Ah, the benchmark wars. I found a news post of mine from June 1997
and made it available at
http://www.math.purdue.edu/~lucier/old.benchmarks
Since then Stalin and Gambit-C have both been improved, but I don't have time
to redo the benchmarks. The main table is:
Benchmark Stalin Gambit-C C
Compile Run Size Compile Run Size Compile Run Size
fft 181 1.9 8052 3.4 2.5 21496 0.3 1.0 5728
fibfp 94 3.1 6852 3.1 5.7 21392 0.2 3.3 5152
mbrot 108 0.5 7564 3.2 6.2 21220 0.2 0.5 5552
nucleic 51744 *** 298800 57.2 9.2 222192 238.4 4.4 130808
pnpoly 108 0.4 7612 3.4 0.6 22088 0.6 0.3 6336
simplex 533 3.7 16798 6.8 5.2 31084 0.8 1.6 9652
sumfp 91 4.7 6744 2.9 40.7 20276 0.2 2.9 5088
***: runtime aborted with message "Segmentation Fault"
The times are runtimes in seconds, the sizes are in bytes. To get the
details you'll have to download the old post. Using Stalin does not
match my style of programming; I rely on separate compilation and dynamic
loading a lot.
On a related note, in a grant proposal from July 1998 I noted that on
the pseudo-knot benchmark [1], (the then-current version of) Gambit-C
produced faster code than (the then-current version of) Sisal; at the
time [1] was written, Sisal was the fastest functional programming system
for scientific computing.
Brad
\ref\no 1
\by Hartel, P. H., Feeley, M., et al.
\paper Benchmarking implementations of functional languages with
``Pseudoknot", a float-intensive benchmark
\jour J. Functional Programming
\vol 6
\yr 1996
\pages 621--655
\endref
--------------------- Object Oriented Numerics List --------------------------
* To subscribe/unsubscribe: use the handy web form at
http://oonumerics.org/oon/
* If this doesn't work, please send a note to owner-oon-list@oonumerics.org
This archive was generated by hypermail 2b29 : Wed Feb 20 2002 - 03:20:10 EST