Intro Results Summary End

Perl::Formance

Ste�en "renormalist" Schwigon

10 Years Dresden Perl Mongers

21 August 2012

Intro Results Summary End

Intro

Results

Summary

End

Intro Results Summary End

Intro

"We suck at benchmarketing."

Intro Results Summary End

Intro

There is hope.

Intro Results Summary End

Intro

Numbers become better.

Intro Results Summary End

Intro

First conclusions now.

Intro Results Summary End

Intro

Oversimpli�cation ahead!

Intro Results Summary End

Functions and OO

Subs and methods: 11% faster

during 5.10..5.14+

Methods 50% slower than subs

selfmade/Moose/Mouse - nearly the same

Moose even faster than selfmade (?)

Intro Results Summary End

Threads

Threads �33% faster

during 5.13

Intro Results Summary End

Regex engines

pathological regexes: all regex engines at similar basic speed

RE2, Lua, Oniguruma, Plan9

POSIX: results under construction

Intro Results Summary End

Regexes

Regexes became �8% slower in 5.14.0

vs. 5.12back to normal in 5.14.2

Intro Results Summary End

Language Shootout

Shootout benchmarks 1% .. 15% faster in 5.14

vs. 5.12

Intro Results Summary End

Results

Show real graphs!

Intro Results Summary End

DPath.dpath

Intro Results Summary End

Fib

Intro Results Summary End

FibOO

Intro Results Summary End

FibMoose

Intro Results Summary End

FibMouse

Intro Results Summary End

Mem.allocate

Intro Results Summary End

Mem.copy

Intro Results Summary End

P6STD.viv

Intro Results Summary End

PerlCritic

Intro Results Summary End

RegexpCommonTS

Intro Results Summary End

Rx.regexes.�eldsplit1

Intro Results Summary End

RxCmp.native

Intro Results Summary End

RxCmp.Lua

Intro Results Summary End

RxCmp.Oniguruma

Intro Results Summary End

RxCmp.RE2

Intro Results Summary End

RxCmp.Plan9

Intro Results Summary End

SpamAssassin.salearn

Intro Results Summary End

Threads.threadstorm

Intro Results Summary End

ThreadsShared.threadstorm

Intro Results Summary End

More numbers at

http://perlformance.net

Intro Results Summary End

Summary

5.12 was slow

5.14 became faster

5.16 kept �at

Micro improvements don't help macro workloads yet

SpamAssassin

Do more variation for deeper understanding

non-threaded

http://perlformance.net

Intro Results Summary End

End