RINGS Fast Bit Compressor.

[toffer]

I really like this one. Keep on going!

[Nania Francesco]

THANKS TOFFER! HI!

[Nania Francesco]

RINGS 1.2 released
- Max 50 MB of memory!
- Impressive compression !
- Implemented FCM engine!
Warning: Only for testing
Copyright ® 2007 by Nania Francesco Antonio (Italy).
All rights reserved.
link:
http://www.winturtle.netsons.org/rings.zip

[Nania Francesco]

SFC Test:
12.105.483 bytes COMP. 5,921 s. - DEC. 6,733
ENWIK8 Test:
25.873.235 bytes COMP. 8,562 s. - DEC. 12,922 s.

[LovePimple]

Thanks Francesco!

[Christian]

enwik9 -> 229.695.548 in 58 seconds!! Awesome compression speed for the resulting filesize, Francesco!
Btw., why is decompression actually slower than compression?

[LovePimple]

Quick test...

A10.jpg > 819,149
AcroRd32.exe > 1,493,651
english.dic > 566,325
FlashMX.pdf > 3,752,096
FP.LOG > 588,980
MSO97.DLL > 1,872,765
ohs.doc > 884,911
rafale.bmp > 784,993
vcfiu.hlp > 716,176
world95.txt > 626,437

Total = 12,105,483 bytes

[Black_Fox1]

Thank you
Results on my usual testset:
14 059 475 - RINGS 1.2
14 785 435 - RINGS 1.0

[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 > 229,695,548 bytes
Elapsed Time: 00:08:02.748 (482.748 Seconds)

[vcore]

Hi , Francesco

RINGS 1.2 is very fast and nice!!

[Nania Francesco]

Thanks to all how much! Hi !

[Nania Francesco]

RINGS 1.3 released
- Max 50 MB of memory!
- PURE POWER !
- new FCM engine!
- More Faster 10-50% !
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!

Mirror: Download

[LovePimple]

Quick test...

A10.jpg > 819,149
AcroRd32.exe > 1,493,651
english.dic > 566,325
FlashMX.pdf > 3,752,096
FP.LOG > 588,980
MSO97.DLL > 1,872,765
ohs.doc > 884,911
rafale.bmp > 799,866
vcfiu.hlp > 716,176
world95.txt > 626,437

Total = 12,120,356 bytes

Speed is impressive!

[Nania Francesco]

Thanks Love Pimple ! Hi!

[LovePimple]

It compresses ENWIK8 to 25,873,235 bytes in 21.9 seconds on my AMD Sempron 2400+.

[Nania Francesco]

RINGS 1.4c released

- Added 9 commands Min=2 MB - Max=512MB
- ULTRA COMPRESSION!
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!

Mirror: Download

[LovePimple]

Quick test...

RINGS c 1

A10.jpg > 819,150
AcroRd32.exe > 1,564,493
english.dic > 599,707
FlashMX.pdf > 3,777,642
FP.LOG > 799,840
MSO97.DLL > 1,991,174
ohs.doc > 926,673
rafale.bmp > 839,574
vcfiu.hlp > 733,449
world95.txt > 716,408

Total = 12,768,110 bytes


RINGS c 4

A10.jpg > 819,150
AcroRd32.exe > 1,493,652
english.dic > 566,326
FlashMX.pdf > 3,751,205
FP.LOG > 613,832
MSO97.DLL > 1,872,766
ohs.doc > 884,912
rafale.bmp > 799,867
vcfiu.hlp > 716,177
world95.txt > 626,438

Total = 12,144,325 bytes


RINGS c 9

A10.jpg > 819,150
AcroRd32.exe > 1,493,652
english.dic > 566,326
FlashMX.pdf > 3,752,097
FP.LOG > 554,280
MSO97.DLL > 1,872,766
ohs.doc > 884,912
rafale.bmp > 799,867
vcfiu.hlp > 716,177
world95.txt > 626,438

Total = 12,085,665 bytes

[LovePimple]

Another quick test...

RINGS c 1

ENWIK8 > 29,038,653 bytes


RINGS c 4

ENWIK8 > 26,442,044 bytes


RINGS c 9

ENWIK8 > 24,591,826 bytes

[Nania Francesco]

Thanks of all tests LovePimple ! Hi !

[toffer]

The speed with this ratio is really imperssive! I still wonder how you got this context mixer so fast.

Keep improving this! Good work!

EDIT: Matt says, that Rings is BWT?! I thought it is CM?

[nimdamsk]

-1 > 2 MB > BWT+MTF+PRE-HUF+ARI
-2 > 4 MB > ???
-3 > 8 MB > BWT+MTF+PRE-HUF+ARI
-4 > 16 MB > ???
-5 > 32 MB > FCM+LZP+FILTERS
-6 > 64 MB > ???
-7 > 128 MB > ???
-8 > 256 MB > ???
-9 > 512 MB > ???

Are memory requirements right? What algoritms are used instead of ???

[Nania Francesco]

For all profiles
-1
.
.
-9
simply BWT ->MTF -> PRE-HUF -> ARI
Example:
0123012301230123 etc.
BWT -> 0000111122223333 etc.
MTF -> 0000100020003000 etc.
PRE-HUF -> 0=bit 0 1=bits 100 2=101 3=1100 etc.
ARI -> COMPRESSED

[toffer]

What about your "fast context mixing" algorithm? Did you always use BWT?

BTW: The dynamic mapping of codes via huffman is a nice idea, which speeds up compression, exspecially after BTW+MTF (low order 0 entropy) and improves it. I had a very beneficial effect after applying just a simpler static mapping of the flat 8 bit codes to 4, 8, 12 bit codes (instead of dynamically built huffman codes).

[Piotr Tarsa]

is it full bwt or schindler transform? grzip uses st4 (shindler transform of order- 4) in its fast modes. do you use pre- bwt stages?

[Nania Francesco]

Old Rings 0.1-0.3 use:
14 MB of buffer -> ARI (Simple FCM + bits mixer ) [2 MB]

Rings 1.0 - 1.4c use:
Fast BWT->MTF->PRE-HUF->ARI
(NEW FCM CORE OPTIMIZED FOR COMPRESSION OF BWT) + ARI [2MB](Simple FCM + bits mixer ) for GIF,JPG,COMPRESSED FILES

[Piotr Tarsa]

what's "fast bwt"? how it differs from "slow" bwt?

[Bulat Ziganshin]

i bet that he uses dark sorting engine which is very good

[Piotr Tarsa]

but it can't be that fast. i bet he uses shindler transform. it's super fast - it's a limited order bwt. with limit = 4 (it's called st4 then) you can do the transform using one std::sort call on normal 32 bit uints (not strings).

also bwt with such memory requirements would provide much better compression.

divsufsort is even faster than dark afaik.