[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [lmi] std::unordered_map vs std::map
From: |
Vadim Zeitlin |
Subject: |
Re: [lmi] std::unordered_map vs std::map |
Date: |
Sat, 6 Oct 2018 23:32:26 +0200 |
On Sat, 6 Oct 2018 21:13:42 +0000 Greg Chicares <address@hidden> wrote:
GC> Interestingly, I see virtually no performance difference in
GC> $make $coefficiency unit_tests unit_test_targets=ledger_test.exe |grep
evalu
GC> make_evaluator() : 1.572e-001 s mean; 150181 us least of 7 runs
GC> if I
GC> %s/unordered_//g
GC> as in the patch below. My measurements:
GC> 150 ms std::map
GC> 148 ms std::unordered_map
For once I can't say that the benchmarking results are really surprising
as I wouldn't expect map lookup time to be a big part of overall time here.
I.e. if the difference is really 2ms (although IME this is well inside the
margin of error), it might still be pretty significant if the total time
spent in map lookup code is, say, 5ms.
Of course, as usual, only really profiling the code would tell for sure.
As you know, I plan to do it anyhow and I could check the difference
between profiling results with and without this patch too when doing it.
Regards,
VZ