Hi,

there is a parallel gzip in testing/ from Mark Adler:
  pkg-get -s http://mirror.opencsw.org/opencsw/testing -U -u pigz
  pkgutil -t http://mirror.opencsw.org/opencsw/testing -i pigz

Some tests:

build8st% ls -l x
-rw-r--r--   1 dam      csw      336455680 Sep  1 17:05 x

File is roughly 336 MB is size

build8st% time pigz x
pigz x  215.55s user 1.91s system 3065% cpu 7.094 total

Compression takes 7 seconds instead 3 1/2 minutes on T5220!

build8st% time unpigz x.gz
unpigz x.gz  10.36s user 1.79s system 143% cpu 8.446 total

Parallel gunzip is only 1,5 times faster

build8st% time pigz -9 x
pigz -9 x  396.91s user 1.96s system 2334% cpu 17.087 total

Higher compressions take 17 seconds instead of 6 1/2 minutes!!

build8st% cksum x.gz
585068046       123219628       x.gz

Write that down.

build8st% time unpigz x.gz
unpigz x.gz  10.34s user 1.79s system 144% cpu 8.413 total
build8st% time gzip -9 x
gzip -9 x  241.84s user 1.56s system 99% cpu 4:03.40 total

Traditional gzip takes less total CPU but 4 minutes total.

build8st% cksum x.gz
206071496       123179310       x.gz

Different checksum, but both archives are valid.



I guess I'll switch package compression in GAR to pigz after
release.


Best regards

  -- Dago

PS: ...and don't forget: it is pronounced "pig-zee"

_______________________________________________
users mailing list
[email protected]
https://lists.opencsw.org/mailman/listinfo/users

Reply via email to