Files

T

Kandula, Venkateshwar reddy e0901eba28 [Samples] Remove thread trace sample dependency on rocprofiler-sdk-amd-comgr. (#555 )

* remove samples dependency on rocprofiler-sdk-amd-comgr.

* add find package for amd_comgr.

---------

Co-authored-by: Venkateshwar Reddy Kandula <vkandula@amd.com>

2025-07-31 10:24:32 -05:00

agent.cpp

Thread trace and Trace Decoder API tests and samples (#416 )

2025-07-22 20:08:12 -05:00

CMakeLists.txt

[Samples] Remove thread trace sample dependency on rocprofiler-sdk-amd-comgr. (#555 )

2025-07-31 10:24:32 -05:00

main.cpp

Thread trace and Trace Decoder API tests and samples (#416 )

2025-07-22 20:08:12 -05:00

README.md

Thread trace and Trace Decoder API tests and samples (#416 )

2025-07-22 20:08:12 -05:00

README.md

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 roctxProfilerResume is called to start thread trace
Stops tracing at roctxProfilerPause
Receives the trace data in shader_data_callback and calls rocprofiler_trace_decode to decode the data
rocprofiler_trace_decode calls parse (a lambda)
parse receives the dedecoded data and increments hitcount/latencies by pc address
At application end, tool_fini calls gen_output_stream to write the top hotspots into thread_trace.log

main.cpp:

Defines a few different kernels and runs them
The first loop iteration warms up the kernels
The second iteration calls roctxProfilerResume to start thread trace
After the loop ends, roctxProfilerPause is called to stop tracing