00:33 < hypaethralist> That came after I found out that I did something wrong
causing my virtual machine to not switch to a different
stack when it should.
00:42 < hypaethralist> I think I'll rewrite the thing to use five stacks
instead of the meager two stacks currently used.
03:00 < hypaethralist> I think I'll use cryptographically secure randomization
to randomly pick a stack to use for each cycle.
03:05 < hypaethralist> Also, limiting the stack sizes to force using multiple
stacks, useless mangling of data, a large number of
stacks and switching each instruction's function to a
preceding one would be helpful features as well, I
imagine.
03:06 < hypaethralist> I wonder how long it would take for people to get
suicidal by coding for this daedal piece of craftmanship.
03:07 < hypaethralist> Stack randomization, memory relocation and deallocation
would also be nice.
03:08 < hypaethralist> As long as it's not controllable during execution.
03:08 < p> Hm. I can think of some solutions to the problmes presented by that.
03:08 < p> ARGH! My sanity!
03:08 < p> Simply have markers in code blocks, and search for them (using only
registers).
03:09 < p> s/code/data/
03:09 < hypaethralist> What registers?
03:09 < p> ...
03:09 < p> argh
03:09 < p> It's worse than x86!
03:10 < hypaethralist> At least it will be a RISC - Randomized Instruction Set
Computer.
03:10 < p> haha
03:11 < p> Do you intend to write programs using it?
03:11 < hypaethralist> Yes.
03:11 < p> Neat. If you manage it, please send me some of them.
03:12 < hypaethralist> Sure. As you probably understand, these will be programs
of which do not produce (easily) predictable outcomes.
03:12 < p> Yes.
03:12 < hypaethralist> WTF?
03:12 < hypaethralist> I fucked up seriously.
03:13 < hypaethralist> Getting in the mood already.
03:13 < hypaethralist> Perhaps I'll even implement a high-level language and
compiler for it.
03:14 < hypaethralist> "while A is not likely to be Z potentially do Q end?"
03:14 < p> Hm. Now I think of it, having a computer that is only
semi-predictable COULD be useful.
03:14 < hypaethralist> For what?
03:15 < p> Not as useful as one that produces predictable output, but it might
run faster (for overclocking for instance).
03:15 < p> s/for//
03:15 < p> Well, you could perform verification on answers.
03:15 < p> To test if they are probably correct.
03:15 < p> And probably repeat if they are not.
03:15 < p> For things where you care about the statistical outcome, that's
often enough.
03:16 < hypaethralist> I wonder if it is possible to somehow randomize up to
the point where predictability arises.
03:17 < hypaethralist> Statistics.