Messages in this thread | | | Date | Thu, 1 Feb 2018 10:38:47 +0100 | From | Peter Zijlstra <> | Subject | Re: [PATCH] x86: Align TLB invalidation info |
| |
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.
| |