happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Merge commit 'refs/pull/14417/head' of github.com:ggerganov/llama.cpp into xsn/ggml_scale_bias

This commit is contained in:
Xuan Son Nguyen 2025-07-09 12:18:36 +02:00
parent 34bacc8365 cd1703a3bc
commit 533016efa5
1 changed files with 25 additions and 19 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

44
ggml/src/ggml-cpu/vec.h
View File

@ -355,27 +355,33 @@ inline static void ggml_vec_mad1_f32(const int n, float * y, const float s, cons
#if defined(GGML_USE_ACCELERATE) #if defined(GGML_USE_ACCELERATE)
vDSP_vsmsa(y, 1, &s, &b, y, 1, n); vDSP_vsmsa(y, 1, &s, &b, y, 1, n);
#elif defined(GGML_SIMD) #elif defined(GGML_SIMD)
// TODO: #if defined(__ARM_FEATURE_SVE) #if defined(__ARM_FEATURE_SVE)
const int np = (n & ~(GGML_F32_STEP - 1)); // scalar ; TODO: Write SVE code
for (int i = 0; i < n; ++i) {
GGML_F32_VEC vs = GGML_F32_VEC_SET1(s); y[i] = y[i]*s + b;
GGML_F32_VEC vb = GGML_F32_VEC_SET1(b);
GGML_F32_VEC ay[GGML_F32_ARR];
for (int i = 0; i < np; i += GGML_F32_STEP) {
for (int j = 0; j < GGML_F32_ARR; j++) {
ay[j] = GGML_F32_VEC_LOAD(y + i + j*GGML_F32_EPR);
ay[j] = GGML_F32_VEC_FMA(ay[j], vs, vb);
GGML_F32_VEC_STORE(y + i + j*GGML_F32_EPR, ay[j]);
} }
} #else
const int np = (n & ~(GGML_F32_STEP - 1));
// leftovers GGML_F32_VEC vs = GGML_F32_VEC_SET1(s);
for (int i = np; i < n; ++i) { GGML_F32_VEC vb = GGML_F32_VEC_SET1(b);
y[i] = y[i]*s + b;
} GGML_F32_VEC ay[GGML_F32_ARR];
for (int i = 0; i < np; i += GGML_F32_STEP) {
for (int j = 0; j < GGML_F32_ARR; j++) {
ay[j] = GGML_F32_VEC_LOAD(y + i + j*GGML_F32_EPR);
ay[j] = GGML_F32_VEC_FMA(ay[j], vs, vb);
GGML_F32_VEC_STORE(y + i + j*GGML_F32_EPR, ay[j]);
}
}
// leftovers
for (int i = np; i < n; ++i) {
y[i] = y[i]*s + b;
}
#endif
#else #else
// scalar // scalar
for (int i = 0; i < n; ++i) { for (int i = 0; i < n; ++i) {