# Invocation command line: # /home/brunst/ws-hpc2020/1.0.2/bin/harness/runhpc -c gnu-taurus --reportable -T base --flagsurl=/home/brunst/ws-hpc2020/1.0.2/config/flags/gcc_flags.xml --define model=mpi --ranks=2040 medium # output_root was not used for this run ############################################################################ ###################################################################### # Example configuration file for the GNU Compilers # # Defines: "model" => "mpi", "acc", "omp", "tgt", "tgtgpu" default "mpi" # "label" => ext base label, default "nv" # # MPI-only Command: # runhpc -c Example_gnu --reportable -T base --define model=mpi --ranks=40 small # # OpenACC Command: # runhpc -c Example_gnu --reportable -T base --define model=acc --ranks=4 small # # OpenMP Command: # runhpc -c Example_gnu --reportable -T base --define model=omp --ranks=1 --threads=40 small # # OpenMP Target Offload to Host Command: # runhpc -c Example_gnu --reportable -T base --define model=tgt --ranks=1 --threads=40 small # # OpenMP Target Offload to NVIDIA GPU Command: # runhpc -c Example_gnu --reportable -T base --define model=tgtnv --ranks=4 small # ####################################################################### %ifndef %{label} # IF label is not set use gnu % define label gnu %endif %ifndef %{model} # IF model is not set use mpi % define model mpi %endif teeout = yes makeflags=-j 24 # Tester Information license_num = 37A tester = Technische Universitaet Dresden test_sponsor = Technische Universitaet Dresden prepared_by = Holger Brunst ###################################################### # SUT Section ###################################################### #include: Example_SUT.inc #include: sut-taurus.inc # ----- Begin inclusion of 'sut-taurus.inc' ############################################################################ ###################################################### # Example configuration information for a # system under test (SUT) Section ###################################################### # General SUT info system_vendor = Bull system_name = Taurus: bullx DLC B720 (Intel Xeon E5-2680 v3) interconnect_fs_syslbl = InfiniBand interconnect_fs_hw_switch_fs_model = SX6025 (36), SX6512 (216) hw_avail = Jan-2015 sw_avail = Sep-2020 # Computation node info # [Node_Description: Hardware] node_compute_syslbl = Taurus Compute Node (Haswell) node_compute_order = 1 node_compute_count = 85 node_compute_purpose = compute node_compute_hw_vendor = Bull node_compute_hw_model = bullx DLC B720 node_compute_hw_cpu_name = Intel Xeon E5-2680 v3 node_compute_hw_ncpuorder = 1,2 chips node_compute_hw_nchips = 2 node_compute_hw_ncores = 24 node_compute_hw_ncoresperchip = 12 node_compute_hw_nthreadspercore = 1 node_compute_hw_cpu_char = Intel Turbo Boost Technology disabled node_compute_hw_cpu_mhz = 2500 node_compute_hw_pcache = 32 KB I + 32 KB D on chip per core node_compute_hw_scache = 256 KB I+D on chip per core node_compute_hw_tcache = 30 MB I+D on chip per chip node_compute_hw_ocache = None node_compute_hw_memory = 64 GB (8 x 8 GB 2Rx8 PC4-2133R-10) node_compute_hw_disk = Micron M510 128GB SSD SATA 6 Gb/s node_compute_hw_other = None #[Node_Description: Accelerator] #[Node_Description: Software] node_compute_hw_adapter_fs_model = Mellanox Technologies MT27600 (MCB193A-FCAT) node_compute_hw_adapter_fs_count = 1 node_compute_hw_adapter_fs_slot_type = PCIe 3.0 x16 node_compute_hw_adapter_fs_data_rate = 56 Gb/s node_compute_hw_adapter_fs_ports_used = 1 node_compute_hw_adapter_fs_interconnect = InfiniBand node_compute_hw_adapter_fs_driver = mlx5_core node_compute_hw_adapter_fs_firmware = 10.16.1200 node_compute_sw_os000 = Red Hat Enterprise Linux Server 7.9 (Maipo) node_compute_sw_os001 = 3.10.0-1127.19.1.el7.x86_64 node_compute_sw_localfile = ext4 node_compute_sw_sharedfile = 4 PB Lustre over Infiniband FDR (56 Gb/s) node_compute_sw_state = Multi-user, run level 3 node_compute_sw_other = None #[Interconnect] interconnect_fs_order = 0 interconnect_fs_purpose = MPI Traffic and File System interconnect_fs_hw_vendor = Mellanox Technologies interconnect_fs_hw_model = Mellanox InfiniBand FDR interconnect_fs_hw_topo = FatTree interconnect_fs_hw_switch_fs_count = 17 interconnect_fs_hw_switch_fs_ports = 36 interconnect_fs_hw_switch_fs_data_rate = 56 Gb/s interconnect_fs_hw_switch_fs_firmware = 9.4.2000, 9.4.5070 ####################################################################### # End of SUT section # If this config file were to be applied to several SUTs, edits would # be needed only ABOVE this point. ###################################################################### # ---- End inclusion of '/home/brunst/ws-hpc2020/1.0.2/config/sut-taurus.inc' #[Software] sw_compiler000 = C/C++/Fortran: Version 8.2.0 of sw_compiler001 = GNU Compilers sw_mpi_library = OpenMPI Version 3.1.3 sw_mpi_other = None system_class = Homogenous Cluster sw_other = None #[General notes] notes_000 =This benchmark result is intended to provide perspective on notes_005 =past