fyi-
trying to use the new release of open mpi (v1.1), make check for meep-
mpi fails in cylindrical.dac:
mpirun -np 2 ./cylindrical.dac
Using MPI version 2.0, 2 processes
Testing cylindrical coords under different splittings...
Checking at r == 0 with m = 0...
Checking at r == 0 with m = 1...
Checking at r == 0 with m = 2...
Trying test pattern with m = 0 and 2 chunks...
First chunk is 0.7 by 0.8
Trying test pattern with m = 0 and 3 chunks...
First chunk is 0.4 by 0.8
Trying test pattern with m = 0 and 4 chunks...
First chunk is 0.7 by 0.4
Trying test pattern with m = 0 and 5 chunks...
First chunk is 0.6 by 0.4
Trying with m = 0 and a splitting into 2 chunks...
Chunks are 0.7 by 0.8
Trying with m = 1 and a splitting into 2 chunks...
Chunks are 0.7 by 0.8
Trying with m = 2 and a splitting into 2 chunks...
Chunks are 0.7 by 0.8
Trying with m = 0 and a splitting into 3 chunks...
Chunks are 0.4 by 0.8
Trying with m = 1 and a splitting into 3 chunks...
Chunks are 0.4 by 0.8
Trying with m = 2 and a splitting into 3 chunks...
Chunks are 0.4 by 0.8
Metallic with m = 0 and a splitting into 2 chunks...
Chunks are 0.7 by 0.8
Metallic with m = 1 and a splitting into 2 chunks...
Chunks are 0.7 by 0.8
Metallic with m = 2 and a splitting into 2 chunks...
Chunks are 0.7 by 0.8
Metallic with m = 0 and a splitting into 3 chunks...
Chunks are 0.4 by 0.8
Metallic with m = 1 and a splitting into 3 chunks...
Chunks are 0.4 by 0.8
Metallic with m = 2 and a splitting into 3 chunks...
Chunks are 0.4 by 0.8
Metallic with m = 0 and a splitting into 4 chunks...
Chunks are 0.7 by 0.4
Metallic with m = 1 and a splitting into 4 chunks...
Chunks are 0.7 by 0.4
Metallic with m = 2 and a splitting into 4 chunks...
Chunks are 0.7 by 0.4
PML with m = 0 and a splitting into 2 chunks...
[mac1:20932] *** An error occurred in MPI_File_open
[mac1:20932] *** on communicator MPI_COMM_WORLD
[mac1:20932] *** MPI_ERR_INFO: invalid info object
[mac1:20932] *** MPI_ERRORS_ARE_FATAL (goodbye)
[mac1:20930] *** An error occurred in MPI_File_open
[mac1:20930] *** on communicator MPI_COMM_WORLD
[mac1:20930] *** MPI_ERR_INFO: invalid info object
[mac1:20930] *** MPI_ERRORS_ARE_FATAL (goodbye)
1 process killed (possibly by Open MPI)
make[2]: *** [cylindrical.done] Error 137
make[1]: *** [check-am] Error 2
make: *** [check-recursive] Error 1
you can trace the error to openmpi_1.1/ompi/mpi/c/file_open.c line
46. this happens on both RH enterprise 4 update 2 (gcc v3.4.4
20050721 (Red Hat 3.4.4-2)) and on Mac OS X 10.4.7 (gcc 4.0.1 (Apple
Computer, Inc. build 5363)). make check on open mpi itself passes.
btw, i reproduced the problem reported in "apparent violation of
Poynting's theorem?" on Mac OS. it's not platform dependent, it appears.
thanks,
gp
------
_______________________________________________
meep-discuss mailing list
[email protected]
http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/meep-discuss
