For loop with a big matrix of 800 MB

Hi all,

I am searching through a big matrix file ( nearly 850 MB) and tried to loop through that matrix but it take a massive amount of time. I wonder if are there any method to run a simple for loop in this "big data" matrix. My code is at belown

clearvars;
load('landorocean_250_W.mat')
tA_world = tall(A_world);
tx_world = tall(x_world);
ty_world = tall(y_world);
[m_1,n_1]=size(A_world);
for i=1:m_1
    for j=1:n_1
        if  gather(tA_world(i,j) == -32768)
            island_250m_1(i,j)=0;
        else
            island_250m_1(i,j)=1;
        end
    end
end
       

Thank you for your time,

Sincerely,

[Jan]

800MB is not really big. Such arrays match in the RAM of modern computers easily.

Loading data directly into the workspace by load without catching the outputs, slows down Matlab, because this impedes the JIT acceleration. Try this:

function test
Data = load('landorocean_250_W.mat')

tA_world = Data.A_world;
tx_world = Data.x_world;
ty_world = Data.y_world;
[m_1,n_1]=size(A_world);
island_250m_1 = zeros(m_1, n_1);  % Pre-allocation!!!
for i=1:m_1
    for j=1:n_1
        if  tA_world(i,j) == -32768
            island_250m_1(i,j)=0;
        else
            island_250m_1(i,j)=1;
        end
    end
end

% Or easier without a loop:
% island_250m_1 = (tA_world ~= -32768);

Do not let arrays grow iteratively, because this is extremely expensive. Pre-allocate the create output arrays instead.

[Trung Ngo]

Hi Jan, thank you for your reply.

I tried preallocation matrix before but it resulted in this error

%Requested  array exceeds maximum array size preference. Creation of arrays greater than this limit may take a long time and cause MATLAB to become unresponsive.

[Jan]

@Trung Ngo: In the opriginal question you wrote ">800mb". Of course this might mean 800 TeraByte also. Then only a tall array can catch this. Please post the maximum size and how much free RAM you have. It matters of ">800MB" means 850MB on a computer with 8GB of RAM, or if it means 800GB on a machine with 2 GB RAM.

[Trung Ngo]

@Jan: Sorry for being unclear. It is 850MB matrix under a computer with 16GB RAM.

Answer 1

[Guillaume]

Using gather as you have done on each individual element of the tall array completely nullifies the advantage of tall arrays. You may as well not have used tall arrays, it would have been faster.

The loop was not needed in the first place:

island_250m_1 = tA_world == -32768;  %creates a tall array the same size as tA_world
%now you can call gather if you want to convert that tall array into a concrete array
island_250m_1 = gather(island_250m_1);

[Trung Ngo]

Thank you so much for your help !