C++Talk.NET C++ language newsgroups   Archives   FAQ   Search   Memberlist   Usergroups   Register   Profile   Log in to check your private messages   Log in  Empty Derived Optimisation (EDO) ???          C++Talk.NET Forum Index -> C++ Language (Moderated) View previous topic :: View next topic   Author Message Matthew Wilson Guest Posted: Sun Aug 24, 2003 11:43 pm    Post subject: Empty Derived Optimisation (EDO) ??? Sorry if this touches on threads it's been beyond me to search, but I was wondering what (a) the language, and (b) implementations have to say about the following: template <typename T> class empty_derived : public T {}; class Simple {}; class Poly { public: virtual ~Poly() {} }; int main() { bool b1 = (sizeof(Simple) == simple(empty_derived<Simple>)); bool b2 = (sizeof(Poly) == simple(empty_derived<Poly>)); return 0; } It is my general experience with Borland, Comeau, CodeWarrior, Digital Mars, GCC, Intel, Visual C++ and Watcom, that only Borland fails to have both b1 and b2 true. In other words, just about all compilers (well Win32 one's anyway) implement what I'm clumsily terming Empty Derived Optimisation. So my question is, is this behaviour a nice feature of the implementations (and not a hard one to achieve, one would think), or is it a recognised part of the language that all compilers *should* provide? Thanks in advance Matthew Wilson STLSoft moderator and C++ monomaniac mailto:admin (AT) stlsoft (DOT) org mailto:stlsoft (AT) hotmail (DOT) com http://www.stlsoft.org news://news.digitalmars.com/c++.stlsoft "If I'm curt with you it's because time is a factor. I think fast, I talk fast, and I need you guys to act fast" -- Mr Wolf ---------------------------------------------------------------------------- [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top llewelly Guest Posted: Mon Aug 25, 2003 9:15 am    Post subject: Re: Empty Derived Optimisation (EDO) ??? [email]stlsoft (AT) hotmail (DOT) com[/email] (Matthew Wilson) writes: Quote: Sorry if this touches on threads it's been beyond me to search, but I was wondering what (a) the language, and (b) implementations have to say about the following: template <typename T>n class empty_derived : public T {}; class Simple {}; class Poly { public: virtual ~Poly() {} }; int main() { bool b1 = (sizeof(Simple) == simple(empty_derived<Simple>)); bool b2 = (sizeof(Poly) == simple(empty_derived<Poly>)); Surely you mean: bool b1 = (sizeof(Simple) == sizeof(empty_derived<Simple>)); bool b2 = (sizeof(Poly) == sizeof(empty_derived<Poly>)); instead? (note sizeof and not simple.) Quote: return 0; } [snip] So my question is, is this behaviour a nice feature of the implementations (and not a hard one to achieve, one would think), or is it a recognised part of the language that all compilers *should* provide? [snip] I think it's an implementator's tradeoff. [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top David B. Held Guest Posted: Mon Aug 25, 2003 9:16 am    Post subject: Re: Empty Derived Optimisation (EDO) ??? "Matthew Wilson" <stlsoft (AT) hotmail (DOT) com> wrote Quote: [...] It is my general experience with Borland, Comeau, CodeWarrior, Digital Mars, GCC, Intel, Visual C++ and Watcom, that only Borland fails to have both b1 and b2 true. In other words, just about all compilers (well Win32 one's anyway) implement what I'm clumsily terming Empty Derived Optimisation. It's called "Empty Base Optimization" (EBO), and the standard allows it, but does not require it (as is the case with most, if not all optimizations in C++). In fact bcc *does* implement EBO, but not consistently. You might have failed to set a compiler switch to turn it on. Quote: So my question is, is this behaviour a nice feature of the implementations (and not a hard one to achieve, one would think), or is it a recognised part of the language that all compilers *should* provide? In an ideal world, all compilers would provide it; but in an ideal world, all compilers would be fast, conforming, and produce small, fast executables. And while single-inheritance EBO seems like it would be easy to implement, the issues get tricky with multiple inheritance. For an example of EBO explicitly used in a library, check out boost::compressed_pair<>. Dave [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Arnold Hendriks Guest Posted: Mon Aug 25, 2003 9:18 am    Post subject: Re: Empty Derived Optimisation (EDO) ??? "Matthew Wilson" <stlsoft (AT) hotmail (DOT) com> wrote Quote: It is my general experience with Borland, Comeau, CodeWarrior, Digital Mars, GCC, Intel, Visual C++ and Watcom, that only Borland fails to have both b1 and b2 true. In other words, just about all compilers (well Win32 one's anyway) implement what I'm clumsily terming Empty Derived Optimisation. Isn't this what I've seen more often called "empty base class optimisation" ? Borland's compiler supports it as well, although it's controlled through a switch (and it's buggy in BCB5 if multiple inheritance is involved) - perhaps you didn't