There is one characteristic of MathCAD 11 that I have made use of when writing my own C code functions.
The default parameter passing requires that the pointers to all the inputs be constant (const).
However, I was implementing an adaptive filter and I needed to see the resulting coefficients, so I called the routine as follows:
LRESULT Nlms( COMPLEXARRAY * const Nlms,
COMPLEXARRAY * const H,
const COMPLEXARRAY * const X,
const COMPLEXARRAY * const Y,
const COMPLEXSCALAR * const Step,
const COMPLEXSCALAR * const Points
)
Not only could I see the resulting updates to the coefficients, but I could repeatedly press F9 over the function and re-run the same data over and over to get better and better results, sometimes while gradually adjusting to smaller and smaller step sizes. This worked because subsequent calls used the coefficients starting at with results of the previous call.
Unfortunately, someone at MathSoft figured out that this door needed to be closed, and in subsequent versions of MatCAD such tricks no longer work.
The technique I use now is to append the coefficients to the end of the output from the adaptive filter, strip them off into a new vector, and then call a new copy of the function with the new coefficient vector. Not a very satisfactory technique as one can appreciate.
I wonder if there is any equivalent game to my earlier technique that I can play to return to this idea in more recent versions?
David Shaw
