On 27/05/16 12:17, Antti Kantee wrote:
In my testing the clock_gettime-to-clock_gettime delay for a 1ns sleep on KVM is ~8us with this patch.
Actually, there's probably a risk that 1ns delay won't sleep at all, but seems like the underlying clock granularity is sucky enough so that a 1ns sleep is possible. Or at least the results are comparable with 100, 1000 and 10000 ns sleeps.
