is there any way to keep the uninitialized variable in parfor and recall it outside

Question

Amirah Algethami 2024-4-29

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2112876-is-there-any-way-to-keep-the-uninitialized-variable-in-parfor-and-recall-it-outside

评论： Manikanta Aditya 2024-5-1

采纳的回答： Manikanta Aditya

在 MATLAB Online 中打开

Hi

This is how it look like my code:

b = [];
parfor i = 1:n
    for j=1:m
        if b && some_condition(i)
            t=do_something(i);
            b = [b;t];
        end
    end...
end

I want to save b matrix for every itration , please help

Best

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Amirah Algethami 2024-4-29

Thank you very much I got it @Manikanta Aditya

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Manikanta Aditya 2024-4-29

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2112876-is-there-any-way-to-keep-the-uninitialized-variable-in-parfor-and-recall-it-outside#answer_1449516

在 MATLAB Online 中打开

Hi @Amirah Algethami

In MATLAB, when using a parfor loop (parallel for loop), you need to understand that each iteration of the loop is executed independently and potentially on different workers. This means that variables like b in your example cannot be directly modified in the manner you're attempting because it breaks the independence of each iteration.

However, you can collect results from each iteration in a way that is compatible with parfor.

n = 10; % Example value for n
m = 5;  % Example value for m
% Preallocate a cell array for the results
results = cell(n, 1);
parfor i = 1:n
    temp_b = []; % Temporary variable for this iteration
    for j = 1:m
        if ~isempty(temp_b) && some_condition(i)
            t = do_something(i);
            temp_b = [temp_b; t];
        end
    end
    results{i} = temp_b; % Store the result of this iteration
end
% Concatenate all the results after the parfor loop
b = vertcat(results{:});

Hope this helps!

6 个评论
显示 4更早的评论隐藏 4更早的评论

Manikanta Aditya 2024-4-30

编辑：Manikanta Aditya 2024-4-30

在 MATLAB Online 中打开

In MATLAB, when using a parfor loop, each iteration of the loop operates independently, and the results from each iteration need to be stored in a way that does not depend on the results of other iterations. This is why a cell array is a good choice for collecting results, as each cell can independently hold the result from each iteration, regardless of size or type. However, your clarification question touches on a scenario where the number of results you want to store may be less than the number of iterations of the loop, due to possibly skipping some iterations based on certain conditions.

If the actual number of results is less than the number of loop iterations (n), you have a few options to handle this situation:

You can still initialize your cell array with n elements (cell(n, 1)) and only fill in the cells for which you have results. After the loop, you can remove the empty cells. This method works well if you don't know in advance how many iterations will produce results.

results = cell(n, 1); % Preallocation
parfor i = 1:n
    % Your loop code here
    % Only assign to results{i} if you have valid results
end
% Remove empty cells after the loop
results = results(~cellfun('isempty', results));

Another approach is to collect results in a temporary variable within each iteration and then use a parfor-compatible method to concatenate these results. This approach might involve using a more complex structure or additional indexing logic to ensure that results are collected correctly.

Thanks!

Manikanta Aditya 2024-4-30

在 MATLAB Online 中打开

Based on your scenario and the issue you're facing with the parfor loop and conditional statement inside it, it seems you want to collect temp_b for each iteration of i when certain conditions are met. The core issue you're encountering is related to how to properly collect and concatenate these results, especially under specific conditions within the nested loop.

Given the constraints and the need to execute conditionally within the parfor loop, let's refine the approach to ensure temp_b is collected correctly based on your condition and then concatenated after the loop.

n = 10; % Example value for n
m = 5;  % Example value for m
% Preallocate a cell array for the results
results = cell(n, 1);
parfor i = 1:n
    temp_b = []; % Temporary variable for this iteration
    for j = 1:m
        % Assuming some_condition(i) is your actual condition
        % and you want to collect results based on this condition.
        if some_condition(i)
            t = do_something(i);
            % Directly append to temp_b if condition is met
            temp_b = [temp_b; t];
        end
    end
    % Store temp_b for this iteration if any condition was met and temp_b was updated
    if ~isempty(temp_b)
        results{i} = temp_b;
    end
end
% Concatenate all the non-empty results after the parfor loop
% Filter out empty cells first
nonEmptyResults = results(~cellfun('isempty', results));
b = vertcat(nonEmptyResults{:});