Parallelization of matrix definiton

Question

Sasha 2014-2-14

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/116137-parallelization-of-matrix-definiton

关闭： MATLAB Answer Bot 2021-8-20

In my project I deal with matrices using up almost all the memory available. Logically the matrix is divided into blocks which can be computed independently from one another. Straight forward use of parfor as follows

parfor a = 1:N,
     [M1] = func(a,par1, par2, par3);
     N(i1(a):i2(a), i3(a):i4(a)) = M1;
end

does not work (the code is an example to illustrate the idea). Is there a way to establish parallel computing of these blocks?

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

此问题已关闭。

Answer 1

Jill Reese 2014-2-14

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/116137-parallelization-of-matrix-definiton#answer_124371

Without more context for your problem I cannot be sure, but you may find using distributed arrays inside an spmd block to be more useful than parfor in this case.

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Sasha 2014-2-14

Thank you for your hint about spmd. If I get it right, matlab can do parallel computing inside the spmd statement if one uses matlab operations with parallel computing support build in. In my case I want to compute 'func' for different a in parallel. Can it be done with spmd?

Context: func produces a matrix ~100x100 elements which is then inserted into a larger matrix N with ~20000x20000 elements. N is a linear system of equations.