The original design was rather painful to work with, and ugly, so I put this together which is a little nicer. The button, switch, chip socket, and cable socket are one ElectroCookie Snappable PCB. I like those PCBs as they have three-hole strips.
I found some decent sockets that fit the old chips (2732A and 2764 EPROMs). In retrospect, I maybe should have got a zero-insertion force sockets where you just drop the chip in and pull the lever to tighten.
With this setup, I can have one ribbon cable going from this board to the Mega. I didn’t know what kind of cable or cable-connectors I needed to fit this 36 pin setup, so I made a cable using the commonplace loose-end Arduino ribbon cable. The pins bend rather easily, but it is better than the previous setup.
I’m still using magnet wire – for which I’ve developed a love/hate relationship. The magnet wire looks nice, is easy to cut, and one roll lasts pretty much forever. But it is rather a pain because (1) you have to boil the coating off the ends of each piece of wire with solder before you use it, and (2) you have to put a blob off solder in the hole first and then reheat it to stick the magnet wire in. Regular insulated wire, that is sized correctly for the hole, is sounding very appealing at this point. But, it works.
Something fun I learned in the process: I wrote a Forth script for testing the pin connections. The output of the script comes from Mega over UART (serial), but I used vte style escape codes to keep the display fixed and updated on one part of the terminal screen.
Forth is a language that was designed to be run on bare-metal – without an underlying operating system. Some interesting quotes from Chuck Moore:
The operating system is another concept that is curious. Operating systems are dauntingly complex and totally unnecessary. It’s a brilliant thing that Bill Gates has done in selling the world on the notion of operating systems. It’s probably the greatest con game the world has ever seen.
Lisp did not address I/O. In fact, C did not address I/O and because it didn’t, it needed an operating system. Forth addressed I/O from the very beginning. I don’t believe in the most common denominator. I think that if you go to a new machine, the only reason it’s a new machine is because it’s different in some way and you want to take advantage of those differences. So, you want to be there at the input-output level so you can do that.
So, I was certainly interested in the idea of Forth running on bare-metal. What I found quickly was byok.
I was able to get it compiled fairly easily using the pre-built toolchain provided by the author. However, I had to delete two lines in the kernel directory Makefile:
The system boots up using Grub. The words display as expected.
I was able to create variables and store/pull data from them. However, a slight oddity is that the @ symbol and the ” symbol are swapped around on the keyboard, which had me confused for about 10 minutes. But I was fine after figuring that out.
The system comes with a nice block editor, for saving a program to block memory, though I think actual disk I/O is not coded yet.
And the dump word is available:
I’m definitely interested in playing around with this some more, and exploring what x86 architecture functionality is accessible with memory reads and writes.
I got this puzzle a while ago as a present. It is a wooden puzzle made up of rotating discs of numbers. The goal is to line up the discs so that all the numbers add up to 42. What makes this especially complicated is that most of the discs have gaps in them, and therefore propagate numbers up from below depending on their position.
As an exercise, I wanted to see if I could write a memory-space efficient Forth program which would solve this puzzle.
In principle, this seemed like a simple idea, as I only need to represent the discs in memory and rotate them methodically, checking along the way for the solution. The trickiest part, however, was figuring out how to represent these discs in memory, while preserving their holes, and figuring out how to overlay them properly.
In the approach I took, there is one 48 byte memory space representing the board as a composite of the discs:
create board 48 allot
The first disc, the bottom one, is easy because it has no holes. This is represented by another 48 byte array.
For the other discs, I will also use byte arrays. However, I have to represent the holes some how. The most practical choice is to use the number zero, which is not used anywhere on the actual board, and conveniently maps to the boolean false value in Forth. So, I create disc1, disc2, and so on. You see that my number of rows shrinks with each disc.
Now, how to manipulate the board? The computationally simple and space-efficient approach is to simple rotate the bytes in place. So, here is my row rotation function, followed by one that rotates a whole disc. (Please forgive some of my inconsistent parameter descriptions…)
: rot-row ( addr -- )
dup 11 + c@ swap ( c a )
11 0 u+do
dup 10 + i - c@ swap ( c c a )
dup 11 + i - ( c c a a )
rot ( c a a c )
swap c! ( c a )
: rot-disc ( addr n )
dup 12 i * +
Having all the discs, and a way to rotate each of them, eventually I would need a procedure to stack them onto the board:
: overlay ( a a u -- )
12 * 0 u+do ( a1 a2 )
dup i + c@ ( a1 a2 c )
dup 0<> if
2 pick ( a1 a2 c a1 )
i + c! ( a1 a2 )
board disc0 4 overlay
board disc1 4 overlay
board 12 + disc2 3 overlay
board 24 + disc3 2 overlay
board 36 + disc4 1 overlay
We are pretty close now. I need a function to check for a solution, which is simple addition of the columns on the board:
: solved? ( -- bool )
12 0 u+do
board i + c@
board 12 i + + c@
board 24 i + + c@
board 36 i + + c@
+ + +
42 <> if drop false leave then
Now, I’ve got to walk through rotating all the discs, checking for a solution in each case. I chose to do this with five nested procedures, which each handle their disc with the appropriate minor variations. (A nested loop would have worked also.)
