結果
| 問題 | No.3677 Global Checksum |
| コンテスト | |
| ユーザー |
 kemuniku
|
| 提出日時 | 2026-09-05 04:51:59 |
| 言語 | Nim (2.2.10) |
| 結果 |
TLE
|
| 実行時間 | - |
| コード長 | 58,081 bytes |
| 記録 | |
| コンパイル時間 | 3,020 ms |
| コンパイル使用メモリ | 142,280 KB |
| 実行使用メモリ | 48,896 KB |
| 最終ジャッジ日時 | 2026-09-05 04:52:13 |
| 合計ジャッジ時間 | 6,810 ms |
|
ジャッジサーバーID （参考情報） |
judge3_0 / judge1_0 |
（要ログイン）
| ファイルパターン | 結果 |
|---|---|
| sample | AC * 3 |
| other | AC * 12 TLE * 1 -- * 7 |
コンパイルメッセージ
--- Self-Recompiling with optimized settings ---
Command: cd /home/judge/data/code && export PATH=$HOME/.nimble/bin:$PATH && nim cpp -d:danger -d:second_compile -d:useMalloc --gc:none --panics:on --opt:speed --checks:off --passC:"-O3 -flto -m64 -march=native -ffast-math" --hints:off -o:a.out /home/judge/data/code/Main.nim
/home/judge/data/code/Main.nim(8, 9) template/generic instantiation of `optimize` from here
(25, 20) Warning: command line(1, 2) Warning: `gc:option` is deprecated; use `mm:option` instead [Deprecated]
command line(1, 2) Warning: `gc:option` is deprecated; use `mm:option` instead [Deprecated]
/home/linuxbrew/.linuxbrew/Cellar/nim/2.2.10/nim/lib/system/fatal.nim(62, 5) Warning: '(ref exceptn)(msg: message)' uses GC'ed memory [GcMem]
/home/linuxbrew/.linuxbrew/Cellar/nim/2.2.10/nim/lib/system.nim(1660, 28) Warning: 'newSeqOfCap(len)' uses GC'ed memory [GcMem]
/home/linuxbrew/.linuxbrew/Cellar/nim/2.2.10/nim/lib/std/syncio.nim(176, 45) Warning: '$strerror(errno)' uses GC'ed memory [GcMem]
/home/linuxbrew/.linuxbrew/Cellar/nim/2.2.10/nim/lib/std/syncio.nim(176, 62) Warning: '`&`("errno: ", $errno, " `", $strerror(errno), "`")' uses GC'ed memory [GcMem]
/home/linuxbrew/.linuxbrew/Cellar/nim/2.2.10/nim/lib/std/syncio.nim(161, 21) Warning: '(ref IOError)(msg: msg, parent: nil)' uses GC'ed memory [GcMem]
(1208, 27) Warning: 'setLen(result, chckRange(chckRange(written, -9223372036854775808,
9223372036854775807), 0, 9223372036854775807))' uses GC'ed memory [GcMem] [User]
ソースコード
import macros;macro ImportExpand(s:untyped):untyped = parseStmt($s[2])
# source: src/cplib/tmpl/sheep.nim
ImportExpand "cplib/tmpl/sheep" <=== "when not declared CPLIB_TMPL_SHEEP:\n const CPLIB_TMPL_SHEEP* = 1\n {.warning[UnusedImport]: off.}\n {.hint[XDeclaredButNotUsed]: off.}\n import algorithm\n import sequtils\n import tables\n import macros\n import math\n import sets\n import strutils\n import strformat\n import sugar\n import heapqueue\n import streams\n import deques\n import bitops\n import std/lenientops\n import options\n when not declared CPLIB_TMPL_FASTIO:\n const CPLIB_TMPL_FASTIO* = 1\n {.passC: \"-mavx2\".}\n when defined(fastioNoMmap):\n {.passC: \"-DCPLIB_FASTIO_NO_MMAP\".}\n import macros\n \n # 入力系\n {.emit: \"\"\"\n #include <cstdio>\n #include <cstddef>\n #include <cstdint>\n #include <cstring>\n #include <immintrin.h>\n #include <string>\n #include <sys/mman.h>\n #include <sys/stat.h>\n \n namespace cplib_fastio_input {\n constexpr std::size_t buffer_size = 1U << 20;\n constexpr std::size_t safe_integer_bytes = 32;\n \n struct InputState {\n alignas(64) char buffer[buffer_size];\n std::size_t cursor;\n std::size_t length;\n const char* mapped;\n bool initialized;\n };\n \n inline InputState& input_state() {\n static InputState state = {};\n return state;\n }\n \n inline std::string& token_storage() {\n static std::string storage;\n return storage;\n }\n \n #if defined(__GNUC__) || defined(__clang__)\n #define CPLIB_FASTIO_ALWAYS_INLINE inline __attribute__((always_inline))\n #define CPLIB_FASTIO_NIM_ALWAYS_INLINE \\\n static inline __attribute__((always_inline))\n #define CPLIB_FASTIO_UNLIKELY(condition) (__builtin_expect(!!(condition), 0))\n #elif defined(_MSC_VER)\n #define CPLIB_FASTIO_ALWAYS_INLINE __forceinline\n #define CPLIB_FASTIO_NIM_ALWAYS_INLINE static __forceinline\n #define CPLIB_FASTIO_UNLIKELY(condition) (condition)\n #else\n #define CPLIB_FASTIO_ALWAYS_INLINE inline\n #define CPLIB_FASTIO_NIM_ALWAYS_INLINE static inline\n #define CPLIB_FASTIO_UNLIKELY(condition) (condition)\n #endif\n \n inline void initialize(InputState& state) {\n if (state.initialized) return;\n state.initialized = true;\n \n #ifndef CPLIB_FASTIO_NO_MMAP\n struct stat st;\n const int fd = fileno(stdin);\n if (fstat(fd, &st) == 0 && S_ISREG(st.st_mode) && st.st_size > 0) {\n void* p = mmap(nullptr, static_cast<std::size_t>(st.st_size),\n PROT_READ, MAP_PRIVATE, fd, 0);\n if (p != MAP_FAILED) {\n state.mapped = static_cast<const char*>(p);\n state.length = static_cast<std::size_t>(st.st_size);\n madvise(const_cast<char*>(state.mapped), state.length,\n MADV_SEQUENTIAL);\n }\n }\n #endif\n }\n \n inline bool refill(InputState& state) {\n state.length =\n fread_unlocked(state.buffer, 1, buffer_size, stdin);\n state.cursor = 0;\n return state.length != 0;\n }\n \n inline int get_char() {\n InputState& state = input_state();\n if (CPLIB_FASTIO_UNLIKELY(!state.initialized)) initialize(state);\n if (state.mapped != nullptr) {\n if (state.cursor == state.length) return -1;\n return static_cast<unsigned char>(state.mapped[state.cursor++]);\n }\n if (state.cursor == state.length && !refill(state)) return -1;\n return static_cast<unsigned char>(state.buffer[state.cursor++]);\n }\n \n inline bool has_eight_digits(const char* source) {\n // 正しい整数入力では、数字・符号・空白を上位4bitだけで判別できる。