addScaled() [2/2]

template<Rank rank, typename Scalar >

void syten::GenericDenseTensorImpl::addScaled ( GenericDenseTensor< rank, Scalar > &  a, GenericDenseTensor< rank, Scalar > const &  b, typename IdentityType< Scalar >::type const  factor, EliminateZeros const  ezeros  )

inline

\( a_i = a_i + f b_i \) for generic dense tensors.

Will fail at runtime if inappropriate dense tensors are supplied.

Parameters

a	the dense tensor to be updated during the call
b	the second dense tensor, to be scaled by the factor and added to a
factor	the scaling factor for the second dense tensor
ezeros	if EliminateZeros::Yes, will set small resulting elements to zero

Remarks

If A is a CUDA tensor, smaller than the threshold size and B is not a CUDA tensor, A is turned into a standard tensor and the default functions are used. If A is a CUDA tensor and either larger than the threshold size or B is also a CUDA tensor, then A is kept as a CUDA tensor and the default functions are used.

If A is not a CUDA tensor but B is a CUDA tensor and smaller than the threshold size, then A and B are made standard tensors and the standard function is used. If A is not a CUDA tensor but B is a CUDA tensor and larger than the threshold size, A is made a CUDA tensor and the CUDA functions are used.

Otherwise, A is made a standard tensor and the default functions are used.

References addScaled(), syten::GenericDenseTensorImpl::GenericDenseTensor< rank, Scalar >::cuda(), syten::EnvVars::cuda_size_to_gpu, syten::GenericDenseTensorImpl::GenericDenseTensor< rank, Scalar >::is_cuda(), syten::GenericDenseTensorImpl::GenericDenseTensor< rank, Scalar >::make_cuda(), syten::GenericDenseTensorImpl::GenericDenseTensor< rank, Scalar >::make_standard(), syten::GenericDenseTensorImpl::GenericDenseTensor< rank, Scalar >::size(), and syten::GenericDenseTensorImpl::GenericDenseTensor< rank, Scalar >::standard().

Here is the call graph for this function: