Re: [PATCH 2/6] fsmonitor: Stop inline'ing mark_fsmonitor_valid / _invalid

From: Ben Peart <peartben@gmail.com>
Date: Jan 8, 2018, 20:27 UTC
Message-ID: <01ad47b4-aa5e-461a-270b-dd60032afbd1@gmail.com>
In-Reply-To: <alpine.DEB.2.21.1.1801042326581.32@MININT-6BKU6QN.europe.corp.microsoft.com>

On 1/4/2018 5:27 PM, Johannes Schindelin wrote:

Show 11 quoted lines

> Hi Alex,
> 
> On Tue, 2 Jan 2018, Alex Vandiver wrote:
> 
>> These were inline'd when they were first introduced, presumably as an
>> optimization for cases when they were called in tight loops.  This
>> complicates using these functions, as untracked_cache_invalidate_path
>> is defined in dir.h.
>>
>> Leave the inline'ing up to the compiler's decision, for ease of use.
>

I'm fine with these not being inline.  I was attempting to minimize the 
performance impact of the fsmonitor code when it was not even turned on. 
  Inlineing these functions allowed it to be kept to a simple test but I 
suspect (especially with modern optimizing compilers) that the overhead 
of calling a function to do that test is negligible.

Show 6 quoted lines

> As a compromise, you could leave the rather simple mark_fsmonitor_valid()
> as inlined function. It should be by far the more-called function, anyway.
> 
> Ciao,
> Johannes
>