Parfor cannot serialize data which is too large

I have a parfor loop which calculates eigenvectors as a function of frequency, and stores all of them for post-processing. The size of the eigenvectors is 924*200 (Temp_zvAi for instance), and the size of zuxAi is 924*200*1311. The eigenvectors are complex, and zuxAi is around 3.8GB. Parfor returns the serialization error. How to get around this please? The code looks like below:
parfor fcount=1:Numsam
[Amat,Bmat]=functionof(fcount);
[Temp_zkA,Temp_zvAi]=eigenSolver (jn1,Amat,Bmat);
Temp_zuxAi=Temp_zvAi(1:nA,1:jn1);
zuxAi(:,:,fcount)=Temp_zuxAi;
Temp_zuyAi=Temp_zvAi(nA+1:2*nA,1:jn1);
zuyAi(:,:,fcount)=Temp_zuyAi;
Temp_zuzAi=Temp_zvAi(2*nA+1:3*nA,1:jn1);
zuzAi(:,:,fcount)=Temp_zuzAi;
end

2 个评论

Your code looks like an in complete example. In particular, none of the Temp_ quantities show any dependence on fcount. For all we can see, they can be pre-computed before the loop and all zu_xyz_Ai(:,:,fcount) will be copies of each other.
Matt, thanks for your answer. The code only shows the places where errors are caught. Temp_ quantities depend on fcount, as Amat and Bmat are matrices dependant on fcount. The frequency dependence is irrelevant to the current problem so I deleted them.
I added one line to the code to clarify this.

请先登录,再进行评论。

 采纳的回答

The limit on PARFOR data transfers was removed in R2013a. Please upgrade if you can.

更多回答(0 个)

类别

帮助中心File Exchange 中查找有关 Parallel for-Loops (parfor) 的更多信息

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by