amd/blis
1
0
Fork 0
mirror of https://github.com/amd/blis.git synced 2026-05-11 09:39:59 +00:00
Code Issues Packages Projects Releases Wiki Activity

Fixed a bug with get_range_weighted would return end = 0 for small problem sizes

Browse Source
This commit is contained in:
Tyler Michael Smith
2015-02-19 22:51:11 -06:00
parent 518a1756cc
commit af32e3a608
1 changed files with 1 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
frame/base/bli_threading.c
View File

@@ -146,11 +146,11 @@ void bli_get_range_weighted( void* thr, dim_t all_start, dim_t all_end, dim_t bl
*start = 0;
*end = all_end - all_start;
double num = size*size / (double) n_way;
if( forward ) {
dim_t curr_caucus = n_way - 1;
dim_t len = 0;
dim_t num = size*size / n_way; // 2xArea per thread?
while(1){
dim_t width = ceil(sqrt( len*len + num )) - len; // The width of the current caucus
width = (width % block_factor == 0) ? width : width + block_factor - (width % block_factor);
@@ -167,7 +167,6 @@ void bli_get_range_weighted( void* thr, dim_t all_start, dim_t all_end, dim_t bl
}
}
else{
dim_t num = size*size / n_way;
while(1){
dim_t width = ceil(sqrt(*start * *start + num)) - *start;
width = (width % block_factor == 0) ? width : width + block_factor - (width % block_factor);