The ascent_pipeline_example2.cpp and ascent_pipeline_example3.cpp examples are currently not working on Alps.

[jfavre]

[jfavre]

@jgphpc can you add a description of the symptoms seen?

[jgphpc]

./ascent_pipeline_example2 fails with: Failing functor: N4vtkm4cont6detail20ScanExclusiveFunctorINS_7worklet4Clip14PointBatchDataEEE

./ascent_pipeline_example3 fails with: device free failed: cudaErrorIllegalAddress

I'll add more details when the system is back

[jgphpc]

example2

uenv start -v default insitu_ascent/0.9.5:2109123735@daint
cp -a /user-tools/env/default/examples/ascent/tutorial/ascent_intro/cpp .
cd cpp
make ASCENT_DIR=/user-tools/env/default/ ascent_pipeline_example2

L1=/user-tools/linux-neoverse_v2/cray-gtl-8.1.32-25u7zwci35lms4zyrodhf24vlfken7xo/lib
LD_LIBRARY_PATH=$L1:$LD_LIBRARY_PATH ./ascent_pipeline_example2

fails with:

Error: TryExecute encountered an error: parallel_for failed: cudaErrorSymbolNotFound: named symbol not found
	- Failing functor: N4vtkm4cont6detail20ScanExclusiveFunctorINS_7worklet4Clip14PointBatchDataEEE
	- Failing device: Cuda

Error: TryExecute encountered an error: parallel_for failed: cudaErrorSymbolNotFound: named symbol not found
	- Failing functor: N4vtkm4cont6detail20ScanExclusiveFunctorINS_7worklet4Clip13CellBatchDataEEE
	- Failing device: Cuda

terminate called after throwing an instance of 'thrust::THRUST_200700_900_NS::system::system_error'
  what():  device free failed: cudaErrorIllegalAddress: an illegal memory access was encountered
Aborted (core dumped)

example3

uenv start -v default insitu_ascent/0.9.5:2109123735@daint
cp -a /user-tools/env/default/examples/ascent/tutorial/ascent_intro/cpp .
cd cpp
make ASCENT_DIR=/user-tools/env/default/ ascent_pipeline_example3

L1=/user-tools/linux-neoverse_v2/cray-gtl-8.1.32-25u7zwci35lms4zyrodhf24vlfken7xo/lib
LD_LIBRARY_PATH=$L1:$LD_LIBRARY_PATH ./ascent_pipeline_example3

fails with:

Error: TryExecute encountered an error: parallel_for failed: cudaErrorSymbolNotFound: named symbol not found
	- Failing functor: N4vtkm4cont6detail20ScanExclusiveFunctorINS_7worklet4Clip14PointBatchDataEEE
	- Failing device: Cuda

Error: TryExecute encountered an error: parallel_for failed: cudaErrorSymbolNotFound: named symbol not found
	- Failing functor: N4vtkm4cont6detail20ScanExclusiveFunctorINS_7worklet4Clip13CellBatchDataEEE
	- Failing device: Cuda

terminate called after throwing an instance of 'thrust::THRUST_200700_900_NS::system::system_error'
  what():  device free failed: cudaErrorIllegalAddress: an illegal memory access was encountered
Aborted (core dumped)

• Debugger:

1,0,main (ascent_pipeline_example3.cpp:99)
1,0,  ascent::Ascent::execute (ascent.cpp:786)
1,0,    ascent::AscentRuntime::Execute (ascent_main_runtime.cpp:2171)
1,0,      flow::Workspace::execute (flow_workspace.cpp:323)
1,0,        ascent::runtime::filters::VTKHMarchingCubes::execute (ascent_runtime_vtkh_filters.cpp:235)
1,0,          vtkh::Filter::Update (Filter.cpp:61)
1,0,            vtkh::MarchingCubes::DoExecute (MarchingCubes.cpp:162)
1,0,              vtkh::vtkmMarchingCubes::Run (vtkmMarchingCubes.cpp:19)
1,0,                vtkm::filter::Filter::Execute(vtkm::cont::DataSet const&)
1,0,                  vtkm::filter::contour::Contour::DoExecute(vtkm::cont::DataSet const&)
1,0,                    vtkm::filter::Filter::Execute(vtkm::cont::DataSet const&)
1,0,                      vtkm::filter::contour::ContourMarchingCells::DoExecute(vtkm::cont::DataSet const&)
1,0,                        vtkm::cont::DataSet vtkm::filter::contour::ContourMarchingCells::DoExecuteDimension<(unsigned char)3>(vtkm::
cont::DataSet const&)
[...]
-> void vtkm::cont::cuda::internal::TaskStrided1DLaunch

[jfavre]

tested with my own Ascent distribution and also on my laptop. I confirm it breaks. I suppose because the data is not on the GPU