Re: Improved memory detection for cgsix

[Michael <macallan%netbsd.org@localhost>]

Hello,

> > Alright, I'll definitely dig up a 4MB board and do some experiments. It
> > would be great if we could use all the memory.

And here's what I just did:
	cgsix_rectfill(sc, 10, 10, 300, 300, 0xe0e0e0e0);
	cgsix_rectfill(sc, 10, 1821, 300, 600, 0x07070707);
	cgsix_bitblt(sc, 0, 1821, 400, 0, 400, 400, CG6_ALU_COPY);

This draws a red square into the visible screen, then draws a blue one
into VRAM just beyond 2MB ( 2MB divided by 1152 -> 1820 ), then blits
that into the visible screen next to the red square.
The idea being, if we wrap around at 2MB the blue square will distort
the visible screen, if we hit unmapped space the attempt to copy the
blue square will get us garbage or errors.
On my 501-2253 I got an undisturbed red square and next to it a blue
one with an uninitialized memory pattern around it, so we didn't wrap
and we didn't hit unmapped space either, which means the blitter can
indeed see the full 4MB.
So, I was wrong about the whole thing. I thought it would at least need
some register voodoo to disable double buffering and allow access to
the full framebuffer, but apparently it doesn't even need that.

have fun
Michael