in two steps:

1. get 24-bit mode working with X (proper cgeight emulation; dumb 
frame buffer).

2. get the SX accelerations working to speed it all up.

I offered a $100 US bounty for this work the last time it came up; 
I'll now double the stakes - $100 US for each of the two items above, 
payable when the source code for it is both demonstrated to work, AND 
committed to the NetBSD CVS source repository.

It occurrs to me that there is another way we can attempt this: make 
Sun's X server work under NetBSD; of course, that requires that our 
Solaris emulation be complete enough (I recall reading somewhere that 
SX acceleration was not supported in SunOS 4.x). Of course, I'd 
rather we had source code, so that we can fix bugs that we find.

	somewhere over the North Atlantic, approaching Greenland,

	Erik <fair@clock.org>