Abstract:
Content Addressable Memory (CAM)accesses a stored data word based on its contents, compares it with a search word and returns the address of the matched word. This comparison and search operationsare done within the memorycell, without the assistance of Arithmetic Logic Unit (ALU). Due to this in-memory search facility, CAM can be a crucial circuit element in high-speed search engine, data base engine, artificial neural network for neuromorphic computingand other applications for Internet of Things (IoT).Ternary Content Addressable Memory (TCAM) can store and compare two bits of data in a single cell.Thus it can be used to represent three states- High, Low and Don‟t Care. This third state can provide Masking, which allows to search a portion of the word, instead of the whole word. Thus TCAMadds more flexibility in the search operation, and hence, has huge demand in IoT applications. In this work, we have proposed a non-volatile memristor-MOSFET hybrid architecture of TCAM cell. The memristormodel used in this work is calibrated to match the physical characterization data of practical Ta-O based memristor. Furthermore, two match-line segmentation schemes- selective charging and pipelining are implemented in order to make the proposed TCAM cell work more reliably and efficiently in larger memory arrays. Finally, corner simulations are done to verify the robustness of the proposed design against process, voltage and temperature variation. The results obtained from the simulation are compared with that of conventional transistor based TCAM design and other memristor-MOSFET hybrid TCAM designs. The comparison shows significant performance improvement based on the following criteria- search time, search energy (per bit) and voltage margin. This reduces the overall search time up to 83%, energy per bit up to 77.81% during match and 99.45% during mismatch; and increases voltage margin up to 81.19% than the conventional TCAM design with MOSFETs. In other words, the proposed design shows significant performance improvement with respect to conventional TCAM designs and more practical applicability than other memristor-MOSFET hybrid TCAM designs. Therefore, this design can be a better alternative for future memory, computing and IoT applications.
