From 6c06d5f989d035b49c488e3575c9678a5686a501 Mon Sep 17 00:00:00 2001
From: danielyou0230 <danielyou0230@gmail.com>
Date: Wed, 26 Aug 2020 11:15:35 -0700
Subject: [PATCH] TFLM: added optimized kernel directory for vexriscv

---
 .../lite/micro/kernels/vexriscv/README.md     | 28 +++++++++++++++++++
 1 file changed, 28 insertions(+)
 create mode 100644 tensorflow/lite/micro/kernels/vexriscv/README.md

diff --git a/tensorflow/lite/micro/kernels/vexriscv/README.md b/tensorflow/lite/micro/kernels/vexriscv/README.md
new file mode 100644
index 00000000000..d135f041404
--- /dev/null
+++ b/tensorflow/lite/micro/kernels/vexriscv/README.md
@@ -0,0 +1,28 @@
+# VexRISC-V
+## Maintainers
+* [danielyou0230](https://github.com/danielyou0230)
+* [tal-x](https://github.com/tcal-x)
+
+## Background
+The optimized kernels for [VexRISC-V](https://github.com/SpinalHDL/VexRiscv)/[Litex](https://github.com/enjoy-digital/litex) are used to run Tensorflow Lite Micro in Zephyr on either
+
+* Digilent Arty board (e.g. Arty A7)
+* [Renode](https://github.com/renode/renode): Open source simulation framework (no hardware required)
+
+To run on Digilent Arty board (FPGA,) you'll also need a soft-CPU gateware for the FPGA, please see [Tensorflow lite demo running in Zephyr on Litex/VexRiscv SoC](https://github.com/antmicro/litex-vexriscv-tensorflow-lite-demo) by Antmicro for more details.
+
+## Info
+To use VexRISC-V optimized kernels instead of reference kernel add `TAGS=vexriscv`
+to the make command. The kernels that doesn't have optimization for a certain micro
+architecture fallback to use TFLM reference kernels.
+
+# Example
+
+To compile the binary file with VexRISC-V optimizations, one can use the following command
+
+```
+make -f tensorflow/lite/micro/tools/make/Makefile \
+TAGS=vexriscv \
+TARGET=zephyr_vexriscv \
+person_detection_int8_bin
+```