\n std::uint64_t bytes;\n std::memcpy(&bytes, source, sizeof(bytes));\n return ((bytes ^ 0x3030303030303030ULL) &\n 0xf0f0f0f0f0f0f0f0ULL) == 0;\n }\n \n inline unsigned parse_eight_digits(const char* source) {\n std::uint64_t digits;\n std::memcpy(&digits, source, sizeof(digits));\n digits ^= 0x3030303030303030ULL;\n digits = (digits * ((10ULL << 8) + 1) >> 8) &\n 0x00ff00ff00ff00ffULL;\n digits = (digits * ((100ULL << 16) + 1) >> 16) &\n 0x0000ffff0000ffffULL;\n return static_cast<unsigned>(\n digits * ((10000ULL << 32) + 1) >> 32);\n }\n \n inline unsigned digit_at(const char* source) {\n return static_cast<unsigned>(static_cast<unsigned char>(*source)) -\n static_cast<unsigned>('0');\n }\n \n template <bool negative>\n CPLIB_FASTIO_ALWAYS_INLINE bool try_parse_digits_unchecked(\n const char*& source, long long& output) {\n long long value = 0;\n if (has_eight_digits(source)) {\n const long long digits =\n static_cast<long long>(parse_eight_digits(source));\n value = negative ? value * 100000000LL - digits\n : value * 100000000LL + digits;\n source += 8;\n if (has_eight_digits(source)) {\n const long long next_digits =\n static_cast<long long>(parse_eight_digits(source));\n value = negative ? value * 100000000LL - next_digits\n : value * 100000000LL + next_digits;\n source += 8;\n for (int pair = 0; pair < 2; ++pair) {\n const unsigned first = digit_at(source);\n if (first >= 10U) {\n output = value;\n return true;\n }\n const unsigned second = digit_at(source + 1);\n if (second >= 10U) {\n ++source;\n output = negative\n ? value * 10 - static_cast<long long>(first)\n : value * 10 + static_cast<long long>(first);\n return true;\n }\n value = negative\n ? value * 100 - static_cast<long long>(first * 10U + second)\n : value * 100 + static_cast<long long>(first * 10U + second);\n source += 2;\n }\n if (digit_at(source) < 10U) return false;\n output = value;\n return true;\n }\n }\n for (;;) {\n const unsigned first = digit_at(source);\n if (first >= 10U) {\n output = value;\n return true;\n }\n const unsigned second = digit_at(source + 1);\n if (second >= 10U) {\n ++source;\n output = negative ? value * 10 - static_cast<long long>(first)\n : value * 10 + static_cast<long long>(first);\n return true;\n }\n value = negative\n ? value * 100 - static_cast<long long>(first * 10U + second)\n : value * 100 + static_cast<long long>(first * 10U + second);\n source += 2;\n }\n }\n \n template <bool negative>\n inline long long parse_digits_bounded(const char*& source,\n const char* end) {\n long long value = 0;\n while (end - source >= 8 && has_eight_digits(source)) {\n const long long digits =\n static_cast<long long>(parse_eight_digits(source));\n value = negative ? value * 100000000LL - digits\n : value * 100000000LL + digits;\n source += 8;\n }\n while (end - source >= 2) {\n const unsigned first = digit_at(source);\n const unsigned second = digit_at(source + 1);\n if (first >= 10U) return value;\n if (second >= 10U) {\n ++source;\n return negative ? value * 10 - static_cast<long long>(first)\n : value * 10 + static_cast<long long>(first);\n }\n value = negative\n ? value * 100 - static_cast<long long>(first * 10U + second)\n : value * 100 + static_cast<long long>(first * 10U + second);\n source += 2;\n }\n if (source != end) {\n const unsigned digit = digit_at(source);\n if (digit < 10U) {\n value = negative ? value * 10 - static_cast<long long>(digit)\n : value * 10 + static_cast<long long>(digit);\n ++source;\n }\n }\n return value;\n }\n \n CPLIB_FASTIO_ALWAYS_INLINE bool try_parse_unsigned_digits_unchecked(\n const char*& source, unsigned long long& output) {\n unsigned long long value = 0;\n if (has_eight_digits(source)) {\n value = static_cast<unsigned long long>(parse_eight_digits(source));\n source += 8;\n if (has_eight_digits(source)) {\n value = value * 100000000ULL +\n static_cast<unsigned long long>(parse_eight_digits(source));\n source += 8;\n for (int pair = 0; pair < 2; ++pair) {\n const unsigned first = digit_at(source);\n if (first >= 10U) {\n output = value;\n return true;\n }\n const unsigned second = digit_at(source + 1);\n if (second >= 10U) {\n ++source;\n output = value * 10ULL + first;\n return true;\n }\n value = value * 100ULL + first * 10U + second;\n source += 2;\n }\n if (digit_at(source) < 10U) return false;\n output = value;\n return true;\n }\n }\n for (;;) {\n const unsigned first = digit_at(source);\n if (first >= 10U) {\n output = value;\n return true;\n }\n const unsigned second = digit_at(source + 1);\n if (second >= 10U) {\n ++source;\n output = value * 10ULL + first;\n return true;\n }\n value = value * 100ULL + first * 10U + second;\n source += 2;\n }\n }\n \n inline unsigned long long parse_unsigned_digits_bounded(\n const char*& source, const char* end) {\n unsigned long long value = 0;\n while (end - source >= 8 && has_eight_digits(source)) {\n value = value * 100000000ULL +\n static_cast<unsigned long long>(parse_eight_digits(source));\n source += 8;\n }\n while (end - source >= 2) {\n const unsigned first = digit_at(source);\n const unsigned second = digit_at(source + 1);\n if (first >= 10U) return value;\n if (second >= 10U) {\n ++source;\n return value * 10ULL + first;\n }\n value = value * 100ULL + first * 10U + second;\n source += 2;\n }\n if (source != end) {\n const unsigned digit = digit_at(source);\n if (digit < 10U) {\n value = value * 10ULL + digit;\n ++source;\n }\n }\n return value;\n }\n \n CPLIB_FASTIO_ALWAYS_INLINE long long read_mapped_at(\n const char*& source, const char* end) {\n while (source != end &&\n static_cast<unsigned char>(*source) <=\n static_cast<unsigned char>(' ')) {\n ++source;\n }\n if (source == end) return 0;\n const bool negative = *source == '-';\n if (negative || *source == '+') ++source;\n if (end - source >= static_cast<std::ptrdiff_t>(safe_integer_bytes)) {\n const char* parsed = source;\n long long value;\n const bool complete = negative\n ? try_parse_digits_unchecked<true>(parsed, value)\n : try_parse_digits_unchecked<false>(parsed, value);\n if (complete) {\n source = parsed;\n return value;\n }\n }\n return negative ? parse_digits_bounded<true>(source, end)\n : parse_digits_bounded<false>(source, end);\n }\n \n CPLIB_FASTIO_ALWAYS_INLINE unsigned long long read_uint_mapped_at(\n const char*& source, const char* end) {\n while (source != end &&\n static_cast<unsigned char>(*source) <=\n static_cast<unsigned char>(' ')) {\n ++source;\n }\n if (source == end) return 0;\n const bool negative = *source == '-';\n if (negative || *source == '+') ++source;\n if (end - source >= static_cast<std::ptrdiff_t>(safe_integer_bytes)) {\n const char* parsed = source;\n unsigned long long value;\n if (try_parse_unsigned_digits_unchecked(parsed, value)) {\n source = parsed;\n return negative ? 