ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-02-08 23:40:10 +00:00
Code Issues Packages Projects Releases Wiki Activity

Fix "make it work for row sizes that are multiple of 4 on NEON"

Browse Source
This commit is contained in:
Kawrakow
2024-07-22 12:28:18 +02:00
parent 847588cc92
commit 005674cecc
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
iqk_mul_mat.cpp
View File

@@ -4231,7 +4231,7 @@ template <int nrc> struct QF16 final : public QF16Base {
for (int iy = 0; iy < nrc_y; ++iy) y[iy] = (const __fp16 *)(cx + iy*bx);
}
IQK_ALWAYS_INLINE Data load1(int iy, int i) const { return load(y[iy] + k_step*i); }
IQK_ALWAYS_INLINE Data load_tail(int iy, int i) const { return load4(y[iy] + k_step*i); }
IQK_ALWAYS_INLINE Data load_tail(int iy, int i) const { return load4(y[iy] + 4*i); }
const __fp16 * y[nrc_y];
};