On Tue, Jan 10, 2012 at 16:40, Noah Watkins <jayhawk@xxxxxxxxxxx> wrote: > The cost of reading the software clock via gettimeofday is dependent on the > OS implementation, and can vary by as much as 2 orders of magnitude (see > attached slide: 1.6us vs 60ns per clock read). > > Older Linux kernels force a thread into the kernel when calling gettimeofday > adding enormous overhead relative to the newer vsyscall implementations that > evaluate gettimeofday within userspace. So turning debug to full blast on older kernels will be painful. I'm personally ok with that, especially when the alternative is implementing all of the TSC logic myself. This is really complex code, and somebody went through the effort of solving the hard problems for us -- let's use their code! -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
