René Scharfe [off-list ref] writes:
Am 05.04.2012 21:17, schrieb Junio C Hamano:
quoted
After seeing "I wrote it myself due to NIH", it strikes me a bit odd that
you still used "start from bunch of singleton sublist, elongating twice
per round as we go" structure from the original.
It's just becasue the dumb bottom-up approach is the most simple way to
implement merge sort.
...
This optimization speeds up the pre-sorted case but slows down the case of
a reversed pre-sorted list because we have to determine the length of the
sublists each time,...
Ah, I somehow missed that point. Thanks.