Skip to content

GitLab

Commit e4bffc94 authored by Adam Osewski's avatar Adam Osewski
Browse files

Add comment depicting B2C mapping pattern.

parent 95a83c6e
Hide whitespace changes
Inline Side-by-side
Showing with 44 additions and 0 deletions
include/ck/tensor_operation/gpu/grid/block_to_ctile_map.hpp
View file @ e4bffc94
...@@ -154,6 +154,50 @@ struct BlockToCTileMap_M00_N0_M01Adapt ...@@ -154,6 +154,50 @@ struct BlockToCTileMap_M00_N0_M01Adapt
index_t idx_M01 = idx_M0 % M01_; index_t idx_M01 = idx_M0 % M01_;
index_t idx_N0_M01_local = idx_N0 + idx_M01 * N0; index_t idx_N0_M01_local = idx_N0 + idx_M01 * N0;
/**
* idxN0
*
* |< mtx N >|
*
* NPerBlock NPerBlock NPerBlock NPerBlock
* N_0 N_1 N_2 N_3
* - |-----------|-----------|-----------|-----|-----|-
* ^ | - - 0 |/----> 2 | | | |
* | | | / | | | | | M_0 MPerBlock
* | M | /| | | | | |
* |-0---|---/-|-----|-----|-----------|-----|-----|-
* | 1 | / | | | blockid | | |
* idxM0 | | | / | V | 5 | | | M_1 MPerBlock
* | - V 1 | - 3 | | | |
* |-----------|-----------|-----------|-----|-----|-
* mtx M | | | | | |
* | | | | | | M_2 MPerBlock
* | | | | | |
* |-----------|-----------|-----------|-----|-----|-
* | | | | | |
* | | | | | | M_3 MPerBlock
* | | | | | |
* |-----------|-----------|-----------|-----|-----|-
* V | | | | | |
* - |-----------|-----------|-----------|-----|-----|- M_4 MPerBlock
* | | | | | |
* |-----------|-----------|-----------|-----|-----|-
* Example:
* assume:
* M0 = 5
* N0 = 4
* block_1d_id = 5
* M01 = 2
*
* idx_N0 = 1
* idx_M0 = 1
* M01_adapt = 2
* idx_M00 = 0
* idx_M01 = 1
* idx_N0_M01_local = 5
* output {1, 2}
*/
return make_tuple(idx_N0_M01_local % M01_adapt + idx_M00 * M01_, return make_tuple(idx_N0_M01_local % M01_adapt + idx_M00 * M01_,
idx_N0_M01_local / M01_adapt); idx_N0_M01_local / M01_adapt);
} }
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment