Join us via Zoom as we welcome Satvik Maurya, a doctoral student at the University of Wisconsin-Madison. He'll be presenting "Managing Classical Processing Requirement for Fault-Tolerant Quantum Computers."
ABSTRACT: Quantum computing has rapidly progressed in recent years resulting in significantly larger systems consisting of 400+ qubits. Such systems have been used to demonstrate the effectiveness of quantum error correction, which improves the usefulness of quantum computers due to reduced error rates. As systems grow larger, the classical FPGA/ASIC-based control hardware required for controlling and decoding errors must be scaled appropriately.
In this talk, Satvik Maurya will discuss how decoders required for quantum error correction can be virtualized to reduce physical hardware requirements. Quantum error correction involves encoding multiple physical qubits into a single logical qubit. Error detection and correction is performed by constantly generating syndromes which in turn require specialized hardware decoders to detect errors in real-time. Brute-force scaling of these decoders for every logical qubit in the system can be very expensive. However, reducing the number of decoders in the system can result in an exponential increase in the memory required to store undecoded syndromes, resulting in an exponential slowdown of the system. To address this trade-off between compute and memory, we propose the notion of virtual decoders for quantum error correction, where decoding is scheduled optimally, resulting in a ~10x reduction in the number of hardware decoders while consuming less than 100MB of memory.
SAVTIK MAURYA is a fourth-year doctoral student at the University of Wisconsin-Madison. He works at the intersection of computer architecture and quantum computing. His work has appeared at architecture conferences ISCA and MICRO. Maurya has won the Hiran Mayuk award and is an alumnus of the Heidelberg Laureate Forum. Prior to grad school, he worked as a protocol transactor developer at Synopsys, Inc.
Read More