July 2020 - mpiwg-fortran - mailman3.cels.anl.gov

Re: [MPIWG Fortran] Question about MPI_Status_f2f08() and _f082f()
by Rolf Rabenseifner 15 Jul '20

15 Jul '20

Dear Hubert and Jeff, > Ø I'm a little curious as to why we conspicuously left it out of mpif.h. I expect, we didn't want that mpif.h must include the overloading of the operators .NE. and .EQ. for all TYPE(MPI_....). I'm not sure whether such declaration is allowed outside of a module. The use of mpif.h was already deprecated, i.e., if somebody wants to use mpi_f08 stuff in old code, he or she must first substitute the include mpif.h by use mpi. > ... But it was never intended that the programmer > does this transformation within the old Fortran subroutine (and old Fortran 77 > compiler or using Fortran 77 language kind wouldn’t support Type(MPI_Status) in > mpif.h). It was intended that you can convert from old INTEGER variable or array to new TYPE(MPI_...) within source code using the mpi module or using the mpi_f08 module. But we excluded mpif.h because it need not to provide compile-time argument checking and it use is therefore "strongly discouraged". Why should we add something to this "strongly discouraged" mpif.h area. And, to add it in a later version of MPI is simple. To remove it later is not backward compatible. This may be another reason for not adding it to mpif.h Best regards Rolf ----- Original Message ----- > From: "MPI-WG Fortran working group" <mpiwg-fortran(a)lists.mpi-forum.org> > To: "MPI-WG Fortran working group" <mpiwg-fortran(a)lists.mpi-forum.org> > Cc: "Hubert Ritzdorf" <Hubert.Ritzdorf(a)EMEA.NEC.COM> > Sent: Wednesday, July 15, 2020 6:15:58 PM > Subject: Re: [MPIWG Fortran] Question about MPI_Status_f2f08() and _f082f() > Ø I'm a little curious as to why we conspicuously left it out of mpif.h. > > > > When I remember correctly, MPI_Status_f2f08() and _f082f() subroutines were for > a smooth transition from Fortran 77 to Fortran 08. > > I.e. if a programmer has changed some functions from old Fortran to Fortran 08 > and uses other libraries or subroutines which still use the old Fortran status > as input or output argument, it was possible to transfer the old Fortran status > within the Fortran 08 subroutine. But it was never intended that the programmer > does this transformation within the old Fortran subroutine (and old Fortran 77 > compiler or using Fortran 77 language kind wouldn’t support Type(MPI_Status) in > mpif.h). > > > > Hubert > > > > > From: mpiwg-fortran [mailto:[email protected]] On Behalf > Of Jeff Squyres (jsquyres) via mpiwg-fortran > Sent: Wednesday, July 15, 2020 5:26 PM > To: MPI Fortran WG <mpiwg-fortran(a)lists.mpi-forum.org> > Cc: Jeff Squyres (jsquyres) <jsquyres(a)cisco.com> > Subject: Re: [MPIWG Fortran] Question about MPI_Status_f2f08() and _f082f() > > > > > Just to bring this thread back on-track about TYPE(MPI_Status)... > > > > > > Off list, I asked Rolf R. about this issue -- he cited the same things I did, > plus a few more: > > > > > > > > > On your question on TYPE(MPI_Status), TYPE(MPI_Comm), ...: > > > MPI-3.1 > > > - page 607 lines 18-24 require these types and the overloaded > > > operators .EQ./.NE. for mpi_f08 module > > > - page 609 lines 34-36 require these types and the overloaded > > > operators .EQ./.NE. for mpi module > > > - there is no such text on page 611-612 on mpif.h > > > > > > And page 802 lines 9-15 also Show that it was never intented to add > > > These types anf routines to old mpif.h. > > > > > > It would be helpful, to add at least on page 657 line 11 > > > > > > C, some in both C and Fortran (only in the mpi_f08 and mpi modules). > > > > > > Can you fix this with an one-vote-bug-fix-issue? > > > Because it is not good if the information must be taken from the change-log. > > > > > > --- > > > Summary: > > > > > > Does this mean that TYPE(MPI_Status) > > > [ and the Fortran routines MPI_STATUS_F2F08 and _F082F ] > > > needs to be defined in > > > > > > - mpif.h? NO > > > - and the mpi module? YES > > > > > > Bug-fix needed in MPI-3.1 page 657 line 11: add > > > "(only in the mpi_f08 and mpi modules)" at the end. > > > > > > I hope this helps. > > > > > > Best regards > > > Rolf > > > > > > So I think there's at least a clarification here: the TYPE(MPI_Status) and > associated functions is -- at a minimum -- supposed to be in the mpi module. > > > > > > I'm a little curious as to why we conspicuously left it out of mpif.h. > > > > > > Bill: this is somewhat counter to the clarification you proposed. > > > > > > Are you ok with this? I think the text in the standard supports what Rolf > proposes. > > > > > > > > > > > > > > > > On Jul 10, 2020, at 12:11 PM, Jeff Squyres (jsquyres) via mpiwg-fortran < [ > mailto:[email protected] | mpiwg-fortran(a)lists.mpi-forum.org ] > > wrote: > > > > > > On Jul 10, 2020, at 11:18 AM, Bill Long < [ mailto:[email protected] | > longb(a)cray.com ] > wrote: > > > > > > > In the “change” section there is this txt: > > • Within the mpi_08 Fortran module, the status was defined as TYPE(MPI_Status). > Additionally, within both the mpi and the mpi_f08 modules, the constants > MPI_STATUS_SIZE, MPI_SOURCE, MPI_TAG, MPI_ERROR, and TYPE(MPI_Status) are > defined. New conversion routines were added: MPI_STATUS_F2F08, > MPI_STATUS_F082F, MPI_Status_c2f08, and MPI_Status_f082c, In mpi.h, the new > type MPI_F08_status, and the external variables MPI_F08_STATUS_IGNORE and > MPI_F08_STATUSES_IGNORE were added. > > > Good point. > > Just to be clear, you're referring to the changelog section in MPI-3.1, > specifically bullet 30 on p802. > > That being said: > > - the changelog is non-binding ...but it does indicate our intent from that time > - the changelog text states that the mpi module has TYPE(MPI_Status) -- but it > does not say it was added to mpif.h > > > > > > > 1) Why would the F08 status be defined different from the C definition? (If that > were the case, conversions between f08 and C would be irrelevant). > > > I remember that there was a lot of discussion about this at the time, which is > what resulted in Figure 17.1. > > I know there were discussions about making the F08 and C statuses the same, but > for some reason we chose not to mandate it. Perhaps we wanted to allow > implementations to do whatever they wanted...? (e.g., allow Status_c2f08 be a > no-op if the implementation wanted to, but not mandate it) > > > > > > > 2) \begin{unpopular} Why are the legacy mpi module and mpif.h still included in > the spec? These are embarrassingly obsolete. If this was fixed, none of the > above mentioned conversion routines would be needed. \end(unpopular} > > > > I would love it if we could ditch -- at a minimum -- mpif.h. > > However, there's oodles of legacy code out there that uses it. That's why even > deprecating it gets shouted down at Forum meetings. > > -- > Jeff Squyres > [ mailto:[email protected] | jsquyres(a)cisco.com ] > > _______________________________________________ > mpiwg-fortran mailing list > [ mailto:[email protected] | mpiwg-fortran(a)lists.mpi-forum.org ] > [ https://lists.mpi-forum.org/mailman/listinfo/mpiwg-fortran | > https://lists.mpi-forum.org/mailman/listinfo/mpiwg-fortran ] > > > > > > > -- > Jeff Squyres > [ mailto:[email protected] | jsquyres(a)cisco.com ] > > > > > > > > Click [ > https://www.mailcontrol.com/sr/HjGlLo600jvGX2PQPOmvUuToodiywuSOXttD3rcwKg2C… > | > here ] to report this email as spam. > > _______________________________________________ > mpiwg-fortran mailing list > mpiwg-fortran(a)lists.mpi-forum.org > https://lists.mpi-forum.org/mailman/listinfo/mpiwg-fortran -- Dr. Rolf Rabenseifner . . . . . . . . . .. email rabenseifner(a)hlrs.de . High Performance Computing Center (HLRS) . phone ++49(0)711/685-65530 . University of Stuttgart . . . . . . . . .. fax ++49(0)711 / 685-65832 . Head of Dpmt Parallel Computing . . . www.hlrs.de/people/rabenseifner . Nobelstr. 19, D-70550 Stuttgart, Germany . . . . (Office: Room 1.307) .

