cuda_size_to_gpu

std::size_t syten::EnvVars::cuda_size_to_gpu = 1024 * 1024

When deciding in an algorithm whether to move a tensor to the GPU or from the GPU, use this value (in bytes) as the threshold.

Tensors larger than this value are moved to the GPU.

Set via SYTEN_CUDA_THRESHOLD.

Referenced by syten::GenericDenseTensorImpl::addScaled(), syten::Pyten::init_env(), syten::GenericDenseTensorImpl::GenericDenseTensor< rank, Scalar >::maybe_make_cuda(), syten::CudaDenseTensorImpl::prodD(), and syten::Cuda::version().