|
| template<typename T, bool IS_FUNC_AB> |
| __device__ __inline__ void | grid_gpu_backend::block_to_cab (const kernel_params ¶ms, const smem_task< T > &task, T *cab) |
| | Decontracts the subblock, going from spherical to cartesian harmonics.
|
| template<typename T, bool IS_FUNC_AB> |
| __global__ | grid_gpu_backend::__launch_bounds__ (64) void calculate_coefficients_two_steps(const kernel_params dev_ |
| | Same as calculate_coefficients, with the spherical to cartesian transformation of block_to_cab done in two steps,.
|
| | grid_gpu_backend::if (number_of_tasks==0) return |
| | grid_gpu_backend::for (int tk=0;tk< number_of_tasks;tk++) |
| template<bool distributed__, bool axis_aligned_> |
| __global__ | grid_gpu_backend::__launch_bounds__ (64) void collocate_separable_kernel(const kernel_params dev_ |
| | grid_gpu_backend::if (glb_task.skip_task) return |
| | grid_gpu_backend::fill_smem_task_reduced (dev_, dev_.first_task+block_index(), task) |
| | grid_gpu_backend::if (tid==0) |
| | grid_gpu_backend::__syncthreads () |
| | grid_gpu_backend::if (axis_aligned_) |
| static size_t | grid_gpu_backend::collocate_separable_smem_size (const int mc, const int lp_max, const int erows) |
| static int | grid_gpu_backend::collocate_separable_erows (const int mc, const int lp_max, const size_t smem_max) |