Synthetic compression benchmark

[giorgiotani]

Hi, I just updated PeaZip project's website with a benchmark page:
http://peazip.sourceforge.net/peazip...benchmark.html

Please note that the benchmark is oriented to generic end users, so it aims to be simple to read and takes in account very simple settings for tested programs (WinZip, WinRar and PeaZip), basically default compression settings and maximum compression settings where it can be interesting.

File tested are (unpacked) Firefox 2.0.0.12 and Google Earth 4.2BZXV (inspired form 7-Zip benchmarks).

I would be glad to shares impressions and thoughts about the method and results.

[LovePimple]

Thanks Giorgio!

[IsName]

Why no tar.lpaq8 results? They should be far better then lpaq5 due to e8/e9 preprocessing.

[giorgiotani]

Why no tar.lpaq8 results? They should be far better then lpaq5 due to e8/e9 preprocessing.

The benchmark is focused on comparison between WinRar, WinZip and PeaZip, which does not yet officially supports Alexander Ratushnyaks .lpaq8

BTW for next release Im working on a mechanism to support custom compression/decompression executables, in order to make simpler to use any arbitrary version of paq* and lpaq* executables in PeaZip.

[AiZ]

Hello,

Good start. But IMHO you should add a line to say which compressors take advantage of your dual core, to be fair in comparing compression times. Or not.


AiZ

[giorgiotani]

But IMHO you should add a line to say which compressors take advantage of your dual core, to be fair in comparing compression times.

Good point, Ill add an "Algorithms" subsection to explain how many threads can be used by each program for each tested format, and consequently what results are influenced by the dual core CPU.

It would also be interesting disabling a core and create a second table to compare the results, or better, testing it on a quad core (I havent, unfortunately) using 1,2 and 4 cores.
It will allow to see better what programs takes more advantages of parallelism, aside the known number of threads: it matters what each thread do, how they are balanced, how they acess memory and how much weights the syncronization overhead.

But as said in the op, I intended the test to be a quick chart for end users (a la 7-Zips homepage comparison table) more than a comprehensive test like Maximumcompression one, so I hope in future to be able to add data without losing in readability for end users.

[giorgiotani]

Updated the syntetic benchmark:
http://peazip.sourceforge.net/peazip...benchmark.html

Added some comments about results and a table comparing speed using one and two cores.