Towards algorithm-aware quantum circuit compilation

A quantum program undergoes several layers of compilation before being executed on the physical device. A vital part of compilation is optimizing the program to get the best possible results from the quantum hardware. This optimization involves several passes - some of which are device-agnostic, and others, such as qubit mapping and routing, which require knowledge of the underlying device architecture and noise model. Currently, however, compilation passes are not cognizant of the structure of the circuits being executed.

In this talk, we describe an algorithm-aware paradigm for quantum compilation. Most quantum applications consist of one or more sub-blocks of standard quantum algorithms and protocols, each of which in turn have a specific circuit structure. In our investigation, we put together a comprehensive set of such algorithms and executed them using a generic set of compilation passes on various devices through Amazon Braket. Next, we ran the same circuits through compiler passes that were specifically tailored for each algorithm and compare the results. Our methodology systematically benchmarks and assesses the impact of algorithm-aware circuit compilation.