/* * xxhsum - Command line interface for xxhash algorithms * Copyright (C) 2013-2020 Yann Collet * * GPL v2 License * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by * the Free Software Foundation; either version 2 of the License, or * (at your option) any later version. * * This program is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU General Public License for more details. * * You should have received a copy of the GNU General Public License along * with this program; if not, write to the Free Software Foundation, Inc., * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. * * You can contact the author at: * - xxHash homepage: https://www.xxhash.com * - xxHash source repository: https://github.com/Cyan4973/xxHash */ /* * xxhsum: * Provides hash value of a file content, or a list of files, or stdin * Display convention is Big Endian, for both 32 and 64 bits algorithms */ /* ************************************ * Compiler Options **************************************/ /* MS Visual */ #if defined(_MSC_VER) || defined(_WIN32) # define _CRT_SECURE_NO_WARNINGS /* removes visual warnings */ #endif /* Under Linux at least, pull in the *64 commands */ #ifndef _LARGEFILE64_SOURCE # define _LARGEFILE64_SOURCE #endif /* ************************************ * Includes **************************************/ #include #include /* malloc, calloc, free, exit */ #include /* strcmp, memcpy */ #include /* fprintf, fopen, ftello64, fread, stdin, stdout, _fileno (when present) */ #include /* stat, stat64, _stat64 */ #include /* stat, stat64, _stat64 */ #include /* clock_t, clock, CLOCKS_PER_SEC */ #include /* assert */ #include /* errno */ #define XXH_STATIC_LINKING_ONLY /* *_state_t */ #include "xxhash.h" #ifdef XXHSUM_DISPATCH # include "xxh_x86dispatch.h" #endif /* ************************************ * OS-Specific Includes **************************************/ #if !defined(_WIN32) && (defined(__unix__) || defined(__unix) || (defined(__APPLE__) && defined(__MACH__)) /* UNIX-like OS */ \ || defined(__midipix__) || defined(__VMS)) # if (defined(__APPLE__) && defined(__MACH__)) || defined(__SVR4) || defined(_AIX) || defined(__hpux) /* POSIX.1-2001 (SUSv3) conformant */ \ || defined(__DragonFly__) || defined(__FreeBSD__) || defined(__NetBSD__) || defined(__OpenBSD__) /* BSD distros */ # define PLATFORM_POSIX_VERSION 200112L # else # if defined(__linux__) || defined(__linux) # ifndef _POSIX_C_SOURCE # define _POSIX_C_SOURCE 200112L /* use feature test macro */ # endif # endif # include /* declares _POSIX_VERSION */ # if defined(_POSIX_VERSION) /* POSIX compliant */ # define PLATFORM_POSIX_VERSION _POSIX_VERSION # else # define PLATFORM_POSIX_VERSION 0 # endif # endif #endif #if !defined(PLATFORM_POSIX_VERSION) # define PLATFORM_POSIX_VERSION -1 #endif #if (defined(__linux__) && (PLATFORM_POSIX_VERSION >= 1)) \ || (PLATFORM_POSIX_VERSION >= 200112L) \ || defined(__DJGPP__) \ || defined(__MSYS__) # include /* isatty */ # define IS_CONSOLE(stdStream) isatty(fileno(stdStream)) #elif defined(MSDOS) || defined(OS2) # include /* _isatty */ # define IS_CONSOLE(stdStream) _isatty(_fileno(stdStream)) #elif defined(WIN32) || defined(_WIN32) # include /* _isatty */ # include /* DeviceIoControl, HANDLE, FSCTL_SET_SPARSE */ # include /* FILE */ static __inline int IS_CONSOLE(FILE* stdStream) { DWORD dummy; return _isatty(_fileno(stdStream)) && GetConsoleMode((HANDLE)_get_osfhandle(_fileno(stdStream)), &dummy); } #else # define IS_CONSOLE(stdStream) 0 #endif #if defined(MSDOS) || defined(OS2) || defined(WIN32) || defined(_WIN32) # include /* _O_BINARY */ # include /* _setmode, _fileno, _get_osfhandle */ # if !defined(__DJGPP__) # include /* DeviceIoControl, HANDLE, FSCTL_SET_SPARSE */ # include /* FSCTL_SET_SPARSE */ # define SET_BINARY_MODE(file) { int const unused=_setmode(_fileno(file), _O_BINARY); (void)unused; } # else # define SET_BINARY_MODE(file) setmode(fileno(file), O_BINARY) # endif #else # define SET_BINARY_MODE(file) #endif #if !defined(S_ISREG) # define S_ISREG(x) (((x) & S_IFMT) == S_IFREG) #endif /* Unicode helpers for Windows to make UTF-8 act as it should. */ #ifdef _WIN32 /* * Converts a UTF-8 string to UTF-16. Acts like strdup. The string must be freed afterwards. * This version allows keeping the output length. */ static wchar_t* utf8_to_utf16_len(const char* str, int* lenOut) { int const len = MultiByteToWideChar(CP_UTF8, 0, str, -1, NULL, 0); if (lenOut != NULL) *lenOut = len; if (len == 0) return NULL; { wchar_t* buf = (wchar_t*)malloc((size_t)len * sizeof(wchar_t)); if (buf != NULL) { if (MultiByteToWideChar(CP_UTF8, 0, str, -1, buf, len) == 0) { free(buf); return NULL; } } return buf; } } /* Converts a UTF-8 string to UTF-16. Acts like strdup. The string must be freed afterwards. */ static wchar_t* utf8_to_utf16(const char *str) { return utf8_to_utf16_len(str, NULL); } /* * Converts a UTF-16 string to UTF-8. Acts like strdup. The string must be freed afterwards. * This version allows keeping the output length. */ static char* utf16_to_utf8_len(const wchar_t *str, int *lenOut) { int len = WideCharToMultiByte(CP_UTF8, 0, str, -1, NULL, 0, NULL, NULL); if (lenOut != NULL) *lenOut = len; if (len == 0) return NULL; { char* const buf = (char*)malloc((size_t)len * sizeof(char)); if (buf != NULL) { if (WideCharToMultiByte(CP_UTF8, 0, str, -1, buf, len, NULL, NULL) == 0) { free(buf); return NULL; } } return buf; } } /* Converts a UTF-16 string to UTF-8. Acts like strdup. The string must be freed afterwards. */ static char *utf16_to_utf8(const wchar_t *str) { return utf16_to_utf8_len(str, NULL); } /* * fopen wrapper that supports UTF-8 * * fopen will only accept ANSI filenames, which means that we can't open Unicode filenames. * * In order to open a Unicode filename, we need to convert filenames to UTF-16 and use _wfopen. */ static FILE* XXH_fopen_wrapped(const char *filename, const wchar_t *mode) { wchar_t* const wide_filename = utf8_to_utf16(filename); if (wide_filename == NULL) return NULL; { FILE* const f = _wfopen(wide_filename, mode); free(wide_filename); return f; } } /* * In case it isn't available, this is what MSVC 2019 defines in stdarg.h. */ #if defined(_MSC_VER) && !defined(__clang__) && !defined(va_copy) # define va_copy(destination, source) ((destination) = (source)) #endif /* * fprintf wrapper that supports UTF-8. * * fprintf doesn't properly handle Unicode on Windows. * * Additionally, it is codepage sensitive on console and may crash the program. * * Instead, we use vsnprintf, and either print with fwrite or convert to UTF-16 * for console output and use the codepage-independent WriteConsoleW. * * Credit to t-mat: https://github.com/t-mat/xxHash/commit/5691423 */ static int fprintf_utf8(FILE *stream, const char *format, ...) { int result; va_list args; va_list copy; va_start(args, format); /* * To be safe, make a va_copy. * * Note that Microsoft doesn't use va_copy in its sample code: * https://docs.microsoft.com/en-us/cpp/c-runtime-library/reference/vsprintf-vsprintf-l-vswprintf-vswprintf-l-vswprintf-l?view=vs-2019 */ va_copy(copy, args); /* Counts the number of characters needed for vsnprintf. */ result = _vscprintf(format, copy); va_end(copy); if (result > 0) { /* Create a buffer for vsnprintf */ const size_t nchar = (size_t)result + 1; char* u8_str = (char*)malloc(nchar * sizeof(u8_str[0])); if (u8_str == NULL) { result = -1; } else { /* Generate the UTF-8 string with vsnprintf. */ result = _vsnprintf(u8_str, nchar - 1, format, args); u8_str[nchar - 1] = '\0'; if (result > 0) { /* * Check if we are outputting to a console. Don't use IS_CONSOLE * directly -- we don't need to call _get_osfhandle twice. */ int fileNb = _fileno(stream); intptr_t handle_raw = _get_osfhandle(fileNb); HANDLE handle = (HANDLE)handle_raw; DWORD dwTemp; if (handle_raw < 0) { result = -1; } else if (_isatty(fileNb) && GetConsoleMode(handle, &dwTemp)) { /* * Convert to UTF-16 and output with WriteConsoleW. * * This is codepage independent and works on Windows XP's * default msvcrt.dll. */ int len; wchar_t *const u16_buf = utf8_to_utf16_len(u8_str, &len); if (u16_buf == NULL) { result = -1; } else { if (WriteConsoleW(handle, u16_buf, (DWORD)len - 1, &dwTemp, NULL)) { result = (int)dwTemp; } else { result = -1; } free(u16_buf); } } else { /* fwrite the UTF-8 string if we are printing to a file */ result = (int)fwrite(u8_str, 1, nchar - 1, stream); if (result == 0) { result = -1; } } } free(u8_str); } } va_end(args); return result; } /* * Since we always use literals in the "mode" argument, it is just easier to append "L" to * the string to make it UTF-16 and avoid the hassle of a second manual conversion. */ # define XXH_fopen(filename, mode) XXH_fopen_wrapped(filename, L##mode) #else # define XXH_fopen(filename, mode) fopen(filename, mode) #endif /* ************************************ * Basic Types **************************************/ #if defined(__cplusplus) /* C++ */ \ || (defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L) /* C99 */ # include typedef uint8_t U8; typedef uint32_t U32; typedef uint64_t U64; # else # include typedef unsigned char U8; # if UINT_MAX == 0xFFFFFFFFUL typedef unsigned int U32; # else typedef unsigned long U32; # endif typedef unsigned long long U64; #endif /* not C++/C99 */ static unsigned BMK_isLittleEndian(void) { const union { U32 u; U8 c[4]; } one = { 1 }; /* don't use static: performance detrimental */ return one.c[0]; } /* ************************************* * Constants ***************************************/ #define LIB_VERSION XXH_VERSION_MAJOR.XXH_VERSION_MINOR.XXH_VERSION_RELEASE #define QUOTE(str) #str #define EXPAND_AND_QUOTE(str) QUOTE(str) #define PROGRAM_VERSION EXPAND_AND_QUOTE(LIB_VERSION) /* Show compiler versions in WELCOME_MESSAGE. CC_VERSION_FMT will return the printf specifiers, * and VERSION will contain the comma separated list of arguments to the CC_VERSION_FMT string. */ #if defined(__clang_version__) /* Clang does its own thing. */ # ifdef __apple_build_version__ # define CC_VERSION_FMT "Apple Clang %s" # else # define CC_VERSION_FMT "Clang %s" # endif # define CC_VERSION __clang_version__ #elif defined(__VERSION__) /* GCC and ICC */ # define CC_VERSION_FMT "%s" # ifdef __INTEL_COMPILER /* icc adds its prefix */ # define CC_VERSION __VERSION__ # else /* assume GCC */ # define CC_VERSION "GCC " __VERSION__ # endif #elif defined(_MSC_FULL_VER) && defined(_MSC_BUILD) /* * MSVC * "For example, if the version number of the Visual C++ compiler is * 15.00.20706.01, the _MSC_FULL_VER macro evaluates to 150020706." * * https://docs.microsoft.com/en-us/cpp/preprocessor/predefined-macros?view=vs-2017 */ # define CC_VERSION_FMT "MSVC %02i.%02i.%05i.%02i" # define CC_VERSION _MSC_FULL_VER / 10000000 % 100, _MSC_FULL_VER / 100000 % 100, _MSC_FULL_VER % 100000, _MSC_BUILD #elif defined(__TINYC__) /* tcc stores its version in the __TINYC__ macro. */ # define CC_VERSION_FMT "tcc %i.%i.%i" # define CC_VERSION __TINYC__ / 10000 % 100, __TINYC__ / 100 % 100, __TINYC__ % 100 #else # define CC_VERSION_FMT "%s" # define CC_VERSION "unknown compiler" #endif /* makes the next part easier */ #if defined(__x86_64__) || defined(_M_AMD64) || defined(_M_X64) # define ARCH_X64 1 # define ARCH_X86 "x86_64" #elif defined(__i386__) || defined(_M_IX86) || defined(_M_IX86_FP) # define ARCH_X86 "i386" #endif /* Try to detect the architecture. */ #if defined(ARCH_X86) # if defined(XXHSUM_DISPATCH) # define ARCH ARCH_X86 " autoVec" # elif defined(__AVX512F__) # define ARCH ARCH_X86 " + AVX512" # elif defined(__AVX2__) # define ARCH ARCH_X86 " + AVX2" # elif defined(__AVX__) # define ARCH ARCH_X86 " + AVX" # elif defined(_M_X64) || defined(_M_AMD64) || defined(__x86_64__) \ || defined(__SSE2__) || (defined(_M_IX86_FP) && _M_IX86_FP == 2) # define ARCH ARCH_X86 " + SSE2" # else # define ARCH ARCH_X86 # endif #elif defined(__aarch64__) || defined(__arm64__) || defined(_M_ARM64) # define ARCH "aarch64 + NEON" #elif defined(__arm__) || defined(__thumb__) || defined(__thumb2__) || defined(_M_ARM) /* ARM has a lot of different features that can change xxHash significantly. */ # if defined(__thumb2__) || (defined(__thumb__) && (__thumb__ == 2 || __ARM_ARCH >= 7)) # define ARCH_THUMB " Thumb-2" # elif defined(__thumb__) # define ARCH_THUMB " Thumb-1" # else # define ARCH_THUMB "" # endif /* ARMv7 has unaligned by default */ # if defined(__ARM_FEATURE_UNALIGNED) || __ARM_ARCH >= 7 || defined(_M_ARMV7VE) # define ARCH_UNALIGNED " + unaligned" # else # define ARCH_UNALIGNED "" # endif # if defined(__ARM_NEON) || defined(__ARM_NEON__) # define ARCH_NEON " + NEON" # else # define ARCH_NEON "" # endif # define ARCH "ARMv" EXPAND_AND_QUOTE(__ARM_ARCH) ARCH_THUMB ARCH_NEON ARCH_UNALIGNED #elif defined(__powerpc64__) || defined(__ppc64__) || defined(__PPC64__) # if defined(__GNUC__) && defined(__POWER9_VECTOR__) # define ARCH "ppc64 + POWER9 vector" # elif defined(__GNUC__) && defined(__POWER8_VECTOR__) # define ARCH "ppc64 + POWER8 vector" # else # define ARCH "ppc64" # endif #elif defined(__powerpc__) || defined(__ppc__) || defined(__PPC__) # define ARCH "ppc" #elif defined(__AVR) # define ARCH "AVR" #elif defined(__mips64) # define ARCH "mips64" #elif defined(__mips) # define ARCH "mips" #elif defined(__s390x__) # define ARCH "s390x" #elif defined(__s390__) # define ARCH "s390" #else # define ARCH "unknown" #endif static