May 2013 - discuss - mailman3.cels.anl.gov

Re: [mpich-discuss] [PATCH v2] test: add attrdeleteget, MPI_Attr_get called from delete_fn
by Pavan Balaji 18 May '13

18 May '13

Hi Jed, You should use MPI_COMM_CREATE_KEYVAL and MPI_COMM_GET_ATTR. KEYVAL_CREATE and ATTR_GET are deprecated. I'm not sure I understand the program. If the comm is invalidated before the delete_fn is called, this is erroneous. But if it is invalidated after the delete_fn is called, this is valid. So what exactly are you trying to do? -- Pavan On 05/17/2013 11:52 PM US Central Time, Jed Brown wrote: > With MS-MPI 64-bit from HPC Pack 2008 and 2012, MPI_Attr_get returns > error code 773 when called from delete_fn on a communicator obtained > from MPI_Comm_split. > > Signed-off-by: Jed Brown <jedbrown(a)mcs.anl.gov> > Signed-off-by: Satish Balay <balay(a)mcs.anl.gov> > --- > Updated to propagate the error code from MPI_Attr_get. > > test/mpi/.gitignore | 1 + > test/mpi/attr/Makefile.am | 1 + > test/mpi/attr/attrdeleteget.c | 36 ++++++++++++++++++++++++++++++++++++ > test/mpi/attr/testlist | 1 + > 4 files changed, 39 insertions(+) > create mode 100644 test/mpi/attr/attrdeleteget.c > > diff --git a/test/mpi/.gitignore b/test/mpi/.gitignore > index 3aadba2..0bdcd15 100644 > --- a/test/mpi/.gitignore > +++ b/test/mpi/.gitignore > @@ -51,6 +51,7 @@ > /attr/attrerr > /attr/attrerrcomm > /attr/attrerrtype > +/attr/attrdeleteget > /attr/attric > /attr/attrorder > /attr/attrordercomm > diff --git a/test/mpi/attr/Makefile.am b/test/mpi/attr/Makefile.am > index a672192..e1d788b 100644 > --- a/test/mpi/attr/Makefile.am > +++ b/test/mpi/attr/Makefile.am > @@ -18,6 +18,7 @@ noinst_PROGRAMS = \ > attrerr \ > attrerrcomm \ > attrerrtype \ > + attrdeleteget \ > attr2type \ > attrorder \ > attrordercomm \ > diff --git a/test/mpi/attr/attrdeleteget.c b/test/mpi/attr/attrdeleteget.c > new file mode 100644 > index 0000000..05afaa4 > --- /dev/null > +++ b/test/mpi/attr/attrdeleteget.c > @@ -0,0 +1,36 @@ > +/* -*- Mode: C; c-basic-offset:4 ; indent-tabs-mode:nil ; -*- */ > +/* > + * > + * (C) 2013 by Argonne National Laboratory. > + * See COPYRIGHT in top-level directory. > + */ > + > +#include <stdio.h> > +#include "mpi.h" > +#include "mpitest.h" > + > +int key = MPI_KEYVAL_INVALID; > +char a[100]; > + > +int delete_fn (MPI_Comm, int, void *, void *); > + > +int main(int argc,char **argv) > +{ > + MPI_Comm scomm; > + > + MTest_Init(&argc,&argv); > + MPI_Comm_split(MPI_COMM_WORLD,1,0,&scomm); > + MPI_Keyval_create(MPI_NULL_COPY_FN,delete_fn,&key,(void*)0); > + MPI_Attr_put(scomm,key,a); > + MPI_Comm_free(&scomm); > + MPI_Finalize(); > + return 0; > +} > + > +int delete_fn(MPI_Comm comm,int keyval,void *attr_val,void *extra_state) > +{ > + int flg; > + void *ptr; > + > + return MPI_Attr_get(comm,key,&ptr,&flg); > +} > diff --git a/test/mpi/attr/testlist b/test/mpi/attr/testlist > index 61bd613..4db94df 100644 > --- a/test/mpi/attr/testlist > +++ b/test/mpi/attr/testlist > @@ -7,6 +7,7 @@ attrend2 1 > attrend2 5 > attrerrcomm 1 > attrerrtype 1 > +attrdeleteget 1 > attr2type 1 > attrorder 1 > attrordercomm 1 > -- Pavan Balaji http://www.mcs.anl.gov/~balaji

