#include "simdutf/westmere/begin.h"

namespace simdutf {

namespace SIMDUTF_IMPLEMENTATION {

namespace {

#ifndef SIMDUTF_WESTMERE_H

  #error "westmere.h must be included"

#endif

using namespace simd;

#if SIMDUTF_FEATURE_ASCII || SIMDUTF_FEATURE_DETECT_ENCODING ||                \

    SIMDUTF_FEATURE_UTF8

simdutf_really_inline bool is_ascii(const simd8x64<uint8_t> &input) {

  return input.reduce_or().is_ascii();

}

#endif // SIMDUTF_FEATURE_ASCII || SIMDUTF_FEATURE_DETECT_ENCODING ||

       // SIMDUTF_FEATURE_UTF8

#if SIMDUTF_FEATURE_UTF8 || SIMDUTF_FEATURE_DETECT_ENCODING

simdutf_really_inline simd8<bool>

must_be_2_3_continuation(const simd8<uint8_t> prev2,

                         const simd8<uint8_t> prev3) {

  simd8<uint8_t> is_third_byte =

      prev2.saturating_sub(0xe0u - 0x80); // Only 111_____ will be >= 0x80

  simd8<uint8_t> is_fourth_byte =

      prev3.saturating_sub(0xf0u - 0x80); // Only 1111____ will be >= 0x80

  return simd8<bool>(is_third_byte | is_fourth_byte);

}

#endif // SIMDUTF_FEATURE_UTF8 || SIMDUTF_FEATURE_DETECT_ENCODING

#if SIMDUTF_FEATURE_UTF8

  #include "westmere/internal/loader.cpp"

#endif // SIMDUTF_FEATURE_UTF8

#if SIMDUTF_FEATURE_UTF16

  #include "westmere/sse_utf16fix.cpp"

#endif // SIMDUTF_FEATURE_UTF16

#if SIMDUTF_FEATURE_UTF16 || SIMDUTF_FEATURE_DETECT_ENCODING

  #include "westmere/sse_validate_utf16.cpp"

#endif // SIMDUTF_FEATURE_UTF16 || SIMDUTF_FEATURE_DETECT_ENCODING

#if SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_LATIN1

  #include "westmere/sse_convert_latin1_to_utf8.cpp"

#endif // SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_LATIN1

#if SIMDUTF_FEATURE_UTF16 && SIMDUTF_FEATURE_LATIN1

  #include "westmere/sse_convert_latin1_to_utf16.cpp"

#endif // SIMDUTF_FEATURE_UTF16 && SIMDUTF_FEATURE_LATIN1

#if SIMDUTF_FEATURE_UTF32 && SIMDUTF_FEATURE_LATIN1

  #include "westmere/sse_convert_latin1_to_utf32.cpp"

#endif // SIMDUTF_FEATURE_UTF32 && SIMDUTF_FEATURE_LATIN1

#if SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_UTF16

  #include "westmere/sse_convert_utf8_to_utf16.cpp"

#endif // SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_UTF16

#if SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_UTF32

  #include "westmere/sse_convert_utf8_to_utf32.cpp"

#endif // SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_UTF32

#if SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_LATIN1

  #include "westmere/sse_convert_utf8_to_latin1.cpp"

#endif // SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_LATIN1

#if SIMDUTF_FEATURE_UTF16 && SIMDUTF_FEATURE_LATIN1

  #include "westmere/sse_convert_utf16_to_latin1.cpp"

#endif // SIMDUTF_FEATURE_UTF16 && SIMDUTF_FEATURE_LATIN1

#if SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_UTF16

  #include "westmere/sse_convert_utf16_to_utf8.cpp"

#endif // SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_UTF16

#if SIMDUTF_FEATURE_UTF16 && SIMDUTF_FEATURE_UTF32

  #include "westmere/sse_convert_utf16_to_utf32.cpp"

#endif // SIMDUTF_FEATURE_UTF16 && SIMDUTF_FEATURE_UTF32

#if SIMDUTF_FEATURE_UTF32 && SIMDUTF_FEATURE_LATIN1

  #include "westmere/sse_convert_utf32_to_latin1.cpp"

#endif // SIMDUTF_FEATURE_UTF32 && SIMDUTF_FEATURE_LATIN1

#if SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_UTF32

  #include "westmere/sse_convert_utf32_to_utf8.cpp"

#endif // SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_UTF32

#if SIMDUTF_FEATURE_UTF16 && SIMDUTF_FEATURE_UTF32

  #include "westmere/sse_convert_utf32_to_utf16.cpp"

#endif // SIMDUTF_FEATURE_UTF16 && SIMDUTF_FEATURE_UTF32

#if SIMDUTF_FEATURE_BASE64

  #include "westmere/sse_base64.cpp"

#endif // SIMDUTF_FEATURE_BASE64

} // unnamed namespace

} // namespace SIMDUTF_IMPLEMENTATION

} // namespace simdutf

#include "generic/buf_block_reader.h"

#if SIMDUTF_FEATURE_UTF8 || SIMDUTF_FEATURE_DETECT_ENCODING

  #include "generic/utf8_validation/utf8_lookup4_algorithm.h"

  #include "generic/utf8_validation/utf8_validator.h"