0ULL - value : value;\n }\n }\n const unsigned long long value =\n parse_unsigned_digits_bounded(source, end);\n return negative ? 0ULL - value : value;\n }\n \n inline long long read_int_stream_slow(InputState& state) {\n bool negative = state.buffer[state.cursor] == '-';\n if (negative || state.buffer[state.cursor] == '+') {\n ++state.cursor;\n if (state.cursor == state.length && !refill(state)) return 0;\n }\n long long value = 0;\n for (;;) {\n while (state.cursor != state.length) {\n const unsigned digit = digit_at(state.buffer + state.cursor);\n if (digit >= 10U) return value;\n value = negative ? value * 10 - static_cast<long long>(digit)\n : value * 10 + static_cast<long long>(digit);\n ++state.cursor;\n }\n if (!refill(state)) return value;\n }\n }\n \n inline unsigned long long read_uint_stream_slow(InputState& state) {\n const bool negative = state.buffer[state.cursor] == '-';\n if (negative || state.buffer[state.cursor] == '+') {\n ++state.cursor;\n if (state.cursor == state.length && !refill(state)) return 0;\n }\n unsigned long long value = 0;\n for (;;) {\n while (state.cursor != state.length) {\n const unsigned digit = digit_at(state.buffer + state.cursor);\n if (digit >= 10U) return negative ? 0ULL - value : value;\n value = value * 10ULL + digit;\n ++state.cursor;\n }\n if (!refill(state)) return negative ? 0ULL - value : value;\n }\n }\n \n CPLIB_FASTIO_ALWAYS_INLINE long long read_int_stream(InputState& state) {\n for (;;) {\n if (state.cursor == state.length && !refill(state)) return 0;\n while (state.cursor != state.length &&\n static_cast<unsigned char>(state.buffer[state.cursor]) <=\n static_cast<unsigned char>(' ')) {\n ++state.cursor;\n }\n if (state.cursor != state.length) break;\n }\n if (state.length - state.cursor < safe_integer_bytes) {\n return read_int_stream_slow(state);\n }\n const char* source = state.buffer + state.cursor;\n const bool negative = *source == '-';\n if (negative || *source == '+') ++source;\n long long value;\n const bool complete = negative\n ? try_parse_digits_unchecked<true>(source, value)\n : try_parse_digits_unchecked<false>(source, value);\n if (!complete) return read_int_stream_slow(state);\n state.cursor = static_cast<std::size_t>(source - state.buffer);\n return value;\n }\n \n CPLIB_FASTIO_ALWAYS_INLINE unsigned long long read_uint_stream(\n InputState& state) {\n for (;;) {\n if (state.cursor == state.length && !refill(state)) return 0;\n while (state.cursor != state.length &&\n static_cast<unsigned char>(state.buffer[state.cursor]) <=\n static_cast<unsigned char>(' ')) {\n ++state.cursor;\n }\n if (state.cursor != state.length) break;\n }\n if (state.length - state.cursor < safe_integer_bytes) {\n return read_uint_stream_slow(state);\n }\n const char* source = state.buffer + state.cursor;\n const bool negative = *source == '-';\n if (negative || *source == '+') ++source;\n unsigned long long value;\n if (!try_parse_unsigned_digits_unchecked(source, value)) {\n return read_uint_stream_slow(state);\n }\n state.cursor = static_cast<std::size_t>(source - state.buffer);\n return negative ? 0ULL - value : value;\n }\n \n CPLIB_FASTIO_ALWAYS_INLINE long long read_int() {\n InputState& state = input_state();\n if (CPLIB_FASTIO_UNLIKELY(!state.initialized)) initialize(state);\n if (state.mapped == nullptr) return read_int_stream(state);\n const char* source = state.mapped + state.cursor;\n const long long value =\n read_mapped_at(source, state.mapped + state.length);\n state.cursor = static_cast<std::size_t>(source - state.mapped);\n return value;\n }\n \n CPLIB_FASTIO_ALWAYS_INLINE unsigned long long read_uint() {\n InputState& state = input_state();\n if (CPLIB_FASTIO_UNLIKELY(!state.initialized)) initialize(state);\n if (state.mapped == nullptr) return read_uint_stream(state);\n const char* source = state.mapped + state.cursor;\n const unsigned long long value =\n read_uint_mapped_at(source, state.mapped + state.length);\n state.cursor = static_cast<std::size_t>(source - state.mapped);\n return value;\n }\n \n template <class T>\n inline void read_int_array(T* output, std::size_t count) {\n InputState& state = input_state();\n if (CPLIB_FASTIO_UNLIKELY(!state.initialized)) initialize(state);\n if (state.mapped != nullptr) {\n const char* source = state.mapped + state.cursor;\n const char* const end = state.mapped + state.length;\n for (std::size_t i = 0; i < count; ++i) {\n output[i] = static_cast<T>(read_mapped_at(source, end));\n }\n state.cursor = static_cast<std::size_t>(source - state.mapped);\n } else {\n for (std::size_t i = 0; i < count; ++i) {\n output[i] = static_cast<T>(read_int_stream(state));\n }\n }\n }\n \n inline bool fastio_is_space(char value) {\n return static_cast<unsigned char>(value) <=\n static_cast<unsigned char>(' ');\n }\n \n inline void skip_spaces(const char*& source, const char* end) {\n while (source != end && fastio_is_space(*source)) ++source;\n }\n \n inline void consume_separator(const char*& source, const char* end) {\n if (source != end) {\n ++source;\n skip_spaces(source, end);\n }\n }\n \n inline unsigned parse_eight_digits_simd(const char* source) {\n const __m128i bytes = _mm_loadl_epi64(\n reinterpret_cast<const __m128i*>(source));\n const __m128i digits = _mm_sub_epi8(bytes, _mm_set1_epi8('0'));\n const __m128i pairs = _mm_maddubs_epi16(\n digits, _mm_setr_epi8(10, 1, 10, 1, 10, 1, 10, 1,\n 0, 0, 0, 0, 0, 0, 0, 0));\n const __m128i quads = _mm_madd_epi16(\n pairs, _mm_setr_epi16(100, 1, 100, 1, 0, 0, 0, 0));\n return static_cast<unsigned>(_mm_cvtsi128_si32(quads)) * 10000U +\n static_cast<unsigned>(_mm_cvtsi128_si32(_mm_srli_si128(quads, 4)));\n }\n \n // 呼び出し元で空白を除去済み。長い先頭ゼロと末尾は境界付き処理へ戻す。