Source Code for Selected GC Benchmarks

These benchmarks are derived from the benchmarks that Lars Hansen used for his thesis on Older-first garbage collection in practice . That thesis contains storage profiles and detailed discussion for most of these benchmarks.


Apart from a run-benchmark procedure, most of these benchmarks are intended to run in any R5RS-conforming implementation of Scheme. (The softscheme benchmark is an exception.) Please report any portability problems that you encounter.

To find the main entry point(s) of a benchmark, search for calls to run-benchmark, which calculates and reports the run time and any other relevant statistics. The run-benchmark procedure is implementation-dependent; see run-benchmark.chez for an example of how to write it.

GC Benchmarks

To obtain a gzip'ed tar file containing source code for all of the benchmarks described below, click here .

Description: A null benchmark for testing the implementation-specific run-benchmark procedure.
Description: Fritz Henglein's algorithm for dynamic type inference. Three inputs are available for this benchmark. In increasing order of size, they are:
  1. dynamic.sch, the code for the benchmark itself
  2. dynamic-input-small.sch, which is macro-expanded code for the Twobit compiler
  3. dynamic-input-large.sch, which is macro-expanded code for the Twobit compiler and SPARC assembler.
Description: Earley's context-free parsing algorithm, as implemented by Marc Feeley, given a simple ambiguous grammar, generating all the parse trees for a short input.
Description: A synthetic benchmark originally written in Java by John Ellis, Pete Kovac, and Hans Boehm.
Description: Enumeration of directed graphs, possibly written by Jim Miller. Makes heavy use of higher-order procedures.
Description: Enumeration of lattices of monotone maps between lattices, obtained from Andrew Wright, possibly written by Wright or Jim Miller.
Description: Bob Boyer's theorem proving benchmark, with a scaling parameter suggested by Boyer, some bug fixes noted by Henry Baker and ourselves, and rewritten to use a more reasonable representation for the database (with constant-time lookups) instead of property lists (which gave linear-time lookups for the most widely distributed form of the boyer benchmark in Scheme).
Description: Marc Feeley et al's Pseudoknot benchmark, revised to use R5RS macros instead of implementation-dependent macro systems.
Description: Zaks's algorithm for generating a list of permutations. This is a diabolical garbage collection benchmark with four parameters M, N, K, and L. The MpermNKL benchmark allocates a queue of size K and then performs M iterations of the following operation: Fill the queue with individually computed copies of all permutations of a list of size N, and then remove the oldest L copies from the queue. At the end of each iteration, the oldest L/K of the live storage becomes garbage, and object lifetimes are distributed uniformly between two volumes that depend upon N, K, and L.
Description: This is the nboyer benchmark with a small but effective tweak: shared consing as implemented by Henry Baker.
Description: Andrew's Wright's soft type inference for Scheme. This software is covered by the GNU GENERAL PUBLIC LICENSE. This benchmark is nonportable because it uses a low-level syntax definition to define a non-hygienic defmacro construct. Requires an input file; the inputs used with the dynamic and twobit benchmarks should be suitable.
Description: A portable version of the Twobit Scheme compiler and Larceny's SPARC assembler, written by Will Clinger and Lars Hansen. Two input files are provided:
  1. twobit-input-short.sch, the nucleic2 benchmark stripped of implementation-specific alternatives to its R4RS macros
  2. twobit.sch, the twobit benchmark itself
Description: The twobit benchmark without the SPARC assembler.

Last updated 4 April 2001.