#endif // SIMDUTF_FEATURE_UTF8 || SIMDUTF_FEATURE_DETECT_ENCODING

#if SIMDUTF_FEATURE_ASCII

  #include "generic/ascii_validation.h"

#endif // SIMDUTF_FEATURE_ASCII

#if SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_UTF16

  // transcoding from UTF-8 to UTF-16

  #include "generic/utf8_to_utf16/valid_utf8_to_utf16.h"

  #include "generic/utf8_to_utf16/utf8_to_utf16.h"

  #include "generic/utf8/utf16_length_from_utf8_bytemask.h"

  // transcoding from UTF-16 to UTF-8

  #include "generic/utf16_to_utf8/utf16_to_utf8_with_replacement.h"

#endif // SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_UTF16

#if SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_UTF32

  #include "generic/utf8_to_utf32/valid_utf8_to_utf32.h"

  #include "generic/utf8_to_utf32/utf8_to_utf32.h"

  #include "generic/utf32.h"

#endif // SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_UTF32

#if SIMDUTF_FEATURE_UTF8

  #include "generic/utf8.h"

#endif // SIMDUTF_FEATURE_UTF8

#if SIMDUTF_FEATURE_UTF16

  #include "generic/utf16.h"

  #include "generic/utf16/utf8_length_from_utf16_bytemask.h"

#endif // SIMDUTF_FEATURE_UTF16

#if SIMDUTF_FEATURE_UTF16 || SIMDUTF_FEATURE_DETECT_ENCODING

  #include "generic/validate_utf16.h"

#endif // SIMDUTF_FEATURE_UTF16 || SIMDUTF_FEATURE_DETECT_ENCODING

#if SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_LATIN1

  #include "generic/utf8_to_latin1/utf8_to_latin1.h"

  #include "generic/utf8_to_latin1/valid_utf8_to_latin1.h"

#endif // SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_LATIN1

#if SIMDUTF_FEATURE_UTF32 || SIMDUTF_FEATURE_DETECT_ENCODING

  #include "generic/validate_utf32.h"

#endif // SIMDUTF_FEATURE_UTF32 || SIMDUTF_FEATURE_DETECT_ENCODING

#if SIMDUTF_FEATURE_BASE64

  #include "generic/base64.h"

  #include "generic/find.h"

  #include "generic/base64lengths.h"

#endif // SIMDUTF_FEATURE_BASE64

//

// Implementation-specific overrides

//

namespace simdutf {

namespace SIMDUTF_IMPLEMENTATION {

#if SIMDUTF_FEATURE_DETECT_ENCODING

simdutf_warn_unused int

implementation::detect_encodings(const char *input,

                                 size_t length) const noexcept {

  // If there is a BOM, then we trust it.

  auto bom_encoding = simdutf::BOM::check_bom(input, length);

  if (bom_encoding != encoding_type::unspecified) {

    return bom_encoding;

  }

  int out = 0;

  uint32_t utf16_err = (length % 2);

  uint32_t utf32_err = (length % 4);

  uint32_t ends_with_high = 0;

  const auto v_d8 = simd8<uint8_t>::splat(0xd8);

  const auto v_f8 = simd8<uint8_t>::splat(0xf8);

  const auto v_fc = simd8<uint8_t>::splat(0xfc);

  const auto v_dc = simd8<uint8_t>::splat(0xdc);

  const __m128i standardmax = _mm_set1_epi32(0x10ffff);

  const __m128i offset = _mm_set1_epi32(0xffff2000);

  const __m128i standardoffsetmax = _mm_set1_epi32(0xfffff7ff);

  __m128i currentmax = _mm_setzero_si128();

  __m128i currentoffsetmax = _mm_setzero_si128();

  utf8_checker c{};

  buf_block_reader<64> reader(reinterpret_cast<const uint8_t *>(input), length);

  while (reader.has_full_block()) {

    simd::simd8x64<uint8_t> in(reader.full_block());

    // utf8 checks

    c.check_next_input(in);

    // utf16le checks

    auto in0 = simd16<uint16_t>(in.chunks[0]);

    auto in1 = simd16<uint16_t>(in.chunks[1]);

    const auto t0 = in0.shr<8>();

    const auto t1 = in1.shr<8>();

    const auto packed1 = simd16<uint16_t>::pack(t0, t1);

    auto in2 = simd16<uint16_t>(in.chunks[2]);

    auto in3 = simd16<uint16_t>(in.chunks[3]);

    const auto t2 = in2.shr<8>();

    const auto t3 = in3.shr<8>();

    const auto packed2 = simd16<uint16_t>::pack(t2, t3);

    const auto surrogates_wordmask_lo = (packed1 & v_f8) == v_d8;

    const auto surrogates_wordmask_hi = (packed2 & v_f8) == v_d8;

    const uint32_t surrogates_bitmask =

        (surrogates_wordmask_hi.to_bitmask() << 16) |

        surrogates_wordmask_lo.to_bitmask();

    const auto vL_lo = (packed1 & v_fc) == v_dc;

    const auto vL_hi = (packed2 & v_fc) == v_dc;

    const uint32_t L = (vL_hi.to_bitmask() << 16) | vL_lo.to_bitmask();

    const uint32_t H = L ^ surrogates_bitmask;

    utf16_err |= (((H << 1) | ends_with_high) != L);

    ends_with_high = (H & 0x80000000) != 0;

