Our efforts at monitoring CPU usage show that the flow processing takes a significant portion of the CPU. This may become more of an issue as additional functionality is added, especially if there's a need for quick response to events. One way to minimize the impact is to optimize the efficiency of flow processing coding, including steps like:
moving constant calculations out of loops
avoiding floating division (multiply by reciprocal instead) because the ESP32 FPU is particularly slow at division
unrolling loops ( although trading off for memory use isn't ideal when we're using a lot of memory at the moment)
reducing unproductive generality. For example, looking up the number of legs on the robot in a variable can be replaced by a macro which can be optimized by the compiler
using inline routines where practical, and not too expensive for memory
using conditionals to remove debugging and diagnostic code for "production" software
using integers rather than floating point when practical
Our efforts at monitoring CPU usage show that the flow processing takes a significant portion of the CPU. This may become more of an issue as additional functionality is added, especially if there's a need for quick response to events. One way to minimize the impact is to optimize the efficiency of flow processing coding, including steps like: