RINGS Fast Bit Compressor.

[Matt Mahoney]

Welcome to the Pareto frontier (an underlined compression time)

http://cs.fit.edu/~mmahoney/compression/text.html

[Bulat Ziganshin]

where is paretto frontier there?

[LovePimple]

RINGS 1.0 released
- Max 50 MB of memory!
- Pure Power!
- New FCM (Fast Context Mixing) engine!
Warning: Only for testing
Copyright ® 2007 by Nania Francesco Antonio (Italy). All rights reserved.
link:
http://www.winturtle.netsons.org/rings.zip

Thanks Francesco!

[LovePimple]

Quick test...

A10.jpg > 819,149
AcroRd32.exe > 1,578,864
english.dic > 619,433
FlashMX.pdf > 3,785,625
FP.LOG > 681,596
MSO97.DLL > 1,957,638
ohs.doc > 959,962
rafale.bmp > 784,993
vcfiu.hlp > 790,888
world95.txt > 643,346

Total = 12,621,494 bytes

[LovePimple]

Another quick test...

Test machine: Intel PIII (Coppermine) @750 MHz, 512 MB RAM, Windows 2000 Pro SP4

Test File: ENWIK9 (1,000,000,000 bytes)

Timed with AcuTimer v1.2


ENWIK9 > 235,897,616 bytes
Elapsed Time: 00:12:25.912 (745.912 Seconds)

[toffer]

Could you give us a describtion of the algorithm?

[Nania Francesco]

Rings use a preprocessor + Context Mixing + Filters !
The results on my Inter Core duo 2 are little different from the visualized pages! Nevertheless I use calculate with integer !

[Nania Francesco]

Excuse me! I confirm the tests in everything and for everything! some files of the SFC tests were corrupted!

[Nania Francesco]

INTEL CORE DUO 2 E6600 2GB RAM

UCLC Test Wave [16 FILES 51.014.416 bytes]

RINGS 1.0 -> 34.574.668 comp. 9,437 s. decomp. 13,859 s.
Hook 1.3 512 mb -> 34.448.856 comp. 15,969 s. decomp. 19,719 s.
CCM 1.30A opt.6 -> 33.402.490 comp. 48,937 s. decomp. 47,725 s.

UCLC Test PGM [24 FILES 9.437.544 bytes]

RINGS 1.0 -> 5.517.396 comp. 1,969 s. decomp. 1,906 s.
Hook 1.3 512 mb -> 5.756.246 comp. 11,157 s. decomp. 11,780 s.
CCM 1.30A opt.6 -> 5.561.545 comp. 28,362 s. decomp. 28,763 s.

UCLC Test TXT [1 FILES 16.013.962 bytes]

Hook 1.3 512 mb -> 3.952.691 comp. 8,844 s. decomp. 10,407 s.
CCM 1.30A opt.6 -> 3.959.687 comp. 8,516 s. decomp. 8,781 s.
RINGS 1.0 -> 4.192.457 comp. 2,141 s. decomp. 3,015 s.

UCLC Test TAR [1 FILES 78.747.600 bytes]

CCM 1.30A opt.6 -> 7.792.412 comp. 30,140 s. decomp. 30,437 s.
Hook 1.3 512 mb -> 10.923.053 comp. 20.375 s. decomp. 22,957 s.
RINGS 1.0 -> 13.616.378 comp. 10,406 s. decomp. 13,641 s.

UCLC Test WTB [25 FILES 5.992.832 bytes]

CCM 1.30A opt.6 -> 3.139.661 comp. 26,906 s. decomp. 27,343 s.
RINGS 1.0 -> 3.321.595 comp. 1,205 s. decomp. 1,482 s.
Hook 1.3 512 mb -> 3.487.280 comp. 10.936 s. decomp. 11,284 s.

[toffer]

I actually meant a more detailed describtion of your algorithm (ignoring filters). It seems to be very fast for CM indeed Thumbs up!

[Simon Berger]

Yes impressive. Thats one rare example where I would want more compression in loss of a little bit speed

[Nania Francesco]

Context Mixing structure is very simple [2 MB] and use Hash for store bit Context! Preprocessor use 48 MB in compression and 40 MB in Decompression!

[toffer]

Which orders do you use? (Or dependent on what do you build your contexts) I guess you use nibble aligned hashing?

How can preprocessors eat up so much memory? Or are you talking about some LZP+CM hybrid, where you use that much for a sliding window+hashing table? And CM just to output tokens (literals, lengths, ...) ?

[Nania Francesco]

Context Mixing Engine use order0 and order1 mix!
LZP not active at moment [ UCLC TAR TEST] !
Preprocessor variable mix oder0.....order4!

[LovePimple]

Excuse me! I confirm the tests in everything and for everything! some files of the SFC tests were corrupted!

I can confirm that vcfiu.hlp is corrupted by RINGS v1.0.

