From -9088805825489566878 X-Google-Language: ENGLISH,ASCII-7-bit X-Google-Thread: f78e5,f5d433881cc29cff X-Google-Attributes: gidf78e5,public From: bparker@gil.com.au (Brian Parker) Subject: Re: Named return value optimisation- reprise. Date: 1997/06/20 Message-ID: <33a8768c.2060634@news.ipswich.gil.com.au>#1/1 X-Deja-AN: 251292084 References: <33a345bf.7425526@news.ipswich.gil.com.au> <33a4e2d5.24720499@news.ipswich.gil.com.au> <33A54D26.F46@acm.org> <33a5f2fc.3455132@news.ipswich.gil.com.au> <5o6rng$5ia@tools.bbnplanet.com> X-Original-Date: Thu, 19 Jun 1997 03:23:08 GMT Organization: Global Info-Links News Server X-Auth: PGPMoose V1.1 PGP comp.std.c++ iQBFAgUBM6o/meEDnX0m9pzZAQEMsAGAkBLZrt/SlfgFVVBKizRCfXsDcbBRgQBv i+ovn+ujITL6OakpAj04iBUw42IPagQA =pBnL Newsgroups: comp.std.c++ Originator: fjh@mundook.cs.mu.OZ.AU On 18 Jun 97 02:13:20 GMT, Barry Margolin wrote: >In article <33a5f2fc.3455132@news.ipswich.gil.com.au>, >Brian Parker wrote: >>As far as I can see, 12.8/15 could only be unambiguously safely >>applied by a compiler for a class with bitwise copy semantics, i.e. >>either the class and all its members and base classes have implicitly >>defined copy constructors, or any user-provided explicit copy >>constructors must involve only simple assignment of members. In these >>cases, though, the copy could already be elided by the "as if" rule. > >I suspect this is the optimization that this section is trying to allow. >However, the as-if rule is not sufficient to permit it, because the as-if >rule is violated if the class or any of its members or bases have side >effects in their destructors (as in the destructor_count++ example that >someone posted). That's true- I forgot about user-provided destructors. My statement above only applies if all destructors are also implicit. > >I think that NRVO should be reworded to specify that if the copy >constructor has bitwise copy semantics, you can't depend on destructors >being run during an assignment. That may be a useful approach. If the copy constructor has bit-wise semantics and the destructor is empty then the optimisations of 12.8/15 could already be done by the "as if" rule, but your proposal would also allow them where there is a non-empty destructor. I agree that that would be quite safe- the only question is whether the additional optimisations allowed by this would be useful in practice. This would certainly solve the problems with 12.8/15, though. Anyone else have any comments on this? ,Brian Parker (bparker@gil.com.au) --- [ 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 ] [ FAQ: http://reality.sgi.com/employees/austern_mti/std-c++/faq.html ] [ Policy: http://reality.sgi.com/employees/austern_mti/std-c++/policy.html ] [ Comments? mailto:std-c++-request@ncar.ucar.edu ]