vector =: 312 [EMAIL PROTECTED] 0 NB. make a 256 element vector
of random #s
matrix =: 312 256 [EMAIL PROTECTED] 0 NB. make a 256 column by 312
row matrix of random #s
doit =: 3 : 0
for_i. i. y do.
output =: vector +/ .* matrix
end.
)
echo 'using function'
echo ((6!:2) 'doit 1000')
NB. ~0.259691 secs
echo 'not using function'
echo 1000 * ( 1000 (6!:2) 'output =: vector +/ .* matrix')
NB. ~0.264517 secs
Doing the same thing is in matlab is ~0.6 secs
NB. Doing the same thing using tvmet (Tiny Vector Matrix
library using Expression Templates for C++) is ~0.36 secs
NB. The reason I am going through this is as follows:
NB. I have about 60 seconds of data that needs to be
processed. Each second has about 5000
NB. frames data.
If I do the processing in C++ using tvmet, it will take
about 1 minute 48 seconds.
If I do the processing in J, it will take about 1 minute 19
seconds.
So to process 1 second of data takes more than 1 minute of
processing time. I have many hours
worth of data to process.
Would moving to a 64 bit quad core processor be a good idea?
Thanks for all the good comments!
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
