[Bug libstdc++/54075] [4.7.1] unordered_map 3x slower than 4.6.2

[plasmahh at gmx dot net]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54075

--- Comment #12 from Dennis Lubert <plasmahh at gmx dot net> 2012-07-26 12:30:00 UTC ---
I can confirm that now the reserve works as I would expect it (causing no
further rehashes). However the amount of rehashes done in the testcase is still
155 (needing 4.5s), while gcc 4.6 only needs 21 (needing 1.5s).

Monitoring the bucket counts on resizes it seems that gcc 4.8 is now much more
fine grained than gcc 4.6. I am unsure if this is expected, deliberate and/or a
good idea.