Hi matt!
On Wednesday 02 Dec 2009 18:29:53 matt wrote:
> On Dec 1, 8:58 pm, practicalp...@gmail.com wrote:
> > Hello,
> >
> > Maybe it's not so suitable to ask this here.
> > But is there a good way (code sample?) to implement a speed test
> > between Perl and C?
> > For a project which handles lots of data we want to know how slower
> > perl is than C.
> >
> > Thanks.
>
> To perform the test, I'd just use 'time':
>
> test.cc:
> int main()
> {
> int a = 1 + 1;
> }
>
> test.pl:
> #!/sw/bin/perl
> $a = 1 + 1;
>
Just a few notes:
In both of these cases, the "1+1" expression should be evaluated at compile-
time by the compilers, because it contains nothing but constants. Furthermore,
g++ may be smart enough to see that you're not doing anything with "a" and to
do nothing in this case.
A more representative example would be one that inputs two integrarl arguments
(from argv/@ARGV or STDIN) and adds them in run-time.
If you're looking for something a bit more larger in scope, see:
http://github.com/shlomif/Project-Euler-Problem-10-Comparison
I've written similar programs in C, Haskell, and Perl to compare Problem #10
in Project Euler. The results were:
<<<<<<<<<<<<<<<
# This was ran on:
# * A Pentium 4 2.4 GHz machine with 2.5 GB of RAM
# * Running Mandriva Linux Cooker (as of 28-November-2009)
# * On top of XFCE from Cooker and very few things running in the background.
Benchmark: timing 500 iterations of c_mine, c_mine_half, c_mine_micro_opt...
c_mine: 15 wallclock secs ( 0.02 usr 0.13 sys + 12.20 cusr 2.37 csys =
14.72 CPU) @ 33.97/s (n=500)
c_mine_half: 10 wallclock secs ( 0.02 usr 0.13 sys + 6.75 cusr 2.30 csys =
9.20 CPU) @ 54.35/s (n=500)
c_mine_micro_opt: 15 wallclock secs ( 0.02 usr 0.13 sys + 11.34 cusr 2.36 csys
= 13.85 CPU) @ 36.10/s (n=500)
Benchmark: timing 20 iterations of haskell_mine, haskell_zeroth, perl-10,
perl-10_half...
haskell_mine: 135 wallclock secs ( 0.00 usr 0.00 sys + 134.71 cusr 0.70 csys =
135.41 CPU) @ 0.15/s (n=20)
haskell_zeroth: 310 wallclock secs ( 0.00 usr 0.01 sys + 307.61 cusr 1.15 csys
= 308.77 CPU) @ 0.06/s (n=20)
perl-10: 95 wallclock secs ( 0.00 usr 0.01 sys + 94.14 cusr 0.30 csys =
94.45 CPU) @ 0.21/s (n=20)
perl-10_half: 73 wallclock secs ( 0.00 usr 0.00 sys + 71.34 cusr 0.36 csys =
71.70 CPU) @ 0.28/s (n=20)
>>>>>>>>>>>>>>>
So the Perl Benchmark.pm module (with some overhead) executed the fastest C
program 54.35 times per second, and the fastest Perl program at 0.28 times per
second - roughly 194.1 times faster. The Haskell program was even worse at
0.15 times per second. I should note that the Haskell program used a different
algorithm, but that both algorithm are functionally equivalent, and that the
algorithm I used for C and Perl used an array / vector of bits and as a result
would likely perform much poorly in Haskell.
> # time bash -c 'for x in {1..100}; do ./a.out; done'
>
> real 0m0.167s
> user 0m0.029s
> sys 0m0.138s
>
> # time bash -c 'for x in {1..100}; do ./test2.pl; done'
>
> real 0m0.435s
> user 0m0.154s
> sys 0m0.247s
>
> So...I've proved that in my specific environment, C is (~3x) faster
> than Perl at adding 1+1...Now as far as what tests you want to
> implement, that's up to you and your specific needs.
>
One should note that there's also the overhead of the bash loop here.
Regards,
Shlomi Fish
--
-----------------------------------------------------------------
Shlomi Fish http://www.shlomifish.org/
Optimising Code for Speed - http://shlom.in/optimise
Chuck Norris read the entire English Wikipedia in 24 hours. Twice.
--
To unsubscribe, e-mail: beginners-unsubscr...@perl.org
For additional commands, e-mail: beginners-h...@perl.org
http://learn.perl.org/
