SpeckでROUNDSが外か中か

speck-rounds-inside.c

/*
こちらのコードは ROUNDS がループの内側。
https://en.wikipedia.org/wiki/Speck_(cipher) とほぼ同じ。
自動ベクトル化はできないみたいです。
 */

#define ROR(x, r) ((x >> r) | (x << (64 - r)))
#define ROL(x, r) ((x << r) | (x >> (64 - r)))
#define R(x, y, k) (x = ROR(x, 8), x += y, x ^= k, y = ROL(y, 3), y ^= x)
#define ROUNDS 32

void encrypt_stream(const uint8_t* data_in, size_t const len, uint8_t* data_out, uint64_t const key[static 2]) {
    const uint64_t* data_in_64 = (const uint64_t *) data_in;
    uint64_t* data_out_64 = (uint64_t *) data_out;

    for (int i = 0; i < len / 8; i += 2) {
        uint64_t y = data_in_64[i];
        uint64_t x = data_in_64[i + 1];
        uint64_t b = key[0];
        uint64_t a = key[1];

        R(x, y, b);
        for (int j = 0; j < ROUNDS - 1; j++) {
            R(a, b, j);
            R(x, y, b);
        }

        data_out_64[i] = y;
        data_out_64[i + 1] = x;
    }
}

speck-rounds-outside.c

/*
こちらのコードは gcc -O3 -fopt-info -o speck.exe main.c すると
main.c:38:9: note: loop turned into non-loop; it never loops.
main.c:38:9: note: loop with 2 iterations completely unrolled
main.c:38:9: note: loop vectorized
のように出力されて自動ベクトル化されます。下記の21行目の所で自動ベクトル化されます。
 */

#define ROR(x, r) ((x >> r) | (x << (64 - r)))
#define ROL(x, r) ((x << r) | (x >> (64 - r)))
#define R(x, y, k) (x = ROR(x, 8), x += y, x ^= k, y = ROL(y, 3), y ^= x)
#define ROUNDS 32

void encrypt_stream(const uint8_t* data_in, size_t const len, uint8_t* data_out, uint64_t const key[static 2]) {
    memcpy(data_out, data_in, len);
    uint64_t* data_out_64 = (uint64_t *) data_out;

    uint64_t b = key[0];
    uint64_t a = key[1];
    for (int j = 0; j < ROUNDS; j++) {
        for (int i = 0; i < len / 8; i += 2) {
            uint64_t y = data_out_64[i];
            uint64_t x = data_out_64[i + 1];

            R(x, y, b);

            data_out_64[i] = y;
            data_out_64[i + 1] = x;
        }
        R(a, b, j);
    }
}