20-aiecompiler-features

AI Engine Development See Vitis™ Development Environment on xilinx.com See Vitis™ AI Development Environment on xilinx.com

AI Engine Compiler Features

Version: Vitis 2024.2

Introduction

This tutorial targets AIE-ML device, but shares test cases and description files with AIE 20-aiecompiler-features.

IMPORTANT: Before beginning the tutorial make sure you have installed the AMD Vitis™ 2024.2 software. The Vitis release includes all the embedded base platforms including the VEK280 base platform that is used in this tutorial.

Set up your PLATFORM_REPO_PATHS environment variable to $XILINX_VITIS/base_platforms.

This tutorial targets VEK280 board for 2024.2/ version.

Objectives

After completing this tutorial, you will be able to:

• Work with multirate design that allow the compiler to handle frame length mismatch between consecutive kernels in a kernel chain.
• Send output data to different other kernels for stream-based and buffer-based I/O.
• Conditionally instantiate graph objects.

This tutorial is based on simple data passthrough to avoid another level of complexity.

Tutorial Sections

This tutorial contains 3 different sections that are independant one from each other.

Follow instructions in AIE Compiler Features. But, run all make commands with ARCH=aie-ml.

Conditional Objects

The following examples illustrate various ways to generate conditional objects at the graph level. All examples are available in different source directories: src1 to src4. They can be compiled using: make CASE=N clean data aie aieviz with N= 1,2,3,4. Type make help to get the list of all available examples.

All these examples can be compiled for the AI Engine ML architecture by including ARCH=aie-ml in the make command.

Case 1

To get the visualization type:

make ARCH=aie-ml CASE=1 clean data aie aieviz

Case 2

To get the visualization type:

make ARCH=aie-ml CASE=2 aie aieviz

Case 3

To get the visualization type:

make ARCH=aie-ml CASE=3 aie aieviz

Case 4

To get the visualization type:

make ARCH=aie-ml CASE=4 aie aieviz

Multirate

UpConv then DownConv (Buffer)

make VERSION=1 ARCH=aie-ml buffer aie

You can run the simulation and visualize the result:

make aiesim aieviz

DownConv then UpConv (Buffer)

In this design the first kernel in the dataflow is the down-converter 5/7 followed by the up-converter.

make ARCH=aie-ml VERSION=2 clean buffer aie aiesim aieviz

Split and Merge (Buffer)

In this new example, two branches are fed with the same PLIO and connected on the output side to a kernel that will compute the difference between the two branch outputs.

make ARCH=aie-ml VERSION=3 clean buffer aie

Let run the simulation and visualize:

make aiesim aieviz

UpConv then DownConv (Stream)

make ARCH=aie-ml VERSION=1 clean stream aie aiesim aieviz

In this configuration, the repetition rates are five (UpConv) and seven (DownConv). The simulation duration is slightly less 40 µs.

DownConv then UpConv (Stream)

make ARCH=aie-ml VERSION=2 clean stream aie aiesim aieviz

In this configuration, the repetition rates are seven (DownConv) and five (UpConv). The simulation duration is slightly more than 40 µs.

Split and Merge (stream)

In this new example, two branches are fed with the same PLIO and connected on the output side to a kernel that will compute the difference between the two branch outputs.

VERSION=3 of this design stalls almost immediately because this design needs FIFOs set up at the input and output of each branch.

VERSION=4 sets these FIFOs and the overall simulation lasts approximately 275 µs.

Multicast

Case 1: Stream and Buffer Multicasting

To explore the graph view and the array view shown by Vitis Analyzer:

make ARCH=aie-ml CASE=1 clean data aie aieviz

Case 2: Multirate Buffer Multicasting

Explore the graph view and the array view shown by the AMD Vitis™ Analyzer:

make ARCH=aie-ml CASE=2 clean data aie aieviz

_{Copyright © 2020–2024 Advanced Micro Devices, Inc}

^{Terms and Conditions}