Adding CUDA prefix sum unit test.

src/TNL/Containers/Algorithms/CudaPrefixSumKernel.h

@@ -274,7 +274,7 @@ struct CudaPrefixSumKernelLauncher
    /****
     * \brief Starts prefix sum in CUDA.
     *
-    * \tparam Operation operation to be peformed on particular elements - addition usually
+    * \tparam Operation operation to be performed on particular elements - addition usually
     * \tparam VolatileOperation - volatile version of Operation
     * \param size is number of elements to be scanned
     * \param blockSize is CUDA block size
@@ -282,7 +282,7 @@ struct CudaPrefixSumKernelLauncher
     * \param deviceOutput is pointer to resulting array, can be the same as input
     * \param operation is instance of Operation
     * \param volatileOperation is instance of VolatileOperation
-    * \param zero is neutral element for fiven Operation
+    * \param zero is neutral element for given Operation
     */
    template< typename Operation,
              typename VolatileOperation >

src/UnitTests/Containers/Algorithms/CMakeLists.txt

+IF( BUILD_CUDA )
+   CUDA_ADD_EXECUTABLE( CudaPrefixSumTest CudaPrefixSumTest.cu
+                        OPTIONS ${CXX_TESTS_FLAGS} )
+   TARGET_LINK_LIBRARIES( CudaPrefixSumTest ${GTEST_BOTH_LIBRARIES} )
+   ADD_TEST( CudaPrefixSumTest ${EXECUTABLE_OUTPUT_PATH}/CudaPrefixSumTest${CMAKE_EXECUTABLE_SUFFIX} )
+ENDIF()

src/UnitTests/Containers/Algorithms/CudaPrefixSumTest.cu

+/***************************************************************************
+                          CudaPrefixSumTest.cu  -  description
+                             -------------------
+    begin                : Jun 6, 2019
+    copyright            : (C) 2019 by Tomas Oberhuber
+    email                : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+/////
+// NOTE: This test serves mainly for testing CUDA prefix sum when more than
+// one CUDA grid is used. To avoid allocation of extremely large vectors and to
+// speed-up this unit test, we decrease the grid size artificialy.
+
+#pragma once
+
+#ifdef HAVE_GTEST
+#include <limits>
+
+#include <TNL/Experimental/Arithmetics/Quad.h>
+#include <TNL/Containers/Vector.h>
+#include <TNL/Containers/VectorView.h>
+#include "VectorTestSetup.h"
+
+#include "gtest/gtest.h"
+
+using namespace TNL;
+using namespace TNL::Containers;
+using namespace TNL::Containers::Algorithms;
+using namespace TNL::Arithmetics;
+