Forth TX16 or ForthX16 is an enhanced port of my older project Forth Model T for Commander X16. It is a completely functional Forth 2012 core for Commander X16 with most features from Forth Model T included. The only two limitations are related to the limited capabilities of the file system on X16 and to the lack of lowercase in C64/X16 charset (it is more nuanced, but it may not be worth effort supporting it). There are also some minor additions and improvements.
The root contains just one assembly file fthtx16.asm. It is proper 6502 assebly in ACME assembler format. Use the following command line to build the binary:
acme --cpu 6502 --outfile fthtx16.prg --format cbm fthtx16.asm
The resulting PRG file can be copied to Commander X16 file system and loaded as usual. It is thoroughly tested with the emulator, but I don't have access to the actual hardware yet, so let me know if it breaks on metal.
Note that the format is compatible with Commodore 64, but it has not been tested on that system. It might actually work, but YMMV.
A prebuilt binary can be found in binaryfolder.
A modified copy of dynamic memory support package can be found in dynamic.
A modified copy of Forth test suite is in tests - copy files from there to the file system of Commander X16 and start it with INCLUDE RUNTESTS.FTH. The current version should run all tests without errors. The runtime on the emulator is about 4 minutes.
A few examples and benchmarks are in other - BENCH.FTH and ERASTO.FTH are old benchmarking programs calculating primes, practically unchanged (BENCH had a few ENDIFs replaced by THENs). RC4TEST.FTH is a sample from the Wikipedia page, unmodified.
There are a few known issues with the current release.
- File handles are not properly disposed on error keeping files open (may not be a problem on actual hardware).
- Opening a file for write does not overwrite an existing file. This is actually the default system behavior, but it feels quite wrong now.
- Implementation of KEY is incorrect, it is supposed to asynchronously get key without echo. It does it synchronously and with echo in this version. This was also the case in Model T, but there the was an excuse that the desired functionality cannot be implemented with standard C features only. It can certainly be implemented on the hardware, but there are some oddities of X16 I still need to trace - the keyboard buffer is not where it was on C64, and the new KERNAL calls are not sufficient to implement this functionality.
Since the switch to the token model, FORGET and MARKER are broken and unsafe, even if the tests don't show anything.WORDS is incomplete, not taking the search order into account.There are a few error conditions that really should be trapped, namely: number of words in CREATE (contrained with token model, limited to 1536); total memory allocation in ALLOT; possibly token in EXECUTE (not that important); a few conditions when FIND fails may not be handled correctly.
This is work in progress. I am planning to do a few rounds of updates. First of all, the Model T was done in a few evenings, there are a number of cut corners and rather questionable implementations there, some cleanups needed. There are known issues that need to be addressed (above). I also want to get the binary smaller - it is at 11K now, but my goal is to get it to 8K and make it relocatable if possible, that would open an opportunity to easily move it to the ROM area. Finally, I should test it with C64, it might just work there. This version won't work on C64 as is as one of the stacks conflicts with C64 memory map. This can be easily fixed, just not a priority at the moment. The interpreter is working on C64 except for file I/O which does not seem to open files correctly.
The other plans include porting this to Atari computers - that would only require a limited number of changes around the file system handling and other I/O.