Late Breaking Results: PyAIE: A Python-based Programming Framework for Versal ACAP Platforms

In order to satisfy the exploding demand from various computation-intensive applications, AMD-Xilinx released Versal, a heterogeneous Adaptive Compute Acceleration Platform (ACAP) with powerful heterogeneous acceleration capabilities. Most AI/ML applications use Python as their primary programming l...

Full description

Saved in:
Bibliographic Details
Published in:2023 60th ACM/IEEE Design Automation Conference (DAC) pp. 1 - 2
Main Authors: Tian, Hongzheng, Yang, Shining, Cha, Yoonha, Huang, Sitao
Format: Conference Proceeding
Language:English
Published: IEEE 09.07.2023
Subjects:
Artificial intelligence
Codes
Computer architecture
Engines
Kernel
Programming
Python
System analysis and design
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In order to satisfy the exploding demand from various computation-intensive applications, AMD-Xilinx released Versal, a heterogeneous Adaptive Compute Acceleration Platform (ACAP) with powerful heterogeneous acceleration capabilities. Most AI/ML applications use Python as their primary programming language these days; yet, AI engine codes as well as numerous Versal ACAP-based acceleration projects must be written in C/C++ with low-level intrinsics in order to take advantage of the Versal ACAP platform. To fill the gap of programming abstractions of applications and Versal ACAP platforms, we propose PyAIE, a Python-based programming framework specifically targeting Versal ACAP architecture. PyAIE allows users to focus on algorithm-level designs without knowledge of the underlying low-level details. PyAIE automatically translates Python code into the optimized C/C++ code for AI Engine kernels, PL kernels, and host, along with configuration script files, thereby completing the entire AI Engine-based system design. To the best of our knowledge, this is the first Python-based programming and compilation flow designed specifically for Versal ACAP platforms.
DOI:10.1109/DAC56929.2023.10247843