tpetra cuda warning looks legit #5698
Labels
impacting: configure or build
The issue is primarily related to configuring or building
pkg: Tpetra
system: gpu
Comments
bathmatt
added
the
type: bug
|
More of them |
mhoemmen
added
impacting: configure or build
|
@trilinos/kokkos |
kddevin
added a commit
that referenced
this issue
Jan 16, 2020
This removes some warnings about calling host functions from host device functions #5698; E.g., warning: calling a __host__ function("std::basic_string<char, std::char_traits<char>, std::allocator<char> >::~basic_string") from a __host__ __device__ function("Tpetra::Details::LocalMap<int, long long, ::Kokkos::Device< ::Kokkos::Serial, ::Kokkos::HostSpace> > ::~LocalMap [subobject]") is not allowed
jmgate
pushed a commit
to tcad-charon/Trilinos
that referenced
this issue
Jan 17, 2020
…s:develop' (d17489d). * trilinos-develop: Tpetra: Missed ifdef guard zoltan2: name change to prevent shadow warnings zoltan2: Change logic for determining gno types to use in tests Simplified now that Trilinos builds only one gno_t Tpetra: More stacked timer fixes Tpetra: Fix overflow for TpetraCore_MatrixMatrix_UnitTests tpetra: removed unused field from FixedHashTable This removes some warnings about calling host functions from host device functions trilinos#5698; E.g., warning: calling a __host__ function("std::basic_string<char, std::char_traits<char>, std::allocator<char> >::~basic_string") from a __host__ __device__ function("Tpetra::Details::LocalMap<int, long long, ::Kokkos::Device< ::Kokkos::Serial, ::Kokkos::HostSpace> > ::~LocalMap [subobject]") is not allowed
jmgate
pushed a commit
to tcad-charon/Trilinos
that referenced
this issue
Jan 18, 2020
…s:develop' (d17489d). * trilinos-develop: SEACAS: go back to lib:fmt 6.0.0 until fix issue on vortex xl/cuda build Disable Teko_testdriver_tpetra_MPI_4 in all atdm 'waterman' builds (trilinos#6463) zoltan2: add missing include file for non-ETI builds Tpetra: Missed ifdef guard zoltan2: name change to prevent shadow warnings zoltan2: Change logic for determining gno types to use in tests Simplified now that Trilinos builds only one gno_t Tpetra: More stacked timer fixes Tpetra: Fix overflow for TpetraCore_MatrixMatrix_UnitTests tpetra: removed unused field from FixedHashTable This removes some warnings about calling host functions from host device functions trilinos#5698; E.g., warning: calling a __host__ function("std::basic_string<char, std::char_traits<char>, std::allocator<char> >::~basic_string") from a __host__ __device__ function("Tpetra::Details::LocalMap<int, long long, ::Kokkos::Device< ::Kokkos::Serial, ::Kokkos::HostSpace> > ::~LocalMap [subobject]") is not allowed tpetra: when looking at trilinos#6158, I saw that, when creating a Vector from a MultiVector, we created subviews for comm buffers, but did not store them. This commit stores them. It also offsets the buffers by the vector j requested from the MultiVector.
jmgate
pushed a commit
to tcad-charon/Trilinos
that referenced
this issue
Jan 19, 2020
…s:develop' (d17489d). * trilinos-develop: SEACAS: go back to lib:fmt 6.0.0 until fix issue on vortex xl/cuda build Disable Teko_testdriver_tpetra_MPI_4 in all atdm 'waterman' builds (trilinos#6463) zoltan2: add missing include file for non-ETI builds Tpetra: Missed ifdef guard zoltan2: name change to prevent shadow warnings zoltan2: Change logic for determining gno types to use in tests Simplified now that Trilinos builds only one gno_t Tpetra: More stacked timer fixes Tpetra: Fix overflow for TpetraCore_MatrixMatrix_UnitTests tpetra: removed unused field from FixedHashTable This removes some warnings about calling host functions from host device functions trilinos#5698; E.g., warning: calling a __host__ function("std::basic_string<char, std::char_traits<char>, std::allocator<char> >::~basic_string") from a __host__ __device__ function("Tpetra::Details::LocalMap<int, long long, ::Kokkos::Device< ::Kokkos::Serial, ::Kokkos::HostSpace> > ::~LocalMap [subobject]") is not allowed tpetra: when looking at trilinos#6158, I saw that, when creating a Vector from a MultiVector, we created subviews for comm buffers, but did not store them. This commit stores them. It also offsets the buffers by the vector j requested from the MultiVector.
