June 2012 - mpi-forum - mailman3.cels.anl.gov

Re: [Mpi-forum] C++ types inaccessible after #281
by Jed Brown 26 Jun '12

26 Jun '12

On Tue, Jun 26, 2012 at 8:19 AM, Jeff Squyres <jsquyres(a)cisco.com> wrote: > On Jun 26, 2012, at 12:14 PM, Jed Brown wrote: > > > On Tue, Jun 26, 2012 at 6:13 AM, Jeff Squyres <jsquyres(a)cisco.com> > wrote: > > There are a few C++-specific datatypes for which it would be worth > introducing new C datatypes (e.g., MPI_CXX_COMPLEX). It's too bad that no > one realized this until now, because this won't make the MPI-3 deadline. > But it could be part of MPI-3.1 or errata, or some such. > > > > Two people in offline discussions have suggested that C bindings to the > std::complex types (I'm not aware of anything else that #281 will lose) > might qualify as ticket 0. > > I'd have to see the exact changes to get a better feel for it, but I would > be a little surprised if these changes could be considered ticket 0, for > two reasons: > > 1. "ticket 0" is reserved for typos or other insignificant changes that do > not change the meaning of the standard. Indeed, the standard would still > be fine without those ticket 0 changes -- the idea is that ticket 0 changes > make the standard read better, etc. > > 2. Adding MPI datatypes for uin64_t (and friends) was a standalone ticket. > I would think that this would be very similar. > Okay. FWIW, I think this request is only to preserve the table "C++-only Named Predefined Datatypes" as in the MPI-2.2 standard, but with namespaced MPI:: names replaced by the MPI_CXX_ prefix (and s/Complex/std::complex/ because that was apparently a typo). From the discussion, it sounds like removing support for these types was an accidental/unrecognized side-effect of bulk-removing the C++ bindings in #281.

1 0

Re: [Mpi-forum] C++ types inaccessible after #281
by Jeff Squyres 26 Jun '12

26 Jun '12

On Jun 26, 2012, at 12:14 PM, Jed Brown wrote: > On Tue, Jun 26, 2012 at 6:13 AM, Jeff Squyres <jsquyres(a)cisco.com> wrote: > There are a few C++-specific datatypes for which it would be worth introducing new C datatypes (e.g., MPI_CXX_COMPLEX). It's too bad that no one realized this until now, because this won't make the MPI-3 deadline. But it could be part of MPI-3.1 or errata, or some such. > > Two people in offline discussions have suggested that C bindings to the std::complex types (I'm not aware of anything else that #281 will lose) might qualify as ticket 0. I'd have to see the exact changes to get a better feel for it, but I would be a little surprised if these changes could be considered ticket 0, for two reasons: 1. "ticket 0" is reserved for typos or other insignificant changes that do not change the meaning of the standard. Indeed, the standard would still be fine without those ticket 0 changes -- the idea is that ticket 0 changes make the standard read better, etc. 2. Adding MPI datatypes for uin64_t (and friends) was a standalone ticket. I would think that this would be very similar. -- Jeff Squyres jsquyres(a)cisco.com For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/

1 0

Re: [Mpi-forum] C++ types inaccessible after #281
by Jed Brown 26 Jun '12

26 Jun '12

On Tue, Jun 26, 2012 at 6:13 AM, Jeff Squyres <jsquyres(a)cisco.com> wrote: > There are a few C++-specific datatypes for which it would be worth > introducing new C datatypes (e.g., MPI_CXX_COMPLEX). It's too bad that no > one realized this until now, because this won't make the MPI-3 deadline. > But it could be part of MPI-3.1 or errata, or some such. Two people in offline discussions have suggested that C bindings to the std::complex types (I'm not aware of anything else that #281 will lose) might qualify as ticket 0.

1 0

Re: [Mpi-forum] [EXTERNAL] Re: C++ types inaccessible after #281
by Jed Brown 26 Jun '12

26 Jun '12

On Tue, Jun 26, 2012 at 6:43 AM, Barrett, Brian W <bwbarre(a)sandia.gov>wrote: > In MPI-2, we supported float, double, and long double as the underlying > type for std::complex with MPI::COMPLEX, MPI::DOUBLE_COMPLEX, and > MPI::LONG_DOUBLE_COMPLEX, respectively. That seems like a good middle > ground... > It's not a middle ground, it's everything supported by the C++ standard.