1 0

Re: [MPIWG Fortran] Question about MPI_Status_f2f08() and _f082f()
by Hubert Ritzdorf 15 Jul '20

15 Jul '20

Ø I'm a little curious as to why we conspicuously left it out of mpif.h. When I remember correctly, MPI_Status_f2f08() and _f082f() subroutines were for a smooth transition from Fortran 77 to Fortran 08. I.e. if a programmer has changed some functions from old Fortran to Fortran 08 and uses other libraries or subroutines which still use the old Fortran status as input or output argument, it was possible to transfer the old Fortran status within the Fortran 08 subroutine. But it was never intended that the programmer does this transformation within the old Fortran subroutine (and old Fortran 77 compiler or using Fortran 77 language kind wouldn’t support Type(MPI_Status) in mpif.h). Hubert From: mpiwg-fortran [mailto:[email protected]] On Behalf Of Jeff Squyres (jsquyres) via mpiwg-fortran Sent: Wednesday, July 15, 2020 5:26 PM To: MPI Fortran WG <mpiwg-fortran(a)lists.mpi-forum.org> Cc: Jeff Squyres (jsquyres) <jsquyres(a)cisco.com> Subject: Re: [MPIWG Fortran] Question about MPI_Status_f2f08() and _f082f() Just to bring this thread back on-track about TYPE(MPI_Status)... Off list, I asked Rolf R. about this issue -- he cited the same things I did, plus a few more: On your question on TYPE(MPI_Status), TYPE(MPI_Comm), ...: MPI-3.1 - page 607 lines 18-24 require these types and the overloaded operators .EQ./.NE. for mpi_f08 module - page 609 lines 34-36 require these types and the overloaded operators .EQ./.NE. for mpi module - there is no such text on page 611-612 on mpif.h And page 802 lines 9-15 also Show that it was never intented to add These types anf routines to old mpif.h. It would be helpful, to add at least on page 657 line 11 C, some in both C and Fortran (only in the mpi_f08 and mpi modules). Can you fix this with an one-vote-bug-fix-issue? Because it is not good if the information must be taken from the change-log. --- Summary: Does this mean that TYPE(MPI_Status) [ and the Fortran routines MPI_STATUS_F2F08 and _F082F ] needs to be defined in - mpif.h? NO - and the mpi module? YES Bug-fix needed in MPI-3.1 page 657 line 11: add "(only in the mpi_f08 and mpi modules)" at the end. I hope this helps. Best regards Rolf So I think there's at least a clarification here: the TYPE(MPI_Status) and associated functions is -- at a minimum -- supposed to be in the mpi module. I'm a little curious as to why we conspicuously left it out of mpif.h. Bill: this is somewhat counter to the clarification you proposed. Are you ok with this? I think the text in the standard supports what Rolf proposes. On Jul 10, 2020, at 12:11 PM, Jeff Squyres (jsquyres) via mpiwg-fortran <mpiwg-fortran(a)lists.mpi-forum.org<mailto:[email protected]>> wrote: On Jul 10, 2020, at 11:18 AM, Bill Long <longb(a)cray.com<mailto:[email protected]>> wrote: In the “change” section there is this txt: • Within the mpi_08 Fortran module, the status was defined as TYPE(MPI_Status). Additionally, within both the mpi and the mpi_f08 modules, the constants MPI_STATUS_SIZE, MPI_SOURCE, MPI_TAG, MPI_ERROR, and TYPE(MPI_Status) are defined. New conversion routines were added: MPI_STATUS_F2F08, MPI_STATUS_F082F, MPI_Status_c2f08, and MPI_Status_f082c, In mpi.h, the new type MPI_F08_status, and the external variables MPI_F08_STATUS_IGNORE and MPI_F08_STATUSES_IGNORE were added. Good point. Just to be clear, you're referring to the changelog section in MPI-3.1, specifically bullet 30 on p802. That being said: - the changelog is non-binding ...but it does indicate our intent from that time - the changelog text states that the mpi module has TYPE(MPI_Status) -- but it does not say it was added to mpif.h 1) Why would the F08 status be defined different from the C definition? (If that were the case, conversions between f08 and C would be irrelevant). I remember that there was a lot of discussion about this at the time, which is what resulted in Figure 17.1. I know there were discussions about making the F08 and C statuses the same, but for some reason we chose not to mandate it. Perhaps we wanted to allow implementations to do whatever they wanted...? (e.g., allow Status_c2f08 be a no-op if the implementation wanted to, but not mandate it) 2) \begin{unpopular} Why are the legacy mpi module and mpif.h still included in the spec? These are embarrassingly obsolete. If this was fixed, none of the above mentioned conversion routines would be needed. \end(unpopular} I would love it if we could ditch -- at a minimum -- mpif.h. However, there's oodles of legacy code out there that uses it. That's why even deprecating it gets shouted down at Forum meetings. -- Jeff Squyres jsquyres(a)cisco.com<mailto:[email protected]> _______________________________________________ mpiwg-fortran mailing list mpiwg-fortran(a)lists.mpi-forum.org<mailto:[email protected]> https://lists.mpi-forum.org/mailman/listinfo/mpiwg-fortran -- Jeff Squyres jsquyres(a)cisco.com<mailto:[email protected]> Click here<https://www.mailcontrol.com/sr/HjGlLo600jvGX2PQPOmvUuToodiywuSOXttD3rcwKg2C…> to report this email as spam.

