[profiling] pre-processing using rocprofiler-sdk profiler
[profiling] perform SoC profiling setup for gfx942
Resolved library (exact match): /rocm-venv/lib/python3.12/site-packages/_rocm_sdk_devel/lib/rocprofiler-sdk/librocprofv3-list-avail.so
Collecting following counters: SQ_ACTIVE_INST_FLAT, GRBM_GUI_ACTIVE, SQ_ACTIVE_INST_VALU, SQ_ACTIVE_INST_SCA, SQ_WAVES, SQ_ACTIVE_INST_VMEM, SQ_ACTIVE_INST_MISC
[profiling] perfmon_coalesce file_count 1
starting "run_profiling" and about to start rocprof's workload
[profiling] performing profiling using rocprofiler-sdk profiler
Rocprofiler-Compute version: 3.7.0
Profiler choice: rocprofiler-sdk
Output directory: /app/projects/rocprofiler-compute/workloads/torch_trace/MI300X_A1
Target: MI300X_A1
Command: python /app/projects/rocprofiler-compute/src/utils/inject_roctx/launch.py --frameworks torch -- sample/simple_net.py
Kernel Selection: None
Dispatch Selection: None
Filtered sections: ['11.2.2', '11.2.3', '11.2.4', '11.2.5']

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Collecting Performance Counters
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Searching for native collector.
ROCm Compute root directory: /app/projects/rocprofiler-compute/src
Searching /app/projects by lib*/rocprofiler-compute/librocprofiler-compute-tool.so for native collector
Generating native tool project using command: cmake -S /app/projects/rocprofiler-compute/src/lib -B /app/projects/rocprofiler-compute/src/lib/_build
-- HIP: Using hipcc from relative path: /rocm/bin/hipcc
-- {fmt} version: 12.1.0
-- Build type:
-- Using the multi-header code from /app/projects/rocprofiler-compute/src/lib/external/json/include/
-- Configuring done
-- Generating done
-- Build files have been written to: /app/projects/rocprofiler-compute/src/lib/_build
Building native tool using command: cmake --build /app/projects/rocprofiler-compute/src/lib/_build --parallel
[  0%] Built target gsl_assert
[  0%] Built target synchronized
Consolidate compiler generated dependencies of target fmt
Consolidate compiler generated dependencies of target pc-sampling-collector
[ 20%] Built target fmt
[ 46%] Built target pc-sampling-collector
Consolidate compiler generated dependencies of target rocprofiler-compute-tool
[100%] Built target rocprofiler-compute-tool
Searching /app/projects/rocprofiler-compute/src by lib/_build/lib/librocprofiler-compute-tool.so for native collector
Using native collector: /app/projects/rocprofiler-compute/src/lib/_build/lib/librocprofiler-compute-tool.so
Using native counter collection tool: /app/projects/rocprofiler-compute/src/lib/_build/lib/librocprofiler-compute-tool.so
Parallel runtime detected: None='None', None=None
[profiling] Iteration multiplexing: Disabled
[Run 1/1][Approximate profiling time left: pending first measurement...]
[profiling] Current input file: /app/projects/rocprofiler-compute/workloads/torch_trace/MI300X_A1/perfmon/pmc_perf_0.yaml
Sending profiler options to run_prof: {'ROCPROF_COUNTER_COLLECTION': '0', 'LD_PRELOAD': '/rocm-venv/lib/python3.12/site-packages/_rocm_sdk_devel/lib/rocprofiler-sdk/librocprofiler-sdk-tool.so:/app/projects/rocprofiler-compute/src/lib/_build/lib/librocprofiler-compute-tool.so', 'ROCPROF_KERNEL_TRACE': '1', 'ROCPROF_OUTPUT_FORMAT': 'rocpd', 'ROCPROF_OUTPUT_PATH': '/app/projects/rocprofiler-compute/workloads/torch_trace/MI300X_A1/out/pmc_1', 'ROCPROF_MARKER_API_TRACE': '1', 'APP_CMD': ['python', '/app/projects/rocprofiler-compute/src/utils/inject_roctx/launch.py', '--frameworks', 'torch', '--', 'sample/simple_net.py']}
pmc file: pmc_perf_0.yaml
