Introduction

[ Home ]

Raspberry Pi is continuously evolving to provide faster and more power efficient hardware.

Unfortunately, optimizing compilers such as GCC and CLANG may not keep pace with this ever increasing complexity of ever changing computer systems anymore and may produce underperforming executable codes while wasting resources and energy (see our vision paper 1).

We developed Collective Knowledge Framework and Repository (CK) to let the community collaboratively optimize (autotune/crowd-tune) various programs across diverse hardware, continuously share best found solutions in the public repository knowledge, and even use machine learning to extrapolate this information and predict better optimizations or hardware designs (see our vision paper 2).

It is even possible to use Android app to participate in such crowd-tuning of shared applications and kernels across numerous and diverse mobile devices.

We will use this CK repository to share results in a reproducible format with optimizations of various realistic application on RPi using compiler flag crowd-tuning powered by CK.

You may also check related motivation behind Collective Knowledge project.