\n inline std::uint32_t read_u32_digits(const char*& source, const char* end) {\n if (source == end) return 0;\n const bool negative = *source == '-';\n if (negative || *source == '+') ++source;\n unsigned long long value = 0;\n if (end - source >= 16) {\n if (has_eight_digits(source)) {\n value = parse_eight_digits_simd(source);\n const unsigned ninth = digit_at(source + 8);\n const unsigned tenth = digit_at(source + 9);\n if (ninth >= 10U) {\n source += 8;\n } else if (tenth >= 10U) {\n value = value * 10ULL + ninth;\n source += 9;\n } else if (digit_at(source + 10) >= 10U) {\n value = value * 100ULL + ninth * 10U + tenth;\n source += 10;\n } else {\n value = parse_unsigned_digits_bounded(source, end);\n }\n } else {\n for (;;) {\n const unsigned first = digit_at(source);\n if (first >= 10U) break;\n const unsigned second = digit_at(source + 1);\n if (second >= 10U) {\n value = value * 10ULL + first;\n ++source;\n break;\n }\n value = value * 100ULL + first * 10U + second;\n source += 2;\n }\n }\n } else {\n value = parse_unsigned_digits_bounded(source, end);\n }\n return static_cast<std::uint32_t>(negative ? 0ULL - value : value);\n }\n \n CPLIB_FASTIO_ALWAYS_INLINE std::uint32_t read_u32() {\n InputState& state = input_state();\n if (CPLIB_FASTIO_UNLIKELY(!state.initialized)) initialize(state);\n if (state.mapped != nullptr) {\n const char* source = state.mapped + state.cursor;\n const char* const end = state.mapped + state.length;\n skip_spaces(source, end);\n const std::uint32_t value = read_u32_digits(source, end);\n consume_separator(source, end);\n state.cursor = static_cast<std::size_t>(source - state.mapped);\n return value;\n }\n const char* source = state.buffer + state.cursor;\n const char* const end = state.buffer + state.length;\n skip_spaces(source, end);\n if (CPLIB_FASTIO_UNLIKELY(end - source <\n static_cast<std::ptrdiff_t>(safe_integer_bytes))) {\n return static_cast<std::uint32_t>(read_uint_stream(state));\n }\n const std::uint32_t value = read_u32_digits(source, end);\n // 長い先頭ゼロがrefill境界をまたぐ場合は、元の位置から読み直す。\n if (CPLIB_FASTIO_UNLIKELY(source == end)) {\n return static_cast<std::uint32_t>(read_uint_stream(state));\n }\n consume_separator(source, end);\n state.cursor = static_cast<std::size_t>(source - state.buffer);\n return value;\n }\n \n inline bool try_read_four_nine_digit_u32(\n const char*& source, const char* end, std::uint32_t* output) {\n if (end - source < 40) return false;\n if (!fastio_is_space(source[9]) || !fastio_is_space(source[19]) ||\n !fastio_is_space(source[29]) || !fastio_is_space(source[39])) {\n return false;\n }\n const __m128i first_two = _mm_unpacklo_epi64(\n _mm_loadl_epi64(reinterpret_cast<const __m128i*>(source)),\n _mm_loadl_epi64(reinterpret_cast<const __m128i*>(source + 10)));\n const __m128i last_two = _mm_unpacklo_epi64(\n _mm_loadl_epi64(reinterpret_cast<const __m128i*>(source + 20)),\n _mm_loadl_epi64(reinterpret_cast<const __m128i*>(source + 30)));\n const __m256i bytes = _mm256_set_m128i(last_two, first_two);\n const __m256i is_digit = _mm256_and_si256(\n _mm256_cmpgt_epi8(bytes, _mm256_set1_epi8('/')),\n _mm256_cmpgt_epi8(_mm256_set1_epi8(':'), bytes));\n if (static_cast<unsigned>(_mm256_movemask_epi8(is_digit)) !=\n 0xffffffffU) {\n return false;\n }\n \n const unsigned digit0 = digit_at(source + 8);\n const unsigned digit1 = digit_at(source + 18);\n const unsigned digit2 = digit_at(source + 28);\n const unsigned digit3 = digit_at(source + 38);\n const bool ninth_digits = (digit0 < 10U) & (digit1 < 10U) &\n (digit2 < 10U) & (digit3 < 10U);\n if (!ninth_digits) return false;\n \n const __m256i digits =\n _mm256_sub_epi8(bytes, _mm256_set1_epi8('0'));\n const __m256i pairs = _mm256_maddubs_epi16(\n digits, _mm256_setr_epi8(\n 10, 1, 10, 1, 10, 1, 10, 1,\n 10, 1, 10, 1, 10, 1, 10, 1,\n 10, 1, 10, 1, 10, 1, 10, 1,\n 10, 1, 10, 1, 10, 1, 10, 1));\n const __m256i quads = _mm256_madd_epi16(\n pairs, _mm256_setr_epi16(\n 100, 1, 100, 1, 100, 1, 100, 1,\n 100, 1, 100, 1, 100, 1, 100, 1));\n const __m256i weighted = _mm256_mullo_epi32(\n quads, _mm256_setr_epi32(\n 10000, 1, 10000, 1, 10000, 1, 10000, 1));\n const __m256i sums =\n _mm256_hadd_epi32(weighted, _mm256_setzero_si256());\n const __m128i first_eight = _mm_unpacklo_epi64(\n _mm256_castsi256_si128(sums),\n _mm256_extracti128_si256(sums, 1));\n const __m128i values = _mm_add_epi32(\n _mm_mullo_epi32(first_eight, _mm_set1_epi32(10)),\n _mm_setr_epi32(static_cast<int>(digit0),\n static_cast<int>(digit1),\n static_cast<int>(digit2),\n static_cast<int>(digit3)));\n _mm_storeu_si128(reinterpret_cast<__m128i*>(output), values);\n source += 40;\n return true;\n }\n \n inline void read_uint_array(std::uint32_t* output, std::size_t count) {\n InputState& state = input_state();\n if (CPLIB_FASTIO_UNLIKELY(!state.initialized)) initialize(state);\n if (state.mapped != nullptr) {\n const char* source = state.mapped + state.cursor;\n const char* const end = state.mapped + state.length;\n skip_spaces(source, end);\n std::size_t i = 0;\n unsigned nine_digit_count = 0;\n // 最初の16要素を読みながら分布を確認し、短い整数ではAVX判定を省く。\n if (count >= 64) {\n for (; i < 16; ++i) {\n nine_digit_count += end - source >= 10 &&\n fastio_is_space(source[9]) && has_eight_digits(source) &&\n digit_at(source + 8) < 10U;\n output[i] = read_u32_digits(source, end);\n consume_separator(source, end);\n }\n }\n if (nine_digit_count >= 12) {\n while (i < count) {\n if (count - i >= 4 && try_read_four_nine_digit_u32(\n source, end, output + i)) {\n i += 4;\n skip_spaces(source, end);\n } else {\n output[i++] = read_u32_digits(source, end);\n if (i != count) consume_separator(source, end);\n }\n }\n } else {\n while (i < count) {\n output[i++] = read_u32_digits(source, end);\n if (i != count) consume_separator(source, end);\n }\n }\n state.cursor = static_cast<std::size_t>(source - state.mapped);\n } else {\n std::size_t i = 0;\n unsigned nine_digit_count = 0;\n if (count >= 64) {\n for (; i < 16; ++i) {\n output[i] = static_cast<std::uint32_t>(read_uint_stream(state));\n nine_digit_count += output[i] >= 100000000U &&\n output[i] < 1000000000U;\n }\n }\n if (nine_digit_count >= 12) {\n while (count - i >= 4) {\n const char* source = state.buffer + state.cursor;\n const char* const end = state.buffer + state.length;\n skip_spaces(source, end);\n if (try_read_four_nine_digit_u32(source, end, output + i)) {\n state.cursor = static_cast<std::size_t>(source - state.buffer);\n i += 4;\n } else {\n state.cursor = static_cast<std::size_t>(source - state.buffer);\n // refillをまたぐ整数は既存のストリーム処理で最後まで読む。