    // utf32le checks

    currentmax = _mm_max_epu32(in.chunks[0], currentmax);

    currentoffsetmax =

        _mm_max_epu32(_mm_add_epi32(in.chunks[0], offset), currentoffsetmax);

    currentmax = _mm_max_epu32(in.chunks[1], currentmax);

    currentoffsetmax =

        _mm_max_epu32(_mm_add_epi32(in.chunks[1], offset), currentoffsetmax);

    currentmax = _mm_max_epu32(in.chunks[2], currentmax);

    currentoffsetmax =

        _mm_max_epu32(_mm_add_epi32(in.chunks[2], offset), currentoffsetmax);

    currentmax = _mm_max_epu32(in.chunks[3], currentmax);

    currentoffsetmax =

        _mm_max_epu32(_mm_add_epi32(in.chunks[3], offset), currentoffsetmax);

    reader.advance();

  }

  uint8_t block[64]{};

  size_t idx = reader.block_index();

  std::memcpy(block, &input[idx], length - idx);

  simd::simd8x64<uint8_t> in(block);

  c.check_next_input(in);

  // utf16le last block check

  auto in0 = simd16<uint16_t>(in.chunks[0]);

  auto in1 = simd16<uint16_t>(in.chunks[1]);

  const auto t0 = in0.shr<8>();

  const auto t1 = in1.shr<8>();

  const auto packed1 = simd16<uint16_t>::pack(t0, t1);

  auto in2 = simd16<uint16_t>(in.chunks[2]);

  auto in3 = simd16<uint16_t>(in.chunks[3]);

  const auto t2 = in2.shr<8>();

  const auto t3 = in3.shr<8>();

  const auto packed2 = simd16<uint16_t>::pack(t2, t3);

  const auto surrogates_wordmask_lo = (packed1 & v_f8) == v_d8;

  const auto surrogates_wordmask_hi = (packed2 & v_f8) == v_d8;

  const uint32_t surrogates_bitmask =

      (surrogates_wordmask_hi.to_bitmask() << 16) |

      surrogates_wordmask_lo.to_bitmask();

  const auto vL_lo = (packed1 & v_fc) == v_dc;

  const auto vL_hi = (packed2 & v_fc) == v_dc;

  const uint32_t L = (vL_hi.to_bitmask() << 16) | vL_lo.to_bitmask();

  const uint32_t H = L ^ surrogates_bitmask;

  utf16_err |= (((H << 1) | ends_with_high) != L);

  // this is required to check for last byte ending in high and end of input

  // is reached

  ends_with_high = (H & 0x80000000) != 0;

  utf16_err |= ends_with_high;

  // utf32le last block check

  currentmax = _mm_max_epu32(in.chunks[0], currentmax);

  currentoffsetmax =

      _mm_max_epu32(_mm_add_epi32(in.chunks[0], offset), currentoffsetmax);

  currentmax = _mm_max_epu32(in.chunks[1], currentmax);

  currentoffsetmax =

      _mm_max_epu32(_mm_add_epi32(in.chunks[1], offset), currentoffsetmax);

  currentmax = _mm_max_epu32(in.chunks[2], currentmax);

  currentoffsetmax =

      _mm_max_epu32(_mm_add_epi32(in.chunks[2], offset), currentoffsetmax);

  currentmax = _mm_max_epu32(in.chunks[3], currentmax);

  currentoffsetmax =

      _mm_max_epu32(_mm_add_epi32(in.chunks[3], offset), currentoffsetmax);

  reader.advance();

  c.check_eof();

  bool is_valid_utf8 = !c.errors();

  __m128i is_zero =

      _mm_xor_si128(_mm_max_epu32(currentmax, standardmax), standardmax);

  utf32_err |= (_mm_test_all_zeros(is_zero, is_zero) == 0);

  is_zero = _mm_xor_si128(_mm_max_epu32(currentoffsetmax, standardoffsetmax),

                          standardoffsetmax);

  utf32_err |= (_mm_test_all_zeros(is_zero, is_zero) == 0);

  if (is_valid_utf8) {

    out |= encoding_type::UTF8;

  }

  if (utf16_err == 0) {

    out |= encoding_type::UTF16_LE;

  }

  if (utf32_err == 0) {

    out |= encoding_type::UTF32_LE;

  }

  return out;

}

#endif // SIMDUTF_FEATURE_DETECT_ENCODING

#if SIMDUTF_FEATURE_UTF8 || SIMDUTF_FEATURE_DETECT_ENCODING

simdutf_warn_unused bool

implementation::validate_utf8(const char *buf, size_t len) const noexcept {

  return westmere::utf8_validation::generic_validate_utf8(buf, len);

}

#endif // SIMDUTF_FEATURE_UTF8 || SIMDUTF_FEATURE_DETECT_ENCODING

#if SIMDUTF_FEATURE_UTF8

simdutf_warn_unused result implementation::validate_utf8_with_errors(

    const char *buf, size_t len) const noexcept {

  return westmere::utf8_validation::generic_validate_utf8_with_errors(buf, len);

}

#endif // SIMDUTF_FEATURE_UTF8

#if SIMDUTF_FEATURE_ASCII

simdutf_warn_unused bool

implementation::validate_ascii(const char *buf, size_t len) const noexcept {

  return westmere::ascii_validation::generic_validate_ascii(buf, len);

}

simdutf_warn_unused result implementation::validate_ascii_with_errors(

    const char *buf, size_t len) const noexcept {

  return westmere::ascii_validation::generic_validate_ascii_with_errors(buf,

                                                                        len);

}

#endif // SIMDUTF_FEATURE_ASCII

#if SIMDUTF_FEATURE_UTF16 && SIMDUTF_FEATURE_ASCII

simdutf_warn_unused bool

implementation::validate_utf16le_as_ascii(const char16_t *buf,

                                          size_t len) const noexcept {

  return westmere::utf16::validate_utf16_as_ascii_with_errors<

             endianness::LITTLE>(buf, len)

             .error == SUCCESS;

}

simdutf_warn_unused bool

implementation::validate_utf16be_as_ascii(const char16_t *buf,

                                          size_t len) const noexcept {

  return westmere::utf16::validate_utf16_as_ascii_with_errors<endianness::BIG>(

             buf, len)

             .error == SUCCESS;

}

#endif // SIMDUTF_FEATURE_UTF16 && SIMDUTF_FEATURE_ASCII

#if SIMDUTF_FEATURE_UTF16 || SIMDUTF_FEATURE_DETECT_ENCODING

simdutf_warn_unused bool

implementation::validate_utf16le(const char16_t *buf,

                                 size_t len) const noexcept {

  if (simdutf_unlikely(len == 0)) {

    // empty input is valid UTF-16. protect the implementation from

    // handling nullptr

    return true;

  }

  const auto res =

      westmere::utf16::validate_utf16_with_errors<endianness::LITTLE>(buf, len);

  if (res.is_err()) {

    return false;

  }

  if (res.count == len)

    return true;

  return scalar::utf16::validate<endianness::LITTLE>(buf + res.count,

                                                     len - res.count);

}

#endif // SIMDUTF_FEATURE_UTF16 || SIMDUTF_FEATURE_DETECT_ENCODING

#if SIMDUTF_FEATURE_UTF16

simdutf_warn_unused bool

implementation::validate_utf16be(const char16_t *buf,

                                 size_t len) const noexcept {

  if (simdutf_unlikely(len == 0)) {

    // empty input is valid UTF-16. protect the implementation from

    // handling nullptr

    return true;

  }

  const auto res =

      westmere::utf16::validate_utf16_with_errors<endianness::BIG>(buf, len);

  if (res.is_err()) {

    return false;

  }

  if (res.count == len)

    return true;

  return scalar::utf16::validate<endianness::BIG>(buf + res.count,

                                                  len - res.count);

}

simdutf_warn_unused result implementation::validate_utf16le_with_errors(

    const char16_t *buf, size_t len) const noexcept {

  const result res =

      westmere::utf16::validate_utf16_with_errors<endianness::LITTLE>(buf, len);

  if (res.count != len) {

    const result scalar_res =

        scalar::utf16::validate_with_errors<endianness::LITTLE>(

            buf + res.count, len - res.count);

    return result(scalar_res.error, res.count + scalar_res.count);

  } else {

    return res;

  }

}

simdutf_warn_unused result implementation::validate_utf16be_with_errors(

    const char16_t *buf, size_t len) const noexcept {

  const result res =

      westmere::utf16::validate_utf16_with_errors<endianness::BIG>(buf, len);

  if (res.count != len) {

    result scalar_res = scalar::utf16::validate_with_errors<endianness::BIG>(

        buf + res.count, len - res.count);

    return result(scalar_res.error, res.count + scalar_res.count);

  } else {

    return res;

  }

}

void implementation::to_well_formed_utf16le(const char16_t *input, size_t len,

                                            char16_t *output) const noexcept {

  return utf16fix_sse<endianness::LITTLE>(input, len, output);

}

void implementation::to_well_formed_utf16be(const char16_t *input, size_t len,

                                            char16_t *output) const noexcept {

  return utf16fix_sse<endianness::BIG>(input, len, output);

}

#endif // SIMDUTF_FEATURE_UTF16

#if SIMDUTF_FEATURE_UTF32 || SIMDUTF_FEATURE_DETECT_ENCODING

simdutf_warn_unused bool

implementation::validate_utf32(const char32_t *buf, size_t len) const noexcept {

  return utf32::validate(buf, len);

}

#endif // SIMDUTF_FEATURE_UTF32 || SIMDUTF_FEATURE_DETECT_ENCODING

#if SIMDUTF_FEATURE_UTF32

simdutf_warn_unused result implementation::validate_utf32_with_errors(

    const char32_t *buf, size_t len) const noexcept {

  return utf32::validate_with_errors(buf, len);

}

#endif // SIMDUTF_FEATURE_UTF32

#if SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_LATIN1

simdutf_warn_unused size_t implementation::convert_latin1_to_utf8(

    const char *buf, size_t len, char *utf8_output) const noexcept {

  std::pair<const char *, char *> ret =

      sse_convert_latin1_to_utf8(buf, len, utf8_output);

  size_t converted_chars = ret.second - utf8_output;

  if (ret.first != buf + len) {

    const size_t scalar_converted_chars = scalar::latin1_to_utf8::convert(

        ret.first, len - (ret.first - buf), ret.second);

    converted_chars += scalar_converted_chars;

