From owner-sc22wg5+sc22wg5-dom9=www.open-std.org@open-std.org Mon Mar 18 08:04:08 2024 Return-Path: X-Original-To: sc22wg5-dom9 Delivered-To: sc22wg5-dom9@www.open-std.org Received: by www.open-std.org (Postfix, from userid 521) id 1AE59356D8A; Mon, 18 Mar 2024 08:04:08 +0100 (CET) Delivered-To: sc22wg5@open-std.org Received: from nag-j.co.jp (bvdeuz19.secure.ne.jp [180.222.80.19]) by www.open-std.org (Postfix) with SMTP id 1D5373568AA for ; Mon, 18 Mar 2024 08:04:04 +0100 (CET) Received: (qmail 24400 invoked from network); 18 Mar 2024 16:04:02 +0900 Received: from unknown (HELO Maru10) (218.42.159.105) by 0 with SMTP; 18 Mar 2024 16:04:02 +0900 From: "Malcolm Cohen" To: "'WG5'" References: In-Reply-To: Subject: RE: [ukfortran] [SC22WG5.6552] bind(C) external linkage issues on MacOS Date: Mon, 18 Mar 2024 16:04:03 +0900 Message-ID: <01e301da7902$75ab3160$61019420$@nag-j.co.jp> MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_NextPart_000_01E4_01DA794D.E5960DB0" X-Mailer: Microsoft Outlook 16.0 Thread-Index: AQKJXIUf4GyabNvhyABgzEFOckKtkq/fSbkg Content-Language: ja Sender: owner-sc22wg5@open-std.org Precedence: bulk This is a multipart message in MIME format. ------=_NextPart_000_01E4_01DA794D.E5960DB0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Perhaps using something equivalent to the gcc "__attribute__((__common__))" on the symbol might help? In C one is stuck with the gcc extension, but in principle the Fortran compiler should be able to do this automatically. Unless this is a slightly different issue, for which the above doesn't work. Cheers, -- ..............Malcolm Cohen, NAG Oxford/Tokyo. From: Jeff Hammond Sent: Saturday, March 16, 2024 2:29 AM To: WG5 Subject: [ukfortran] [SC22WG5.6552] bind(C) external linkage issues on MacOS This is both an FYI for those of you who use MacOS but also a possible identification of a problem in how we specify 19.5.1.5 Linkage association. Apple has removed "-commons use_dylibs" in recent versions of MacOS. This flag sets a non-default behavior of COMMON symbols (which include linkage associated bind(C) symbols that are not using obsolescent features) to be "U" (undefined) instead of "S" (uninitialized, small) https://linux.die.net/man/1/nm. This flag is required to get a bind(C) module variable that points to a C global variable to work properly, as is potentially required by MPI. There are many reports of failures online with both GCC and Intel compilers (I can only verify the former). It has been noted that our definition of linkage association may not be adequate to interact with C shared libraries. I don't understand this topic well enough to have an opinion either way. Has anybody else run into this or thought about it enough to know if there is a way to enhance our definition of linkage association so as to not require Fortran-specific linker options and behave more like C extern in shared libraries? Thanks, Jeff jhammond@Jeff-M2 fortran % gcc -fPIC -shared extern2.c -o libxxx.so && gfortran -c extern.F90 && ld extern.o libxxx.so -L/opt/homebrew/Cellar/gcc/13.2.0/lib/gcc/current/ -lgfortran -commons use_dylibs -o extern && ./extern ; nm extern | grep MPI MPIR_F08_MPI_IN_PLACE=0 &MPIR_F08_MPI_IN_PLACE=0x1003a4000 &MPIR_F08_MPI_IN_PLACE=4298784768 LOC(MPI_IN_PLACE)= 4298784768 LOC(buf)= 6168523360 sendbuf=0x1003a4000, sendbuf=4298784768 sendbuf is MPI_IN_PLACE? yes recvbuf=0x16fac3260, recvbuf=6168523360 *count=1, *datatype=2, *op=3, *comm=4 911 U _MPIR_F08_MPI_IN_PLACE U _MPI_Allreduce jhammond@Jeff-M2 fortran % gcc -fPIC -shared extern2.c -o libxxx.so && gfortran -c extern.F90 && ld extern.o libxxx.so -L/opt/homebrew/Cellar/gcc/13.2.0/lib/gcc/current/ -lgfortran -o extern && ./extern ; nm extern | grep MPI MPIR_F08_MPI_IN_PLACE=0 &MPIR_F08_MPI_IN_PLACE=0x102b1c000 &MPIR_F08_MPI_IN_PLACE=4340170752 LOC(MPI_IN_PLACE)= 4339777664 LOC(buf)= 6127137376 sendbuf=0x102abc080, sendbuf=4339777664 sendbuf is MPI_IN_PLACE? no recvbuf=0x16d34b260, recvbuf=6127137376 *count=1, *datatype=2, *op=3, *comm=4 911 0000000100008080 S _MPIR_F08_MPI_IN_PLACE U _MPI_Allreduce ==> extern2.c <== #include #include int MPIR_F08_MPI_IN_PLACE; void p(void) { printf("MPIR_F08_MPI_IN_PLACE=%d &MPIR_F08_MPI_IN_PLACE=%p &MPIR_F08_MPI_IN_PLACE=%zu\n", MPIR_F08_MPI_IN_PLACE, &MPIR_F08_MPI_IN_PLACE, (intptr_t)&MPIR_F08_MPI_IN_PLACE); } void MPI_Allreduce(void ** sendbuf, void ** recvbuf, int * count, int * datatype, int * op, int * comm, int * ierror) { printf("sendbuf=%p, sendbuf=%zu\n", sendbuf, (intptr_t)sendbuf); printf("sendbuf is MPI_IN_PLACE? %s\n", (intptr_t)sendbuf==(intptr_t)&MPIR_F08_MPI_IN_PLACE ? "yes" : "no"); printf("recvbuf=%p, recvbuf=%zu\n", recvbuf, (intptr_t)recvbuf); printf("*count=%d, *datatype=%d, *op=%d, *comm=%d\n", *count, *datatype, *op, *comm); *ierror = 911; } ==> extern.F90 <== module mpi use iso_c_binding !type(c_ptr), bind(C,name="MPI_F_IN_PLACE") :: MPI_IN_PLACE integer(c_int), bind(C, name="MPIR_F08_MPI_IN_PLACE"), target :: MPI_IN_PLACE interface subroutine p() bind(C,name="p") end subroutine end interface interface SUBROUTINE MPI_ALLREDUCE(SENDBUF, RECVBUF, COUNT, DATATYPE, OP, COMM, IERROR) & bind(C,name="MPI_Allreduce") use iso_c_binding import :: MPI_IN_PLACE !DEC$ ATTRIBUTES NO_ARG_CHECK :: sendbuf,recvbuf !GCC$ ATTRIBUTES NO_ARG_CHECK :: sendbuf,recvbuf !$PRAGMA IGNORE_TKR sendbuf,recvbuf !DIR$ IGNORE_TKR sendbuf,recvbuf !IBM* IGNORE_TKR sendbuf,recvbuf INTEGER(kind=c_int) :: SENDBUF(*), RECVBUF(*) INTEGER(kind=c_int) :: COUNT, DATATYPE, OP, COMM, IERROR END SUBROUTINE MPI_ALLREDUCE end interface end module mpi program main use mpi implicit none real :: buf(100) integer :: ierror call p buf = 17 print*,'LOC(MPI_IN_PLACE)=',LOC(MPI_IN_PLACE) print*,'LOC(buf)=',LOC(buf) call MPI_ALLREDUCE(MPI_IN_PLACE,buf,1,2,3,4,ierror) print*,ierror end program main ------=_NextPart_000_01E4_01DA794D.E5960DB0 Content-Type: text/html; charset="us-ascii" Content-Transfer-Encoding: quoted-printable