const int g_nbBits = (int)(sizeof(void*)*8); static const char g_lename[] = "little endian"; static const char g_bename[] = "big endian"; #define ENDIAN_NAME (BMK_isLittleEndian() ? g_lename : g_bename) static const char author[] = "Yann Collet"; #define WELCOME_MESSAGE(exename) "%s %s by %s, compiled as %i-bit %s %s with " CC_VERSION_FMT " \n", \ exename, PROGRAM_VERSION, author, g_nbBits, ARCH, ENDIAN_NAME, CC_VERSION #define KB *( 1<<10) #define MB *( 1<<20) #define GB *(1U<<30) static size_t XXH_DEFAULT_SAMPLE_SIZE = 100 KB; #define NBLOOPS 3 /* Default number of benchmark iterations */ #define TIMELOOP_S 1 #define TIMELOOP (TIMELOOP_S * CLOCKS_PER_SEC) /* target timing per iteration */ #define TIMELOOP_MIN (TIMELOOP / 2) /* minimum timing to validate a result */ #define XXHSUM32_DEFAULT_SEED 0 /* Default seed for algo_xxh32 */ #define XXHSUM64_DEFAULT_SEED 0 /* Default seed for algo_xxh64 */ #define MAX_MEM (2 GB - 64 MB) static const char stdinName[] = "-"; typedef enum { algo_xxh32=0, algo_xxh64=1, algo_xxh128=2 } AlgoSelected; static const AlgoSelected g_defaultAlgo = algo_xxh64; /* required within main() & usage() */ /* <16 hex char> <'\0'> * '4096' is typical Linux PATH_MAX configuration. */ #define DEFAULT_LINE_LENGTH (sizeof(XXH64_hash_t) * 2 + 2 + 4096 + 1) /* Maximum acceptable line length. */ #define MAX_LINE_LENGTH (32 KB) /* ************************************ * Display macros **************************************/ #ifdef _WIN32 #define DISPLAY(...) fprintf_utf8(stderr, __VA_ARGS__) #define DISPLAYRESULT(...) fprintf_utf8(stdout, __VA_ARGS__) #else #define DISPLAY(...) fprintf(stderr, __VA_ARGS__) #define DISPLAYRESULT(...) fprintf(stdout, __VA_ARGS__) #endif #define DISPLAYLEVEL(l, ...) do { if (g_displayLevel>=l) DISPLAY(__VA_ARGS__); } while (0) static int g_displayLevel = 2; /* ************************************ * Local variables **************************************/ static U32 g_nbIterations = NBLOOPS; /* ************************************ * Benchmark Functions **************************************/ static clock_t BMK_clockSpan( clock_t start ) { return clock() - start; /* works even if overflow; Typical max span ~ 30 mn */ } static size_t BMK_findMaxMem(U64 requiredMem) { size_t const step = 64 MB; void* testmem = NULL; requiredMem = (((requiredMem >> 26) + 1) << 26); requiredMem += 2*step; if (requiredMem > MAX_MEM) requiredMem = MAX_MEM; while (!testmem) { if (requiredMem > step) requiredMem -= step; else requiredMem >>= 1; testmem = malloc ((size_t)requiredMem); } free (testmem); /* keep some space available */ if (requiredMem > step) requiredMem -= step; else requiredMem >>= 1; return (size_t)requiredMem; } static U64 BMK_GetFileSize(const char* infilename) { int r; #if defined(_MSC_VER) struct _stat64 statbuf; r = _stat64(infilename, &statbuf); #else struct stat statbuf; r = stat(infilename, &statbuf); #endif if (r || !S_ISREG(statbuf.st_mode)) return 0; /* No good... */ return (U64)statbuf.st_size; } /* * Allocates a string containing s1 and s2 concatenated. Acts like strdup. * The result must be freed. */ static char* XXH_strcatDup(const char* s1, const char* s2) { assert(s1 != NULL); assert(s2 != NULL); { size_t len1 = strlen(s1); size_t len2 = strlen(s2); char* buf = (char*)malloc(len1 + len2 + 1); if (buf != NULL) { /* strcpy(buf, s1) */ memcpy(buf, s1, len1); /* strcat(buf, s2) */ memcpy(buf + len1, s2, len2 + 1); } return buf; } } /* use #define to make them constant, required for initialization */ #define PRIME32 2654435761U #define PRIME64 11400714785074694797ULL /* * Fills a test buffer with pseudorandom data. * * This is used in the sanity check - its values must not be changed. */ static void BMK_fillTestBuffer(U8* buffer, size_t len) { U64 byteGen = PRIME32; size_t i; assert(buffer != NULL); for (i=0; i>56); byteGen *= PRIME64; } } /* * A secret buffer used for benchmarking XXH3's withSecret variants. * * In order for the bench to be realistic, the secret buffer would need to be * pre-generated. * * Adding a pointer to the parameter list would be messy. */ static U8 g_benchSecretBuf[XXH3_SECRET_SIZE_MIN]; /* * Wrappers for the benchmark. * * If you would like to add other hashes to the bench, create a wrapper and add * it to the g_hashesToBench table. It will automatically be added. */ typedef U32 (*hashFunction)(const void* buffer, size_t bufferSize, U32 seed); static U32 localXXH32(const void* buffer, size_t bufferSize, U32 seed) { return XXH32(buffer, bufferSize, seed); } static U32 localXXH64(const void* buffer, size_t bufferSize, U32 seed) { return (U32)XXH64(buffer, bufferSize, seed); } static U32 localXXH3_64b(const void* buffer, size_t bufferSize, U32 seed) { (void)seed; return (U32)XXH3_64bits(buffer, bufferSize); } static U32 localXXH3_64b_seeded(const void* buffer, size_t bufferSize, U32 seed) { return (U32)XXH3_64bits_withSeed(buffer, bufferSize, seed); } static U32 localXXH3_64b_secret(const void* buffer, size_t bufferSize, U32 seed) { (void)seed; return (U32)XXH3_64bits_withSecret(buffer, bufferSize, g_benchSecretBuf, sizeof(g_benchSecretBuf)); } static U32 localXXH3_128b(const void* buffer, size_t bufferSize, U32 seed) { (void)seed; return (U32)(XXH3_128bits(buffer, bufferSize).low64); } static U32 localXXH3_128b_seeded(const void* buffer, size_t bufferSize, U32 seed) { return (U32)(XXH3_128bits_withSeed(buffer, bufferSize, seed).low64); } static U32 localXXH3_128b_secret(const void* buffer, size_t bufferSize, U32 seed) { (void)seed; return (U32)(XXH3_128bits_withSecret(buffer, bufferSize, g_benchSecretBuf, sizeof(g_benchSecretBuf)).low64); } static U32 localXXH3_stream(const void* buffer, size_t bufferSize, U32 seed) { XXH3_state_t state; (void)seed; XXH3_64bits_reset(&state); XXH3_64bits_update(&state, buffer, bufferSize); return (U32)XXH3_64bits_digest(&state); } static U32 localXXH128_stream(const void* buffer, size_t bufferSize, U32 seed) { XXH3_state_t state; (void)seed; XXH3_128bits_reset(&state); XXH3_128bits_update(&state, buffer, bufferSize); return (U32)(XXH3_128bits_digest(&state).low64); } typedef struct { const char* name; hashFunction func; } hashInfo; #define NB_HASHFUNC 10 static const hashInfo g_hashesToBench[NB_HASHFUNC] = { { "XXH32", &localXXH32 }, { "XXH64", &localXXH64 }, { "XXH3_64b", &localXXH3_64b }, { "XXH3_64b w/seed", &localXXH3_64b_seeded }, { "XXH3_64b w/secret", &localXXH3_64b_secret }, { "XXH128", &localXXH3_128b }, { "XXH128 w/seed", &localXXH3_128b_seeded }, { "XXH128 w/secret", &localXXH3_128b_secret }, { "XXH3_stream", &localXXH3_stream }, { "XXH128_stream", &localXXH128_stream }, }; #define NB_TESTFUNC (1 + 2 * NB_HASHFUNC) static char g_testIDs[NB_TESTFUNC] = { 0 }; static const char k_testIDs_default[NB_TESTFUNC] = { 0, 1 /*XXH32*/, 0, 1 /*XXH64*/, 0, 1 /*XXH3*/, 0, 0, 0, 0, 0, 1 /*XXH128*/ }; #define HASHNAME_MAX 29 static void BMK_benchHash(hashFunction h, const char* hName, int testID, const void* buffer, size_t bufferSize) { U32 nbh_perIteration = (U32)((300 MB) / (bufferSize+1)) + 1; /* first iteration conservatively aims for 300 MB/s */ unsigned iterationNb, nbIterations = g_nbIterations + !g_nbIterations /* min 1 */; double fastestH = 100000000.; assert(HASHNAME_MAX > 2); DISPLAYLEVEL(2, "\r%80s\r", ""); /* Clean display line */ for (iterationNb = 1; iterationNb <= nbIterations; iterationNb++) { U32 r=0; clock_t cStart; DISPLAYLEVEL(2, "%2u-%-*.