I've been messing around with the emulator a bit, playing with programs that I used to type in for fun all the time back in the day.
The first computer I ever used on a regular basis was the TRS-80 Model III down at the Radio Shack in a nearby shopping center. And it had some pseudographics commands. SET, and RESET. They're very similar to the PSET command.
The program that I am thinking about at the moment was the one for filling the screen with random pixels until the whole screen was full. It took a while, but eventually, every pixel would be filled in, and at that point I'd hit BREAK and move on to the next activity.
But while the TRS-80 BASIC would eventually fill the screen, X16 BASIC won't. Using exactly the same 128×48 grid that the old Z80 system used, trying to use the RND() function to fill the screen results in a few pixels always being left over.
I know that the available routines have a fairly short periodicity, so I tried to enhance the randomness by recursive randomizing. But still, even using RND(RND(1)*32768)*[number of pixels] for both X and Y, this is the best it will do for a 128×48 grid:
I ran that at top speed for an hour, and that's as far as it got. Those white pixels never get filled in. If I try to fill the entire SCREEN $80 grid of 320×240, it's much worse.
I would use RND(0) as the inner seed, but it's my understanding that it's even more repetitive, at least, going by the C64 Wiki's article on the function.
While I don't know much about the TRS-80 version of the function, and I know it's still a pseudo-random number generator, it seems to be a little better at covering a wider range of values more comprehensively.
Is this just something we're stuck with, and will just have to code our own, or is there a way to shoehorn in a better PRNG algorithm into the stock interpreter? Or is there a workaround to get better results out of the existing RND() code?