1 0

Re: [MPIWG Fortran] Question about MPI_Status_f2f08() and _f082f()
by William Gropp 15 Jul '20

15 Jul '20

I’m ok with a clarification that excludes TYPE(MPI_Status) from mpif.h , as Rolf proposes. Bill William Gropp Director and Chief Scientist, NCSA Thomas M. Siebel Chair in Computer Science University of Illinois Urbana-Champaign > On Jul 15, 2020, at 10:25 AM, Jeff Squyres (jsquyres) via mpiwg-fortran <mpiwg-fortran(a)lists.mpi-forum.org> wrote: > > Just to bring this thread back on-track about TYPE(MPI_Status)... > > Off list, I asked Rolf R. about this issue -- he cited the same things I did, plus a few more: > > On your question on TYPE(MPI_Status), TYPE(MPI_Comm), ...: > MPI-3.1 > - page 607 lines 18-24 require these types and the overloaded > operators .EQ./.NE. for mpi_f08 module > - page 609 lines 34-36 require these types and the overloaded > operators .EQ./.NE. for mpi module > - there is no such text on page 611-612 on mpif.h > > And page 802 lines 9-15 also Show that it was never intented to add > These types anf routines to old mpif.h. > > It would be helpful, to add at least on page 657 line 11 > > C, some in both C and Fortran (only in the mpi_f08 and mpi modules). > > Can you fix this with an one-vote-bug-fix-issue? > Because it is not good if the information must be taken from the change-log. > > --- > Summary: > > Does this mean that TYPE(MPI_Status) > [ and the Fortran routines MPI_STATUS_F2F08 and _F082F ] > needs to be defined in > > - mpif.h? NO > - and the mpi module? YES > > Bug-fix needed in MPI-3.1 page 657 line 11: add > "(only in the mpi_f08 and mpi modules)" at the end. > > I hope this helps. > > Best regards > Rolf > > So I think there's at least a clarification here: the TYPE(MPI_Status) and associated functions is -- at a minimum -- supposed to be in the mpi module. > > I'm a little curious as to why we conspicuously left it out of mpif.h. > > Bill: this is somewhat counter to the clarification you proposed. > > Are you ok with this? I think the text in the standard supports what Rolf proposes. > > > >> On Jul 10, 2020, at 12:11 PM, Jeff Squyres (jsquyres) via mpiwg-fortran <mpiwg-fortran(a)lists.mpi-forum.org <mailto:[email protected]>> wrote: >> >> On Jul 10, 2020, at 11:18 AM, Bill Long <longb(a)cray.com <mailto:[email protected]>> wrote: >>> >>> In the “change” section there is this txt: >>> >>> • Within the mpi_08 Fortran module, the status was defined as TYPE(MPI_Status). Additionally, within both the mpi and the mpi_f08 modules, the constants MPI_STATUS_SIZE, MPI_SOURCE, MPI_TAG, MPI_ERROR, and TYPE(MPI_Status) are defined. New conversion routines were added: MPI_STATUS_F2F08, MPI_STATUS_F082F, MPI_Status_c2f08, and MPI_Status_f082c, In mpi.h, the new type MPI_F08_status, and the external variables MPI_F08_STATUS_IGNORE and MPI_F08_STATUSES_IGNORE were added. >> >> Good point. >> >> Just to be clear, you're referring to the changelog section in MPI-3.1, specifically bullet 30 on p802. >> >> That being said: >> >> - the changelog is non-binding ...but it does indicate our intent from that time >> - the changelog text states that the mpi module has TYPE(MPI_Status) -- but it does not say it was added to mpif.h >> >>> 1) Why would the F08 status be defined different from the C definition? (If that were the case, conversions between f08 and C would be irrelevant). >> >> I remember that there was a lot of discussion about this at the time, which is what resulted in Figure 17.1. >> >> I know there were discussions about making the F08 and C statuses the same, but for some reason we chose not to mandate it. Perhaps we wanted to allow implementations to do whatever they wanted...? (e.g., allow Status_c2f08 be a no-op if the implementation wanted to, but not mandate it) >> >>> 2) \begin{unpopular} Why are the legacy mpi module and mpif.h still included in the spec? These are embarrassingly obsolete. If this was fixed, none of the above mentioned conversion routines would be needed. \end(unpopular} >> >> >> I would love it if we could ditch -- at a minimum -- mpif.h. >> >> However, there's oodles of legacy code out there that uses it. That's why even deprecating it gets shouted down at Forum meetings. >> >> -- >> Jeff Squyres >> jsquyres(a)cisco.com <mailto:[email protected]> >> >> _______________________________________________ >> mpiwg-fortran mailing list >> mpiwg-fortran(a)lists.mpi-forum.org >> https://lists.mpi-forum.org/mailman/listinfo/mpiwg-fortran > > > -- > Jeff Squyres > jsquyres(a)cisco.com <mailto:[email protected]> > _______________________________________________ > mpiwg-fortran mailing list > mpiwg-fortran(a)lists.mpi-forum.org > https://lists.mpi-forum.org/mailman/listinfo/mpiwg-fortran

