CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-30 08:42:00 +00:00
Code Issues Packages Projects Releases Wiki Activity

ggml: fix loongarch quantize_row_q8_1 error (#14827)

Browse Source
This commit is contained in:
lixing-star
2025-07-23 14:39:51 +08:00
committed by GitHub
parent 14c28dfc50
commit 6c88b3bb25
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
ggml/src/ggml-cpu/arch/loongarch/quants.c
View File

@@ -544,7 +544,7 @@ void quantize_row_q8_1(const float * GGML_RESTRICT x, void * GGML_RESTRICT vy, i
__m128 max4 = __lsx_vfmax_s( lasx_extractf128( max_abs, 1 ), lasx_extractf128( max_abs, 0) ); __m128 max4 = __lsx_vfmax_s( lasx_extractf128( max_abs, 1 ), lasx_extractf128( max_abs, 0) );
max4 = __lsx_vfmax_s( max4, (__m128)__lsx_vpickod_d((__m128i) max4, (__m128i)max4 ) ); max4 = __lsx_vfmax_s( max4, (__m128)__lsx_vpickod_d((__m128i) max4, (__m128i)max4 ) );
__m128 tmp = max4; __m128 tmp = max4;
max4 = __lsx_vfmax_s( max4, (__m128)__lsx_vextrins_w((__m128i)tmp, (__m128i)max4, 0x10 )); max4 = __lsx_vfmax_s( max4, (__m128)__lsx_vextrins_w((__m128i)tmp, (__m128i)max4, 0x1 ));
const float max_scalar = ((v4f32)max4)[0]; const float max_scalar = ((v4f32)max4)[0];
// Quantize these floats // Quantize these floats