1 0

Re: [Mpi-forum] C++ types inaccessible after #281
by Jeff Hammond 26 Jun '12

26 Jun '12

Interesting, because http://www.cplusplus.com/reference/std/complex/complex/, which I recognize is not the C++ standard, just says that std::complex is specialized for those types, not restricted to them. Clang was happy to compile std::complex<int> and related for me just now, so we might want to worry about users doing this since we all know MPI users don't read the MPI standard so why should we expect MPI-C++ users to read both the MPI and C++ standards? JeffS was not necessarily suggesting we replace the C++ standard with a human being. He was suggesting someone write the proposal for this feature. I don't see how a standards document is going to produce draft text for the MPI standard. Do you? :-) Jeff On Tue, Jun 26, 2012 at 10:05 AM, Jed Brown <jedbrown(a)mcs.anl.gov> wrote: > On Tue, Jun 26, 2012 at 6:33 AM, Jeff Hammond <jhammond(a)alcf.anl.gov> wrote: >> >> So one issue to consider here is that std::complex is a template, not >> a single type. ï¿½One can presumably stick lots of types inside the >> brackets. > > > Good thing the C++ standards committee nipped that one in the bud: > > 26.4 > The effect of instantiating the template complex for any type other than > float, double, or long doubleï¿½is unspecified. The specializations > complex<float>, complex<double>, and complex<long double> areï¿½literal types > (3.9). > >> >> So the MPI Forum needs to figure out which of the existing C99 types >> that will work with this. ï¿½Do we support std::complex<long double>? >> How about std::complex<uint64_t>? > > > No > >> >> >> My reference is http://www.cplusplus.com/reference/std/complex/complex/, >> btw. > > > As a standards guy, why not use a standard (or a final draft, which you can > have for free)? > > http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3242.pdf > > _______________________________________________ > mpi-forum mailing list > mpi-forum(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum -- Jeff Hammond Argonne Leadership Computing Facility University of Chicago Computation Institute jhammond(a)alcf.anl.gov / (630) 252-5381 http://www.linkedin.com/in/jeffhammond https://wiki.alcf.anl.gov/parts/index.php/User:Jhammond

1 0

Re: [Mpi-forum] C++ types inaccessible after #281
by Jed Brown 26 Jun '12

26 Jun '12

On Tue, Jun 26, 2012 at 6:33 AM, Jeff Hammond <jhammond(a)alcf.anl.gov> wrote: > So one issue to consider here is that std::complex is a template, not > a single type. One can presumably stick lots of types inside the > brackets. > Good thing the C++ standards committee nipped that one in the bud: *26.4* *The effect of instantiating the template complex for any type other than float, double, or long double is unspecified. The specializations complex<float>, complex<double>, and complex<long double> are literal types (3.9).* > So the MPI Forum needs to figure out which of the existing C99 types > that will work with this. Do we support std::complex<long double>? > How about std::complex<uint64_t>? > No > > My reference is http://www.cplusplus.com/reference/std/complex/complex/, > btw. > As a standards guy, why not use a standard (or a final draft, which you can have for free)? http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3242.pdf

1 0

Re: [Mpi-forum] [EXTERNAL] Re: C++ types inaccessible after #281
by Barrett, Brian W 26 Jun '12

