Two-Qubit Gate Synthesis via Linear Programming for Heterogeneous Instruction Sets

We present GULPS (Global Unitary Linear Programming Synthesis), a segmented Cartan trajectory method that compiles arbitrary two-qubit unitaries into native gate sets by decomposing the synthesis problem into locally solvable sub-problems. Each segment corresponds to a depth-2 circuit synthesized from a linear program over canonical gate invariants, subject to quantum Littlewood-Richardson (QLR) constraints. The intermediate invariants are stitched together via nonlinear least-squares optimization to recover the local operations between segments. This hybrid LP-numerical method enables robust synthesis across parameterized instruction sets. As quantum hardware continues to evolve, GULPS provides a scalable, ISA-aware compilation strategy suitable for integration into platforms like Qiskit.