\n output[i++] = static_cast<std::uint32_t>(read_uint_stream(state));\n }\n }\n }\n for (; i < count; ++i) {\n output[i] = static_cast<std::uint32_t>(read_uint_stream(state));\n }\n }\n }\n \n template <class T>\n inline void read_uint_array(T* output, std::size_t count) {\n InputState& state = input_state();\n if (CPLIB_FASTIO_UNLIKELY(!state.initialized)) initialize(state);\n if (state.mapped != nullptr) {\n const char* source = state.mapped + state.cursor;\n const char* const end = state.mapped + state.length;\n for (std::size_t i = 0; i < count; ++i) {\n output[i] = static_cast<T>(read_uint_mapped_at(source, end));\n }\n state.cursor = static_cast<std::size_t>(source - state.mapped);\n } else {\n for (std::size_t i = 0; i < count; ++i) {\n output[i] = static_cast<T>(read_uint_stream(state));\n }\n }\n }\n \n inline const char* read_token(std::size_t* output_length) {\n InputState& state = input_state();\n if (CPLIB_FASTIO_UNLIKELY(!state.initialized)) initialize(state);\n if (state.mapped != nullptr) {\n const char* source = state.mapped + state.cursor;\n const char* const end = state.mapped + state.length;\n while (source != end &&\n static_cast<unsigned char>(*source) <=\n static_cast<unsigned char>(' ')) {\n ++source;\n }\n const char* const token = source;\n while (source != end &&\n static_cast<unsigned char>(*source) >\n static_cast<unsigned char>(' ')) {\n ++source;\n }\n state.cursor = static_cast<std::size_t>(source - state.mapped);\n *output_length = static_cast<std::size_t>(source - token);\n return token;\n }\n \n for (;;) {\n if (state.cursor == state.length && !refill(state)) {\n *output_length = 0;\n return \"\";\n }\n while (state.cursor != state.length &&\n static_cast<unsigned char>(state.buffer[state.cursor]) <=\n static_cast<unsigned char>(' ')) {\n ++state.cursor;\n }\n if (state.cursor != state.length) break;\n }\n \n const std::size_t token_begin = state.cursor;\n while (state.cursor != state.length &&\n static_cast<unsigned char>(state.buffer[state.cursor]) >\n static_cast<unsigned char>(' ')) {\n ++state.cursor;\n }\n if (state.cursor != state.length) {\n *output_length = state.cursor - token_begin;\n return state.buffer + token_begin;\n }\n \n std::string& storage = token_storage();\n storage.assign(state.buffer + token_begin,\n state.length - token_begin);\n while (refill(state)) {\n while (state.cursor != state.length &&\n static_cast<unsigned char>(state.buffer[state.cursor]) >\n static_cast<unsigned char>(' ')) {\n ++state.cursor;\n }\n storage.append(state.buffer, state.cursor);\n if (state.cursor != state.length) break;\n }\n *output_length = storage.size();\n return storage.c_str();\n }\n \n #undef CPLIB_FASTIO_UNLIKELY\n #undef CPLIB_FASTIO_ALWAYS_INLINE\n } // namespace cplib_fastio_input\n \"\"\".}\n \n proc fastioGetChar(): cint {.importcpp: \"cplib_fastio_input::get_char()\", nodecl, inline.}\n proc fastioReadInt(): clonglong {.importcpp: \"cplib_fastio_input::read_int()\", nodecl, inline.}\n proc fastioReadUInt(): culonglong {.importcpp: \"cplib_fastio_input::read_uint()\", nodecl, inline.}\n proc fastioReadUInt32(): uint32 {.importcpp: \"cplib_fastio_input::read_u32()\", nodecl, inline.}\n proc fastioReadSignedArray[T: SomeSignedInt](values: ptr T, count: csize_t) {.importcpp: \"cplib_fastio_input::read_int_array(@)\", nodecl, inline.}\n proc fastioReadUnsignedArray[T: SomeUnsignedInt](values: ptr T, count: csize_t) {.importcpp: \"cplib_fastio_input::read_uint_array(@)\", nodecl, inline.}\n proc fastioReadToken(length: ptr csize_t): cstring\n {.importcpp: \"cplib_fastio_input::read_token(@)\", nodecl, inline.}\n \n type FastioInteger = int | int8 | int16 | int32 | int64 |\n uint | uint8 | uint16 | uint32 | uint64\n \n when NimMajor >= 2:\n template fastioNewSeqUninit(T: typedesc, length: int): untyped =\n newSeqUninit[T](length)\n \n template fastioNewStringUninit(length: int): untyped =\n newStringUninit(length)\n else:\n template fastioNewSeqUninit(T: typedesc, length: int): untyped =\n newSeqUninitialized[T](length)\n \n template fastioNewStringUninit(length: int): untyped =\n newString(length)\n \n proc ii(): int {.inline,\n codegenDecl: \"CPLIB_FASTIO_NIM_ALWAYS_INLINE $# $#$#\".} =\n fastioReadInt().int\n proc lii(N: int): seq[int] {.inline.} =\n result = fastioNewSeqUninit(int, N)\n if N > 0:\n fastioReadSignedArray(addr result[0], N.csize_t)\n \n # 型だけなら1要素、長さと型ならseqとして読み込む。\n proc input[T: FastioInteger](valueType: typedesc[T]): T {.inline,\n codegenDecl: \"CPLIB_FASTIO_NIM_ALWAYS_INLINE $# $#$#\".} =\n when T is range:\n {.error: \"input supports only primitive integer types\".}\n elif T is SomeSignedInt:\n T(fastioReadInt())\n elif T is uint32:\n fastioReadUInt32()\n else:\n T(fastioReadUInt())\n \n proc input[T: FastioInteger](N: int, valueType: typedesc[T]): seq[T] {.inline.} =\n when T is range:\n {.error: \"input supports only primitive integer types\".}\n else:\n result = fastioNewSeqUninit(T, N)\n if N > 0:\n when T is SomeSignedInt:\n fastioReadSignedArray(addr result[0], N.csize_t)\n else:\n fastioReadUnsignedArray(addr result[0], N.csize_t)\n \n proc si(): string {.inline.