Question

In C++, given vector<T> src, dst, both already sorted, is there a more efficient way to merge the contents of src into dst than

size_t n = dst.size();
dst.insert(dst.end(), src.begin(), src.end());
std::inplace_merge(dst.begin(), dst.begin() + n, dst.end());

? In the case I care about, T is a small (12-16 bytes, depending on ABI) POD structure, but each vector contains millions of elements, so the total amount of memory in play is tens to hundreds of megabytes.

Answer 1

I would at least try:

std::vector<T> tmp;
tmp.reserve(src.size() + dst.size()); // commenters are probably right about this
std::merge(src.begin(), src.end(), dst.begin(), dst.end(), std::back_inserter(tmp));
src.swap(tmp);

But I suspect that much depends on the nature of T, the size of src and dst, and why we need to optimize.