1 0

Re: [MPIWG Fortran] Question about MPI_Status_f2f08() and _f082f()
by Jeff Squyres (jsquyres) 15 Jul '20

15 Jul '20

Just to bring this thread back on-track about TYPE(MPI_Status)... Off list, I asked Rolf R. about this issue -- he cited the same things I did, plus a few more: On your question on TYPE(MPI_Status), TYPE(MPI_Comm), ...: MPI-3.1 - page 607 lines 18-24 require these types and the overloaded operators .EQ./.NE. for mpi_f08 module - page 609 lines 34-36 require these types and the overloaded operators .EQ./.NE. for mpi module - there is no such text on page 611-612 on mpif.h And page 802 lines 9-15 also Show that it was never intented to add These types anf routines to old mpif.h. It would be helpful, to add at least on page 657 line 11 C, some in both C and Fortran (only in the mpi_f08 and mpi modules). Can you fix this with an one-vote-bug-fix-issue? Because it is not good if the information must be taken from the change-log. --- Summary: Does this mean that TYPE(MPI_Status) [ and the Fortran routines MPI_STATUS_F2F08 and _F082F ] needs to be defined in - mpif.h? NO - and the mpi module? YES Bug-fix needed in MPI-3.1 page 657 line 11: add "(only in the mpi_f08 and mpi modules)" at the end. I hope this helps. Best regards Rolf So I think there's at least a clarification here: the TYPE(MPI_Status) and associated functions is -- at a minimum -- supposed to be in the mpi module. I'm a little curious as to why we conspicuously left it out of mpif.h. Bill: this is somewhat counter to the clarification you proposed. Are you ok with this? I think the text in the standard supports what Rolf proposes. On Jul 10, 2020, at 12:11 PM, Jeff Squyres (jsquyres) via mpiwg-fortran <mpiwg-fortran(a)lists.mpi-forum.org<mailto:[email protected]>> wrote: On Jul 10, 2020, at 11:18 AM, Bill Long <longb(a)cray.com<mailto:[email protected]>> wrote: In the “change” section there is this txt: • Within the mpi_08 Fortran module, the status was defined as TYPE(MPI_Status). Additionally, within both the mpi and the mpi_f08 modules, the constants MPI_STATUS_SIZE, MPI_SOURCE, MPI_TAG, MPI_ERROR, and TYPE(MPI_Status) are defined. New conversion routines were added: MPI_STATUS_F2F08, MPI_STATUS_F082F, MPI_Status_c2f08, and MPI_Status_f082c, In mpi.h, the new type MPI_F08_status, and the external variables MPI_F08_STATUS_IGNORE and MPI_F08_STATUSES_IGNORE were added. Good point. Just to be clear, you're referring to the changelog section in MPI-3.1, specifically bullet 30 on p802. That being said: - the changelog is non-binding ...but it does indicate our intent from that time - the changelog text states that the mpi module has TYPE(MPI_Status) -- but it does not say it was added to mpif.h 1) Why would the F08 status be defined different from the C definition? (If that were the case, conversions between f08 and C would be irrelevant). I remember that there was a lot of discussion about this at the time, which is what resulted in Figure 17.1. I know there were discussions about making the F08 and C statuses the same, but for some reason we chose not to mandate it. Perhaps we wanted to allow implementations to do whatever they wanted...? (e.g., allow Status_c2f08 be a no-op if the implementation wanted to, but not mandate it) 2) \begin{unpopular} Why are the legacy mpi module and mpif.h still included in the spec? These are embarrassingly obsolete. If this was fixed, none of the above mentioned conversion routines would be needed. \end(unpopular} I would love it if we could ditch -- at a minimum -- mpif.h. However, there's oodles of legacy code out there that uses it. That's why even deprecating it gets shouted down at Forum meetings. -- Jeff Squyres jsquyres(a)cisco.com<mailto:[email protected]> _______________________________________________ mpiwg-fortran mailing list mpiwg-fortran(a)lists.mpi-forum.org https://lists.mpi-forum.org/mailman/listinfo/mpiwg-fortran -- Jeff Squyres jsquyres(a)cisco.com<mailto:[email protected]>

1 0

Re: [MPIWG Fortran] Question about MPI_Status_f2f08() and _f082f()
by Gilles Gouaillardet 13 Jul '20

