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

1.10: Apparent segfault in precompilation serialization #56947

Open
wsmoses opened this issue Jan 4, 2025 · 5 comments · May be fixed by #56973
Open

1.10: Apparent segfault in precompilation serialization #56947

wsmoses opened this issue Jan 4, 2025 · 5 comments · May be fixed by #56973

Comments

@wsmoses
Copy link
Contributor

wsmoses commented Jan 4, 2025 •
edited
Loading

I'm not sure how to properly debug this.

But essentially here (EnzymeAD/Reactant.jl#353 (comment)) we're trying to add precompilation to reactant. Everything is fine on 1.11. On 1.10, loading of the serializatized object hits a julia assertion error in particular here:

julia/src/method.c

Line 1248 in 4976d05

assert(key == 0);

This data isn't written to after mmaping in the compiled code, and where it asserts so I presume it's a bug in Julia's precompilation.

cc @gbaraldi @vchuravy @vtjnash @mofeing
@vchuravy
Copy link
Member

vchuravy commented Jan 6, 2025 •
edited
Loading

What is a MWE for this? The referenced branch got merged and it is unclear what should be run to reproduce the issue.

@mofeing
Copy link
Contributor

mofeing commented Jan 6, 2025

I think you just need to comment this if so you run the branch on 1.10 https://github.com/EnzymeAD/Reactant.jl/blob/75c31406631d3a31502fafee39bd36d734684865/src/Precompile.jl#L42-L43

@vchuravy
Copy link
Member

vchuravy commented Jan 6, 2025

No joy.

julia> using Reactant
2025-01-06 13:50:09.817878: E external/xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: CUDA_ERROR_NO_DEVICE: no CUDA-capable device is detected
AssertionError("No visible GPU devices.")

on 1.10-debug with that branch commented out.

@vchuravy
Copy link
Member

vchuravy commented Jan 6, 2025

Pure speculation, but it might be #54738

@wsmoses
Copy link
Contributor Author

wsmoses commented Jan 6, 2025

EnzymeAD/Reactant.jl#484

@wsmoses wsmoses linked a pull request Jan 6, 2025 that will close this issue
Open
@mofeing mofeing mentioned this issue Jan 6, 2025
Closed
@giordano giordano linked a pull request Jan 6, 2025 that will close this issue
Backport "serialization: fix relocatability bug" to 1.10 #56973
Open
@wsmoses wsmoses mentioned this issue Jan 8, 2025
Open
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging a pull request may close this issue.

Backport "serialization: fix relocatability bug" to 1.10 JuliaLang/julia
3 participants
@vchuravy @wsmoses @mofeing