Re: Timing 2.4 vs. 3.x

• To: netcdfgroup@xxxxxxxxxxxxxxxx
• Subject: Re: Timing 2.4 vs. 3.x
• From: jps@xxxxxxxx (John Sheldon)
• Date: Mon, 2 Jun 1997 13:58:10 -0400 (EDT)

> >    Has anybody timed NetCDF v2.x compared to NetCDF v3.x? For both
> > read & write.  Is the difference large enough to justify
> > recompiling?
> 
> On what platform?  We haven't done actual timings on them, but Sun
> and SGI times don't seem to have changed for us (and were small to
> begin with).  Cray vector times, however, are _way_ down -- by better 
> than a factor of 10.

Our most recent tests of 2.4.2 vs 3.3 on our Cray T90 show a reduction
in CP time of about 40% for writes and 20% for reads.  More important,
perhaps, is the "fix" for the pre-fill option, which used to consume
more than 100x the CP time of the actual data write.

I am not sure why Carlie is getting a factor of 10 speedup on his Cray.
Optimization differences?

John P. Sheldon
(jps@xxxxxxxx) 
Geophysical Fluid Dynamics Laboratory/NOAA 
Princeton University/Forrestal Campus/Rte. 1 
P.O. Box 308 
Princeton, NJ, USA  08542

(609) 987-5053 office
(609) 987-5063 fax

• References:
  • Re: Timing 2.4 vs. 3.x
    • From: Carlie Coats