26 Jun '12

In MPI-2, we supported float, double, and long double as the underlying type for std::complex with MPI::COMPLEX, MPI::DOUBLE_COMPLEX, and MPI::LONG_DOUBLE_COMPLEX, respectively. That seems like a good middle ground... Brian On 6/26/12 8:33 AM, "Jeff Hammond" <jhammond(a)alcf.anl.gov> wrote: >So one issue to consider here is that std::complex is a template, not >a single type. One can presumably stick lots of types inside the >brackets. > >So the MPI Forum needs to figure out which of the existing C99 types >that will work with this. Do we support std::complex<long double>? >How about std::complex<uint64_t>? > >My reference is http://www.cplusplus.com/reference/std/complex/complex/, >btw. > >Jeff > >On Tue, Jun 26, 2012 at 9:20 AM, Jeff Squyres <jsquyres(a)cisco.com> wrote: >> Not for many years. :-) (/me puts finger on nose) >> >> I originally authored the C++ bindings in the mid-90's (ouch!), when I >>was *much* more active in the C++ community than I am today. >> >> >> >> On Jun 26, 2012, at 10:16 AM, Jeff Hammond wrote: >> >>> Um, aren't you sort of the C++ expert attending the Forum right now? >>> >>> :-) >>> >>> Jeff >>> >>> On Tue, Jun 26, 2012 at 9:13 AM, Jeff Squyres <jsquyres(a)cisco.com> >>>wrote: >>>> Yes, I think that Jed has a good point here. >>>> >>>> There are a few C++-specific datatypes for which it would be worth >>>>introducing new C datatypes (e.g., MPI_CXX_COMPLEX). It's too bad >>>>that no one realized this until now, because this won't make the MPI-3 >>>>deadline. But it could be part of MPI-3.1 or errata, or some such. >>>> >>>> Is there a C++ person out there who could draft a proposal about this? >>>> >>>> >>>> On Jun 26, 2012, at 1:02 AM, Jed Brown wrote: >>>> >>>>> I'm all for getting rid of the C++ bindings and see no compelling >>>>>reason for anyone to use them. >>>>> >>>>> Predefined types are inconsequential to MPI-1 functionality, but >>>>>they are special for one-sided, so availability of types is highly >>>>>relevant to applications. This is a non-cosmetic change imposed by >>>>>#281 that doesn't seem to have been discussed. >>>>> >>>>> On Jun 25, 2012 8:51 PM, "Fab Tillier" <ftillier(a)microsoft.com> >>>>>wrote: >>>>> Ah, thanks, sorry for being dense (I'm not familiar with the C++ >>>>>bindings as we don't ship them for MSMPI). >>>>> >>>>> Ticket 281 removes the entire C++ bindings from the standard, >>>>>MPI::COMPLEX included. You'd need to convince your MPI library >>>>>vendor to keep supplying you with an MPI-2.2 compliant version of the >>>>>C++ bindings (which is still allowed by ticket #281). >>>>> >>>>> Hopefully someone with more experience with the C++ bindings can >>>>>chime in, but I'd expect the answer to your original question is yes, >>>>>ticket #281 intentionally removes all C++ support from MPI-3. >>>>> >>>>> -Fab >>>>> >>>>> Jed Brown wrote on Mon, 25 Jun 2012 at 21:31:55 >>>>> >>>>>> MPI::COMPLEX is not the same as MPI_C_COMPLEX (and the latter is not >>>>>> available on systems without C99). >>>>> >>>>>> On Jun 25, 2012 8:21 PM, "Fab Tillier" <ftillier(a)microsoft.com> >>>>>>wrote: >>>>>> Hi Jed, >>>>>> >>>>>> Maybe I'm not following you, but what predefined MPI datatype >>>>>>supports >>>>>> std::complex today? My understanding is that there is no support >>>>>>for >>>>>> std::complex in MPI today, and that ticket 281 does not change this >>>>>>at >>>>>> all. >>>>>> >>>>>> Thanks, >>>>>> -Fab >>>>>> >>>>>> Jed Brown wrote on Mon, 25 Jun 2012 at 20:42:53 >>>>>> >>>>>>> I see that std::complex will not be accessible in a predefined >>>>>>>type if >>>>>>> #281 passes in its current form. Do you intend for there to be a >>>>>>> supported/recommended way for a C++ caller to get a complex >>>>>>>predefined >>>>>>> (so that one-sided can be used) type? Note