Research
Trident three-level memory design cuts yoked surface code load verification time
Gunagya Singh Mamak, Suhas Vittal and Moinuddin K. Qureshi (Georgia Institute of Technology), with Kabir Dubey and Kaitlin N. Smith (Northwestern University), propose Trident, a three-level memory hierarchy for fault-tolerant quantum computers. Verifying qubits loaded from two-dimensional Yoked Surface Codes (YSCs) currently requires measuring many checks called yokes, taking 178 cycles. Trident adds an intermediate tier of faster, less dense one-dimensional YSCs. Simulations show speedups of up to 4.2x with a 4.9% increase in physical qubit cost, and the authors describe two variants, demand-driven Trident-D (2.8x) and prefetch-driven Trident-P, which delivers the peak gains. The work is posted on arXiv and is not yet peer reviewed.
Why it mattersLong verification times for loading data from compact quantum memory limit the performance of fault-tolerant machines. The reported results come from simulations under ideal conditions that do not yet account for real hardware imperfections or decoherence during prefetching, and the gains depend on frequently used data sitting in the fastest tier.}