This bounced, so I'm forwarding it
-Christopher
------- Forwarded Message
From: Bart Kienhuis <[EMAIL PROTECTED]>
To: Jeff Patterson <[EMAIL PROTECTED]>
Subject: Re: [Ptolemy] Re: Quantizer bug
Cc: Stephen Andrew Neuendorffer <[EMAIL PROTECTED]>,
Zoltan Kemenczy <[EMAIL PROTECTED]>, [EMAIL PROTECTED],
[EMAIL PROTECTED], Sean Simmons <[EMAIL PROTECTED]>,
[EMAIL PROTECTED]
Date: Wed, 10 Apr 2002 19:00:27 +0200
Hello I create the FixPoint class back in June 1999. The challenge
of the FixPoint class was to use BigInteger to get a really platform
independent implementation of fix points. At the time I have build
the FixPoint class, good data was -- and still is -- missing that I
could use to debug or enhance the functional behavior of Fixpoint.
For the Fixpoint there are only a number of Unit tests, but there
are not functional tests. This makes it hard to get the functionality
of FixPoint debugged. However, 2+2=5 errors should have been removed...
Finding good data to debug functionally fix points is not trivial. I
recently spoke to people that are contributing their fix point
classes to SystemC. They used real hardware to debug their fix point
classes! They implemented add/substract/multiply/divide in an FPGA
and looked at the input/output behavior to debug the behavior of their
fix point.
Perhaps someone has a good list of known input/output values in various
fix point formats (preferably in some IEEE standard way) using the four
different rounding modes of the IEEE. This can then be used by someone
to check the functional behavior of the FixPoint class and its
quantizers.
Bart Kienhuis
_______________________________________________
Ptolemy maillist - [EMAIL PROTECTED]
http://www.gigascale.org/ptolemy/listinfo/ptolemy
------- End of Forwarded Message
----------------------------------------------------------------------------
Posted to the ptolemy-hackers mailing list. Please send administrative
mail for this list to: [EMAIL PROTECTED]
