Works badly in little endian systems unless unaligned access is allowed

[fasf]

Precisely, it does nothing then.

The reason is:

#if BYTE_ORDER == LITTLE_ENDIAN && UNALIGNED_WORD_ACCESS
    {
        uint64_t *data64 = (uint64_t *)data;
        while (data64 != (uint64_t *) end) {
        m = *data64++;
        SIP_2_ROUND(m, v0, v1, v2, v3);
        }
    }
#elif BYTE_ORDER == BIG_ENDIAN
    for (; data != end; data += sizeof(uint64_t)) {
    m = U8TO64_LE(data);
    SIP_2_ROUND(m, v0, v1, v2, v3);
    }
#endif

Which should be:

#if BYTE_ORDER == LITTLE_ENDIAN && UNALIGNED_WORD_ACCESS
    {
        uint64_t *data64 = (uint64_t *)data;
        while (data64 != (uint64_t *) end) {
        m = *data64++;
        SIP_2_ROUND(m, v0, v1, v2, v3);
        }
    }
#else
    for (; data != end; data += sizeof(uint64_t)) {
    m = U8TO64_LE(data);
    SIP_2_ROUND(m, v0, v1, v2, v3);
    }
#endif