option faster than unique?

I am working on a 955*1 double vector xy where I need to generate it 101 times and once generated, each time I need to remove repeating elements from it, sort in ascending order and get index of remaining elements. I am using unique for this. (i.e.[xyuni, ind] = unique( xy );) This operation takes almost 2 sec. I need to make this faster. For that, I tried to use intersect but it is slower than unique. I downloaded count_unique from http://www.mathworks.com.au/matlabcentral/fileexchange/23333-determine-and-count-unique-values-of-an-array. But this only removes repeating elements and sorts but does not provide index. I tried tip from http://www.mathworks.com/matlabcentral/answers/127660-how-would-i-create-my-own-unique-function-without-using-the-built-in-unique and created myUnique using sort but again it does not provide index. Please let me know if there exist any other option to make this operation faster. Many thanks.

[Uday]

Can you paste the data and the code to visualize what you are trying to explain ?

Answer 1

[Sean de Wolski]

Perhaps I'm missing something but on my machine (fairly wimpy laptop) this is taking a few hundredths of a second

tic
for ii = 1:100
  [uv,id] = unique(randi(400,1000,1));
end
toc

Elapsed time is 0.013925 seconds

So I doubt unique is the bottleneck and even if it is, I doubt you'll find a faster solution without C or Fortran hardwired to your inputs. Have you profiled your code?

docsearch profile code