Home / Class/ convert Class — pytorch Architecture

convert Class — pytorch Architecture

Architecture documentation for the convert class in vec_int.h from the pytorch codebase.

Entity Profile

Source Code

aten/src/ATen/cpu/vec/sve/vec_int.h lines 309–323

template <>
inline void convert(const int32_t* src, int64_t* dst, int64_t n) {
  const int64_t fraction = n % Vectorized<int64_t>::size();
  svbool_t pg_32 = svwhilelt_b32(0ull, Vectorized<int64_t>::size());
  svbool_t pg_64 = svwhilelt_b64(0ull, Vectorized<int64_t>::size());
#pragma unroll
  for (int64_t i = 0; i < n - fraction; i += Vectorized<int64_t>::size())
    svst1_s64(pg_64, dst + i, svunpklo_s64(svldnt1_s32(pg_32, src + i)));
#pragma unroll
  for (int64_t i = n - fraction; i < n; i += Vectorized<int64_t>::size()) {
    pg_32 = svwhilelt_b32(i, n);
    pg_64 = svwhilelt_b64(i, n);
    svst1_s64(pg_64, dst + i, svunpklo_s64(svldnt1_s32(pg_32, src + i)));
  }
}

Analyze Your Own Codebase

Get architecture documentation, dependency graphs, and domain analysis for your codebase in minutes.

Try Supermodel Free