On 8/12/06, Ken Snider <ksnider at flarn.com> wrote:
> Gregory Maxwell wrote:
>
> > Please no bzip2.
> > The problem with bzip2 is that it is every bit as slow to compress as
> > it is to decompress.
>
> Ok, lets do a quick comparison for x86 (Run on a 2.8Ghz Xeon):

> 7zr a -t7z -m0=lzma -mx=9 -mfb=64 -md=32m -ms=on:

Cheat cheat. You couldn't have possibly made 7zr slower for
compression.  :) The default shouldn't perform much worse.

I can cheat too, can you spot how I'm cheating?

Compression test on ~450MB of Wikipedia text.

[gmaxwell at floodlamp tmp]$ ls -l big
-rw-rw-r-- 1 gmaxwell gmaxwell 447868953 Aug 12 22:12 big
[gmaxwell at floodlamp tmp]$ time bzip2 -k big

real    5m10.537s
user    5m9.819s
sys     0m0.492s
[gmaxwell at floodlamp tmp]$ time lzma -c big > big.lzma

real    3m57.193s
user    3m56.395s
sys     0m0.532s
[gmaxwell at floodlamp tmp]$ ls -l big.*
-rw-rw-r-- 1 gmaxwell gmaxwell 22813560 Aug 12 22:12 big.bz2
-rw-rw-r-- 1 gmaxwell gmaxwell   499021 Aug 12 22:23 big.lzma


big.lzma is available from:
http://myrandomnode.dyndns.org/big.lzma

Reply via email to