Skip to content

GPU Access

Workbench supports GPU based simulators powered by NVIDIA's cuQuantum. These simulators (e.g., state vector, tensor network, matrix product state) let you run larger and faster quantum computing simulations. In order to make the most of these simulators, you will need to use a GPU specific QDE.

How to use a GPU QDE in Construct

  1. Request access by emailing construct@psiquantum.com ✉. Please include the following information in your email.
    1. Subject: Construct GPU access
    2. List the names of those who want access
    3. List the emails of those who want access
    4. State your organization (or indicate you are an individual user)

After you send the email, the Construct team will respond next steps and additional details. 2. Create a QDE with the GPU template 1. Navigate to construct.psiquantum.com/ ⧉ 2. Click QDE 3. Click Create new QDE 4. Select the QDE with CudaQ template 5. Click Create 6. Wake and open the new QDE 3. Update your Workbench code with a single change: use a CUDA-Q filter preset (CUDAQ_STATE_VEC orCUDAQ_TENSORNET) when creating a QPU.

from psiqworkbench import QPU
from psiqworkbench.filter_presets import CUDAQ_STATE_VEC, CUDAQ_TENSORNET

qpu = QPU(num_qubits=n, filters=CUDAQ_STATE_VEC)

Alternatively, you can use environment variables to automatically have all existing Workbench code use the CUDA-Q state vector simulator instead of Workbench CPU state vector simulator. Simply open a terminal window and run the following command:

export OVERRIDE_SV_TO_CUDAQ_SV=1

See Simulating Workbench Programs tutorial ⧉ for more information on using CUDA-Q simulators with Workbench.