that C99 is not a >>>>>>>subset of >>>>>>> any C++ and C99 complex cannot be used portably from C++ (even with >>>>>>> compiler suites that happen to support both). >>>>>>> >>>>>>> On Jun 25, 2012 7:07 PM, "Fab Tillier" <ftillier(a)microsoft.com >>>>>>> <mailto:[email protected]> > wrote: Hi Jed, >>>>>>> >>>>>>> Jed Brown wrote on Mon, 25 Jun 2012 at 16:43:54 >>>>>>> >>>>>>>> Is it intended that #281 would make std::complex inaccessible? >>>>>>>>Should >>>>>>>> there be a MPI_CXX_COMPLEX, MPI_CXX_DOUBLE_COMPLEX, >>>>>>>> MPI_CXX_LONG_DOUBLE_COMPLEX, and (possibly) MPI_CXX_BOOL? >>>>>>> >>>>>>> I don't think ticket #281 does anything with respect to >>>>>>>std::complex >>>>>>> accessibility. The MPI standard defines the MPI_C_COMPLEX and >>>>>>>family >>>>>>> as mapping to the C99 _Complex type (see Table 3.2, Predefined MPI >>>>>>> datatypes corresponding to C datatypes). The standard is clear >>>>>>>that >>>>>>> there are no corresponding C++ bindings. >>>>>>> >>>>>>>> Note that C99 complex is not a substitute because Microsoft does >>>>>>>>not >>>>>>>> implement it and it has different semantics regarding numerical >>>>>>>> stability. (Well, C99 semantics are specified and useful, >>>>>>>>std::complex >>>>>>>> stability is not specified by any C++ standard and in practice, >>>>>>>>is not >>>>>>>> implemented in a stable way.) This is not such a big deal for use >>>>>>>>with >>>>>>>> collectives because user-defined MPI_Ops can be used, but only >>>>>>>> predefined types can be used with one-sided operations, so >>>>>>>>presence of >>>>>>>> complex datatypes (or lack thereof) is relevant to applications. >>>>>>> >>>>>>> The fact that Microsoft does not support C99 is lame. You can >>>>>>>probably >>>>>>> get away using std::complex as a stand-in for C99 _Complex (but >>>>>>>then >>>>>>> you're on your own), or move to a compiler that does (I believe the >>>>>>> Intel compiler supports C99 on Windows). >>>>>>> >>>>>>> -Fab >>>>>>> >>>>>> >>>>> _______________________________________________ >>>>> mpi-forum mailing list >>>>> mpi-forum(a)lists.mpi-forum.org >>>>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum >>>> >>>> >>>> -- >>>> Jeff Squyres >>>> jsquyres(a)cisco.com >>>> For corporate legal information go to: >>>>http://www.cisco.com/web/about/doing_business/legal/cri/ >>>> >>>> >>>> _______________________________________________ >>>> mpi-forum mailing list >>>> mpi-forum(a)lists.mpi-forum.org >>>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum >>> >>> >>> >>> -- >>> Jeff Hammond >>> Argonne Leadership Computing Facility >>> University of Chicago Computation Institute >>> jhammond(a)alcf.anl.gov / (630) 252-5381 >>> http://www.linkedin.com/in/jeffhammond >>> https://wiki.alcf.anl.gov/parts/index.php/User:Jhammond >>> >>> _______________________________________________ >>> mpi-forum mailing list >>> mpi-forum(a)lists.mpi-forum.org >>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum >> >> >> -- >> Jeff Squyres >> jsquyres(a)cisco.com >> For corporate legal information go to: >>http://www.cisco.com/web/about/doing_business/legal/cri/ >> >> >> _______________________________________________ >> mpi-forum mailing list >> mpi-forum(a)lists.mpi-forum.org >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum > > > >-- >Jeff Hammond >Argonne Leadership Computing Facility >University of Chicago Computation Institute >jhammond(a)alcf.anl.gov / (630) 252-5381 >http://www.linkedin.com/in/jeffhammond >https://wiki.alcf.anl.gov/parts/index.php/User:Jhammond > >_______________________________________________ >mpi-forum mailing list >mpi-forum(a)lists.mpi-forum.org >http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum > > -- Brian W. Barrett Dept. 1423: Scalable System Software Sandia National Laboratories

