EngineX-Ascend/enginex-ascend-910-llama.cpp
10
0
Fork 0
Code Issues Pull Requests Actions 4 Projects Releases Wiki Activity

ggml-cpu: clean up s390x SIMD (#15855)

Browse Source 
* ggml-cpu: clean up s390x simd

Signed-off-by: Aaron Teo <aaron.teo1@ibm.com>
(cherry picked from commit 0da4b6aa07d96b758812d17b2c82267632fa4ba5)
Signed-off-by: Aaron Teo <aaron.teo1@ibm.com>

* ggml-cpu: fix hsum data types

Signed-off-by: Aaron Teo <aaron.teo1@ibm.com>

---------

Signed-off-by: Aaron Teo <aaron.teo1@ibm.com>
This commit is contained in:
Aaron Teo
2025-09-08 02:18:28 +08:00
committed by GitHub
parent c97b5e5854
commit d36e61c580
2 changed files with 63 additions and 60 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
ggml/src/ggml-cpu/ggml-cpu-impl.h
View File

@@ -483,11 +483,16 @@ inline static int16x8_t vec_padd_s16(int16x8_t a, int16x8_t b) {
/**
* @see https://github.com/ggml-org/llama.cpp/pull/14037
*/
inline static float vec_hsum(float32x4_t v) {
inline static float vec_hsum_f32x4(float32x4_t v) {
float32x4_t v_temp = v + vec_reve(v);
return v_temp[0] + v_temp[1];
}
inline static int32_t vec_hsum_i32x4(int32x4_t v) {
int32x4_t v_temp = v + vec_reve(v);
return v_temp[0] + v_temp[1];
}
inline static int32x4_t ggml_vec_dot(int32x4_t acc, int8x16_t a, int8x16_t b) {
const int16x8_t p = vec_mule(a, b) + vec_mulo(a, b);
return acc + (vec_unpackh(p) + vec_unpackl(p));