The worst case for LZTurbo?

[m^2]

While testing how programs perform on a stripped Quake found something interesting...

"LZTurbo -11" 1227415 0.156
"LZTurbo -12" 1216635 9.750
"LZTurbo -13" 1182871 13.641
"LZTurbo -14" 1164885 27.734
"LZTurbo -15" 1155096 54.937
"LZTurbo -16" 1149690 109.625
"LZTurbo -17" 1146899 209.860
"LZTurbo -18" 1145159 847.688

Then I interrupted the compression process. It turned out that the problem is wqpro.exe.
Any idea why is it a problem? It's just a single 560KB file...
No other compressor I tested had issues.

[IsName]

On older forum,Bulat has proposed that it's "naive binary tree implementation". You can search old forum, but you'll mostly find discussion on GPL violation.

[Bulat Ziganshin]

at least cabarc had the same problem on pdf file in some (calgary?) corpus used in ACT. the problem is that with data like this:

00001
00002
00003
00004
00005

you binary tree becomes a list,a ctually. and snce you don't cutoff long searches, speed becomes o(n^2)

ps: try cabarc on this file

[m^2]

at least cabarc had the same problem on pdf file in some (calgary?) corpus used in ACT. the problem is that with data like this:

00001
00002
00003
00004
00005

you binary tree becomes a list,a ctually. and snce you don't cutoff long searches, speed becomes o(n^2)

ps: try cabarc on this file

I tried like 40 compressors with all supported switches already, cabarc too.
Only LZTurbo has problems.

[Bulat Ziganshin]

if you interested, tru to break file in two parts and then repeat this recursively. at the last end you may find rather small part which makes the problem

[m^2]

This is something strange. Can anybody try the attached file?

[LovePimple]

A new version (0.94) has been released...

http://lzturbo.web.officelive.com/documents/lzturbo.zip

[nanoflooder]

Thanks LovePimple!

[Nania Francesco]

Great job for you! Thanks!