*s : %10u ->\r", iterationNb, HASHNAME_MAX, HASHNAME_MAX, hName, (unsigned)bufferSize); cStart = clock(); while (clock() == cStart); /* starts clock() at its exact beginning */ cStart = clock(); { U32 u; for (u=0; u (double)(4000U<<20)) nbh_perSecond = (double)(4000U<<20); /* avoid overflow */ nbh_perIteration = (U32)nbh_perSecond; } /* g_nbIterations==0 => quick evaluation, no claim of accuracy */ if (g_nbIterations>0) { iterationNb--; /* new round for a more accurate speed evaluation */ continue; } } if (ticksPerHash < fastestH) fastestH = ticksPerHash; if (fastestH>0.) { /* avoid div by zero */ DISPLAYLEVEL(2, "%2u-%-*.*s : %10u -> %8.0f it/s (%7.1f MB/s) \r", iterationNb, HASHNAME_MAX, HASHNAME_MAX, hName, (unsigned)bufferSize, (double)1 / fastestH, ((double)bufferSize / (1 MB)) / fastestH); } } { double nbh_perSecond = (1 / fastestH) + 1; if (nbh_perSecond > (double)(4000U<<20)) nbh_perSecond = (double)(4000U<<20); /* avoid overflow */ nbh_perIteration = (U32)nbh_perSecond; } } DISPLAYLEVEL(1, "%2i#%-*.*s : %10u -> %8.0f it/s (%7.1f MB/s) \n", testID, HASHNAME_MAX, HASHNAME_MAX, hName, (unsigned)bufferSize, (double)1 / fastestH, ((double)bufferSize / (1 MB)) / fastestH); if (g_displayLevel<1) DISPLAYLEVEL(0, "%u, ", (unsigned)((double)1 / fastestH)); } /*! * BMK_benchMem(): * buffer: Must be 16-byte aligned. * The real allocated size of buffer is supposed to be >= (bufferSize+3). * returns: 0 on success, 1 if error (invalid mode selected) */ static void BMK_benchMem(const void* buffer, size_t bufferSize) { assert((((size_t)buffer) & 15) == 0); /* ensure alignment */ BMK_fillTestBuffer(g_benchSecretBuf, sizeof(g_benchSecretBuf)); { int i; for (i = 1; i < NB_TESTFUNC; i++) { int const hashFuncID = (i-1) / 2; assert(g_hashesToBench[hashFuncID].name != NULL); if (g_testIDs[i] == 0) continue; /* aligned */ if ((i % 2) == 1) { BMK_benchHash(g_hashesToBench[hashFuncID].func, g_hashesToBench[hashFuncID].name, i, buffer, bufferSize); } /* unaligned */ if ((i % 2) == 0) { /* Append "unaligned". */ char* const hashNameBuf = XXH_strcatDup(g_hashesToBench[hashFuncID].name, " unaligned"); assert(hashNameBuf != NULL); BMK_benchHash(g_hashesToBench[hashFuncID].func, hashNameBuf, i, ((const char*)buffer)+3, bufferSize); free(hashNameBuf); } } } } static size_t BMK_selectBenchedSize(const char* fileName) { U64 const inFileSize = BMK_GetFileSize(fileName); size_t benchedSize = (size_t) BMK_findMaxMem(inFileSize); if ((U64)benchedSize > inFileSize) benchedSize = (size_t)inFileSize; if (benchedSize < inFileSize) { DISPLAY("Not enough memory for '%s' full size; testing %i MB only...\n", fileName, (int)(benchedSize>>20)); } return benchedSize; } static int BMK_benchFiles(const char*const* fileNamesTable, int nbFiles) { int fileIdx; for (fileIdx=0; fileIdx 10 KB) { DISPLAYLEVEL(1, "%u KB", (unsigned)(keySize >> 10)); } else { DISPLAYLEVEL(1, "%u bytes", (unsigned)keySize); } DISPLAYLEVEL(1, "... \n"); BMK_benchMem(alignedBuffer, keySize); free(buffer); } return 0; } /* ************************************************ * Self-test: * ensure results consistency accross platforms *********************************************** */ static void BMK_checkResult32(XXH32_hash_t r1, XXH32_hash_t r2) { static int nbTests = 1; if (r1!=r2) { DISPLAY("\rError: 32-bit hash test %i: Internal sanity check failed!\n", nbTests); DISPLAY("\rGot 0x%08X, expected 0x%08X.\n", (unsigned)r1, (unsigned)r2); DISPLAY("\rNote: If you modified the hash functions, make sure to either update the values\n" "or temporarily comment out the tests in BMK_sanityCheck.\n"); exit(1); } nbTests++; } static void BMK_checkResult64(XXH64_hash_t r1, XXH64_hash_t r2) { static int nbTests = 1; if (r1!=r2) { DISPLAY("\rError: 64-bit hash test %i: Internal sanity check failed!\n", nbTests); DISPLAY("\rGot 0x%08X%08XULL, expected 0x%08X%08XULL.\n", (unsigned)(r1>>32), (unsigned)r1, (unsigned)(r2>>32), (unsigned)r2); DISPLAY("\rNote: If you modified the hash functions, make sure to either update the values\n" "or temporarily comment out the tests in BMK_sanityCheck.\n"); exit(1); } nbTests++; } static void BMK_checkResult128(XXH128_hash_t r1, XXH128_hash_t r2) { static int nbTests = 1; if ((r1.low64 != r2.low64) || (r1.high64 != r2.high64)) { DISPLAY("\rError: 128-bit hash test %i: Internal sanity check failed.\n", nbTests); DISPLAY("\rGot { 0x%08X%08XULL, 0x%08X%08XULL }, expected { 0x%08X%08XULL, 0x%08X%08XULL } \n", (unsigned)(r1.low64>>32), (unsigned)r1.low64, (unsigned)(r1.high64>>32), (unsigned)r1.high64, (unsigned)(r2.low64>>32), (unsigned)r2.low64, (unsigned)(r2.high64>>32), (unsigned)r2.high64 ); DISPLAY("\rNote: If you modified the hash functions, make sure to either update the values\n" "or temporarily comment out the tests in BMK_sanityCheck.\n"); exit(1); } nbTests++; } static void BMK_testXXH32(const void* data, size_t len, U32 seed, U32 Nresult) { XXH32_state_t *state = XXH32_createState(); size_t pos; assert(state != NULL); if (len>0) assert(data != NULL); BMK_checkResult32(XXH32(data, len, seed), Nresult); (void)XXH32_reset(state, seed); (void)XXH32_update(state, data, len); BMK_checkResult32(XXH32_digest(state), Nresult); (void)XXH32_reset(state, seed); for (pos=0; pos0) assert(data != NULL); BMK_checkResult64(XXH64(data, len, seed), Nresult); (void)XXH64_reset(state, seed); (void)XXH64_update(state, data, len); BMK_checkResult64(XXH64_digest(state), Nresult); (void)XXH64_reset(state, seed); for (pos=0; pos> 40); } void BMK_testXXH3(const void* data, size_t len, U64 seed, U64 Nresult) { if (len>0) assert(data != NULL); { U64 const Dresult = XXH3_64bits_withSeed(data, len, seed); BMK_checkResult64(Dresult, Nresult); } /* check that the no-seed variant produces same result as seed==0 */ if (seed == 0) { U64 const Dresult = XXH3_64bits(data, len); BMK_checkResult64(Dresult, Nresult); } /* streaming API test */ { XXH3_state_t* const state = XXH3_createState(); assert(state != NULL); /* single ingestion */ (void)XXH3_64bits_reset_withSeed(state, seed); (void)XXH3_64bits_update(state, data, len); BMK_checkResult64(XXH3_64bits_digest(state), Nresult); /* random ingestion */ { size_t p = 0; (void)XXH3_64bits_reset_withSeed(state, seed); while (p < len) { size_t const modulo = len > 2 ? len : 2; size_t l = (size_t)(BMK_rand()) % modulo; if (p + l > len) l = len - p; (void)XXH3_64bits_update(state, (const char*)data+p, l); p += l; } BMK_checkResult64(XXH3_64bits_digest(state), Nresult); } /* byte by byte ingestion */ { size_t pos; (void)XXH3_64bits_reset_withSeed(state, seed); for (pos=0; pos0) assert(data != NULL); { U64 const Dresult = XXH3_64bits_withSecret(data, len, secret, secretSize); BMK_checkResult64(Dresult, Nresult); } /* streaming API test */ { XXH3_state_t *state = XXH3_createState(); assert(state != NULL); (void)XXH3_64bits_reset_withSecret(state, secret, secretSize); (void)XXH3_64bits_update(state, data, len); BMK_checkResult64(XXH3_64bits_digest(state), Nresult); /* random ingestion */ { size_t p = 0; (void)XXH3_64bits_reset_withSecret(state, secret, secretSize); while (p < len) { size_t const modulo = len > 2 ? len : 2; size_t l = (size_t)(BMK_rand()) % modulo; if (p + l > len) l = len - p; (void)XXH3_64bits_update(state, (const char*)data+p, l); p += l; } BMK_checkResult64(XXH3_64bits_digest(state), Nresult); } /* byte by byte ingestion */ { size_t pos; (void)XXH3_64bits_reset_withSecret(state, secret, secretSize); for (pos=0; pos 2 ? len : 2; size_t l = (size_t)(BMK_rand()) % modulo; if (p + l > len) l = len - p; (void)XXH3_128bits_update(state, (const char*)data+p, l); p += l; } BMK_checkResult128(XXH3_128bits_digest(state), Nresult); } /* byte by byte ingestion */ { size_t pos; (void)XXH3_128bits_reset_withSeed(state, seed); for (pos=0; pos0) assert(data != NULL); { XXH128_hash_t const Dresult = XXH3_128bits_withSecret(data, len, secret, secretSize); BMK_checkResult128(Dresult, Nresult); } /* streaming API test */ { XXH3_state_t* const state = XXH3_createState(); assert(state != NULL); (void)XXH3_128bits_reset_withSecret(state, secret, secretSize); (void)XXH3_128bits_update(state, data, len); BMK_checkResult128(XXH3_128bits_digest(state), Nresult); /* random ingestion */ { size_t p = 0; (void)XXH3_128bits_reset_withSecret(state, secret, secretSize); while (p < len) { size_t const modulo = len > 2 ? len : 2; size_t l = (size_t)(BMK_rand()) % modulo; if (p + l > len) l = len - p; (void)XXH3_128bits_update(state, (const char*)data+p, l); p += l; } BMK_checkResult128(XXH3_128bits_digest(state), Nresult); } /* byte by byte ingestion */ { size_t pos; (void)XXH3_128bits_reset_withSecret(state, secret, secretSize); for (pos=0; pos= 7 + secretSize); BMK_testXXH3_withSecret(NULL, 0, secret, secretSize, 0x6775FD10343C92C3ULL); /* empty string */ BMK_testXXH3_withSecret(sanityBuffer, 1, secret, secretSize, 0xC3382C326E24E3CDULL); /* 1 - 3 */ BMK_testXXH3_withSecret(sanityBuffer, 6, secret, secretSize, 0x82C90AB0519369ADULL); /* 4 - 8 */ BMK_testXXH3_withSecret(sanityBuffer, 12, secret, secretSize, 0x14631E773B78EC57ULL); /* 9 - 16 */ BMK_testXXH3_withSecret(sanityBuffer, 24, secret, secretSize, 0xCDD5542E4A9D9FE8ULL); /* 17 - 32 */ BMK_testXXH3_withSecret(sanityBuffer, 48, secret, secretSize, 0x33ABD54D094B2534ULL); /* 33 - 64 */ BMK_testXXH3_withSecret(sanityBuffer, 80, secret, secretSize, 0xE687BA1684965297ULL); /* 65 - 96 */ BMK_testXXH3_withSecret(sanityBuffer, 195, secret, secretSize, 0xA057273F5EECFB20ULL); /* 129-240 */ BMK_testXXH3_withSecret(sanityBuffer, 403, secret, secretSize, 0xF9C0BA5BA3AF70B8ULL); /* one block, last stripe is overlapping */ BMK_testXXH3_withSecret(sanityBuffer, 512, secret, secretSize, 0x7896E65DCFA09071ULL); /* one block, finishing at stripe boundary */ BMK_testXXH3_withSecret(sanityBuffer,2048, secret, secretSize, 0xD6545DB87ECFD98CULL); /* >= 2 blocks, at least one scrambling */ BMK_testXXH3_withSecret(sanityBuffer,2367, secret, secretSize, 0x857320340D953686ULL); /* >= 2 blocks, at least one scrambling, last stripe unaligned */ BMK_testXXH3_withSecret(sanityBuffer,64*10*3, secret, secretSize, 0xD4989A002E9850ABULL); /* exactly 3 full blocks, not a multiple of 256 */ } /* XXH128 */ { XXH128_hash_t const expected = { 0x1F17545BCE1061F1ULL, 0x07FD4E968E916AE1ULL }; BMK_testXXH128(NULL, 0, 0, expected); /* empty string */ } { XXH128_hash_t const expected = { 0x7282E631387D51ACULL, 0x8743B0A8131AB9E6ULL }; BMK_testXXH128(NULL, 0, PRIME32, expected); } { XXH128_hash_t const expected = { 0xB936EBAE24CB01C5ULL, 0x2554B05763A71A05ULL }; BMK_testXXH128(sanityBuffer, 1, 0, expected); /* 1-3 */ } { XXH128_hash_t const expected = { 0xCA57C628C04B45B8ULL, 0x916831F4DCD21CF9ULL }; BMK_testXXH128(sanityBuffer, 1, PRIME32, expected); /* 1-3 */ } { XXH128_hash_t const expected = { 0x3E7039BDDA43CFC6ULL, 0x082AFE0B8162D12AULL }; BMK_testXXH128(sanityBuffer, 6, 0, expected); /* 4-8 */ } { XXH128_hash_t const expected = { 0x269D8F70BE98856EULL, 0x5A865B5389ABD2B1ULL }; BMK_testXXH128(sanityBuffer, 6, PRIME32, expected); /* 4-8 */ } { XXH128_hash_t const expected = { 0x061A192713F69AD9ULL, 0x6E3EFD8FC7802B18ULL }; BMK_testXXH128(sanityBuffer, 12, 0, expected); /* 9-16 */ } { XXH128_hash_t const expected = { 0x9BE9F9A67F3C7DFBULL, 0xD7E09D518A3405D3ULL }; BMK_testXXH128(sanityBuffer, 12, PRIME32, expected); /* 9-16 */ } { XXH128_hash_t const expected = { 0x1E7044D28B1B901DULL, 0x0CE966E4678D3761ULL }; BMK_testXXH128(sanityBuffer, 24, 0, expected); /* 17-32 */ } { XXH128_hash_t const expected = { 0xD7304C54EBAD40A9ULL, 0x3162026714A6A243ULL }; BMK_testXXH128(sanityBuffer, 24, PRIME32, expected); /* 17-32 */ } { XXH128_hash_t const expected = { 0xF942219AED80F67BULL, 0xA002AC4E5478227EULL }; BMK_testXXH128(sanityBuffer, 48, 0, expected); /* 33-64 */ } { XXH128_hash_t const expected = { 0x7BA3C3E453A1934EULL, 0x163ADDE36C072295ULL }; BMK_testXXH128(sanityBuffer, 48, PRIME32, expected); /* 33-64 */ } { XXH128_hash_t const expected = { 0x5E8BAFB9F95FB803ULL, 0x4952F58181AB0042ULL }; BMK_testXXH128(sanityBuffer, 81, 0, expected); /* 65-96 */ } { XXH128_hash_t const expected = { 0x703FBB3D7A5F755CULL, 0x2724EC7ADC750FB6ULL }; BMK_testXXH128(sanityBuffer, 81, PRIME32, expected); /* 65-96 */ } { XXH128_hash_t const expected = { 0xF1AEBD597CEC6B3AULL, 0x337E09641B948717ULL }; BMK_testXXH128(sanityBuffer, 222, 0, expected); /* 129-240 */ } { XXH128_hash_t const expected = { 0xAE995BB8AF917A8DULL, 0x91820016621E97F1ULL }; BMK_testXXH128(sanityBuffer, 222, PRIME32, expected); /* 129-240 */ } { XXH128_hash_t const expected = { 0xCDEB804D65C6DEA4ULL, 0x1B6DE21E332DD73DULL }; BMK_testXXH128(sanityBuffer, 403, 0, expected); /* one block, last stripe is overlapping */ } { XXH128_hash_t const