Speed up QC-LDPC decoder through memory subsystem optimization

Yajie Li,Dake Liu,Xinbing Zhou,Wei Chen
DOI: https://doi.org/10.1109/iwecai55315.2022.00027
2022-01-01
Abstract:In this paper, we proposed a QC-LDPC decoder for 5G NR, WiMAX, and WLAN applications. The decoder can flexibly support arbitrary code length and rate under 5G NR standard. This design follows programmable ASIP (Application Specific Instruction-set Processor) design method and uses data parallel SIMD (Single Instruction Multiple Data) architecture. We doubled the logic computing speed to improve the throughout instead of using double speed memory architecture. Compared with the design mentioned in [1], the original design could achieve the highest QC-LDPC decoding throughput of 533Mbps (12 SISO in parallel) when the clock frequency is 200MHz (memory clock frequency is 400MHz). The existing design saved 47% of the decoding clock cycle without lossing decoding circuit quality. We achieved a maximum QC-LDPC decoding throughput of 1984 Mbps (with 48 SISO in parallel) when the system frequency is 400MHz (memory clock frequency is also 400MHz). Further, through circuit optimization, the upper limit of the system clock is promoted from 344MHz to 1.1GHz. The throughput can possibly be upto 5456Mbps. Finally, through the logic synthesis using 28nm SMIC CMOS cell library, the design is physically verified, and the logic gate count is reported 1716K. It is similar to that of the original design per SISO.
What problem does this paper attempt to address?