Linux Bogomips is calculated by executing a simple loop:
- count down a register
- jump back if register > 0
For ARM Cortex A8, you need 2 Instructions for each loop iteration. You will get 800.000.000 loop iterations == 1.600.000.000 Instructions/second.
The bogomips loop is executed from L1 cache, so the memory interface has nothing to do with the bogomips number.