enable it? [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top John Torjo Guest Posted: Tue Aug 26, 2003 8:12 am    Post subject: Re: Empty Derived Optimisation (EDO) ??? Quote: int main() { bool b1 = (sizeof(Simple) == simple(empty_derived<Simple>)); bool b2 = (sizeof(Poly) == simple(empty_derived<Poly>)); return 0; } sizeof instead of simple. Quote: It is my general experience with Borland, Comeau, CodeWarrior, Digital Mars, GCC, Intel, Visual C++ and Watcom, that only Borland fails to have both b1 and b2 true. In other words, just about all compilers (well Win32 one's anyway) implement what I'm clumsily terming Empty Derived Optimisation. So my question is, is this behaviour a nice feature of the implementations (and not a hard one to achieve, one would think), or is it a recognised part of the language that all compilers *should* provide? It's not a requirement, the C++ Standard allows it though. It's called Empty Base class optimization. In generic programming, it's *quite* useful. That's why boost::compressed_pair exists. Best, John [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Matthew Wilson Guest Posted: Tue Aug 26, 2003 8:19 am    Post subject: Re: Empty Derived Optimisation (EDO) ??? Quote: It's called "Empty Base Optimization" (EBO), and the standard allows it, but does not require it (as is the case with most, if not all optimizations in C++). In fact bcc *does* implement EBO, but not consistently. You might have failed to set a compiler switch to turn it on. AFAIUI EBO is not what's going on here. Maybe my example gave the wrong impression. I'll rephrase template <typename T> class empty_derived : public T { // No per-instance (non-static) member variables // No explicitly virtual methods, i.e. no use of virtual keyword }; class Simple { int i; }; class Poly { public: virtual ~Poly() {} int i; }; int main() { bool b1 = (sizeof(Simple) == sizeof(empty_derived<Simple>)); bool b2 = (sizeof(Poly) == sizeof(empty_derived<Poly>)); return 0; } EBO would only apply if the base classes Simple and Poly were empty, which they're not. It's the derived class empty_derived that is empty hence E*D*O. (btw, I'm not allied to the term. It's just a place-holder until someone enlightens me.) Matthew Wilson STLSoft moderator and C++ monomaniac mailto:matthew (AT) stlsoft (DOT) org http://www.stlsoft.org news://news.digitalmars.com/c++.stlsoft "But if less is more, think how much more more will be!" -- Dr Frazier Crane ------------------------------------------------------------------------------- [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top David B. Held Guest Posted: Tue Aug 26, 2003 7:08 pm    Post subject: Re: Empty Derived Optimisation (EDO) ??? "Matthew Wilson" <stlsoft (AT) hotmail (DOT) com> wrote Quote: [...] AFAIUI EBO is not what's going on here. Maybe my example gave the wrong impression. I'll rephrase [...] EBO would only apply if the base classes Simple and Poly were empty, which they're not. It's the derived class empty_derived that is empty hence E*D*O. Yes, the fact that your original example had all empty classes made it impossible to know that this was the precise distinction you were making. Anyway, after looking at 9.3, 10.5 and a few other places, I suspect that this optimization is allowed, but not explicitly mentioned or implied in the standard. That's because the standard doesn't say how big a derived class has to be other than that it has to have non-zero size (like all class types). Which means that anything which observes all the other rules for layout/size/storage are legal, and I'm not aware of any rules which constrain the derived class in this case to be larger. I would say that because the base subobject and derived complete object can share the same address, that EDO must be legal (except possibly in the case of MI). Dave [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Daveed Vandevoorde Guest Posted: Sat Aug 30, 2003 7:36 pm    Post subject: Re: Empty Derived Optimisation (EDO) ??? [email]stlsoft (AT) hotmail (DOT) com[/email] (Matthew Wilson) wrote: [...] Quote: So my question is, is this behaviour a nice feature of the implementations (and not a hard one to achieve, one would think), or is it a recognised part of the language that all compilers *should* provide? The former. (Note that the base could be a POD, whereas the derived class is certainly not.) Daveed [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Display posts from previous: All Posts1 Day7 Days2 Weeks1 Month3 Months6 Months1 Year Oldest FirstNewest First         C++Talk.NET Forum Index -> C++ Language (Moderated) All times are GMT Page 1 of 1   Jump to: Select a forum C/C++----------------C LanguageC Language (Moderated)C++ Language (Moderated)C++ language (comp.lang.c++)C++ language, library and standards Non-English C/C++ Newsgroups----------------C++ (German)C++ (French)  You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum Powered by phpBB © 2001, 2006 phpBB Group SEO toolkit © 2004-2006 webmedic.