TalkChess.com

[nczempin]

I've received my kit, and have started to put it together. I didn't have an RS-232 cable handy, so I had to wait until tonight to start actually doing some stuff with it.
Because I don't have a TV in the hotel room I'm staying in, I can't test the video/audio output until the weekend.
I will play around blinking a few LEDs, and perhaps connect the 7-segment display that comes with the "extended" Pico (2.0).

In the meantime, I thought a bit about techniques to save memory ruthlessly. I don't care much if I lose speed by some of these techniques; I can always get back a bit of speed if it turns out I'm not even using all those 136 bytes of RAM.

So, here are some ideas:

1. If I disallow sub-promotion and only allow a maximum of 5 queens, I can pack a move into 1 byte instead of two:
1 queen, 28 moves max: 28
2 rooks, 14 moves max: 28
2 bishops, 14 moves max: 28
2 knights, 8 moves max: 16
1 king, 10 moves max: 10
8 pawns, 4 moves max: 32

So I have 142 possible codes using just one queen, and 28 more per additional queen, for five queens I'd need 254 states. Just enough for 1 byte.

Sub-promotion would be much costlier, and occurs virtually never in real games. So I figure it's an adequate tradeoff.

2. For the state of the board I need 4 bits per square * 64 squares = 32 bytes. Using an "indexed" scheme similar to the move packing above I could get it down to 28 (although I don't remember how I did this right now). Plus 3 bits for en passant (one of eight) and 4 bits for castling rights, so make it a byte; we'll find a use for that extra bit

If there are any flaws in my calculations, please let me know...