Asked by hung lequang
on 4 Oct 2012

*Hello everyone,

Can anyone show me how I can avoid following for loops, so that my run can be much faster? Any help is highly appreciated. Thanks!*

c2=0.7; c1=0.3; x1=rand(1,100)*10; x3=rand(1,100)*10; x3p=rand(1,100)*10; x=[0.0000,0.2560,0.4980,0.7590,1.0000,1.2400,1.5000,1.7600,2.0100]; y=[0.3,0.133,0.0777,0.0695,0.0920,0.1078,0.0829,0.0807,0.0936]; for ix1=1:length(x1) for ix3=1:length(x3) for ix3p=1:length(x3p) r= sqrt(x1(ix1)^2 + (x3(ix3)-x3p(ix3p))^2); if r <= 10 g = fit(x',y','splineinterp'); Prs_22x(ix1,ix3,ix3p)= c2-c1+g(r/R); else Prs_22x(ix1,ix3,ix3p)= c2-c1+c1^2; end end end end

*No products are associated with this question.*

Answer by Matt Fig
on 4 Oct 2012

Edited by Matt Fig
on 4 Oct 2012

Accepted answer

Just taking this outside all loops will **GREATLY** speed it up.

g = fit(x',y','splineinterp');

Beyond that, you could do this (I assume R is a scalar):

P = bsxfun(@(x,y) (x-y),x3,reshape(x3p,1,1,numel(x3p))); P = bsxfun(@(x,y) sqrt(x.^2+y.^2),x1.',P); I = P<=10; P(I) = c2-c1+g(P(I)/R); P(~I) = c2-c1+c1^2;

Opportunities for recent engineering grads.

## 0 Comments