42 lines
1.6 KiB
Plaintext
42 lines
1.6 KiB
Plaintext
This directory contains a set of R6RS benchmarks. Some were
|
|
originally collected by Richard Gabriel, while others were
|
|
collected or written by Marc Feeley and Will Clinger.
|
|
Abdulaziz Ghuloum converted about 50 of these benchmarks to
|
|
R6RS libraries. R6RS libraries are non-portable by design,
|
|
however, so Clinger rewrote the benchmarks as R6RS top-level
|
|
programs and added a script for running the benchmarks on
|
|
Unix systems. Clinger also added new benchmarks for R6RS.
|
|
|
|
Files and directories:
|
|
|
|
* bench : a shell script for running benchmarks
|
|
* src : contains R6RS code for the benchmarks
|
|
* inputs : contains inputs for the benchmarks
|
|
* outputs : will hold the outputs of some benchmarks
|
|
|
|
For succinct instructions on running benchmarks, run the
|
|
bench script without any arguments:
|
|
|
|
% ./bench
|
|
|
|
The bench script creates a /tmp/larcenous directory to hold
|
|
the source code constructed for the benchmarks.
|
|
|
|
The bench script appends its results to files with names
|
|
like results.Ikarus, results.Larceny, and so forth.
|
|
|
|
Will
|
|
|
|
================================================================
|
|
|
|
NOTE:
|
|
The nbody, trav1, and trav2 benchmarks have been dropped because
|
|
the depend upon a non-portable order of evaluation. The sumloop
|
|
benchmark has been dropped because it was essentially the same
|
|
as the sum benchmark. The boyer benchmark has been replaced by
|
|
the nboyer and sboyer benchmarks, which are fundamentally better
|
|
benchmarks, with fewer bugs, and scalable. The gcold benchmark
|
|
has been dropped temporarily because its initialization phase is
|
|
so long compared to the benchmark phase, and the R6RS provides
|
|
no portable way to time those phases separately.
|