[mpich-discuss] MPICH 3.4.1 and macOS 11.2 (aka Big Sur)

Eric Borisch eborisch at gmail.com
Fri Feb 19 10:33:43 CST 2021


Yes: (Again, only testing locally)

$ MPIR_CVAR_OFI_SKIP_IPV6=1 mpiexec -np 2  ./cpi
Process 0 of 2 is on <HOSTNAME>
Process 1 of 2 is on <HOSTNAME>
pi is approximately 3.1415926544231318, Error is 0.0000000008333387
wall clock time = 0.000268

while without:

$ mpiexec -np 2  ./cpi
Assertion failed in file src/mpid/ch4/netmod/ofi/ofi_init.c at line
1988: mapped_table[i] != FI_ADDR_NOTAVAIL
Assertion failed in file src/mpid/ch4/netmod/ofi/ofi_init.c at line
1988: mapped_table[i] != FI_ADDR_NOTAVAIL
0   libpmpi.12.dylib                    0x000000010d3e8b04
MPL_backtrace_show + 52

Thanks! I'll patch our installation to crowbar that to 1 for now, if
that seems reasonable.

Thanks,
  - Eric


On Fri, Feb 19, 2021 at 7:56 AM Zhou, Hui <zhouh at anl.gov> wrote:
>
> Does setting environment variable `MPIR_CVAR_OFI_SKIP_IPV6=1` work?
>
>
>
> --
> Hui Zhou
>
>
>
>
>
> From: Eric Borisch via discuss <discuss at mpich.org>
> Date: Thursday, February 18, 2021 at 7:36 PM
> To: discuss at mpich.org <discuss at mpich.org>
> Cc: Eric Borisch <eborisch at gmail.com>
> Subject: Re: [mpich-discuss] MPICH 3.4.1 and macOS 11.2 (aka Big Sur)
>
> Blaise,
>
>
>
> This is 3.4.1 and CH4?
>
>
>
> With —enable-fast=all, asserts are disabled; can you see if it will compile and execute mpi binaries when built without that option? I suspect it will catch the invalid condition (with the non-elided assert), and would likewise crash if it was used in actual distributed (non-localhost) execution.
>
>
>
> Let me know; if it doesn’t crash for you I would love to figure out why.
>
>
>
>  - Eric
>
>
>
> On Thu, Feb 18, 2021 at 3:59 PM Blaise A Bourdin via discuss <discuss at mpich.org> wrote:
>
> Hi,
>
> Absolutely no issues here when using gcc10 from homebrew both on intel and ARM.
>
> Here is my configuration options:
>
> ./configure --prefix=$MPICH_HOME        \
>
>             --enable-fast=all,O3        \
>
>             --enable-g=dbg              \
>
>             --enable-romio              \
>
>             --enable-shared             \
>
>             --enable-sharedlibs=gcc-osx \
>
>             --with-pm=hydra             \
>
>             CC=gcc-10 CXX=g++-10 FC=gfortran-10 F77=gfortran-10 \
>
>             CFLAGS="-fgnu89-inline -Wno-deprecated" \
>
>             CXXFLAGS="-Wno-deprecated" \
>
>             FFLAGS="-fallow-argument-mismatch"
>
>
>
> Blaise
>
>
>
> On Feb 18, 2021, at 3:53 PM, Allen Sanderson via discuss <discuss at mpich.org> wrote:
>
>
>
> Hello,
>
>
>
> First post to the discussion board. I made the mistake of upgrading to macOS 11.2 (aka Big Sur). I was able to build MPICH 3.4.1 successfully and integrate it with an application. However, when I ran, MPICH crashed. Here is the initial output:
>
>
>
> Abort(1) on node 0: Internal error
>
> 0   libpmpi.12.dylib                    0x00000001068c2159 MPL_backtrace_show + 57
>
> 1   libpmpi.12.dylib                    0x000000010684e874 MPIR_Assert_fail + 36
>
> 2   libpmpi.12.dylib                    0x0000000106897bc8 MPIDI_OFI_mpi_init_hook + 4472
>
> 3   libpmpi.12.dylib                    0x00000001068790d3 MPID_Init + 3091
>
> 4   libpmpi.12.dylib                    0x00000001066ae70b MPIR_Init_thread + 219
>
> 5   libmpi.12.dylib                     0x00000001064bb897 MPI_Init + 279
>
>
>
> I have not yet dug into the issue any further. Before doing so I wanted to first check to see if others have successfully built and run via macOS 11.2. If so I would appreciate any pointers. If it is known that it will not yet work, that would be good information as well so not to spend anytime digging.
>
>
>
> I should note that on another machine, I was wiser and upgraded to macOS 10.5.7 (aka Catalina) and built and ran with 3.3.1 that worked fine running in parallel.
>
>
>
> Cheers,
>
>
>
> Allen
>
>
>
>
>
> Allen Sanderson
>
> SCI Institute
>
> University of Utah
>
>
>
>
>
>
> _______________________________________________
> discuss mailing list     discuss at mpich.org
> To manage subscription options or unsubscribe:
> https://lists.mpich.org/mailman/listinfo/discuss
>
>
>
> --
>
> A.K. & Shirley Barton Professor of  Mathematics
>
> Adjunct Professor of Mechanical Engineering
>
> Adjunct of the Center for Computation & Technology
>
> Louisiana State University, Lockett Hall Room 344, Baton Rouge, LA 70803, USA
>
> Tel. +1 (225) 578 1612, Fax  +1 (225) 578 4276 Web http://www.math.lsu.edu/~bourdin
>
>
>
> _______________________________________________
> discuss mailing list     discuss at mpich.org
> To manage subscription options or unsubscribe:
> https://lists.mpich.org/mailman/listinfo/discuss


More information about the discuss mailing list