Stalls and Spequlation: Pipelined Execution for Fault Tolerant Quantum Computation

Abstract

Fault-tolerant quantum computation requires the coordinated action of three distinct systems: classical control logic, quantum hardware, and classical error decoders. Current scheduling models treat logical operations as atomic, hiding the fact that these subsystems operate sequentially and spend significant time idle. We present a pipelined execution framework that decomposes each logical operation into its component stages i.e. Control, Execute, and Decode. Building on this, we discuss some speculation strategies that allow successor operations to begin processing before their predecessors have completed decoding. We evaluate our framework on several common benchmarks and show that pipelining with speculation reduces total pipeline steps by 20-40% compared to a no-speculation baseline. The most aggressive strategy consistently outperforms conservative alternatives, even though partial rollback is needed at times, because the per-rollback penalty is small relative to the parallelism gained. We further show that speculation facilitates load balancing by distributing work more evenly across the heterogeneous subsystems of a fault-tolerant quantum computer, converting idle time into useful computation while also saving on execution time.