I'm practicing with qiskit experiments. Try to work it with GPU installed in my system. But it takes 100% performance of CPU and didn't consider the GPU.
Kindly help me to run the qiskit codes with GPU.
Thanks..
3
1 Answer
$\begingroup$
$\endgroup$
4
According to qiskit-aer README, you can install qiskit-aer-gpu to utilize GPU for simulation. This package requires CUDA® 10.1 or newer previously installed. CUDA® itself would require a set of specific GPU drivers. On Linux, you can install by pip install qiskit-aer-gpu, on other platforms you need to build the package from source.
Note that qiskit-aer and qiskit-aer-gpu are mutually exclusive. They contain the same code but the latter supports CUDA. If you install both at the same it will interfere with each other. See this answer for more details.
Furthermore, you can use cusvaer or cuStateVec from NVIDIA’s cuQuantum SDK (requires CUDA as well) with Qiskit to utilize GPU for simulation. See some code examples here and the cuQuantum documentation.
-
$\begingroup$ This page states: "If you want to install our GPU supported simulators, you have to install this other package:
pip install qiskit-aer-gpuThis will overwrite your currentqiskit-aerpackage installation giving you the same functionality found in the canonicalqiskit-aerpackage, plus the ability to run the GPU supported simulators: statevector, density matrix, and unitary." If this information is correct (no need to uninstallqiskit-aeranymore), can you update your answer to reflect so please? $\endgroup$ Feb 19 at 15:11 -
$\begingroup$ I am not sure it’s needed to uninstall qiskit-aer and I can’t test I because I don’t have a machine with NVIDIA GPU… $\endgroup$ Feb 20 at 13:01
-
$\begingroup$ No problem, I'll try to test it and will tell you the result. Don't hesitate to ping me if I forget to do so in several days, I think this question could be useful to a large number of people, so it might be worth the effort! $\endgroup$ Feb 20 at 13:18
-
cudaErrorNoKernelImageForDevice$\endgroup$