Perhaps using something = equivalent to the gcc “__attribute__((__common__))” on the = symbol might help? In C one is stuck with the gcc extension, but in = principle the Fortran compiler should be able to do this = automatically.

 

Unless this is a slightly = different issue, for which the above doesn’t = work…

 

Cheers,

<= div>

--

..............Malcolm Cohen, NAG Oxford/Tokyo.

 

From: Jeff = Hammond <jehammond@nvidia.com>
Sent: Saturday, March = 16, 2024 2:29 AM
To: WG5 = <sc22wg5@open-std.org>
Subject: [ukfortran] = [SC22WG5.6552] bind(C) external linkage issues on = MacOS

 

This is both an FYI for those of = you who use MacOS but also a possible identification of a problem in how = we specify 19.5.1.5 Linkage association.

 

Apple = has removed “-commons use_dylibs” in recent versions of = MacOS.  This flag sets a non-default behavior of COMMON symbols = (which include linkage associated bind(C) symbols that are not using = obsolescent features) to be “U” (undefined) instead of = “S” (uninitialized, small) https://linux.die.net/man/1/nm.

 

This = flag is required to get a bind(C) module variable that points to a C = global variable to work properly, as is potentially required by = MPI.  There are many reports of failures online with both GCC and = Intel compilers (I can only verify the former).

 

It has = been noted that our definition of linkage association may not be = adequate to interact with C shared libraries.  I don’t = understand this topic well enough to have an opinion either = way.

 

Has = anybody else run into this or thought about it enough to know if there = is a way to enhance our definition of linkage association so as to not = require Fortran-specific linker options and behave more like C extern in = shared libraries?

 

Thanks,

Jeff

 

jhammond@Jeff-M2 fortran % gcc -fPIC -shared = extern2.c -o libxxx.so && gfortran -c extern.F90 && ld = extern.o libxxx.so = -L/opt/homebrew/Cellar/gcc/13.2.0/lib/gcc/current/  -lgfortran = -commons use_dylibs -o extern = && ./extern ; nm extern | grep MPI

MPIR_F08_MPI_IN_PLACE=3D0 = &MPIR_F08_MPI_IN_PLACE=3D0x1003a4000 = &MPIR_F08_MPI_IN_PLACE=3D4298784768

LOC(MPI_IN_PLACE)=3D    &n= bsp;      4298784768

LOC(buf)=3D      = ;     6168523360

sendbuf=3D0x1003a4000, = sendbuf=3D4298784768

sendbuf is MPI_IN_PLACE? = yes

recvbuf=3D0x16fac3260, = recvbuf=3D6168523360

*count=3D1, *datatype=3D2, = *op=3D3, *comm=3D4

       &nbs= p; 911

      =            U = _MPIR_F08_MPI_IN_PLACE

       &nbs= p;         U = _MPI_Allreduce

 

 

jhammond@Jeff-M2 fortran % gcc -fPIC -shared = extern2.c -o libxxx.so && gfortran -c extern.F90 && ld = extern.o libxxx.so = -L/opt/homebrew/Cellar/gcc/13.2.0/lib/gcc/current/  = -lgfortran  -o extern && ./extern ; nm extern | grep = MPI

MPIR_F08_MPI_IN_PLACE=3D0 = &MPIR_F08_MPI_IN_PLACE=3D0x102b1c000 = &MPIR_F08_MPI_IN_PLACE=3D4340170752

LOC(MPI_IN_PLACE)=3D    &n= bsp;      4339777664

LOC(buf)=3D      = ;     6127137376

sendbuf=3D0x102abc080, = sendbuf=3D4339777664

sendbuf is MPI_IN_PLACE? = no

recvbuf=3D0x16d34b260, = recvbuf=3D6127137376

*count=3D1, *datatype=3D2, = *op=3D3, *comm=3D4

       &nbs= p; 911

0000000100008080 S = _MPIR_F08_MPI_IN_PLACE

       &nbs= p;         U = _MPI_Allreduce

 