1 0

Re: [Mpi-forum] C++ types inaccessible after #281
by Jeff Hammond 26 Jun '12

26 Jun '12

So one issue to consider here is that std::complex is a template, not a single type. One can presumably stick lots of types inside the brackets. So the MPI Forum needs to figure out which of the existing C99 types that will work with this. Do we support std::complex<long double>? How about std::complex<uint64_t>? My reference is http://www.cplusplus.com/reference/std/complex/complex/, btw. Jeff On Tue, Jun 26, 2012 at 9:20 AM, Jeff Squyres <jsquyres(a)cisco.com> wrote: > Not for many years. ï¿½:-) ï¿½(/me puts finger on nose) > > I originally authored the C++ bindings in the mid-90's (ouch!), when I was *much* more active in the C++ community than I am today. > > > > On Jun 26, 2012, at 10:16 AM, Jeff Hammond wrote: > >> Um, aren't you sort of the C++ expert attending the Forum right now? >> >> :-) >> >> Jeff >> >> On Tue, Jun 26, 2012 at 9:13 AM, Jeff Squyres <jsquyres(a)cisco.com> wrote: >>> Yes, I think that Jed has a good point here. >>> >>> There are a few C++-specific datatypes for which it would be worth introducing new C datatypes (e.g., MPI_CXX_COMPLEX). ï¿½It's too bad that no one realized this until now, because this won't make the MPI-3 deadline. ï¿½But it could be part of MPI-3.1 or errata, or some such. >>> >>> Is there a C++ person out there who could draft a proposal about this? >>> >>> >>> On Jun 26, 2012, at 1:02 AM, Jed Brown wrote: >>> >>>> I'm all for getting rid of the C++ bindings and see no compelling reason for anyone to use them. >>>> >>>> Predefined types are inconsequential to MPI-1 functionality, but they are special for one-sided, so availability of types is highly relevant to applications. This is a non-cosmetic change imposed by #281 that doesn't seem to have been discussed. >>>> >>>> On Jun 25, 2012 8:51 PM, "Fab Tillier" <ftillier(a)microsoft.com> wrote: >>>> Ah, thanks, sorry for being dense (I'm not familiar with the C++ bindings as we don't ship them for MSMPI). >>>> >>>> Ticket 281 removes the entire C++ bindings from the standard, MPI::COMPLEX included. ï¿½You'd need to convince your MPI library vendor to keep supplying you with an MPI-2.2 compliant version of the C++ bindings (which is still allowed by ticket #281). >>>> >>>> Hopefully someone with more experience with the C++ bindings can chime in, but I'd expect the answer to your original question is yes, ticket #281 intentionally removes all C++ support from MPI-3. >>>> >>>> -Fab >>>> >>>> Jed Brown wrote on Mon, 25 Jun 2012 at 21:31:55 >>>> >>>>> MPI::COMPLEX is not the same as MPI_C_COMPLEX (and the latter is not >>>>> available on systems without C99). >>>> >>>>> On Jun 25, 2012 8:21 PM, "Fab Tillier" <ftillier(a)microsoft.com> wrote: >>>>> Hi Jed, >>>>> >>>>> Maybe I'm not following you, but what predefined MPI datatype supports >>>>> std::complex today? ï¿½My understanding is that there is no support for >>>>> std::complex in MPI today, and that ticket 281 does not change this at >>>>> all. >>>>> >>>>> Thanks, >>>>> -Fab >>>>> >>>>> Jed Brown wrote on Mon, 25 Jun 2012 at 20:42:53 >>>>> >>>>>> I see that std::complex will not be accessible in a predefined type if >>>>>> #281 passes in its current form. Do you intend for there to be a >>>>>> supported/recommended way for a C++ caller