<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-1" http-equiv="Content-Type">
<title></title>
</head>
<body bgcolor="#ffffff" text="#000000">
I've just completed a modification to the NCMPI option of IOR,
specifically so I could get benchmark results writing an arbitrarily
large file to a substantial Lustre FS from only 128 processors.<br>
<br>
IOR is written in C, and, due to my lack of recent experience with
Fortran, I stuck with that.<br>
<br>
I have some regard in migrating to Fortran, however, as I have an
involvement with the CAM code from UCAR, where the have written a new
I/O library on top of PNetCDF.<br>
<br>
Could you describe the research or code on which you are working? I'm
particularly interested in whether you have an abstract, I/0 layer; or
is your I/O integrated throughout you code?<br>
<br>
<br>
Marty Barnaby<br>
<br>
<br>
<a class="moz-txt-link-abbreviated" href="mailto:robl@mcs.anl.gov">robl@mcs.anl.gov</a> wrote:
<blockquote cite="mid:20080312153302.GD29640@mcs.anl.gov" type="cite">
<pre wrap="">Hi Brian
I know you sent this to me a month ago, but I've been pretty busy and
haven't had a chance to look at this yet. Let's see if the wider
parallel-netcdf mailing list has some feedback for you.
On Tue, Feb 12, 2008 at 09:27:50PM -0600, <a class="moz-txt-link-abbreviated" href="mailto:bdtaylo1@uiuc.edu">bdtaylo1@uiuc.edu</a> wrote:
</pre>
<blockquote type="cite">
<pre wrap="">Mr. Latham,
Hello! I am using the parallel netCDF library in a couple of
Fortran 90 CFD codes I've written for my research. I wrote a F90
module that wraps a large portion of the F77 API provided by
parallel netCDF library. I have done my best to follow the serial
netCDF F90 API in putting together this module. I thought this
might be of interest to you.
There are a good number of issues that need to be addressed, such as
how to define the required constants (I just copied them out of
"pnetcdf.inc"), the non- standard but widely supported usage of
byte-storage variable specification (integer*1, integer*2, real*4,
etc), and how to obtain the MPI_OFFSET_TYPE constant (either via
configure/preprocessor tricks, via the "use mpi" statement as I have
done, or via "include 'mpif.h'"). If you're interested, I can give
you a complete rundown of the issues I've found.
Feel free to contact me if you have any questions or if I can be of
further assistance.
Thanks, Brian Taylor
</pre>
</blockquote>
<pre wrap=""><!---->
Thanks for the starting point. Let's see if we can address the
remaining issues. Has anybody else on this list roled their own f90
bindings?
==rob
--
Rob Latham
Mathematics and Computer Science Division A215 0178 EA2D B059 8CDF
Argonne National Lab, IL USA B29D F333 664A 4280 315B
</pre>
</blockquote>
<br>
</body>
</html>