} =\n var length: csize_t\n let source = fastioReadToken(addr length)\n result = fastioNewStringUninit(length.int)\n if length != 0:\n copyMem(addr result[0], source, length.int)\n \n # 出力系\n {.emit: \"\"\"\n #include <cstdio>\n #include <cstddef>\n #include <cstdint>\n #include <cstring>\n #include <type_traits>\n \n namespace cplib_fastio_output {\n struct StdoutBuffer {\n static constexpr std::size_t capacity = 1U << 16;\n alignas(64) char data[capacity];\n \n StdoutBuffer() { setvbuf(stdout, data, _IOFBF, capacity); }\n };\n \n static StdoutBuffer stdout_buffer;\n \n struct FourDigits {\n char data[10000][4];\n constexpr FourDigits() : data{} {\n for (unsigned i = 0; i < 10000; ++i) {\n data[i][0] = static_cast<char>('0' + i / 1000);\n data[i][1] = static_cast<char>('0' + i / 100 % 10);\n data[i][2] = static_cast<char>('0' + i / 10 % 10);\n data[i][3] = static_cast<char>('0' + i % 10);\n }\n }\n };\n \n constexpr FourDigits four_digit_table{};\n \n inline const FourDigits& four_digits() {\n return four_digit_table;\n }\n \n inline char* reserve_bytes(std::FILE* output, std::size_t size) {\n #if defined(__GLIBC__)\n // stdio自身のバッファを使い、echo/write/flushFileとの出力順を保つ。\n if (output->_IO_write_ptr != nullptr &&\n output->_IO_write_end != nullptr &&\n static_cast<std::size_t>(output->_IO_write_end -\n output->_IO_write_ptr) >= size) {\n return output->_IO_write_ptr;\n }\n #endif\n return nullptr;\n }\n \n inline void commit_bytes(std::FILE* output, char* end) {\n #if defined(__GLIBC__)\n output->_IO_write_ptr = end;\n #else\n (void)output;\n (void)end;\n #endif\n }\n \n inline char* write_small(char* output, unsigned value,\n const FourDigits& table) {\n if (value >= 1000) {\n std::memcpy(output, table.data[value], 4);\n return output + 4;\n }\n if (value >= 100) {\n std::memcpy(output, table.data[value] + 1, 3);\n return output + 3;\n }\n if (value >= 10) {\n std::memcpy(output, table.data[value] + 2, 2);\n return output + 2;\n }\n *output++ = static_cast<char>('0' + value);\n return output;\n }\n \n inline char* write_four(char* output, unsigned value,\n const FourDigits& table) {\n std::memcpy(output, table.data[value], 4);\n return output + 4;\n }\n \n inline char* write_unsigned_32(char* output, std::uint32_t value,\n const FourDigits& table) {\n if (value < 10000U) return write_small(output, value, table);\n const std::uint32_t quotient = value / 10000U;\n const unsigned low = static_cast<unsigned>(value - quotient * 10000U);\n if (quotient < 10000U) {\n output = write_small(output, quotient, table);\n return write_four(output, low, table);\n }\n const unsigned high = quotient / 10000U;\n const unsigned middle = quotient - high * 10000U;\n output = write_small(output, high, table);\n output = write_four(output, middle, table);\n return write_four(output, low, table);\n }\n \n inline char* write_unsigned_64(char* output, std::uint64_t value,\n const FourDigits& table) {\n if (value < 10000ULL) {\n return write_small(output, static_cast<unsigned>(value), table);\n }\n const std::uint64_t quotient1 = value / 10000ULL;\n const unsigned chunk1 = static_cast<unsigned>(value - quotient1 * 10000ULL);\n if (quotient1 < 10000ULL) {\n output = write_small(output, static_cast<unsigned>(quotient1), table);\n return write_four(output, chunk1, table);\n }\n const std::uint64_t quotient2 = quotient1 / 10000ULL;\n const unsigned chunk2 = static_cast<unsigned>(quotient1 - quotient2 * 10000ULL);\n if (quotient2 < 10000ULL) {\n output = write_small(output, static_cast<unsigned>(quotient2), table);\n output = write_four(output, chunk2, table);\n return write_four(output, chunk1, table);\n }\n const std::uint64_t quotient3 = quotient2 / 10000ULL;\n const unsigned chunk3 = static_cast<unsigned>(quotient2 - quotient3 * 10000ULL);\n if (quotient3 < 10000ULL) {\n output = write_small(output, static_cast<unsigned>(quotient3), table);\n output = write_four(output, chunk3, table);\n output = write_four(output, chunk2, table);\n return write_four(output, chunk1, table);\n }\n const std::uint64_t quotient4 = quotient3 / 10000ULL;\n const unsigned chunk4 = static_cast<unsigned>(quotient3 - quotient4 * 10000ULL);\n output = write_small(output, static_cast<unsigned>(quotient4), table);\n output = write_four(output, chunk4, table);\n output = write_four(output, chunk3, table);\n output = write_four(output, chunk2, table);\n return write_four(output, chunk1, table);\n }\n \n template <class Unsigned>\n inline char* write_unsigned_dispatch(char* output, Unsigned value,\n const FourDigits& table,\n std::true_type) {\n return write_unsigned_32(output, static_cast<std::uint32_t>(value), table);\n }\n \n template <class Unsigned>\n inline char* write_unsigned_dispatch(char* output, Unsigned value,\n const FourDigits& table,\n std::false_type) {\n return write_unsigned_64(output, static_cast<std::uint64_t>(value), table);\n }\n \n template <class Unsigned>\n inline char* write_unsigned(char* output, Unsigned value,\n const FourDigits& table) {\n return write_unsigned_dispatch(output, value, table,\n std::integral_constant<bool, (sizeof(Unsigned) <= 4)>{});\n }\n \n template <class Integer>\n inline std::size_t join_integers(\n const Integer* values, std::size_t count, char* output,\n const char* separator, std::size_t separator_length) {\n const FourDigits& table = four_digits();\n char* cursor = output;\n using Unsigned = typename std::make_unsigned<Integer>::type;\n for (std::size_t i = 0; i < count; ++i) {\n const Integer value = values[i];\n Unsigned magnitude = static_cast<Unsigned>(value);\n if (std::is_signed<Integer>::value && value < 0) {\n *cursor++ = '-';\n magnitude = Unsigned(0) - magnitude;\n }\n cursor = write_unsigned(cursor, magnitude, table);\n if (i + 1 != count) {\n std::memcpy(cursor, separator, separator_length);\n cursor += separator_length;\n }\n }\n return static_cast<std::size_t>(cursor - output);\n }\n \n class BufferedWriter {\n public:\n static constexpr std::size_t capacity = 1U << 16;\n \n explicit BufferedWriter(std::FILE* output)\n : length_(0), output_(output) {}\n \n inline char* reserve_integer() {\n constexpr std::size_t max_integer_length = 21;\n if (capacity - length_ < max_integer_length) flush();\n return data_ + length_;\n }\n \n inline void commit(char* end) {\n length_ = static_cast<std::size_t>(end - data_);\n }\n \n inline void append(const char* source, std::size_t size) {\n if (size <= capacity - length_) {\n