8000 Question about multi-GPU support · Issue #635 · quantumlib/qsim · GitHub
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content
8000

Question about multi-GPU support #635

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its 8000 maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Open
jakurzak opened this issue Nov 7, 2023 · 3 comments
Open

Question about multi-GPU support #635

jakurzak opened this issue Nov 7, 2023 · 3 comments
Assignees
@sergeisakov
Labels
area/gpu Involves CUDA, GPUs, and similar

Comments

@jakurzak
Copy link
Contributor
jakurzak commented Nov 7, 2023

Is there multi-GPU support in Qsim or there any plans to implement it?
Is there any intention to implement distributed memory capabilities?
What can I do when the number of qubits is such that the memory of a single GPU is exceeded?
Any feedback appreciated!
@sergeisakov
Copy link
Collaborator

There is no native multi-GPU support in qsim and there are no plans to implement it. Note that there is multi-GPU support in qsimcirq via NVIDIA cuQuantum Appliance.

@jakurzak
Copy link
Contributor Author
jakurzak commented Nov 7, 2023

Sorry for a naive question, but what about the technique described here: https://quantumai.google/qsim/tutorials/qsimcirq?authuser=1#advanced_applications_distributed_execution

82C7
@95-martin-orion
Copy link
Collaborator

What about the technique described here: {...}

The "circuit cutting" technique applied by qsimh is equivalent to running multiple completely separate simulations and combining the final results. No communication takes place during circuit execution. Note also that qsimh is fairly out of date - it does not have a Cirq interface or support running on GPU (in contrast with qsim.)

Depending on the exact circuit you want to simulate, you might still be able to use this technique to distribute the work and avoid exceeding GPU memory limits. However, you would need to manually construct the cut circuits and feed them into qsim (not qsimh) for simulation.

@mhucka mhucka added the area/gpu Involves CUDA, GPUs, and similar label May 11, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@sergeisakov sergeisakov

Labels
area/gpu Involves CUDA, GPUs, and similar
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
4 participants
@mhucka @95-martin-orion @sergeisakov @jakurzak
0