We reported last year on a new announcement from NVIDIA that they had developed a quantum simulation platform called cuQuantum which uses their GPUs to provide a significant improvement in simulation performance versus running them on standard microprocessors. Subsequently, Xanadu developed a GPU-accelerated simulator for PennyLane called lightning.gpu which uses cuQuantum and now Amazon Web Services (AWS) is supporting this combination in their Amazon Braket Hybrid Jobs module for supporting classical/quantum workloads. This will allow support for very high performance and error free simulations of hybrid algorithms such as variational quantum eigensolver (VQE), quantum approximate optimization algorithms (QAOA), or quantum machine learning (QML). Quantum algorithm developers will be able to develop and simulate programs for various quantum applications using the simulators and then eventually move large instances of those programs to full quantum processors when more powerful quantum processors are available. To make it less complex for the user, Amazon’s Braket software can automatically parallelize the code so it can run the calculation over multiple GPUs for improved latency. This capability can also lower both the cost and the performance of quantum simulations versus some of the other circuit simulators that AWS has available. Additional information about this new capability is described in blog postings on the AWS website here and here.

May 5, 2022