std::memcpy(data_ + length_, source, size);\n length_ += size;\n return;\n }\n flush();\n if (size >= capacity) {\n fwrite_unlocked(source, 1, size, output_);\n } else {\n std::memcpy(data_, source, size);\n length_ = size;\n }\n }\n \n inline void flush() {\n if (length_ != 0) {\n fwrite_unlocked(data_, 1, length_, output_);\n length_ = 0;\n }\n }\n \n private:\n char data_[capacity];\n std::size_t length_;\n std::FILE* output_;\n };\n \n template <class Integer>\n inline void print_integers(std::FILE* output, const Integer* values,\n std::size_t count,\n const char* separator,\n std::size_t separator_length) {\n const FourDigits& table = four_digits();\n BufferedWriter writer(output);\n using Unsigned = typename std::make_unsigned<Integer>::type;\n for (std::size_t i = 0; i < count; ++i) {\n char* cursor = writer.reserve_integer();\n const Integer value = values[i];\n Unsigned magnitude = static_cast<Unsigned>(value);\n if (std::is_signed<Integer>::value && value < 0) {\n *cursor++ = '-';\n magnitude = Unsigned(0) - magnitude;\n }\n cursor = write_unsigned(cursor, magnitude, table);\n writer.commit(cursor);\n if (i + 1 != count) writer.append(separator, separator_length);\n }\n writer.append(\"\\n\", 1);\n writer.flush();\n }\n \n template <class Integer>\n inline void print_one(std::FILE* output, Integer value) {\n const FourDigits& table = four_digits();\n char buffer[22];\n char* const reserved = reserve_bytes(output, sizeof(buffer));\n char* const begin = reserved == nullptr ? buffer : reserved;\n char* cursor = begin;\n using Unsigned = typename std::make_unsigned<Integer>::type;\n Unsigned magnitude = static_cast<Unsigned>(value);\n if (std::is_signed<Integer>::value && value < 0) {\n *cursor++ = '-';\n magnitude = Unsigned(0) - magnitude;\n }\n cursor = write_unsigned(cursor, magnitude, table);\n *cursor++ = '\\n';\n if (reserved != nullptr) {\n commit_bytes(output, cursor);\n } else {\n fwrite_unlocked(buffer, 1,\n static_cast<std::size_t>(cursor - buffer), output);\n }\n }\n \n } // namespace cplib_fastio_output\n \"\"\".}\n \n proc fastioJoinInts[T: SomeInteger](values: ptr T, count: csize_t,\n output: ptr char, separator: cstring, separatorLen: csize_t): csize_t\n {.importcpp: \"cplib_fastio_output::join_integers(@)\", nodecl.}\n proc fastioPrintInts[T: SomeInteger](output: File, values: ptr T,\n count: csize_t, separator: cstring, separatorLen: csize_t)\n {.importcpp: \"cplib_fastio_output::print_integers(@)\", nodecl.}\n proc fastioPrintInt[T: SomeInteger](output: File, value: T)\n {.importcpp: \"cplib_fastio_output::print_one(@)\", nodecl.}\n \n proc print_internal(prop: tuple[f: File, sepc: string, endc: string,\n flush: bool], args: openArray[string]) =\n for i in 0 ..< args.len:\n prop.f.write(args[i])\n if i != args.len - 1:\n prop.f.write(prop.sepc)\n else:\n prop.f.write(prop.endc)\n if prop.flush:\n prop.f.flushFile()\n \n proc print*(prop: tuple[f: File, sepc: string, endc: string, flush: bool],\n args: varargs[string, `$`]) =\n print_internal(prop, args)\n \n proc fastioPrintWithSeparator(sep: string, args: varargs[string, `$`]) =\n print_internal((f: stdout, sepc: sep, endc: \"\\n\", flush: false), args)\n \n proc fastioAppendInteger[T: SomeInteger](destination: var string, value: T) =\n var magnitude: uint64\n when T is SomeSignedInt:\n let signedValue = value.int64\n if signedValue < 0:\n destination.add('-')\n magnitude = uint64(-(signedValue + 1)) + 1'u64\n else:\n magnitude = signedValue.uint64\n else:\n magnitude = value.uint64\n \n if magnitude == 0:\n destination.add('0')\n return\n var digits: array[20, char]\n var count = 0\n while magnitude != 0:\n digits[count] = char(ord('0') + int(magnitude mod 10))\n magnitude = magnitude div 10\n inc count\n while count != 0:\n dec count\n destination.add(digits[count])\n \n # 整数は4桁テーブルを使うC++フォーマッタへまとめて渡す。\n proc fastioJoinImpl[T](a: openArray[T], sep: string): string =\n when nimvm:\n result = newStringOfCap(a.len * 4)\n for i, value in a:\n if i != 0:\n result.add(sep)\n when T is SomeInteger:\n fastioAppendInteger(result, value)\n else:\n result.add($value)\n else:\n if a.len == 0:\n return \"\"\n when T is SomeInteger and sizeof(T) in [4, 8]:\n const digits = when sizeof(T) == 8: 20\n elif T is SomeSignedInt: 11\n else: 10\n result = fastioNewStringUninit(\n a.len * digits + (a.len - 1) * sep.len)\n let written = fastioJoinInts(unsafeAddr a[0],\n a.len.csize_t, addr result[0], sep.cstring, sep.len.csize_t)\n result.setLen(written.int)\n elif compiles(T.umod()) and compiles(a[0].val()):\n # Montgomery表現を含め、公開値へ正規化してから一括変換する。\n var canonical = fastioNewSeqUninit(uint32, a.len)\n for i, value in a:\n canonical[i] = value.val.uint32\n result = fastioJoinImpl(canonical, sep)\n else:\n result = newStringOfCap(a.len * 4)\n for i, value in a:\n if i != 0:\n result.add(sep)\n when T is SomeInteger:\n fastioAppendInteger(result, value)\n else:\n result.add($value)\n \n proc join*[T: not string](a: openArray[T], sep: string = \"\"): string {.inline.} =\n fastioJoinImpl(a, sep)\n \n proc fastioPrintArrayImpl[T](a: openArray[T], sep: string) =\n if a.len == 0:\n stdout.write('\\n')\n return\n when T is SomeInteger and sizeof(T) in [4, 8]:\n fastioPrintInts(stdout, unsafeAddr a[0], a.len.csize_t,\n sep.cstring, sep.len.csize_t)\n elif compiles(T.umod()) and compiles(a[0].val()):\n var canonical = fastioNewSeqUninit(uint32, a.len)\n for i, value in a:\n canonical[i] = value.val.uint32\n fastioPrintArrayImpl(canonical, sep)\n else:\n stdout.write(fastioJoinImpl(a, sep))\n stdout.write('\\n')\n \n proc fastioPrintOneImpl[T](value: T, sep: string) =\n when T is SomeInteger and sizeof(T) in [4, 8]:\n fastioPrintInt(stdout, value)\n elif compiles(T.umod()) and compiles(value.val()):\n fastioPrintInt(stdout, value.val.uint32)\n else:\n fastioPrintWithSeparator(sep, value)\n \n proc fastioPrintIntegerMany[T: SomeInteger](sep: string,\n values: varargs[T]) =\n fastioPrintArrayImpl(values, sep)\n \n # Python風に print(*X) と書くと、Xを空白区切りで1行に出力する。