to get a complex predefined >>>>>> (so that one-sided can be used) type? Note that C99 is not a subset of >>>>>> any C++ and C99 complex cannot be used portably from C++ (even with >>>>>> compiler suites that happen to support both). >>>>>> >>>>>> On Jun 25, 2012 7:07 PM, "Fab Tillier" <ftillier(a)microsoft.com >>>>>> <mailto:[email protected]> > wrote: Hi Jed, >>>>>> >>>>>> Jed Brown wrote on Mon, 25 Jun 2012 at 16:43:54 >>>>>> >>>>>>> Is it intended that #281 would make std::complex inaccessible? Should >>>>>>> there be a MPI_CXX_COMPLEX, MPI_CXX_DOUBLE_COMPLEX, >>>>>>> MPI_CXX_LONG_DOUBLE_COMPLEX, and (possibly) MPI_CXX_BOOL? >>>>>> >>>>>> I don't think ticket #281 does anything with respect to std::complex >>>>>> accessibility. ï¿½The MPI standard defines the MPI_C_COMPLEX and family >>>>>> as mapping to the C99 _Complex type (see Table 3.2, Predefined MPI >>>>>> datatypes corresponding to C datatypes). ï¿½The standard is clear that >>>>>> there are no corresponding C++ bindings. >>>>>> >>>>>>> Note that C99 complex is not a substitute because Microsoft does not >>>>>>> implement it and it has different semantics regarding numerical >>>>>>> stability. (Well, C99 semantics are specified and useful, std::complex >>>>>>> stability is not specified by any C++ standard and in practice, is not >>>>>>> implemented in a stable way.) This is not such a big deal for use with >>>>>>> collectives because user-defined MPI_Ops can be used, but only >>>>>>> predefined types can be used with one-sided operations, so presence of >>>>>>> complex datatypes (or lack thereof) is relevant to applications. >>>>>> >>>>>> The fact that Microsoft does not support C99 is lame. ï¿½You can probably >>>>>> get away using std::complex as a stand-in for C99 _Complex (but then >>>>>> you're on your own), or move to a compiler that does (I believe the >>>>>> Intel compiler supports C99 on Windows). >>>>>> >>>>>> -Fab >>>>>> >>>>> >>>> _______________________________________________ >>>> mpi-forum mailing list >>>> mpi-forum(a)lists.mpi-forum.org >>>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum >>> >>> >>> -- >>> Jeff Squyres >>> jsquyres(a)cisco.com >>> For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/ >>> >>> >>> _______________________________________________ >>> mpi-forum mailing list >>> mpi-forum(a)lists.mpi-forum.org >>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum >> >> >> >> -- >> Jeff Hammond >> Argonne Leadership Computing Facility >> University of Chicago Computation Institute >> jhammond(a)alcf.anl.gov / (630) 252-5381 >> http://www.linkedin.com/in/jeffhammond >> https://wiki.alcf.anl.gov/parts/index.php/User:Jhammond >> >> _______________________________________________ >> mpi-forum mailing list >> mpi-forum(a)lists.mpi-forum.org >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum > > > -- > Jeff Squyres > jsquyres(a)cisco.com > For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/ > > > _______________________________________________ > mpi-forum mailing list > mpi-forum(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum -- Jeff Hammond Argonne Leadership Computing Facility University of Chicago Computation Institute jhammond(a)alcf.anl.gov / (630) 252-5381 http://www.linkedin.com/in/jeffhammond https://wiki.alcf.anl.gov/parts/index.php/User:Jhammond