1 0

Re: [mpich-discuss] [PATCH] config: change PAC_FUNC_NEEDS_DECL to check by assignment
by Pavan Balaji 18 May '13

18 May '13

Hi Jed, Thanks. This looks OK to me. Can you send the patch as an attachment, so I can apply it? There's also the autoconf macro AC_CHECK_DECLS, which I was planning to migrate to after some testing. I haven't investigated what it does internally, though. -- Pavan On 05/17/2013 10:41 PM US Central Time, Jed Brown wrote: > The old method attempts to create a failed compile when the function > *is* already declared, by creating an incompatible declaration. That is > insufficient with clang-3.2, which only warns in case of incompatible > redeclaration of a library function: > > conftest.c:44:5: warning: incompatible redeclaration of library function 'strncmp' > int strncmp(double, int, double, const char *); > ^ > /usr/include/string.h:143:12: note: 'strncmp' is a builtin with type 'int (const char *, const char *, size_t)' > extern int strncmp (const char *__s1, const char *__s2, size_t __n) > ^ > 1 warning generated. > > Instead, we take the opposite approach, generating a failed compile when > the library function has *not* been declared in the header. > > void (*fptr)(void) = (void(*)(void))strncmp; > > Signed-off-by: Jed Brown <jedbrown(a)mcs.anl.gov> > --- > confdb/aclocal_cc.m4 | 21 +++++++++------------ > 1 file changed, 9 insertions(+), 12 deletions(-) > > diff --git a/confdb/aclocal_cc.m4 b/confdb/aclocal_cc.m4 > index 7a92a8a..10964e1 100644 > --- a/confdb/aclocal_cc.m4 > +++ b/confdb/aclocal_cc.m4 > @@ -1205,25 +1205,22 @@ dnl Sets 'NEEDS_<funcname>_DECL' if 'funcname' is not declared by the > dnl headerfiles. > dnl > dnl Approach: > -dnl Try to compile a program with the function, but passed with an incorrect > -dnl calling sequence. If the compilation fails, then the declaration > -dnl is provided within the header files. If the compilation succeeds, > -dnl the declaration is required. > +dnl Attempt to assign library function to function pointer. If the function > +dnl is not declared in a header, this will fail. Use a non-static global so > +dnl the compiler does not warn about an unused variable. > dnl > -dnl We use a 'double' as the first argument to try and catch varargs > -dnl routines that may use an int or pointer as the first argument. > +dnl Simply calling the function is not enough because C89 compilers allow > +dnl calls to implicitly-defined functions. Re-declaring a library function > +dnl with an incompatible prototype is also not sufficient because some > +dnl compilers (notably clang-3.2) only produce a warning in this case. > dnl > -dnl There is one difficulty - if the compiler has been instructed to > -dnl fail on implicitly defined functions, then this test will always > -dnl fail. > -dnl > dnl D*/ > AC_DEFUN([PAC_FUNC_NEEDS_DECL],[ > AC_CACHE_CHECK([whether $2 needs a declaration], > pac_cv_func_decl_$2,[ > AC_TRY_COMPILE([$1 > -int $2(double, int, double, const char *);],[int a=$2(1.0,27,1.0,"foo");], > -pac_cv_func_decl_$2=yes,pac_cv_func_decl_$2=no)]) > +void (*fptr)(void) = (void(*)(void))$2;],[], > +pac_cv_func_decl_$2=no,pac_cv_func_decl_$2=yes)]) > if test "$pac_cv_func_decl_$2" = "yes" ; then > changequote(<<,>>)dnl > define(<<PAC_FUNC_NAME>>, translit(NEEDS_$2_DECL, [a-z *], [A-Z__]))dnl > -- Pavan Balaji http://www.mcs.anl.gov/~balaji

1 0

Re: [mpich-discuss] [PATCH v2] test: add attrdeleteget, MPI_Attr_get called from delete_fn
by Jed Brown 18 May '13