expected = { 0x6259F6ECFD6443FDULL, 0xBED311971E0BE8F2ULL }; BMK_testXXH128(sanityBuffer, 403, PRIME64, expected); /* one block, last stripe is overlapping */ } { XXH128_hash_t const expected = { 0x1443B8153EBEE367ULL, 0x98EC7E48CD872997ULL }; BMK_testXXH128(sanityBuffer, 512, 0, expected); /* one block, finishing at stripe boundary */ } { XXH128_hash_t const expected = { 0x43FDC6823A52F1F2ULL, 0x2F748A4F194E1EF0ULL }; BMK_testXXH128(sanityBuffer, 512, PRIME64, expected); /* one block, finishing at stripe boundary */ } { XXH128_hash_t const expected = { 0xF4258501BE8E0623ULL, 0x6930A2267A755B20ULL }; BMK_testXXH128(sanityBuffer,2048, 0, expected); /* two blocks, finishing at block boundary */ } { XXH128_hash_t const expected = { 0x10CC56C2FA0AD9ACULL, 0xD0D7A3C2EEF2D892ULL }; BMK_testXXH128(sanityBuffer,2048, PRIME32, expected); /* two blocks, finishing at block boundary */ } { XXH128_hash_t const expected = { 0x5890AE7ACBB84A7EULL, 0x85C327B377AA7E62ULL }; BMK_testXXH128(sanityBuffer,2240, 0, expected); /* two blocks, ends at stripe boundary */ } { XXH128_hash_t const expected = { 0x205E6D72DCCBD2AAULL, 0x62B70214DB075235ULL }; BMK_testXXH128(sanityBuffer,2240, PRIME32, expected); /* two blocks, ends at stripe boundary */ } { XXH128_hash_t const expected = { 0xCB37AEB9E5D361EDULL, 0xE89C0F6FF369B427ULL }; BMK_testXXH128(sanityBuffer,2367, 0, expected); /* two blocks, last stripe is overlapping */ } { XXH128_hash_t const expected = { 0x6F5360AE69C2F406ULL, 0xD23AAE4B76C31ECBULL }; BMK_testXXH128(sanityBuffer,2367, PRIME32, expected); /* two blocks, last stripe is overlapping */ } /* XXH128 with custom Secret */ { const void* const secret = sanityBuffer + 7; const size_t secretSize = XXH3_SECRET_SIZE_MIN + 11; assert(sizeof(sanityBuffer) >= 7 + secretSize); { XXH128_hash_t const expected = { 0x58607A62EE3D62F5ULL, 0x339965195B2942D1ULL }; BMK_testXXH128_withSecret(NULL, 0, secret, secretSize, expected); /* empty string */ } { XXH128_hash_t const expected = { 0xC3382C326E24E3CDULL, 0x2A2771BA8906D2EBULL }; BMK_testXXH128_withSecret(sanityBuffer, 1, secret, secretSize, expected); /* 1-3 */ } { XXH128_hash_t const expected = { 0x0B61C8ACA7D4778FULL, 0x376BD91B6432F36DULL }; BMK_testXXH128_withSecret(sanityBuffer, 6, secret, secretSize, expected); /* 4-8 */ } { XXH128_hash_t const expected = { 0xAF82F6EBA263D7D8ULL, 0x90A3C2D839F57D0FULL }; BMK_testXXH128_withSecret(sanityBuffer, 12, secret, secretSize, expected); /* 9-16 */ } } /* secret generator */ { verifSample_t const expected = { { 0xB8, 0x26, 0x83, 0x7E } }; BMK_testSecretGenerator(NULL, 0, expected); } { verifSample_t const expected = { { 0x25, 0x82, 0x62, 0x9A } }; BMK_testSecretGenerator(sanityBuffer, 1, expected); } { verifSample_t const expected = { { 0xDA, 0x2A, 0x12, 0x11 } }; BMK_testSecretGenerator(sanityBuffer, XXH3_SECRET_SIZE_MIN - 1, expected); } { verifSample_t const expected = { { 0x7E, 0x48, 0x0C, 0xA7 } }; BMK_testSecretGenerator(sanityBuffer, XXH3_SECRET_DEFAULT_SIZE + 500, expected); } DISPLAYLEVEL(3, "\r%70s\r", ""); /* Clean display line */ DISPLAYLEVEL(3, "Sanity check -- all tests ok\n"); } /* ******************************************************** * File Hashing **********************************************************/ /* for support of --little-endian display mode */ static void BMK_display_LittleEndian(const void* ptr, size_t length) { const U8* const p = (const U8*)ptr; size_t idx; for (idx=length-1; idx 0) { switch(hashType) { case algo_xxh32: (void)XXH32_update(&state32, buffer, readSize); break; case algo_xxh64: (void)XXH64_update(&state64, buffer, readSize); break; case algo_xxh128: (void)XXH3_128bits_update(&state128, buffer, readSize); break; default: assert(0); } } if (ferror(inFile)) { DISPLAY("Error: a failure occurred reading the input file.\n"); exit(1); } } { Multihash finalHash; switch(hashType) { case algo_xxh32: finalHash.xxh32 = XXH32_digest(&state32); break; case algo_xxh64: finalHash.xxh64 = XXH64_digest(&state64); break; case algo_xxh128: finalHash.xxh128 = XXH3_128bits_digest(&state128); break; default: assert(0); } return finalHash; } } /* algo_xxh32, algo_xxh64, algo_xxh128 */ static const char* XSUM_algoName[] = { "XXH32", "XXH64", "XXH128" }; static const char* XSUM_algoLE_name[] = { "XXH32_LE", "XXH64_LE", "XXH128_LE" }; static const size_t XSUM_algoLength[] = { 4, 8, 16 }; #define XSUM_TABLE_ELT_SIZE(table) (sizeof(table) / sizeof(*table)) typedef void (*XSUM_displayHash_f)(const void*, size_t); /* display function signature */ static void XSUM_printLine_BSD_internal(const char* filename, const void* canonicalHash, const AlgoSelected hashType, const char* algoString[], XSUM_displayHash_f f_displayHash) { assert(0 <= hashType && hashType <= XSUM_TABLE_ELT_SIZE(XSUM_algoName)); { const char* const typeString = algoString[hashType]; const size_t hashLength = XSUM_algoLength[hashType]; DISPLAYRESULT("%s (%s) = ", typeString, filename); f_displayHash(canonicalHash, hashLength); DISPLAYRESULT("\n"); } } static void XSUM_printLine_BSD_LE(const char* filename, const void* canonicalHash, const AlgoSelected hashType) { XSUM_printLine_BSD_internal(filename, canonicalHash, hashType, XSUM_algoLE_name, BMK_display_LittleEndian); } static void XSUM_printLine_BSD(const char* filename, const void* canonicalHash, const AlgoSelected hashType) { XSUM_printLine_BSD_internal(filename, canonicalHash, hashType, XSUM_algoName, BMK_display_BigEndian); } static void XSUM_printLine_GNU_internal(const char* filename, const void* canonicalHash, const AlgoSelected hashType, XSUM_displayHash_f f_displayHash) { assert(0 <= hashType && hashType <= XSUM_TABLE_ELT_SIZE(XSUM_algoName)); { const size_t hashLength = XSUM_algoLength[hashType]; f_displayHash(canonicalHash, hashLength); DISPLAYRESULT(" %s\n", filename); } } static void XSUM_printLine_GNU(const char* filename, const void* canonicalHash, const AlgoSelected hashType) { XSUM_printLine_GNU_internal(filename, canonicalHash, hashType, BMK_display_BigEndian); } static void XSUM_printLine_GNU_LE(const char* filename, const void* canonicalHash, const AlgoSelected hashType) { XSUM_printLine_GNU_internal(filename, canonicalHash, hashType, BMK_display_LittleEndian); } typedef enum { big_endian, little_endian} Display_endianess; typedef enum { display_gnu, display_bsd } Display_convention; typedef void (*XSUM_displayLine_f)(const char*, const void*, AlgoSelected); /* line display signature */ static XSUM_displayLine_f XSUM_kDisplayLine_fTable[2][2] = { { XSUM_printLine_GNU, XSUM_printLine_GNU_LE }, { XSUM_printLine_BSD, XSUM_printLine_BSD_LE } }; static int XSUM_hashFile(const char* fileName, const AlgoSelected hashType, const Display_endianess displayEndianess, const Display_convention convention) { size_t const blockSize = 64 KB; XSUM_displayLine_f const f_displayLine = XSUM_kDisplayLine_fTable[convention][displayEndianess]; FILE* inFile; Multihash hashValue; assert(displayEndianess==big_endian || displayEndianess==little_endian); assert(convention==display_gnu || convention==display_bsd); /* Check file existence */ if (fileName == stdinName) { inFile = stdin; fileName = "stdin"; SET_BINARY_MODE(stdin); } else { inFile = XXH_fopen( fileName, "rb" ); } if (inFile==NULL) { DISPLAY("Error: Could not open '%s': %s. \n", fileName, strerror(errno)); return 1; } /* Memory allocation & streaming */ { void* const buffer = malloc(blockSize); if (!buffer) { DISPLAY("\nError: Out of memory.