search

trilinos / Trilinos

Primary repository for the Trilinos Project
https://trilinos.org/
Other
1.21k stars 565 forks source link

ifpack2: broken unit tests w/ cuda 12.4 + h100 gpus #13398

Open vasylivy opened 2 months ago

vasylivy commented 2 months ago

Hi,

Reporting broken units w/ cuda 12.4 + h100 gpus. See reported configurations here https://github.com/trilinos/Trilinos/issues/13397. 

Ifpack2_unit_tests_MPI_4 

w/ configuration 1

p=2 | 37. Ifpack2OverlapGraph_int_longlong_OverlapGraphTest0_UnitTest ... [Passed] (0.0158 sec)
(ptr->cuda_stream_synchronize_wrapper(stream)) error( cudaErrorMisalignedAddress): misaligned address

w/ configuration 2

p=0 | 35. Ifpack2MDF_double_int_longlong_Test1_UnitTest ... [Passed] (0.0129 sec)
(cuda_instance->cuda_event_record_wrapper( CudaInternal::constantMemReusablePerDevice[cuda_device])) error( cudaErrorIllegalAddress): an illegal memory access was encountered

Can someone take a look and try to reproduce these errors?

Thanks,

Yaro

github-actions[bot] commented 2 months ago

Automatic mention of the @trilinos/ifpack2 team