18 May '13

Fab Tillier <ftillier(a)microsoft.com> writes: >>+int main(int argc,char **argv) >>+{ >>+ MPI_Comm scomm; >>+ >>+ MTest_Init(&argc,&argv); >>+ MPI_Comm_split(MPI_COMM_WORLD,1,0,&scomm); >>+ MPI_Keyval_create(MPI_NULL_COPY_FN,delete_fn,&key,(void*)0); >>+ MPI_Attr_put(scomm,key,a); >>+ MPI_Comm_free(&scomm); > > The communicator handle will become invalid at some point after you > call MPI_Comm_free (but before the call returns). I didn't see in the > standard any mention of whether using the communicator in the destroy > callback is acceptable. Why pass the communicator as an argument to delete_fn if the MPI implementation is allowed to have already made it invalid? Since delete_fn is called by MPI_Comm_free, any use of the first argument passed to delete_fn would be erroneous. IMO, the presence of the argument and lack of any comments stating that it should never be used imply that it was intended to be valid. It is valid with all other MPI implementations that we've used (most of them). > There's the usual vague notion of the call to MPI_Comm_free being > erroneous if the delete callback fails. Yes, the test relies on that. The problem is MPI_Attr_get returning an error when called from delete_fn, presumably due to eager invalidation of the communicator. >>+ MPI_Finalize(); >>+ return 0; >>+} >>+ >>+int delete_fn(MPI_Comm comm,int keyval,void *attr_val,void *extra_state) >>+{ >>+ int flg; >>+ void *ptr; >>+ >>+ return MPI_Attr_get(comm,key,&ptr,&flg); > > Why aren't you using attr_val and extra_state here, rather than > calling MPI_Attr_get? Is this just a symptom of the test case being > reduced, and the original code attempts to access a different > attribute on that communicator? It's a consequence of reducing, but PETSc can restructure (by using separate destructors for each of the two relevant keyvals) so that it is not necessary to call MPI_Attr_get in a delete_fn. I'll make that change, but I think it is worth fixing in MS-MPI as well.

1 0

Re: [mpich-discuss] [PATCH v2] test: add attrdeleteget, MPI_Attr_get called from delete_fn
by Fab Tillier 18 May '13

18 May '13

Hi Jed, Satish, Comments inline to the code. Cheers, -Fab Jed Brown wrote on Fri, 17 May 2013 at 21:52:38 >With MS-MPI 64-bit from HPC Pack 2008 and 2012, MPI_Attr_get returns >error code 773 when called from delete_fn on a communicator obtained >from MPI_Comm_split. > >Signed-off-by: Jed Brown <jedbrown(a)mcs.anl.gov> >Signed-off-by: Satish Balay <balay(a)mcs.anl.gov> >--- >Updated to propagate the error code from MPI_Attr_get. <snip> >diff --git a/test/mpi/attr/attrdeleteget.c b/test/mpi/attr/attrdeleteget.c >new file mode 100644 >index 0000000..05afaa4 >--- /dev/null >+++ b/test/mpi/attr/attrdeleteget.c >@@ -0,0 +1,36 @@ >+/* -*- Mode: C; c-basic-offset:4 ; indent-tabs-mode:nil ; -*- */ >+/* >+ * >+ * (C) 2013 by Argonne National Laboratory. >+ * See COPYRIGHT in top-level directory. >+ */ >+ >+#include <stdio.h> >+#include "mpi.h" >+#include "mpitest.h" >+ >+int key = MPI_KEYVAL_INVALID; >+char a[100]; >+ >+int delete_fn (MPI_Comm, int, void *, void *); >+ >+int main(int argc,char **argv) >+{ >+ MPI_Comm scomm; >+ >+ MTest_Init(&argc,&argv); >+ MPI_Comm_split(MPI_COMM_WORLD,1,0,&scomm); >+ MPI_Keyval_create(MPI_NULL_COPY_FN,delete_fn,&key,(void*)0); >+ MPI_Attr_put(scomm,key,a); >+ MPI_Comm_free(&scomm); The communicator handle will become invalid at some point after you call MPI_Comm_free (but before the call returns). I didn't see in the standard any mention of whether using the communicator in the destroy callback is acceptable. There's the usual vague notion of the call to MPI_Comm_free being erroneous if the delete callback fails. >+ MPI_Finalize(); >+ return 0; >+} >+ >+int delete_fn(MPI_Comm comm,int keyval,void *attr_val,void *extra_state) >+{ >+ int flg; >+ void *ptr; >+ >+ return MPI_Attr_get(comm,key,&ptr,&flg); Why aren't you using attr_val and extra_state here, rather than calling MPI_Attr_get? Is this just a symptom of the test case being reduced, and the original code attempts to access a different attribute on that communicator? I'd like to understand the original use case better, though, if you don't mind. >+}

