nviennot/core-to-core-latency

Does not take into account cmpxchg instruction latency for x86

geranim0 opened this issue · 0 comments

Current code calculates

latency(cmpxchg) + latency(cacheline), not just latency(cacheline)

latency(cmpxchg) for modern processors should be in the order of 1-2ns, therefore results presented have a small error in them