amd/blis
1
0
Fork 0
mirror of https://github.com/amd/blis.git synced 2026-05-13 02:25:39 +00:00
Code Issues Packages Projects Releases Wiki Activity

AOCL Dynamic tuning for DAXPYV

Browse Source 
- Existing logic is not picking the ideal number
  of threads for some problem sizes.
- Problem size and their corresponding ideal number
  of threads are retuned for daxpy in aocl dynamic.

AMD-Internal: [CPUPL-3484]
Change-Id: Ice874ceef0a1815383f74f1a4b9677677b276af7
This commit is contained in:
Shubham Sharma
2023-05-30 07:25:20 +00:00
parent fa77d0415a
commit 9607f207da
1 changed files with 7 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
frame/base/bli_rntm.c
View File

@@ -1741,16 +1741,18 @@ static void aocl_daxpyv_dynamic
case BLIS_ARCH_ZEN2:
case BLIS_ARCH_ZEN3:
if ( n_elem <= 100 )
if ( n_elem <= 4000 )
*nt_ideal = 1;
else if (n_elem <= 10000)
*nt_ideal = 2;
else if (n_elem <= 250000)
else if (n_elem <= 11000)
*nt_ideal = 4;
else if (n_elem <= 300000)
*nt_ideal = 8;
else if (n_elem <= 750000)
*nt_ideal = 16;
else if (n_elem <= 2000000)
else if (n_elem <= 2600000)
*nt_ideal = 32;
else if (n_elem <= 4000000)
*nt_ideal = 64;
else
// For sizes in this range, AOCL dynamic does not make any change
*nt_ideal = -1;