Asip data-plane processor for multi-standard wireless protocol processing

Evolving Multi-Protocol Multi-Band Software Defined Radio (SDR) devices aim at supporting multiple protocols seamlessly and efficiently. The design of such radios necessitates flexibility in physical layer processing, flexibility in routing packets through processing engines and flexibility in radio frequency reception/transmission. This dissertation addresses an efficient implementation of flexible physical layer processing (PHY) for Interleaving, De-Interleaving and linear Minimum Mean Square Error (MMSE) detection in Multiple Input Multiple Output (MIMO) receivers through Application Specific Instruction Set Processors (ASIPs). The thesis defines and develops a WINLAB cognitive radio (WiNC2R) compatible data-plane ASIP architecture along with suitable hardware-software partitioning of the Processing Engine unit.;Given the requirement of very significant design time and the lack of the flexibility after design, dedicated ASIC for PHY may not be a viable option although it has the best performance among all available options. The software application running on general purpose processor cannot satisfy the throughput requirements of the wireless standards. ASIPs provide a better trade-off between flexibility and performance, with the advantage of considerably lower design time than ASICs. We design an efficient multi-standard (802.11a, 802.16e/m) supporting Interleaver/De-Interleaver ASIP, satisfying the throughput requirements for all the modulation-schemes/data-rates in both of the standards. It can be programmed to scale for supporting future wireless standards (that use Block Interleaving/De-Interleaving). We also study viability of a flexible MIMO MMSE detector ASIP supporting variable MR (Number of receiving antennas) * MT (Number of transmitting antennas) operations. We have analyzed the implementation of an hardware-centric algorithm for MIMO detection on an ASIP and also improved its performance with the help of techniques such as fixed point implementation, Single Instruction Multiple Data (SIMD) and Very Long Instruction Word (VLIW). Analysis of the design performance results for MIMO ASIP indicates the limitations of hardware-implementation-specific algorithms on ASIP. We also provide the account of design decisions such as custom ports, memory interfaces and registers that are added to the data-plane processor ASIPs in order to substitute them for dedicated hardware engines in the WiNC2R platform.