[PATCH] PR libstdc++/81476 Optimise vector insertion from input iterators

[Jonathan Wakely <jwakely at redhat dot com>]

The PR shows a fairly pathological case where ranges of InputIterators
are repeatedly inserted at the start of a vector. Each insertion from
an InputIterator moves every element after the insertion point by a
single position. So if we insert 1000 elements at the start we move
each existing element 1000 times to reach its final position.

This patch changes insertions that aren't at the end to use a
temporary vector, which starts empty so insertions are at the end, and
so fast. Then we do a range insert to copy everything from that vector
at once. That uses random access iterators, so we know how far to move
the existing elements and can just move them to their final position
in one step.

There are other possible strategies, such as inserting the input range
at the end of *this and then rotating them into place once we have
inserted them all. However, when the insertions trigger a reallocation
we need to transfer all the existing elements (not only those past the
insertion point). The patch I've committed only has to move the
existing elements if the final insertion from the temporary vector to
*this causes a reallocation.

	PR libstdc++/81476
	* include/bits/vector.tcc (vector::_M_range_insert<_InputIterator>):
	Only insert elements one-by-one when inserting at the end.
	* testsuite/performance/23_containers/insert/81476.cc: New.

Tested x86_64-linux, committed to trunk.

Attachment: patch.txt
Description: Text document