MCUXpresso_MIMXRT1052xxxxB/middleware/eiq/glow
Yilin Sun 6baf4427ce
Updated to v2.15.000
Signed-off-by: Yilin Sun <imi415@imi.moe>
2024-03-18 23:15:10 +08:00
..
bundle_utils Updated to v2.15.000 2024-03-18 23:15:10 +08:00
examples/common Updated to v2.15.000 2024-03-18 23:15:10 +08:00
LICENSE.txt Updated to v2.15.000 2024-03-18 23:15:10 +08:00
readme.txt Updated to v2.15.000 2024-03-18 23:15:10 +08:00

readme.txt

eIQ Glow Utils 1.1.0

Content
-------

1. Introduction
2. Directory structure
3. Project examples
4. Requirements
5. Documentation

1. Introduction
---------------
Glow is a Machine Learning compiler for neural network graphs. It is designed
to optimize the neural network graphs and generate code for various hardware
devices (available at https://github.com/pytorch/glow). Glow comes in two flavors:
Just in Time (JIT) compilation, where compilation is performed at runtime just
before the model is executed and Ahead of Time (AOT) compilation, where compilation
is performed offline and generates an object file (bundle) which is later linked
in the application code. The Glow compiler is a separate tool which is installed
with a dedicated Windows Installer.
The Glow Utils package from the MCUXpresso SDK provides a set of helper functions
that allows the integration of the Glow AOT bundles generated by the Glow compiler.

2. Directory structure
----------------------
<MCUXpresso-SDK-root>
|-- boards
|   -- <board>
|      -- eiq_examples        - eIQ project examples
|         -- glow_lenet_mnist - Lenet MNIST image recognition project example
|         -- glow_cifar10     - CIFAR-10 image recognition project example
|-- middleware
    -- eiq
       -- glow
          -- bundle_utils     - Utility functions for integrating Glow AOT bundles
          -- examples         - Utility scripts for integrating Glot AOT bundles

3. Project examples
-------------------
The package contains several example applications by integrating bundles
generated by the Glow compiler. The example projects are found in:
- <MCUXpresso-SDK-root>/<board>/eiq_examples/glow_lenet_mnist
- <MCUXpresso-SDK-root>/<board>/eiq_examples/glow_cifar10
The following <board> configurations are currently supported:
- evkbimxrt1050
- evkmimxrt1060
- evkbmimxrt1060
- evkmimxrt1160
- evkmimxrt1170
The following toolchains are currently supported:
- MCUXpresso IDE
- IAR Embedded Workbench for ARM
- Keil uVision MDK
- ArmGCC - GNU Tools ARM Embedded

4. Requirements
---------------
In order to run the examples and use the Glow AOT capabilities provided in the
MCUXpresso SDK one must install separately the Glow compiler. The Glow Windows
Installer reference and user guide can be found in the Glow documentation.

5. Documentation
----------------
5.1. Getting started with Glow AOT:
     <MCUXpresso-SDK-root>/docs/EIQ/eIQ Glow Ahead of Time User Guide.pdf