Notes on Efficiency

Here we indicate some key features that can significantly affect efficiency. Some of them are unsound in general, and hence extreme care should be taken when using them. Novice programmers may (and probably should) skip this section entirely.

• Indexing  
  CLP( ) employs first argument indexing for constructed functor terms as well as real numbers. Using indexing can result in significant speedups.
• Tail recursion  
  Last call optimization is employed, and hence procedures that are tail-recursive will not increase local stack usage.
• Logical disjunction ``;/2'' and if-then-else ``->/2''    
  These are implemented at the meta-level and hence are not particularly efficient.
• Dynamic code  
  Dynamic code is slower than static code and is also less deterministic. Cuts can be used to make it more deterministic. Also, since dynamic code is compiled, asserting large terms may not be very fast.
• Garbage collection  
  Not implemented as yet.
• Implicit equalities  
  The solving of inequalities that imply some implicit equations can be controlled using implicit/0, noimplicit/0, partial_implicit/0 (see Section 5.1.7).
• Asserting a rule   
  The predicate assert/1 involves incorporating the constraints that relate the variables in that rule (see Section 3.4.2). This is less efficient than if the constraints were not taken into account. The fassert family of special predicates (``fast assert'') performs assertion without incorporating arithmetic constraints (see Section 5.1.7), as in PROLOG.