I canot see any long, "un-signable" complicated "terms and conditions" to posting here, so I guess I can post here.

This is a test post to see if it works.

By the way, I apparently discovered how to store incredible amounts of data in a very minimal amount of space, as it were.

I'm writing a small online book on data compression that might be useful for anyone developing algorithms or software. Comments and corrections welcome.

http://mattmahoney.net/dc/dce.html

Hi :

>>If you know of an (experimental) program not listed here >>or know of a switch combination yielding better results >>than the one listed please let me know!.

check this out :
http://groups.google.co.uk/group/comp.compression/browse_thread/thread/56e3b97913192e1b/837b7cb91dac658a?hl=en#837b7cb91dac658a

....However with quite random file such as reasonable size WinZip final compressed file , this early version ranking.exe ( download the .exe : https://www.box.net/shared/static/rm79sk26oy.zip )
often can reduce the compressed winzip file even further final winzipped file probably quite random but not always complete random .... its assumed winzip has the best state-of-art compressions available (?), cost 1bit to indicate if further compressable thus compressed reduced further with
ranking.exe

its simple enough confirm verify with own files, winzip them & then further compress with ranking.exe

Why don't you give IZARC a try ? http://izarc.org/

Ultra-fast LZ has been released! Check it out!
http://encode.dreamhosters.com/attachment.php?attachmentid=1189&d=1265054281

I have developed an interesting lossless data compression algorithm. I am looking for some adivce on how to commercialize it. It can perform better than many exisitng data compression lagorithms. some of the data already published in International journals and conferences.

>Any news on your benchmark site?
Nope, still closed... and will stay so at least for the foreseeable future.

Jan Ondrus has released paq8pxpre which combines precomp with paq8px_v67. It compresses flashmx.pdf to 1588181 and SFC to 6778517. http://encode.dreamhosters.com/showthread.php?p=10719#post10719

we are looking for matlab codes for lossless data decompression in launch vehicle telemetry system. please do help us..

I've sent the Excel sheet table with different Efficiency formula (and the capability to set parameters such as size rate increase) to : info at maximumcompression dot com

Hope it can be useful in any way.

Have you tried xz utils http://tukaani.org/xz/? I couldn't find it in any of the comparisons. It is supposed to eventually be the successor of gzip.

> Viktor, it's lzrw4 aka rolz in http://www.cbloom.com/src/lzrw.zip
Good one! I checked the file, and indeed the graph is there in ascii form. Not 100% sure this is the one I'm looking for. The author mentions several papers, and I checked them all, except one:

[Langdon84] Langdon G.G., "On Parsing Versus Mixed-Order Model Structures for Data Compression", IBM Research Report RJ-4163 (46091) 1/18/84, IBM Research Laboratory, San Jose, CA 95193, 1984.

"In [Langdon84], he has a more thorough byte at the problem, formalizing the whole case and proving that dictionary algorithms can always be emulated by Markov algorithms, but not the converse."

This sounds like the paper I'm looking for, however I failed to find it on IEEE/ACM. Does anyone have a ps/pdf copy of this paper on disk, or know a site that hosts this paper?

Wonderful site, clear and interesting information.
I frequently looked at the efficiency scores; since scores increase exponentially with size (2X every 10% increase) and linearly with time, I asked some question to myself:

- What would happen if I set a different score with 5% or 20% increase instead of 10%?
- Compression and decompression are independent features, then a vector sum (sqrt(comp_time^2+decomp_time^2)) can be more significant than a linear sum; what happen in that case?
- Considering extension to the Moore law, also processor performance increase exponentially (and consequently processing time decrease exponentially with new processors). What would happen if I use an exponential increase for both size and time (e.g. 2X increase when time is 10X, or 5X)?
- If I set different Efficiency scores, after having normalized them to the same min/max value, I can compute an overall average score just using sum or vector sum, to identify the best overall compressor in all the different efficiency expressions.

Well, I took the data from you table, did the computation in an excel sheet and WOW...
... the different efficiency rules (vector sum, values of percentage increase, exponential time increase) were mostly consistent to the original Efficiency formula: that means your results are good and significant.

Great job!

Viktor, it's lzrw4 aka rolz in http://www.cbloom.com/src/lzrw.zip

Hello. I've been trying to find one particular publication on the topic of comparing LZ and PPM. I got the document some 5 years ago (probably off of ACM), lost it, and couldn't find it again. I don't remember its name nor its author; I know it set up some sort of equivalence between LZ, and PPM where the accumulated context is discarded on each "match". It had a small graph that showed the context lengths in a zigzag pattern while processing some example input (the only thing I can distinctly recall).

I asked Werner about it, and he suggested I post here. Is anyone familiar with the abovementioned paper?

is gzip -9 better than gzip -lz77 for compression?

sbc: are you tried to add -of? and try -b1, it may be much faster due to fitting in L2 cache

about sbc - it spends too much time analyzing the data. please try "-m1 -b0 -of"

Thanks for test! Excuse me for error!
For me best flashzip option is -m0 -c7 -s (solid)!
Hi!

thank you for update, it's amazing to see how each speed improvement in FA reflected in better results. i've seen morning'23 results were not so good, what you have changed in test platform?

we all wish that you will continue to support up-to-date tests. please be less lazy! are you automated your testing using my scripts or something else? meantime, i've estableished my own little test: http://freearc.org/HFCB.aspx and w/o nightly scripts it would be impossible to perform all these tests

for further tests i propose: freearc -m1, nanozip -cf/-cF/cd, rar -m1 -md256 -s, rar -m2 -md256 -s, csc -m0..m3 -d7 (instead of current tests), pigz -1/-6/-9, pbzip2 -1/-2/-9, 7zip lzma2 test from HFCB table, winrk best asymmeric mode, paq8px67, paq8pf

plus i propose to remove several modes that aren't interesting: rar default (=normal 4mb non-solid), ccm default, sbc -b9 (sbc -m1 -b3 or more will be more interesting)

and, can you please unoficially test for me freearc -mex1..5 modes, as you can see on HFCB page, they should provide huge step forward, with 1.5-2x better efficiency. if possible, they should be tested on RAM disk, especially fastest modes that provides compression speed up to 150mb/s

also, can you please

Hello Werner! Good to hear from you again, thanks for the new update

Thanks for the update!

And could have a test of my csc31 -m0 -d5 after main works done? I think it should take only a few seconds.

I suppose there is a little misprint in latest news.
Should be WinRAR 3.91, not 3.61

Please in the next release please add
- Zhuff
- Flashzip 0.99b5b
- Hook 1.4
- bcm 0.10
Hi! Thanks!

Hello all. After some (too) big delay I'm back testing compressors. I just updated the site with some new ones. I'm not yet finished as 7-Zip 9, LZTurbo 0.95, WinZIP 14 and LZP2 are still to be tested.

Did I miss some important releases?. Also what's the latest update on PAQ8Q?. What is the latest high performance version? Got a bit lost with all the PAQ forks going on 6 months ago

Hey every one. I have been a follower of this board for quite a long time teehee :D Just wished to finally say hi to everyone! I cannot wait till new year! By the way, does anybody know if you can watch Avatar movie online on this site like it states I can? I very much wanna watch it but i am quite broke atm teehee :P