Re: Inlining and estimate_num_insns

[Gabriel Dos Reis <gdr at integrable-solutions dot net>]

Richard Guenther <richard.guenther@gmail.com> writes:

[...]

| > Of course, if we can make the compiler automatically do the right thing,
| > that's great.  What I'm proposing is a method that would be possibly
| > acceptable for 4.0 that would give people an easy to understand way of
| > getting the compiler to do what they want: put "inline" on functions
| > they really want to have inlined.  That's in addition to the existing
| > heuristics.  I think that's a plausible way of mitigating the problem
| > for people in the short term.
| 
| People already know to use __attribute__((always_inline)) (ugh!), I don't

If your only compiler is G++, then that might probably be sufficient.  But,
in industrial setting, we do not have that luxury.  Typically, one has
a set of compilers to target.  I'm of the opinion that we should not
require people to use non-standard syntax just to achieve what the
existing standard keyword has been invented for.  But, I suppose we
already had that argument.

-- Gaby