1 0

[PATCH v2] test: add attrdeleteget, MPI_Attr_get called from delete_fn
by Jed Brown 17 May '13

17 May '13

With MS-MPI 64-bit from HPC Pack 2008 and 2012, MPI_Attr_get returns error code 773 when called from delete_fn on a communicator obtained from MPI_Comm_split. Signed-off-by: Jed Brown <jedbrown(a)mcs.anl.gov> Signed-off-by: Satish Balay <balay(a)mcs.anl.gov> --- Updated to propagate the error code from MPI_Attr_get. test/mpi/.gitignore | 1 + test/mpi/attr/Makefile.am | 1 + test/mpi/attr/attrdeleteget.c | 36 ++++++++++++++++++++++++++++++++++++ test/mpi/attr/testlist | 1 + 4 files changed, 39 insertions(+) create mode 100644 test/mpi/attr/attrdeleteget.c diff --git a/test/mpi/.gitignore b/test/mpi/.gitignore index 3aadba2..0bdcd15 100644 --- a/test/mpi/.gitignore +++ b/test/mpi/.gitignore @@ -51,6 +51,7 @@ /attr/attrerr /attr/attrerrcomm /attr/attrerrtype +/attr/attrdeleteget /attr/attric /attr/attrorder /attr/attrordercomm diff --git a/test/mpi/attr/Makefile.am b/test/mpi/attr/Makefile.am index a672192..e1d788b 100644 --- a/test/mpi/attr/Makefile.am +++ b/test/mpi/attr/Makefile.am @@ -18,6 +18,7 @@ noinst_PROGRAMS = \ attrerr \ attrerrcomm \ attrerrtype \ + attrdeleteget \ attr2type \ attrorder \ attrordercomm \ diff --git a/test/mpi/attr/attrdeleteget.c b/test/mpi/attr/attrdeleteget.c new file mode 100644 index 0000000..05afaa4 --- /dev/null +++ b/test/mpi/attr/attrdeleteget.c @@ -0,0 +1,36 @@ +/* -*- Mode: C; c-basic-offset:4 ; indent-tabs-mode:nil ; -*- */ +/* + * + * (C) 2013 by Argonne National Laboratory. + * See COPYRIGHT in top-level directory. + */ + +#include <stdio.h> +#include "mpi.h" +#include "mpitest.h" + +int key = MPI_KEYVAL_INVALID; +char a[100]; + +int delete_fn (MPI_Comm, int, void *, void *); + +int main(int argc,char **argv) +{ + MPI_Comm scomm; + + MTest_Init(&argc,&argv); + MPI_Comm_split(MPI_COMM_WORLD,1,0,&scomm); + MPI_Keyval_create(MPI_NULL_COPY_FN,delete_fn,&key,(void*)0); + MPI_Attr_put(scomm,key,a); + MPI_Comm_free(&scomm); + MPI_Finalize(); + return 0; +} + +int delete_fn(MPI_Comm comm,int keyval,void *attr_val,void *extra_state) +{ + int flg; + void *ptr; + + return MPI_Attr_get(comm,key,&ptr,&flg); +} diff --git a/test/mpi/attr/testlist b/test/mpi/attr/testlist index 61bd613..4db94df 100644 --- a/test/mpi/attr/testlist +++ b/test/mpi/attr/testlist @@ -7,6 +7,7 @@ attrend2 1 attrend2 5 attrerrcomm 1 attrerrtype 1 +attrdeleteget 1 attr2type 1 attrorder 1 attrordercomm 1 -- 1.8.2.3

1 0

Re: [mpich-discuss] [PATCH] test: add attrdeleteget, MPI_Attr_get called from delete_fn
by Jed Brown 17 May '13

17 May '13

Jed Brown <jedbrown(a)mcs.anl.gov> writes: > + MPI_Attr_get(comm,key,&ptr,&flg); > + return MPI_SUCCESS; Looks like I consolidated too much from the version Satish actually ran. MPI_Attr_get is not aborting, it's just returning nonzero. Rewriting the above as return MPI_Attr_get(comm,key,&ptr,&flg); is sufficiont. Updated version of the patch coming.

