A quantum assisted intermediate representation generation using llvm

Abstract

The present invention is related to a quantum assisted intermediate representation generation using LLVM. Current QIRs have three key problems: (1) poor optimization pipelines that do not use classical compiler techniques, (2) an inability to create a standard toolchain that will enable cross platform interoperability, and (3) a lack of formal verification to verify the correctness of circuit rewrites. Such issues are even more serious in hybrid quantum-classical algorithms, and smooth integration is extremely important. Quantum systems tend to compute quantum functions as black boxes and thus cannot be optimized and may experience compilation failure. Classical compilers, which have been developed over decades, are well matured and well connected but quantum compilers are not. There is recent interest in the scalability of LLVM-based QIR, although it is claimed that it has no measurement semantics, conflicting intrinsic definitions, and does not provide enough classical-quantum interaction representation. Multi-level IR hierarchies, entanglement-sensitive dataflow analysis and type-theoretic verification all have promise and bring new trade-offs between verification completeness and its performance overhead. These limitations are considered in this invention where it provides some improvements in superposition optimization, hardware-software co-design, and time-domain optimizations. It works to bridge semantic gaps, advance interoperability, and combine quantum compilation with more mature classical infrastructures towards efficient, verifiable and consistent near-term quantum applications.