Float buffer allocation uses unnecessary memory

[0xMihir]

The first part of the rf_heart_rate_and_oxygen_saturation code allocates two float arrays to perform calculations on. The default buffer size of 100 results in 800 bytes (2 arrays 4 bytes 100 elements) of memory usage, making many libraries that require some dynamic memory impossible to use in conjunction with this one.

I know that it's the best convention to leave parameters unmodified, but when working on underpowered microcontrollers like the Arduino Uno, performing all calculations on a single float buffer makes this library use a lot less memory. I have already modified the code to work this way, but I wanted to hear your thoughts before opening a PR.

[aromring]

I am not exactly sure what you are referring to, so I am guessing it's about an_x and an_y arrays, correct? Indeed, some calculations can be performed separately for IR and RED channels using one array, but what about the rf_Pcorrelation() call? It needs both at the same time.

[0xMihir]

Rather than creating the an_x and an_y arrays, I modified the code to modify the arrays passed in the function's parameter. I realized that saying single float buffer was unclear.

[aromring]

Thank you for good contribution, but I cannot accept the pull request as is because it seems incomplete - it does not contain matching changes to RD117_ARDUINO.ino. It still contains: uint32_t aun_ir_buffer[BUFFER_SIZE]; which is passed to your version of rf_heart_rate_and_oxygen_saturation() as float pointer. Perhaps there is an implicit conversion from pointer to uint32_t but I would not bet that all compilers do it.

[0xMihir]

Yep, I'll change it to float.

[aromring]

Hi 0xMihir, I had to revert your pull request, sorry, since it would break the code in this repository. In fact, other functions in other parts of the code still use uint32_t incompatible with float. For example, maxim_max30102_read_fifo() in max30102.cpp or the original MAXIM algorithm in algorithm.cpp.