Is it possible to parallelize permpos?

I found this code on file exchange: http://www.mathworks.com/matlabcentral/fileexchange/11216-permpos

And it does exactly what I want. However, it stalls about when permpos(16,32) is there a way to parallelize this so that it will run faster?

Answer 1

[Jan]

See Answers: generating exhaustive permutations for a list of C-Mex files for the memory efficient creation of large permutations/combinations. If your data have integer values, use an integer type to save memory and run-time.

You can create one permutation after the other without repetitions. See. e.g. FEX: nextperm, but as far as I can see, this does not allow repetitions of the elements.

Other useful submission:

• FEX: nextvector
• FEX: oneperm
• FEX: next permutation/combination
• FEX: nextvector

You see, searching the FEX is a good idea.

[Jenn Lee]

Thank you very much, I'll look into these

Answer 2

[Walter Roberson]

It appears to me that permpos(16,32) would be getting pretty close to the point of filling 32 bits worth of address space (4 Gb). How big do you need to go, and what are you going to do with the output when you get it? Are you using a 64 bit version of MATLAB ?

[Jenn Lee]

It is 64-bit 2011 version.

I take the output and compare each vector's distance from all hadamard vectors of equal size.

[Walter Roberson]

And generating all of them at once is an efficient algorithm??

Anyhow you did not answer about what size you wanted to go to?

[Jenn Lee]

Hm, I haven't figured out how big I need to go yet.

I need to be unique, and I'm not sure how to know they are unique unless I have all of the generated at once.