Messages in this thread | | | Date | Mon, 25 Jan 2021 14:55:29 -0500 | From | Steven Rostedt <> | Subject | Re: [PATCH 1/3] tracing: Merge irqflags + preempt counter. |
| |
On Mon, 25 Jan 2021 20:52:28 +0100 Sebastian Andrzej Siewior <bigeasy@linutronix.de> wrote:
> On 2021-01-25 14:03:23 [-0500], Steven Rostedt wrote: > > > > I was thinking about the inlining for two reasons. One was to consolidate > > the logic in the header file, as they are small functions. And two, inlined > > functions tend to be faster than non-inlined functions. Thus, I wasn't > > looking at this from a size point of view, but since this is called by all > > events, including function tracer, being efficient is a requirement. > > In the ftrace_syscall_enter() example I made, flags were not evaluated > three times but only once. This should do more in terms of performance > compare to simply inline it.
Oh, I know. It's one of the things I like about your series. But that doesn't make it less of a reason to inline it ;-)
> > > > I can post the irqflags-merge and the inlinining as two seprate patches > > > and you can then decide if you merge the two patches or drop the > > > inlining. > > > > Feel free to send it as separate patches. I'd like to have the inlining. > > Just sent as an extra patch. In case you have a benchmark, I'm curious > ;) >
Yep, looking at it now. Thanks!
-- Steve
| |