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

Update search for bitcode files for rocm 3.9 #6865

Merged
merged 1 commit into from
Nov 7, 2020
Merged

Conversation

t-vi
Copy link
Contributor

@t-vi t-vi commented Nov 6, 2020

rocm 3.9 moved the bitcodes, we adapt to that.
As this gives opaque error messages that are hard to debug (loading the module fails with could not initialize shared object but does not tell you about the missing symbols), we tighten the checks at this stage:

  • we become more strict with missing bitcodes,
  • we let the linker fail loudly for unresolved symbols.

@masahi

rocm 3.9 moved the bitcodes, we adapt to that.
As this gives opaque error messages that are hard to debug
(loading the module fails with could not initialize shared object
but does not tell you about the missing symbols), we tighten
the checks at this stage:
- we become more strict with missing bitcodes,
- we let the linker fail loudly for unresolved symbols.
@t-vi
Copy link
Contributor Author

t-vi commented Nov 6, 2020

I could use a hint regarding the CI failure.

@jroesch
Copy link
Member

jroesch commented Nov 6, 2020

It looks like it could be flaky networking? networking code can be flaky even on localhost. I re-ran pipeline any thoughts @tqchen or @areusch

@areusch
Copy link
Contributor

areusch commented Nov 6, 2020

@jroesch @t-vi best guess is a race condition in the test--it launches an rpc.Server, doesn't wait for it to start listening, then immediately connects. you can see in the captured teardown stdio that the "listening" message is printed, so i'd guess the test could use like a sleep(0.1) or better yet a retry on the connect call.

@masahi masahi merged commit 89ce1ed into apache:main Nov 7, 2020
@masahi
Copy link
Member

masahi commented Nov 7, 2020

Thanks @t-vi

@t-vi
Copy link
Contributor Author

t-vi commented Nov 7, 2020

Thank you @masahi and @jroesch and @areusch!

trevor-m pushed a commit to trevor-m/tvm that referenced this pull request Dec 2, 2020
rocm 3.9 moved the bitcodes, we adapt to that.
As this gives opaque error messages that are hard to debug
(loading the module fails with could not initialize shared object
but does not tell you about the missing symbols), we tighten
the checks at this stage:
- we become more strict with missing bitcodes,
- we let the linker fail loudly for unresolved symbols.
trevor-m pushed a commit to trevor-m/tvm that referenced this pull request Dec 4, 2020
rocm 3.9 moved the bitcodes, we adapt to that.
As this gives opaque error messages that are hard to debug
(loading the module fails with could not initialize shared object
but does not tell you about the missing symbols), we tighten
the checks at this stage:
- we become more strict with missing bitcodes,
- we let the linker fail loudly for unresolved symbols.
trevor-m pushed a commit to neo-ai/tvm that referenced this pull request Dec 4, 2020
rocm 3.9 moved the bitcodes, we adapt to that.
As this gives opaque error messages that are hard to debug
(loading the module fails with could not initialize shared object
but does not tell you about the missing symbols), we tighten
the checks at this stage:
- we become more strict with missing bitcodes,
- we let the linker fail loudly for unresolved symbols.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants