MPI refactoring: removed MpiCommunicator from solvers: Merson, GMRES, Linear/Traits.h (5e7005a6) · Commits · TNL / tnl-dev

src/TNL/Solvers/Linear/GMRES.h

+0 −3

Original line number	Diff line number	Diff line
		@@ -23,10 +23,7 @@ class GMRES
		: public LinearSolver< Matrix >
		{
		using Base = LinearSolver< Matrix >;

		// compatibility shortcuts
		using Traits = Linear::Traits< Matrix >;
		using CommunicatorType = typename Traits::CommunicatorType;

		public:
		using RealType = typename Base::RealType;

src/TNL/Solvers/Linear/GMRES_impl.h

+5 −5

Original line number	Diff line number	Diff line
		@@ -510,7 +510,7 @@ hauseholder_generate( const int i,
		norm_yi_squared = 2 * (normz * normz + std::fabs( y_ii ) * normz);
		}
		// no-op if the problem is not distributed
		CommunicatorType::Bcast( &norm_yi_squared, 1, 0, Traits::getCommunicationGroup( *this->matrix ) );
		MPI::Bcast( &norm_yi_squared, 1, 0, Traits::getCommunicationGroup( *this->matrix ) );

		// XXX: normalization is slower, but more stable
		// y_i *= 1.0 / std::sqrt( norm_yi_squared );
		@@ -534,7 +534,7 @@ hauseholder_generate( const int i,
		i,
		aux );
		// no-op if the problem is not distributed
		CommunicatorType::Allreduce( aux, i, MPI_SUM, Traits::getCommunicationGroup( *this->matrix ) );
		MPI::Allreduce( aux, i, MPI_SUM, Traits::getCommunicationGroup( *this->matrix ) );

		// [T_i]_{0..i-1} = - T_{i-1} * t_i * aux
		for( int k = 0; k < i; k++ ) {
		@@ -559,7 +559,7 @@ hauseholder_apply_trunc( HostView out,
		HostView YL_i( &YL[ i * (restarting_max + 1) ], restarting_max + 1 );
		Algorithms::MultiDeviceMemoryOperations< Devices::Host, DeviceType >::copy( YL_i.getData(), Traits::getLocalView( y_i ).getData(), YL_i.getSize() );
		// no-op if the problem is not distributed
		CommunicatorType::Bcast( YL_i.getData(), YL_i.getSize(), 0, Traits::getCommunicationGroup( *this->matrix ) );
		MPI::Bcast( YL_i.getData(), YL_i.getSize(), 0, Traits::getCommunicationGroup( *this->matrix ) );

		// NOTE: aux = t_i * (y_i, z) = 1 since t_i = 2 / \|\|y_i\|\|^2 and
		// (y_i, z) = \|\|z_trunc\|\|^2 + \|z_i\| \|\|z_trunc\|\| = \|\|y_i\|\|^2 / 2
		@@ -579,7 +579,7 @@ hauseholder_apply_trunc( HostView out,
		}

		// no-op if the problem is not distributed
		CommunicatorType::Bcast( out.getData(), i + 1, 0, Traits::getCommunicationGroup( *this->matrix ) );
		MPI::Bcast( out.getData(), i + 1, 0, Traits::getCommunicationGroup( *this->matrix ) );
		}

		template< typename Matrix >
		@@ -634,7 +634,7 @@ hauseholder_cwy_transposed( VectorViewType z,
		i + 1,
		aux );
		// no-op if the problem is not distributed
		Traits::CommunicatorType::Allreduce( aux, i + 1, MPI_SUM, Traits::getCommunicationGroup( *this->matrix ) );
		MPI::Allreduce( aux, i + 1, MPI_SUM, Traits::getCommunicationGroup( *this->matrix ) );

		// aux = T_i^T * aux
		// Note that T_i^T is lower triangular, so we can overwrite the aux vector with the result in place

src/TNL/Solvers/Linear/Traits.h

+3 −7

Original line number	Diff line number	Diff line
		@@ -12,7 +12,7 @@

		#pragma once

		#include <TNL/Communicators/MpiCommunicator.h>
		#include <TNL/MPI/Wrappers.h>
		#include <TNL/Containers/Vector.h>
		#include <TNL/Containers/VectorView.h>
		#include <TNL/Containers/DistributedVector.h>
		@@ -26,8 +26,6 @@ namespace Linear {
		template< typename Matrix >
		struct Traits
		{
		using CommunicatorType = Communicators::MpiCommunicator;

		using VectorType = Containers::Vector
		< typename Matrix::RealType,
		typename Matrix::DeviceType,
		@@ -51,7 +49,7 @@ struct Traits
		static ConstLocalViewType getConstLocalView( ConstVectorViewType v ) { return v; }
		static LocalViewType getLocalView( VectorViewType v ) { return v; }

		static typename CommunicatorType::CommunicationGroup getCommunicationGroup( const Matrix& m ) { return CommunicatorType::AllGroup; }
		static MPI_Comm getCommunicationGroup( const Matrix& m ) { return MPI::AllGroup(); }
		static void startSynchronization( VectorViewType v ) {}
		static void waitForSynchronization( VectorViewType v ) {}
		};
		@@ -59,8 +57,6 @@ struct Traits
		template< typename Matrix, typename Communicator >
		struct Traits< Matrices::DistributedMatrix< Matrix, Communicator > >
		{
		using CommunicatorType = Communicator;

		using VectorType = Containers::DistributedVector
		< typename Matrix::RealType,
		typename Matrix::DeviceType,
		@@ -96,7 +92,7 @@ struct Traits< Matrices::DistributedMatrix< Matrix, Communicator > >
		static ConstLocalViewType getConstLocalView( ConstVectorViewType v ) { return v.getConstLocalView(); }
		static LocalViewType getLocalView( VectorViewType v ) { return v.getLocalView(); }

		static typename CommunicatorType::CommunicationGroup getCommunicationGroup( const Matrices::DistributedMatrix< Matrix, Communicator >& m ) { return m.getCommunicationGroup(); }
		static MPI_Comm getCommunicationGroup( const Matrices::DistributedMatrix< Matrix, Communicator >& m ) { return m.getCommunicationGroup(); }
		static void startSynchronization( VectorViewType v ) { v.startSynchronization(); }
		static void waitForSynchronization( VectorViewType v ) { v.waitForSynchronization(); }
		};

src/TNL/Solvers/ODE/Merson_impl.h

+5 −5

Original line number	Diff line number	Diff line
		@@ -13,7 +13,7 @@
		#include <TNL/Devices/Host.h>
		#include <TNL/Devices/Cuda.h>
		#include <TNL/Config/ParameterContainer.h>
		#include <TNL/Communicators/MpiCommunicator.h>
		#include <TNL/MPI/Wrappers.h>

		#include "Merson.h"

		@@ -156,7 +156,7 @@ bool Merson< Problem, SolverMonitor >::solve( DofVectorPointer& _u )
		{
		const RealType localError =
		max( currentTau / 3.0 * abs( 0.2 * k1 -0.9 * k3 + 0.8 * k4 -0.1 * k5 ) );
		Problem::CommunicatorType::Allreduce( &localError, &error, 1, MPI_MAX, Problem::CommunicatorType::AllGroup );
		MPI::Allreduce( &localError, &error, 1, MPI_MAX, MPI::AllGroup() );
		}

		if( adaptivity == 0.0 \|\| error < adaptivity )