How does matlab compute hypergeom?

I want to know what method Matlab uses to compute hypergeom (generalised hypergeometric function) if I input say floats rather than symbolic variables. For instance does it simply use the power series or is there some sort of continued fraction approximation?

Answer 1

[Walter Roberson]

Historical method: https://www.mathworks.com/matlabcentral/answers/350039-is-there-any-way-to-speed-up-the-calculation-of-the-confluent-hypergeometric-function

You can read the current method:

strjoin(regexp(char(evalin(symengine,'expose(hypergeom)')), '\\n', 'split'), '\n')

[Matt]

Thanks, that's helpful. However, I am not quite sure what the code (that I get when I type in the above) is doing? Am I right in thinking that your previous answer still hold - i.e. matlab uses maple for as many digits as necessary?

[Walter Roberson]

MATLAB these days calls the MuPAD symbolic engine instead of Maple. The code you would have seen is MuPAD code. MuPAD was originally written in Germany. specfunc::Wurzelbehandlung appears to translate as "root treatment".

Most of the code should be easy to mentally translate. [] is indexing, () is bracketing or function calls, := is assignment, = is comparison. nops() is more or less length() . {} is set construction. {op(l1)} more or less converts the list of items in l1 into a set, so it is like unique(l1).