Subject:
|
IC floating point bug
|
Newsgroups:
|
lugnet.robotics.handyboard
|
Date:
|
Wed, 29 Jul 1998 13:58:01 GMT
|
Original-From:
|
Skye Legon <SLEGON@avoidspamWATFAST.UWATERLOO.CA>
|
Viewed:
|
847 times
|
| |
| |
Has any new information on the IC floating point bug come up? I have
been coding neural networks into my robot and was having problems.
A simple 'test' network of simple values ran fine, but when I manually
entered a real evolved network with many significant digits I got
garbage results. I assumed I had a problem somewhere with my code and
have been tearing out my hair, but if it is just a problem with
entering long literal strings of digits...
So what is up with NewtonLabs? I am an honest-to-gosh paying owner
of IC 3.2, and emailed their support address months ago with a question
about "how much global variable space is possible in IC", and have heard
nothing. Will they ignore this problem too?
Mildly annoyed, Skye.
PS:
Does anyone *else* know how much global variable space is available?
From the FAQ:
User program object code plus user global variables: 16320 bytes available
My code is about 10K in size, but I have problems with assigning more
than about 60 bytes of global variable space. It does *not* seem
limited by available user space.
+-----------------------------------------+----------------------------+
| Skye Legon | University of Waterloo |
| Systems Design Engineering | __/ __/ __/ __/ |
| Pattern Analysis & Machine Intelligence | __/ __/ __/ __/ __/ |
| 143 Columbia St. West, Unit E-4 | __/ __/ __/ ____/ __/ |
| Waterloo Ontario CANADA N2L 3L2 | __/__/__/ __/__/\__/__/ |
| +1(519)888-9249 | ______/ ___/ \___/ |
| slegon@uwaterloo.ca | DC 2620, 888-4567 x5192 |
+-----------------------------------------+----------------------------+
|
|
1 Message in This Thread:
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|