Re: SPARCStation VSIMM

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

Hello,

On Wed, 27 Oct 2021 21:19:53 -0600
Aaron <aranderson%gmail.com@localhost> wrote:

> Thank you to everyone who replied and for the great info below
> Michael! As you suggested, it seems the Southland and Fujitsu have
> both passed into legend, but I did manage to track down an 8MB SX!
> Paid more than I'd have liked, but nowhere near the current eBay rate
> at least.

Nice!
There were a few companies that made SBus graphics cards, some of them
were just rebadged cg3 and cg6, sometimes with vga connectors instead
of 13w3. Unfortunately most of what shows up on ebay falls into this
category :/ 

> Just to confirm a few things, it works great at 1280x1024@24bpp (known
> by many, but since it was cited as a potential concern earlier in the
> thread). Increasing to 24-bit also resolved the NetSurf blank window
> issue so I'm partying like it's 1994.

With the 8MB VSIMM you can theoretically do 1920x1080 in 24bit.
Seriously. Look at the words defined on the cgfourteen node, there
should be a few r1280x1024x66 style video mode definitions, and one
'hdtv'. Imagine trying to get a monitor supporting that in 1994...
That said, the suncg14 driver in -current supports a good part of the
most common xrender operations ( basically, what's needed to render
anti-aliased text and images with transparency ), it's disabled by
default since there aren't many people testing it. Options "xrender"
"true" will enable it.
If you decide to do so please send feedback on whatever issues you may
run into.
Matter of fact I'd like to see both dmesg and Xorg.0.log, mostly
because there may be hardware revisions out there that I haven't seen,
which may need special handling. That actually happened a while ago,
someone tried the driver on an SX with a different hardware revision
than what I have and it indeed caused problems. Turned out mine just
ignores the low address bits in 32bit writes ( you send load/store
instructions to SX by writing the instruction into SX's address space
at an address corresponding to the memory address you want to access,
that way sending both instruction and data in one write ) - the driver
didn't zero these bits since thats extra work, the other hw revision
rejected unaligned writes, so I had to adapt the driver.
Don't expect a huge speed boost, this is still early 1990s graphics hw.
It is probably the oldest graphics processor we can (ab)use for xrender
operations though.

have fun
Michael