Re: Parallel GZIP --> speed instead of 2 CPUs running

[Osamu Aoki <osamu@debian.org>]

Hi,

I am wonderig why Ron Johnson replied to my post with just bzip2 bench
marks.  Maybe I was unclear about myself.

>> Osamu:
>> The full story is here on my amd64 core2duo system:
>>
>> Summary: (with -1 option for speed, real time for compression)
>>
>>           Binary files            Text file
>>           time(s)  size           time(s)  size
>>   cp      0.060    13166592       0.129    40323584                     
>>   lzop    0.448    12727856       2.526    28777352
>>   gzip    0.911    12689884       3.526    27281520
>>   pbzip2  2.027    12804577       4.800    25928838
>>   lzma    3.808    12772852       7.636    25926966
>>   bzip2   4.358    12804175       9.938    25919654
>>

This is my real time bench mark summary. I am not worried about minor
errors due to disk I/O (caching) since my test copied files first and
most likely cached them on memory.

Key points to note are
 * compression speed is very dependent to algoism.
 * bzip2 algorism finishes faster (1/2) if SMP is uses via pbzip2 program
 * bzip2 is very slow even with help of 2nd CPU.

I used lowest compression ratio.  All these algorisms do decent
compression job despite major speed difference. (See fie sizes)

Osamu

PS:  I did not understand intent of "Total" time written in Ron Johnson's post.