Re: [PATCH 2/3] wildmatch: support "no FNM_PATHNAME" mode
- From
Nguyen Thai Ngoc Duy <pclouds@gmail.com>
- Date
- Dec 20, 2012, 13:34 UTC
- Message-ID
- <CACsJy8AFAKEeeFMp1EF+=QyZydoyVRFqPdAKrZbPe2h4_bf=fg@mail.gmail.com>
- In-Reply-To
- <CACsJy8C9Nhdq_xBAOxtmLcUnrUioAMvWCPk=sBZKOzFY2WR+iA@mail.gmail.com>
On Thu, Dec 20, 2012 at 8:55 AM, Nguyen Thai Ngoc Duy <pclouds@gmail.com> wrote:
Show 6 quoted lines
>> As long as simpler patterns fnmatch() groks (namely, '?', '*', and >> '[class]' wildcards only) are not slowed down by replacing it with >> wildmatch(), that is, of course. > > I'm concerned about performance vs fnmatch too. I'll probably write a > small program to exercise both and measure it with glibc.
I'll send out proper patches later this weekend, but unless I make big mistakes it seems wildmatch is faster than fnmatch from glibc 2.14.1 on x86-64 :) Maybe I did make mistakes. The test matches every line in /tmp/filelist.txt (which is ls-files from linux-2.6.git) 2000 times. "pathname" means FNM_PATHNAME.
$ export LANG=C $ ./test-wildmatch-perf /tmp/filelist.txt 'Documentation/*' 2000 wildmatch 1s 528394us fnmatch 2s 588396us $ ./test-wildmatch-perf /tmp/filelist.txt 'drivers/*' 2000 wildmatch 1s 957243us fnmatch 3s 134839us $ ./test-wildmatch-perf /tmp/filelist.txt 'Documentation/*' 2000 pathname wildmatch 1s 548575us fnmatch 2s 629680us $ ./test-wildmatch-perf /tmp/filelist.txt 'drivers/*' 2000 pathname wildmatch 2s 142377us fnmatch 3s 318295us $ ./test-wildmatch-perf /tmp/filelist.txt '[Dd]ocu[Mn]entation/*' 2000 wildmatch 1s 748505us fnmatch 3s 224414us $ ./test-wildmatch-perf /tmp/filelist.txt '[Dd]o?u[Mn]en?ati?n/*' 2000 wildmatch 1s 743268us fnmatch 3s 278034us $ ./test-wildmatch-perf /tmp/filelist.txt '[Dd]o?u[Mn]en?ati?n/*' 2000 pathname wildmatch 1s 745151us fnmatch 3s 314127us $ ./test-wildmatch-perf /tmp/filelist.txt nonsense 2000 pathname wildmatch 1s 310727us fnmatch 2s 279123us
-- Duy