How can I accelerate this function with GPU arrayfun

Tong Wu

2018 8 8

0 个回答

12 次查看（30 天）

0 个投票

Hi,

I want to accelerate a function like below with GPU. The real code is too complex so I put a simplified version here.

The main problem of this code is at the 5th line of nestedFcn: Indexing is not supported.

And the 1st line of nestedFcn : The size and shape for all variables must be the same. Variable 'test2' differs from 'test'.

Is there any way to do that?

 % code
function gtest
Np=10;
test = zeros(Np);
test(1,1)=1;
E = rand(1,Np);
gpuVector = gpuArray.colon(1, Np);
    function out = nestedFcn(k)
        test2 = test*acos(E(k));
        temp = rdivide(1,(rand()-test2));
        % need a diag function below
        for i_d=1:Np
            out(i_d,1) = temp(i_d,i_d);
        end
    end
% Call arrayfun.
arrayfun(@nestedFcn, gpuVector)
end