Performance and evaluation of LISP systems
Performance and evaluation of LISP systems
Theoretical Computer Science
The semantics of destructive LISP
The semantics of destructive LISP
Lisp and Symbolic Computation
Is there a use for linear logic?
PEPM '91 Proceedings of the 1991 ACM SIGPLAN symposium on Partial evaluation and semantics-based program manipulation
Proving memory management invariants for a language based on linear logic
LFP '92 Proceedings of the 1992 ACM conference on LISP and functional programming
Lively linear Lisp: “look ma, no garbage!”
ACM SIGPLAN Notices
Computational interpretations of linear logic
Theoretical Computer Science - Special volume of selected papers of the Sixth Workshop on the Mathematical Foundations of Programming Semantics, Kingston, Ont., Canada, May 1990
A method for overlapping and erasure of lists
Communications of the ACM
Mechanisms for compile-time enforcement of security
POPL '83 Proceedings of the 10th ACM SIGACT-SIGPLAN symposium on Principles of programming languages
Proceedings of the 5th ACM Conference on Functional Programming Languages and Computer Architecture
The Boyer benchmark meets linear logic
ACM SIGPLAN Lisp Pointers
Linear logic and permutation stacks—the Forth shall be first
ACM SIGARCH Computer Architecture News - Special issue: panel sessions of the 1991 workshop on multithreaded computers
| Hi-index | 0.00 |
The Gabriel FRPOLY benchmark was rewritten in a "linear" fragment of Common Lisp and is competitive with the standard FRPOLY benchmark code. This linear FRPOLY is considerably more perspicuous than the standard code, while its running time is only 6% longer than that of the standard FRPOLY code. Linear FRPOLY recovers all of its garbage, and its "high water mark" space requirement is very probably smaller than that of the standard code. In the expansion of (x+y+z+1)15, the standard FRPOLY does 48,892 new conses, while the linear FRPOLY does only 4821 new conses---i.e., it does only about 10% of the consing of the standard FRPOLY code.We also tested versions of FRPOLY in which squarings were not used for exponentiation. This non-linear FRPOLY does 38,780 conses, and takes only 59% of the time of the non-linear squaring FRPOLY. The linear non-squaring FRPOLY takes only 62% of the time of the linear squaring FRPOLY, and cuts the new consing to 3988 cells. A slightly slower version cuts the new consing to 2590 cells---only 567 cells (28%) more than are used in the result.