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.

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

Runtime not re-selected after driver upgrade #1877

Closed
maleadt opened this issue Apr 19, 2023 · 4 comments · Fixed by JuliaPackaging/Yggdrasil#7514
Closed

Runtime not re-selected after driver upgrade #1877

maleadt opened this issue Apr 19, 2023 · 4 comments · Fixed by JuliaPackaging/Yggdrasil#7514
Labels
bug Something isn't working installation CUDA is easy to install, right?

Comments

@maleadt
Copy link
Member

maleadt commented Apr 19, 2023

We're currently embedding the selected driver in the precompilation image. For example, here I just upgraded my driver, but didn't recompile the Runtime JLL:

❯ julia +1.9
julia> using CUDA
CUDA.versioninfo()

julia> CUDA.versioninfo()
CUDA runtime 11.8, artifact installation
CUDA driver 12.1
NVIDIA driver 530.41.3


❯ jl +1.9 --pkgimages=no --compiled-modules=no
julia> using CUDA
  Downloaded artifact: CUDA_Runtime

julia> CUDA.versioninfo()
CUDA runtime 12.1, artifact installation
CUDA driver 12.1
NVIDIA driver 530.41.3

That's not great. cc @vchuravy

@maleadt maleadt added bug Something isn't working installation CUDA is easy to install, right? labels Apr 19, 2023
@maleadt
Copy link
Member Author

maleadt commented Apr 19, 2023

Forcibly recompiling the relevant packages didn't help:

julia> Base.compilecache(Base.PkgId(Base.UUID("76a88914-d11a-5bdc-97e0-2f5a05c973a2"), "CUDA_Runtime_jll"))
[ Info: Precompiling CUDA_Runtime_jll [76a88914-d11a-5bdc-97e0-2f5a05c973a2]
("/home/tim/Julia/depot/compiled/v1.9/CUDA_Runtime_jll/Hs50y_u396q.ji", nothing)

julia> Base.compilecache(Base.PkgId(Base.UUID("052768ef-5323-5732-b1bb-66c8b64840ba"), "CUDA"))
[ Info: Precompiling CUDA [052768ef-5323-5732-b1bb-66c8b64840ba]
("/home/tim/Julia/depot/compiled/v1.9/CUDA/oWw5k_u396q.ji", nothing)

❯ jl +1.9 --pkgimages=no
julia> using CUDA

julia> CUDA.versioninfo()
CUDA runtime 11.8, artifact installation
CUDA driver 12.1
NVIDIA driver 530.41.3

Removing all images did:

❯ rm -rf /home/tim/Julia/depot/compiled/v1.9/CUDA*

❯ jl +1.9 --pkgimages=no
julia> using CUDA
[ Info: Precompiling CUDA [052768ef-5323-5732-b1bb-66c8b64840ba]

julia> CUDA.versioninfo()
CUDA runtime 12.1, artifact installation
CUDA driver 12.1
NVIDIA driver 530.41.3

@MariusDrulea
Copy link

Just had this issue. I have added CUDA.jl while using nouveau driver on Ubuntu, then switched to an Nvidia Driver. I had to remove the compiled CUDA stuff, just like in the previous comment.

@vchuravy
Copy link
Member

vchuravy commented Oct 7, 2023

We could use Base.include_dependency to invalidate the cache-files when a file on disk changes.

x-ref: JuliaLang/julia#49866 (comment)

@maleadt
Copy link
Member Author

maleadt commented Oct 9, 2023

We could use Base.include_dependency to invalidate the cache-files when a file on disk changes.

I know, but manually invalidating the CUDA_Runtime_jll compilecache didn't help; see initial post. Adding an include dependency does seem to work, even though that just invalidates the compilecache... Not sure what's happening, but I added the change in JuliaPackaging/Yggdrasil#7514

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
bug Something isn't working installation CUDA is easy to install, right?
Projects
None yet
Development

Successfully merging a pull request may close this issue.

3 participants