Procedure

This directory contains an idiosyncratic collection of benchmark programs. Each benchmark has been executed without input or output, on a sequence (usually 20) of input sizes.

Each benchmark has been run for FISh and perhaps some other languages, especially Ocaml. Ocaml is particularly addressed because among higher-order, strongly-typed languages it is one of the fastest at array programming. Each benchmark has a graph plotting user times against input size. One example of system times, for mapping, is plotted.

FISh programs were first compiled into C using the action %run -q -c. The C code was then transferred to the C directory, where it was compiled using the GNU C compiler gcc version 2.8.0 for sparc-sun-solaris2.6 with standard optimisation, i.e.

gcc -O

The executables were placed in directory Bin and executed, with user, system and real times recorded in times.summary.

Ocaml source code was placed in Source/benchmark.ml whose "main" function benchmark was then used in files test01.ml etc. Executables were produced on the same machine as the FISh code, using optimised Ocaml compiler with unsafe array bounds and maximum in-lining, i.e.

ocamlopt -unsafe -inline 100

The executables were placed in directory Bin and executed, with user, system and real times recorded in times.summary.

Page Last Updated: Friday, 07-Aug-1998 20:51:14 EST

Please feel free to send any comments.