If I run my program under perf on linux, it will easily give me the total count of instructions executed. Is it possible to somehow (maybe also with perf but that’s not a requirement) get a count of instructions executed, broken down by instruction? e.g. it might look like this (but with presumably many more instructions and much higher numbers)

      8 cmp
      4 mov
      3 jne
      2 xor
      2 shr
      2 ret
      2 jb
      2 jae
      1 movzx
      1 jmp
      1 inc
      1 cmovb
      1 add

my motivation is to diff the outputs for two versions of a program as one angle of analyzing performance differences.