Opens a larger view. Escape closes it.

hardware-counters

gromacs_c64_profile.txt

CrayPat/X:  Version 23.09.0 Revision 6034a9414 sles15.4_x86_64  08/01/23 18:55:19

Number of PEs (MPI ranks):   64
                           
Numbers of PEs per Node:     64
                           
Numbers of Threads per PE:    1
                           
Number of Cores per Socket:  64

Execution start time:  Tue Jul 28 08:42:52 2026

System name and speed:  nid002766  2.182 GHz (nominal)

AMD   Rome                 CPU  Family: 23  Model: 49  Stepping:  0

Core Performance Boost:  All 64 PEs have CPB capability

Current path to data file:
  /work/project/project/user/runs/sweep/gromacs_c64//exp_A   (RTS)


Notes for table 1:

  This table shows functions that have significant exclusive time,
    averaged across ranks.
  For further explanation, see the "General table notes" below, or 
    use:  pat_report -v -O profile ...

Table 1:  Profile by Function Group and Function

  Time% |       Time | Imb. Time |  Imb. |    Calls | Group
        |            |           | Time% |          |  Function
        |            |           |       |          |   PE=HIDE
       
 100.0% | 208.351803 |        -- |    -- | 27,213.4 | Total
|--------------------------------------------------------------------
|  77.4% | 161.338056 | 34.117532 | 17.7% |      1.0 | USER
||-------------------------------------------------------------------
||  77.4% | 161.338056 | 34.117532 | 17.7% |      1.0 | main
||===================================================================
|  13.1% |  27.207924 |        -- |    -- |    638.2 | MPI_SYNC
||-------------------------------------------------------------------
||  12.8% |  26.656156 | 25.779186 | 96.7% |     77.5 | MPI_Bcast(sync)
||===================================================================
|   9.5% |  19.804360 |        -- |    -- | 26,574.2 | MPI
||-------------------------------------------------------------------
||   4.7% |   9.823013 |  3.889333 | 28.8% |  4,225.6 | MPI_Waitall
||   2.1% |   4.471408 |  3.330443 | 43.4% |  6,279.7 | MPI_Sendrecv
||   1.3% |   2.729228 |  0.046547 |  1.7% |     24.2 | MPI_Comm_split
|====================================================================

========================  Additional details  ========================

General table notes:

    The default notes for a table are based on the default definition of
    the table, and do not account for the effects of command-line options
    that may modify the content of the table.
    
    Detailed notes, produced by the pat_report -v option, do account for
    all command-line options, and also show how data is aggregated, and
    if the table content is limited by thresholds, rank selections, etc.
    
    An imbalance metric in a line is based on values in main threads
    across multiple ranks, or on values across all threads, as applicable.
    
    An imbalance percent in a line is relative to the maximum value
    for that line across ranks or threads, as applicable.
    
    If the number of Calls for a function is shown as "--", then that
    function was not traced and the other values in its line summarize
    the data collected for functions that it calls and that were traced.
    
Experiment:  trace

Original path to data file:
  /mnt/lustre/a2fs-work3/work/project/project/user/runs/sweep/gromacs_c64/exp_A/xf-files   (RTS)

Original program:
  /mnt/lustre/a2fs-work3/work/project/project/user/builds/gromacs/install/bin/gmx_mpi

Instrumented with:
  pat_build -g mpi -w -o gmx_mpi+pat install/bin/gmx_mpi

Instrumented program:
  /work/project/project/user/builds/gromacs/gmx_mpi+pat

Program invocation:
  /work/project/project/user/builds/gromacs/gmx_mpi+pat mdrun -notunepme -dlb yes -resethway -noconfout -nsteps 512 -s /work/project/project/user/builds/gromacs/benchRIB.tpr -g md_A.log

