PAT is a generic data structure supporting vlsi concepts. It allows representation of a sequence of simulation patterns in a data structure.
The goal of PAT is to define an ad hoc single data structure (object), with well known fixed meaning for each concept manipulated in pattern description. So any tools that needs to access simulation patterns can be build upon it, using these structures.
For each object (data structure), a set of functions has been defined in order to add, delete or access it. These functions are defined in a low-level library libPat106.a. Some other functions are higher level. They are defined in a high-level library libPhl106.a. There are also some parsers-drivers. A parser reads a pattern description file in a given format and build up a complete PAT data base. A driver creates a text file in a given format from a PAT data base. A pat format parser-driver is available in the libPpt106.a library.
Pattern description envolves the following notions (for further details see pat(5)):
port controlled by the user
signal or port observed by the user
forcing the value of a register
- pattern sequence
list of values to be forced on inputs or to be checked as the result of the simulation on outputs
The input-output list is represented by two data structures: pagrp contains informations relative to actual or virtual arraies and, paiol represents a one-bit input-output. Both are simply chained lists. However, to make a more efficient access to paiol structures, they are organized in a table. So, a paiol can also be addressed using an offset.
Another group of data structures is used to represent a pattern. paini contains values to be forced into registers before processing the pattern. An event (a modification of value) on an input-output is given by a paevt structure. A third structure, papat gathers all information concerning a pattern (including initializations - paini - and events - paevt).
A pacom is used to represent a comment.
Finally, all these structures can be accessed through a unique structure, paseq, designed to represent a complete sequence of simulation patterns.
In order to use PAT, concepts on use of libraries are needed. And, since these libraries are under development, the code is subject to change.
To enable work, a static version of each library is always present for the user. Libraries and header files are suffixed by a number (the library's version). The programmer can prefer to work with an earlier version of a library rather than the most recent one. However, it is recommended to adapt softwares to libraries as soon as possible in order to spotlight potential compatibility problems before old libraries are removed.
A makefile is necessary for all PAT applications. This is required because any soft must be easily recompilable, and knowing the needed libraries for a PAT based program is not an easy matter. It can't be achieved by an other way, so do use makefile.
In terms of software organization, PAT is splitted into two libraries for the basic functions, a header file for structures and variable declarations and, up to now, one parser-driver library for pat format.
It is to be noticed that library's version number is incremented when, at least, one function composing the library has been modified. Therefore, PAT libraries may evolve independently form each other.
Here follows the list of the libraries and their contents.
pat_addpaseq, pat_addpagrp, pat_addpaiol, pat_addpapat, pat_addpaevt, pat_addpaini, pat_addpacom, pat_frepaini, pat_frepaiol, pat_frepacom, pat_frepapat, pat_frepaevt,
pat_lodpaseq, pat_savpaseq, pat_drvpat, pat_prspat
The libraries are organized in such a way that no cycle can appear through calls. It means that, if in a given library a function calls another function in another library, then none of the functions of this one does call any function that belongs to the first library.
In order to use PAT libraries, one needs something like this in its makefile:
HEADER = -I/labo/include
LIB = -L/labo/lib -lPat106 -lPhl106 -lPpt106 -lMut315