  }

  return converted_chars;

}

#endif // SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_LATIN1

#if SIMDUTF_FEATURE_UTF16 && SIMDUTF_FEATURE_LATIN1

simdutf_warn_unused size_t implementation::convert_latin1_to_utf16le(

    const char *buf, size_t len, char16_t *utf16_output) const noexcept {

  std::pair<const char *, char16_t *> ret =

      sse_convert_latin1_to_utf16<endianness::LITTLE>(buf, len, utf16_output);

  if (ret.first == nullptr) {

    return 0;

  }

  size_t converted_chars = ret.second - utf16_output;

  if (ret.first != buf + len) {

    const size_t scalar_converted_chars =

        scalar::latin1_to_utf16::convert<endianness::LITTLE>(

            ret.first, len - (ret.first - buf), ret.second);

    if (scalar_converted_chars == 0) {

      return 0;

    }

    converted_chars += scalar_converted_chars;

  }

  return converted_chars;

}

simdutf_warn_unused size_t implementation::convert_latin1_to_utf16be(

    const char *buf, size_t len, char16_t *utf16_output) const noexcept {

  std::pair<const char *, char16_t *> ret =

      sse_convert_latin1_to_utf16<endianness::BIG>(buf, len, utf16_output);

  if (ret.first == nullptr) {

    return 0;

  }

  size_t converted_chars = ret.second - utf16_output;

  if (ret.first != buf + len) {

    const size_t scalar_converted_chars =

        scalar::latin1_to_utf16::convert<endianness::BIG>(

            ret.first, len - (ret.first - buf), ret.second);

    if (scalar_converted_chars == 0) {

      return 0;

    }

    converted_chars += scalar_converted_chars;

  }

  return converted_chars;

}

#endif // SIMDUTF_FEATURE_UTF16 && SIMDUTF_FEATURE_LATIN1

#if SIMDUTF_FEATURE_UTF32 && SIMDUTF_FEATURE_LATIN1

simdutf_warn_unused size_t implementation::convert_latin1_to_utf32(

    const char *buf, size_t len, char32_t *utf32_output) const noexcept {

  std::pair<const char *, char32_t *> ret =

      sse_convert_latin1_to_utf32(buf, len, utf32_output);

  if (ret.first == nullptr) {

    return 0;

  }

  size_t converted_chars = ret.second - utf32_output;

  if (ret.first != buf + len) {

    const size_t scalar_converted_chars = scalar::latin1_to_utf32::convert(

        ret.first, len - (ret.first - buf), ret.second);

    if (scalar_converted_chars == 0) {

      return 0;

    }

    converted_chars += scalar_converted_chars;

  }

  return converted_chars;

}

#endif // SIMDUTF_FEATURE_UTF32 && SIMDUTF_FEATURE_LATIN1

#if SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_LATIN1

simdutf_warn_unused size_t implementation::convert_utf8_to_latin1(

    const char *buf, size_t len, char *latin1_output) const noexcept {

  utf8_to_latin1::validating_transcoder converter;

  return converter.convert(buf, len, latin1_output);

}

simdutf_warn_unused result implementation::convert_utf8_to_latin1_with_errors(

    const char *buf, size_t len, char *latin1_output) const noexcept {

  utf8_to_latin1::validating_transcoder converter;

  return converter.convert_with_errors(buf, len, latin1_output);

}

simdutf_warn_unused size_t implementation::convert_valid_utf8_to_latin1(

    const char *buf, size_t len, char *latin1_output) const noexcept {

  return westmere::utf8_to_latin1::convert_valid(buf, len, latin1_output);

}

#endif // SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_LATIN1

#if SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_UTF16

simdutf_warn_unused size_t implementation::convert_utf8_to_utf16le(

    const char *buf, size_t len, char16_t *utf16_output) const noexcept {

  utf8_to_utf16::validating_transcoder converter;

  return converter.convert<endianness::LITTLE>(buf, len, utf16_output);

}

simdutf_warn_unused size_t implementation::convert_utf8_to_utf16be(

    const char *buf, size_t len, char16_t *utf16_output) const noexcept {

  utf8_to_utf16::validating_transcoder converter;

  return converter.convert<endianness::BIG>(buf, len, utf16_output);

}

simdutf_warn_unused result implementation::convert_utf8_to_utf16le_with_errors(

    const char *buf, size_t len, char16_t *utf16_output) const noexcept {

  utf8_to_utf16::validating_transcoder converter;

  return converter.convert_with_errors<endianness::LITTLE>(buf, len,

                                                           utf16_output);

}

simdutf_warn_unused result implementation::convert_utf8_to_utf16be_with_errors(

    const char *buf, size_t len, char16_t *utf16_output) const noexcept {

  utf8_to_utf16::validating_transcoder converter;

  return converter.convert_with_errors<endianness::BIG>(buf, len, utf16_output);

}

simdutf_warn_unused size_t implementation::convert_valid_utf8_to_utf16le(

    const char *input, size_t size, char16_t *utf16_output) const noexcept {

  return utf8_to_utf16::convert_valid<endianness::LITTLE>(input, size,

                                                          utf16_output);

}

simdutf_warn_unused size_t implementation::convert_valid_utf8_to_utf16be(

    const char *input, size_t size, char16_t *utf16_output) const noexcept {

  return utf8_to_utf16::convert_valid<endianness::BIG>(input, size,

                                                       utf16_output);

