Cyclops Tensor Framework
parallel arithmetic on multidimensional arrays
|
Go to the source code of this file.
Namespaces | |
CTF_int | |
Functions | |
void | CTF_int::padded_reshuffle (int const *sym, distribution const &old_dist, distribution const &new_dist, char *tsr_data, char **tsr_cyclic_data, algstrct const *sr, CommData ord_glb_comm) |
Reshuffle elements using key-value pair read/write. More... | |
int ** | CTF_int::compute_bucket_offsets (distribution const &old_dist, distribution const &new_dist, int const *len, int const *old_phys_edge_len, int const *old_virt_lda, int const *old_offsets, int *const *old_permutation, int const *new_phys_edge_len, int const *new_virt_lda, int forward, int old_virt_np, int new_virt_np, int const *old_virt_edge_len) |
computes offsets for redistribution targets along each edge length More... | |
void | CTF_int::calc_cnt_displs (int const *sym, distribution const &old_dist, distribution const &new_dist, int new_nvirt, int np, int const *old_virt_edge_len, int const *new_virt_lda, int64_t *send_counts, int64_t *recv_counts, int64_t *send_displs, int64_t *recv_displs, CommData ord_glb_comm, int idx_lyr, int *const *bucket_offset) |
assigns keys to an array of values More... | |
double | CTF_int::blres_est_time (int64_t tot_sz, int nv0, int nv1) |
estimates execution time, given this processor sends a receives tot_sz across np procs More... | |
void | CTF_int::block_reshuffle (distribution const &old_dist, distribution const &new_dist, char *tsr_data, char *&tsr_cyclic_data, algstrct const *sr, CommData glb_comm) |
Reshuffle elements by block given the global phases stay the same. More... | |
int | CTF_int::can_block_reshuffle (int order, int const *old_phase, mapping const *map) |
determines if tensor can be permuted by block More... | |