1 0

[PATCH] test: add attrdeleteget, MPI_Attr_get called from delete_fn
by Jed Brown 17 May '13

17 May '13

MS-MPI 64-bit from HPC Pack 2008 and 2012 errors when MPI_Attr_get is called from delete_fn on a communicator obtained from MPI_Comm_split. Signed-off-by: Jed Brown <jedbrown(a)mcs.anl.gov> Signed-off-by: Satish Balay <balay(a)mcs.anl.gov> --- This represents a reduced test case from a reported PETSc error when using split communicators on MS-MPI. PETSc uses a reference-counted "inner" communicator to manage tags between objects. When the user destroys the outer communicator, the inner communicator must be detached to avoid holding a dangling reference. I'm submitting this bug report as a patch to MPICH in hopes that it makes it downstream to Microsoft and other vendors so that this problem does not occur again. test/mpi/.gitignore | 1 + test/mpi/attr/Makefile.am | 1 + test/mpi/attr/attrdeleteget.c | 37 +++++++++++++++++++++++++++++++++++++ test/mpi/attr/testlist | 1 + 4 files changed, 40 insertions(+) create mode 100644 test/mpi/attr/attrdeleteget.c diff --git a/test/mpi/.gitignore b/test/mpi/.gitignore index 3aadba2..0bdcd15 100644 --- a/test/mpi/.gitignore +++ b/test/mpi/.gitignore @@ -51,6 +51,7 @@ /attr/attrerr /attr/attrerrcomm /attr/attrerrtype +/attr/attrdeleteget /attr/attric /attr/attrorder /attr/attrordercomm diff --git a/test/mpi/attr/Makefile.am b/test/mpi/attr/Makefile.am index a672192..e1d788b 100644 --- a/test/mpi/attr/Makefile.am +++ b/test/mpi/attr/Makefile.am @@ -18,6 +18,7 @@ noinst_PROGRAMS = \ attrerr \ attrerrcomm \ attrerrtype \ + attrdeleteget \ attr2type \ attrorder \ attrordercomm \ diff --git a/test/mpi/attr/attrdeleteget.c b/test/mpi/attr/attrdeleteget.c new file mode 100644 index 0000000..7ed66d0 --- /dev/null +++ b/test/mpi/attr/attrdeleteget.c @@ -0,0 +1,37 @@ +/* -*- Mode: C; c-basic-offset:4 ; indent-tabs-mode:nil ; -*- */ +/* + * + * (C) 2013 by Argonne National Laboratory. + * See COPYRIGHT in top-level directory. + */ + +#include <stdio.h> +#include "mpi.h" +#include "mpitest.h" + +int key = MPI_KEYVAL_INVALID; +char a[100]; + +int delete_fn (MPI_Comm, int, void *, void *); + +int main(int argc,char **argv) +{ + MPI_Comm scomm; + + MTest_Init(&argc,&argv); + MPI_Comm_split(MPI_COMM_WORLD,1,0,&scomm); + MPI_Keyval_create(MPI_NULL_COPY_FN,delete_fn,&key,(void*)0); + MPI_Attr_put(scomm,key,a); + MPI_Comm_free(&scomm); + MPI_Finalize(); + return 0; +} + +int delete_fn(MPI_Comm comm,int keyval,void *attr_val,void *extra_state) +{ + int flg; + void *ptr; + + MPI_Attr_get(comm,key,&ptr,&flg); + return MPI_SUCCESS; +} diff --git a/test/mpi/attr/testlist b/test/mpi/attr/testlist index 61bd613..4db94df 100644 --- a/test/mpi/attr/testlist +++ b/test/mpi/attr/testlist @@ -7,6 +7,7 @@ attrend2 1 attrend2 5 attrerrcomm 1 attrerrtype 1 +attrdeleteget 1 attr2type 1 attrorder 1 attrordercomm 1 -- 1.8.2.3

