commit | 7e0775474aa06ecd264c5cec299f73e66c37af9e | [log] [tgz] |
---|---|---|
author | Henrik Hoglind <henrik.hoglind@arm.com> | Fri Aug 21 15:19:11 2020 +0200 |
committer | Henrik Hoglind <henrik.hoglind@arm.com> | Tue Aug 25 10:22:33 2020 +0200 |
tree | c1b14ccf6cd3162833a37f55c4caac29d8b209df | |
parent | 68013163b2633114de71c76fd2b1c5c9fc13571c [diff] |
Add options to Tensorflow for compiler optimization level Change-Id: I2c9b685b69dc3068b92d07a1c7840b0906954582
The core software is built with CMake. It is recommended to build out of tree like illustrated below.
$ mkdir build $ cd build $ cmake .. -DCMAKE_TOOLCHAIN_FILE=<toolchain> -DCMAKE_SYSTEM_PROCESSOR=cortex-m<nr><features> $ make
Available build options can be listed with cmake -LH ..
.
Supported CPU targets are any of the Cortex-M processors with any of the supported features, for example cortex-m33+nodsp+nofp. A toolchain file is required to cross compile the software.