13 Jul '20

Doing so would break all the MPI applications what use the 'implicit' directive implicit none include 'mpif.h' is fine, but implicit none use mpi is not, it should be use mpi implicit none On Tue, Jul 14, 2020 at 1:45 AM Jeff Hammond via mpiwg-fortran <mpiwg-fortran(a)lists.mpi-forum.org> wrote: > > When did Fortran start requiring function declarations? One of the issues we might see in legacy Fortran code is use of MPI functions without the header at all. > > Would it be legal to specify the contents of mpif.h be nothing more than "use mpi"? > > Jeff > > > On Mon, Jul 13, 2020 at 8:53 AM Bill Long <longb(a)cray.com> wrote: >> >> Well, as I pointed out, vendors are not going to actually delete old stuff. But that should not be a reason to avoid cleaning up the spec. >> >> There are ways to “fix” the mpif.h problem: >> >> 1) Assuming that the legacy MPI module provides the same capabilities as mpif.h, you could jus supply an empty mpif.h file (avoids having to delete the include line from the source), and then have a compiler option to “auto-use” a module and have the compilation include that option, specifying the MPI module. (Advantage of not changing the actual source file.) >> >> 2) Have a program that checks for an include of mpif.h and deleting that line and including a USE statement instead that specifies the MPI module. (Disadvantage that the mpif.h file might be included indirectly, by being part of another file that is included or part of a user module.) >> >> 3) Same as 2 only check for actual calls to MPI routines or uses of MPI data types in the source file. I there are none, do not add the USE MPI statement. Compile with the option that forces IMPLICIT NONE to be sure there were no missed references to MPI stuff. >> >> 4) Much more difficult - have a whole program analyzer that would remove all of the MPI (and SHMEM) and replace it with the native stuff built into Fortran. Probably the right path for new code, and bets done by a person for an old code. >> >> The Cray/HPE Fortran compiler already provides the tools to implement option 1. >> >> Cheers, >> Bill >> >> >> > On Jul 10, 2020, at 9:52 PM, Jeff Hammond <jeff.science(a)gmail.com> wrote: >> > >> > To help aid in the depreciation of mpif.h, perhaps Cray/HPE can contribute a source-to-source translation tool that can be recommended to the MPI user community. >> > >> > Jeff >> > >> > On Fri, Jul 10, 2020 at 9:54 AM Bill Long via mpiwg-fortran <mpiwg-fortran(a)lists.mpi-forum.org> wrote: >> > >> > >> > > On Jul 10, 2020, at 11:11 AM, Jeff Squyres (jsquyres) <jsquyres(a)cisco.com> wrote: >> > > >> > >> >> > >> 2) \begin{unpopular} Why are the legacy mpi module and mpif.h still included in the spec? These are embarrassingly obsolete. If this was fixed, none of the above mentioned conversion routines would be needed. \end(unpopular} >> > > >> > > >> > > I would love it if we could ditch -- at a minimum -- mpif.h. >> > > >> > > However, there's oodles of legacy code out there that uses it. That's why even deprecating it gets shouted down at Forum meetings. >> > >> > We have the same discussions in the Fortran committee about deleting things, or at least making them “obsolescent”, which seems functionally similar to deprecated in MPI. In the end, we specify certain archaic and now-redundant features as deleted with full knowledge that implementors will not remove the capabilities from compilers because of old codes. However, the Intent is to guide programmers writing new code in the direction of using the new (hopefully better) alternatives. And, relevant to the present discussion, deleting certain old stuff makes the standard document cleaner. >> > >> > Fundamentally, MPI is a C library. Other languages, like C++ and Python, don’t require separate bindings specified in the MPI spec because they can call C library routines. The whole mpi_f08 module exercise was to demonstrate that Fortran can now do the same. The whole idea of specifying a separate Fortran MPI interface needs to be rethought. Perhaps something like >> > >> > “This version of MPI specifies the contents of a vendor-required Fortran module for MPI that allows access to the underlying C specification. The separate language binding for Fortran has been deleted, but is still available for historical reference in the MPI-3 and earlier specifications.” >> > >> > That should allow for a significant simplification of the MPI specification document. >> > >> > >> > Cheers, >> > Bill >> > >> > >> > Bill Long longb(a)cray.com >> > Principal Engineer, Fortran Technical Support & voice: 651-605-9024 >> > Bioinformatics Software Development fax: 651-605-9143 >> > Cray, a Hewlett Packard Enterprise company/ 2131 Lindau Lane/ Suite 1000/ Bloomington, MN 55425 >> > >> > >> > >> > _______________________________________________ >> > mpiwg-fortran mailing list >> > mpiwg-fortran(a)lists.mpi-forum.org >> > https://lists.mpi-forum.org/mailman/listinfo/mpiwg-fortran >> > -- >> > Jeff Hammond >> > jeff.science(a)gmail.com >> > http://jeffhammond.github.io/ >> >> Bill Long longb(a)cray.com >> Principal Engineer, Fortran Technical Support & voice: 651-605-9024 >> Bioinformatics Software Development fax: 651-605-9143 >> Cray, a Hewlett Packard Enterprise company/ 2131 Lindau Lane/ Suite 1000/ Bloomington, MN 55425 >> >> >> > > > -- > Jeff Hammond > jeff.science(a)gmail.com > http://jeffhammond.github.io/ > _______________________________________________ > mpiwg-fortran mailing list > mpiwg-fortran(a)lists.mpi-forum.org > https://lists.mpi-forum.org/mailman/listinfo/mpiwg-fortran

1 0

Re: [MPIWG Fortran] Question about MPI_Status_f2f08() and _f082f()
by Bill Long 13 Jul '20

