[Intel-gfx] [PATCH 11/12] sched: use for_each_if in topology.h

[Daniel Vetter]

On Mon, Jul 09, 2018 at 05:12:58PM +0200, Peter Zijlstra wrote:
> On Mon, Jul 09, 2018 at 05:00:07PM +0200, Daniel Vetter wrote:
> > On Mon, Jul 9, 2018 at 12:36 PM, Peter Zijlstra <peterz at infradead.org> wrote:
> > > On Mon, Jul 09, 2018 at 10:36:49AM +0200, Daniel Vetter wrote:
> 
> > >>  #define for_each_node_with_cpus(node)                        \
> > >>       for_each_online_node(node)                      \
> > >> -             if (nr_cpus_node(node))
> > >> +             for_each_if (nr_cpus_node(node))
> > >
> > > Not having gotten any of the other patches, I'm not really sure what
> > > this does and such, but improve readability it does not :/
> > 
> > Patch 1 in this series, which I dumped onto lkml as a whole:
> > 
> > https://lkml.org/lkml/2018/7/9/179
> 
> Right, so while I don't object to being Cc'ed to the whole series, I do
> mind not being Cc'ed to at least the generic bits required to understand
> the patch I do have to look at.
> 
> > Imo it does improve readability for the if (!cond) {} else pattern.
> > And (assuming my grep fu isn't too badly wrong) most places in the
> > kernel do use this pattern in for_each macros, so I guess its a real
> > thing. We've definitely hit it plenty in drm iterators (but we seem to
> > like if() checks in iterator macros maybe a bit too much).
> > 
> > I'm happy to drop this patch tough if you deem it offensive.
> 
> I'd just like to understand it better; what compiler complains about
> this and is the warning otherwise useful? These things don't seem
> mentioned in that initial patch either.
> 
> IOW I suppose I'm asking for the justification of this churn. If it's
> really needed and useful so be it, but so far I'm not seeing any.
> 
> At a while guess I'd say this is something new in gcc-8 (and while I
> have that installed on some machines, it doesn't seem to be the default,
> and so I've not actually seen its output). But is the warning actually
> useful, should we not just kill the warning like we tend to do some
> really silly ones.

for_each_something(foo)
	if (foo->bla)
		call_bla(foo);
	else
		call_default(foo);

Totally contrived, but this complains. Liberally sprinkling {} also shuts
up the compiler, but it's a bit confusing given that a plain for {;;} is
totally fine. And it's confusing since at first glance the compiler
complaining about nested if and ambigous else doesn't make sense since
clearly there's only 1 if there.

Wrt this being useful or not: We've had it for a while in drm, and Andy
and Yishen where rolling yet another open coded version of this on a patch
that flew past me on dri-devel. So I pointed them at the for_each_if() we
have and typed this series to move it to kernel.h.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch