Concerns about MatrixMath.h Library

Projects Programming
1

This is a fantastic library, however I am concerned why it requires matrices to be casted to float* for each function? I am not sure, but I believe this is very demanding on the micro, in terms of process time.

I ask this question because I am doing matrix math in real-time (for state-space modelling). I hardly use pointers so I am unaware why they are used here.

2

Are you referring to the float* in the function definitions, such as this one?:

void MatrixPrint(float* A, int m, int n, String label);

If so, it just means that you pass the address of the matrix A to the function. e.g. you might have a code snippet like this:

float A[3][5];
.
.
.
MatrixPrint(A, 3, 5, String("whatever"));

[+edit] i.e. the name of the array is already the address of the array. No cast needed.

Pete

3

It also means that your arrays should be floats already. What kind of arrays are you using?

4

I followed the example that was provided in the library.

For instance, if I have 2 matrices,

const double A[2][2] = {
 {1.95, -0.9512},
 {1, 0}
};

const double B[2][1] = {
 {1},
 {0}
};

And variable matrix,

double temp[2][1] = {
 {0},
 {0}
};

Using the library, I would multiply them and store the result into temp[][],

Matrix.Multiply((float*)A, (float*)B, 2, 2, 1, (float*)temp);

In my application, I require values in my matrices with precision of at least 8 decimal places.

This thread is ultimately leading to, what appropriate data types can I use with this library.

5

On the Arduino float == double. And 8 significant float figures is pushing it.

6

Which Arduino are you using? On Nano/UNO and similar, a double is the same as float (single precision).
But, if the library uses float, you aren't going to get the precision you need. The precision of float is only about 7 digits.

Pete

7

KeithRB:
On the Arduino float == double. And 8 significant float figures is pushing it.

Yes, thank you, I realized the float == double. I haven't finished my state space modelling and in most cases the precision I need is 4 decimal places. I can prevent my model from requiring such precision.

el_supremo:
Which Arduino are you using? On Nano/UNO and similar, a double is the same as float (single precision).
But, if the library uses float, you aren't going to get the precision you need. The precision of float is only about 7 digits.

Pete

I am using an Arduino Due. Thank you for letting me know.

8

OK. AFAIK, on the Due, "double" is double precision. But if the library uses float internally, that's not going to help you.

Pete

9

Though it should be pretty easy to re-write the library for double. Search and replace should do it.