jmgate
pushed a commit
to tcad-charon/Trilinos
that referenced
this issue
Jan 20, 2020
…s:develop' (d17489d). * trilinos-develop: tpetra: In trilinos#6598, @mhoemmen recommended this change of offset SEACAS: go back to lib:fmt 6.0.0 until fix issue on vortex xl/cuda build Disable Teko_testdriver_tpetra_MPI_4 in all atdm 'waterman' builds (trilinos#6463) zoltan2: add missing include file for non-ETI builds Tpetra: Missed ifdef guard zoltan2: name change to prevent shadow warnings zoltan2: Change logic for determining gno types to use in tests Simplified now that Trilinos builds only one gno_t Tpetra: More stacked timer fixes Tpetra: Fix overflow for TpetraCore_MatrixMatrix_UnitTests tpetra: removed unused field from FixedHashTable This removes some warnings about calling host functions from host device functions trilinos#5698; E.g., warning: calling a __host__ function("std::basic_string<char, std::char_traits<char>, std::allocator<char> >::~basic_string") from a __host__ __device__ function("Tpetra::Details::LocalMap<int, long long, ::Kokkos::Device< ::Kokkos::Serial, ::Kokkos::HostSpace> > ::~LocalMap [subobject]") is not allowed tpetra: when looking at trilinos#6158, I saw that, when creating a Vector from a MultiVector, we created subviews for comm buffers, but did not store them. This commit stores them. It also offsets the buffers by the vector j requested from the MultiVector.
jmgate
pushed a commit
to tcad-charon/Trilinos
that referenced
this issue
Jan 21, 2020
…s:develop' (d17489d). * trilinos-develop: tpetra: In trilinos#6598, @mhoemmen recommended this change of offset Tpetra::CrsMatrix: Add Kokkos kernel labels; expose debug code Tpetra::CrsMatrix: Remove values2D_ Tpetra::CrsGraph: Remove gblInds2D_ Tpetra::CrsGraph: Remove lclInds2D_ Tpetra::CrsMatrix: Remove unused method allocateValues2D Tpetra: Use verbosePrintCountThreshold in copyOffsets Tpetra::Details::Behavior: Add longRowMinNumEntries Tpetra::Details::Behavior: Factor out size_t reading SEACAS: go back to lib:fmt 6.0.0 until fix issue on vortex xl/cuda build Disable Teko_testdriver_tpetra_MPI_4 in all atdm 'waterman' builds (trilinos#6463) zoltan2: add missing include file for non-ETI builds Tpetra: Missed ifdef guard zoltan2: name change to prevent shadow warnings zoltan2: Change logic for determining gno types to use in tests Simplified now that Trilinos builds only one gno_t Tpetra: More stacked timer fixes Tpetra: Fix overflow for TpetraCore_MatrixMatrix_UnitTests tpetra: removed unused field from FixedHashTable This removes some warnings about calling host functions from host device functions trilinos#5698; E.g., warning: calling a __host__ function("std::basic_string<char, std::char_traits<char>, std::allocator<char> >::~basic_string") from a __host__ __device__ function("Tpetra::Details::LocalMap<int, long long, ::Kokkos::Device< ::Kokkos::Serial, ::Kokkos::HostSpace> > ::~LocalMap [subobject]") is not allowed Framework: updating the autotester env to remove dependency on atdm-env tpetra: when looking at trilinos#6158, I saw that, when creating a Vector from a MultiVector, we created subviews for comm buffers, but did not store them. This commit stores them. It also offsets the buffers by the vector j requested from the MultiVector.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Bug Report
@trilinos/tpetra
In tracking down some of these cuda issues in tpetra I'm seeing this warning which could be an error. Want to make sure it is not an issue
I'm seeing this on vortex
The text was updated successfully, but these errors were encountered: