On 16/04/19 13:04, Liran Alon wrote: >> A 1 nanosecond advance (3-5 clock cycles) is even shorter than the time >> taken to execute kvm_read_l1_tsc. I would just replace it with >> cpu_relax(); I can do the change when applying. > > Paolo, there is also another approach Sean and I were discussing. > I think it is more elegant. > See previous comments in this thread. Yep, I've seen it now. Paolo
