|
| subroutine, public | dbt_block::dbt_iterator_start (iterator, tensor) |
| | Generalization of block_iterator_start for tensors.
|
| subroutine, public | dbt_block::dbt_iterator_stop (iterator) |
| | Generalization of block_iterator_stop for tensors.
|
| pure integer function, public | dbt_block::ndims_iterator (iterator) |
| | Number of dimensions.
|
| subroutine, public | dbt_block::dbt_iterator_next_block (iterator, ind_nd, blk_size, blk_offset) |
| | iterate over nd blocks of an nd rank tensor, index only (blocks must be retrieved by calling dbt_get_block on tensor).
|
| integer function, public | dbt_block::dbt_iterator_num_blocks (iterator) |
| | Generalization of block_iterator_num_blocks for tensors.
|
| logical function, public | dbt_block::dbt_iterator_blocks_left (iterator) |
| | Generalization of block_iterator_blocks_left for tensors.
|
| subroutine, public | dbt_block::dbt_get_block_pattern (tensor, dimensions, pattern) |
| | Collect the global stored-block pattern projected onto two tensor dimensions. Called outside an OpenMP team; iteration and reduction are internal.
|
| subroutine, public | dbt_block::destroy_block (block) |
| elemental logical function, public | dbt_block::checker_tr (row, column) |
| | Determines whether a transpose must be applied.
|
| subroutine, public | dbt_block::dbt_get_block_data (tensor, ind, sizes, data, found) |
| | Get a block into the given contiguous data (arbitrary tensor rank), without allocating any intermediate block.
|
| subroutine, public | dbt_block::dbt_put_block_data (tensor, ind, sizes, data, summation) |
| | Put a block from the given contiguous data (arbitrary tensor rank), without allocating any intermediate block.
|
| subroutine, public | dbt_block::dbt_copy_block (tensor_in, tensor_out, ind, summation, ind_out) |
| | Copy a local block between tensors with the same index mapping and matching block sizes, without an intermediate copy or reshape. Both blocks must belong to the calling rank, and the tensors must have separate block storage.
|