应用于数据密集型场景的功耗-面积高效的基于串行IMPLY的4:2压缩器
Power-Area Efficient Serial IMPLY-based 4:2 Compressor Applied in Data-Intensive Applications
AI总结:
针对数据密集型应用的存储墙瓶颈,提出基于串行忆阻IMPLY的4:2压缩器并构建新型乘法器,在面积、能耗和速度等指标上较现有方案有显著提升。
摘要:
在数据密集型应用中,处理器与存储器之间的数据传输已成为设计瓶颈。存内计算(Processing-In-Memory, PIM)是突破存储墙瓶颈的可行方案。4:2压缩器适用于实现处理器的关键算术电路,包括乘法器。一些面积高效的忆阻结构,例如串行架构中的实质蕴含(Material Implication, IMPLY),与交叉开关阵列兼容。本文提出了一种基于串行忆阻IMPLY的4:2压缩器,并将其用于构建新型4位和8位乘法器。研究从延迟、面积和能耗三个维度对所提电路进行了评估。与现有串行压缩器相比,所提4:2压缩器的算法在面积、能耗和速度上分别优化了36%、17%和15%。与基于XOR/MUX设计的4:2压缩器构建的串行乘法器相比,所提4位和8位乘法器的延迟分别降低了7.3%和10%,能耗最多降低12%。
英文摘要:
The data transfer between a processor and memory has become a design bottleneck in data-intensive applications. Processing-In-Memory (PIM) is a practical approach to overcome the memory wall bottleneck. The 4:2 compressor is suitable for implementing the processor's crucial arithmetic circuits, including multiplier. Some area-efficient memristive structures, like Material Implication (IMPLY) in serial architecture, are compatible with the crossbar array. This paper proposes a serial memristive IMPLY-based 4:2 compressor, which is applied to present new 4-bit and 8-bit multipliers. The proposed circuits are evaluated regarding latency, area, and energy consumption. Compared to the existing serial compressor, the proposed 4:2 compressor's algorithm improves the area, energy consumption, and speed by 36%, 17%, and 15%, respectively. The proposed 4-bit and 8-bit multipliers are improved by 7.3% and 10%, respectively, regarding the latency, and reduced energy consumption by up to 12%, compared to the serial multiplier based on a 4:2 compressor with XOR/MUX design.
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。