Tried it before and found it completely useless. The only place I cared about using parallel compression / decompression was on large files, and since it does everything in memory it dragged my system into a swapping hell every time that I used it.
You tried it, although it was just released today?
since it does everything in memory it dragged my system into a swapping hell
Well, it just keeps pending blocks in memory. It should be quite easy to bound the worst-case memory consumption (at some cost to parallelism), as they note in the article.
> You tried it, although it was just released today?
The version from clik was released today. Parallel bzip2 has been around for at least a couple of years. In fact, just doing a search for parallel bzip2 will lead you right to both an SMP/threaded version as well as one that uses MPI. I am guessing this is what was being referred to regarding the memory-consumption issues.
Jeff Atwood (from Coding Horror) ran a comparison of bzip2 on 8 cores against 7-Zip on 2 cores in February. bzip2 ran faster but generated larger files.
"The only reason to go bzip2 is if you need the file compressed as rapidly as possible to a reasonable size.
On pure efficiency / energy, 7zip wins, despite bzip2 using 4x the CPU time."
5 comments
[ 2.0 ms ] story [ 13.8 ms ] threadsince it does everything in memory it dragged my system into a swapping hell
Well, it just keeps pending blocks in memory. It should be quite easy to bound the worst-case memory consumption (at some cost to parallelism), as they note in the article.
http://compression.ca/pbzip2/
Though looking again, it seems like it's a separate implementation.
The version from clik was released today. Parallel bzip2 has been around for at least a couple of years. In fact, just doing a search for parallel bzip2 will lead you right to both an SMP/threaded version as well as one that uses MPI. I am guessing this is what was being referred to regarding the memory-consumption issues.
"The only reason to go bzip2 is if you need the file compressed as rapidly as possible to a reasonable size.
On pure efficiency / energy, 7zip wins, despite bzip2 using 4x the CPU time."
http://www.codinghorror.com/blog/archives/001231.html