\n template `*`*[T](values: openArray[T]): string =\n fastioJoinImpl(values, \" \")\n \n # 最後の文字列引数をsepと誤認しないよう、名前付きsepはマクロで処理する。\n macro print*(args: varargs[untyped]): untyped =\n var sep = newLit(\" \")\n var hasSep = false\n var values: seq[NimNode]\n for arg in args:\n if arg.kind == nnkExprEqExpr and arg[0].eqIdent(\"sep\"):\n if hasSep:\n error(\"sep can only be specified once\", arg)\n sep = arg[1]\n hasSep = true\n else:\n values.add(arg)\n var splatValues: NimNode\n if values.len == 1:\n if values[0].kind == nnkPrefix and values[0][0].eqIdent(\"*\"):\n splatValues = values[0][1]\n elif values[0].kind in nnkCallKinds and values[0].len == 3 and\n values[0][0].eqIdent(\"fastioJoinImpl\"):\n # オーバーロード解決時に *values が先に展開された場合。\n splatValues = values[0][1]\n if not splatValues.isNil:\n result = newCall(bindSym\"fastioPrintArrayImpl\", splatValues, sep)\n elif values.len == 1:\n result = newCall(bindSym\"fastioPrintOneImpl\", values[0], sep)\n else:\n let integerCall = newCall(bindSym\"fastioPrintIntegerMany\", sep)\n let fallbackCall = newCall(bindSym\"fastioPrintWithSeparator\", sep)\n for value in values:\n integerCall.add(value)\n fallbackCall.add(value)\n result = quote do:\n when compiles(`integerCall`):\n `integerCall`\n else:\n `fallbackCall`\n \n macro getSymbolName(x: typed): string = x.toStrLit\n macro debug*(args: varargs[untyped]): untyped =\n when defined(debug):\n result = newNimNode(nnkStmtList, args)\n template prop(e: string = \"\"): untyped = (f: stderr, sepc: \"\", endc: e, flush: true)\n for i, arg in args:\n if arg.kind == nnkStrLit:\n result.add(quote do: print(prop(), \"\\\"\", `arg`, \"\\\"\"))\n else:\n result.add(quote do: print(prop(\": \"), getSymbolName(`arg`)))\n result.add(quote do: print(prop(), `arg`))\n if i != args.len - 1: result.add(quote do: print(prop(), \", \"))\n else: result.add(quote do: print(prop(), \"\\n\"))\n else:\n return (quote do: discard)\n #chmin,chmax\n template `max=`(x, y) =\n let yVal = y # yが計算式の場合に評価を1回にするため\n if x < yVal:\n x = yVal\n\n template `min=`(x, y) =\n let yVal = y\n if x > yVal:\n x = yVal\n proc chmin[T](x: var T, y: T):bool=\n if x > y:\n x = y\n return true\n return false\n proc chmax[T](x: var T, y: T):bool=\n if x < y:\n x = y\n return true\n return false\n #bit演算\n proc `%`*(x: int, y: int): int =\n result = x mod y\n if y > 0 and result < 0: result += y\n if y < 0 and result > 0: result += y\n proc `//`*(x: int, y: int): int{.inline.} =\n result = x div y\n if y > 0 and result * y > x: result -= 1\n if y < 0 and result * y < x: result -= 1\n proc `%=`(x: var int, y: int): void = x = x%y\n proc `//=`(x: var int, y: int): void = x = x//y\n proc `**`(x: int, y: int): int = x^y\n proc `**=`(x: var int, y: int): void = x = x^y\n proc `^`(x: int, y: int): int = x xor y\n proc `|`(x: int, y: int): int = x or y\n proc `&`(x: int, y: int): int = x and y\n proc `>>`(x: int, y: int): int = x shr y\n proc `<<`(x: int, y: int): int = x shl y\n proc `~`(x: int): int = not x\n proc `^=`(x: var int, y: int): void = x = x ^ y\n proc `&=`(x: var int, y: int): void = x = x & y\n proc `|=`(x: var int, y: int): void = x = x | y\n proc `>>=`(x: var int, y: int): void = x = x >> y\n proc `<<=`(x: var int, y: int): void = x = x << y\n proc `[]`(x: int, n: int): bool = (x and (1 shl n)) != 0\n #便利な変換\n proc `!`(x: char, a = '0'): int = int(x)-int(a)\n #定数\n when not declared CPLIB_UTILS_CONSTANTS:\n const CPLIB_UTILS_CONSTANTS* = 1\n const INF32*: int32 = 1001000027.int32\n const INF64*: int = int(3300300300300300491)\n \n const INF = INF64\n #converter\n\n #range\n iterator range(start: int, ends: int, step: int): int =\n var i = start\n if step < 0:\n while i > ends:\n yield i\n i += step\n elif step > 0:\n while i < ends:\n yield i\n i += step\n iterator range(ends: int): int = (for i in 0..<ends: yield i)\n iterator range(start: int, ends: int): int = (for i in\n start..<ends: yield i)\n\n proc dump[T](arr:seq[seq[T]])=\n for i in 0..<len(arr):\n echo arr[i]\n\n proc sum(slice:HSlice[int,int]):int=\n return (slice.a+slice.b)*len(slice)//2\n \n proc `<`[T](l,r:seq[T]):bool=\n for i in 0..<min(len(l),len(r)):\n if l[i] > r[i]:\n return false\n elif l[i] < r[i]:\n return true\n return len(l) < len(r)\n \n # Yes/No\n proc yes*(b: bool = true): void = print(if b: \"Yes\" else: \"No\")\n\n template dblock(body: untyped) =\n when defined(debug):\n block:\n body\n"
# source: src/cplib/tmpl/optimize.nim
ImportExpand "cplib/tmpl/optimize" <=== "when not declared CPLIB_TMPL_OPTIMIZE:\n const CPLIB_TMPL_OPTIMIZE* = 1\n import macros\n import strutils\n import std/compilesettings\n macro optimize*(arg: static string = \"\"\"nim c -d:danger -d:second_compile -d:useMalloc --gc:arc --panics:on --opt:speed --checks:off --passC:\"-flto -m64 -march=native -ffast-math\" --hints:off \"\"\") =\n let isSecond = defined(second_compile)\n let isDebug = defined(debug)\n\n if (not isSecond) and (not isDebug):\n if \"-d:second_compile\" notin arg:\n error(\"plz add -d:second_compile\")\n let sourcePath = querySetting(SingleValueSetting.projectFull)\n let projectDir = sourcePath[0..<sourcePath.rfind('/')]\n let outFile = querySetting(SingleValueSetting.outFile)\n let outFlag = if outFile.len > 0: \"-o:\" & outFile & \" \" else: \"-o:a.out \"\n var cmd = \"cd \" & projectDir & \" && export PATH=$HOME/.nimble/bin:$PATH && \" & arg\n cmd.add(outFlag & sourcePath)\n\n echo \"--- Self-Recompiling with optimized settings ---\"\n echo \"Command: \", cmd\n echo \"\\n\\n\\n\"\n\n let output = staticExec(cmd)\n warning(output)\n echo \"\\n\\n\\n\"\n\n quit(0)"
{.define: fastioNoMmap.}
optimize("""nim cpp -d:danger -d:second_compile -d:useMalloc --gc:none --panics:on --opt:speed --checks:off --passC:"-O3 -flto -m64 -march=native -ffast-math" --hints:off """)
var H,W = ii()
var A = input(H*W,uint32)
var S = newseqwith(H,uint32(0))
for i in 0..<(H):
for j in 0..<(W):
S[i] += A[i*W+j]
var T = S.sum()
S.applyit((it + T))
print(S.join("\n"))