Mark the execution of BuildCudaEngine with AnnotatedTraceMe to collect the GPU execution time for the routine. Mark the execution of ExecuteNativeSegment, ExecuteCalibration and ComputeAsync with TraceMe to collect the CPU execution time for the routines. We may want to fine tune this later. PiperOrigin-RevId: 347490797 Change-Id: Id6906ea0d433133c5ec5d2b2d234525645bb3c9d |
||
---|---|---|
.. | ||
annotated_traceme.h | ||
BUILD | ||
connected_traceme.h | ||
profiler_factory.cc | ||
profiler_factory.h | ||
profiler_interface.h | ||
profiler_lock.cc | ||
profiler_lock.h | ||
profiler_session.cc | ||
profiler_session.h | ||
scoped_annotation_test.cc | ||
scoped_annotation.h | ||
traceme_encode_test.cc | ||
traceme_encode.h | ||
traceme.h |