MemGen: An Open-Source Framework for Autonomous Generation of Memory Macros

Autor: Shourya Gupta, Sumanth Kamineni, Benton H. Calhoun
Rok vydání: 2021
Předmět:
Zdroj: CICC
Popis: Static Random-Access Memories (SRAMs) form an integral part of System-on-Chips (SoCs), wireless sensor nodes and other Internet-of-Things (IoT) devices. SRAM has a large, multi-dimensional design space that includes different bit-cell designs, peripheral assist-circuit designs, operating voltages, and frequency targets. Custom design of memories for any application in this broad design space is a tedious, iterative, and time-consuming process. Commercial memory compilers (CMCs) [1 – 3] provide an automated alternative, but CMCs often have a limited design space usually emphasizing high performance and may not be readily available due to cost or licensing issues, especially for newer technologies. To address these issues and allow easy, autonomous, and versatile generation of optimized memory macros, we present MemGen (“Memory Macro Generator”), an open-source memory macro generation framework that creates tapeout-ready integrated memories across a broad range of voltages, frequencies, and capacities. The new framework uses a template and cell-based design methodology and leverages the conventional digital tool flow to generate optimized memories based on high-level user intent, making it highly modular, process-portable, and easily augmentable. We demonstrate the framework’s capability by generating multiple memories for various use cases in a planar 65nm and in a 12nm FinFET process. MemGen is also verified by fabricating 64kbit and 128kbit 65nm auto-generated memories.
Databáze: OpenAIRE