i've discovered the best way to profile stuff in hardware is to toggle GPIOs in "breakpoints". It has almost no overhead and doesn't interfere with timing-critical operations. Then just watch the GPIOs with your oscilloscope
@MightyPork Agreed. We have a lot of execution flag toggles (inline function calls that #ifdef to empty functions in release).
The ARM ETM is extremely good for complex things where you'd need 20 GPIOs to build a flame chart, but requires $$$ trace pods.
@MightyPork Segger's Ozone runs on Mac, Win, Linux and seems to be well-liked. But, again, gotta have the $$ pod.
(We use IAR's debugger for most things but with Segger pod hardware and support software. I'm contemplating switching back to Ozone because IAR has had some Problems shipping working scripts for some of ST's new silicon.)
@jond (ST has the same problems with their own silicon ...)
@jond and even if i had one it's gonna be windows only most likely