Performing random numbers generator from a generic discrete distribution

This function extracts random numbers distributed over a discrete set; the PDF is user-defined

This function extracts a scalar/vector/matrix of random numbers with discrete Probability Distribution Function.
The PDF is specified by the user as a input vector.

This function is designed to be fast, and it is implemented within a .mex file

Following Olivier B. comments (that I acknowledge for his comments), I performed cross-comparisons with randp. gDiscrPdfRnd is faster with a ratio that increases with the number of number, i.e. for about 3 times faster for 10^6 numbers to over 40 times faster for 10^7 numbers.
Moreover, for large random arrays, randp seriously surcharges the RAM memory, whereas gDiscrPdfRnd limits thememory use to what is essential (tanksto the coding). In what follows the details of thecomparison are given.

>> tic;R = randp([1 3 2],1000000,1);toc

elapsed_time =

0.4840

>> tic;R = gDiscrPdfRnd([1 3 2],1000000,1);toc

elapsed_time =

0.1570

>> tic;R = randp([1 3 2],10000000,1);toc

elapsed_time =

68.5780

>> tic;R = gDiscrPdfRnd([1 3 2],10000000,1);toc

elapsed_time =

1.6410

>> 68.5780/1.6410

ans =

41.7904