cuda_transpose_kernel()

template<Rank rank, typename Scalar >

void syten::CudaDenseTensorImpl::cuda_transpose_kernel	(	CudaConstSpan< Scalar >	inp,
		CudaMutSpan< Scalar >	out,
		ConstSpan< IndexNumber >	perm,
		ConstSpan< Index >	dim,
		Conj	do_conj
	)

Wrapper around the CUDA transpose kernels which sets everything up such that the functions in cuda_transpose_impl_cukrn.h really only have to launch the kernels.

References syten::Cuda::CudaMutSpan< T >::begin(), cuda_transpose_kernel_impl(), cukrn_transpose_max_rank, syten::Cuda::CudaPtr< T >::dev(), syten::RepRegister::dim(), syten::Cuda::CudaPtr< T >::get(), syten::pi, syten::rank(), syten::Cuda::select_device(), syten::Cuda::CudaMutSpan< T >::size(), syten::SpanImpl::ConstSpan< T >::size(), syten::Cuda::CudaStream::str(), syten::Cuda::CudaStream::sync_destroy(), SYTEN_ASSERT, and cukrn_transpose_array::values.

Here is the call graph for this function: