Can't easily be comapred.

1 - there are a huge number of different 8051 chips out there.

2 - the use of sleep, interrupts, algorithms etc will greatly affect the number of instructions needed.

3 - power consumption depends on supply voltages, and load on processor pins.

4 - Recompiling similar code for different processor architectures will mean different compilers. But one compiler with one set of optimization flags for a specific architecture will not produce the same results as another compiler with another set of optimizations and different way of generating code.