: solve4 ( -- bool )
12 0 u+do
solved? if drop true leave else disc4 1 rot-disc then
: solve3 ( -- bool )
12 0 u+do
solve4 if drop true leave else disc3 2 rot-disc then
: solve2 ( -- bool )
12 0 u+do
solve3 if drop true leave else disc2 3 rot-disc then
: solve1 ( -- bool )
12 0 u+do
solve2 if drop true leave else disc1 4 rot-disc then
: solve ( -- bool )
12 0 u+do
solve1 if drop true leave else disc0 4 rot-disc then
Now, I just need to run the solve procedure. It should return -1 (true) and then I can print the board with the command board 48 dump, which prints the board memory.
Unfortunately, I did this, and after a few seconds, the program instead return 0 (false) meaning there is no solution. Naturally, I expected there was some fault in my coding, and I dived into debugging. After an hour of carefully checking code, and inserting debugging code here and there, I was still getting the same result, and getting discouraged.
At one point, I inserted some code that would, at least, allow me to see the closest solution. I found one solution that had all columns adding up to 42, and one column adding up to 47. I did a quick Internet search, and found this revelation on the Project Genius website!
Yes, indeed, I happened to own one of the defective early models, which had a misprint. I edited my disc0 array to read 3 instead of 8 in that location (already corrected above). The edited program found the solution quickly.
One must translate between hexidecimal to decimal, and then map the four arrays of twelve bytes onto the board, which gives you the solution (SPOILER WARNING!)
The Forth code above is provided under the GPLv3+ license:
Copyright 2020 Christopher Howard
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program. If not, see <https://www.gnu.org/licenses/>.
Arduino-FVM comes with only a few Arduino-function words, basically just a few for working with the digital pins. So the test for Forth was going to be: how difficult is it to access other microchip functionality using memory reads and writes? An encouraging first step was to implement SPI TX, using the information in the 328 datasheet.
First I needed some constants for the register addresses and bit numbers:
Next I needed to set up the data direction bits for the MOSI, SCK, and SS pins, as well as set some bits in the SPCR register to control mode and communication frequency:
: setup-master-spi ( -- )
1 DD_MOSI lshift
1 DD_SCK lshift
1 DD_SS lshift
1 SPE lshift
1 MSTR lshift
1 SPR0 lshift
And here is a function for transmitting a single byte, after dropping the byte on the stack:
: tx-master ( ch -- )
SPSR c@ 1 SPIF lshift and
And here is a demo procedure for sending byte 0xA6 repeatedly:
Now we can see that byte on the oscilloscope, from MOSI pin 11:
Byte 0xA6 equals 10100110, which you can understand from the signal image if you know that (1) signal low represents “1” and high represents “0”; (2) the X-scale is 2 microseconds, with one bit per microsecond; and (3) the first bit is the one microsecond of low signal just to the left of the Y axis.
Edit after more research: the approach below might not be best possible approach when working with GPIO pins, as it might not be taking advantage of special opcodes for this purpose. Also, due to the two byte FVM cell size, I think there might be some potential for strange effectsin nearby registers of a different type (like, the PINC register being after the PORTB register). Heading back to the data sheets…
Edit 2: It looks like FORTH (including Arduino-FVM) has the C@ and C! commands which are single-byte (char) versions of @ and !.
The Atmel specific keywords that come with Arduino-FVM are rather limited – basically just PINMODE, DIGITALREAD, and DIGITALWRITE. So, to do much of interest you would want to start playing around with the atmel microprocessor registers. This is mostly straightforward, however, as the registers are accessible simply as locations in ram, meaning you just write or read bits to ram to get what you want. Doing this in FORTH, you need a few tools
@ word: read from a memory location
! word: write to a memory location
HEX word: switch to hexidecimal interpretation of the stack input/output (DECIMAL to switch back).
Pinout for your arduino board (available on the internet)
Data sheet for your specific microcontroller (available on the internet)
Here is a quick example of lighting up the LED through a register write. First, we need to know our pin number and also which register we are dealing with, available from the pinout.
There we have pin number 13 (decimal) for the LED, and register PB5. First, to keep this post a bit shorter, we we simply use the PINMODE keyword to set pin 13 to output mode:
true 13 pinmode
Now we look at the datasheet to get the memory location for register PB5.
So, PORTB register is at 0x25 (hex) and bit five is what we want. Now, let’s see what is already in that memory location. On my chip, I got:
0x25 @ . 0
Just keep in mind however, as you can see using the CELL command, that you have actually read in two bytes. So, if you get something like 0x2000, you are only interested in the first byte (the right-most 00).
As a general practice, I only want to write to the bit of interest, preserving other bit states. So, I drop a 1 for that bit on the stack, and use an OR to light up just that bit. A 1 in bit 5 is hex 0x20.
0x25 @ 0x20 or 0x25 !
And the LED lights up. We can see now what we have in register memory:
0x25 @ . 20
That is a trivial example, but I think you would need to go into this memory “peeking” and “poking” in order to do more advanced things like control the interrupt system or special purpose pins. At least, I find the idea more appealing than having to write C interface functions on the backend. Of course, you would want to hide the details inside nice forth functions.