Re: perfomance of clear vs swap

[Krishanu Debnath]

Paul Pluzhnikov wrote:
> krishanu.debnath@gmail.com writes:
>
> > I have a call to hash_map::clear() function which takes long time.
>
> It is somewhat likely that you are mis-interpreting the results
> that you observed.
>
> It is hard to imagine a situation in which creating a hash_map
> takes less time than clear()ing it.

I admit this is a corner case.

> Using g++4.0.0, creating a 10,000,000 entry hash_map takes 116
> seconds on my machine; clearing via either clear() or swap() with
> temporary takes 8 seconds.
>
> > Now runtime drops significantly, 10 fold less.
> > What's exactly cause this run time reduction?
>
> User error, most likely.
>
> If you are convinced that you really see what you claim to see,
> post a complete compilable test case that demonstrates that.
>
> Cheers,

I will once I get free time. :-)
You may be interested on the discussion going on comp.lang.c++.moderated.

http://groups.google.co.in/group/comp.lang.c++.moderated/browse_frm/thread/db399126443f69e7/c44bac7e9547a6f0?lnk=st&q=clear+vs+swap&rnum=1&hl=en#c44bac7e9547a6f0


Krishanu