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

cudaErrorInvalidConfiguration with debug build #1647

Closed
stanmoore1 opened this issue May 29, 2018 · 6 comments
Closed

cudaErrorInvalidConfiguration with debug build #1647

stanmoore1 opened this issue May 29, 2018 · 6 comments
Assignees
Labels
Bug Broken / incorrect code; it could be Kokkos' responsibility, or others’ (e.g., Trilinos)
Milestone

Comments

@stanmoore1
Copy link
Contributor

stanmoore1 commented May 29, 2018

With either KOKKOS_DEBUG=yes or running my executable in cuda-gdb, I'm seeing an invalid configuration error. It appears to be related to the ViewCopy-2D algorithm, which may not be choosing the correct block size:

KokkosP: Allocate<Cuda> name: atom:improper_type pointer: 0x10017000c80 size: 0
KokkosP: Executing parallel-for kernel on device 0 with unique execution identifier 856
KokkosP: Kokkos::ViewCopy-2D
terminate called after throwing an instance of 'std::runtime_error'
  what():  cudaGetLastError() error( cudaErrorInvalidConfiguration): invalid configuration argument ../../lib/kokkos/core/src/Cuda/Kokkos_CudaExec.hpp:401
Traceback functionality not available
@ndellingwood
Copy link
Contributor

@crtrott possibly related to the Panzer example errors in trilinos/Trilinos#2827

@crtrott crtrott added the Bug Broken / incorrect code; it could be Kokkos' responsibility, or others’ (e.g., Trilinos) label May 30, 2018
@crtrott crtrott added this to the 2018 July milestone May 30, 2018
@crtrott
Copy link
Member

crtrott commented Aug 13, 2018

Possibly fixed with the changes to the recommended team_size stuff.

@mhoemmen
Copy link
Contributor

@kddevin reported something funny about ViewCopy performance (on host?) for 2-D Views in Tpetra::MultiVector. I forget the issue number at the moment. Could that possibly be related?

@ibaned ibaned assigned ndellingwood and unassigned crtrott Aug 22, 2018
@ibaned
Copy link
Contributor

ibaned commented Aug 22, 2018

@ndellingwood assigned to use @crtrott 's new more accurate block size predictor to determine tile sizes at the right place in the code.

@crtrott
Copy link
Member

crtrott commented Sep 30, 2018

@stanmoore1 can you check whether this was fixed now?

@stanmoore1
Copy link
Contributor Author

I can't reproduce this, even with an older version of Kokkos. The compiler and CUDA version on the testbed could have changed since then, and I also can't remember exactly what input I was using. So this ticket can be closed. If I see it again I'll let you know.

@ibaned ibaned added InDevelop Bug Broken / incorrect code; it could be Kokkos' responsibility, or others’ (e.g., Trilinos) and removed Bug Broken / incorrect code; it could be Kokkos' responsibility, or others’ (e.g., Trilinos) labels Oct 30, 2018
@crtrott crtrott closed this as completed Nov 5, 2018
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Bug Broken / incorrect code; it could be Kokkos' responsibility, or others’ (e.g., Trilinos)
Projects
None yet
Development

No branches or pull requests

5 participants