1 0

Re: [Mpi-forum] C++ types inaccessible after #281
by Jeff Squyres 26 Jun '12

26 Jun '12

Not for many years. :-) (/me puts finger on nose) I originally authored the C++ bindings in the mid-90's (ouch!), when I was *much* more active in the C++ community than I am today. On Jun 26, 2012, at 10:16 AM, Jeff Hammond wrote: > Um, aren't you sort of the C++ expert attending the Forum right now? > > :-) > > Jeff > > On Tue, Jun 26, 2012 at 9:13 AM, Jeff Squyres <jsquyres(a)cisco.com> wrote: >> Yes, I think that Jed has a good point here. >> >> There are a few C++-specific datatypes for which it would be worth introducing new C datatypes (e.g., MPI_CXX_COMPLEX). It's too bad that no one realized this until now, because this won't make the MPI-3 deadline. But it could be part of MPI-3.1 or errata, or some such. >> >> Is there a C++ person out there who could draft a proposal about this? >> >> >> On Jun 26, 2012, at 1:02 AM, Jed Brown wrote: >> >>> I'm all for getting rid of the C++ bindings and see no compelling reason for anyone to use them. >>> >>> Predefined types are inconsequential to MPI-1 functionality, but they are special for one-sided, so availability of types is highly relevant to applications. This is a non-cosmetic change imposed by #281 that doesn't seem to have been discussed. >>> >>> On Jun 25, 2012 8:51 PM, "Fab Tillier" <ftillier(a)microsoft.com> wrote: >>> Ah, thanks, sorry for being dense (I'm not familiar with the C++ bindings as we don't ship them for MSMPI). >>> >>> Ticket 281 removes the entire C++ bindings from the standard, MPI::COMPLEX included. You'd need to convince your MPI library vendor to keep supplying you with an MPI-2.2 compliant version of the C++ bindings (which is still allowed by ticket #281). >>> >>> Hopefully someone with more experience with the C++ bindings can chime in, but I'd expect the answer to your original question is yes, ticket #281 intentionally removes all C++ support from MPI-3. >>> >>> -Fab >>> >>> Jed Brown wrote on Mon, 25 Jun 2012 at 21:31:55 >>> >>>> MPI::COMPLEX is not the same as MPI_C_COMPLEX (and the latter is not >>>> available on systems without C99). >>> >>>> On Jun 25, 2012 8:21 PM, "Fab Tillier" <ftillier(a)microsoft.com> wrote: >>>> Hi Jed, >>>> >>>> Maybe I'm not following you, but what predefined MPI datatype supports >>>> std::complex today? My understanding is that there is no support for >>>> std::complex in MPI today, and that ticket 281 does not change this at >>>> all. >>>> >>>> Thanks, >>>> -Fab >>>> >>>> Jed Brown wrote on Mon, 25 Jun 2012 at 20:42:53 >>>> >>>>> I see that std::complex will not be accessible in a predefined type if >>>>> #281 passes in its current form. Do you intend for there to be a >>>>> supported/recommended way for a C++ caller to get a complex predefined >>>>> (so that one-sided can be used) type? Note that C99 is not a subset of >>>>> any C++ and C99 complex cannot be used portably from C++ (even with >>>>> compiler suites that happen to support both). >>>>> >>>>> On Jun 25, 2012 7:07 PM, "Fab Tillier" <ftillier(a)microsoft.com >>>>> <mailto:[email protected]> > wrote: Hi Jed, >>>>> >>>>> Jed Brown wrote on Mon, 25 Jun 2012 at 16:43:54 >>>>> >>>>>> Is it intended that #281 would make std::complex inaccessible? Should >>>>>> there be a MPI_CXX_COMPLEX, MPI_CXX_DOUBLE_COMPLEX, >>>>>> MPI_CXX_LONG_DOUBLE_COMPLEX, and (possibly) MPI_CXX_BOOL? >>>>> >>>>> I don't think ticket #281 does anything with respect to std::complex >>>>> accessibility. The MPI standard defines the MPI_C_COMPLEX and family >>>>> as mapping to the C99 _Complex type (see Table 3.2, Predefined MPI >>>>> datatypes corresponding to C datatypes). The standard is clear that >>>>> there are no corresponding C++ bindings. >>>>> >>>>>> Note that C99 complex is not a substitute because Microsoft does not >>>>>> implement it and it has different semantics regarding numerical >>>>>> stability. (Well, C99 semantics are specified and useful, std::complex >>>>>> stability is not specified by any C++ standard and in practice, is not >>>>>> implemented in a stable way.) This is not such a big deal for use with >>>>>> collectives because user-defined MPI_Ops can be used, but only >>>>>> predefined types can be used with one-sided operations, so presence of >>>>>> complex datatypes (or lack thereof) is relevant to applications. >>>>> >>>>> The fact that Microsoft does not support C99 is lame. You can probably >>>>> get away using std::complex as a stand-in for C99 _Complex (but then >>>>> you're on your own), or move to a compiler that does (I believe the >>>>> Intel compiler supports C99 on Windows). >>>>> >>>>> -Fab >>>>> >>>> >>> _______________________________________________ >>> mpi-forum mailing list >>> mpi-forum(a)lists.mpi-forum.org >>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum >> >> >> -- >> Jeff Squyres >> jsquyres(a)cisco.com >> For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/ >> >> >> _______________________________________________ >> mpi-forum mailing list >> mpi-forum(a)lists.mpi-forum.org >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum > > > > -- > Jeff Hammond > Argonne Leadership Computing Facility > University of Chicago Computation Institute > jhammond(a)alcf.anl.gov / (630) 252-5381 > http://www.linkedin.com/in/jeffhammond > https://wiki.alcf.anl.gov/parts/index.php/User:Jhammond > > _______________________________________________ > mpi-forum mailing list > mpi-forum(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum -- Jeff Squyres jsquyres(a)cisco.com For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/