[toffer]

Preprocessor variable mix oder0.....order4!

Could you further explain this? I wouldnt call that a preprocessor?!

[Nania Francesco]

Excuse me LovePimple, my files of SFC (in my computer) are corrupted! My automatic-tester (MOC) confirm VCFIU.HLP are not corrupted!

[Nania Francesco]

For me is impossible that VCFIU.HLP is corrupt! it is to make effortless an immediate test:

C:>rings c vcfiu.hlp t.a
compress vcfiu.hlp 4121418 bytes to t.a 790888 bytes time= 0.64 s.

C:>rings d t.a test.hlp
decompress t.a 790888 bytes to test.hlp 4121418 bytes time= 0.69 s.

C:>rings c test.hlp t.a
compress test.hlp 4121418 bytes to t.a 790888 bytes time= 0.56 s.

[Simon Berger]

I proofed it, they are the same

[Nania Francesco]

Thanks Simon!

[LovePimple]

For me is impossible that VCFIU.HLP is corrupt! it is to make effortless an immediate test:

C:>rings c vcfiu.hlp t.a
compress vcfiu.hlp 4121418 bytes to t.a 790888 bytes time= 0.64 s.

C:>rings d t.a test.hlp
decompress t.a 790888 bytes to test.hlp 4121418 bytes time= 0.69 s.

C:>rings c test.hlp t.a
compress test.hlp 4121418 bytes to t.a 790888 bytes time= 0.56 s.

Apologies! It was an error in my batch file that caused the problem.

All files are OK!

[Nania Francesco]

Thanks LovePimple!
I owe nevertheless to confirm that the files of this type
file name=Text one.txt
file name=xxx yyy.bmp
etc.
that they contain one or more empty spaces is not compressed well!

[Matt Mahoney]

I tested rings 1.0 on SFC and verified decompression is correct on all files.

819,149 a10.jpg.rings
1,578,864 acrord32.exe.rings
619,433 english.dic.rings
3,785,625 FlashMX.pdf.rings
681,596 fp.log.rings
1,957,638 mso97.dll.rings
959,962 ohs.doc.rings
784,993 rafale.bmp.rings
790,888 vcfiu.hlp.rings
643,346 world95.txt.rings
12,621,494 bytes

[Matt Mahoney]

where is paretto frontier there?

That means there is no program that compresses both faster and smaller. On LTCB I underline the compression time. I also have Pareto frontiers for decompression time and memory.

[Nania Francesco]

RINGS 1.1 released
- Max 50 MB of memory!
- SuperSonic!
- New FCM (Fast Context Mixing) engine+LZP!
Warning: Only for testing
Copyright ® 2007 by Nania Francesco Antonio (Italy). All rights reserved.
link:
http://www.winturtle.netsons.org/rings.zip

[LovePimple]

Thanks Francesco!

[LovePimple]

Quick test...

A10.jpg > 819,149
AcroRd32.exe > 1,541,087
english.dic > 590,806
FlashMX.pdf > 3,770,482
FP.LOG > 637,759
MSO97.DLL > 1,926,950
ohs.doc > 899,384
rafale.bmp > 784,993
vcfiu.hlp > 772,137
world95.txt > 658,755

Total = 12,401,502 bytes

[LovePimple]

Another quick test...

Test machine: Intel PIII (Coppermine) @750 MHz, 512 MB RAM, Windows 2000 Pro SP4

Test File: ENWIK9 (1,000,000,000 bytes)

Timed with AcuTimer v1.2


ENWIK9 > 238,353,988 bytes
Elapsed Time: 00:09:02.522 (542.522 Seconds)


Compression speed is quite impressive!

[Nania Francesco]

INTEL CORE 2 DUO E6600 2GB RAM

CALGARY CORPUS
865.183 COMP. 0,388 SEC. DEC. 0,503 SEC.
CANTERBURY CORPUS
469.825 COMP. 0,313 SEC. DEC. 0,328 SEC.
UCLC PGM FILES
5.532.803 COMP 1,891 SEC. DEC 1,984 SEC.
UCLC WAVE FILES
34.771.034 COMP. 9,154 SEC. DEC. 14,360 DEC.
UCLC WTB [OTHELLO]
3.278.458 COMP. 0,983 SEC. DEC. 1,313 SEC.
UCLC MULTI LANG. TEXT
3.373.722 COMP. 1,250 SEC. DEC. 1,795 SEC.
UCLC HALF LIFE MOD
35.601.588 COMP. 21,970 SEC. DEC. 30,221 SEC.
UCLC LARGE TEXT
4.369.707 COMP. 1,560 SEC. DEC. 2,610 SEC.
UCLC UT2004MAP
3.184.257 COMP. 1,201 SEC. DEC. 1,812 SEC.

[Nania Francesco]

Thanks LovePimple! Hi !