arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2410.07546cs.AR
收藏

BRAM 是极限:打破迷思、确立标准并构建可扩展的 PIM 加速器

The BRAM is the Limit: Shattering Myths, Shaping Standards, and Building Scalable PIM Accelerators

MD Arafat Kabir, Tendayi Kamucheka, Nathaniel Fredricks, Joel Mandebi, Jason Bakos, Miaoqing Huang, David Andrews

AI总结:

针对现有FPGA基PIM架构时钟低、扩展差等问题,本文确立PIM设计黄金标准,提出IMAGine存内加速GEMV引擎,实现BRAM满频满容量扩展,性能优于现有同类加速器及部分商用芯片。

摘要:

近年来涌现出许多基于 FPGA 的存内计算(Processor-in-Memory,PIM)架构,它们承诺能实现极高的并行度,但实际性能却未达预期,原因在于最高时钟频率降低、处理单元无法扩展至 BRAM 的最大容量,以及对大型归约操作的硬件支持极少。本文首先确立了我们认为基于 PIM 的 FPGA 设计应遵循的一套“黄金标准”设计目标。该黄金标准既是对比不同技术节点和厂商系列下开发的 PIM 的绝对指标,也是设计者的理想目标。\n 随后我们提出了 IMAGine——一款存内加速 GEMV 引擎,以此作为案例研究,证明黄金标准可在实践中实现。IMAGine 作为存在性证明,打破了围绕 FPGA 时钟与性能扩展限制的若干普遍认知迷思。具体而言,IMAGine 以 BRAM 的最高频率运行,且可扩展至占用 100% 的可用 BRAM。对比分析显示,其执行速度优于现有基于 FPGA 的 PIM 型 GEMV 引擎,时钟速度快 2.65 倍至 3.2 倍。我们展示了在 AMD Alveo U55 上的实现方案,其系统时钟速度达 737 MHz,为 GEMV 运算提供 64K 个位串行乘累加(multiply-accumulate,MAC)单元。这使 IMAGine 成为最快的基于 PIM 的 GEMV 覆盖架构,性能甚至超过迄今已报道的定制化 PIM 型 FPGA 加速器。此外,它的时钟速度超越 TPU v1-v2 与阿里巴巴含光 800,同时提供数量相当或更多的 MAC 单元。

英文摘要:

Many recent FPGA-based Processor-in-Memory (PIM) architectures have appeared with promises of impressive levels of parallelism but with performance that falls short of expectations due to reduced maximum clock frequencies, an inability to scale processing elements up to the maximum BRAM capacity, and minimal hardware support for large reduction operations. In this paper, we first establish what we believe should be a "Gold Standard" set of design objectives for PIM-based FPGA designs. This Gold Standard was established to serve as an absolute metric for comparing PIMs developed on different technology nodes and vendor families as well as an aspirational goal for designers. We then present IMAGine, an In-Memory Accelerated GEMV engine used as a case study to show the Gold Standard can be realized in practice. IMAGine serves as an existence proof that dispels several myths surrounding what is normally accepted as clocking and scaling FPGA performance limitations. Specifically, IMAGine clocks at the maximum frequency of the BRAM and scales to 100% of the available BRAMs. Comparative analyses are presented showing execution speeds over existing PIM-based GEMV engines on FPGAs and achieving a 2.65x - 3.2x faster clock. An AMD Alveo U55 implementation is presented that achieves a system clock speed of 737 MHz, providing 64K bit-serial multiply-accumulate (MAC) units for GEMV operation. This establishes IMAGine as the fastest PIM-based GEMV overlay, outperforming even the custom PIM-based FPGA accelerators reported to date. Additionally, it surpasses TPU v1-v2 and Alibaba Hanguang 800 in clock speed while offering an equal or greater number of MAC units.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