experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Add Documentation for profiling in TFLite Micro.

Browse Source 
PiperOrigin-RevId: 340916590
Change-Id: I481964c8a9d58fe7b180d259e6008e378a3a5c9a
This commit is contained in:
Nat Jeffries 2020-11-05 13:24:49 -08:00 committed by TensorFlower Gardener
parent 72da592385
commit a27578f7a2
2 changed files with 47 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
tensorflow/lite/micro/README.md
View File

@ -1,3 +1,5 @@
<!-- mdformat off(b/169948621#comment2) -->
<!-- <!--
Semi-automated TOC generation with instructions from Semi-automated TOC generation with instructions from
https://github.com/ekalinin/github-markdown-toc#auto-insert-and-update-toc https://github.com/ekalinin/github-markdown-toc#auto-insert-and-update-toc
@ -52,6 +54,7 @@ For developers that are interested in more details of the internals of the
project, we have additional documentation in the [docs](docs/) folder. project, we have additional documentation in the [docs](docs/) folder.
* [Benchmarks](benchmarks/README.md) * [Benchmarks](benchmarks/README.md)
* [Profiling](docs/profiling.md)
* [Memory Management](docs/memory_management.md) * [Memory Management](docs/memory_management.md)
* [New Platform Support](docs/new_platform_support.md) * [New Platform Support](docs/new_platform_support.md)
* [Software Emulation with Renode](docs/renode.md) * [Software Emulation with Renode](docs/renode.md)

44
tensorflow/lite/micro/docs/profiling.md Normal file
View File

@ -0,0 +1,44 @@
<!-- mdformat off(b/169948621#comment2) -->
<!--
Semi-automated TOC generation with instructions from
https://github.com/ekalinin/github-markdown-toc#auto-insert-and-update-toc
-->
<!--ts-->
* [Profiling](#profiling)
* [API](#api)
* [Per-Op Profiling](#per-op-profiling)
* [Subroutine Profiling](#subroutine-profiling)
<!-- Added by: njeff, at: Wed 04 Nov 2020 04:35:07 PM PST -->
<!--te-->
# Profiling
This doc outlines how to use the TFLite Micro profiler to gather information
about per-op invoke duration and to use the profiler to identify bottlenecks
from within operator kernels and other TFLite Micro routines.
## API
The MicroInterpreter class constructor contains and optional profiler argument.
This profiler must be an instance of the tflite::Profiler class, and should
implement the BeginEvent and EndEvent methods. There is a default implementation
in tensorflow/lite/micro/micro_profiler.cc which can be used for most purposes.
## Per-Op Profiling
There is a feature in the MicroInterpreter to enable per-op profiling. To enable
this, provide a MicroProfiler to the MicroInterpreter's constructor then build
with a non-release build to disable the NDEBUG define surrounding the
ScopedOperatorProfile within the MicroInterpreter.
## Subroutine Profiling
In order to further dig into performance of specific routines, the MicroProfiler
can be used directly from the TFLiteContext or a new MicroProfiler can be
created if the TFLiteContext is not available where the profiling needs to
happen. The MicroProfiler's BeginEvent and EndEvent can be called directly, or
wrapped using a [ScopedProfile](../../lite/core/api/profiler.h).