13 Jul '20

> On Jul 13, 2020, at 11:37 AM, Jeff Hammond <jeff.science(a)gmail.com> wrote: > > When did Fortran start requiring function declarations? Visible interfaces are only required if the arguments or the function has specific characteristics. (PURE function, optional arguments, …). F66 functions do not have any of these characteristics, so do not require interfaces. (Though providing them anyway is a good programming practice.) > One of the issues we might see in legacy Fortran code is use of MPI functions without the header at all. > > Would it be legal to specify the contents of mpif.h be nothing more than "use mpi”? No. unfortunately that will not work, because the include line can be in a part of the program were a USE statement is not legal. Cheers, Bill > > Jeff > > > On Mon, Jul 13, 2020 at 8:53 AM Bill Long <longb(a)cray.com> wrote: > Well, as I pointed out, vendors are not going to actually delete old stuff. But that should not be a reason to avoid cleaning up the spec. > > There are ways to “fix” the mpif.h problem: > > 1) Assuming that the legacy MPI module provides the same capabilities as mpif.h, you could jus supply an empty mpif.h file (avoids having to delete the include line from the source), and then have a compiler option to “auto-use” a module and have the compilation include that option, specifying the MPI module. (Advantage of not changing the actual source file.) > > 2) Have a program that checks for an include of mpif.h and deleting that line and including a USE statement instead that specifies the MPI module. (Disadvantage that the mpif.h file might be included indirectly, by being part of another file that is included or part of a user module.) > > 3) Same as 2 only check for actual calls to MPI routines or uses of MPI data types in the source file. I there are none, do not add the USE MPI statement. Compile with the option that forces IMPLICIT NONE to be sure there were no missed references to MPI stuff. > > 4) Much more difficult - have a whole program analyzer that would remove all of the MPI (and SHMEM) and replace it with the native stuff built into Fortran. Probably the right path for new code, and bets done by a person for an old code. > > The Cray/HPE Fortran compiler already provides the tools to implement option 1. > > Cheers, > Bill > > > > On Jul 10, 2020, at 9:52 PM, Jeff Hammond <jeff.science(a)gmail.com> wrote: > > > > To help aid in the depreciation of mpif.h, perhaps Cray/HPE can contribute a source-to-source translation tool that can be recommended to the MPI user community. > > > > Jeff > > > > On Fri, Jul 10, 2020 at 9:54 AM Bill Long via mpiwg-fortran <mpiwg-fortran(a)lists.mpi-forum.org> wrote: > > > > > > > On Jul 10, 2020, at 11:11 AM, Jeff Squyres (jsquyres) <jsquyres(a)cisco.com> wrote: > > > > > >> > > >> 2) \begin{unpopular} Why are the legacy mpi module and mpif.h still included in the spec? These are embarrassingly obsolete. If this was fixed, none of the above mentioned conversion routines would be needed. \end(unpopular} > > > > > > > > > I would love it if we could ditch -- at a minimum -- mpif.h. > > > > > > However, there's oodles of legacy code out there that uses it. That's why even deprecating it gets shouted down at Forum meetings. > > > > We have the same discussions in the Fortran committee about deleting things, or at least making them “obsolescent”, which seems functionally similar to deprecated in MPI. In the end, we specify certain archaic and now-redundant features as deleted with full knowledge that implementors will not remove the capabilities from compilers because of old codes. However, the Intent is to guide programmers writing new code in the direction of using the new (hopefully better) alternatives. And, relevant to the present discussion, deleting certain old stuff makes the standard document cleaner. > > > > Fundamentally, MPI is a C library. Other languages, like C++ and Python, don’t require separate bindings specified in the MPI spec because they can call C library routines. The whole mpi_f08 module exercise was to demonstrate that Fortran can now do the same. The whole idea of specifying a separate Fortran MPI interface needs to be rethought. Perhaps something like > > > > “This version of MPI specifies the contents of a vendor-required Fortran module for MPI that allows access to the underlying C specification. The separate language binding for Fortran has been deleted, but is still available for historical reference in the MPI-3 and earlier specifications.” > > > > That should allow for a significant simplification of the MPI specification document. > > > > > > Cheers, > > Bill > > > > > > Bill Long longb(a)cray.com > > Principal Engineer, Fortran Technical Support & voice: 651-605-9024 > > Bioinformatics Software Development fax: 651-605-9143 > > Cray, a Hewlett Packard Enterprise company/ 2131 Lindau Lane/ Suite 1000/ Bloomington, MN 55425 > > > > > > > > _______________________________________________ > > mpiwg-fortran mailing list > > mpiwg-fortran(a)lists.mpi-forum.org > > https://lists.mpi-forum.org/mailman/listinfo/mpiwg-fortran > > -- > > Jeff Hammond > > jeff.science(a)gmail.com > > http://jeffhammond.github.io/ > > Bill Long longb(a)cray.com > Principal Engineer, Fortran Technical Support & voice: 651-605-9024 > Bioinformatics Software Development fax: 651-605-9143 > Cray, a Hewlett Packard Enterprise company/ 2131 Lindau Lane/ Suite 1000/ Bloomington, MN 55425 > > > > > > -- > Jeff Hammond > jeff.science(a)gmail.com > http://jeffhammond.github.io/ Bill Long longb(a)cray.com Principal Engineer, Fortran Technical Support & voice: 651-605-9024 Bioinformatics Software Development fax: 651-605-9143 Cray, a Hewlett Packard Enterprise company/ 2131 Lindau Lane/ Suite 1000/ Bloomington, MN 55425

1 0

Re: [MPIWG Fortran] Question about MPI_Status_f2f08() and _f082f()
by Jeff Hammond 13 Jul '20

