parallel processing time is bigger,

Question

Julio 2014-2-21

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/117132-parallel-processing-time-is-bigger

评论： Julio 2014-2-23

I have a code comparing a to process using parfor and for, but the time of the parfor is bigger thans using "for". here below the program:

Count=12000000;
MSB=A(2:2:Count);
LSB=A(1:2:Count);
DiffB=zeros(Count/2,1);
B=DiffB;
tic
for i=2:2:Count
    B(i/2)=A(i)*256 + A(i-1);
end
for j=1:size(B,2)-1
    if((B(j+1)==0) && (B(j)==4095))
        DiffB(j)=1;
    else
        DiffB(j)=B(j+1)-B(j);
    end
end
t1=toc
tic
parfor i = 1:Count/2           
    B(i)=MSB(i)*256 + LSB(i);
end
C=B;
parfor j=1:size(B,2)-1
    if((B(j+1)==0) && (C(j)==4095))
           DiffB(j)=1
    else
        DiffB(j)=B(j+1)-C(j);
    end
end
t2=toc
matlabpool close
disp('Tiempo differencia')
t2-t1

time for "FOR" is 0.6016 sec and the time for "PARFOR" 5.5039 sec can someone can tell me what did I make wrong?

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Julio 2014-2-21

NOte: before execute the code, I execute matlabpool('open',2); and it works

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Matt J 2014-2-21

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/117132-parallel-processing-time-is-bigger#answer_125254

Are you running inside a script or in an mfile function? Try both.

2 个评论
显示无隐藏无

Matt J 2014-2-21

编辑：Matt J 2014-2-21

But aside from that, your compute-to-memory-access ratio looks very small. You are doing very little computation inside the loops for every element of data (LSB(i), MSB(i), etc...) that you are sending to the workers. PARFOR is intended for a situations where you are doing a lot more computational work inside the loops.

Julio 2014-2-23