Adding env var for counter definitions: ROCPROFILER_METRICS_PATH=/tmp/rocprof_compute_sdk_config_k7itk34_
rocprof sdk env vars: {'ROCPROFILER_METRICS_PATH': '/tmp/rocprof_compute_sdk_config_k7itk34_', 'ROCPROF_COUNTER_COLLECTION': '0', 'LD_PRELOAD': '/rocm-venv/lib/python3.12/site-packages/_rocm_sdk_devel/lib/rocprofiler-sdk/librocprofiler-sdk-tool.so:/app/projects/rocprofiler-compute/src/lib/_build/lib/librocprofiler-compute-tool.so', 'ROCPROF_KERNEL_TRACE': '1', 'ROCPROF_OUTPUT_FORMAT': 'rocpd', 'ROCPROF_OUTPUT_PATH': '/app/projects/rocprofiler-compute/workloads/torch_trace/MI300X_A1/out/pmc_1', 'ROCPROF_MARKER_API_TRACE': '1', 'ROCPROF_COUNTERS': 'pmc: SQ_ACTIVE_INST_FLAT SQ_ACTIVE_INST_MISC SQ_ACTIVE_INST_SCA SQ_ACTIVE_INST_VALU SQ_ACTIVE_INST_VMEM SQ_WAVES GRBM_GUI_ACTIVE', 'ROCPROF_AGENT_INDEX': 'absolute'}
rocprof sdk user provided command: ['python', '/app/projects/rocprofiler-compute/src/utils/inject_roctx/launch.py', '--frameworks', 'torch', '--', 'sample/simple_net.py']
   |-> [rocprofiler-sdk] WARNING: All log messages before absl::InitializeLog() is called are written to STDERR
   |-> [rocprofiler-sdk] I0000 00:00:1781922426.490045   38090 tool.cpp:3972] [PPID=37772][PID=38090][TID=38090][rocprofv3] rocprofv3_main(6, 'python /app/projects/rocprofiler-compute/src/utils/inject_roctx/launch.py --frameworks torch -- sample/simple_net.py', ...)
   |-> [rocprofiler-sdk] [rocprofiler-compute] [rocprofiler_configure] (priority=1) is using rocprofiler-sdk v1.3.2 (1.3.2)
   |-> [rocprofiler-sdk] W0620 02:27:06.771724   38090 simple_timer.cpp:55] [rocprofv3] tool initialization ::     0.246891 sec
   |-> [rocprofiler-sdk] [rocprofiler-compute] In tool init
   |-> [rocprofiler-sdk] W0620 02:27:06.821557   38090 registration.cpp:1085] Failed to invoke rocprofiler-register propagation. This is normal if runtimes have not initialized yet, or if rocprofiler-register is not loaded. Runtimes that initialize after this call will be automatically profiled.
   |-> [rocprofiler-sdk] W0620 02:27:06.821676   38090 simple_timer.cpp:55] [rocprofv3] 'python /app/projects/rocprofiler-compute/src/utils/inject_roctx/launch.py --frameworks torch -- sample/simple_net.py' ::     0.000000 sec
   |-> [rocprofiler-sdk] W0620 02:27:32.937999   38090 tool.cpp:2945] HSA version 1.21.0 initialized (instance=0)
   |-> [rocprofiler-sdk] W0620 02:27:32.954383   38090 tool.cpp:2945] MARKER (ROCTx) version 1.3.2 initialized (instance=0)
   |-> [rocprofiler-sdk] Training completed
   |-> [rocprofiler-sdk] W0620 02:27:41.590164   38090 simple_timer.cpp:55] [rocprofv3] 'python /app/projects/rocprofiler-compute/src/utils/inject_roctx/launch.py --frameworks torch -- sample/simple_net.py' ::    34.768487 sec
   |-> [rocprofiler-sdk] W0620 02:27:41.605408   38090 generateRocpd.cpp:1054] writing SQL database for process 38090 on node 3166496885
   |-> [rocprofiler-sdk] E0620 02:27:41.606126   38090 generateRocpd.cpp:1078] Opened result file: /app/projects/rocprofiler-compute/workloads/torch_trace/MI300X_A1/out/pmc_1/b84e5a1bd2f0/38090_results.db (UUID=000233e4-d7b1-77b1-bc57-4b5e8997857f)
   |-> [rocprofiler-sdk] W0620 02:27:42.026260   38090 simple_timer.cpp:55] SQLite3 generation :: rocpd_string             ::     0.352958 sec
   |-> [rocprofiler-sdk] W0620 02:27:42.026295   38090 simple_timer.cpp:55] SQLite3 generation :: rocpd_info_node          ::     0.000005 sec
   |-> [rocprofiler-sdk] W0620 02:27:42.026640   38090 simple_timer.cpp:55] SQLite3 generation :: rocpd_info_process       ::     0.000342 sec
