Cross-platform support via ifdefs

Projects Programming
1

I'm working on Tensorflow for microcontrollers and the same C++ source is used to build for (among others) Linux as well as Arduino.

I would like to use #ifdefs to separate out the platform specific code.

For example:

void my_function() {

... // common code to both Arduino and Linux

#ifdef MICRO_BUILD // embedded target

... // Arduino-specific implementation.

#else // Linux

... // More general implementation, with say dynamic memory allocation or Linux-only matrix math libraries.

#endif

... // common code to both Arduino and Linux

}

This pattern works fine with the arduino-cli:

arduino-cli compile --build-properties compiler.cpp.extra_flags=-DMICRO_BUILD

However, I can not figure out how to specify -DMICRO_BUILD for the Tensorflow Arduino library when it is built via the Arduino IDE as part of a sketch.

Is there a way to specify some custom cflags for a library? Maybe something similar to the ldflags that can be specified in library.properties so that any sketch that uses the Tensorflow Arduino library will build that library with -DMICRO_BUILD?

2

You could use #ifdef ARDUINO, it's defined by the Arduino builder.

Also, all hardware platforms and architectures have their own defines, so you could even differentiate between ARM and AVR, for example.

Pieter

3

Thanks! I knew about the hardware platform defines but a blanket ifdef ARDUINO does the trick for me.