1 0

[PATCH] config: change PAC_FUNC_NEEDS_DECL to check by assignment
by Jed Brown 17 May '13

17 May '13

The old method attempts to create a failed compile when the function *is* already declared, by creating an incompatible declaration. That is insufficient with clang-3.2, which only warns in case of incompatible redeclaration of a library function: conftest.c:44:5: warning: incompatible redeclaration of library function 'strncmp' int strncmp(double, int, double, const char *); ^ /usr/include/string.h:143:12: note: 'strncmp' is a builtin with type 'int (const char *, const char *, size_t)' extern int strncmp (const char *__s1, const char *__s2, size_t __n) ^ 1 warning generated. Instead, we take the opposite approach, generating a failed compile when the library function has *not* been declared in the header. void (*fptr)(void) = (void(*)(void))strncmp; Signed-off-by: Jed Brown <jedbrown(a)mcs.anl.gov> --- confdb/aclocal_cc.m4 | 21 +++++++++------------ 1 file changed, 9 insertions(+), 12 deletions(-) diff --git a/confdb/aclocal_cc.m4 b/confdb/aclocal_cc.m4 index 7a92a8a..10964e1 100644 --- a/confdb/aclocal_cc.m4 +++ b/confdb/aclocal_cc.m4 @@ -1205,25 +1205,22 @@ dnl Sets 'NEEDS_<funcname>_DECL' if 'funcname' is not declared by the dnl headerfiles. dnl dnl Approach: -dnl Try to compile a program with the function, but passed with an incorrect -dnl calling sequence. If the compilation fails, then the declaration -dnl is provided within the header files. If the compilation succeeds, -dnl the declaration is required. +dnl Attempt to assign library function to function pointer. If the function +dnl is not declared in a header, this will fail. Use a non-static global so +dnl the compiler does not warn about an unused variable. dnl -dnl We use a 'double' as the first argument to try and catch varargs -dnl routines that may use an int or pointer as the first argument. +dnl Simply calling the function is not enough because C89 compilers allow +dnl calls to implicitly-defined functions. Re-declaring a library function +dnl with an incompatible prototype is also not sufficient because some +dnl compilers (notably clang-3.2) only produce a warning in this case. dnl -dnl There is one difficulty - if the compiler has been instructed to -dnl fail on implicitly defined functions, then this test will always -dnl fail. -dnl dnl D*/ AC_DEFUN([PAC_FUNC_NEEDS_DECL],[ AC_CACHE_CHECK([whether $2 needs a declaration], pac_cv_func_decl_$2,[ AC_TRY_COMPILE([$1 -int $2(double, int, double, const char *);],[int a=$2(1.0,27,1.0,"foo");], -pac_cv_func_decl_$2=yes,pac_cv_func_decl_$2=no)]) +void (*fptr)(void) = (void(*)(void))$2;],[], +pac_cv_func_decl_$2=no,pac_cv_func_decl_$2=yes)]) if test "$pac_cv_func_decl_$2" = "yes" ; then changequote(<<,>>)dnl define(<<PAC_FUNC_NAME>>, translit(NEEDS_$2_DECL, [a-z *], [A-Z__]))dnl -- 1.8.2.3

1 0

Re: [mpich-discuss] mpich 3.0.4 build fails on OS X Mountain lion
by Jeff Hammond 15 May '13

