Asked by the cyclist
on 23 Apr 2012

In the Stanford online course in probabilistic graphical models, they use their own linear congruential random number generator. The results from that generator differ between R2012a and earlier versions of MATLAB, and I boiled it down to a difference in the calculation of mod() for a very large number:

mod(25754210418041856,179424673)

Since the first argument is 17 digits, and therefore not accurately stored as an integer, it is no surprise to me that mod() might struggle; but I am wondering if anyone can give me some more specific insight as to why the behavior may have changed from version R2011b to R2012a.

Answer by Richard Brown
on 23 Apr 2012

Accepted answer

@Walter Roberson is correct, in R2012a MATLAB automatically casts the number into a `uint64`. For this piece of code

[mod(25754210418041856,179424673), mod(uint64(25754210418041856),179424673)]

In R2011b I get this:

149629568 149629566

In R2012a I get this:

149629566 149629566

Richard Brown
on 23 Apr 2012

Also, from the release notes for R2012a:

"The following functions show improved performance:

Arithmetic and similar basic math functions for double, single, and integer data types."

Opportunities for recent engineering grads.

## 1 Comment

## Walter Roberson (view profile)

Direct link to this comment:http://www.mathworks.com/matlabcentral/answers/36236#comment_75126

Perhaps R2012a automatically used uint64?