[sci.electronics] Sparc and Minix

thomas@trane.UUCP (Thomas Driemeyer) (09/20/90)

A free sample of a Sparc CPU has found its way to me, and I'd like to
build a simple (and cheap) computer around it and port Minix to it.

Currently, I am hunting for the Sparc System Developer's Guide (or
whatever it's called) because the data sheets I have are unusable.
Then, I'll put together a simple system with no MMU and no FP, at
least 4 MB of RAM and SCSI and serial interfaces. A memory-mapped
32-bit/pixel graphics board will eventually follow.

To port Minix, I'll probably get the newest gcc/gas to cross-compile.
I want the whole thing to be as simple as possible, RISC on all levels.
MMUs don't make sense because there is no affordable OS that could use
one, and FP can come later when I find a use for it. Don't know how to
build a reasonable cache yet.
I can't mass-produce boards, but when I'm done, I'll put the schematics
and the Minix patches (Minix is copyrighted) in the public domain. Minix
is fairly easy to port, I've done it before.

Is this a dumb idea? Did anyone do something like this? Does anybody
want it? Is anyone interested in helping, preferrably in the SF Bay Area?
Someone with a really good scope, maybe? :-)

Thomas Driemeyer
pacbell!trane!thomas		... don't use pyramid!trane

Disclaimer: This has nothing to do with my company, I'm doing this for fun.

adrie@philica.ica.philips.nl (Adrie Koolen) (09/24/90)

In article <872@trane.UUCP> thomas@trane.UUCP (Thomas Driemeyer) writes:
>A free sample of a Sparc CPU has found its way to me, and I'd like to
>build a simple (and cheap) computer around it and port Minix to it.

Then you're in luck, because Minix has already been ported to the
SparcStation 1!

>Currently, I am hunting for the Sparc System Developer's Guide (or
>whatever it's called) because the data sheets I have are unusable.
>Then, I'll put together a simple system with no MMU and no FP, at
>least 4 MB of RAM and SCSI and serial interfaces. A memory-mapped
>32-bit/pixel graphics board will eventually follow.

The SparcStation 1 contains an MMU, and my Minix port depends on it.
Sun's MMU is proprietary, so you can't use (or buy) it, but I suggest
you to look at the Cypress reference MMU. It's different from Sun's
MMU, but without MMU, you got a hard time to port Minix. When you
can use high speed logic devices, you could considder a simple
segmentation unit (like in the I8088).

>To port Minix, I'll probably get the newest gcc/gas to cross-compile.
>I want the whole thing to be as simple as possible, RISC on all levels.
>MMUs don't make sense because there is no affordable OS that could use
>one, and FP can come later when I find a use for it. Don't know how to
>build a reasonable cache yet.

Then think about it fast! Without a cache, you'll have to run right out
of the DRAMS. When not using (fast) page-mode, you can't read/write
more than 6 to 8 times per usec! When using (fast) page-mode, you'll
have to design some kind of small cache anyhow.

>I can't mass-produce boards, but when I'm done, I'll put the schematics
>and the Minix patches (Minix is copyrighted) in the public domain. Minix
>is fairly easy to port, I've done it before.

I didn't find it easy to port it to the SparcStation 1. Ofcourse, I had
to learn the hardware of the SS1, and it took me some time to get the
info from Sun. You design your own system, so you should know your
hardware. There are two kinds of ports: to a new processor and to an
existing processor, but a different computer. The first kind of port
is definately more difficult than the second one. Ports of the first
kind are the Atari port (Johan Stevenson), the NS32032 (???) and the
Sparc (me). The ports of the second kind are the Amiga, Macintosh and
80386 ports (and maybe the incompatible PC boards). NOTE: I don't want
to offend someone, but this is the way I see it.

>Is this a dumb idea? Did anyone do something like this? Does anybody
>want it? Is anyone interested in helping, preferrably in the SF Bay Area?
>Someone with a really good scope, maybe? :-)

I'd like to help, but I think that Aachen in West-Germany is somewhat out
of range! Any way, success with your project!

>Thomas Driemeyer
>pacbell!trane!thomas		... don't use pyramid!trane

Adrie Koolen (adrie@ica.philips.nl)
Philips Innovation Centre Aachen