Bug #1956
tests/charm++/sdag/migration and tests/charm++/sdag/anytimeMigration fail on mpi-win-x86_64-smp with debug options (-g -O0)
Start date:
08/07/2018
Due date:
% Done:
0%
Description
Charm build command: ./build LIBS mpi-win-x86_64 smp --enable-error-checking --without-romio --suffix=debug -j8 -g -O0 |& tee build_result_debug
tests/charm++/sdag/migration:
make -C migration test OPTS='' TESTOPTS='' make[1]: Entering directory '/home/nikhil/nitin/charm/mpi-win-x86_64-smp-debug/tests/charm++/sdag/migration' ../../../../bin/testrun +p2 ./test1 Running on 2 processors: ./test1 charmrun> /cygdrive/c/Program Files/Microsoft MPI/Bin/mpiexec -n 2 ./test1 Charm++> Running on MPI version: 2.0 Charm++> level of thread support used: MPI_THREAD_FUNNELED (desired: MPI_THREAD_FUNNELED) Charm++> Running in SMP mode: 2 processes, 1 worker threads (PEs) + 1 comm threads per process, 0 PEs total Charm++> The comm. thread both sends and receives messages Charm++ warning> fences and atomic operations not available in native assembly Converse/Charm++ Commit ID: v6.8.2-853-g4146bf788 Charm++> Disabling isomalloc because mmap() does not work. CharmLB> Load balancer assumes all CPUs are same. Charm++> Running on 1 hosts (1 sockets x 4 cores x 2 PUs = 8-way SMP) Charm++> cpu topology info is gathered in 0.000 seconds. Running Parallel on 2 processors for 25 elements Migrating from PE 0 Migrating from PE 0 Migrating from PE 0 Migrating from PE 0 Migrating from PE 0 Migrating from PE 0 Migrating from PE 0 Migrating from PE 0 Migrating from PE 0 Migrating from PE 0 Migrating from PE 0 Migrating from PE 0 Migrating from PE 0 (0, 0): received finished message (0, 1): received finished message (0, 2): received finished message (0, 3): received finished message (0, 4): received finished message (1, 0): received finished message (1, 1): received finished message (1, 2): received finished message (1, 3): received finished message (1, 4): received finished message (2, 0): received finished message (2, 1): received finished message (2, 2): received finished message Migrating from PE 1 called PUP for cell unpacking or sizing called PUP for cell packing Migrating from PE 1 called PUP for cell unpacking or sizing called PUP for cell unpacking or sizing called PUP for cell packing Migrating from PE 1 called PUP for cell unpacking or sizing called PUP for cell unpacking or sizing called PUP for cell packing Migrating from PE 1 called PUP for cell unpacking or sizing called PUP for cell packing called PUP for cell unpacking or sizing Migrating from PE 1 called PUP for cell unpacking or sizing called PUP for cell unpacking or sizing called PUP for cell packing Migrating from PE 1 called PUP for cell unpacking or sizing called PUP for cell unpacking or sizing called PUP for cell packing called PUP for cell unpacking or sizing job aborted: [ranks] message [0] terminated [1] process exited without calling finalize ---- error analysis ----- [1] on CS-DEXTERITY ./test1 ended prematurely and may have crashed. exit code 0xc0000417 ---- error analysis ----- make[1]: *** [Makefile:28: test] Error 127 make[1]: Leaving directory '/home/nikhil/nitin/charm/mpi-win-x86_64-smp-debug/tests/charm++/sdag/migration'
tests/charm++/sdag/anytimeMigration:
make[1]: Entering directory '/home/nikhil/nitin/charm/mpi-win-x86_64-smp-debug/tests/charm++/sdag/anytimeMigration' ../../../../bin/charmc migration.ci ../../../../bin/charmc migration.C -o migration migration.C moduleinit5796.C Ignored Unrecognized argument -Wl,--export-dynamic ../../../../bin/testrun +p2 ./migration Running on 2 processors: ./migration charmrun> /cygdrive/c/Program Files/Microsoft MPI/Bin/mpiexec -n 2 ./migration Charm++> Running on MPI version: 2.0 Charm++> level of thread support used: MPI_THREAD_FUNNELED (desired: MPI_THREAD_FUNNELED) Charm++> Running in SMP mode: 2 processes, 1 worker threads (PEs) + 1 comm threads per process, 0 PEs total Charm++> The comm. thread both sends and receives messages Charm++ warning> fences and atomic operations not available in native assembly Converse/Charm++ Commit ID: v6.8.2-853-g4146bf788 Charm++> Disabling isomalloc because mmap() does not work. CharmLB> Load balancer assumes all CPUs are same. Charm++> Running on 1 hosts (1 sockets x 4 cores x 2 PUs = 8-way SMP) Charm++> cpu topology info is gathered in 0.000 seconds. running SDAG migration test job aborted: [ranks] message [0] terminated [1] process exited without calling finalize ---- error analysis ----- [1] on CS-DEXTERITY ./migration ended prematurely and may have crashed. exit code 0xc0000417 ---- error analysis ----- make[1]: *** [Makefile:23: test] Error 127 make[1]: Leaving directory '/home/nikhil/nitin/charm/mpi-win-x86_64-smp-debug/tests/charm++/sdag/anytimeMigration'
History
#1 Updated by Eric Bohm 7 months ago
On a related note, the format of testing for mpi-win-smp seems to be running all tests with PPN = 1. Which means its not really exercising the multiple worker thread aspect that motivates use of the smp target. So all of PPN>1 could be broken and nightly build would not tell us. This creates a strong expectation shortfall. When looking at autobuild for mpi-win-smp we are misled into believing we have a good test for it, when we are in fact barely hitting any of the interesting smp use cases.
#2 Updated by Evan Ramos 6 months ago
- Target version deleted (
6.9.0)
#4 Updated by Evan Ramos 3 months ago
Was this fixed by #1960?