• Bug#1102612: mpich 4.3 not initialising multiple processes

    From Drew Parsons@21:1/5 to All on Fri Apr 11 16:00:02 2025
    Package: mpich
    Version: 4.3.0-5
    Followup-For: Bug #1102612

    One data point might be relevant. The bug reported here manifests in
    the amd64 and arm64 tests of armci-mpi.

    But the armci-mpi tests are passing cleanly on the 32-bit arches, armhf and
    the others.

    I note that the build time mpich errors for armci-mpi, seen in https://buildd.debian.org/status/fetch.php?pkg=armci-mpi&arch=amd64&ver=0.4-5&stamp=1744327219&raw=0
    refer to UCX:

    FAIL: benchmarks/ping-pong
    ==========================
    [1744327153.607644] [sbuild:19884:0] sock.c:513 UCX WARN unable to read somaxconn value from /proc/sys/net/core/somaxconn file
    [0] ARMCI Error: This benchmark should be run on at least two processes

    That suggests to me the issue might be in one of the auxiliary
    libraries e.g. libucx0. Perhaps it's a red herring. My trivial test
    fails but does not trigger a UCX warning.


    One other point, I mentioned that in my first run of the trivial test,
    it "passed", emitting pmix warnings. I can reproduce that by running
    the mpich test using mpiexec.openmpi instead of mpiexec.mpich. Of
    course one should not launch it with the wrong mpi that way. I think I
    must have used mpiexec instead of mpiexec.mpich in that first test.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)