Non-volatile Memory Application to Quantum Error Correction with Non-uniformly Quantized CiM

Non-volatile memory (NVM)-based Computation-in-Memory (CiM) decoder is proposed for Quantum Error Correction (QEC). When the QEC circuitry is placed at cryogenic temperature, the power and area budgets are limited in order to save the cooling power. Compared with the SRAM-based CiM, NVMs have the advantage of saving memory size, power consumption, and data transfer to load the weights. Through the investigation on bit error rate (BER) tolerance, it is revealed that the accuracy of QEC can be maintained even with the high weight BER (up to 0.3%), opening the door to NVM application to QEC. In addition, two types of non-uniform weight quantization schemes (gate focused and layer focused) are proposed through the comprehensive investigation on bit-precision sensitivity. In the proposed schemes, the memory size is reduced by over 40% compared with the uniformly quantized CiM.