🚀 go-pugleaf

Our group is acquiring a number of 3/110C machines, and we are interested in
running X on them.  I have been following the discussion, and have some
questions about the 110C frame buffer organization and how it affects the
X implementation.

*** MY IMPRESSION *** of the 110C, is that it is really sort of a "1 1/2"
frame buffer machine (please correct me if I am wrong).  In addition to
an eight bit deep lookup table driven color frame buffer, the machine
also has a single pixel deep "text" black and white frame buffer.  A "control"
bitplane is used to select which buffer will be displayed (note: I do not
know whether the control plane selects between the two buffers, or whether
it masks the color buffer).  I suspect that the SUN engineers who designed
the machine found that:

	1. Most color graphics (line drawing, etc.) were easier with
	   an 8 bit contiguous pixel frame buffer organization
	   (I assume SUN uses contigous pixels... If they used stacked
	   color planes, the control plane would be unnecessary).

	2. Doing text with a contigous pixel buffer was too slow.

	3. 2 megabits (256K) of memory for the B/W buffer and control plane was
	   cheaper than putting in high powered hardware.

If my understanding is correct, then the real question is "how can the
X server best make use of the 110C hardware".  All of the notices I have
seen so far appear to address the question "how can we get the X server
to draw in color".

Now, with some clever programming it should be possible to mask the architecture
under the PIXRECT library.  If SUN has done this, then they have already
solved the first order problem.  If not, then something needs to be
done on top of the PIXRECT library.

If PIXRECT already masks the hardware, then the questions becomes:

	1. What assumptions does PIXRECT make and how do they interact
	   with the X11 implementation?

	2. Does SUN (though the "emulation modes") allow access to
	   the different frame buffers.

	3. How could the V11 server (no reason to work on V10.4
	   unless performance is really bad) take advantage of the
	   architecture.

	   For example, one of the features of V11 allows a pixmap
	   depth to be specified for a window.  An obvious approach
	   would be to special case a one bit deep window with no
	   color map.  Would this give a performance boost for text
	   windows?

	   Would it be worth while trying to model the machine as a
	   ten bit deep buffer with a strange, partially fixed,
	   color map?

-- Jim
-- 
(Grapevine) mayer.wbst			(Arpa) mayer.wbst@Xerox.com
(NS) mayer:wbst128:xerox		(Phone) (716) 422-2496
(UUCP) rochester!rocksanne!mayer