一句话：提出 Cohort，以软件为中心的异构 SoC 加速方案，避免加速器编程抽象泄漏过多硬件细节，提升通用性。
论文指出现有加速器设计过度追求硬件极致优化，硬件与软件不匹配常被留给软件修复。当前加速器编程抽象泄漏硬件细节，程序员需修改数据格式并手动管理内存与一致性，这不仅要求深入硬件知识，也损害了通用性。为此，Cohort 提出面向异构 SoC 的软件导向加速思路，试图以更软件为中心的抽象来编程加速器，减少对底层硬件细节的暴露，从而提升可编程性与通用性。给出的摘要未提供具体实现机制与实验评估。
Philosophically, our approaches to acceleration focus on the extreme. We must optimise accelerators to the maximum, leaving software to fix any hardware-software mismatches. Today’s software abstractions for programming accelerators leak hardware details, requiring changes to data formats and manual memory and coherence management, among other issues. This harms generality and requires deep hardware knowledge to efficiently program accelerators, a state which we consider hardware-oriented.
@inproceedings{paperbot10054,
title = {Cohort: Software-Oriented Acceleration for Heterogeneous SoCs},
author = {Tianrui Wei and Nazerke Turtayeva and Marcelo Orenes-Vera and Omkar Lonkar and Jonathan Balkind},
booktitle = {ASPLOS},
year = {2023},
doi = {10.1145/3582016.3582059}
}