[Zlib-devel] pigz 2.1.4 output differs for -p 1 and -p 2

[Mark Adler]

On Nov 14, 2008, at 12:11 AM, Pilorz Wojciech wrote:
> Actually in this case (as well as in a number of other tests if I  
> remember correctly)
> the output from parallel version is slightly shorter:
>

Hmm.  I guess sometimes the longer distance matches are a bad choice,  
even though the match length is longer, due to more bits required to  
code the distance.

I could get -p 1 to produce the same result, but it would slow down  
the non-parallel case to call deflateSetDictionary() for each block.   
What I should really do is figure out how to fix  
deflateSetDictionary() so it feeds in all 32K bytes.

Yes, I need to get consistency somehow for -R to be useful.

Mark