Calgary Challlenge update

Calgary Challlenge update

Post by Matt Mahon » Tue, 13 Jan 2004 13:44:10


I submitted a RAR archive of 645,667 bytes to the Calgary Challenge
( http://www.yqcomputer.com/ ) beating the previous best entry of 653,720
bytes. The new entry was created with PAQC, a custom version of PAQ6. For
those interested, PAQC source code can be found at
http://www.yqcomputer.com/ ~mmahoney/compression/#challenge

The previous entry was a SLIM variant by Serge Voskoboynikov in Nov. 2002.
He has improved SLIM by quite a bit since then, so I would not be surprised
if he quickly retakes the lead. However Leo Broukhis has now set a minimum
payout of $100.01 which would require an entry of 634,566 bytes, so it will
take some work.

PAQC is a general purpose archiver although I don't expect the compression
to be quite as good as PAQ6 except on the Calgary corpus. The main
improvements are a better model for pic (about 3K, to 26K), and slightly
better text compression due to removing interference from unneeded models.
However Serge's earlier submission compressed pic to 20K, and current
versions of SLIM are better at text compression than PAQC as well.

My submission was the first under a rule allowing a source code
decompressor. I did not make a big effort to shrink the program. I think
with some work I could probably reduce it from 6K to 3K or better after RAR
compression.

-- Matt Mahoney
 
 
 

Calgary Challlenge update

Post by Phil Carmo » Tue, 13 Jan 2004 16:53:41

"Matt Mahoney" < XXXX@XXXXX.COM > writes:


I trust that those that meet you in person equip you with a grand total
of $72 worth of congratulatory beer!

Nice work, Matt. I too look east for a response with bated breath!

Phil
--
Unpatched IE vulnerability: history.back method caching
Description: cross-domain scripting, cookie/data/identity
theft, command execution
Reference: http://www.yqcomputer.com/
Exploit: http://www.yqcomputer.com/

 
 
 

Calgary Challlenge update

Post by David A. S » Tue, 13 Jan 2004 21:08:33

"Matt Mahoney" < XXXX@XXXXX.COM > wrote in



Way to go its a good thing.


David A. Scott
--
My Crypto code
http://www.yqcomputer.com/
http://www.yqcomputer.com/
http://www.yqcomputer.com/
My Compression code http://www.yqcomputer.com/
**TO EMAIL ME drop the roman "five" **
Disclaimer:I am in no way responsible for any of the statements
made in the above text. For all I know I might be drugged.
As a famous person once said "any cryptograhic
system is only as strong as its weakest link"
 
 
 

Calgary Challlenge update

Post by Matt Mahon » Wed, 14 Jan 2004 00:29:18


surprised
minimum
will

Oops, checking my math, you would actually need 642,619 bytes, only a 3048
byte improvement. I think this could be achieved fairly easily, given that
the previous entry was 6K better on pic.

-- Matt Mahoney
 
 
 

Calgary Challlenge update

Post by xleob » Wed, 14 Jan 2004 02:04:04


That's right. But in Serge's case compressing source code instead of the binary
may be not as beneficial (he says there is inline asm, etc).
If not for the new rule, you'd have had to rewrite
PAQC in C to achieve a decent decompressor size. The object code of
C++ STL is a nightmare size-wise.

Leo
 
 
 

Calgary Challlenge update

Post by Matt Mahon » Wed, 14 Jan 2004 14:04:00


2002.
3048
that
binary

True, g++ produces about 90K after UPX, though Eugene Shelwien produced some
.exes about 12K. There is a big advantage to using a source level
decompressor in that you have complete control over it. To get the same
benefit in an executable I think you'd have to write it all in assembler.

-- Matt Mahoney