|Re: Rijndael on a 42s?!?|
Message #10 Posted by Thomas Okken on 30 Aug 2007, 11:15 p.m.,
in response to message #9 by Howard Owen
It depends on how the acceleration is implemented on the real calculator, and if that/those feature(s) are emulated in emu42.
As far as I know, accelerated HP calcs simply have a modified master clock, which makes *everything* run faster, i.e. all the hardware -- not like speeding up a PC by overclocking the CPU, where only the CPU runs faster, while memory and peripherals still run at the same speed as before. So, theoretically, an HP-42S whose clock has been cranked up by a factor x from the normal 1 MHz, should do *everything* faster by a factor x. I don't know about the HP-42S' "Fast Mode", though -- that may well be different. Christoph, if you're listening, could you shed some light as to how that may work in Emu42?
aes.raw and aesprep.raw are 3357 bytes together -- that's almost half the HP-42S' total user memory. Pretty darn big by HP-42S standards, but you should still be able to do a great deal even with those things loaded.
I don't know if the AES code would fit alongside my current stuff.
It should be possible to shave off some bytes by using a matrix to hold the constants from aesprep.raw, and writing a small program to load the required registers from that; in a matrix, each number takes up 8 bytes, while in a program, it can be significantly more (and the numbers in aesprep.raw are definitely more than 8 bytes each -- count 1 byte each for each digit, minus sign, decimal point, and exponent sign, plus one extra byte for the "null" byte that is automatically inserted in front of each number).