R help - Jul 2007 - code optimization tips

Hi,

Being new to R I'm asking for some advice on how to optimize the  
performance of the following piece of code:

> alpha_c <- function(lambda=600e-9,alpha_s=1e-14,N=400,spacing=1e-7){
>
> k<-2*pi/lambda
> ri<-c(0,0) # particle at the origin
> x<-c(-N:N)
> positions <- function(N) {
>    reps <- 2*N+1
>    matrix(c(rep(-N:N, each = reps), rep(-N:N, times = reps)),
>           nrow = 2, byrow = TRUE)
> }
> rj<-positions(N)*spacing # all positions in the 2N x 2N array
> rj<-rj[1:2,-((dim(rj)[2]-1)/2+1)] # remove rj=(0,0)
>
> mod<-function(x){sqrt(x[1]^2+x[2]^2)} # modulus
>
> sij <-function(rj){
> rij=mod(rj-ri)
> cos_ij=rj[1]/rij
> sin_ij=rj[2]/rij
>
> A<-(1-1i*k*rij)*(3*cos_ij^2-1)*exp(1i*k*rij)/(rij^3)
> B<-k^2*sin_ij^2*exp(1i*k*rij)/rij
>
> sij<-A+B
> }
>
> s_ij<-apply(rj,2,sij)
> S<-sum(s_ij)
> alpha_s/(1-alpha_s*S)
> }
> alpha_c()

This function is to be called for a few tens of values of lambda in a  
'for' loop, and possibly a couple of different N and spacing (their  
magnitude is typically around the default one).

This can be a bit slow ??? not that I would expect otherwise --- and  
I wonder if there is something I could do to optimize it (vectorize  
with respect to the lambda parameter?, change the units of the  
problem to deal with numbers closer to unity?,...)

Best regards,

baptiste

First question is why are you defining the functions within the main
function each time?  Why don't you define them once outside?

On 7/23/07, baptiste Augui? <ba208 at exeter.ac.uk>
wrote:
> Hi,
>
> Being new to R I'm asking for some advice on how to optimize the
> performance of the following piece of code:
>
>
> > alpha_c <-
function(lambda=600e-9,alpha_s=1e-14,N=400,spacing=1e-7){
> >
> > k<-2*pi/lambda
> > ri<-c(0,0) # particle at the origin
> > x<-c(-N:N)
> > positions <- function(N) {
> >    reps <- 2*N+1
> >    matrix(c(rep(-N:N, each = reps), rep(-N:N, times = reps)),
> >           nrow = 2, byrow = TRUE)
> > }
> > rj<-positions(N)*spacing # all positions in the 2N x 2N array
> > rj<-rj[1:2,-((dim(rj)[2]-1)/2+1)] # remove rj=(0,0)
> >
> > mod<-function(x){sqrt(x[1]^2+x[2]^2)} # modulus
> >
> > sij <-function(rj){
> > rij=mod(rj-ri)
> > cos_ij=rj[1]/rij
> > sin_ij=rj[2]/rij
> >
> > A<-(1-1i*k*rij)*(3*cos_ij^2-1)*exp(1i*k*rij)/(rij^3)
> > B<-k^2*sin_ij^2*exp(1i*k*rij)/rij
> >
> > sij<-A+B
> > }
> >
> > s_ij<-apply(rj,2,sij)
> > S<-sum(s_ij)
> > alpha_s/(1-alpha_s*S)
> > }
> > alpha_c()
>
>
> This function is to be called for a few tens of values of lambda in a
> 'for' loop, and possibly a couple of different N and spacing (their
> magnitude is typically around the default one).
>
> This can be a bit slow ??? not that I would expect otherwise --- and
> I wonder if there is something I could do to optimize it (vectorize
> with respect to the lambda parameter?, change the units of the
> problem to deal with numbers closer to unity?,...)
>
> Best regards,
>
> baptiste
>
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>

-- 
Jim Holtman
Cincinnati, OH
+1 513 646 9390

What is the problem you are trying to solve?