add more registration step description

Author: jzhoulon

rfcs/20210513-pluggable-profiler-for-tensorflow.md

@@ -69,9 +69,10 @@ The table below summarizes all structures defined and the functionality they inv
 #### Registration
 Core TensorFlow will register a new ProfilerInterface with [ProfilerFactory](https://github.com/tensorflow/tensorflow/blob/master/tensorflow/core/profiler/lib/profiler_factory.h#L29).
 1. Core TensorFlow loads the function `TF_InitProfiler` from the plugin's dynamic library installed under "…python_dir.../site-packages/tensorflow-plugins".
-2. Core TensorFlow populates `TF_ProfilerRegistrationParams` and passes it in a call to `TF_InitProfiler`. Core TensorFlow owns the memory of `TF_ProfilerRegistrationParams`'s `profiler` and `profiler_fns`.
+2. Core TensorFlow populates `TF_ProfilerRegistrationParams` and passes it in a call to `TF_InitProfiler`. Core TensorFlow owns the memory of `TF_ProfilerRegistrationParams`'s `profiler` and `profiler_fns` struct.
     * In `TF_InitProfiler`, the plugin populates `TF_ProfilerRegistrationParams`'s `profiler` and `profiler_fns`.
-3. Core Tensorflow will create a `PluggableProfiler` during [ProfilerSession](https://github.com/tensorflow/tensorflow/blob/master/tensorflow/core/profiler/lib/profiler_session.cc#L109) setup.
+3. Core TensorFlow registers a profiler creation function to ProfilerFactory based on `profiler` and `profiler_fns` populated by the plugin during the initialization time.
+4. Core Tensorflow will create a `PluggableProfiler` during [ProfilerSession](https://github.com/tensorflow/tensorflow/blob/master/tensorflow/core/profiler/lib/profiler_session.cc#L109) setup.
 
 #### Protobuf class
 Profiler uses `XSpace` to store the performance data collected by backends. With this type of data structure, TensorFlow's profiler tools can generate various views of performance, such as timeline, memory consumption, performance of every TensorFlow op and set of summaries. `XSpace` is C++ object generated by protobuf toolchain with a predefined structure in [xplane.proto](https://github.com/tensorflow/tensorflow/blob/master/tensorflow/core/profiler/protobuf/xplane.proto#L9).