Memristive hybrid on-chip memory for a low-power RISC-V processor - Design and Implementation (HYB-RISC)

用于低功耗 RISC-V 处理器的忆阻混合片上存储器 - 设计和实现 (HYB-RISC)

• 批准号: 536099247
• 负责人: Professor Dr.-Ing. Dietmar Fey
• 金额: --
• 依托单位: Chair of Computer Science 3 (Hardware Architectures)
• 依托单位国家: 德国
• 项目类别: Priority Programmes
• 资助国家: 德国
• 项目状态: 未结题
• 来源: https://gepris.dfg.de/gepris/projekt/536099247?language=en
• 关键词: Memristive; hybrid; chip; memory; low

The aim of the project is to demonstrate that the use of memristive ReRAM memories in close combination with flip-flops and SRAM memory in so-called hybrid non-volatile registers and memories can noticeably reduce energy consumption and power consumption in future processors. This also requires adaptation of existing design tools so that complicated designs such as a processor can use memristive memories, precisely RRAMs, for automatic synthesis. This has not been the case to date. The proof is to be provided by means of a simulation system to be developed, which will use online learning in neural networks to show that energy and power consumption are noticeably reduced by more energy-efficient reading in RRAMs and energy-efficient writing in RRAMs. The proof that the necessary extension of the design programs carried out in the project works correctly is to be provided by the production and measurement of synthesized chips.

该项目的目的是证明，在所谓的混合非挥发性登记册中，使用互联网记忆与触发器和SRAM记忆密切相结合，并且记忆可以明显地减少未来处理器中的能耗和功耗。这也需要适应现有的设计工具，以便像处理器这样的复杂设计可以使用回忆的记忆（精确的RRAM）进行自动合成。迄今为止并非如此。该证明是通过要开发的模拟系统提供的，该模拟系统将在神经网络中使用在线学习，以表明通过RRAMS中的更节能的阅读和RRAMS中的节能写作明显降低了能源和功耗。可以通过合成芯片的生产和测量来提供的项目正确工作的设计程序必要扩展的证据。