18b5be92d6
When you try to call the `Reduce` operation in the CUDA device adapter with a sufficently complex interator type, you get a compile error that says `error: cannot pass an argument with a user-provided copy-constructor to a device-side kernel launch`. This appears to be a bug in either nvcc or Thrust. I believe it is related to the following reported issues: * https://github.com/thrust/thrust/issues/928 * https://github.com/thrust/thrust/issues/1044 Work around this problem by making a special condition for calling `Reduce` with an `ArrayHandleMultiplexer` that calls the generic algorithm in `DeviceAdapterAlgorithmGeneral` instead of the algorithm in Thrust.
148 lines
5.2 KiB
C++
148 lines
5.2 KiB
C++
//============================================================================
|
|
// Copyright (c) Kitware, Inc.
|
|
// All rights reserved.
|
|
// See LICENSE.txt for details.
|
|
//
|
|
// This software is distributed WITHOUT ANY WARRANTY; without even
|
|
// the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR
|
|
// PURPOSE. See the above copyright notice for more information.
|
|
//============================================================================
|
|
#ifndef vtk_m_cont_testing_TestingArrayHandleMultiplexer_h
|
|
#define vtk_m_cont_testing_TestingArrayHandleMultiplexer_h
|
|
|
|
#include <vtkm/cont/ArrayCopy.h>
|
|
#include <vtkm/cont/ArrayHandleConstant.h>
|
|
#include <vtkm/cont/ArrayHandleCounting.h>
|
|
#include <vtkm/cont/ArrayHandleImplicit.h>
|
|
#include <vtkm/cont/ArrayHandleMultiplexer.h>
|
|
#include <vtkm/cont/ArrayHandleUniformPointCoordinates.h>
|
|
#include <vtkm/cont/DeviceAdapterAlgorithm.h>
|
|
#include <vtkm/cont/RuntimeDeviceTracker.h>
|
|
|
|
#include <vtkm/BinaryOperators.h>
|
|
|
|
#include <vtkm/cont/testing/Testing.h>
|
|
|
|
namespace vtkm
|
|
{
|
|
namespace cont
|
|
{
|
|
namespace testing
|
|
{
|
|
|
|
template <typename T>
|
|
struct TestValueFunctor
|
|
{
|
|
VTKM_EXEC_CONT T operator()(vtkm::Id index) const { return TestValue(index, T()); }
|
|
};
|
|
|
|
template <typename DeviceAdapter>
|
|
class TestingArrayHandleMultiplexer
|
|
{
|
|
static constexpr vtkm::Id ARRAY_SIZE = 10;
|
|
|
|
template <typename... Ts0, typename... Ts1>
|
|
static void CheckArray(const vtkm::cont::ArrayHandleMultiplexer<Ts0...>& multiplexerArray,
|
|
const vtkm::cont::ArrayHandle<Ts1...>& expectedArray)
|
|
{
|
|
using T = typename std::remove_reference<decltype(multiplexerArray)>::type::ValueType;
|
|
|
|
vtkm::cont::printSummary_ArrayHandle(multiplexerArray, std::cout);
|
|
VTKM_TEST_ASSERT(test_equal_portals(multiplexerArray.ReadPortal(), expectedArray.ReadPortal()),
|
|
"Multiplexer array gave wrong result in control environment");
|
|
|
|
vtkm::cont::ArrayHandle<T> copy;
|
|
vtkm::cont::ArrayCopy(multiplexerArray, copy);
|
|
VTKM_TEST_ASSERT(test_equal_portals(copy.ReadPortal(), expectedArray.ReadPortal()),
|
|
"Multiplexer did not copy correctly in execution environment");
|
|
}
|
|
|
|
static void BasicSwitch()
|
|
{
|
|
std::cout << std::endl << "--- Basic switch" << std::endl;
|
|
|
|
using ValueType = vtkm::FloatDefault;
|
|
|
|
using ArrayType1 = vtkm::cont::ArrayHandleConstant<ValueType>;
|
|
ArrayType1 array1(TestValue(0, vtkm::FloatDefault{}), ARRAY_SIZE);
|
|
|
|
using ArrayType2 = vtkm::cont::ArrayHandleCounting<ValueType>;
|
|
ArrayType2 array2(TestValue(1, vtkm::FloatDefault{}), 1.0f, ARRAY_SIZE);
|
|
|
|
auto array3 = vtkm::cont::make_ArrayHandleImplicit(TestValueFunctor<ValueType>{}, ARRAY_SIZE);
|
|
using ArrayType3 = decltype(array3);
|
|
|
|
vtkm::cont::ArrayHandleMultiplexer<ArrayType1, ArrayType2, ArrayType3> multiplexer;
|
|
|
|
std::cout << "Check array1" << std::endl;
|
|
multiplexer = array1;
|
|
CheckArray(multiplexer, array1);
|
|
|
|
std::cout << "Check array2" << std::endl;
|
|
multiplexer = array2;
|
|
CheckArray(multiplexer, array2);
|
|
|
|
std::cout << "Check array3" << std::endl;
|
|
multiplexer = array3;
|
|
CheckArray(multiplexer, array3);
|
|
}
|
|
|
|
static void Reduce()
|
|
{
|
|
// Regression test for an issue with compiling ArrayHandleMultiplexer with the thrust reduce
|
|
// algorithm on CUDA. Most likely related to:
|
|
// https://github.com/thrust/thrust/issues/928
|
|
// https://github.com/thrust/thrust/issues/1044
|
|
std::cout << std::endl << "--- Reduce" << std::endl;
|
|
|
|
using ValueType = vtkm::Vec3f;
|
|
using MultiplexerType = vtkm::cont::ArrayHandleMultiplexer<
|
|
vtkm::cont::ArrayHandleConstant<ValueType>,
|
|
vtkm::cont::ArrayHandleCounting<ValueType>,
|
|
vtkm::cont::ArrayHandle<ValueType>,
|
|
vtkm::cont::ArrayHandleUniformPointCoordinates,
|
|
vtkm::cont::ArrayHandleCartesianProduct<vtkm::cont::ArrayHandle<vtkm::FloatDefault>,
|
|
vtkm::cont::ArrayHandle<vtkm::FloatDefault>,
|
|
vtkm::cont::ArrayHandle<vtkm::FloatDefault>>>;
|
|
|
|
MultiplexerType multiplexer =
|
|
vtkm::cont::ArrayHandleCounting<ValueType>(vtkm::Vec3f(1), vtkm::Vec3f(1), ARRAY_SIZE);
|
|
|
|
using Algorithm = vtkm::cont::DeviceAdapterAlgorithm<DeviceAdapter>;
|
|
|
|
{
|
|
std::cout << "Basic Reduce" << std::endl;
|
|
ValueType result = Algorithm::Reduce(multiplexer, ValueType(0.0));
|
|
VTKM_TEST_ASSERT(test_equal(result, ValueType(0.5 * (ARRAY_SIZE * (ARRAY_SIZE + 1)))));
|
|
}
|
|
|
|
{
|
|
std::cout << "Reduce with custom operator" << std::endl;
|
|
vtkm::Vec<ValueType, 2> initial(ValueType(10000), ValueType(0));
|
|
vtkm::Vec<ValueType, 2> result =
|
|
Algorithm::Reduce(multiplexer, initial, vtkm::MinAndMax<ValueType>{});
|
|
VTKM_TEST_ASSERT(test_equal(result[0], ValueType(1)));
|
|
VTKM_TEST_ASSERT(
|
|
test_equal(result[1], ValueType(static_cast<vtkm::FloatDefault>(ARRAY_SIZE))));
|
|
}
|
|
}
|
|
|
|
static void TestAll()
|
|
{
|
|
BasicSwitch();
|
|
Reduce();
|
|
}
|
|
|
|
public:
|
|
static int Run(int argc, char* argv[])
|
|
{
|
|
vtkm::cont::ScopedRuntimeDeviceTracker device(DeviceAdapter{});
|
|
return vtkm::cont::testing::Testing::Run(TestAll, argc, argv);
|
|
}
|
|
};
|
|
}
|
|
}
|
|
} // namespace vtkm::cont::testing
|
|
|
|
#endif //vtk_m_cont_testing_TestingArrayHandleMultiplexer_h
|