 

=3D=3D> extern2.c = <=3D=3D

#include = <stdio.h>

#include = <stdint.h>

 

int = MPIR_F08_MPI_IN_PLACE;

 

void = p(void)

{

    = printf("MPIR_F08_MPI_IN_PLACE=3D%d &MPIR_F08_MPI_IN_PLACE=3D%p = &MPIR_F08_MPI_IN_PLACE=3D%zu\n",

       &nbs= p;    MPIR_F08_MPI_IN_PLACE,   = &MPIR_F08_MPI_IN_PLACE,   = (intptr_t)&MPIR_F08_MPI_IN_PLACE);

}

 

void = MPI_Allreduce(void ** sendbuf, void ** recvbuf,

       &nbs= p;           int * = count, int * datatype,

       &nbs= p;           int * op, = int * comm, int * ierror)

{

    printf("sendbuf=3D%p, = sendbuf=3D%zu\n", sendbuf, = (intptr_t)sendbuf);

    = printf("sendbuf is MPI_IN_PLACE? = %s\n",

       &nbs= p;   = (intptr_t)sendbuf=3D=3D(intptr_t)&MPIR_F08_MPI_IN_PLACE ? = "yes" : "no");

    printf("recvbuf=3D%p, = recvbuf=3D%zu\n", recvbuf, = (intptr_t)recvbuf);

    = printf("*count=3D%d, *datatype=3D%d, *op=3D%d, = *comm=3D%d\n",

       &nbs= p;    *count, *datatype, *op, = *comm);

    *ierror =3D = 911;

}

 

=3D=3D> extern.F90 = <=3D=3D

module mpi

    use = iso_c_binding

    !type(c_ptr), = bind(C,name=3D"MPI_F_IN_PLACE") :: = MPI_IN_PLACE

    = integer(c_int), bind(C, name=3D"MPIR_F08_MPI_IN_PLACE"), = target :: MPI_IN_PLACE

    = interface

        = subroutine p() bind(C,name=3D"p")

        = end subroutine

    end = interface

    = interface

        = SUBROUTINE MPI_ALLREDUCE(SENDBUF, RECVBUF, COUNT, DATATYPE, OP, COMM, = IERROR) &

       &nbs= p;           = bind(C,name=3D"MPI_Allreduce")

       &nbs= p;    use iso_c_binding

       &nbs= p;    import :: MPI_IN_PLACE

       &nbs= p;    !DEC$ ATTRIBUTES NO_ARG_CHECK :: = sendbuf,recvbuf

       &nbs= p;    !GCC$ ATTRIBUTES NO_ARG_CHECK :: = sendbuf,recvbuf

       &nbs= p;    !$PRAGMA IGNORE_TKR = sendbuf,recvbuf

       &nbs= p;    !DIR$ IGNORE_TKR = sendbuf,recvbuf

       &nbs= p;    !IBM* IGNORE_TKR = sendbuf,recvbuf

       &nbs= p;    INTEGER(kind=3Dc_int) :: SENDBUF(*), = RECVBUF(*)

       &nbs= p;    INTEGER(kind=3Dc_int) :: COUNT, DATATYPE, OP, COMM, = IERROR

        = END SUBROUTINE MPI_ALLREDUCE

    end = interface

end module mpi

 

program = main

    use = mpi

    implicit = none

    real :: = buf(100)

    integer :: = ierror

    call = p

    buf =3D = 17

    = print*,'LOC(MPI_IN_PLACE)=3D',LOC(MPI_IN_PLACE)

    = print*,'LOC(buf)=3D',LOC(buf)

    call = MPI_ALLREDUCE(MPI_IN_PLACE,buf,1,2,3,4,ierror)

    = print*,ierror

end program = main

------=_NextPart_000_01E4_01DA794D.E5960DB0--