This is not a full review, but rather a few comments on the implementation of reorder(x) functions. What you are doing is casting pointers around and assigning individual bytes, which might not really effective: inline void reorder(int64_t source, int64_t& target) { const char* s (reinterpret_cast<const char*>(&source)); char * t (reinterpret_cast<char*>(&target) + sizeof(target) - 1); *t = *s; *--t = *++s; *--t = *++s; *--t = *++s; *--t = *++s; *--t = *++s; *--t = *++s; *--t = *++s; } it does eight increments, eight decrements, one addition, one substraction and eight assignments. maybe something like: inline void reorder(int64_t source, int64_t& target) { target = ((source << 0x38) & 0xFF00000000000000) | ((source << 0x28) & 0x00FF000000000000) | ((source << 0x18) & 0x0000FF0000000000) | ((source << 0x08) & 0x000000FF00000000) | ((source >> 0x08) & 0x00000000FF000000) | ((source >> 0x18) & 0x0000000000FF0000) | ((source >> 0x28) & 0x000000000000FF00) | ((source >> 0x38) & 0x00000000000000FF); } would be more efficient?