Speaker
Description
With the computing power far superior to classical computers, quantum computing has attracted increasing attention and interest all over the world. Many prototypes are constructed like IBM Osprey, USTC Jiuzhang 2.0 to demonstrate the advantage of quantum computing, and corresponding developing SDKs are proposed to make the utmost of these devices.
Qiskit is one of the most popular quantum computing developing kits from IBM, which can run quantum circuits on IBM’s physical machines or on simulators in digital computers. Quantum computing simulation for large and/or deep quantum circuits requires quite huge computing resource and memory, and Qiskit can accelerate or simulations or extend the scales on NVIDIA GPUs with the support of CUDA-backend Thrust library.
In the past few years, AMD’s heterogeneous computing framework – ROCm similar to CUDA, has supported many software like TensorFlow, pyTorch, while qiskit has not been ported still.
To expand the usability of qiskit, we made some efforts to porting qiskit to ROCm platform, and performed some benchmark tests on both ROCm and CUDA platform. In this work, we use ROCm 5.0 and tests are performed on a cluster with ROCm accelerating devices, and for comparison, we use CUDA 11.7 and a NVIDIA V100 cluster. The performances on ROCm and CUDA platform are comparable in a way, while more efforts are needed to optimizing the performance.
| Consider for long presentation | No |
|---|
