Comparison of implementations of the function s20_hash (original, Petya Red, Petya Green)

petya_s20_hash.cpp

//original (from: https://github.com/alexwebr/salsa20/blob/master/salsa20.c#L59)
static void s20_hash(uint8_t seq[static 64])
{
  int i;
  uint32_t x[16];
  uint32_t z[16];

  // Create two copies of the state in little-endian format
  // First copy is hashed together
  // Second copy is added to first, word-by-word
  for (i = 0; i < 16; ++i)
    x[i] = z[i] = s20_littleendian(seq + (4 * i));

  for (i = 0; i < 10; ++i)
    s20_doubleround(z);

  for (i = 0; i < 16; ++i) {
    z[i] += x[i];
    s20_rev_littleendian(seq + (4 * i), z[i]);
  }
}

//reconstructed:

//Petya1 (red):
//@8F7Ah
static void s20_hash(uint8_t seq[static 64])
{
  int i;
  uint8_t x[32];
  uint8_t z[32];

  for (i = 0; i < 16; ++i) {
    z[i*2] = s20_littleendian(seq + (4 * i));
    x[i*2] = s20_littleendian(seq + (4 * i));
  }

  for (i = 0; i < 10; ++i)
    s20_doubleround(z);

  for (i = 0; i < 16; ++i) {
    z[i*2] += x[i*2];
    s20_rev_littleendian(seq + (4 * i), (WORD)(z + i * 2));
  }
}

//Petya2 (green):
//@9862h
static void s20_hash(uint8_t seq[static 64])
{
  int i;
  uint8_t i4;
  uint8_t x[64]
  uint8_t z[64];

  for (i = 0; i < 16; ++i) {
    i4 = i*4;

    uint16_t ax = s20_littleendian(seq + (4 * i));
    x[i4] = ax;
    x[i4+2] = (ax >> (sizeof(ax) * 8 - 1)) ? 0xFF : 0;; //sign bit extension
    
    ax = s20_littleendian(seq + (4 * i));
    z[i4] = ax;
    z[i4+2] = (ax >> (sizeof(ax) * 8 - 1)) ? 0xFF : 0; //sign bit extension
  }

  for (i = 0; i < 10; ++i)
    s20_doubleround(z);

  for (i = 0; i < 16; ++i) {
    z[i*4] += x[i*4];
    s20_rev_littleendian(seq + (4 * i), (DWORD)(z + i * 4));
  }
}