On Tue, Mar 29, 2011 at 18:52, Torsten Förtsch <[email protected]> wrote:
> Meanwhile Ævar uploaded his own quite specialized smaps parser
> (Linux::Smaps::Tiny) to CPAN. Here is what I get in comparison:
>
> perl -Ilib -MLinux::Smaps -MLinux::Smaps::Tiny \
> -MBenchmark=:all,:hireswallclock -e '
> cmpthese timethese -5, {
> tiny=>sub{Linux::Smaps::Tiny::get_smaps_summary},
> smaps=>sub{scalar Linux::Smaps->new->all}
> }
> '
> Benchmark: running smaps, tiny for at least 5 CPU seconds...
> smaps: 5.0982 wallclock secs ( 4.56 usr + 0.47 sys = 5.03 CPU) @
> 365.81/s (n=1840)
> tiny: 5.51047 wallclock secs ( 4.63 usr + 0.83 sys = 5.46 CPU) @
> 591.58/s (n=3230)
> Rate smaps tiny
> smaps 366/s -- -38%
> tiny 592/s 62% --
I've just uploaded 0.06 of Linux::Smaps::Tiny. It'll use C XS code to
parse the file by default, falling back to the pure-perl version if
there's no compiler on the system:
$ perl -Ilib -MLinux::Smaps -MLinux::Smaps::Tiny
-MLinux::Smaps::Tiny::PP -MBenchmark=:all,:hireswallclock -e '
cmpthese timethese -5, {
tiny=>sub{Linux::Smaps::Tiny::get_smaps_summary},
tiny_pp=>sub{Linux::Smaps::Tiny::PP::__get_smaps_summary},
smaps=>sub{scalar Linux::Smaps->new->all}
}
'
Benchmark: running smaps, tiny, tiny_pp for at least 5 CPU seconds...
smaps: 5.3729 wallclock secs ( 4.42 usr + 0.93 sys = 5.35
CPU) @ 792.52/s (n=4240)
tiny: 5.42358 wallclock secs ( 2.94 usr + 2.46 sys = 5.40
CPU) @ 2082.59/s (n=11246)
tiny_pp: 5.39001 wallclock secs ( 4.02 usr + 1.36 sys = 5.38
CPU) @ 1029.18/s (n=5537)
Rate smaps tiny_pp tiny
smaps 793/s -- -23% -62%
tiny_pp 1029/s 30% -- -51%
tiny 2083/s 163% 102% --