15 May '13

Well, assuming that the problem you see is the same as is discussed on that website, you can: (1) --disable-shared (2) use llvm-gcc instead of clang (3) upgrade to Xcode 4.5.2+ Jeff On Wed, May 15, 2013 at 11:42 AM, Mathieu MORLIGHEM <Mathieu.Morlighem(a)uci.edu> wrote: > Hi Jeff, > > thanks for the link, I will try it. Does this mean that I cannot build shared libs? > Also, this workaround is for mpich2, and the compilation with --enable-shared of mpich2 (version 1.5) works well on my machine, I tried it again this morning… > > Thanks > Mathieu > > On May 15, 2013, at 9:33 AM, Jeff Hammond <jhammond(a)alcf.anl.gov> wrote: > >> I recall that this was addressed a month or two ago on this list but I >> can't find the direct reference. I can, however, point you to >> https://github.com/mxcl/homebrew/pull/16299, which explains the >> details and how you can work around it. >> >> Jeff >> >> On Wed, May 15, 2013 at 11:28 AM, Mathieu MORLIGHEM >> <Mathieu.Morlighem(a)uci.edu> wrote: >>> Hello, >>> >>> I am trying to compile mpich 3.0.4 on my machine: >>> >>> uname -a >>> Darwin shackleton.ess.uci.edu 12.3.0 Darwin Kernel Version 12.3.0: Sun Jan >>> 6 22:37:10 PST 2013; root:xnu-2050.22.13~1/RELEASE_X86_64 x86_64 >>> >>> I successfully installed mpich 1.5 several times with the following options: >>> ./configure --prefix="$ISSM_DIR/externalpackages/mpich2/install" >>> --enable-shared >>> >>> However, with the same options, the build of mpich 3.0.4 at linking with a >>> segmentation fault: >>> >>> Making all in . >>> make[3]: Nothing to be done for `all-am'. >>> Making all in . >>> GEN lib/libmpich.la >>> collect2: ld terminated with signal 11 [Segmentation fault: 11] >>> make[2]: *** [lib/libmpich.la] Error 1 >>> make[1]: *** [all-recursive] Error 1 >>> make: *** [all] Error 2 >>> >>> with VERBOSE=1: >>> /bin/sh ./libtool --tag=F77 --mode=link gfortran -O2 -L./src/mpl >>> -L./src/openpa/src -version-info 10:4:0 -Wl,-flat_namespace -o >>> lib/libmpich.la -rpath >>> /Users/mmorligh/Desktop/issmuci/trunk-jpl/externalpackages/mpich/install/lib >>> src/mpi/attr/attrutil.lo src/mpi/attr/dup_fn.lo src/mpi/coll/allred_group.lo >>> src/mpi/coll/barrier_group.lo ………… src/mpi_t/.libs/pvar_stop.o >>> src/mpi_t/.libs/pvar_write.o src/binding/f77/.libs/setbot.o >>> src/binding/f77/.libs/setbotf.o >>> -Wl,-force_load,src/mpi/romio/.libs/libromio.a >>> -Wl,-force_load,src/mpi/romio/.libs/libpromio.a -L./src/mpl >>> -L./src/openpa/src >>> /Users/mmorligh/Desktop/issmuci/trunk-jpl/externalpackages/mpich/src/src/mpl/.libs/libmpl.dylib >>> /Users/mmorligh/Desktop/issmuci/trunk-jpl/externalpackages/mpich/src/src/openpa/src/.libs/libopa.dylib >>> -lpthread -O2 -Wl,-flat_namespace -install_name >>> /Users/mmorligh/Desktop/issmuci/trunk-jpl/externalpackages/mpich/install/lib/libmpich.10.dylib >>> -compatibility_version 11 -current_version 11.4 -Wl,-single_module >>> collect2: ld terminated with signal 11 [Segmentation fault: 11] >>> make[2]: *** [lib/libmpich.la] Error 1 >>> make[1]: *** [all-recursive] Error 1 >>> make: *** [all] Error 2 >>> >>> Has anybody reported a similar issue? Do you think this is a problem with >>> libtool? >>> Thanks a lot for your help >>> >>> -- >>> Mathieu Morlighem >>> >>> Assistant Project Scientist >>> Department of Earth System Science >>> University of California, Irvine >>> >>> >>> _______________________________________________ >>> discuss mailing list discuss(a)mpich.org >>> To manage subscription options or unsubscribe: >>> https://lists.mpich.org/mailman/listinfo/discuss >> >> >> >> -- >> 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 >> ALCF docs: http://www.alcf.anl.gov/user-guides >> _______________________________________________ >> discuss mailing list discuss(a)mpich.org >> To manage subscription options or unsubscribe: >> https://lists.mpich.org/mailman/listinfo/discuss >> > > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss -- 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 ALCF docs: http://www.alcf.anl.gov/user-guides

