/*
 * Copyright (c) Meta Platforms, Inc. and affiliates.
 * All rights reserved.
 *
 * This source code is licensed under the BSD-style license found in the
 * LICENSE file in the root directory of this source tree.
 */

#pragma once

#include <executorch/kernels/optimized/vec/vec256/vec256.h>

namespace executorch {
namespace vec {

// See Note [CPU_CAPABILITY namespace]
inline namespace CPU_CAPABILITY {

inline Vectorized<bool> convert_to_bool(Vectorized<int8_t> x) {
  __at_align__ bool buffer[x.size()];
  x.ne(Vectorized<int8_t>(0)).store(buffer);

  Vectorized<bool> ret;
  static_assert(x.size() == ret.size(), "");
  std::memcpy(ret, buffer, ret.size() * sizeof(bool));
  return ret;
}

template <>
inline Vectorized<bool> Vectorized<bool>::loadu(const void* ptr) {
  // See NOTE [Loading boolean values]
  return convert_to_bool(Vectorized<int8_t>::loadu(ptr));
}

template <>
inline Vectorized<bool> Vectorized<bool>::loadu(const void* ptr, int64_t count) {
  // See NOTE [Loading boolean values]
  return convert_to_bool(Vectorized<int8_t>::loadu(ptr, count));
}

} // namespace CPU_CAPABILITY

} // namespace vec
} // namespace executorch