There is an interesting benchmarks "contest" at
http://shootout.alioth.debian.org/
which includes a broad set of benchmark performance comparisons
applied across a variety of programming languages.
As a stress relief, I've cobbled up a tentative submission for the
"partial-sums" benchmark for the J language -- which currently has no
coverage at all, surprisingly.
Here is the link to the partial-sums benchmark comparison:
http://shootout.alioth.debian.org/gp4/benchmark.php?
test=partialsums&lang=all
I've included my tentative submission below.
Before submitting this script to the contest, I would like to get it
reviewed and possibly improved (it seems to already outperform all
the other systems in the benchmark -- but I've not done a thorough
evaluation either).
I would also invite others to contribute J solutions to the other
benchmarks -- which would be an excellent way to see how J solutions
can be applied to common programming problems.
If you are interested, this FAQ is worth reading:
http://shootout.alioth.debian.org/gp4/faq.php
Thanks
--
Alan
-----
NB. partial-sums
NB.
NB. The Computer Language Benchmarks Game
NB. http://shootout.alioth.debian.org/
NB.
NB. contributed by Alan K. Stebbens <[EMAIL PROTECTED]>
NB. modified by
NB.
NB.
NB. This script implements the partial-sums benchmark,
NB. as defined at:
NB. http://shootout.alioth.debian.org/gp4/benchmark.php?
test=partialsums&lang=all
NB.
NB.
NB. The "correct" output file for this benchmark test is defined
NB. at http://shootout.alioth.debian.org/gp4/iofile.php?
test=partialsums&lang=all&file=output
NB.
NB. I've included the correct output here, for easy reference:
NB.
NB. 3.000000000 (2/3)^k
NB. 314.770573775 k^-0.5
NB. 0.999960002 1/k(k+1)
NB. 30.314520404 Flint Hills 1/( k^3 * sin(k)^2 )
NB. 42.994899946 Cookson Hills 1/( k^3 * cos(k)^2 )
NB. 10.703866769 Harmonic 1/k
NB. 1.644894068 Riemann Zeta 1/(k^2)
NB. 0.693127181 Alternating Harmonic -1^(k+1) / k
NB. 0.785388163 Gregory -1^(k+1) / ( 2k - 1 )
time =: 6!:2
k =: i. 25000 NB. default input (for testing)
sum =: 0j9&":@(+/) NB. 9 digits precision sum
sum1 =: (2%3)&^ NB. (2/3)^k
sum2 =: (_0.5&(^~))@>: NB. k^(-0.5)
sum3 =: (%@(] * >:))@>: NB. 1/(k * (k + 1))
sin =: 1&o.
cos =: 2&o.
sum4 =: %@((]^&3)*(*:@sin))@>: NB. 1/((k^3)*(sin(k)^2))
sum5 =: %@((]^&3)*(*:@cos))@>: NB. 1/((k^3)*(cos(k)^2))
sum6 =: %@>: NB. 1/k
sum7 =: [EMAIL PROTECTED]:@>: NB. 1/(k^2)
sum8 =: ((_1&^@>:)%])@>: NB. -1^(k+1) / k
sum9 =: ((_1&^@>:)%(_1&[EMAIL PROTECTED]:))@>: NB. -1^(k+1) / (2k - 1)
out =: 1!:2
sum_for =: 3 : 0
if. y-:'' do. k=.i.25000
else. k=.i.y end.
s=.,:''
s=.s,(sum sum1 k),' (2/3)^k'
s=.s,(sum sum2 k),' k^-0.5'
s=.s,(sum sum3 k),' 1/k(k+1)'
s=.s,(sum sum4 k),' Flint Hills'
s=.s,(sum sum5 k),' Cookson Hills'
s=.s,(sum sum6 k),' Harmonic'
s=.s,(sum sum7 k),' Riemann Zeta'
s=.s,(sum sum8 k),' Alternating Harmonic'
s=.s,(sum sum9 k),' Gregory'
s NB. Show results
)
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
