Squeeze Chart 2007

[Mike]

Stephan, I start a new topic where you can notify update of your bench.
It 'll be easier to follow.
I also have a request : could you make the test's date more visible on your page ? When you open main benchmark page, you see first revision (REV 6.0 - template ?)... you must search the date on the right end of the page (in small character)

[Stephan Busch]

Yes, you're right. I will put the date above Revision.
I have added 2 new photographs on hall of fame page:
Marcel Lemke (WinACE) and Nania (Hook).
Currently, PAQ8l is being tested. Hook 0.8c and Quad 1.10
as well as CCM 1.20a are already tested. Those results will
be online this afternoon (but PAQ8l only partly).
Thanks for this topic, Mike

Yours, Stephan

[Stephan Busch]

Hey Guys,

the promised main chart is now online

Yours,
Stephan

[Mike]

Thanks

[Christian]

Thank you very much, Stephan!

[encode]

Thank you Stephan!

[Nania Francesco]

Thank you Stephan! your benchmark is perfect!

[Stephan Busch]

Dear readers,

I just wanted to let you know that I have finished testing PAQ8L;
atm Hook 0.8e is running and then Quad. I will set the new results
online this evening. I woke up later today because I watched the
fight Maske vs Hill last night..


Cheers

[encode]

Boxen: Henry Maske vs. Virgil Hill - Die Revanche

Thanks for info!

[Nania Francesco]

We wait for all with eagerness! we wait for all this evening!

[Stephan Busch]

And not to forget: The brand new M99 from Michael Mansicalco,
that includes MSufsort 3.1beta, arithmetic coding.. from this algorithm the new M03 will be made and it is rumored that M03
will outperform PPMonstr..

More this evenig

[Christian]

And not to forget: The brand new M99 from Michael Mansicalco,
that includes MSufsort 3.1beta, arithmetic coding.. from this algorithm the new M03 will be made and it is rumored that M03
will outperform PPMonstr..

Dont tease us like this. M99 is totally new to me. Cant wait to see first results.

[michael maniscalco]

Dont tease us like this. M99 is totally new to me. Cant wait to see first results.

Actually, I wrote M99 back in 1999. I had not bothered to write a full implementation with BWT because I got destracted with MSufSort to create the BWT and then I invented M03 and had not found time.

I decided to implement it fully before moving on the M03, which is the current project. I *expect* M03 to perform in the same league as PPMonstr in most cases. I never promised that it would outperform it.

There is lots of room for improving M99 but I am focusing on M03 instead because I know that its much stronger. So its not likely I will find the time to go back and tune M99 for compression or speed.

Currently, its almost as strong as Dark for compression and, in fast mode its as fast as bzip2 with better compression at the same blocksizes.

But M03 is context based compression of the BWT and will be far stronger for compression than M99 can achieve. It wont likely be as fast though.

[Stephan Busch]

I am testing M99 right now, and it's far better than M99 from 2001,
but unfortunately, I can't use switch combination -m 32m -a;
M99 tells me then that it could not find source file. Without '-a'
option everything runs just fine.

[Christian]

Hi Michael!

Actually, I wrote M99 back in 1999.

Please excuse my mistake. I am interested in compression but Ive somehow lost the overview (because there are so many compressors).

But M03 is context based compression of the BWT and will be far stronger for compression than M99 can achieve. It wont likely be as fast though.

Thank you for this explanation. Your context-based BWT algorithm seems to be very powerful. Im really looking forward to your next release of M03.

Best regards,

Christian

[Nania Francesco]

Hi Michael! Very good work whit M99 ! Faster Compressor!

[Stephan Busch]

The new benchmark is online now

[Fallon]

I have added 2 new photographs on hall of fame page

Your hall of fame, Stephan, is a good effort to give a face to history.
And ... some inappropriate comments about bad hairdays from Darwinian- to our present times rise to my lips.

http://www.michael-maniscalco.com/compression.htm
Will have a look at the new version. Thanks.
(for bodily compression, I prefer the Heckler Koch MSG-90A1 to the Barrett M99. The Russians may like their SV-98 best.).

Boxen: Henry Maske vs. Virgil Hill - Die Revanche

Maske won on points and has retired again right away. With a new bank account.

[Stephan Busch]

and for mind compression the most widely used tools are alcohol and cigarettes

Yes, but Maske is already rich: He has some Mc Donalds stores which he supervises and he also has some car-selling houses..

[Christian]

The new benchmark is online now

Thank you very much, Stephan!
Your benchmark is awesome and very informative! But I think Im repeating myself.

[Stephan Busch]

Thanks for the flowers

But also CCM is awesome and it is my favorite at present

[encode]

Strange thing - theoretically, PIM 1.50 must has the same compression as PIM 1.40BETA - bit to bit identical. At your test the results are slightly different! A bug? Anyone can confirm that PIM 1.50 has different compression than 1.40?

[encode]

I've made a few tests - compression of 1.50 is the SAME as with 1.40BETA! I compressed many files, checked compressed size, md5 checksums of archives - all okay! Probably you changed file names in your test?

[encode]

And one more important thing:

PIM 1.50 is NOT beta!
QUAD 1.10 is NOT beta!

At your test we can see:
PIM 1.50 Beta (2007)
QUAD 1.10 Beta (2007)



Please fix that!

[Stephan Busch]

Hi Ilia,

I am almost sure I used Pim 1.50 for the test and regarding the testsets: They are always identical, on the same testsets were Pim
1.40 running.. Indeed strange.. I will check this again and fix the writing errors. Please excuse me, Ilia..

[encode]

I will check this again and fix the writing errors.

Thank you very much Stephan!

[Stephan Busch]

I have retested PIM 1.50 on all testsets and got the same results.
So the archives are not the same compared to Pim 1.40 beta.
But I can remove 1.40 results if you want me to do..

[encode]

I have retested PIM 1.50 on all testsets and got the same results.
So the archives are not the same compared to Pim 1.40 beta.

Probably, something wrong with PIM 1.40BETA...

But I can remove 1.40 results if you want me to do..

Okay!

Anyway, as announced, PIM 1.50 is some kind of final version (however, the popularity of PIM is growing up with a logarithmic progression, so, it is possible some day Ill back to its development).

For me now QUAD is more important. QUAD 1.11 has a slightly higher compression, however, I stuck in the middle with Optimal Parsing... But the main thing Ive done in this version is the serious compression speed improvements in both modes. Thus QUAD with new Normal mode has an essentially new place, being faster than Fast LZMA and at the same time with Higher compression...

[Stephan Busch]

cool, I will test Quad 1.11 asap

[Stephan Busch]

The fixed benchmark is online. Thanks again for reporting, Ilia