// Copyright 2025 Google LLC
//
// This source code is licensed under the BSD-style license found in the
// LICENSE file in the root directory of this source tree.

#ifndef XNNPACK_YNNPACK_BASE_SIMD_X86_AVX2_H_
#define XNNPACK_YNNPACK_BASE_SIMD_X86_AVX2_H_

#include <cstdint>

#include "ynnpack/base/base.h"
#include "ynnpack/base/simd/vec.h"
#include "ynnpack/base/simd/x86_avx2_base.h"  // IWYU pragma: export
#include "ynnpack/base/simd/x86_avx_base.h"  // IWYU pragma: export

namespace ynn {

namespace simd {

using f32x16 = vec<float, 16>;
using s32x16 = vec<int32_t, 16>;
using s32x32 = vec<int32_t, 32>;

YNN_ALWAYS_INLINE s32x16 convert(s8x16 a, int32_t) {
  return {
      s32x8{_mm256_cvtepi8_epi32(a.v)},
      s32x8{_mm256_cvtepi8_epi32(_mm_srli_si128(a.v, 8))},
  };
}

YNN_ALWAYS_INLINE s32x16 convert(u8x16 a, int32_t) {
  return {
      s32x8{_mm256_cvtepu8_epi32(a.v)},
      s32x8{_mm256_cvtepu8_epi32(_mm_srli_si128(a.v, 8))},
  };
}

}  // namespace simd

}  // namespace ynn

#include "ynnpack/base/simd/generic.inc"  // IWYU pragma: export

#endif  // XNNPACK_YNNPACK_BASE_SIMD_X86_AVX2_H_