CSC3 FINAL

[Fu Siyuan]

After so much little improvements done. I finally decided to finish this version. Now more or less I am satisfied with its performance, compared to my former works.


About the program:

- LZ77 (1M -- 128M dictionary) + range coder
- Hash3+HC4 match finder + flexible parsing, some tricks added.
- E89transform by Shelwien. A simple 3-channel Delta on RGB or 4-channel Delta on WAV, not so good, but better than nothing. They are turned on automaticly by file extension or manually.

- Three mode: -m1..-m3 from fast to high compression.
- typical speed: 2-5 M/s compression 15-40 M/s decompression (Single Core on 2.+ GHZ CPU)

Usage: csc3 [options] file1 file2
Options:
-m1 Fast -m2 Efficient -m3 Strong
-d0 .. -d7 Dictionary Size:1M..128M (2^n)
-fe Exe filter, -fd# delta with #channels, -fo Turn Off the filter.
This parameter is recommended when compressing unknown extensions.
-t DO NOT OUTPUT ( Maybe for test ).
file2 isn't necessary in such situation.
Default: -m2 -d4 (16M) Auto filter by extension.

example: csc3 -m2 -d5 -fe e:\d\t\tar2 r:\out.csc


Special thanks:
Osman , Shelwien who are very patient to me and give me many experiences
which can not be obtained on books. 8 months ago I knew nothing about DC.
And Bulat who is the most experienced LZ-developer on the forum also gave me some practical advices.
nanoflooder helped me made some tests.

The code:
Though the code was completely done by myself (except e89). Little idea was original invented. I think it's better to post the source. It's programmed under visucal c++ 2008 express edition.
The code may looks very ugly .


I still have so much to improve. Filters&detection BinaryTreeMF OptimalParsing are the main for me. The programming have began, however there is still long distance to become practical.

This version is hoped to be final, but I can't guarantee no bug even many tests done on my own computer. So If you found bugs, tell me as soon as possible please. Hope it will work on squeezechart's wikipedia.

Replys and advices always be welcome!

[Nania Francesco]

Intel COre duo 2 E6600
Am trying him/it for MOC but doesn't work in all settings! The program cannot be launched!
My test results
C:\>CSC3
Impossibile eseguire il programma specificato.

C:\>CSC3 -m1 c:\test\fp.log
Impossibile eseguire il programma specificato.

C:\>CSC3 -m1 c:\test\fp.log t.a
Impossibile eseguire il programma specificato.

C:\>CSC3 -m1 c:\test\fp.log c:\t.a
Impossibile eseguire il programma specificato.

C:\>CSC3 -m1 -fe c:\test\fp.log c:\t.a
Impossibile eseguire il programma specificato.

C:\>CSC3 -m1 -d1 -fe c:\test\fp.log c:\t.a
Impossibile eseguire il programma specificato.

[Fu Siyuan]

Intel COre duo 2 E6600
Am trying him/it for MOC but doesn't work in all settings! The program cannot be launched!
My test results
C:\>CSC3
Impossibile eseguire il programma specificato.
...
.

Sry. Though I don't know what does "Impossibile eseguire il programma specificato." mean. I guess it is the compile problem. May be you lack of some runtime library. I uploaded a new compiled file. This should be works!

[Nania Francesco]

CSC3 tested and added in the next release of MOC single File compressor ! No errors !

[Fu Siyuan]

Thanks Nania! I'm looking forward to the results.

[Matt Mahoney]

Test results. http://mattmahoney.net/dc/text.html#2634

[Fu Siyuan]

Thanks Matt! But CSC is the real program name. 2 or 3 are the versions. So if csc4 comes out, it should be merged with csc3 csc2.

[Matt Mahoney]

Thanks for the explanation. I combined csc2 and csc3 results into one entry in LTCB. Also, both versions are listed in the generic test. http://mattmahoney.net/dc/uiq/

[LovePimple]

After so much little improvements done. I finally decided to finish this version. Now more or less I am satisfied with its performance, compared to my former works.


About the program:

- LZ77 (1M -- 128M dictionary) + range coder
- Hash3+HC4 match finder + flexible parsing, some tricks added.
- E89transform by Shelwien. A simple 3-channel Delta on RGB or 4-channel Delta on WAV, not so good, but better than nothing. They are turned on automaticly by file extension or manually.

- Three mode: -m1..-m3 from fast to high compression.
- typical speed: 2-5 M/s compression 15-40 M/s decompression (Single Core on 2.+ GHZ CPU)

Usage: csc3 [options] file1 file2
Options:
-m1 Fast -m2 Efficient -m3 Strong
-d0 .. -d7 Dictionary Size:1M..128M (2^n)
-fe Exe filter, -fd# delta with #channels, -fo Turn Off the filter.
This parameter is recommended when compressing unknown extensions.
-t DO NOT OUTPUT ( Maybe for test ).
file2 isn't necessary in such situation.
Default: -m2 -d4 (16M) Auto filter by extension.

example: csc3 -m2 -d5 -fe e:\d\t\tar2 r:\out.csc


Special thanks:
Osman , Shelwien who are very patient to me and give me many experiences
which can not be obtained on books. 8 months ago I knew nothing about DC.
And Bulat who is the most experienced LZ-developer on the forum also gave me some practical advices.
nanoflooder helped me made some tests.

The code:
Though the code was completely done by myself (except e89). Little idea was original invented. I think it's better to post the source. It's programmed under visucal c++ 2008 express edition.
The code may looks very ugly .


I still have so much to improve. Filters&detection BinaryTreeMF OptimalParsing are the main for me. The programming have began, however there is still long distance to become practical.

This version is hoped to be final, but I can't guarantee no bug even many tests done on my own computer. So If you found bugs, tell me as soon as possible please. Hope it will work on squeezechart's wikipedia.

Replys and advices always be welcome!

Thanks! Both compiles work perfectly on my PC's.

[LovePimple]

I have no idea why, but compiling with GCC seems to have resulted in a slower executable with better compression. I have altered the source code slightly to stop compiler warnings, and to remove the message about using -h to view the help text.

[Fu Siyuan]

I have no idea why, but compiling with GCC seems to have resulted in a slower executable with better compression. I have altered the source code slightly to stop compiler warnings, and to remove the message about using -h to view the help text.

Thanks, LP

however

31825244 vs 31824715 on enwik8 with same parameters...

Why the same code occur different result? So strange for me!

[Bulat Ziganshin]

Why the same code occur different result? So strange for me!

unitialized vars?

[Fu Siyuan]

unitialized vars?

Maybe. Shelwien said so too.