Usage of Implementation details in lorenz-parameters.cu

[derdotte]

Hello, it came to light that in line 85 of the lorenz-parameters example

odeint-v2/examples/thrust/lorenz_parameters.cu

Lines 70 to 90 in db8b39a

	void operator()( const State &x , Deriv &dxdt , value_type t ) const
	{
	thrust::for_each(
	thrust::make_zip_iterator( thrust::make_tuple(
	boost::begin( x ) ,
	boost::begin( x ) + m_N ,
	boost::begin( x ) + 2 * m_N ,
	m_beta.begin() ,
	boost::begin( dxdt ) ,
	boost::begin( dxdt ) + m_N ,
	boost::begin( dxdt ) + 2 * m_N ) ) ,
	thrust::make_zip_iterator( thrust::make_tuple(
	boost::begin( x ) + m_N ,
	boost::begin( x ) + 2 * m_N ,
	boost::begin( x ) + 3 * m_N ,
	m_beta.begin() ,
	boost::begin( dxdt ) + m_N ,
	boost::begin( dxdt ) + 2 * m_N ,
	boost::begin( dxdt ) + 3 * m_N ) ) ,
	lorenz_functor() );
	}

and in line 152

odeint-v2/examples/thrust/lorenz_parameters.cu

Lines 125 to 162 in db8b39a

	{
	thrust::for_each(
	thrust::make_zip_iterator( thrust::make_tuple(
	thrust::make_zip_iterator( thrust::make_tuple(
	boost::begin( x ) ,
	boost::begin( x ) + m_N ,
	boost::begin( x ) + 2 * m_N ,
	boost::begin( x ) + 3 * m_N ,
	boost::begin( x ) + 4 * m_N ,
	boost::begin( x ) + 5 * m_N ) ) ,
	m_beta.begin() ,
	thrust::make_zip_iterator( thrust::make_tuple(
	boost::begin( dxdt ) ,
	boost::begin( dxdt ) + m_N ,
	boost::begin( dxdt ) + 2 * m_N ,
	boost::begin( dxdt ) + 3 * m_N ,
	boost::begin( dxdt ) + 4 * m_N ,
	boost::begin( dxdt ) + 5 * m_N ) )
	) ) ,
	thrust::make_zip_iterator( thrust::make_tuple(
	thrust::make_zip_iterator( thrust::make_tuple(
	boost::begin( x ) + m_N ,
	boost::begin( x ) + 2 * m_N ,
	boost::begin( x ) + 3 * m_N ,
	boost::begin( x ) + 4 * m_N ,
	boost::begin( x ) + 5 * m_N ,
	boost::begin( x ) + 6 * m_N ) ) ,
	m_beta.begin() ,
	thrust::make_zip_iterator( thrust::make_tuple(
	boost::begin( dxdt ) + m_N ,
	boost::begin( dxdt ) + 2 * m_N ,
	boost::begin( dxdt ) + 3 * m_N ,
	boost::begin( dxdt ) + 4 * m_N ,
	boost::begin( dxdt ) + 5 * m_N ,
	boost::begin( dxdt ) + 6 * m_N ) )
	) ) ,
	lorenz_perturbation_functor() );
	}

implementation details of the thrust::make_zip_iterator are being used. Specifically this code should not work under the normal behavior of zip_iterators, as the shortest element in the iterator tuple is m_beta. This however still works because of how the implementation of thrust`s zip_iterator works.
Specifically referencing the answer to this question https://stackoverflow.com/a/75212854/20624687 .

This should get fixed as people use this when learning about the capabilities of boost and thrust using this website:
https://www.boost.org/doc/libs/1_81_0/libs/numeric/odeint/doc/html/boost_numeric_odeint/tutorial/using_cuda__or_openmp__tbb_______via_thrust.html