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

Russ Rew: Re: 960909: NetCDF vs. Raw IEEE I/O speed

Oops, forgot to CC: support-netcdf on this ...

------- Forwarded Message

Date:    Mon, 09 Sep 1996 08:38:42 -0600
From:    Russ Rew <address@hidden>
To:      address@hidden (John Sheldon)
Subject: Re: 960909: NetCDF vs. Raw IEEE I/O speed 

>From: address@hidden (John Sheldon)
>Subject: NetCDF vs. Raw IEEE I/O speed
>Organization: .
>Keywords: 199609090420.AA23337

Hi John,

> I thought you all would be interested in some timing tests I've done
> comparing netCDF I/O speed with flat IEEE I/O on the T90 (Unicos 9.1).
> Some strange results regarding IEEE I/O, itself, surfaced.  I also
> tried to assess how much of the speed penalty is incurred when using
> our comprehensive routine READNC_1, compared to a series of raw netCDF
> API calls.

I don't know much about the specific results you are seeing, but the
recently-announced 3.1a prerelease of the C interface for netCDF-3 might
be a better base on which to optimize.  It's structure is cleaner, and
it already is faster by a factor of 2 than netCDF 2.4 on non-Cray
platforms.  Although it introduces a new API, it supports the current
netCDF API with a thin compatibility layer.  You can get it at 

    ftp://ftp.unidata.ucar.edu/pub/netcdf/netcdf-3.1a.tar.Z

- --Russ

------- End of Forwarded Message