convert Class — pytorch Architecture

Source Code

aten/src/ATen/cpu/vec/sve/vec_float.h lines 652–664

template <>
inline void convert(const float* src, float* dst, int64_t n) {
  const int64_t fraction = n % Vectorized<float>::size();
#pragma unroll
  for (int64_t i = 0; i < n - fraction; i += Vectorized<float>::size()) {
    svst1_f32(ptrue, dst + i, svldnt1_f32(ptrue, src + i));
  }
#pragma unroll
  for (int64_t i = n - fraction; i < n; i += Vectorized<float>::size()) {
    svbool_t pg = svwhilelt_b32(i, n);
    svst1_f32(pg, dst + i, svldnt1_f32(pg, src + i));
  }
}