Buffets: An Efficient and Composable Storage Idiom for Explicit Decoupled Data Orchestration
Buffets: An Efficient and Composable Storage Idiom for Explicit Decoupled Data Orchestration
复制标题
DOI：
10.1145/3297858.3304025
复制
发表时间：
2019-04
期刊：
影响因子：
--
通讯作者：
Michael Pellauer;Y. Shao;Jason Clemons;N. Crago;Kartik Hegde;Rangharajan Venkatesan;S. Keckler;Christopher W. Fletcher;J. Emer
中科院分区：
文献类型：
--
作者：
Michael Pellauer;Y. Shao;Jason Clemons;N. Crago;Kartik Hegde;Rangharajan Venkatesan;S. Keckler;Christopher W. Fletcher;J. Emer
加速器在定制片内缓冲上花费大量的面积和精力。不幸的是，这些解决方案与特定的设计紧密相连，阻碍了其他加速器或领域的可重用性。我们提出自助餐，一个有效的和可组合的存储习惯用法的需要，是独立于任何特定的设计的加速器。Buffets有几个显著的特征，包括高效的解耦填充和访问，以及细粒度的同步、分层组合和高效的多播。我们在RTL中实现了缓冲区，并表明它们在8KB RAM上只增加了2%的控制开销。与DMA管理的双缓冲暂存器和缓存相比，在一系列工作负载中，缓冲区分别将能量延迟产品提高了1.53倍和5.39倍。
Accelerators spend significant area and effort on custom on-chip buffering. Unfortunately, these solutions are strongly tied to particular designs, hampering re-usability across other accelerators or domains. We present buffets, an efficient and composable storage idiom for the needs of accelerators that is independent of any particular design. Buffets have several distinguishing characteristics, including efficient decoupled fills and accesses with fine-grained synchronization, hierarchical composition, and efficient multi-casting. We implement buffets in RTL and show that they only add 2% control overhead over an 8KB RAM. When compared with DMA-managed double-buffered scratchpads and caches across a range of workloads, buffets improve energy-delay-product by 1.53x and 5.39x, respectively.