Exit Status:  0 for 64 PEs

Memory pagesize:  4 KiB

Memory hugepagesize:  Not Available

Programming environment:  CRAY

Runtime environment variables:
  CRAYPAT_COMPILER_OPTIONS=1
  CRAYPAT_GCC_LIB_PATH=/opt/cray/pe/gcc-libs
  CRAYPAT_LD_LIBRARY_PATH=/opt/cray/pe/gcc-libs/opt/cray/pe/perftools/23.09.0/lib64
  CRAYPAT_OPTS_EXECUTABLE=libexec64/opts
  CRAYPAT_ROOT=/opt/cray/pe/perftools/23.09.0
  CRAYPE_VERSION=2.7.23
  CRAY_BINUTILS_VERSION=/opt/cray/pe/cce/16.0.1
  CRAY_CC_VERSION=16.0.1
  CRAY_DSMML_VERSION=0.2.2
  CRAY_FFTW_VERSION=3.3.10.5
  CRAY_FTN_VERSION=16.0.1
  CRAY_LIBSCI_VERSION=23.09.1.1
  CRAY_MPICH_VERSION=8.1.27
  CRAY_PERFTOOLS_VERSION=23.09.0
  FFTW_VERSION=3.3.10.5
  LIBSCI_VERSION=23.09.1.1
  LMOD_FAMILY_COMPILER_VERSION=16.0.1
  LMOD_FAMILY_CRAYPE_CPU_VERSION=false
  LMOD_FAMILY_CRAYPE_NETWORK_VERSION=false
  LMOD_FAMILY_CRAYPE_VERSION=2.7.23
  LMOD_FAMILY_LIBFABRIC_VERSION=1.12.1.2.2.0.0
  LMOD_FAMILY_MPI_VERSION=8.1.27
  LMOD_FAMILY_PERFTOOLS_VERSION=false
  LMOD_FAMILY_PRGENV_VERSION=8.4.0
  LMOD_VERSION=8.7.19
  MPICH_DIR=/opt/cray/pe/mpich/8.1.27/ofi/crayclang/14.0
  OMP_NUM_THREADS=1
  PAT_RT_EXPDIR_BASE=/work/project/project/user/runs/sweep/gromacs_c64
  PAT_RT_EXPDIR_NAME=exp_A
  PAT_RT_EXPDIR_REPLACE=1
  PAT_RT_PERFCTR=PAPI_FP_OPS,PAPI_FP_INS,PAPI_TOT_INS,PAPI_TOT_CYC,cray_rapl:::PACKAGE_ENERGY,cray_rapl:::PP0_ENERGY,cray_zenl3:::UNC_L3_CACHE_MISSES,cray_zenl3:::UNC_L3_CACHE_REQUESTS,cray_zenl3:::UNC_L3_MISS_LATENCY
  PAT_RT_PERFCTR_DISABLE_COMPONENTS=nvml,rocm_smi
  PERFTOOLS_VERSION=23.09.0
  PMI_CONTROL_PORT=30742
  PMI_JOBID=14534151
  PMI_LOCAL_RANK=0
  PMI_LOCAL_SIZE=64
  PMI_RANK=0
  PMI_SHARED_SECRET=CHANGE_ME
  PMI_SIZE=64
  PMI_UNIVERSE_SIZE=64

Report time environment variables:
    CRAYPAT_ROOT=/opt/cray/pe/perftools/23.09.0

Number of MPI control variables collected:  136

  (To see the list, specify: -s mpi_cvar=show)

Report command line options:  -O profile

Operating system:
  Linux 5.14.21-150400.24.228-default #1 SMP PREEMPT_DYNAMIC Tue Jul 7 09:49:36 UTC 2026 (9ba75fc)

Hardware performance counter events:
   PAPI_TOT_INS           Instructions completed
   PAPI_FP_INS            Floating point instructions
   PAPI_TOT_CYC           Total cycles
   PAPI_FP_OPS            Floating point operations
   UNC_L3_CACHE_REQUESTS  Requests to L3 cache
   UNC_L3_MISS_LATENCY    Accumulated L3 miss latency (total cycles for all transactions divided by 16)
   UNC_L3_CACHE_MISSES    L3 cache misses by request type
   PACKAGE_ENERGY         Energy used by chip package
   PP0_ENERGY             Energy used by all cores in package

Estimated minimum instrumentation overhead per call of a traced function,
  which was subtracted from the data shown in this report
  (for raw data, use the option:  -s overhead=include):
    Time  1.817  microsecs

Number of traced functions that were called:  30

  (To see the list, specify:  -s traced_functions=show)