Re: [boost] serialization performance

From: Felipe Magno de Almeida (felipe.m.almeida_at_[hidden])
Date: 2005-11-12 13:31:25

On 11/12/05, Robert Ramey <ramey_at_[hidden]> wrote:
[snip - Martin Slater wrote]

> A couple of observations.
>
> I believe the intel machine has hardware instructions which implement
> strcmp and that compilers support them.  So even if strcmp is the
> bottleneck, I wouldn't expect it to show up on the profiler unless
> some sort of inlining were turned off.  Or maybe the vtune profiler
> has special provision for these cases somewhere.

If you check for extended_type_info equality using strcmp very much, I
cant see why it wouldnt be a bottleneck. I dont know the details of
the serialization library, but my guess is that it must find the
extended_type_info which matches the type given, depending on the
complexity of your find algorithm, it may have unnecessary
comparitions, that would point strcmp as the bottleneck.

>
> I did check to verify that the strcmp in the type-id lookup has been
> removed.  Instead we just make sure there is only one instance
> of a particular extended_type_info record so that we can
> just compare the addresses.  There are still some optimizations
> to be implemented - but I can't predict how much they will
> speed up anything.

IMHO, profilers are essential before any work on optimization.
Probably lots of optimizations wont even be needed, while others will
make the real difference in speed.
I have experienced that reusing containers(as someone already posted
in this mailing list as a possible solution to some bottlenecks in the
serialization library) to have huge impact in performance.

>
> Robert Ramey

best regards,
--
   Felipe Magno de Almeida
Developer from synergy and Computer Science student from State
University of Campinas(UNICAMP).
Unicamp: http://www.ic.unicamp.br
Synergy: http://www.synergy.com.br
"There is no dark side of the moon really. Matter of fact it's all dark."