Michael Völker

Very nice, simplistic code.

However, if the input is a complex column vector, you accidentally conjugate it, due to this line:
x = x';

Could you please repair that?
--> x = x.';

Charles,

for my own record and to make it machine-searchable, I wrote earlier how I was able to compile mtimesx. Taking the liberties to quote myself:

> On a 64Bit Debian based Linux I managed to compile it with
> mex -DDEFINEUNIX CFLAGS="\$CFLAGS -march=native" -largeArrayDims -lmwblas -lmwlapack -lgomp mtimesx.c
> using gcc-4.5.

Can you try that, or did you already try?

Michael

Where MATLAB's atan2d() is already present, only a single line is required to make the result [0...360]:

ang = mod( atan2d(y,x), 360 );

If atan2d() is not present but atan2() is, it is sufficient to use

ang = mod( atan2(y,x) .* 180./pi, 360 );

And if somehow neither atan2() nor atan2d() were present and they had to be implemented manually, I would expect them to behave like the original functions.

The most important lack of function is the inability to work on arrays.

Hi gang,

I actually need that for masking 2D data, only. Or to help me out in some quick simulations.
But since I only used simple Pythagoras, I thought I'd make it more flexible with n-dimensional support.

And yes, I am really not surprised that you can use it for a 3D ROI. :-)

Jan Simon,

> I will definitely avoid to overload built-in functions

the same is (usually) true for me.

But the unbelievable 64-step-colormap-issue in current MATLAB *is* in fact a reason to change MATLAB's behaviour for all platforms where size(colormap) == 64.

At the moment, I see no other usecase for myself, but in these pretty rare cases, it is nice to have the possibility to patch MATLAB...