Re: CG14 in 8-bit color

Hello,

On Fri, 3 Dec 2021 18:55:15 +0100
Romain Dolbeau <[email protected]> wrote:

> Le ven. 3 déc. 2021 à 18:25, Michael <[email protected]> a écrit :
> > As promised I finally sat down to make suncg14 work in 8bit.  
> 
> Oh nice :-) Though as the only SPARCstation 24-bits display that isn't
> entirely mythical, it seems weird to use it in 8-bits.
> Even for the higher resolution as Malte does...

If all you need is amiwm and a bunch of xterms / nedits...

> Annoyingly, I couldn't get my pseudo-cg6 in the SBusFPGA to do
> 1920x1200, it seems the 193+MHz pixel clock is too high (a faster FPGA
> would definitely work though). The 148.5MHz of 1920x1080 is fine now.
> (except I still can't for the life of me figure out why my version of
> the cg3 or cg6 causes X11 to corrupt UVM, when I don't change the
> software at all).

IIRC the tgx+ topped out at 1600x1200, according to its firmware.
No clue about the X11 issue, does it happen during framebuffer or
register access?
X ( as in xf86-video-suncg6 ) shouldn't use the chip in any other way
than the kernel does, we map the framebuffer, registers etc., and the
EXA driver doesn't even use the colour expanding goop the kernel uses
for character drawing, it's just host blits, copies and rectangles.

> > That's a weirdness in SX's STBS ( STore Bytes from Scalar )
> > instruction, if you store a bunch of bytes from a scalar you need to
> > repeat the byte-to-be-stored in all four sub-bytes of the source
> > register. Apparently that saved the designers a few gates on the chip,  
> 
> Well that's less routing of data from the LSB to the others (and
> associated control), so it figures out it would be cheaper in silicon.

Yeah, that's more or less what the manual says.

> Is there some easy way to broadcast a byte to all the bytes in the
> (vector) registers, or does it have to be handled by the host?

Possible, but annoying. It's just ~two SPARC instructions and it's done
rather infrequently, so no big deal. Having the SX do it would be a bit
of a waste.

> They probably didn't expect much accelerated use in 8-bits -
> presumably the primary design point was 8-bits console with
> accelerated scrolling and 24-bits X11 with full acceleration.

Yeah, and 1152x900 wasn't much of a limitation in the early 1990s
either.

> > I fixed a handful pixels-are-32bit (...)  
> 
> By the way, thank you so much for doing all that work for
> NetBSD/sparc, it is very much appreciated!

It's always nice to find out there's still interest in these machines :)

have fun
Michael