Re: [PATCH] x86: Align TLB invalidation info

[Peter Zijlstra]

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.