Re: ncdigest V1 #390

• To: netcdfgroup@xxxxxxxxxxxxxxxx
• Subject: Re: ncdigest V1 #390
• From: John Caron <caron@xxxxxxxx>
• Date: Mon, 02 Jun 1997 10:10:06 -0600

> > I tested 3.1alpha against 2.4 six months ago.
> > reading large (~10^5 floats) arrays was sped up by a factor of 2, I
> > assume due to efficiency at the xdr layer.
> 
>   is the speedup you're talking about due to 3.3 itself or the
> version 3 calls only.  In other words, if I continue to use
> the 2.4.3 calls that are in 3.3 will I still get the speed
> improvement?

that test used the old calls.  I think it was an SGI machine, but maybe
it was AIX.

if the speed is due to improved xdr layer, which I strongly suspect
(the old sun xdr implementation used a subroutine call for every array
element!)
you will get the speedup wherever the 3.3 replaces an inefficient
implementation.

The only way to be sure is to time your particular case.