Hi All, I have x10-trunk building with MVAPICH and running on our cluster at OSC. I have been able to run x10rt_basic.mpi as well as examples from my tutorial using DistArrays.
I turned my attention to the HPC Challenge implementations in the benchmarks repository. I've hit problems with FT and LU. LU first, since its the one I really wanted to try. I've built it with the pgas_sockets runtime, and it runs: dhu...@opt-login01 1021%> make LU make -C LU lu-opt make[1]: Entering directory `/nfs/07/dhudak/devel/x10/20101124r1/benchmarks/hpcc/LU' mkdir -p opt-gen/rc7 x10c++ -STATIC_CALLS -O -NO_CHECKS -d opt-gen -report postcompile=5 -x10rt pgas_sockets rc7/*.x10 -post '# -O3 # # -L/usr/local/goto-blas-1.15-gfortran/lib -lgoto' -o lu-opt Output files: [rc7/Timer.h, rc7/BlockedArray__Block.cc, rc7/BlockedArray__Block.h, rc7/BlockedArray.inc, rc7/Timer.inc, rc7/BlockedArray__ArrayView.inc, rc7/BlockedArray__ArrayView.cc, rc7/BlockedArray.cc, rc7/BlockedArray__ArrayView.h, rc7/essl_natives.h, rc7/LU.inc, rc7/BlockedArray__Block.inc, rc7/essl_natives.cc, rc7/LU.h, rc7/LU.cc, rc7/Timer.cc, rc7/BlockedArray.h] Executing post-compiler g++ -O3 -g -I/nfs/07/dhudak/devel/x10/20101122r1/x10-trunk/x10.dist/include -Iopt-gen -I. -O2 -DNDEBUG -DNO_PLACE_CHECKS -finline-functions -DNO_CHECKS -Wno-long-long -Wno-unused-parameter -pthread -msse2 -mfpmath=sse -o /nfs/07/dhudak/devel/x10/20101124r1/benchmarks/hpcc/LU/lu-opt rc7/BlockedArray__ArrayView.cc rc7/BlockedArray.cc rc7/BlockedArray__Block.cc rc7/essl_natives.cc rc7/LU.cc rc7/Timer.cc -L/nfs/07/dhudak/devel/x10/20101122r1/x10-trunk/x10.dist/lib -lx10 -DX10_USE_BDWGC -lgc -lx10rt_pgas_sockets -lpthread -ldl -lm -lpthread -Wl,--rpath -Wl,/nfs/07/dhudak/devel/x10/20101122r1/x10-trunk/x10.dist/lib -Wl,-export-dynamic -lrt -L/usr/local/goto-blas-1.15-gfortran/lib -lgoto make[1]: Leaving directory `/nfs/07/dhudak/devel/x10/20101124r1/benchmarks/hpcc/LU' make -C LU lu-dbg make[1]: Entering directory `/nfs/07/dhudak/devel/x10/20101124r1/benchmarks/hpcc/LU' mkdir -p dbg-gen/rc7 x10c++ -d dbg-gen -report postcompile=5 -x10rt pgas_sockets rc7/*.x10 -post '# # # -L/usr/local/goto-blas-1.15-gfortran/lib -lgoto' -o lu-dbg Output files: [rc7/Timer.h, rc7/BlockedArray__Block.cc, rc7/BlockedArray__Block.h, rc7/BlockedArray.inc, rc7/Timer.inc, rc7/BlockedArray__ArrayView.inc, rc7/BlockedArray__ArrayView.cc, rc7/BlockedArray.cc, rc7/BlockedArray__ArrayView.h, rc7/essl_natives.h, rc7/LU.inc, rc7/BlockedArray__Block.inc, rc7/essl_natives.cc, rc7/LU.h, rc7/LU.cc, rc7/Timer.cc, rc7/BlockedArray.h] Executing post-compiler g++ -g -I/nfs/07/dhudak/devel/x10/20101122r1/x10-trunk/x10.dist/include -Idbg-gen -I. -Wno-long-long -Wno-unused-parameter -pthread -msse2 -mfpmath=sse -o /nfs/07/dhudak/devel/x10/20101124r1/benchmarks/hpcc/LU/lu-dbg rc7/BlockedArray__ArrayView.cc rc7/BlockedArray.cc rc7/BlockedArray__Block.cc rc7/essl_natives.cc rc7/LU.cc rc7/Timer.cc -L/nfs/07/dhudak/devel/x10/20101122r1/x10-trunk/x10.dist/lib -lx10 -DX10_USE_BDWGC -lgc -lx10rt_pgas_sockets -lpthread -ldl -lm -lpthread -Wl,--rpath -Wl,/nfs/07/dhudak/devel/x10/20101122r1/x10-trunk/x10.dist/lib -Wl,-export-dynamic -lrt -L/usr/local/goto-blas-1.15-gfortran/lib -lgoto make[1]: Leaving directory `/nfs/07/dhudak/devel/x10/20101124r1/benchmarks/hpcc/LU' dhu...