logo
ResearchBunny Logo
End-to-end programmable computing systems

Computer Science

End-to-end programmable computing systems

Y. Xiao, G. Ma, et al.

Discover a groundbreaking framework, PGL, that revolutionizes the management of algorithm complexity in autonomous systems, developed by Yao Xiao and team. With remarkable speedups achieved through advanced program representation learning, this research unveils a future where code execution is optimized across diverse hardware seamlessly.

00:00
00:00
Playback language: English
Abstract
To manage the increasing complexity of algorithms in applications like autonomous systems, a unified, end-to-end, programmable graph representation learning (PGL) framework is proposed. PGL mines program complexity from high-level programs down to low-level virtual machine intermediate representation, extracts computational patterns, and predicts optimal code segment execution on heterogeneous hardware. Using multifractal features from code graphs and graph representation learning, PGL achieves automatic parallelization and processor assignment. Evaluations show significant speedups (6.42x and 2.02x) compared to thread-based execution and state-of-the-art techniques, respectively.
Publisher
Communications Engineering
Published On
Nov 24, 2023
Authors
Yao Xiao, Guixiang Ma, Nesreen K. Ahmed, Mihai Capotă, Theodore L. Willke, Shahin Nazarian, Paul Bogdan
Tags
graph representation learning
algorithm complexity
autonomous systems
parallelization
code optimization
multifractal features
heterogeneous hardware

Related Publications

Explore these studies to deepen your understanding of the subject.

End-to-end differentiability and tensor processing unit computing to accelerate materials’ inverse design
Engineering and Technology

End-to-end differentiability and tensor processing unit computing to accelerate materials’ inverse design

H. Liu, Y. Liu, et al.

An end-to-end deep learning framework for translating mass spectra to de-novo molecules
Chemistry

An end-to-end deep learning framework for translating mass spectra to de-novo molecules

E. E. Litsa, V. Chenthamarakshan, et al.

The conterminous United States are projected to become more prone to flash floods in a high-end emissions scenario
Earth Sciences

The conterminous United States are projected to become more prone to flash floods in a high-end emissions scenario

Z. Li, S. Gao, et al.

A framework to predict the price of energy for the end-users with applications to monetary and energy policies
Economics

A framework to predict the price of energy for the end-users with applications to monetary and energy policies

S. G. Baratsas, A. M. Niziolek, et al.

Listen, Learn & Level Up
Over 10,000 hours of research content in 25+ fields, available in 12+ languages.
No more digging through PDFs, just hit play and absorb the world's latest research in your language, on your time.
listen to research audio papers with researchbunny