13 Jul '20

When did Fortran start requiring function declarations? One of the issues we might see in legacy Fortran code is use of MPI functions without the header at all. Would it be legal to specify the contents of mpif.h be nothing more than "use mpi"? Jeff On Mon, Jul 13, 2020 at 8:53 AM Bill Long <longb(a)cray.com> wrote: > Well, as I pointed out, vendors are not going to actually delete old > stuff. But that should not be a reason to avoid cleaning up the spec. > > There are ways to “fix” the mpif.h problem: > > 1) Assuming that the legacy MPI module provides the same capabilities as > mpif.h, you could jus supply an empty mpif.h file (avoids having to delete > the include line from the source), and then have a compiler option to > “auto-use” a module and have the compilation include that option, > specifying the MPI module. (Advantage of not changing the actual source > file.) > > 2) Have a program that checks for an include of mpif.h and deleting that > line and including a USE statement instead that specifies the MPI module. > (Disadvantage that the mpif.h file might be included indirectly, by being > part of another file that is included or part of a user module.) > > 3) Same as 2 only check for actual calls to MPI routines or uses of MPI > data types in the source file. I there are none, do not add the USE MPI > statement. Compile with the option that forces IMPLICIT NONE to be sure > there were no missed references to MPI stuff. > > 4) Much more difficult - have a whole program analyzer that would remove > all of the MPI (and SHMEM) and replace it with the native stuff built into > Fortran. Probably the right path for new code, and bets done by a person > for an old code. > > The Cray/HPE Fortran compiler already provides the tools to implement > option 1. > > Cheers, > Bill > > > > On Jul 10, 2020, at 9:52 PM, Jeff Hammond <jeff.science(a)gmail.com> > wrote: > > > > To help aid in the depreciation of mpif.h, perhaps Cray/HPE can > contribute a source-to-source translation tool that can be recommended to > the MPI user community. > > > > Jeff > > > > On Fri, Jul 10, 2020 at 9:54 AM Bill Long via mpiwg-fortran < > mpiwg-fortran(a)lists.mpi-forum.org> wrote: > > > > > > > On Jul 10, 2020, at 11:11 AM, Jeff Squyres (jsquyres) < > jsquyres(a)cisco.com> wrote: > > > > > >> > > >> 2) \begin{unpopular} Why are the legacy mpi module and mpif.h still > included in the spec? These are embarrassingly obsolete. If this was > fixed, none of the above mentioned conversion routines would be needed. > \end(unpopular} > > > > > > > > > I would love it if we could ditch -- at a minimum -- mpif.h. > > > > > > However, there's oodles of legacy code out there that uses it. That's > why even deprecating it gets shouted down at Forum meetings. > > > > We have the same discussions in the Fortran committee about deleting > things, or at least making them “obsolescent”, which seems functionally > similar to deprecated in MPI. In the end, we specify certain archaic and > now-redundant features as deleted with full knowledge that implementors > will not remove the capabilities from compilers because of old codes. > However, the Intent is to guide programmers writing new code in the > direction of using the new (hopefully better) alternatives. And, relevant > to the present discussion, deleting certain old stuff makes the standard > document cleaner. > > > > Fundamentally, MPI is a C library. Other languages, like C++ and > Python, don’t require separate bindings specified in the MPI spec because > they can call C library routines. The whole mpi_f08 module exercise was to > demonstrate that Fortran can now do the same. The whole idea of specifying > a separate Fortran MPI interface needs to be rethought. Perhaps something > like > > > > “This version of MPI specifies the contents of a vendor-required Fortran > module for MPI that allows access to the underlying C specification. The > separate language binding for Fortran has been deleted, but is still > available for historical reference in the MPI-3 and earlier specifications.” > > > > That should allow for a significant simplification of the MPI > specification document. > > > > > > Cheers, > > Bill > > > > > > Bill Long > longb(a)cray.com > > Principal Engineer, Fortran Technical Support & voice: 651-605-9024 > > Bioinformatics Software Development fax: > 651-605-9143 > > Cray, a Hewlett Packard Enterprise company/ 2131 Lindau Lane/ Suite > 1000/ Bloomington, MN 55425 > > > > > > > > _______________________________________________ > > mpiwg-fortran mailing list > > mpiwg-fortran(a)lists.mpi-forum.org > > https://lists.mpi-forum.org/mailman/listinfo/mpiwg-fortran > > -- > > Jeff Hammond > > jeff.science(a)gmail.com > > http://jeffhammond.github.io/ > > Bill Long > longb(a)cray.com > Principal Engineer, Fortran Technical Support & voice: 651-605-9024 > Bioinformatics Software Development fax: 651-605-9143 > Cray, a Hewlett Packard Enterprise company/ 2131 Lindau Lane/ Suite > 1000/ Bloomington, MN 55425 > > > > -- Jeff Hammond jeff.science(a)gmail.com http://jeffhammond.github.io/

1 0

Re: [MPIWG Fortran] Question about MPI_Status_f2f08() and _f082f()
by Bill Long 13 Jul '20

13 Jul '20

Well, as I pointed out, vendors are not going to actually delete old stuff. But that should not be a reason to avoid cleaning up the spec. There are ways to “fix” the mpif.h problem: 1) Assuming that the legacy MPI module provides the same capabilities as mpif.h, you could jus supply an empty mpif.h file (avoids having to delete the include line from the source), and then have a compiler option to “auto-use” a module and have the compilation include that option, specifying the MPI module. (Advantage of not changing the actual source file.) 2) Have a program that checks for an include of mpif.h and deleting that line and including a USE statement instead that specifies the MPI module. (Disadvantage that the mpif.h file might be included indirectly, by being part of another file that is included or part of a user module.) 3) Same as 2 only check for actual calls to MPI routines or uses of MPI data types in the source file. I there are none, do not add the USE MPI statement. Compile with the option that forces IMPLICIT NONE to be sure there were no missed references to MPI stuff. 4) Much more difficult - have a whole program analyzer that would remove all of the MPI (and SHMEM) and replace it with the native stuff built into Fortran. Probably the right path for new code, and bets done by a person for an old code. The Cray/HPE Fortran compiler already provides the tools to implement option 1. Cheers, Bill > On Jul 10, 2020, at 9:52 PM, Jeff Hammond <jeff.science(a)gmail.com> wrote: > > To help aid in the depreciation of mpif.h, perhaps Cray/HPE can contribute a source-to-source translation tool that can be recommended to the MPI user community. > > Jeff > > On Fri, Jul 10, 2020 at 9:54 AM Bill Long via mpiwg-fortran <mpiwg-fortran(a)lists.mpi-forum.org> wrote: > > > > On Jul 10, 2020, at 11:11 AM, Jeff Squyres (jsquyres) <jsquyres(a)cisco.com> wrote: > > > >> > >> 2) \begin{unpopular} Why are the legacy mpi module and mpif.h still included in the spec? These are embarrassingly obsolete. If this was fixed, none of the above mentioned conversion routines would be needed. \end(unpopular} > > > > > > I would love it if we could ditch -- at a minimum -- mpif.h. > > > > However, there's oodles of legacy code out there that uses it. That's why even deprecating it gets shouted down at Forum meetings. > > We have the same discussions in the Fortran committee about deleting things, or at least making them “obsolescent”, which seems functionally similar to deprecated in MPI. In the end, we specify certain archaic and now-redundant features as deleted with full knowledge that implementors will not remove the capabilities from compilers because of old codes. However, the Intent is to guide programmers writing new code in the direction of using the new (hopefully better) alternatives. And, relevant to the present discussion, deleting certain old stuff makes the standard document cleaner. > > Fundamentally, MPI is a C library. Other languages, like C++ and Python, don’t require separate bindings specified in the MPI spec because they can call C library routines. The whole mpi_f08 module exercise was to demonstrate that Fortran can now do the same. The whole idea of specifying a separate Fortran MPI interface needs to be rethought. Perhaps something like > > “This version of MPI specifies the contents of a vendor-required Fortran module for MPI that allows access to the underlying C specification. The separate language binding for Fortran has been deleted, but is still available for historical reference in the MPI-3 and earlier specifications.” > > That should allow for a significant simplification of the MPI specification document. > > > Cheers, > Bill > > > Bill Long longb(a)cray.com > Principal Engineer, Fortran Technical Support & voice: 651-605-9024 > Bioinformatics Software Development fax: 651-605-9143 > Cray, a Hewlett Packard Enterprise company/ 2131 Lindau Lane/ Suite 1000/ Bloomington, MN 55425 > > > > _______________________________________________ > mpiwg-fortran mailing list > mpiwg-fortran(a)lists.mpi-forum.org > https://lists.mpi-forum.org/mailman/listinfo/mpiwg-fortran > -- > Jeff Hammond > jeff.science(a)gmail.com > http://jeffhammond.github.io/ Bill Long longb(a)cray.com Principal Engineer, Fortran Technical Support & voice: 651-605-9024 Bioinformatics Software Development fax: 651-605-9143 Cray, a Hewlett Packard Enterprise company/ 2131 Lindau Lane/ Suite 1000/ Bloomington, MN 55425

