CS 4448 - Spring 1998
Object-Oriented Programming and Design
Discussion 7.2.1
by
Charles Griffis

Garbage Collection in C++: No panacea, but useful
and
Memory allocation and C compatibility
by
Andrew Koenig

Garbage Collection in C++ addresses issues concerning the utilization of garbage collection in C++ and the benefits and pitfalls derived therefrom.  Benefits include: increased programmability and software reliability.  Some concerns raised in the class discussion and in the paper itself include: memory management/system performance and pointer maintenance.

• Garbage collection reduces programming overhead wrt memory allocation and de-allocation.  In general this reduces programming error while increasing software maintainability.

While the benefits to the programmer are obvious, problems associated with garbage collection are generally difficult to discover and/or correct.

• Garbage collection is often implemented differently from one system to another.
  
• Garbage collectors run periodically to reclaim memory resources, often running when memory is demanded most,  thus increasing load and possibly crippling a system with limited resources.
  
• Software which utilizes garbage collection tends to be less efficient than software that doesn't.
  
• Garbage collection maintains a reference count to allocated resources that prevents an object from being deleted when a reference to it is removed and others still exist.  Although this is an important benefit it can prevent a resource from being deleted altogether if the object(s) in question represent a link looping back upon itself or from one object to another.
  

C++:  Memory allocation and C compatibility addresses issues concerning the portability and interchangeability of C's "malloc" and "free" and the C++ methods "new" and "delete".

• In general these functions are not mutually compatible.
  
• C's "malloc" should not be used to allocate memory in C++ for anything but "built-in" types and scalars since it cannot determine the size of user defined objects which require constructors for memory allocation.  The same is true for releasing "non-built-in" types.  C's "delete" cannot determine the size of an object allocated memory through the C++ "new" method.
  
• Great care must be taken by the programmer to ensure portability between code utilizing both languages' memory allocation and release operations.
  

Copyright © University of Colorado. All rights reserved.
Revised: March 5, 1998