[g++] make_pair inlining failure?

Marc Glisse marc.glisse@inria.fr
Tue Nov 29 14:00:00 GMT 2011


On Tue, 29 Nov 2011, Robert Jones wrote:

> On Mon, Nov 28, 2011 at 2:42 PM, Marc Glisse <marc.glisse@inria.fr> wrote:
>> The set of circumstances where a compiler is allowed to elide a copy is extremely restrictive.
>
> Really... how restrictive exactly? Would you take issue with anything stated in
>
> http://cpp-next.com/archive/2009/08/want-speed-pass-by-value/

No, that page was carefully written by a knowledgeable person so the 
examples would fit right into those circumstances. And those examples are 
basically the only cases where copy elision is allowed. Yes, some of us 
would like to change that. Note that in C++11, you at least get a move 
instead of a copy.

-- 
Marc Glisse



More information about the Gcc-help mailing list