Optimizing Quantum Compilation via High-Level Quantum Instructions

Abstract: Current quantum programming is dominated by low-level, circuit-centric approaches that limit the potential for compiler optimization. This work presents how a high-level programming construct provides compilers with the semantic information needed for advanced optimizations. We introduce a novel optimization that leverages a quantum-specific instruction to automatically substitute quantum gates with more efficient, approximate decompositions, a process that is transparent to the programmer and significantly reduces quantum resource requirements. Furthermore, we show how this instruction guarantees the correct uncomputation of auxiliary qubits, enabling safe, dynamic quantum memory management. We illustrate these concepts by implementing a V-chain decomposition of the multi-controlled NOT gate, showing that our high-level approach not only simplifies the code but also enables the compiler to generate a circuit with up to a 50% reduction in CNOT gates. Our results suggest that high-level abstractions are crucial for unlocking a new class of powerful compiler optimizations, paving the way for more efficient quantum computation.