Berkeley Pascal is designed for interactive instructional use and runs on the PDP 11 family of computers. A version of pxp for the VAX 11/780 is nearly complete. The Pascal execution profiler pxp can be used to generate restructured source program listings which are annotated with counts of the number of times each statement in the profiled parts of the program was executed for a particular run of the program. A number of options of pxp provide easy control over the parts of the program to be profiled. pxp can also be used to produce restructured program listings, i.e. as a pretty-printer.

The PXP Implementation Notes describe the design decisions which lead to the present form of pxp and give details of the internal organization. The notes conclude with thoughts on similar facilities for future systems.

Table of Contents

• Front Page


• 1. Design Considerations
  • 1.1. Goals
  • 1.2. Placement of statement counters
    • 1.2.1. Basic considerations
    • 1.2.2 Goto statements
    • 1.2.3. Counter placement rules
    • 1.2.4. Partial program counting
  • 1.3. Producing the profile
  • 1.4. Abnormal termination
  • 1.5. Formatting of the program
    • 1.5.1. Pretty printers
    • 1.5.2. Comments and output compression
    • 1.5.3. Bushy if-then-elses
    • 1.5.4. Begin-end pairs and well-bracketing
    • 1.5.5. Indenting of structured levels
    • 1.5.6. Expressions
    • 1.5.7. Procedures and functions
    • 1.5.8. Case statements
    • 1.5.9. Labels and miscellania
    • 1.5.10. Summary
  • 1.6. Profile Data Presentation
    • 1.6.1. Count interpretation
    • 1.6.2. Data compression


• 2. Implementation
  • 2.1. External data structures
    • 2.1.1. Format of the profile data file
    • 2.1.2. Core data files
    • 2.1.3. Shared text programs
    • 2.1.4. Extraction methods
  • 2.2. Internal data structures
    • 2.2.1. External data copies
    • 2.2.2. Profile cluster data structures
    • 2.2.3. The profile cluster primitives
    • 2.2.4. Profile-printing cluster interface
    • 2.2.5. The printing cluster
    • 2.2.6. Printing cluster primitives
  • 2.3. Producing the profile
    • 2.3.1. Main routine and option processing
    • 2.3.2. Procedure and function level processing
    • 2.3.3. Statement processing
    • 2.3.4. If-then-else's
    • 2.3.5. Expression processing
  • 2.4. Multiple file processing


• 3. Automatic program formatting
  • 3.1. Motivation
  • 3.2. Implementation
    • 3.2.1. The kinds of comments
  • 3.3. Examples of comment types
    • 3.3.1. Data structures (overview)
    • 3.3.2. Comment processing (overview)


• 4. Conclusions
  • 4.1. Design
  • 4.2. Implementation
  • 4.3. Profiling
  • 4.4. Reformatting
  • 4.5. Future systems


• References