@opt-login01 1016%> mpirun -np 4 ./lu-opt 128 64 2 2 LU Starting: M 128 B 64 px 2 py 2 Main loop starting ... diff 0.0 1 diff 0.0 3 diff 4.8E-25 0 diff 4.6E-25 2 difference 4.8E-25 ok) diff=4.8E-25 Timer(0) TOTAL #invocations=1 Time=0.10502 seconds Timer(1) PANEL #invocations=2 Time=0.063584 seconds Timer(2) SWAPROWS #invocations=2 Time=0.015885 seconds Timer(3) TRISOLVE #invocations=2 Time=0.002328 seconds Timer(4) UPDATE #invocations=2 Time=0.00231 seconds Timer(5) PANEL-PIVOT #invocations=64 Time=0.017302 seconds Timer(6) PANEL-SWAP #invocations=64 Time=0.013163 seconds Timer(7) PANEL-BCAST #invocations=64 Time=0.005222 seconds Timer(8) PANEL-UPDATE #invocations=64 Time=3.4E-4 seconds Timer(10) ROW-BROADCAST # invocations=3 Time=5.1E-4 seconds Timer(11) COL-BROADCAST # invocations=65 Time=0.005448 seconds Time= 0.105624 seconds Rate= 0.044498674543664 GFlops However, when I build the same code with the MPI runtime, it hangs. After a make LU-clean, I changed the Makefile to: export X10RTTRANSPORT ?= -x10rt mpi ...and did: dhu...@opt-login01 1033%> make LU make -C LU lu-opt make[1]: Entering directory `/nfs/07/dhudak/devel/x10/20101124r1/benchmarks/hpcc/LU' mkdir -p opt-gen/rc7 x10c++ -STATIC_CALLS -O -NO_CHECKS -d opt-gen -report postcompile=5 -x10rt mpi rc7/*.x10 -post '# -O3 # # -L/usr/local/goto-blas-1.15-gfortran/lib -lgoto' -o lu-opt Output files: [rc7/Timer.h, rc7/BlockedArray__Block.cc, rc7/BlockedArray__Block.h, rc7/BlockedArray.inc, rc7/Timer.inc, rc7/BlockedArray__ArrayView.inc, rc7/BlockedArray__ArrayView.cc, rc7/BlockedArray.cc, rc7/BlockedArray__ArrayView.h, rc7/essl_natives.h, rc7/LU.inc, rc7/BlockedArray__Block.inc, rc7/essl_natives.cc, rc7/LU.h, rc7/LU.cc, rc7/Timer.cc, rc7/BlockedArray.h] Executing post-compiler mpicxx -O3 -g -I/nfs/07/dhudak/devel/x10/20101122r1/x10-trunk/x10.dist/include -Iopt-gen -I. -O2 -DNDEBUG -DNO_PLACE_CHECKS -finline-functions -DNO_CHECKS -Wno-long-long -Wno-unused-parameter -pthread -msse2 -mfpmath=sse -o /nfs/07/dhudak/devel/x10/20101124r1/benchmarks/hpcc/LU/lu-opt rc7/BlockedArray__ArrayView.cc rc7/BlockedArray.cc rc7/BlockedArray__Block.cc rc7/essl_natives.cc rc7/LU.cc rc7/Timer.cc -L/nfs/07/dhudak/devel/x10/20101122r1/x10-trunk/x10.dist/lib -lx10 -DX10_USE_BDWGC -lgc -lx10rt_mpi -ldl -lm -lpthread -Wl,--rpath -Wl,/nfs/07/dhudak/devel/x10/20101122r1/x10-trunk/x10.dist/lib -Wl,-export-dynamic -lrt -L/usr/local/goto-blas-1.15-gfortran/lib -lgoto make[1]: Leaving directory `/nfs/07/dhudak/devel/x10/20101124r1/benchmarks/hpcc/LU' make -C LU lu-dbg make[1]: Entering directory `/nfs/07/dhudak/devel/x10/20101124r1/benchmarks/hpcc/LU' mkdir -p dbg-gen/rc7 x10c++ -d dbg-gen -report postcompile=5 -x10rt mpi rc7/*.x10 -post '# # # -L/usr/local/goto-blas-1.15-gfortran/lib -lgoto' -o lu-dbg Output files: [rc7/Timer.h, rc7/BlockedArray__Block.cc, rc7/BlockedArray__Block.h, rc7/BlockedArray.inc, rc7/Timer.inc, rc7/BlockedArray__ArrayView.inc, rc7/BlockedArray__ArrayView.cc, rc7/BlockedArray.cc, rc7/BlockedArray__ArrayView.h, rc7/essl_natives.h, rc7/LU.inc, rc7/BlockedArray__Block.inc, rc7/essl_natives.cc, rc7/LU.h, rc7/LU.cc, rc7/Timer.cc, rc7/BlockedArray.h] Executing post-compiler mpicxx -g -I/nfs/07/dhudak/devel/x10/20101122r1/x10-trunk/x10.dist/include -Idbg-gen -I. -Wno-long-long -Wno-unused-parameter -pthread -msse2 -mfpmath=sse -o /nfs/07/dhudak/devel/x10/20101124r1/benchmarks/hpcc/LU/lu-dbg rc7/BlockedArray__ArrayView.cc rc7/BlockedArray.cc rc7/BlockedArray__Block.cc rc7/essl_natives.cc rc7/LU.cc rc7/Timer.cc -L/nfs/07/dhudak/devel/x10/20101122r1/x10-trunk/x10.dist/lib -lx10 -DX10_USE_BDWGC -lgc -lx10rt_mpi -ldl -lm -lpthread -Wl,--rpath -Wl,/nfs/07/dhudak/devel/x10/20101122r1/x10-trunk/x10.dist/lib -Wl,-export-dynamic -lrt -L/usr/local/goto-blas-1.15-gfortran/lib -lgoto make[1]: Leaving directory `/nfs/07/dhudak/devel/x10/20101124r1/benchmarks/hpcc/LU' Then, I qsubbed a job and got the following output: dhu...@opt-login01 1034%> cat LU-x10.o4468726 LU Starting: M 128 B 64 px 2 py 2 Main loop starting ... =>> PBS: job killed: walltime 645 exceeded limit 600 ...any suggestions on how to proceed? FT is hitting a syntax error, but the code in question looks right. dhu...@opt-login01 1017%> make FT make -C FT fft-opt make[1]: Entering directory `/nfs/07/dhudak/devel/x10/20101124r1/benchmarks/hpcc/FT' mkdir -p opt-gen/FT x10c++ -O -NO_CHECKS -d opt-gen -report postcompile=5 -x10rt mpi fft.x10 Random.x10 -post '# -O3 # # ' -o fft-opt /nfs/07/dhudak/devel/x10/20101124r1/benchmarks/hpcc/FT/fft.x10:3: Syntax error: Token "*" expected instead of this input /nfs/07/dhudak/devel/x10/20101124r1/benchmarks/hpcc/FT/fft.x10:3-15: Syntax error: Misplaced constructs(s): "ERROR_TOKEN" /nfs/07/dhudak/devel/x10/20101124r1/benchmarks/hpcc/FT/fft.x10:23: Syntax error: Token(s) inserted to complete scope: "= VariableInitializer" /nfs/07/dhudak/devel/x10/20101124r1/benchmarks/hpcc/FT/fft.x10:23: Syntax error: Token(s) inserted to complete scope: ";" /nfs/07/dhudak/devel/x10/20101124r1/benchmarks/hpcc/FT/fft.x10:123: Syntax error: Token "=>" expected instead of this input 5 errors. make[1]: *** [fft-opt] Error 1 make[1]: Leaving directory `/nfs/07/dhudak/devel/x10/20101124r1/benchmarks/hpcc/FT' make: *** [FT-opt] Error 2 STREAM and RA built without complaint, but I haven't gotten the chance to run them yet. Thanks, Dave --- David E. Hudak, Ph.D. dhu...@osc.edu Program Director, HPC Engineering Ohio Supercomputer Center http://www.osc.edu ------------------------------------------------------------------------------ Increase Visibility of Your 3D Game App & Earn a Chance To Win $500! Tap into the largest installed PC base & get more eyes on your game by optimizing for Intel(R) Graphics Technology. Get started today with the Intel(R) Software Partner Program. Five $500 cash prizes are up for grabs. http://p.sf.net/sfu/intelisp-dev2dev _______________________________________________ X10-users mailing list X10-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/x10-users
