Quote from: x^2 on September 03, 2013, 11:40:51 PM

Quote from: ZirconiumX on September 02, 2013, 06:22:54 PM

One thing I've realised. All of the AES candidates are based around encryption from a key. If we go for a quark-style hash system, then if we are being secure, we need 14 different keys.

Something for donators to suggest?

Matthew:out

Should I start a donation fund?

If you want to. I'm not forcing you to.

Quote from: smeagol on September 03, 2013, 09:11:31 PM

Quote from: ZirconiumX on September 02, 2013, 06:22:54 PM

One thing I've realised. All of the AES candidates are based around encryption from a key. If we go for a quark-style hash system, then if we are being secure, we need 14 different keys.

Something for donators to suggest?

Matthew:out

Are we doing the random hashing algos too? (like first sha, then whirlpool, then md5, then random, then random again)

As far as I can work out, each pass of Quark's "random hashing" simply checks whether the least significant 3 bits in the second most significant 32-bit word are zero. If they are, then it runs one hashing algo, else, it runs a different one. It screws PGO massively.

I like your idea, but I think we can make a different scheme which makes optimisation even more awkward.

Some pseudocode:

Code:

void Hash(int* input, int* output)
{
   imt mask = 8;
   int i = 0;

   for (; i < 8; i++) {
   switch (input[5] & 8) {
   case 0:
   Rijndael(input, input);
   case 1:
   Square(input, input);
   case 2:
   Serpent(input, input);
   break;
   case 3:
   RC6(input, input);
   case 4:
   MARS(input, input);
   case 5:
   Twofish(input, input);
   break;
   case 6:
   Twofish(input, input);
   case 7:
   Rotate256(input, input, 3);
   }

   memcpy(...);
}

That ought to make GPU mining difficult. With up to 24 rounds of pseudorandom hashing algos, and a rotate in there for good measure, it may well make CPU mining difficult, too.

Matthew:out