}

#endif // SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_UTF16

#if SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_UTF32

simdutf_warn_unused size_t implementation::convert_utf8_to_utf32(

    const char *buf, size_t len, char32_t *utf32_output) const noexcept {

  utf8_to_utf32::validating_transcoder converter;

  return converter.convert(buf, len, utf32_output);

}

simdutf_warn_unused result implementation::convert_utf8_to_utf32_with_errors(

    const char *buf, size_t len, char32_t *utf32_output) const noexcept {

  utf8_to_utf32::validating_transcoder converter;

  return converter.convert_with_errors(buf, len, utf32_output);

}

simdutf_warn_unused size_t implementation::convert_valid_utf8_to_utf32(

    const char *input, size_t size, char32_t *utf32_output) const noexcept {

  return utf8_to_utf32::convert_valid(input, size, utf32_output);

}

#endif // SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_UTF32

#if SIMDUTF_FEATURE_UTF16 && SIMDUTF_FEATURE_LATIN1

simdutf_warn_unused size_t implementation::convert_utf16le_to_latin1(

    const char16_t *buf, size_t len, char *latin1_output) const noexcept {

  std::pair<const char16_t *, char *> ret =

      sse_convert_utf16_to_latin1<endianness::LITTLE>(buf, len, latin1_output);

  if (ret.first == nullptr) {

    return 0;

  }

  size_t saved_bytes = ret.second - latin1_output;

  if (ret.first != buf + len) {

    const size_t scalar_saved_bytes =

        scalar::utf16_to_latin1::convert<endianness::LITTLE>(

            ret.first, len - (ret.first - buf), ret.second);

    if (scalar_saved_bytes == 0) {

      return 0;

    }

    saved_bytes += scalar_saved_bytes;

  }

  return saved_bytes;

}

simdutf_warn_unused size_t implementation::convert_utf16be_to_latin1(

    const char16_t *buf, size_t len, char *latin1_output) const noexcept {

  std::pair<const char16_t *, char *> ret =

      sse_convert_utf16_to_latin1<endianness::BIG>(buf, len, latin1_output);

  if (ret.first == nullptr) {

    return 0;

  }

  size_t saved_bytes = ret.second - latin1_output;

  if (ret.first != buf + len) {

    const size_t scalar_saved_bytes =

        scalar::utf16_to_latin1::convert<endianness::BIG>(

            ret.first, len - (ret.first - buf), ret.second);

    if (scalar_saved_bytes == 0) {

      return 0;

    }

    saved_bytes += scalar_saved_bytes;

  }

  return saved_bytes;

}

simdutf_warn_unused result

implementation::convert_utf16le_to_latin1_with_errors(

    const char16_t *buf, size_t len, char *latin1_output) const noexcept {

  std::pair<result, char *> ret =

      sse_convert_utf16_to_latin1_with_errors<endianness::LITTLE>(

          buf, len, latin1_output);

  if (ret.first.error) {

    return ret.first;

  } // Can return directly since scalar fallback already found correct

    // ret.first.count

  if (ret.first.count != len) { // All good so far, but not finished

    result scalar_res =

        scalar::utf16_to_latin1::convert_with_errors<endianness::LITTLE>(

            buf + ret.first.count, len - ret.first.count, ret.second);

    if (scalar_res.error) {

      scalar_res.count += ret.first.count;

      return scalar_res;

    } else {

      ret.second += scalar_res.count;

    }

  }

  ret.first.count =

      ret.second -

      latin1_output; // Set count to the number of 8-bit code units written

  return ret.first;

}

simdutf_warn_unused result

implementation::convert_utf16be_to_latin1_with_errors(

    const char16_t *buf, size_t len, char *latin1_output) const noexcept {

  std::pair<result, char *> ret =

      sse_convert_utf16_to_latin1_with_errors<endianness::BIG>(buf, len,

                                                               latin1_output);

  if (ret.first.error) {

    return ret.first;

  } // Can return directly since scalar fallback already found correct

    // ret.first.count

  if (ret.first.count != len) { // All good so far, but not finished

    result scalar_res =

        scalar::utf16_to_latin1::convert_with_errors<endianness::BIG>(

            buf + ret.first.count, len - ret.first.count, ret.second);

    if (scalar_res.error) {

      scalar_res.count += ret.first.count;

      return scalar_res;

    } else {

      ret.second += scalar_res.count;

    }

  }

  ret.first.count =

      ret.second -

      latin1_output; // Set count to the number of 8-bit code units written

  return ret.first;

}

simdutf_warn_unused size_t implementation::convert_valid_utf16be_to_latin1(

    const char16_t *buf, size_t len, char *latin1_output) const noexcept {

  // optimization opportunity: we could provide an optimized function.

  return convert_utf16be_to_latin1(buf, len, latin1_output);

}

simdutf_warn_unused size_t implementation::convert_valid_utf16le_to_latin1(

    const char16_t *buf, size_t len, char *latin1_output) const noexcept {

  // optimization opportunity: we could provide an optimized function.

  return convert_utf16le_to_latin1(buf, len, latin1_output);

