Re: Inlining and copy-elision
Ofer Porat ha scritto:
Consider the code
=====
struct T {};
inline void g(T const& p)
{
T q(p);
}
int main()
{
g(T());
}
=====
If the function g() is not inlined, the copy construction of q may not
be elided, because the temporary passed to g() is bound to a reference
(p).
What happens if g() is inlined? Since p is only used once in g(), the
compiler might eliminate the p parameter altogether. Is the compiler
allowed to elide the copy construction of q and instead
value-initialize q directly?
According to my interpretation, in this specific case the answer is yes.
Notice that the criteria set in 12.8/15 are about omitting the copy
"even if the copy constructor and/or destructor for the object have side
effects." If the copy ctor and/or dtor do *not* have side-effect, as it
happens in this case, then the presence of absence of the copy is not
part of the "observable behavior" so the compiler can practically do
whatever it want: it can elide the copy or not, the program can't tell
the difference anyway (timed benchmarks unfortunately do not count as
"observations"). Of course the semantic (accessibility, etc.) must still
be respected, but that't another story.
HTH,
Ganesh
---
[ comp.std.c++ is moderated. To submit articles, try just posting with ]
[ your news-reader. If that fails, use mailto:std-c++@ncar.ucar.edu ]
[ --- Please see the FAQ before posting. --- ]
[ FAQ: http://www.comeaucomputing.com/csc/faq.html ]