Baraldi, Giovanni
4ca156e572
* Adding test and samples to decoder
* Fix sample
* Formatting
* Fix multi test
* Disable sample
* Fix tests
* Format
* Version fix
* Locking the decoder
* Add atomic
* Review comments
* Format
* Adding readme
* merge conflict and adding PCS+ATT test
* Review comments
* Properly disable PCS test
* Update tests/rocprofv3/advanced-thread-trace/CMakeLists.txt
* Adding back env var test
* Name fix
* Preload sample
* Addressing review comments
* Update docs
---------
Co-authored-by: Giovanni Baraldi <gbaraldi@amd.com>
[ROCm/rocprofiler-sdk commit: e898079a13]
1.1 KiB
1.1 KiB
Thread Trace and ROCprof Trace Decoder
Services
- Thread trace in device profiling mode
- ROCprof Trace Decoder decodes the received thread trace data
- Thread trace start/stop using roctx
Properties
agent.cpp:
- Configures thread trace in all GPU agents found with
rocprofiler_configure_device_thread_trace_service - Waits until
roctxProfilerResumeis called to start thread trace - Stops tracing at
roctxProfilerPause - Receives the trace data in
shader_data_callbackand callsrocprofiler_trace_decodeto decode the data rocprofiler_trace_decodecallsparse(a lambda)parsereceives the dedecoded data and increments hitcount/latencies by pc address- At application end,
tool_finicallsgen_output_streamto write the top hotspots intothread_trace.log
main.cpp:
- Defines a few different kernels and runs them
- The first loop iteration warms up the kernels
- The second iteration calls
roctxProfilerResumeto start thread trace - After the loop ends,
roctxProfilerPauseis called to stop tracing