Skip to content

fix: resolve SIGSEGV on HIP builds #3992

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

Merged
castelletto1 merged 4 commits into from
Mar 9, 2026
Merged

fix: resolve SIGSEGV on HIP builds #3992

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
9bdf8c1
Fix SIGSEGV from Comm. tools on HIP builds
victorapm Mar 5, 2026
99c290a
Switch default to vendor SpGEMM in hypre for HIP builds
victorapm Mar 5, 2026
3e93977
Version header minor fixes
victorapm Mar 5, 2026
86974da
Add hipMemInfo APIs
victorapm Mar 6, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 5 additions & 0 deletions src/coreComponents/common/MemoryInfos.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -50,6 +50,11 @@ MemoryInfos::MemoryInfos( umpire::MemoryResourceTraits::resource_type resourceTy
case umpire::MemoryResourceTraits::resource_type::um:
#if defined( GEOS_USE_CUDA )
cudaMemGetInfo( &m_availableMemory, &m_totalMemory );
#elif defined( GEOS_USE_HIP )
{
hipError_t const err = hipMemGetInfo( &m_availableMemory, &m_totalMemory );
GEOS_WARNING_IF( err != hipSuccess, GEOS_FMT( "HIP error found: {} at {}:{}", hipGetErrorString( err ), __FILE__, __LINE__ ) );
}
#else
GEOS_WARNING( "Unknown device physical memory size getter for this compiler." );
m_physicalMemoryHandled = 0;
Expand Down
4 changes: 4 additions & 0 deletions src/coreComponents/linearAlgebra/interfaces/hypre/HypreInterface.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -55,7 +55,11 @@ void HypreInterface::initialize()
HYPRE_Initialize();
#if GEOS_USE_HYPRE_DEVICE == GEOS_USE_HYPRE_CUDA || GEOS_USE_HYPRE_DEVICE == GEOS_USE_HYPRE_HIP
HYPRE_SetExecutionPolicy( HYPRE_EXEC_DEVICE );
#if GEOS_USE_HYPRE_DEVICE == GEOS_USE_HYPRE_CUDA
HYPRE_SetSpGemmUseVendor( 0 );
#else
HYPRE_SetSpGemmUseVendor( 1 );
#endif
HYPRE_DeviceInitialize();
#endif
HYPRE_SetMemoryLocation( hypre::memoryLocation );
Expand Down
30 changes: 15 additions & 15 deletions src/coreComponents/mainInterface/version.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -72,13 +72,13 @@ void outputVersionInfo()

GEOS_LOG_RANK_0( "GEOS version: " << getVersion() );

GEOS_LOG_RANK_0( " - c++ compiler: " << getCppCompilerIdString() );
GEOS_LOG_RANK_0( " - C++ compiler: " << getCppCompilerIdString() );

std::string const gpuCompilerIdString = getGpuCompilerIdString();
GEOS_LOG_RANK_0_IF( !gpuCompilerIdString.empty(), gpuCompilerIdString );

#if defined(_OPENMP)
GEOS_LOG_RANK_0( " - openmp version: " << _OPENMP );
GEOS_LOG_RANK_0( " - OpenMP version: " << _OPENMP );
#endif

#if defined(GEOS_USE_MPI)
Expand Down Expand Up @@ -107,39 +107,39 @@ void outputVersionInfo()
#endif

#if defined(umpire_VERSION)
GEOS_LOG_RANK_0( " - umpire version: " << STRINGIZE( umpire_VERSION ) );
GEOS_LOG_RANK_0( " - Umpire version: " << STRINGIZE( umpire_VERSION ) );
#endif

#if defined(chai_VERSION)
GEOS_LOG_RANK_0( " - chai version: " << STRINGIZE( chai_VERSION ) );
GEOS_LOG_RANK_0( " - CHAI version: " << STRINGIZE( chai_VERSION ) );
#endif

#if defined(adiak_VERSION)
GEOS_LOG_RANK_0( " - adiak version: " << STRINGIZE( adiak_VERSION ) );
#endif

#if defined(caliper_VERSION)
GEOS_LOG_RANK_0( " - caliper version: " << STRINGIZE( caliper_VERSION ) );
GEOS_LOG_RANK_0( " - Caliper version: " << STRINGIZE( caliper_VERSION ) );
#endif

#if defined(metis_VERSION)
GEOS_LOG_RANK_0( " - METIS version: " << STRINGIZE( metis_VERSION ) );
#endif

#if defined(parmetis_VERSION)
GEOS_LOG_RANK_0( " - PARAMETIS version: " << STRINGIZE( parmetis_VERSION ) );
GEOS_LOG_RANK_0( " - PARMETIS version: " << STRINGIZE( parmetis_VERSION ) );
#endif