1 0

Re: [MPIWG Fortran] Question about MPI_Status_f2f08() and _f082f()
by Jeff Hammond 10 Jul '20

10 Jul '20

To help aid in the depreciation of mpif.h, perhaps Cray/HPE can contribute a source-to-source translation tool that can be recommended to the MPI user community. Jeff On Fri, Jul 10, 2020 at 9:54 AM Bill Long via mpiwg-fortran < mpiwg-fortran(a)lists.mpi-forum.org> wrote: > > > > On Jul 10, 2020, at 11:11 AM, Jeff Squyres (jsquyres) < > jsquyres(a)cisco.com> wrote: > > > >> > >> 2) \begin{unpopular} Why are the legacy mpi module and mpif.h still > included in the spec? These are embarrassingly obsolete. If this was > fixed, none of the above mentioned conversion routines would be needed. > \end(unpopular} > > > > > > I would love it if we could ditch -- at a minimum -- mpif.h. > > > > However, there's oodles of legacy code out there that uses it. That's > why even deprecating it gets shouted down at Forum meetings. > > We have the same discussions in the Fortran committee about deleting > things, or at least making them “obsolescent”, which seems functionally > similar to deprecated in MPI. In the end, we specify certain archaic and > now-redundant features as deleted with full knowledge that implementors > will not remove the capabilities from compilers because of old codes. > However, the Intent is to guide programmers writing new code in the > direction of using the new (hopefully better) alternatives. And, relevant > to the present discussion, deleting certain old stuff makes the standard > document cleaner. > > Fundamentally, MPI is a C library. Other languages, like C++ and Python, > don’t require separate bindings specified in the MPI spec because they can > call C library routines. The whole mpi_f08 module exercise was to > demonstrate that Fortran can now do the same. The whole idea of specifying > a separate Fortran MPI interface needs to be rethought. Perhaps something > like > > “This version of MPI specifies the contents of a vendor-required Fortran > module for MPI that allows access to the underlying C specification. The > separate language binding for Fortran has been deleted, but is still > available for historical reference in the MPI-3 and earlier specifications.” > > That should allow for a significant simplification of the MPI > specification document. > > > Cheers, > Bill > > > Bill Long > longb(a)cray.com > Principal Engineer, Fortran Technical Support & voice: 651-605-9024 > Bioinformatics Software Development fax: 651-605-9143 > Cray, a Hewlett Packard Enterprise company/ 2131 Lindau Lane/ Suite > 1000/ Bloomington, MN 55425 > > > > _______________________________________________ > mpiwg-fortran mailing list > mpiwg-fortran(a)lists.mpi-forum.org > https://lists.mpi-forum.org/mailman/listinfo/mpiwg-fortran > -- Jeff Hammond jeff.science(a)gmail.com http://jeffhammond.github.io/

1 0

Re: [MPIWG Fortran] Question about MPI_Status_f2f08() and _f082f()
by Bill Long 10 Jul '20

10 Jul '20

> On Jul 10, 2020, at 11:11 AM, Jeff Squyres (jsquyres) <jsquyres(a)cisco.com> wrote: > >> >> 2) \begin{unpopular} Why are the legacy mpi module and mpif.h still included in the spec? These are embarrassingly obsolete. If this was fixed, none of the above mentioned conversion routines would be needed. \end(unpopular} > > > I would love it if we could ditch -- at a minimum -- mpif.h. > > However, there's oodles of legacy code out there that uses it. That's why even deprecating it gets shouted down at Forum meetings. We have the same discussions in the Fortran committee about deleting things, or at least making them “obsolescent”, which seems functionally similar to deprecated in MPI. In the end, we specify certain archaic and now-redundant features as deleted with full knowledge that implementors will not remove the capabilities from compilers because of old codes. However, the Intent is to guide programmers writing new code in the direction of using the new (hopefully better) alternatives. And, relevant to the present discussion, deleting certain old stuff makes the standard document cleaner. Fundamentally, MPI is a C library. Other languages, like C++ and Python, don’t require separate bindings specified in the MPI spec because they can call C library routines. The whole mpi_f08 module exercise was to demonstrate that Fortran can now do the same. The whole idea of specifying a separate Fortran MPI interface needs to be rethought. Perhaps something like “This version of MPI specifies the contents of a vendor-required Fortran module for MPI that allows access to the underlying C specification. The separate language binding for Fortran has been deleted, but is still available for historical reference in the MPI-3 and earlier specifications.” That should allow for a significant simplification of the MPI specification document. Cheers, Bill Bill Long longb(a)cray.com Principal Engineer, Fortran Technical Support & voice: 651-605-9024 Bioinformatics Software Development fax: 651-605-9143 Cray, a Hewlett Packard Enterprise company/ 2131 Lindau Lane/ Suite 1000/ Bloomington, MN 55425

1 0