When talking about code generation for highly constrained real time application, a "good" C code is often not sufficient. The most critical functions of the system (e.g. FFT) must be implemented with assembly code i.e. optimized for the target processor. In order to keep the advantages of the C code (e.g. portability, maintainability), the application is implemented with a library, and if possible with a standard library (e.g. VSIPL).
For example, within Ptolemy Classic we have implemented a subset of VSIPL, both in the SDF and CGC domain. A standard C implementation of VSIPL has been linked with Ptolemy in order to run the simulation. The code generated includes calls to VSIPL functions, and is linked with a different implementation of VSIPL, which is optimized for the target processor.
Coming back to Ptolemy II, I'm wondering if your approach is suitable for such kind of implementation? I think that it would be interesting to be able to call a C function (VSIPL is written in C, so far) or a Java function (write VSIPL in Java) within the actors, and keep the function call as is, during the code generation process. Any comment?
Patrick
-- Patrick MEYER THALES Naval Systems Netherlands BU Radar & Sensors - Processing Zuidelijke Havenweg 40 7554 RR Hengelo Ov; The Netherlands mailto:[EMAIL PROTECTED] Phone: (+31) 74 2483973 Fax: (+31) 74 2484018---------------------------------------------------------------------------- Posted to the ptolemy-hackers mailing list. Please send administrative mail for this list to: [EMAIL PROTECTED]
