Re: [Gnash-dev] Re: about perfomance

[Yuriy Zhuravlev]

 -O2 optimization std::* tools very fast(templates...). Need test on Linux and curent gcc version(gcc-4.1.1)

2006/7/25, Sandro Santilli <address@hidden >:

Guys, Vitaly did some performance tests for tulrich containers
vs. standard ones. I'm happy to see that for most cases the
standard containers are faster, so we can safely switch to
std::string, for example, instead of tu_string (we'd need
a variation for tu_stringi - case insensitive comparison).

The only case in which tulrich's seems better is std::vector
vs. tulrich's array. We decided with Vitaly to check uses
case by case and optimize use of std::vector and also make
sure that's the correct container to use.

In any case, only using standard containers seems a good
solution to me.

Profiling results follow. Comments welcome.

--strk;


On Mon, Jul 24, 2006 at 02:34:34PM +0400, Vitaly Alexeev wrote:
> Hi,
>   Appendix contains some piece of program code that shows difference between "std::" and Thatcher's classes.On Win32 it have shown the next results (ticks)
>
>   std::vector - 2899
> tulrich array - 112
> std::string - 882
> tu_string - 3218
> std:hash - 213
> tulrich hash - 971
>
>   We can see,that "std::vector" must be changed to "tulrich array","tu_string" must be changed to "std::string" and "tulrich hash" must be changed to "std::hash".What do you think about it?
>
> Vitaly.


_______________________________________________
Gnash-dev mailing list
address@hidden
http://lists.gnu.org/mailman/listinfo/gnash-dev