\n"); fclose(inFile); return 1; } /* Stream file & update hash */ hashValue = XSUM_hashStream(inFile, hashType, buffer, blockSize); fclose(inFile); free(buffer); } /* display Hash value in selected format */ switch(hashType) { case algo_xxh32: { XXH32_canonical_t hcbe32; (void)XXH32_canonicalFromHash(&hcbe32, hashValue.xxh32); f_displayLine(fileName, &hcbe32, hashType); break; } case algo_xxh64: { XXH64_canonical_t hcbe64; (void)XXH64_canonicalFromHash(&hcbe64, hashValue.xxh64); f_displayLine(fileName, &hcbe64, hashType); break; } case algo_xxh128: { XXH128_canonical_t hcbe128; (void)XXH128_canonicalFromHash(&hcbe128, hashValue.xxh128); f_displayLine(fileName, &hcbe128, hashType); break; } default: assert(0); /* not possible */ } return 0; } /* * XSUM_hashFiles: * If fnTotal==0, read from stdin instead. */ static int XSUM_hashFiles(const char*const * fnList, int fnTotal, AlgoSelected hashType, Display_endianess displayEndianess, Display_convention convention) { int fnNb; int result = 0; if (fnTotal==0) return XSUM_hashFile(stdinName, hashType, displayEndianess, convention); for (fnNb=0; fnNb= (size_t)*lineMax) { char* newLineBuf = NULL; size_t newBufSize = (size_t)*lineMax; newBufSize += (newBufSize/2) + 1; /* x 1.5 */ if (newBufSize > MAX_LINE_LENGTH) newBufSize = MAX_LINE_LENGTH; if (len+1 >= newBufSize) return GetLine_exceedMaxLineLength; newLineBuf = (char*) realloc(*lineBuf, newBufSize); if (newLineBuf == NULL) return GetLine_outOfMemory; *lineBuf = newLineBuf; *lineMax = (int)newBufSize; } if (c == '\n') break; (*lineBuf)[len++] = (char) c; } (*lineBuf)[len] = '\0'; return result; } /* * Converts one hexadecimal character to integer. * Returns -1 if the given character is not hexadecimal. */ static int charToHex(char c) { int result = -1; if (c >= '0' && c <= '9') { result = (int) (c - '0'); } else if (c >= 'A' && c <= 'F') { result = (int) (c - 'A') + 0x0a; } else if (c >= 'a' && c <= 'f') { result = (int) (c - 'a') + 0x0a; } return result; } /* * Converts canonical ASCII hexadecimal string `hashStr` * to the big endian binary representation in unsigned char array `dst`. * * Returns CANONICAL_FROM_STRING_INVALID_FORMAT if hashStr is not well formatted. * Returns CANONICAL_FROM_STRING_OK if hashStr is parsed successfully. */ static CanonicalFromStringResult canonicalFromString(unsigned char* dst, size_t dstSize, const char* hashStr) { size_t i; for (i = 0; i < dstSize; ++i) { int h0, h1; h0 = charToHex(hashStr[i*2 + 0]); if (h0 < 0) return CanonicalFromString_invalidFormat; h1 = charToHex(hashStr[i*2 + 1]); if (h1 < 0) return CanonicalFromString_invalidFormat; dst[i] = (unsigned char) ((h0 << 4) | h1); } return CanonicalFromString_ok; } /* * Parse single line of xxHash checksum file. * Returns PARSE_LINE_ERROR_INVALID_FORMAT if the line is not well formatted. * Returns PARSE_LINE_OK if the line is parsed successfully. * And members of parseLine will be filled by parsed values. * * - line must be terminated with '\0'. * - Since parsedLine.filename will point within given argument `line`, * users must keep `line`s content when they are using parsedLine. * * xxHash checksum lines should have the following format: * * <8, 16, or 32 hexadecimal char> <'\0'> */ static ParseLineResult parseLine(ParsedLine* parsedLine, const char* line) { const char* const firstSpace = strchr(line, ' '); if (firstSpace == NULL) return ParseLine_invalidFormat; parsedLine->filename = NULL; parsedLine->xxhBits = 0; switch (firstSpace - line) { case 8: { XXH32_canonical_t* xxh32c = &parsedLine->canonical.xxh32; if (canonicalFromString(xxh32c->digest, sizeof(xxh32c->digest), line) != CanonicalFromString_ok) { return ParseLine_invalidFormat; } parsedLine->xxhBits = 32; break; } case 16: { XXH64_canonical_t* xxh64c = &parsedLine->canonical.xxh64; if (canonicalFromString(xxh64c->digest, sizeof(xxh64c->digest), line) != CanonicalFromString_ok) { return ParseLine_invalidFormat; } parsedLine->xxhBits = 64; break; } case 32: { XXH128_canonical_t* xxh128c = &parsedLine->canonical.xxh128; if (canonicalFromString(xxh128c->digest, sizeof(xxh128c->digest), line) != CanonicalFromString_ok) { return ParseLine_invalidFormat; } parsedLine->xxhBits = 128; break; } default: return ParseLine_invalidFormat; break; } /* note : skipping second separation character, which can be anything, * allowing insertion of custom markers such as '*' */ parsedLine->filename = firstSpace + 2; return ParseLine_ok; } /*! * Parse xxHash checksum file. */ static void parseFile1(ParseFileArg* parseFileArg) { const char* const inFileName = parseFileArg->inFileName; ParseFileReport* const report = &parseFileArg->report; unsigned long lineNumber = 0; memset(report, 0, sizeof(*report)); while (!report->quit) { LineStatus lineStatus = LineStatus_hashFailed; ParsedLine parsedLine; memset(&parsedLine, 0, sizeof(parsedLine)); lineNumber++; if (lineNumber == 0) { /* This is unlikely happen, but md5sum.c has this error check. */ DISPLAY("%s: Error: Too many checksum lines\n", inFileName); report->quit = 1; break; } { GetLineResult const getLineResult = getLine(&parseFileArg->lineBuf, &parseFileArg->lineMax, parseFileArg->inFile); if (getLineResult != GetLine_ok) { if (getLineResult == GetLine_eof) break; switch (getLineResult) { case GetLine_ok: case GetLine_eof: /* These cases never happen. See above getLineResult related "if"s. They exist just for make gcc's -Wswitch-enum happy. */ assert(0); break; default: DISPLAY("%s:%lu: Error: Unknown error.\n", inFileName, lineNumber); break; case GetLine_exceedMaxLineLength: DISPLAY("%s:%lu: Error: Line too long.\n", inFileName, lineNumber); break; case GetLine_outOfMemory: DISPLAY("%s:%lu: Error: Out of memory.\n", inFileName, lineNumber); break; } report->quit = 1; break; } } if (parseLine(&parsedLine, parseFileArg->lineBuf) != ParseLine_ok) { report->nImproperlyFormattedLines++; if (parseFileArg->warn) { DISPLAY("%s:%lu: Error: Improperly formatted checksum line.