PPMX - a new PPM encoder

[encode]

OK, check out my very new and DRAFT yet PPM encoder. It's quite basic and has a very few special tricks at this early stage of development, however.

No filters, just pure algo...

All in all it is an order-7 PPM with a few tricks. Currently it is designed for speed, so it has no such things like symbol masking, SEE, SSE, etc. However it uses a special and my very own escape method...

Enjoy!

P.S.
Finally someone did not CM or LZ related...

[Shelwien]

http://compression.ru/download/sources/cm/dummyppm.rar
http://compression.ru/download/sources/cm/dum_ppm.rar

[encode]

http://compression.ru/download/sourc...odec-1.0.1.zip

[Shelwien]

Thats kinda unrelated

[encode]

[encode]

ENWIK8 -> 24,369,312 bytes
ENWIK9 -> 213,206,926 bytes

world95.txt -> 535,733 bytes
fp.log -> 630,767 bytes

3200.txt -> 4,148,657 bytes

[Shelwien]

Also this, quoting ppmd readme:
"[2] Very descriptive M.R.Nelson's COMP-2 program (PPMd is based on it)."
http://shelwien.googlepages.com/nelson_comp2.rar

[m^2]

Bookstar result:
10318337B, 33.422s.
About the same size and 3 times slower than 7z -m0=PPMD -mx3

[Shelwien]

Its really cool if the size is the same, btw.
Because there's a known redundancy (no symbol masking) in ppmx.
So compression ratio can be improved, and speed is a matter of working time.
...Which is also true for ppmd though

[osmanturan]

SFC benchmark on my laptop (Core2Duo 2.2 GHz, 2 GB RAM, Vista Business x64 SP1):

PPMX (540 MB memory usage)
PPMd var.J rev.1 (-o7 -m540)

A10.jpg (842,468 bytes)
PPMX -> 878,972 @ 470 KB/s (1.752 seconds)
PPMd -> 833,772 @ 767 KB/s (1.072 seconds)

AcroRd32.exe (3,870,784 bytes)
PPMX -> 1,733,372 @ 985 KB/s (3.836 seconds)
PPMd -> 1,549,310 @ 1,844 KB/s (2.049 seconds)

english.dic (4,067,439 bytes)
PPMX ->   988,675 @ 1,483 KB/s (2.678 seconds)
PPMd -> 1,067,526 @ 3,841 KB/s (1.034 seconds)

FlashMX.pdf (4,526,946 bytes)
PPMX -> 3.808.788 @ 861 KB/s (5.137 seconds)
PPMd -> 3.721.723 @ 966 KB/s (4.576 seconds)

FP.LOG (20,617,071 bytes)
PPMX -> 630,767 @ 3,603 KB/s (5.588 seconds)
PPMd -> 620,661 @ 25,105 KB/s (0.802 seconds)

MSO97.DLL (3.782.416 bytes)
PPMX -> 2,050,096 @ 961 KB/s (3.843 seconds)
PPMd -> 1,872,522 @ 1,514 KB/s (2.439 seconds)

ohs.doc (4,168,192 bytes)
PPMX -> 913,950 @ 1,415 KB/s (2.876 seconds)
PPMd -> 833,816 @ 4,022 KB/s (1.012 seconds)

rafale.bmp (4,149,414 bytes)
PPMX -> 886,972 @ 1,266 KB/s (3.201 seconds)
PPMd -> 784,493 @ 4,126 KB/s (0.982 seconds)

vcfiu.hlp (4,121,418 bytes)
PPMX -> 735,942 @ 1,726 KB/s (2.332 seconds)
PPMd -> 668,759 @ 4,702 KB/s (0.856 seconds)

world95.txt (2,988,578 bytes)
PPMX -> 535,733 @ 1,366 KB/s (2.136 seconds)
PPMd -> 474,445 @ 3,923 KB/s (0.744 seconds)

Total
PPMX -> 13,163,267 bytes
PPMd -> 12,427,027 bytes

According to your fixed size memory usage, you probably used a hashed statistics which makes your coder very slow. Also, compression can be dropped due to hashing. So, I highly advice to use tree like statistics.

[LovePimple]

OK, check out my very new and DRAFT yet PPM encoder. It's quite basic and has a very few special tricks at this early stage of development, however.

No filters, just pure algo...

All in all it is an order-7 PPM with a few tricks. Currently it is designed for speed, so it has no such things like symbol masking, SEE, SSE, etc. However it uses a special and my very own escape method...

Enjoy!

P.S.
Finally someone did not CM or LZ related...

Thanks Ilia!

[encode]

And this will happen if add masking (same coder and model + symbol masking):

book1 -> 231,223 bytes
world95.txt -> 519,126 bytes
ENWIK8 -> 23,898,278 bytes
bookstar -> 10,028,609 bytes
3200.txt -> 4,009,262 bytes

And if add an extra order-12 model:

book1 -> 224,351 bytes
world95.txt -> 492,022 bytes
ENWIK8 -> 22,916,946 bytes
bookstar -> 9,679,378 bytes
3200.txt -> 3,955,708 bytes

[SvenBent]

yay more tools to test

has been a god week first new rep and now this.
i got a sick day today so i might do alot of testing.

antthing speciel yo wan to have teste

otherwise I'll just compare to 7-zip ppmd

[encode]

LTCB:
http://cs.fit.edu/~mmahoney/compression/text.html#2132

[encode]

Added an extra compression:

ENWIK8 -> 22,580,291 bytes
ENWIK9 -> 194,298,469 bytes
BOOK1 -> 227,280 bytes
world95.txt -> 493,122 bytes
bookstar -> 9,613,763 bytes
3200.txt -> 3,927,886 bytes