MONSTER OF COMPRESSION - New Benchmark

[Cyan]

Hello Nania

I wish i could submit you a newer version of LZP2, which is available at this website :
http://phantasie.tonempire.net/utili...ws-t86.htm#133

The compression rate is improved at a negligible speed cost, so it may fare better.
Note that it doesn't change the whole picture : LZP2 is still a too-fast compressor, with a too low compression rate. Well, it is its category


Regards

[Fu Siyuan]

Hello Nania

I wish i could submit you a newer version of LZP2, which is available at this website :
http://phantasie.tonempire.net/utili...ws-t86.htm#133

The compression rate is improved at a negligible speed cost, so it may fare better.
Note that it doesn't change the whole picture : LZP2 is still a too-fast compressor, with a too low compression rate. Well, it is its category


Regards

Quick test on enwik9 with
Athlon X2 4000+ (2.10GHz) 2GB DDR800 Windows XP SP2

R:\>f:\d\timer r:\lzp2 f:\d\t\enwik9 f:\enwik9.lzp2

Timer 8.00 : Igor Pavlov : Public domain : 2008-11-25
LZP2 Oct 10 2009, compression software by Yann Collet
Compressed Size : 598076398 (59.81%)
Compression Time : 6.85s ==> 146MB/s
Total Time : 107.31s ( Read wait : 76.52s // Write wait : 23.93s )

Kernel Time = 2.265 = 00:00:02.265 = 2%
User Time = 6.828 = 00:00:06.828 = 6%
Process Time = 9.093 = 00:00:09.093 = 8%
Global Time = 107.312 = 00:01:47.312 = 100%

R:\>f:\d\timer -u f:\enwik9.lzp2 c:\enwik9out

Timer 8.00 : Igor Pavlov : Public domain : 2008-11-25

Can't find program. Error code = 2

R:\>f:\d\timer r:\lzp2 -u f:\enwik9.lzp2 c:\enwik9

Timer 8.00 : Igor Pavlov : Public domain : 2008-11-25
LZP2 Oct 10 2009, compression software by Yann Collet
Regenerated Size : 1000000000 Bytes
Decoding Time : 4.08s ==> 244MB/s
Total Time : 87.44s ( Read wait : 50.35s // Write wait : 33.01s )

Kernel Time = 3.031 = 00:00:03.031 = 3%
User Time = 4.031 = 00:00:04.031 = 4%
Process Time = 7.062 = 00:00:07.062 = 8%
Global Time = 87.441 = 00:01:27.441 = 100%

R:\>

Data No error

[Nania Francesco]

there are no problems, I add LZP2 in the next release!

[Nania Francesco]

LZP2 crash in file *.DLL compression! Please verify!

[Cyan]

Thanks for the testings

I've investigated from your bug report and finally found a file on which the error condition can be reproduced.
It's a pretty obvious flaw in end-condition testing. I'm surprised it never happened on my previous set of test files...
Anyway, this is corrected.

An updated version is available at website,
http://lzp2.dnsalias.com

[Edit] : the new end-condition had a remaining bug, resulting in decompressed file slightly altered under a certain set of conditions (mostly affecting large *.exe files). This is corrected. The direct-link is therefore removed, so that you get the updated version from website

[Nania Francesco]

MOC test passed for LZP2 0.7c ! Hi!

[Cyan]

thanks !

minor comment : as i notice you provide a flag of author's country in your charts, note that i'm actually leaving in France, not US (yet)

[Matt Mahoney]

LZP2 makes the Pareto frontier on LTCB. http://mattmahoney.net/dc/text.html

[Nania Francesco]

Excuse moi ! Naturellement vive la France!

[PiPPoNe92]

Ciao nania!! Anche tu italiano?

[Nania Francesco]

Si anche io sono italiano !
Yes! I am Italian!

[Nania Francesco]

Last MOC 2009 update is on-line ! enjoy!

[Bulat Ziganshin]

btw, please try to use -m9d option instead of -m9x - it should be more efficient (-m9x uses more memory for decompression what's fine for "best decomporession efficiency" benchmarks)

[michael maniscalco]

M03 didn't make the list?

[Nania Francesco]

Sincerely they are at work for the new MOC that from 2010 will be always the same group of files and I hope for a lot of time doesn't change! If I have some time I adjourn your applications!

[Nania Francesco]

I have need your help (a list of particular files also) that would like to insert in the new MOC 2010! I will also choose among the links of file free that will give me! Hi thanks!
LIMIT of MOC 2010 2GB!