performance using the historical hardware and/or notes_010 =software described on this result page. notes_015 = notes_020 =The system as described on this result page was formerly notes_025 =generally available. At the time of this publication, it may notes_030 =not be shipping, and/or may not be supported, and/or may fail notes_035 =to meet other tests of General Availability described in the notes_040 =SPEC HPG Policy document, http://www.spec.org/hpg/policy.html notes_045 = notes_050 =This measured result may not be representative of the result notes_055 =that would be measured were this benchmark run with hardware notes_060 =and software available as of the publication date. ####################################################################### # End of SUT section ###################################################################### ###################################################################### # The header section of the config file. Must appear # before any instances of "section markers" (see below) # # ext = how the binaries you generated will be identified # tune = specify "base" or "peak" or "all" %ifndef %{tudprof} label = %{label}_%{model} %else label = %{label}_%{model}_%{tudprof} %endif tune = base output_format = text use_submit_for_speed = 1 # Compiler Settings default: CC = mpicc CXX = mpicxx FC = mpif90 %if %{tudprof} eq 'scorep' CC = scorep --mpp=mpi --instrument-filter=${SPEC}/scorep.filter mpicc CXX = scorep --mpp=mpi --instrument-filter=${SPEC}/scorep.filter mpicxx FC = scorep --mpp=mpi --instrument-filter=${SPEC}/scorep.filter mpif90 %endif # Compiler Version Flags CC_VERSION_OPTION = --version CXX_VERSION_OPTION = --version FC_VERSION_OPTION = --version # enable non-official patches to this kit #strict_rundir_verify = 0 # MPI options and binding environment, dependent upon Model being run # Adjust to match your system # OpenMP (CPU) Settings %if %{model} eq 'omp' preENV_OMP_PROC_BIND=true preENV_OMP_PLACES=cores %endif #OpenMP Targeting Host Settings %if %{model} eq 'tgt' preENV_OMP_PROC_BIND=true preENV_OMP_PLACES=0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,226,27,28,29,30,31,32,33,34,35,36,37,38,39 %endif #MPIRUN_OPTS = --bind-to none -q MPIRUN_OPTS= submit = timeout 2h srun ${MPIRUN_OPTS} -n $ranks -c $threads $command # MPI Workaround for mca issues in sph_exa #preOMPI_MCA_topo=basic # Score-P performance profiling %if %{tudprof} eq 'scorep' ## only record calls to main and MPI functions (runtime filtering) ## runtime filtering was replaced by compile-time filtering (see above) # preENV_SCOREP_FILTERING_FILE=/home/brunst/ws-hpc2020/kit91/scorep.filter ## set buffer memory size for profile/trace preENV_SCOREP_TOTAL_MEMORY=64MB ## enable profile recording preENV_SCOREP_ENABLE_PROFILING=true ## set to 'true' to enable detailed trace file recording preENV_SCOREP_ENABLE_TRACING=false ## collect memory consumption per node preENV_SCOREP_METRIC_RUSAGE=ru_maxrss ## uncomment to record cycle counter for scheduling analysis preENV_SCOREP_METRIC_PAPI=PAPI_TOT_CYC %endif ####################################################################### # Optimization # Note that SPEC baseline rules require that all uses of a given compiler # use the same flags in the same order. See the SPEChpc Run Rules # for more details # http://www.spec.org/hpc2021/Docs/runrules.html # # OPTIMIZE = flags applicable to all compilers # FOPTIMIZE = flags appliable to the Fortran compiler # COPTIMIZE = flags appliable to the C compiler # CXXOPTIMIZE = flags appliable to the C++ compiler # # See your compiler manual for information on the flags available # for your compiler # Compiler flags applied to all models default=base=default: COPTIMIZE = -Ofast -march=native -lm # use -mcpu=native for ARM CXXOPTIMIZE = -Ofast -march=native -std=c++14 FOPTIMIZE = -Ofast -march=native -fno-stack-protector FPORTABILITY = -ffree-line-length-none %if %{model} eq 'mpi' pmodel=MPI %endif # OpenACC flags %if %{model} eq 'acc' pmodel=ACC OPTIMIZE += -fopenacc -foffload=-lm %endif # OpenMP (CPU) flags %if %{model} eq 'omp' pmodel=OMP OPTIMIZE += -fopenmp %endif # OpenMP Targeting host flags %if %{model} eq 'tgt' pmodel=TGT OPTIMIZE += -fopenmp %endif # OpenMP Targeting Nvidia GPU flags %if %{model} eq 'tgtnv' pmodel=TGT OPTIMIZE += -fopenmp -fopenmp-targets=nvptx64-nvidia-cuda %endif # No peak flags set, so make peak use the same flags as base default=peak=default: basepeak=1 ####################################################################### # Portability ####################################################################### 519.clvleaf_t,619.clvleaf_s,719.clvleaf_m,819.clvleaf_l=default=default: # Not needed anymore? #PORTABILITY += -DSPEC_GNU_FLUSH # The following section was added automatically, and contains settings that # did not appear in the original configuration file, but were added to the # raw file after the run. default: flagsurl000 = http://www.spec.org/hpc2021/flags/gcc.xml notes_submit_000 = srun -n $ranks -c $threads $command