Skip to content
New issue

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

Disable or make automatic prefecthing of unified memory optional #2618

Open
maleadt opened this issue Jan 11, 2025 · 0 comments · May be fixed by #2626
Open

Disable or make automatic prefecthing of unified memory optional #2618

maleadt opened this issue Jan 11, 2025 · 0 comments · May be fixed by #2626
Labels
cuda array Stuff about CuArray. good first issue Good for newcomers performance How fast can we go?

Comments

@maleadt
Copy link
Member

maleadt commented Jan 11, 2025

As diagnosed in #2615, automatically prefetching unified memory arrays is counterproductive when the array is used on multiple devices.

For the time being, we should probably disable this prefetching when multiple devices are available on the system.
@maleadt maleadt added cuda array Stuff about CuArray. good first issue Good for newcomers performance How fast can we go? labels Jan 11, 2025
@efaulhaber efaulhaber mentioned this issue Jan 16, 2025
Draft
@vchuravy vchuravy linked a pull request Jan 20, 2025 that will close this issue
Open
@maleadt maleadt linked a pull request Jan 20, 2025 that will close this issue
Don't prefetch on multi-device systems #2626
Open
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
cuda array Stuff about CuArray. good first issue Good for newcomers performance How fast can we go?
Projects
None yet
Milestone
No milestone
Development

Successfully merging a pull request may close this issue.

Don't prefetch on multi-device systems JuliaGPU/CUDA.jl
1 participant
@maleadt