I had started a thread about callback performance here and I’ve just added a link to may sources:
If you’re using a STM32 MCU, then the HWTimer class may work for you. It isn’t heavily tested, I’ve used it with F407 and F411.
You want to use it also for a stepper driver, there the overhead of the Mbed Timer is significant. I will improve it further, also the CThunk costs a lot of cycles.
My StepperController has some fixed settings and needs also some more work. But acceleration and deceleration works fine, I have used the ramp function from the Arduino AccelStepper Lib. The calculation is using float math, so a MCU with FPU should be used.