Chris@16: /*
Chris@16:  [auto_generated]
Chris@16:  boost/numeric/odeint/external/mkl/mkl_operations.hpp
Chris@16: 
Chris@16:  [begin_description]
Chris@16:  Wrapper classes for intel math kernel library types.
Chris@16:  Get a free, non-commercial download of MKL at
Chris@16:  http://software.intel.com/en-us/articles/non-commercial-software-download/
Chris@16:  [end_description]
Chris@16: 
Chris@101:  Copyright 2010-2011 Mario Mulansky
Chris@101:  Copyright 2011-2013 Karsten Ahnert
Chris@16: 
Chris@16:  Distributed under the Boost Software License, Version 1.0.
Chris@16:  (See accompanying file LICENSE_1_0.txt or
Chris@16:  copy at http://www.boost.org/LICENSE_1_0.txt)
Chris@16:  */
Chris@16: 
Chris@16: 
Chris@16: #ifndef BOOST_NUMERIC_ODEINT_EXTERNAL_MKL_MKL_OPERATIONS_HPP_INCLUDED
Chris@16: #define BOOST_NUMERIC_ODEINT_EXTERNAL_MKL_MKL_OPERATIONS_HPP_INCLUDED
Chris@16: 
Chris@16: #include <iostream>
Chris@16: 
Chris@16: #include <mkl_cblas.h>
Chris@16: #include <boost/numeric/odeint/algebra/default_operations.hpp>
Chris@16: 
Chris@16: /* exemplary example for writing bindings to the Intel MKL library
Chris@16:  * see test/mkl for how to use mkl with odeint
Chris@16:  * this is a quick and dirty implementation showing the general possibility.
Chris@16:  * It works only with containers based on double and sequential memory allocation.
Chris@16:  */
Chris@16: 
Chris@16: namespace boost {
Chris@16: namespace numeric {
Chris@16: namespace odeint {
Chris@16: 
Chris@16: /* only defined for doubles */
Chris@16: struct mkl_operations
Chris@16: {
Chris@16:     //template< class Fac1 , class Fac2 > struct scale_sum2;
Chris@16: 
Chris@16:     template< class F1 = double , class F2 = F1 >
Chris@16:     struct scale_sum2
Chris@16:     {
Chris@16:         typedef double Fac1;
Chris@16:         typedef double Fac2;
Chris@16:         const Fac1 m_alpha1;
Chris@16:         const Fac2 m_alpha2;
Chris@16: 
Chris@16:         scale_sum2( const Fac1 alpha1 , const Fac2 alpha2 ) : m_alpha1( alpha1 ) , m_alpha2( alpha2 ) { }
Chris@16: 
Chris@16:         template< class T1 , class T2 , class T3 >
Chris@16:         void operator()( T1 &t1 , const T2 &t2 , const T3 &t3) const
Chris@16:         {   // t1 = m_alpha1 * t2 + m_alpha2 * t3;
Chris@16:             // we get Containers that have size() and [i]-access
Chris@16: 
Chris@16:             const int n = t1.size();
Chris@16:             //boost::numeric::odeint::copy( t1 , t3 );
Chris@16:             if( &(t2[0]) != &(t1[0]) )
Chris@16:             {
Chris@16:                 cblas_dcopy( n , &(t2[0]) , 1 , &(t1[0]) , 1 );
Chris@16:             }
Chris@16:             cblas_dscal( n , m_alpha1 , &(t1[0]) , 1 );
Chris@16:             cblas_daxpy( n , m_alpha2 , &(t3[0]) , 1 , &(t1[0]) , 1 );
Chris@16:             //daxpby( &n , &m_alpha2 , &(t3[0]) , &one , &m_alpha1 , &(t1[0]) , &one );
Chris@16:         }
Chris@16:     };
Chris@16: 
Chris@16:     template< class F1 = double , class F2 = F1 , class F3 = F2 >
Chris@16:     struct scale_sum3
Chris@16:     {
Chris@16:         typedef double Fac1;
Chris@16:         typedef double Fac2;
Chris@16:         typedef double Fac3;
Chris@16:         const Fac1 m_alpha1;
Chris@16:         const Fac2 m_alpha2;
Chris@16:         const Fac3 m_alpha3;
Chris@16: 
Chris@16:         scale_sum3( const Fac1 alpha1 , const Fac2 alpha2 , const Fac3 alpha3 )
Chris@16:             : m_alpha1( alpha1 ) , m_alpha2( alpha2 ) , m_alpha3( alpha3 ) { }
Chris@16: 
Chris@16:         template< class T1 , class T2 , class T3 , class T4 >
Chris@16:         void operator()( T1 &t1 , const T2 &t2 , const T3 &t3 , const T4 &t4 ) const
Chris@16:         {   // t1 = m_alpha1 * t2 + m_alpha2 * t3 + m_alpha3 * t4;
Chris@16:             // we get Containers that have size() and [i]-access
Chris@16: 
Chris@16:             const int n = t1.size();
Chris@16:             //boost::numeric::odeint::copy( t1 , t3 );
Chris@16:             if( &(t2[0]) != &(t1[0]) )
Chris@16:             {
Chris@16:                 cblas_dcopy( n , &(t2[0]) , 1 , &(t1[0]) , 1 );
Chris@16:             }
Chris@16:             cblas_dscal( n , m_alpha1 , &(t1[0]) , 1 );
Chris@16:             cblas_daxpy( n , m_alpha2 , &(t3[0]) , 1 , &(t1[0]) , 1 );
