As a sibling comment points out, DSPs can be an exception. I’m far from an expert on them, but CPUs that try to run all instructions in the same fixed amount of time used to accomplish that by avoiding complex addressing modes and by omitting the really slow instructions (division and instructions that push/pop multiple registers onto/from the stack being prime examples).
IIRC, some of them also accomplished it by running some of the simpler instructions slower than necessary (raw speed isn’t as important as being hard real time isn many applications)
