[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: ramblings about old hardware, gzip, bz2, and pentium opts

Michael Neuffer <neuffer@alpha.mz.rhein-main.de> writes:

> * Jason Gunthorpe (jgg@ualberta.ca) [991130 03:45]:
> > 
> > On Mon, 29 Nov 1999, Joey Hess wrote:
> > 
> > > > A better test is to use gzip on a large file from cache or some other
> > > > integer program.
> >  
> > > Well I tried that some time ago and I was unable to find any significant
> > > speed differances. (Although making sure gzip gets the large file from cache
> > > is a PITA).
> > 
> > I remember the Stampede people claiming quite a gain on gzip with pgcc vs
> > gcc? Could their results simply have been tainted by the fact that gcc
> > 2.7.2 is really bad? :>
> I know it doesn't quite fit here, but with SUNs native compiler for
> SPARC you can indeed measure a speedup of almost 600% if you
> compile gzip optimized for Ultra SPARC (v9) instead of normal
> SPARC architecture.
> We measured that almost 1 year ago mainly on large files (DB dumps
> of about 1GB file size).
> Since then we use architecture specific versions for things like gzip
> and it makes a temendous difference in many occasions. The difference 
> between 20 minutes and 2 hours to gzip, transfer and archive DB dumps 
> can be and is indeed (for us at my current customer site at least) 
> business critical.

Make the gzip package create a gzip and gzip_ultra_sparc deb packages
and send the patches to the maintainer. I`m sure he will include that
in the distribution if done cleanly. Same maybe for gcc, xaos,
xfractin and ssystem.

May the Source be with you.

Reply to: