The TRS8BIT 2022 contest is a wordle implementation for any of the TRS-80 machines. There are two tracks they call "Beauty and the Beast". I went for Beauty – there's no way my T102 can fit 12,960 5-letter words, 64,800 bytes, in my 32 KB T102. Even compacted into 5-bit "letters", that's 40 KB...
But I did manage to compact things a little. The first set of words is a list of 2,313 words. I cut it down to 2,312, to fit into 289 blocks of 25 bytes, encoding 40 bytes each, ie 8 words. At the moment, the code and the dictionary take up 8,584 bytes only. There's a bug in an edge case, but even after fixing that, I should be under 9 KB...
I split the code and the dictionary, so that I could upload the dictionary once – it takes long enough...– and the code as often as needed. To do this, I created my own development environment from almost scratch. I took a simple 8085 assembler that had been long abandoned, and modernized it, adding compile options, the undocumented instructions, etc. It produces various files, the most important of which is the the binary, a .co file. This is uploaded to the T102 via a combination of a BASIC program on the T102 and a custom GUI application I made with Xojo, LOAD_CO. It works pretty well so far, and I am quite happy with my home-grown toolchain... Among other things this toolchain enables the compilation and sideloading of apps at customized addresses in RAM, without having to calculate them: you give the compiler the HIMEM address (ie the lowest available address), and it calculates the ORG address, and the new HIMEM. This can then be used with LOAD_CO.
