Re: [boost] [local] Help for the Alternatives section

Subject: Re: [boost] [local] Help for the Alternatives section
From: Thomas Klimpel (Thomas.Klimpel_at_[hidden])
Date: 2011-03-28 13:26:02

lcaminiti wrote:
> Thomas Heller-7 wrote:
> > Just did a quick test ... code is attached ...
> >
> > $ time g++  -O3 -I. -Wall -Wextra add.cpp -o add
> >
> > real	0m0.657s
> > user	0m0.583s
> > sys	0m0.067s
> >
> > $ time ./add
> > 1e+12
> >
> > real	0m35.641s
> > user	0m35.618s
> > sys	0m0.017s
> >
> > $ time g++ -O3 -I. -Wall -Wextra add_boost_phoenix.cpp -o add_phoenix
> >
> > real	0m3.385s
> > user	0m3.160s
> > sys	0m0.217s
> > thomas_at_sunshine ~/programming/local $ time ./add_phoenix
> > 1e+12
> >
> > real	0m6.648s
> > user	0m6.643s
> > sys	0m0.007s
> >
> 
> This is not a fare comparison because add.cpp contains also local
> blocks and exits while add_boost_phoenix.cpp contains only local
> functions. The comparison should be made between add_boost_local.cpp
> and add_boost_phoenix.cpp (see below). The example names got a bit
> confusing and
> I will rename add.cpp to add_funciton_block_exit.cpp to avoid similar
> misunderstandings in the future.

Sorry, I didn't get the impression that there was some misunderstanding, or that the comparison was unfair.

I took from that comparison that Boost.Phoenix takes 6 times longer to compile, but runs 6 times faster than Boost.Local.

> A quick compile and run time comparison using Cygwin. I have ran `time`
> [snip]
> Run-Time ($ time )
> ---------
> Boost.Local (add_boost_local.cpp)
>     real    =  0.547 s    user    =  0.030 s    system  =  0.030 s
> Local Classes (add_local_class.cpp)
>     real    =  0.532 s    user    =  0.030 s    system  =  0.061 s
> C++0x Lambas (add_cpp0x_lambda.cpp)
>     Not available.
> Boost.Lambda (add_boost_lambda.cpp)
>     real    =  0.438 s    user    =  0.030 s    system  =  0.046s
> Boost.Phoenix (add_boost_phoenix.cpp)
>     real    =  0.500 s    user    =  0.030 s    system  =  0.046 s
> 
> These Boost.Local compile and run time performances look decent to me
> when compared with the other approaches (including the local class functor).

Boost.Local has the worst run time of all listed approaches. Why do you think that the performance is decent? For these attached tests, the local function seems to be called only three times. For the test from Thomas Heller, 1000000 * 10000 = 10^10 times.

Thomas Heller wrote:
> Are you serious? Please write real benchmarks.

Indeed, something seems very strange here.

Regards,
Thomas