On Wed, Jan 31, 2018 at 09:38:46PM -0800, Nadav Amit wrote: > I used ftrace to measure the execution time of flush_tlb_func_remote() on a > 2-socket Haswell machine, using a microbenchmark I wrote for some research > project.
However cool ftrace is, it is _really_ bad for such uses. The cost of using ftrace is many many time higher than any change you could affect by this. A microbench and/or perf is what you should use for this.
