CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-14 11:07:10 +00:00
Code Issues Packages Projects Releases Wiki Activity

ggml-cpu : add RISC-V RVV (Zvfh) optimization for FP16 to FP32 conversion (#17161)

Browse Source 
Signed-off-by: Wang Yang <yangwang@iscas.ac.cn>
This commit is contained in:
ixgbe
2025-11-11 19:41:51 +08:00
committed by GitHub
parent 73460f6278
commit ca4844062b
1 changed files with 7 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
ggml/src/ggml-cpu/ggml-cpu.c
View File

@@ -3274,6 +3274,13 @@ void ggml_cpu_fp16_to_fp32(const ggml_fp16_t * x, float * y, int64_t n) {
__m128 y_vec = _mm_cvtph_ps(x_vec); __m128 y_vec = _mm_cvtph_ps(x_vec);
_mm_storeu_ps(y + i, y_vec); _mm_storeu_ps(y + i, y_vec);
} }
#elif defined(__riscv_zvfh)
for (int vl; i < n; i += vl) {
vl = __riscv_vsetvl_e16m1(n - i);
vfloat16m1_t vx = __riscv_vle16_v_f16m1((_Float16 *)&x[i], vl);
vfloat32m2_t vy = __riscv_vfwcvt_f_f_v_f32m2(vx, vl);
__riscv_vse32_v_f32m2(&y[i], vy, vl);
}
#endif #endif
for (; i < n; ++i) { for (; i < n; ++i) {