Hi Rob,
> > Features of the 3.6.2 release (from the netCDF FAQ):
> >
> > * as much as 22% speedup for applications accessing floats or ints
> > on little-endian platforms.
>
> Apologies if this isn't the right forum, but this item really caught
> our attention. As someone working on parallel-netcdf, I'd sure like
> to hear more about anything that speeds things up 22%. What did you
> do to achieve this? ...
This resulted from unrolling a loop in the function swapn4b() in
libsrc/ncx.c (the source was actually libsrc/ncx.m4). If your
compiler already unrolls loops aggressively, you may not see this much
speedup.
--Russ
_____________________________________________________________________
Russ Rew UCAR Unidata Program
russ@xxxxxxxxxxxxxxxx http://www.unidata.ucar.edu
==============================================================================
To unsubscribe netcdfgroup, visit:
http://www.unidata.ucar.edu/mailing-list-delete-form.html
==============================================================================