}

#endif // SIMDUTF_FEATURE_UTF16 && SIMDUTF_FEATURE_LATIN1

#if SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_UTF16

simdutf_warn_unused size_t implementation::convert_utf16le_to_utf8(

    const char16_t *buf, size_t len, char *utf8_output) const noexcept {

  std::pair<const char16_t *, char *> ret =

      sse_convert_utf16_to_utf8<endianness::LITTLE>(buf, len, utf8_output);

  if (ret.first == nullptr) {

    return 0;

  }

  size_t saved_bytes = ret.second - utf8_output;

  if (ret.first != buf + len) {

    const size_t scalar_saved_bytes =

        scalar::utf16_to_utf8::convert<endianness::LITTLE>(

            ret.first, len - (ret.first - buf), ret.second);

    if (scalar_saved_bytes == 0) {

      return 0;

    }

    saved_bytes += scalar_saved_bytes;

  }

  return saved_bytes;

}

simdutf_warn_unused size_t implementation::convert_utf16be_to_utf8(

    const char16_t *buf, size_t len, char *utf8_output) const noexcept {

  std::pair<const char16_t *, char *> ret =

      sse_convert_utf16_to_utf8<endianness::BIG>(buf, len, utf8_output);

  if (ret.first == nullptr) {

    return 0;

  }

  size_t saved_bytes = ret.second - utf8_output;

  if (ret.first != buf + len) {

    const size_t scalar_saved_bytes =

        scalar::utf16_to_utf8::convert<endianness::BIG>(

            ret.first, len - (ret.first - buf), ret.second);

    if (scalar_saved_bytes == 0) {

      return 0;

    }

    saved_bytes += scalar_saved_bytes;

  }

  return saved_bytes;

}

template <endianness big_endian>

simdutf_really_inline full_result convert_utf16_to_utf8_with_details(

    const char16_t *buf, size_t len, char *utf8_output) {

  std::pair<result, char *> ret =

      westmere::sse_convert_utf16_to_utf8_with_errors<big_endian>(buf, len,

                                                                  utf8_output);

  if (ret.first.error) {

    return full_result(ret.first.error, ret.first.count,

                       size_t(ret.second - utf8_output));

  }

  if (ret.first.count != len) {

    full_result sres =

        scalar::utf16_to_utf8::convert_with_errors<big_endian, false>(

            buf + ret.first.count, len - ret.first.count, ret.second, 0);

    return full_result(sres.error, ret.first.count + sres.input_count,

                       size_t(ret.second - utf8_output) + sres.output_count);

  }

  return full_result(error_code::SUCCESS, len,

                     size_t(ret.second - utf8_output));

}

Unexecuted instantiation: simdutf::full_result simdutf::westmere::convert_utf16_to_utf8_with_details<(simdutf::endianness)0>(char16_t const*, unsigned long, char*)

Unexecuted instantiation: simdutf::full_result simdutf::westmere::convert_utf16_to_utf8_with_details<(simdutf::endianness)1>(char16_t const*, unsigned long, char*)

simdutf_warn_unused result implementation::convert_utf16le_to_utf8_with_errors(

    const char16_t *buf, size_t len, char *utf8_output) const noexcept {

  return convert_utf16_to_utf8_with_details<endianness::LITTLE>(buf, len,

                                                                utf8_output);

}

simdutf_warn_unused result implementation::convert_utf16be_to_utf8_with_errors(

    const char16_t *buf, size_t len, char *utf8_output) const noexcept {

  return convert_utf16_to_utf8_with_details<endianness::BIG>(buf, len,

                                                             utf8_output);

}

simdutf_warn_unused size_t implementation::convert_valid_utf16le_to_utf8(

    const char16_t *buf, size_t len, char *utf8_output) const noexcept {

  return convert_utf16le_to_utf8(buf, len, utf8_output);

}

simdutf_warn_unused size_t implementation::convert_valid_utf16be_to_utf8(

    const char16_t *buf, size_t len, char *utf8_output) const noexcept {

  return convert_utf16be_to_utf8(buf, len, utf8_output);

}

#endif // SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_UTF16

#if SIMDUTF_FEATURE_UTF32 && SIMDUTF_FEATURE_LATIN1

simdutf_warn_unused size_t implementation::convert_utf32_to_latin1(

    const char32_t *buf, size_t len, char *latin1_output) const noexcept {

  std::pair<const char32_t *, char *> ret =

      sse_convert_utf32_to_latin1(buf, len, latin1_output);

  if (ret.first == nullptr) {

    return 0;

  }

  size_t saved_bytes = ret.second - latin1_output;

  // if (ret.first != buf + len) {

  if (ret.first < buf + len) {

    const size_t scalar_saved_bytes = scalar::utf32_to_latin1::convert(

        ret.first, len - (ret.first - buf), ret.second);

    if (scalar_saved_bytes == 0) {

      return 0;

    }

    saved_bytes += scalar_saved_bytes;

  }

  return saved_bytes;

}

simdutf_warn_unused result implementation::convert_utf32_to_latin1_with_errors(

    const char32_t *buf, size_t len, char *latin1_output) const noexcept {

  // ret.first.count is always the position in the buffer, not the number of

  // code units written even if finished

  std::pair<result, char *> ret =

      westmere::sse_convert_utf32_to_latin1_with_errors(buf, len,

                                                        latin1_output);

  if (ret.first.count != len) {

    result scalar_res = scalar::utf32_to_latin1::convert_with_errors(

        buf + ret.first.count, len - ret.first.count, ret.second);

    if (scalar_res.error) {

      scalar_res.count += ret.first.count;

      return scalar_res;

