I don't see how these assumptions really help - 'greatest program ever written' is a bit bombastic but working chess program in a few hundred bytes of z80 code is still quite impressive.
Reading keyboard is reading a register/known address, not loading the keyboard library, allocating a structure and jumping into an OS routine
Or calling a ROM routine. Reading the keyboard directly at the lowest level would probably just make the program bigger
Clear screen is writing zeros to a known address space
Generally, another ROM routine call.
Drawing is writing a 1 to a known address space
Not really. Just about all these machines had color. The Apple ][ had a fun non-linear video buffer layout, to boot.
Reading the keyboard directly at the lowest level would probably just make the program bigger
Generally, another ROM routine call.
Thats @tezza's point. (Edit For Clarity): Since its just one instruction, you can shrink the size of your program, making fitting it into a small space less impressive (though, not unimpressive).
Not a lot of text in a chess program.
Fair enough, so that one doesn't help, but the other two do.
For reference to my original post: I wasn't trying to deny that this program is amazing... it is.
I was intending to illustrate that you could do a lot more with 1k historically than you can nowadays, and that some of the worst offenders to bulking out a process don't apply on older computers in assembly.
That doesn't take away from the brilliance of AI in 1k, but rather I think the focus should be on that and not the chessy bit.
Except it's not one instruction - you still have to put the result somewhere, potentially save registers (of which there is a great dearth) and so forth. My point is none of the things listed (aside from not really being part of 'the architecture') are significant factors on top of being largely inaccurate. What does make a difference is that the instructions themselves are short, the addresses are short, alignment is on the byte, etc. Can you fit a lot into 1k? Sure, if you're clever, you can apparently fit a whole chess program in less. But the fact that, say, there's just your process or that Huffman encoding is neat has next to nothing to do with it.
Heh, well, I'm not clear what is unclear. The poster said 'well, the program is short because of these N things'. Except none of the N things were accurate or relevant. There are architectural factors that much more significantly affect the size of the program, but poster didn't mention any of them. My point is tezza either didn't know or didn't think what s/he was posting about. It's sort of like having the discussion "So, airplanes fly because of magic" - "Well, no actually airplanes fly because of [a bunch of physics]" - "Right, you're making my point, airplanes fly!".
Okay, I feel like there is a real break in communication and I'm curious to explore it. If it gets too tedious for you, feel free to ignore me.
Going back to the N = 5 things. Three seem to help, two don't.
The helpful ones:
* Reading keyboard is reading a register/known address, not loading the keyboard library, allocating a structure and jumping into an OS routine. *
Doens't this make make the program smaller? Getting the key entered is just a read instruction. Sure, there maybe things that happen afterwards (as you said) but just getting the key is one instruction i.e. read <some memory spot>
* Clear screen is writing zeros to a known address space *
Again, instead of doing a number of instructions to clear the screen (I have no idea how many this takes on a modern computer) you just need to write do write 0 <memory location>
* Drawing is writing a 1 to a known address space *
My understanding is that on modern computers this is a complex operation. But, in this it appears that drawing is again a single operation.
In each case, the point @tezza is making is: Hey, all you people who haven't ever used these crazy old machines, these are the types of things that make it easier to create a very small chess program (in terms of size of the program itself, not the resources it uses). All these things that nowadays require a lot of instructions only required a single instruction on this machine.
The other two points, I agree, are probably not helpful.
Using the keyboard reading as an example, it seems whoever designed this machine said "one way to reduce the size of programs is to make it easy to read keyboard inputs. Lets have the hardeware store that in a known address." Thus, the fact that you can do this does help reduce the size of the program.
Am I wrong? Let me know. I've never worked on these machines so I could be completely off base. A friend of mine used to wonder if it would be instructive to learn about programming under these limitations so I'm genuinly curious.
Note: I'm not objecting to your points about what reduced the size of the program. I'm just trying to figure out why you think @tezza's points don't help.
Dude, you're going over the top, under the bottom and around the sides.
8bit coding was very different. I listed some common tricks not necessarily specifically ZX81, nor necessarily used in this chess game, but illustrative of common scenarios.
Thus you could fit much more in 1k than you can today.
Please don't 'dude' me. My breath is perfectly even. And at this point I don't think I was even replying to you. The things you said make make a program short are simply inaccurate and not the significant factor in making programs shorter. That's all. Nobody made their programs shorter by the 'tricks' you describe. And it's not a ZX81 thing. To read the keyboard on, say, a 6502 apple ][ you have to do a store to clear the strobe, a read to get a value and a check on the most significant bit of the value to check for keypress and then you need to mask out the actual value. I don't need to go into endless detail about the rest of them. The primary factor is the size of the instructions and the size of their operands (8 and 16 bit addresses).
Some of the 'results' that need storing can be stored on the screen itself. An example of a difference between these 8bit machines and today's machines that @tezza is talking about.
I had a ZX81. Drawing was indeed done by writing directly to a special address space. However, you didn't write pixels directly but rather bytes which where rendered as characters on the screen by the system.
In fact, if you look at the source code in the actual article, you can see the graphical characters used to draw the board right in the Basic code listing. :)
You're linking to information on a ZX Spectrum which is not a ZX81. The ZX81 had a monochrome character mapped display with no pixel access (well - there were tricks that could sort of give you pixel access but they came much later than the time 1k chess was written)
Sorry about that... i didn't mention colour depth originally, someone else brought it up (trying to look smart).
6502 was my platform, so I missed the exact Z* platform...
I was sharing general info on assembly for 8bit, not specific to one platform (ZX81). This seems to have incensed some people who took a narrow view on what sort of comment was permitted.
IMHO this 128 Byte(!) 3d demo is even more impressive: http://pouet.net/prod.php?which=53871 (Click on the Youtube link at the right to see it running)
the 1k is aided and abetted by the architecture of the computer.
* Reading keyboard is reading a register/known address, not loading the keyboard library, allocating a structure and jumping into an OS routine.
* Clear screen is writing zeros to a known address space
* Drawing is writing a 1 to a known address space
* Yours is the only process on the machine
* Text compresses well with Huffman
With those assumptions a lot more can fit in the 1K