How to Reduce Computation Time

Question

Tez 2017-8-22

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/353582-how-to-reduce-computation-time

编辑： per isakson 2017-8-23

Hello,

I am new to matlab. I am analyzing satellite data.each variable is about 400x31000 in dimension. After selecting data in those variable which i want i end up with a matrix of same dimension where data is in bunches , that is there will be data in some columns. i want the mean of these bunches separately.I wrote a code to do it but it takes more than 3 hours to run.I need help to reduce the time.

q=1;
nover1=NaN(size(nover));
nover2=NaN(size(nover));
nover3=NaN(size(nover));
meantime2=NaN(size(mean_ctime));
date2=NaN(size(date));
mlat1=NaN(size(mean_lat));
mlat2=NaN(size(mean_lat));
mlon1=NaN(size(mean_lon));
mlon2=NaN(size(mean_lon));
for i=1:size(alt,1);
    for j=1:size(alt,2);
              if alt(i,j)>80 && alt(i,j)<100;
                  if mean_lat(1,j)>0 && mean_lat(1,j)<20;
                            mlat1(1,j)=mean_lat(1,j);
                            mlon1(1,j)=mean_clon(1,j);
                            nover1(i,j)=nover(i,j);
                            meantime1(1,j)=mean_ctime(1,j);
                            date1(1,j)=date(1,j);
                  end
              end
      end
end
for j=1:size(alt,2);
               if mean_clon(1,j)>60 && mean_clon(1,j)<120;
                   mlat2(1,j)=mlat1(1,j);
                   mlon2(1,j)=mlon1(1,j);
                   nover2(:,j)=nover1(:,j);
                   meantime2(1,j)=meantime1(1,j);
                   date2(1,j)=date1(1,j);
               else
                   nover2(nover2==0)=NaN;
                   meantime2(meantime2==0)=NaN;
                   date2(date2==0)=NaN;
                   nno=nonzeros(nover2);
                   ntime=nonzeros(meantime2);
                   ndate=nonzeros(date2);
                   mno_42(1,q)=nanmean(nno); % the final no ver values
                   mtime_42(1,q)=nanmean(ntime); % the final time values
                   mdate_42(1,q)=nanmean(ndate); % the final date values
                   q=q+1;
                   nover2=NaN(size(nover));
                   meantime2=NaN(size(mean_time));
                   date2=NaN(size(date));
               end
end

2 个评论
显示无隐藏无

KSSV 2017-8-22

What form the satellite data is in?

Tez 2017-8-22

netcdf files

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

per isakson 2017-8-22

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/353582-how-to-reduce-computation-time#answer_278795

编辑：per isakson 2017-8-23

The code analyzer box in the upper right corner tells you that there is potential for improving the code. Hovering over the small bars in the right column will give you hints on how.

"preallocation for speed" see: Preallocation. That alone will work wonders on the speed of your code.

If that isn't enough see Profile to Improve Performance

Without being able to run the code it is difficult to be more specific than this.