|
Message-ID: <6c8f8023b5cd4980b5b6af2d813929ee@TiSMBX01.TiSnetworks.net> Date: Sat, 19 Jul 2014 06:39:22 +0000 From: Pyrex <pyrex@...networks.net> To: "john-users@...ts.openwall.com" <john-users@...ts.openwall.com> Subject: RE: Help Compiling John on ARM w/ MPI & CUDA So I managed to get a few of my cluster systems online. I am going to post some of the data I collected while running a test against a WPA key. Do I need to change the commands I am passing to "mpirun" or do I pretty much have it as optimized as it can be? Also the OpenMP error what is going on in that case and does it concern me at this time? Thank You -Pyrex ----------------------------------------------------------------- mpirun -hostfile mpi.hosts -x LD_LIBRARY_PATH=/usr/local/cuda-6.0/lib:$LD_LIBRARY_PATH -n 5 ~/src/john/run/john -format=wpapsk-cuda ~/wrk/wpa-hash.wrk Loaded 1 password hash (wpapsk-cuda, WPA/WPA2 PSK [PBKDF2-SHA1 CUDA]) MPI in use, disabling OMP (see doc/README.mpi) Warning: OpenMP is disabled; a non-OpenMP build may be faster Node numbers 1-5 of 5 (MPI) Note: minimum length forced to 8 Send SIGUSR1 to mpirun for status mpirun: Forwarding signal 10 to job 3 0g 0:00:01:27 3/3 0g/s 6248p/s 6248c/s 6248C/s berempid..berceyoo 4 0g 0:00:01:26 3/3 0g/s 6262p/s 6262c/s 6262C/s 024572113..091999011 5 0g 0:00:01:26 3/3 0g/s 6286p/s 6286c/s 6286C/s br080662..br048854 1 0g 0:00:01:27 3/3 0g/s 6258p/s 6258c/s 6258C/s chumity5..carrubee 2 0g 0:00:01:26 3/3 0g/s 6231p/s 6231c/s 6231C/s pauggla1..pautiley ________________________________________ From: Pyrex Sent: Friday, July 18, 2014 9:15 AM To: john-users@...ts.openwall.com Subject: RE: [john-users] Help Compiling John on ARM w/ MPI & CUDA When running it utilizing MPI you do not need to specify the device? ex: "--device=0" Also, do I have to utilize OpenMPI or will it work with MPICH2? -Pyrex ________________________________________ From: magnum Sent: Friday, July 18, 2014 2:56 AM To: john-users@...ts.openwall.com Subject: Re: [john-users] Help Compiling John on ARM w/ MPI & CUDA On 2014-07-18 02:54, Pyrex wrote: > From: Lukas Odzioba<mailto:lukas.odzioba@...il.com> >> Just for the record, modify Makefile.in: >> NVCC_FLAGS = -c -Xptxas -v -fmad=true -arch sm_10 >> to >> NVCC_FLAGS = -c -Xptxas -v -fmad=true -arch sm_32 >> >> and then ./configure --disable-mpi && make -s worked for me. The --disable-mpi is a no-op, it defaults to disabled. > I will append that to my notes, how goes the MPI portion of it? Provided MPI libs, headers and runtime are installed in system paths, you'd use --enable-mpi: ./configure --enable-mpi && make -s clean && make -sj8 You'd run it eg. like this (provided all hosts has one CUDA device #0): mpirun -host host1,host2 ./john -format=md5crypt-cuda -test See doc/README.mpi and doc/README-CUDA magnum No malware was found: Scanned by TiSnetworks E-Mail Security Systems. = No malware was found: Scanned by TiSnetworks E-Mail Security Systems.
Powered by blists - more mailing lists
Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.