[rescue] [OT] S: IBM /370 or /390 card

[Romain Dolbeau]

On 9/17/23 15:58, Mouse via rescue wrote:
> That's encouraging!  I'll have to find a work machine I can clone the
> repo from and have a look.

Be warned, both the PCB and a lot of the HW design are the work of a 
newbie - I started the project knowing next to nothing about hardware 
design during Covid lockdown. Had to occupy myself, wanted to do my own 
SBus for years... My newer designs should be a bit less terrible.

Also the hardware design makes extensive use of Litex 
(<https://github.com/enjoy-digital/litex/>), and so is written in a 
dialect of Migen - a python-based hardware definition language.

Principle of operation is easy enough - the SBus is bridged to the 
internal Wishbone bus via a Finite State Machine, and most of the 
devices live on the Wishbone and are built mostly from Litex components. 
There's some optimizations to bypass the Wishbone bottlenecks that were 
added on the go (e.g. the DMA engine for the RAM disk, 240 MiB of really 
fast swap!)

> Personally, I would pay extra to _not_ have HDMI.

It's my Highly Desirable Macintosh Interface, who just so happen to be 
compatible with those other 'hdmi' display devices ;-)

VGA is unfortunately not a 'viable' option due to the number of I/O 
required from the FPGA. Also analog signaling is a pain. HDMI can make 
do with 4 differential pairs and some external voltage protection (the 
FPGA can output the required TMDS signaling out-of-the-box).

> Not that I'm thrilled by the FPGA world either, but that's a lower
> level of distaste (it's "just" closed source, not actively crippling
> what end users can do with devices they own).

The Xilinx Artix-7 I use is pretty nice, but indeed will require the 
proprietary Xilinx toolchain. Some FPGAs now have open-source toolchains 
based on reverse engineering (e.g. the ECP5).

If you're interested in the subject of trustworthiness/openness of the 
hardware, look at e.g. <https://www.contrib.andrew.cmu.edu/~somlo/BTCP/>.

There's currently a lot of missing pieces, but with the opening of some 
PDKs from the likes of Skywater or GloFo 
(<https://github.com/google/skywater-pdk>, 
<https://github.com/google/gf180mcu-pdk>) and initiative like OpenRoad 
(<https://theopenroadproject.org/>), perhaps someday open hardware will 
be able to go for dedicated ASICs... for now it's still a bit complex 
and expensive for a hobbyist though :-)

Cordially,

-- 
Romain Dolbeau