close all; clearvars; clc; delete(gcp('nocreate'));

%% Parameters: Parallel computation and gpu
global comput
comput.Paral = 1; % |1| use parallel computation (required Parallel Computing Toolbox).
comput.gpu = 1; %|1| use GPU device (required Parallel Computing Toolbox).

%% Initialization: parallel processing
if(license('test', 'Distrib_Computing_Toolbox') == 0)
    comput.gpu = 0;
    comput.Paral = 0;
    fprintf('Distrib_Computing_Toolbox not available: GPU and multicore capabilities can not be used\n');
end
% GPU device
comput.gpuHandle = [];
if(comput.gpu)
    for ii = 1:gpuDeviceCount
        g = gpuDevice(ii);
        if(str2double(g.ComputeCapability) >= 3)
           comput.gpuHandle = g;
           fprintf('Using GPU\n');
           break;
        end
    end
    fprintf('Not using GPU\n');
else
    fprintf('Not using GPU\n');
end
% Cluster processing
if(comput.Paral)
    parpool;
    p = gcp('nocreate'); % If no pool, do not create new one.
    fprintf('Using %d CPU\n', p.NumWorkers);
%     parpool('local', 4);
else
   fprintf('Using 1 CPU\n');
end

%% GPU processing: matrix multiplication
if(~isempty(comput.gpuHandle))
    try
        a = randn(100, 100, 'single', 'gpuArray');
        % Alternatively
        % a = randn(100, 100, 'single');
        % a = gpuArray(a);
        b = gather(a*a);
    catch
        warning('GPU computation is not possible, using CPU computation.\n');
        comput.gpuHandle = [];
        a = randn(100, 100, 'single');
        b = a*a;
    end
else
    a = randn(100, 100, 'single');
    b = a*a;
end

%% Parallel loop processing: : matrix multiplication.
b = cell(10,1);
if(comput.Paral == 1)
  parfor i = 1:10
    a = randn(i, i, 'single');
    b{i} = a*a;
  end
else
  for i = 1:10
    a = randn(i, i, 'single');
    b{i} = a*a;
  end
end