#if defined(scotch_VERSION)
GEOS_LOG_RANK_0( " - scotch version: " << STRINGIZE( scotch_VERSION ) );
GEOS_LOG_RANK_0( " - Scotch version: " << STRINGIZE( scotch_VERSION ) );
#endif

#if defined(superlu_dist_VERSION)
GEOS_LOG_RANK_0( " - superlu_dist version: " << STRINGIZE( superlu_dist_VERSION ) );
GEOS_LOG_RANK_0( " - SuperLU_Dist version: " << STRINGIZE( superlu_dist_VERSION ) );
#endif

#if defined(suitesparse_VERSION)
GEOS_LOG_RANK_0( " - suitesparse version: " << STRINGIZE( suitesparse_VERSION ) );
GEOS_LOG_RANK_0( " - SuiteSparse version: " << STRINGIZE( suitesparse_VERSION ) );
#endif

#if defined(hypre_VERSION)
Expand All @@ -151,7 +151,7 @@ void outputVersionInfo()
#endif

#if defined(petsc_VERSION)
GEOS_LOG_RANK_0( " - petsc version: " << STRINGIZE( petsc_VERSION ) );
GEOS_LOG_RANK_0( " - PETSc version: " << STRINGIZE( petsc_VERSION ) );
#endif

#if defined(Python3_VERSION)
Expand All @@ -167,11 +167,11 @@ void outputVersionInfo()
defined(GEOS_USE_HYPRE) && \
( GEOS_USE_HYPRE_DEVICE == GEOS_USE_HYPRE_CPU )
GEOS_LOG_RANK_0( "" );
GEOS_LOG_RANK_0( "**************************************************" );
GEOS_LOG_RANK_0( "* WARNING!!! *" );
GEOS_LOG_RANK_0( "* *" );
GEOS_LOG_RANK_0( "* GEOS has GPU support enabled, but not HYPRE! *" );
GEOS_LOG_RANK_0( "**************************************************" );
GEOS_LOG_RANK_0( "*******************************************************" );
GEOS_LOG_RANK_0( "* WARNING!! WARNING!! WARNING!! *" );
GEOS_LOG_RANK_0( "* *" );
GEOS_LOG_RANK_0( "* GEOS has GPU support enabled, but HYPRE does not! *" );
GEOS_LOG_RANK_0( "*******************************************************" );
GEOS_LOG_RANK_0( "" );
#endif

Expand Down
14 changes: 13 additions & 1 deletion src/coreComponents/mesh/mpiCommunications/CommunicationTools.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1019,11 +1019,15 @@ void CommunicationTools::asyncSendRecv( stdVector< NeighborCommunicator > & neig
parallelDeviceEvents & events )
{
GEOS_MARK_FUNCTION;
#if !defined( GEOS_USE_HIP )
if( onDevice )
{
waitAllDeviceEvents( events );
}

#else
GEOS_UNUSED_VAR( onDevice );
GEOS_UNUSED_VAR( events );
#endif

// could swap this to test and make this function call async as well, only launch the sends/recvs for
// those we've already recv'd sizing for, go back to some usefule compute / launch some other compute, then
Expand Down Expand Up @@ -1162,10 +1166,14 @@ void CommunicationTools::finalizeUnpack( ObjectManagerBase & manager,

// poll mpi for completion then wait 10 nanoseconds 6,000,000,000 times (60 sec timeout)
GEOS_ASYNC_WAIT( 6000000000, 10, asyncUnpack( manager, neighbors, icomm, onDevice, events ) );
#if !defined( GEOS_USE_HIP )
if( onDevice )
{
waitAllDeviceEvents( events );
}
#else
GEOS_UNUSED_VAR( onDevice );
#endif

MpiWrapper::waitAll( icomm.size(),
icomm.mpiSendBufferSizeRequest(),
Expand All @@ -1187,10 +1195,14 @@ void CommunicationTools::finalizeUnpack( MeshLevel & mesh,

// poll mpi for completion then wait 10 nanoseconds 6,000,000,000 times (60 sec timeout)
GEOS_ASYNC_WAIT( 6000000000, 10, asyncUnpack( mesh, neighbors, icomm, onDevice, events, op ) );
#if !defined( GEOS_USE_HIP )
if( onDevice )
{
waitAllDeviceEvents( events );
}
#else
GEOS_UNUSED_VAR( onDevice );
#endif

MpiWrapper::waitAll( icomm.size(),
icomm.mpiSendBufferSizeRequest(),
Expand Down
Loading