// Mantid Repository : https://github.com/mantidproject/mantid // // Copyright © 2017 ISIS Rutherford Appleton Laboratory UKRI, // NScD Oak Ridge National Laboratory, European Spallation Source, // Institut Laue - Langevin & CSNS, Institute of High Energy Physics, CAS // SPDX - License - Identifier: GPL - 3.0 + #pragma once #include "MantidAPI/Algorithm.h" #include "MantidAPI/DllConfig.h" namespace Mantid { namespace API { /** Base class for algorithms that treat all spectra independently, i.e., we can trivially parallelize over the spectra without changes. The assumption is that we have one input and one output workspace. The storage mode is just propagated from input to output. When a specific algorithm is determined to be trivially parallel (this is a manual process), the only required change to add MPI support is to inherit from this class instead of Algorithm. Inheriting from DistributedAlgorithm instead of from Algorithm provides the necessary overriden method(s) to allow running an algorithm with MPI. This works under the following conditions: 1. The algorithm must input workspaces with compatible storage modes. StorageMode::Distributed is not compatible with StorageMode::MasterOnly, but all combinations with StorageMode::Cloned are considered compatible. 2. No output files may be written since filenames would clash. Algorithms that do not modify spectra in a workspace may also use this base class to support MPI. For example, modifications of the instrument are handled in a identical manner on all MPI ranks, without requiring changes to the algorithm, other than setting the correct execution mode via the overloads provided by DistributedAlgorithm. @author Simon Heybrock @date 2017 */ class MANTID_API_DLL DistributedAlgorithm : public Algorithm { protected: Parallel::ExecutionMode getParallelExecutionMode( const std::map<std::string, Parallel::StorageMode> &storageModes) const override; }; } // namespace API } // namespace Mantid