    } else {

      ret.second += scalar_res.count;

    }

  }

  ret.first.count =

      ret.second -

      latin1_output; // Set count to the number of 8-bit code units written

  return ret.first;

}

simdutf_warn_unused size_t implementation::convert_valid_utf32_to_latin1(

    const char32_t *buf, size_t len, char *latin1_output) const noexcept {

  // optimization opportunity: we could provide an optimized function.

  return convert_utf32_to_latin1(buf, len, latin1_output);

}

#endif // SIMDUTF_FEATURE_UTF32 && SIMDUTF_FEATURE_LATIN1

#if SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_UTF32

simdutf_warn_unused size_t implementation::convert_utf32_to_utf8(

    const char32_t *buf, size_t len, char *utf8_output) const noexcept {

  std::pair<const char32_t *, char *> ret =

      sse_convert_utf32_to_utf8(buf, len, utf8_output);

  if (ret.first == nullptr) {

    return 0;

  }

  size_t saved_bytes = ret.second - utf8_output;

  if (ret.first != buf + len) {

    const size_t scalar_saved_bytes = scalar::utf32_to_utf8::convert(

        ret.first, len - (ret.first - buf), ret.second);

    if (scalar_saved_bytes == 0) {

      return 0;

    }

    saved_bytes += scalar_saved_bytes;

  }

  return saved_bytes;

}

simdutf_warn_unused result implementation::convert_utf32_to_utf8_with_errors(

    const char32_t *buf, size_t len, char *utf8_output) const noexcept {

  // ret.first.count is always the position in the buffer, not the number of

  // code units written even if finished

  std::pair<result, char *> ret =

      westmere::sse_convert_utf32_to_utf8_with_errors(buf, len, utf8_output);

  if (ret.first.count != len) {

    result scalar_res = scalar::utf32_to_utf8::convert_with_errors(

        buf + ret.first.count, len - ret.first.count, ret.second);

    if (scalar_res.error) {

      scalar_res.count += ret.first.count;

      return scalar_res;

    } else {

      ret.second += scalar_res.count;

    }

  }

  ret.first.count =

      ret.second -

      utf8_output; // Set count to the number of 8-bit code units written

  return ret.first;

}

#endif // SIMDUTF_FEATURE_UTF8 && SIMDUTF_FEATURE_UTF32

#if SIMDUTF_FEATURE_UTF16 && SIMDUTF_FEATURE_UTF32

simdutf_warn_unused size_t implementation::convert_utf16le_to_utf32(

    const char16_t *buf, size_t len, char32_t *utf32_output) const noexcept {

  std::pair<const char16_t *, char32_t *> ret =

      sse_convert_utf16_to_utf32<endianness::LITTLE>(buf, len, utf32_output);

  if (ret.first == nullptr) {

    return 0;

  }

  size_t saved_bytes = ret.second - utf32_output;

  if (ret.first != buf + len) {

    const size_t scalar_saved_bytes =

        scalar::utf16_to_utf32::convert<endianness::LITTLE>(

            ret.first, len - (ret.first - buf), ret.second);

    if (scalar_saved_bytes == 0) {

      return 0;

    }

    saved_bytes += scalar_saved_bytes;

  }

  return saved_bytes;

}

simdutf_warn_unused size_t implementation::convert_utf16be_to_utf32(

    const char16_t *buf, size_t len, char32_t *utf32_output) const noexcept {

  std::pair<const char16_t *, char32_t *> ret =

      sse_convert_utf16_to_utf32<endianness::BIG>(buf, len, utf32_output);

  if (ret.first == nullptr) {

    return 0;

  }

  size_t saved_bytes = ret.second - utf32_output;

  if (ret.first != buf + len) {

    const size_t scalar_saved_bytes =

        scalar::utf16_to_utf32::convert<endianness::BIG>(

            ret.first, len - (ret.first - buf), ret.second);

    if (scalar_saved_bytes == 0) {

      return 0;

    }

    saved_bytes += scalar_saved_bytes;

  }

  return saved_bytes;

}

simdutf_warn_unused result implementation::convert_utf16le_to_utf32_with_errors(

    const char16_t *buf, size_t len, char32_t *utf32_output) const noexcept {

  // ret.first.count is always the position in the buffer, not the number of

  // code units written even if finished

  std::pair<result, char32_t *> ret =

      westmere::sse_convert_utf16_to_utf32_with_errors<endianness::LITTLE>(

          buf, len, utf32_output);

  if (ret.first.error) {

    return ret.first;

  } // Can return directly since scalar fallback already found correct

    // ret.first.count

  if (ret.first.count != len) { // All good so far, but not finished

    result scalar_res =

        scalar::utf16_to_utf32::convert_with_errors<endianness::LITTLE>(

            buf + ret.first.count, len - ret.first.count, ret.second);

    if (scalar_res.error) {

      scalar_res.count += ret.first.count;

      return scalar_res;

    } else {

      ret.second += scalar_res.count;

    }

  }

  ret.first.count =

      ret.second -

      utf32_output; // Set count to the number of 8-bit code units written

  return ret.first;

}