Quantum computing - a revolutionary emerging technology - is capable of delivering outstanding computational performance in certain areas of science and technology. A major challenge today is the precise identification of its field of applicability.
In this context, emulating quantum algorithms on classical computers is a useful approach. Current quantum platforms operate on very short time intervals (called pulses), due to a phenomenon of decoherence. The current quantum computing paradigm (see figure) is therefore based on a hybrid model, where a classical computer delegates the execution of very short quantum tasks to a coprocessor. However, heavy and powerful quantum algorithms, likely to bring great added value to a simulation, can nevertheless be studied and evaluated by running them on an emulator operating in the classical side.
These observations have motivated the development of a quantum emulation environment that can be used as a pedagogical and research tool. The desire to integrate into hybrid simulations demanded a high level of computational performance. Most of the development work was carried out on the Jean Zay platform at IDRIS, with an allocation of 100,000 CPU hours and 100,000 GPU hours. The software environment consists of four C++ libraries, implementing four different parallel contexts (shared or distributed memory, with or without GPU support). They are fully portable, running on supercomputers or laptops.
Some architectural choices bring significant performance increases. In addition to GPU activation, the libraries support the operation of quantum gates operating on multiple qubits. This service, combined with an optimization strategy based on "gate fusion", enables us to replace any quantum circuit with a single equivalent multiqubit quantum gate. In some quantum chemistry simulations, GPU activation plus gate fusion results in a 50x performance boost, compared with CPU code without GPU and gate fusion.
The libraries have very detailed documentation, which is in preparation. They also offer a large number of codes featuring examples, as well as some large simulations in several areas of science and technology.