Remember: I'm not an expert at this stuff! That being said:
Do this second:
I would look deeper into what exactly the numerix functions are doing.
Do this first:
Try the simplest expressions you can. Start with a straight Python and
FORTRAN comparison to make sure you understand what is happening.
Build up from the simplest parts first...
My opinion only, of course...
Dan Lewis
Assistant Professor
Materials Research Center, Room 110
Rensselaer Polytechnic Institute
http://www.rpi.edu/~lewisd2
On Jun 8, 2006, at 12:23 PM, Damm, Edward F. (E. Buddy) wrote:
Good idea Dan,
I made it just a little complicated so I could make sure I was
using the
same Numerix functions (Pi, exp, arctan)....
Here are the results:
From FORTRAN
M**p = 5.170837517496476E-002
From Python
M**p = 0.0517083688947
My simple codes are attached. This example gives about the same
differences (wrt to how many decimals before deviation) as I tend
to see
in my FiPy phase field model against my bench mark Fortran code. Even
though these differences are in the 8th decimal place, I am seeing
significant differences in my run comparision.
Buddy
-----Original Message-----
From: [email protected] [mailto:[EMAIL PROTECTED] On Behalf Of Daniel Lewis
Sent: Thursday, June 08, 2006 11:03 AM
To: Multiple recipients of list
Subject: Precision Suggestion
Maybe trying a simple calculation (X + n*epsilon ??) in FORTRAN and
FiPy
would be an instructive yet empirical approach?
It is simpler than dealing with complicated equations...
Dan Lewis
Assistant Professor
Materials Research Center, Room 110
Rensselaer Polytechnic Institute
http://www.rpi.edu/~lewisd2
-----------------------------------------
This message and any attachments are intended for the individual or
entity named above. If you are not the intended recipient, please
do not forward, copy, print, use or disclose this communication to
others; also please notify the sender by replying to this message,
and then delete it from your system. The Timken Company / The
Timken Corporation
<DPTest.py>
<DPTest.f>