1 0

Re: [mpich-discuss] mpich 3.0.4 build fails on OS X Mountain lion
by Mathieu MORLIGHEM 15 May '13

15 May '13

Hi Jeff, thanks for the link, I will try it. Does this mean that I cannot build shared libs? Also, this workaround is for mpich2, and the compilation with --enable-shared of mpich2 (version 1.5) works well on my machine, I tried it again this morning… Thanks Mathieu On May 15, 2013, at 9:33 AM, Jeff Hammond <jhammond(a)alcf.anl.gov> wrote: > I recall that this was addressed a month or two ago on this list but I > can't find the direct reference. I can, however, point you to > https://github.com/mxcl/homebrew/pull/16299, which explains the > details and how you can work around it. > > Jeff > > On Wed, May 15, 2013 at 11:28 AM, Mathieu MORLIGHEM > <Mathieu.Morlighem(a)uci.edu> wrote: >> Hello, >> >> I am trying to compile mpich 3.0.4 on my machine: >> >> uname -a >> Darwin shackleton.ess.uci.edu 12.3.0 Darwin Kernel Version 12.3.0: Sun Jan >> 6 22:37:10 PST 2013; root:xnu-2050.22.13~1/RELEASE_X86_64 x86_64 >> >> I successfully installed mpich 1.5 several times with the following options: >> ./configure --prefix="$ISSM_DIR/externalpackages/mpich2/install" >> --enable-shared >> >> However, with the same options, the build of mpich 3.0.4 at linking with a >> segmentation fault: >> >> Making all in . >> make[3]: Nothing to be done for `all-am'. >> Making all in . >> GEN lib/libmpich.la >> collect2: ld terminated with signal 11 [Segmentation fault: 11] >> make[2]: *** [lib/libmpich.la] Error 1 >> make[1]: *** [all-recursive] Error 1 >> make: *** [all] Error 2 >> >> with VERBOSE=1: >> /bin/sh ./libtool --tag=F77 --mode=link gfortran -O2 -L./src/mpl >> -L./src/openpa/src -version-info 10:4:0 -Wl,-flat_namespace -o >> lib/libmpich.la -rpath >> /Users/mmorligh/Desktop/issmuci/trunk-jpl/externalpackages/mpich/install/lib >> src/mpi/attr/attrutil.lo src/mpi/attr/dup_fn.lo src/mpi/coll/allred_group.lo >> src/mpi/coll/barrier_group.lo ………… src/mpi_t/.libs/pvar_stop.o >> src/mpi_t/.libs/pvar_write.o src/binding/f77/.libs/setbot.o >> src/binding/f77/.libs/setbotf.o >> -Wl,-force_load,src/mpi/romio/.libs/libromio.a >> -Wl,-force_load,src/mpi/romio/.libs/libpromio.a -L./src/mpl >> -L./src/openpa/src >> /Users/mmorligh/Desktop/issmuci/trunk-jpl/externalpackages/mpich/src/src/mpl/.libs/libmpl.dylib >> /Users/mmorligh/Desktop/issmuci/trunk-jpl/externalpackages/mpich/src/src/openpa/src/.libs/libopa.dylib >> -lpthread -O2 -Wl,-flat_namespace -install_name >> /Users/mmorligh/Desktop/issmuci/trunk-jpl/externalpackages/mpich/install/lib/libmpich.10.dylib >> -compatibility_version 11 -current_version 11.4 -Wl,-single_module >> collect2: ld terminated with signal 11 [Segmentation fault: 11] >> make[2]: *** [lib/libmpich.la] Error 1 >> make[1]: *** [all-recursive] Error 1 >> make: *** [all] Error 2 >> >> Has anybody reported a similar issue? Do you think this is a problem with >> libtool? >> Thanks a lot for your help >> >> -- >> Mathieu Morlighem >> >> Assistant Project Scientist >> Department of Earth System Science >> University of California, Irvine >> >> >> _______________________________________________ >> discuss mailing list discuss(a)mpich.org >> To manage subscription options or unsubscribe: >> https://lists.mpich.org/mailman/listinfo/discuss > > > > -- > 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 > ALCF docs: http://www.alcf.anl.gov/user-guides > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss >

1 0