textfiles/programming/AI/news13.txt

67 lines
3.2 KiB
Plaintext

This response was prepared in response to a question about where
the membership functions and rules of a fuzzy logic unit are
stored in the assembly code generated by Fide.
Aptronix has a methodology we use in the generation of the assembly code.
We call this methodology TVFI which stands for the Truth Value
Flow Inference method. This is a proprietary method which was
developed by the Fuzzy Logic experts at Aptronix to permit faster
execution of the assembly code as well as the most efficient use
of memory possible.
There are three major steps that occur during the generation
of assembly code. First, FIL, (fuzzy inference language) is a
high level language that allows humans to easily define a fuzzy
inference unit. This language is designed to be a user-friendly
interface for humans. Secondly, the FIL source code is converted
into the format of the standard data structure of Aptronix. This
data structure is a medium between the easy human interface of
FIL and assembly code. The data structure is designed to be
compatible with hardware and has a format such that it can
easily generate code for almost any target chip. Thus, as FIL
is the interface that allows humans to work easily with fuzzy
logic, the standard data structre allows hardware to use fuzzy
logic easily.
Finally, the FIU (fuzzy inference unit) stored in the
standard data structure format is converted into assembly code.
It is impossible to trace directly from this assembly code line
for line to the source code written in the FIL. This is because
different parts of rules and membership functions may be combined
as part of the TVFI methodology. We have the capability to
optimize the target- specific code generated completely for speed
or completely for efficient use of memory depending upon your
application. Custom RTC (Real Time code convertors) can be
designed by Aptronix upon customer request.
To summarize, the assembly code generated by Fide is
optimized for the fuzzy logic aspect. We know how to optimize
the code because the Aptronix engineers have used fuzzy logic in
applications for many years and are experts on fuzzy logic and
its applications. Thus, there is no way to optimize assembly
code generated from a fuzzy inference unit more than with the
TVFI methodology of Aptronix and the standard data structure.
The individual instructions for the target chip contained in our
assembly code can be modified for optimization however.
(Although this should be done carefully). For instance, three
instructions executing a load and multiply sequence can be
replaced by one instruction accomplishing the same sequence.
This does not interfere with the assembly code for the fuzzy
inference unit.
Scott Irwin
Software Engineer
Customer Support