possible solution to speed up with power function?

Is there any solution to speed up this code? The power function is too expensive. Any help is much appreciated.

tic
N = 1000000;
Linput = 500;

t = 1:Linput;
mTT = rand(N,1);
alf = rand(N,1);

beta = mTT./alf;
faktor = 1./(beta.^alf.*gamma(alf));
A =  t.^(alf-1);
B = -t./beta ;                 
C = faktor.*A.*exp(B); 
toc
Elapsed time is 104.574729 seconds.

Answer 1

[Deepak]

Hi Le Duy Nguyen,

To my understanding, you want to speed up the code provided by you that includes mathematical operations.

To optimize the code, Element wise operations such as divide, multiply, power can be performed by using the inbuilt method “bsxfun” of MATLAB.

“bsxfun” performs element-wise operations to two arrays, in optimized manner.

Also, we can precompute the “gamma(alf)” values to use them to calculate the “factor” which will speed up the process.

Attaching the documentation of “bsxfun” for reference – https://www.mathworks.com/help/matlab/ref/bsxfun.html

Here is the updated code snippet –

tic
N = 1000000;
Linput = 500;
t = 1:Linput;
mTT = rand(N,1);
alf = rand(N,1);
beta = mTT ./ alf;

gamma_alf = gamma(alf);

factor = 1 ./ (beta.^alf .* gamma_alf);
A = bsxfun(@power, t, alf-1);
B = -bsxfun(@rdivide, t, beta);

C = bsxfun(@times, factor, A) .* exp(B);
toc