一句话：提出 Mind mappings 以高效搜索算法到加速器的最优映射，避免低效映射带来的乘性效率开销。
现代计算依赖专用化来满足性能与效率需求，而设计专用硬件架构的核心挑战是算法到硬件的映射空间搜索。已有工作表明，低效映射会带来乘性效率开销；同时该搜索空间不仅庞大，而且非凸、非光滑，使先进搜索技术难以适用。因此，先前工作只能依赖专家选择或次优启发式。本文提出 Mind mappings，旨在实现高效的算法-加速器映射空间搜索。摘要未具体说明其技术路线与实验评估，但核心动机是克服上述搜索障碍，避免低效映射造成的效率损失。
Modern day computing increasingly relies on specialization to satiate growing performance and efficiency requirements. A core challenge in designing such specialized hardware architectures is how to perform mapping space search, i.e., search for an optimal mapping from algorithm to hardware. Prior work shows that choosing an inefficient mapping can lead to multiplicative-factor efficiency overheads. Additionally, the search space is not only large but also non-convex and non-smooth, precluding advanced search techniques. As a result, previous works are forced to implement mapping space search using expert choices or sub-optimal search heuristics.
@inproceedings{paperbot6104,
title = {Mind mappings: enabling efficient algorithm-accelerator mapping space search},
author = {Kartik Hegde and Po-An Tsai and Sitao Huang and Vikas Chandra and Angshuman Parashar and Christopher W. Fletcher},
booktitle = {ASPLOS},
year = {2021},
doi = {10.1145/3445814.3446762}
}