Hacker News Reader

Comment by oconnor663 | Hacker News Reader

The computer this runs on is Ben Eater's 6502 breadboard kit, which I highly recommend to anyone with the slightest interest in wires and NAND gates: https://eater.net/6502. After the hello world program was working, I wanted to try something more substantial, so here's the world's slowest implementation of BLAKE3 :) If you have Rust and Cargo installed, you can run `make emulate` to test out the program in your terminal.

View on HN

Ha, this is neat, I haven't seen the permutation table presented this way before, good stuff.

    permute:
    ; Rather than permuting the 32-bit words stored in the
    ; COMPRESS_MSG array, we permute the 8-bit pointers stored
    ; in the MPTRS array.
    ;
    ; The permutation is defined to be:
    ; 2, 6, 3, 10, 7, 0, 4, 13, 1, 11, 12, 5, 9, 14, 15, 8
    ;
    ; If we walk those assignments, we get two loops:
    ; 0<-2<-3<-10<-12<-9<-11<-5<-0
    ; 1<-6<-4<-7<-13<-14<-15<-8<-1

I'm pleased with that part, but I'm ashamed of my ror12 function :p I'd love it if anyone could suggest a better way to do it.

http://6502.org/source/general/SWN.html swaps the nybbles in A in 8 bytes/12 cycles. That’s not quite what you need, but may give you ideas.

Oh that looks perfect. I'll have to try it out today and report back.

If you want to be on the far end of the speed/space trade off you could simply use a lookup table. That should speed things up considerably. You'd index it by byte (split the tables, save one shift left) and then recombine the results.

Great work! You might like https://github.com/janroesner/sixty5o2 as a lightweight OS for the BE6502, which makes program changes way easier.

Reply on news.ycombinator.com