1 0

Re: [Mpi-forum] C++ types inaccessible after #281
by Jeff Hammond 26 Jun '12

26 Jun '12

Um, aren't you sort of the C++ expert attending the Forum right now? :-) Jeff On Tue, Jun 26, 2012 at 9:13 AM, Jeff Squyres <jsquyres(a)cisco.com> wrote: > Yes, I think that Jed has a good point here. > > There are a few C++-specific datatypes for which it would be worth introducing new C datatypes (e.g., MPI_CXX_COMPLEX). ï¿½It's too bad that no one realized this until now, because this won't make the MPI-3 deadline. ï¿½But it could be part of MPI-3.1 or errata, or some such. > > Is there a C++ person out there who could draft a proposal about this? > > > On Jun 26, 2012, at 1:02 AM, Jed Brown wrote: > >> I'm all for getting rid of the C++ bindings and see no compelling reason for anyone to use them. >> >> Predefined types are inconsequential to MPI-1 functionality, but they are special for one-sided, so availability of types is highly relevant to applications. This is a non-cosmetic change imposed by #281 that doesn't seem to have been discussed. >> >> On Jun 25, 2012 8:51 PM, "Fab Tillier" <ftillier(a)microsoft.com> wrote: >> Ah, thanks, sorry for being dense (I'm not familiar with the C++ bindings as we don't ship them for MSMPI). >> >> Ticket 281 removes the entire C++ bindings from the standard, MPI::COMPLEX included. ï¿½You'd need to convince your MPI library vendor to keep supplying you with an MPI-2.2 compliant version of the C++ bindings (which is still allowed by ticket #281). >> >> Hopefully someone with more experience with the C++ bindings can chime in, but I'd expect the answer to your original question is yes, ticket #281 intentionally removes all C++ support from MPI-3. >> >> -Fab >> >> Jed Brown wrote on Mon, 25 Jun 2012 at 21:31:55 >> >> > MPI::COMPLEX is not the same as MPI_C_COMPLEX (and the latter is not >> > available on systems without C99). >> >> > On Jun 25, 2012 8:21 PM, "Fab Tillier" <ftillier(a)microsoft.com> wrote: >> > Hi Jed, >> > >> > Maybe I'm not following you, but what predefined MPI datatype supports >> > std::complex today? ï¿½My understanding is that there is no support for >> > std::complex in MPI today, and that ticket 281 does not change this at >> > all. >> > >> > Thanks, >> > -Fab >> > >> > Jed Brown wrote on Mon, 25 Jun 2012 at 20:42:53 >> > >> >> I see that std::complex will not be accessible in a predefined type if >> >> #281 passes in its current form. Do you intend for there to be a >> >> supported/recommended way for a C++ caller to get a complex predefined >> >> (so that one-sided can be used) type? Note that C99 is not a subset of >> >> any C++ and C99 complex cannot be used portably from C++ (even with >> >> compiler suites that happen to support both). >> >> >> >> On Jun 25, 2012 7:07 PM, "Fab Tillier" <ftillier(a)microsoft.com >> >> <mailto:[email protected]> > wrote: Hi Jed, >> >> >> >> Jed Brown wrote on Mon, 25 Jun 2012 at 16:43:54 >> >> >> >>> Is it intended that #281 would make std::complex inaccessible? Should >> >>> there be a MPI_CXX_COMPLEX, MPI_CXX_DOUBLE_COMPLEX, >> >>> MPI_CXX_LONG_DOUBLE_COMPLEX, and (possibly) MPI_CXX_BOOL? >> >> >> >> I don't think ticket #281 does anything with respect to std::complex >> >> accessibility. ï¿½The MPI standard defines the MPI_C_COMPLEX and family >> >> as mapping to the C99 _Complex type (see Table 3.2, Predefined MPI >> >> datatypes corresponding to C datatypes). ï¿½The standard is clear that >> >> there are no corresponding C++ bindings. >> >> >> >>> Note that C99 complex is not a substitute because Microsoft does not >> >>> implement it and it has different semantics regarding numerical >> >>> stability. (Well, C99 semantics are specified and useful, std::complex >> >>> stability is not specified by any C++ standard and in practice, is not >> >>> implemented in a stable way.) This is not such a big deal for use with >> >>> collectives because user-defined MPI_Ops can be used, but only >> >>> predefined types can be used with one-sided operations, so presence of >> >>> complex datatypes (or lack thereof) is relevant to applications. >> >> >> >> The fact that Microsoft does not support C99 is lame. ï¿½You can probably >> >> get away using std::complex as a stand-in for C99 _Complex (but then >> >> you're on your own), or move to a compiler that does (I believe the >> >> Intel compiler supports C99 on Windows). >> >> >> >> -Fab >> >> >> > >> _______________________________________________ >> mpi-forum mailing list >> mpi-forum(a)lists.mpi-forum.org >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum > > > -- > Jeff Squyres > jsquyres(a)cisco.com > For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/ > > > _______________________________________________ > mpi-forum mailing list > mpi-forum(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum -- Jeff Hammond Argonne Leadership Computing Facility University of Chicago Computation Institute jhammond(a)alcf.anl.gov / (630) 252-5381 http://www.linkedin.com/in/jeffhammond https://wiki.alcf.anl.gov/parts/index.php/User:Jhammond

1 0