Parallel Tensor Network Simulator QTensor*

We present a parallel quantum circuit simulator QTensor* for running on large supercomputers with the eventual goal to run at scale on exa-scale supercomputers Aurora and Frontier. The simulator is based on the tensor network representation of quantum circuits and designed to run efficiently on both CPUs and GPUs. We implemented NumPy, PyTorch, and CuPy backends and benchmarked the codes to find the optimal allocation of tensor simulations to either a CPU or a GPU. We also present a dynamic mixed backend to achieve optimal performance. To demonstrate the performance, we simulate QAOA circuits for computing the MaxCut energy expectation. Our method achieves 176 times speedup on a GPU over the NumPy baseline on a CPU for the benchmarked QAOA circuits to solve MaxCut problem on a 3-regular graph of size 30 with depth p=4.