Přidán parametr build scriptů --have-mic=yes, je podmíněný použitím icpc a vypíná CUDU (nastaví cuda=no než se začna cuda kontrolovat)

Implementace Array Operations pro MIC, vedoucí k zpřehlednění kódu -> kód použitý pro srovnávací test MIC-krakow, MIC-sharedpoiters, MIC-ArrayOperations

Nově zavedená stará implementace, lze přeložit s MIC i s CUDA, snad... Stále využívá Specializaci Array pro MIC.

Conflicts:
	build
	src/TNL/Assert.h
	src/TNL/Containers/Algorithms/ArrayOperations.h
	src/TNL/Containers/Algorithms/VectorOperations.h
	src/TNL/Containers/Array.h
	src/TNL/Containers/SharedArray.h
	src/TNL/Containers/SharedVector.h
	src/TNL/Containers/StaticArray.h
	src/TNL/File.h
	src/TNL/File_impl.h
	src/TNL/Matrices/CSR_impl.h
	src/TNL/Matrices/MatrixSetter.h
	src/TNL/Meshes/GridDetails/GridEntity_impl.h
	src/TNL/Meshes/GridDetails/NeighbourGridEntityGetter1D_impl.h
	src/TNL/Meshes/GridDetails/NeighbourGridEntityGetter2D_impl.h
	src/TNL/Meshes/GridDetails/NeighbourGridEntityGetter3D_impl.h
	src/TNL/Meshes/GridEntity.h
	src/TNL/Solvers/ODE/Euler_impl.h
	src/core/arrays/tnlArray_impl.h
	src/core/arrays/tnlSharedArray_impl.cpp
	src/core/arrays/tnlSharedArray_impl.h
	src/core/arrays/tnlStaticArray1D_impl.h
	src/core/arrays/tnlStaticArray2D_impl.h
	src/core/arrays/tnlStaticArray3D_impl.h
	src/core/arrays/tnlStaticArray_impl.h
	src/core/mfuncs.h
	src/core/tnlCuda.h
	src/core/tnlDevice.h
	src/core/tnlObject.h
	src/core/vectors/tnlSharedVector_impl.h
	src/core/vectors/tnlStaticVector.h
	src/core/vectors/tnlStaticVector1D_impl.h
	src/core/vectors/tnlStaticVector2D_impl.h
	src/core/vectors/tnlStaticVector3D_impl.h
	src/core/vectors/tnlStaticVector_impl.h
	src/core/vectors/tnlVectorOperationsHost_impl.h
	src/functions/tnlFunctionEvaluator_impl.h
	src/functions/tnlMeshFunction.h
	src/functions/tnlMeshFunctionEvaluator_impl.h
	src/functions/tnlMeshFunction_impl.h
	src/matrices/tnlMatrixSetter_impl.h
	src/mesh/grids/CMakeLists.txt
	src/mesh/grids/tnlGridTraverser.h
	src/mesh/grids/tnlGridTraverserCUDA_impl.h
	src/solvers/pde/tnlBoundaryConditionsSetter.h
	src/solvers/pde/tnlBoundaryConditionsSetter_impl.h
	src/solvers/pde/tnlExplicitUpdater.h
	src/solvers/pde/tnlExplicitUpdater_impl.h
	src/solvers/pde/tnlLinearSystemAssembler_impl.h
	src/solvers/tnlBuildConfigTags.h
	src/solvers/tnlSolverConfig_impl.h
	src/solvers/tnlSolverInitiator_impl.h
	tests/benchmarks/heat-equation-benchmark/tnlTestNeighbourGridEntitiesStorage.h
	tests/unit-tests/core/arrays/CMakeLists.txt
	tests/unit-tests/core/vectors/CMakeLists.txt

Apparently there is an abs function defined in <cstdlib>, which works
only on integers. We want the std::abs overloads defined in <cmath>,
which are used by TNL::abs.

clang rocks!

http://stackoverflow.com/questions/29136607/c-what-is-wrong-with-abs#29136774

The methods did not work with zero array size, so extra checks were
added.

Curiously clang complained about it, but nvcc didn't.

The original algorithm used space as one of the separators between types
in template parameter lists, so as a result "A< short int >" was parsed
into a list of ("A", "short", "int").

This prevented scaling of GMRES on small grids (up to 240x240)!

We must pass the CUDA_HOST_COMPILER variable manually to use the
compiler configured by the user.