Awesome
ICOZIP
ICO Zip is intended to be a demonstration repository, demonstrating how the ZipCPU may be placed onto an ICO board.
Particular goals of this project include:
- Generate the main project files using AutoFPGA
- Demonstrate the ZipCPU running tttt, a 4x4x4 tic-tac-toe game
A unique feature of this project is that the Raspberry Pi that will host the ICO board will only have a minor role in the project. My intention is to use the Raspberry Pi only as a TCP/IP stack the provides access to 1) a command port on the FPGA, 2) a serial port within the FPGA, and perhaps even 3) a means of downloading new bit files to the FPGA. If all goes well, then, one should only need to run the netpport program on the Raspberry Pi, and after that everything should be accessable through the TCP/IP port numbers given in port.h.
This actually has a very profound consequence. By allowing the user to have access to the internals of the FPGA, peripheral development may be done and verified prior to placing the CPU into the FPGA. Hence, by the time the CPU is placed on board, all the parts and pieces of the design will already be known to work, leaving the only new variable that of the CPU itself.
Status
This project is a work in progress, but it is now close enough that if you have an interest in it then I would welcome you to try it.
As of 20170526, the basic, uart, and pmodled examples are fully functional. While much work has taken place on the parallel port testing modules (using sw/netpport)--they just aren't reliable enough (yet) to depend upon.
UPDATE: As of 20180731 ...
-
The simulator is now fully functional
-
The CPU test runs. Steps to running the CPU test include:
-
Adjusting sw/host/port.h to contain the network name of your Raspberry Pi. This can be 'localhost' if you intend to do all your work on the Pi.
-
To build the design, you'll need to have verilator, yosys, arachne-pnr, icestorm, wiringPi and the icotools installed. (wiringPi and icotools are ARM-based packages, to be installed on the Pi.) You'll also need the ZipCPU toolchain and AutoFPGA installed and in your path
-
Once the prerequisites have been installed, type 'make' from the main directory.
-
Repeat the build in the sw/host directory from your Raspberry Pi, if this wasn't your build host.
-
Copy the arm-netpport program to the main directory of the Pi.
-
Using icoprog from the icotools package, load the design to the board.
icoprog -p < icozip.bin
You may need to usesudo
depending on how your wiringPi library is set up on your Pi. -
From the RPi, run arm-netpport. Again, you may need to use
sudo
for this, depending on your setup. -
You can then interact with the FPGA using arm-wbregs from another ARM terminal, or pc-wbregs from a PC based host connected over the network. For example,
pc-wbregs version
will return the date of the build. -
To run the CPU test, telnet into the RPi at port 8364 to get access to the CPU's console, as in
telnet rpi 8364
. -
In another window, type
pc-zipload -n rpi -r sw/board/cputest
whererpi
is the network name/address of your Raspberry Pi. Alternatively, you can typearm-zipload -r sw/board/cputest
from a Pi. -
Look for the results of the CPU test in the console window where you telnet'ed into the CPU on port 8364.
-
Building the design will also build a simulation in the sim/verilated directory. The main program file for this simulation is automaster_tb.cpp. (The main_tb.cpp file is created by AutoFPGA, and so automaster_tb.cpp is the actual top level program.) You can interact with this simulation just as you would with the CPU. The name of the simulation will either be
pc-main_tb
orarm-main_tb
depending on whether you are running from your host computer or the Pi.pc-main_tb
can take an argument-t trace.vcd
to force it to create a trace. It can also take the name of a ZipCPU program, which it will then load and automatically run. -
I haven't tested flash based programs yet, such as the 4x4x4 tic-tac-toe program mentioned above.
UPDATE: As of 20180801 ...
-
Flash based programs now run
-
The ZipCPU debugger has been ported to this architecture
-
The CPU can now run "Hello, World!" using the newlib C-library, and even starting by copying the program from flash to SRAM
-
After running cputest, the following are the additional steps necessary to run hello world:
-
The default Makefile should cause a build in sw/zlib, then sw/board
-
This will build hello
-
To run it, load the board and start arm-netpport as you did with the cputest above
-
To capture the output,
telnet rpi 8364
, or whatever the network address of your Raspberry Pi is. -
Then run
pc-zipload -r sw/board/hello
to load and run the program. -
Watch your telnet session to see the result
- You can also run this in a simulation, by running either
pc-main_tb path/to/hello
, orpc-main_tb
and then runningpc-zipload -r path/to/hello
. You may need to add a-n localhost
parameter to yourzipload
command, and you will need to telnet into your local simulation host, but otherwise running from simulation is the same as from the hardware.