it's possible to velocize this loop?

orig=Sis(i).dailyprof;
xx=orig;
yy=Sis(i).Ntradess;
x1=find(xx);


last=-1;
for i=numel(x1):-1:1
  if yy(x1(i))~=last
      last=yy(x1(i));
      xx(x1(i))
      xx(x1(i))=xx(x1(i))-k;
      xx(x1(i))
  end     
end    

class(xx) double
class(YY) double
%%CHECK!
%k=[orig xx yy];

[ProblemSolver]

Please read this and ask your question again properly:

https://www.mathworks.com/matlabcentral/answers/6200-tutorial-how-to-ask-a-question-on-answers-and-get-a-fast-answer

Answer 1

[Bruno Luong]

@piero Your question/code is awfully asked: no comment, no description, cannot run as it is, debug code left over, etc....

load('matlab_orig.mat')
load('matlab_yy.mat')
xx = orig;
k = 100;

x1 = find(xx);
last=-1;
for i=numel(x1):-1:1
  if yy(x1(i))~=last
      last=yy(x1(i));
      xx(x1(i))=xx(x1(i))-k;
  end     
end

xx2 = orig;
x1 = find(xx2);
yyx1 = yy(x1);
last = x1([diff(yyx1(:))~=0; true]);
xx2(last) = xx2(last) - k;

isequal(xx, xx2)

ans = logical
   1

[piero]

always top! Thank you

Answer 2

[ProblemSolver]

However, to the understanding what you are looking for is VECTORIZING the loop:

First, you should change 'find to logical indexing:

x1 = logical(orig);

Then, pre-allocate xx and yy to avoid repeated memory allocations:

xx = zeros(size(orig));
yy = zeros(size(orig));

The MAIN part I guess what you are looking for:

last = [0; yy(x1(1:end-1))]; 
mask = [yy(x1) ~= last];
xx(x1) = orig(x1) - k;
xx(x1(mask)) = xx(x1(mask)) - k;

Now you can use bsxfun to vectorize the difference operation:

mask = [true; diff(yy(x1))~=0]; 
xx(x1) = orig(x1) - k*double(bsxfun(@times, mask, ones(sum(x1),1)));

For optimization, the class checks should be avoided and just use 'double'

xx = double(zeros(size(orig)));
yy = double(zeros(size(orig)));

[piero]

thanks but it give me an error..

Arrays have incompatible sizes for this operation.

Error in untitled12 (line 30)

mask = [yy(x1) ~= last];

i attachment file