   |-> [rocprofiler-sdk] W0620 02:27:42.031465   38090 simple_timer.cpp:55] SQLite3 generation :: rocpd_info_agent         ::     0.004817 sec
   |-> [rocprofiler-sdk] W0620 02:27:42.074856   38090 simple_timer.cpp:55] SQLite3 generation :: rocpd_info_pmc           ::     0.043385 sec
   |-> [rocprofiler-sdk] W0620 02:27:42.384408   38090 simple_timer.cpp:55] SQLite3 generation :: rocpd kernel info        ::     0.309537 sec
   |-> [rocprofiler-sdk] W0620 02:27:42.387446   38090 simple_timer.cpp:55] SQLite3 generation :: rocpd_region             ::     0.002865 sec
   |-> [rocprofiler-sdk] W0620 02:27:42.396494   38090 simple_timer.cpp:55] SQLite3 generation :: rocpd_kernel_dispatch    ::     0.009036 sec
   |-> [rocprofiler-sdk] W0620 02:27:42.396514   38090 simple_timer.cpp:55] SQLite3 generation :: rocpd_pmc_event          ::     0.000006 sec
   |-> [rocprofiler-sdk] W0620 02:27:42.396519   38090 simple_timer.cpp:55] SQLite3 generation :: rocpd_memory_copy        ::     0.000002 sec
   |-> [rocprofiler-sdk] W0620 02:27:42.396525   38090 simple_timer.cpp:55] SQLite3 generation :: rocpd_memory_allocate    ::     0.000004 sec
   |-> [rocprofiler-sdk] W0620 02:27:42.414889   38090 simple_timer.cpp:55] SQLite3 generation :: rocpd_info_pmc: kfd      ::     0.018361 sec
   |-> [rocprofiler-sdk] W0620 02:27:42.414903   38090 simple_timer.cpp:55] SQLite3 generation :: rocpd_pmc_event: kfd     ::     0.000000 sec
   |-> [rocprofiler-sdk] W0620 02:27:42.415557   38090 simple_timer.cpp:55] SQLite3 generation :: rocpd_flush_pending      ::     0.000642 sec
   |-> [rocprofiler-sdk] W0620 02:27:42.415827   38090 simple_timer.cpp:55] SQLite3 generation :: SQL indexing             ::     0.000260 sec
   |-> [rocprofiler-sdk] W0620 02:27:42.424101   38090 simple_timer.cpp:55] SQLite3 generation :: total                    ::     0.818695 sec
   |-> [rocprofiler-sdk] W0620 02:27:42.426475   38090 simple_timer.cpp:55] [rocprofv3] output generation ::     0.835099 sec
   |-> [rocprofiler-sdk] W0620 02:27:42.429300   38090 simple_timer.cpp:55] [rocprofv3] tool finalization ::     0.838820 sec
   |-> [rocprofiler-sdk] [rocprofiler-compute] In tool fini
   |-> [rocprofiler-sdk] [rocprofiler-compute] [write_counters] Counter collection data has been written to: /app/projects/rocprofiler-compute/workloads/torch_trace/MI300X_A1/out/pmc_1/38090_native_counter_collection.csv
Finishing subprocess of pmc file(s), the time taken is 0 m 37.772451400756836 sec
Updated rocpd db /app/projects/rocprofiler-compute/workloads/torch_trace/MI300X_A1/out/pmc_1/b84e5a1bd2f0/38090_results.db with native tool counters.
Intermediate results_*.csv generation from rocpd databases is deprecated and will be replaced with automatic .db file retention in a future release.
[ml api trace] Moved counter collection and marker trace files to workload dir for ML API trace creation.
[Counter Collection: ] /app/projects/rocprofiler-compute/workloads/torch_trace/MI300X_A1/ml_api_trace_pmc_perf_0_counter_collection.csv
[Marker API Trace: ] /app/projects/rocprofiler-compute/workloads/torch_trace/MI300X_A1/ml_api_trace_pmc_perf_0_marker_api_trace.csv
The time of run_prof of /app/projects/rocprofiler-compute/workloads/torch_trace/MI300X_A1/perfmon/pmc_perf_0.yaml is 0 min 39.89185667037964 sec
PC sampling data collection skipped as --pc-sampling is not specified.
finished "run_profiling" and finished rocprof's workload, time taken was 0 m 42.85194277763367 sec
[profiling] performing post-processing using rocprofiler-sdk profiler
[profiling] perform SoC post processing for gfx942
[roofline] Skipping roofline
time taken for "post_processing" was 0 seconds
