On Sat, 3 Mar 2018, Martin Sebor wrote:
> On 03/02/2018 01:05 AM, Richard Biener wrote:
> > On Thu, 1 Mar 2018, Martin Sebor wrote:
> >
> > > While testing my recent changes to the handling of attributes
> > > on C++ templates I noticed that the -finline-limit=N option
> > > is not recognized in attribute or pragma optimize. In response
> > > to the bug I raised, Richard you said the option is deprecated,
> > > so I went ahead and documented the deprecation in the manual
> > > and also added a deprecation warning. I also added it to
> > > Optimization options. The attached patch reflects these
> > > changes.
> > >
> > > I also tried to put together a test to verify that the option
> > > works as expected, both in attributes and in pragmas. I wasn't
> > > able to get it to work reliably or sensibly.
> > >
> > > In C, #pragma optimize ("inline-limit=1") has a global effect
> > > on all functions in the file, even those lexically before the
> > > pragma. In C++, the pragma has no effect. Both of these
> > > effects are contrary to my reading of the manual.
> > >
> > > Attribute optimize ("inline-limit") behaves similarly, which
> > > seems even more unexpected and even more contrary to the manual
> > > which by referring to Function Specific Option Pragmas suggests
> > > that both the pragma and especially the attribute apply
> > > optimizations to just the functions they're specified for.
> >
> > It can't really work so please do _not_ add the option to
> > the set of Optimize options. Inlining is an IPA task so
> > "per function" parameters do not make sense at all.
>
> It does work as I expect for other inlining options (e.g., for
> -finline-functions or -finline-functions-called-once) so either
> there's something special/different about -finline-limit or that
> it works for the others is an accident. Can you please clarify
> which it is?
It works for the others because those flags can be properly
tested per function. Since -finline-limit just maps to
some --params and those are not per function it cannot work
similarly for that. Adding 'Optimization' to -finline-limit
will just cause that flag to be kept per function without
any effect given nothing tests that flag but only the params.
Consider
#pragma GCC optimize("-finline-limit=10")
void foo () { }
#pragma GCC optimize("-finline-limit=100")
where we'll likely just have the 100 limit at the end.
I'd rather _not_ expose this kind of brokeness to the user.
> Either way, wherever they come into play, the limitations you
> refer to should be mentioned in the GCC warning when an option
> that can't or may not work is used in attribute or #pragma
> optimize, and it should also be explained in the manual. How
> else are users supposed to know?
They know because the option doesn't work there but only
on the command line?
> As it is, since -finline-limit is listed among optimization
> options, and since attribute and #pragma optimize say they
> accept optimization options, it's a bug that GCC rejects it
> as if it were invalid.
That's a documentation bug then. Specifially it says
"This pragma allows you to set global optimization options for functions
defined later in the source file."
that's not very specific and I don't read into it that you
can add all and only those options listed in the
"Optimization Options" list. After all there's no cross-reference.
Specifically it says "global" optimization options -- which is
a subset of "optimization options". I'd have expected
"local" optimization options here but that's besides the point.
I'm fine with documenting for -finline-limit that it cannot be used
in the pragma or attribute but I expect others to be in the same
category.
Richard.