Chris@16:             //daxpby( &n , &m_alpha2 , &(t3[0]) , &one , &m_alpha1 , &(t1[0]) , &one );
Chris@16:             cblas_daxpy( n , m_alpha3 , &(t4[0]) , 1 , &(t1[0]) , 1 );
Chris@16:         }
Chris@16:     };
Chris@16: 
Chris@16:     template< class F1 = double , class F2 = F1 , class F3 = F2 , class F4 = F3 >
Chris@16:     struct scale_sum4
Chris@16:     {
Chris@16:         typedef double Fac1;
Chris@16:         typedef double Fac2;
Chris@16:         typedef double Fac3;
Chris@16:         typedef double Fac4;
Chris@16:         const Fac1 m_alpha1;
Chris@16:         const Fac2 m_alpha2;
Chris@16:         const Fac3 m_alpha3;
Chris@16:         const Fac4 m_alpha4;
Chris@16: 
Chris@16:         scale_sum4( const Fac1 alpha1 , const Fac2 alpha2 , const Fac3 alpha3 , const Fac4 alpha4 )
Chris@16:             : m_alpha1( alpha1 ) , m_alpha2( alpha2 ) , m_alpha3( alpha3 ) , m_alpha4( alpha4 ) { }
Chris@16: 
Chris@16:         template< class T1 , class T2 , class T3 , class T4 , class T5 >
Chris@16:         void operator()( T1 &t1 , const T2 &t2 , const T3 &t3 , const T4 &t4 , const T5 &t5 ) const
Chris@16:         {   // t1 = m_alpha1 * t2 + m_alpha2 * t3 + m_alpha3 * t4 + m_alpha4 * t5;
Chris@16:             // we get Containers that have size() and [i]-access
Chris@16: 
Chris@16:             const int n = t1.size();
Chris@16:             //boost::numeric::odeint::copy( t1 , t3 );
Chris@16:             if( &(t2[0]) != &(t1[0]) )
Chris@16:             {
Chris@16:                 cblas_dcopy( n , &(t2[0]) , 1 , &(t1[0]) , 1 );
Chris@16:             }
Chris@16: 
Chris@16:             cblas_dscal( n , m_alpha1 , &(t1[0]) , 1 );
Chris@16:             cblas_daxpy( n , m_alpha2 , &(t3[0]) , 1 , &(t1[0]) , 1 );
Chris@16:             //daxpby( &n , &m_alpha2 , &(t3[0]) , &one , &m_alpha1 , &(t1[0]) , &one );
Chris@16:             cblas_daxpy( n , m_alpha3 , &(t4[0]) , 1 , &(t1[0]) , 1 );
Chris@16:             cblas_daxpy( n , m_alpha4 , &(t5[0]) , 1 , &(t1[0]) , 1 );
Chris@16:         }
Chris@16:     };
Chris@16: 
Chris@16: 
Chris@16:     template< class F1 = double , class F2 = F1 , class F3 = F2 , class F4 = F3 , class F5 = F4 >
Chris@16:     struct scale_sum5
Chris@16:     {
Chris@16:         typedef double Fac1;
Chris@16:         typedef double Fac2;
Chris@16:         typedef double Fac3;
Chris@16:         typedef double Fac4;
Chris@16:         typedef double Fac5;
Chris@16:         const Fac1 m_alpha1;
Chris@16:         const Fac2 m_alpha2;
Chris@16:         const Fac3 m_alpha3;
Chris@16:         const Fac4 m_alpha4;
Chris@16:         const Fac5 m_alpha5;
Chris@16: 
Chris@16:         scale_sum5( const Fac1 alpha1 , const Fac2 alpha2 , const Fac3 alpha3 , const Fac4 alpha4 , const Fac5 alpha5 )
Chris@16:             : m_alpha1( alpha1 ) , m_alpha2( alpha2 ) , m_alpha3( alpha3 ) , m_alpha4( alpha4 ) , m_alpha5( alpha5 ) { }
Chris@16: 
Chris@16:         template< class T1 , class T2 , class T3 , class T4 , class T5 , class T6   >
Chris@16:         void operator()( T1 &t1 , const T2 &t2 , const T3 &t3 , const T4 &t4 , const T5 &t5 , const T6 &t6 ) const
Chris@16:         {   // t1 = m_alpha1 * t2 + m_alpha2 * t3 + m_alpha3 * t4 + m_alpha4 * t5 + m_alpha5 * t6;
Chris@16:             // we get Containers that have size() and [i]-access
Chris@16: 
Chris@16:             const int n = t1.size();
Chris@16:             //boost::numeric::odeint::copy( t1 , t3 );
Chris@16:             if( &(t2[0]) != &(t1[0]) )
Chris@16:             {
Chris@16:                 cblas_dcopy( n , &(t2[0]) , 1 , &(t1[0]) , 1 );
Chris@16:             }
Chris@16: 
Chris@16:             cblas_dscal( n , m_alpha1 , &(t1[0]) , 1 );
Chris@16:             cblas_daxpy( n , m_alpha2 , &(t3[0]) , 1 , &(t1[0]) , 1 );
Chris@16:             //daxpby( &n , &m_alpha2 , &(t3[0]) , &one , &m_alpha1 , &(t1[0]) , &one );
Chris@16:             cblas_daxpy( n , m_alpha3 , &(t4[0]) , 1 , &(t1[0]) , 1 );
Chris@16:             cblas_daxpy( n , m_alpha4 , &(t5[0]) , 1 , &(t1[0]) , 1 );
Chris@16:             cblas_daxpy( n , m_alpha5 , &(t6[0]) , 1 , &(t1[0]) , 1 );
Chris@16:         }
Chris@16:     };
Chris@16: 
Chris@16: };
Chris@16: 
Chris@16: } // odeint
Chris@16: } // numeric
Chris@16: } // boost
Chris@16: 
Chris@16: #endif // BOOST_NUMERIC_ODEINT_EXTERNAL_MKL_MKL_OPERATIONS_HPP_INCLUDED