\n", inFileName, lineNumber); } continue; } if (report->xxhBits != 0 && report->xxhBits != parsedLine.xxhBits) { /* Don't accept xxh32/xxh64 mixed file */ report->nImproperlyFormattedLines++; report->nMixedFormatLines++; if (parseFileArg->warn) { DISPLAY("%s: %lu: Error: Multiple hash types in one file.\n", inFileName, lineNumber); } continue; } report->nProperlyFormattedLines++; if (report->xxhBits == 0) { report->xxhBits = parsedLine.xxhBits; } do { FILE* const fp = XXH_fopen(parsedLine.filename, "rb"); if (fp == NULL) { lineStatus = LineStatus_failedToOpen; break; } lineStatus = LineStatus_hashFailed; switch (parsedLine.xxhBits) { case 32: { Multihash const xxh = XSUM_hashStream(fp, algo_xxh32, parseFileArg->blockBuf, parseFileArg->blockSize); if (xxh.xxh32 == XXH32_hashFromCanonical(&parsedLine.canonical.xxh32)) { lineStatus = LineStatus_hashOk; } } break; case 64: { Multihash const xxh = XSUM_hashStream(fp, algo_xxh64, parseFileArg->blockBuf, parseFileArg->blockSize); if (xxh.xxh64 == XXH64_hashFromCanonical(&parsedLine.canonical.xxh64)) { lineStatus = LineStatus_hashOk; } } break; case 128: { Multihash const xxh = XSUM_hashStream(fp, algo_xxh128, parseFileArg->blockBuf, parseFileArg->blockSize); if (XXH128_isEqual(xxh.xxh128, XXH128_hashFromCanonical(&parsedLine.canonical.xxh128))) { lineStatus = LineStatus_hashOk; } } break; default: break; } fclose(fp); } while (0); switch (lineStatus) { default: DISPLAY("%s: Error: Unknown error.\n", inFileName); report->quit = 1; break; case LineStatus_failedToOpen: report->nOpenOrReadFailures++; if (!parseFileArg->statusOnly) { DISPLAYRESULT("%s:%lu: Could not open or read '%s': %s.\n", inFileName, lineNumber, parsedLine.filename, strerror(errno)); } break; case LineStatus_hashOk: case LineStatus_hashFailed: { int b = 1; if (lineStatus == LineStatus_hashOk) { /* If --quiet is specified, don't display "OK" */ if (parseFileArg->quiet) b = 0; } else { report->nMismatchedChecksums++; } if (b && !parseFileArg->statusOnly) { DISPLAYRESULT("%s: %s\n", parsedLine.filename , lineStatus == LineStatus_hashOk ? "OK" : "FAILED"); } } break; } } /* while (!report->quit) */ } /* Parse xxHash checksum file. * Returns 1, if all procedures were succeeded. * Returns 0, if any procedures was failed. * * If strictMode != 0, return error code if any line is invalid. * If statusOnly != 0, don't generate any output. * If warn != 0, print a warning message to stderr. * If quiet != 0, suppress "OK" line. * * "All procedures are succeeded" means: * - Checksum file contains at least one line and less than SIZE_T_MAX lines. * - All files are properly opened and read. * - All hash values match with its content. * - (strict mode) All lines in checksum file are consistent and well formatted. */ static int checkFile(const char* inFileName, const Display_endianess displayEndianess, U32 strictMode, U32 statusOnly, U32 warn, U32 quiet) { int result = 0; FILE* inFile = NULL; ParseFileArg parseFileArgBody; ParseFileArg* const parseFileArg = &parseFileArgBody; ParseFileReport* const report = &parseFileArg->report; if (displayEndianess != big_endian) { /* Don't accept little endian */ DISPLAY( "Check file mode doesn't support little endian\n" ); return 0; } /* note: stdinName is special constant pointer. It is not a string. */ if (inFileName == stdinName) { /* * Note: Since we expect text input for xxhash -c mode, * we don't set binary mode for stdin. */ inFileName = "stdin"; inFile = stdin; } else { inFile = XXH_fopen( inFileName, "rt" ); } if (inFile == NULL) { DISPLAY("Error: Could not open '%s': %s\n", inFileName, strerror(errno)); return 0; } parseFileArg->inFileName = inFileName; parseFileArg->inFile = inFile; parseFileArg->lineMax = DEFAULT_LINE_LENGTH; parseFileArg->lineBuf = (char*) malloc((size_t)parseFileArg->lineMax); parseFileArg->blockSize = 64 * 1024; parseFileArg->blockBuf = (char*) malloc(parseFileArg->blockSize); parseFileArg->strictMode = strictMode; parseFileArg->statusOnly = statusOnly; parseFileArg->warn = warn; parseFileArg->quiet = quiet; if ( (parseFileArg->lineBuf == NULL) || (parseFileArg->blockBuf == NULL) ) { DISPLAY("Error: : memory allocation failed \n"); exit(1); } parseFile1(parseFileArg); free(parseFileArg->blockBuf); free(parseFileArg->lineBuf); if (inFile != stdin) fclose(inFile); /* Show error/warning messages. All messages are copied from md5sum.c */ if (report->nProperlyFormattedLines == 0) { DISPLAY("%s: no properly formatted xxHash checksum lines found\n", inFileName); } else if (!statusOnly) { if (report->nImproperlyFormattedLines) { DISPLAYRESULT("%lu %s are improperly formatted\n" , report->nImproperlyFormattedLines , report->nImproperlyFormattedLines == 1 ? "line" : "lines"); } if (report->nOpenOrReadFailures) { DISPLAYRESULT("%lu listed %s could not be read\n" , report->nOpenOrReadFailures , report->nOpenOrReadFailures == 1 ? "file" : "files"); } if (report->nMismatchedChecksums) { DISPLAYRESULT("%lu computed %s did NOT match\n" , report->nMismatchedChecksums , report->nMismatchedChecksums == 1 ? "checksum" : "checksums"); } } /* Result (exit) code logic is copied from * gnu coreutils/src/md5sum.c digest_check() */ result = report->nProperlyFormattedLines != 0 && report->nMismatchedChecksums == 0 && report->nOpenOrReadFailures == 0 && (!strictMode || report->nImproperlyFormattedLines == 0) && report->quit == 0; return result; } static int checkFiles(const char*const* fnList, int fnTotal, const Display_endianess displayEndianess, U32 strictMode, U32 statusOnly, U32 warn, U32 quiet) { int ok = 1; /* Special case for stdinName "-", * note: stdinName is not a string. It's special pointer. */ if (fnTotal==0) { ok &= checkFile(stdinName, displayEndianess, strictMode, statusOnly, warn, quiet); } else { int fnNb; for (fnNb=0; fnNb='0') && (**stringPtr <='9')) { if (result > max) return 1; /* overflow error */ result *= 10; result += (U32)(**stringPtr - '0'); (*stringPtr)++ ; } if ((**stringPtr=='K') || (**stringPtr=='M')) { U32 const maxK = ((U32)(-1)) >> 10; if (result > maxK) return 1; /* overflow error */ result <<= 10; if (**stringPtr=='M') { if (result > maxK) return 1; /* overflow error */ result <<= 10; } (*stringPtr)++; /* skip `K` or `M` */ if (**stringPtr=='i') (*stringPtr)++; if (**stringPtr=='B') (*stringPtr)++; } *value = result; return 0; } /*! * readU32FromChar(): * @return: unsigned integer value read from input in `char` format. * allows and interprets K, KB, KiB, M, MB and MiB suffix. * Will also modify `*stringPtr`, advancing it to position where it stopped reading. * Note: function will exit() program if digit sequence overflows */ static U32 readU32FromChar(const char** stringPtr) { U32 result; if (readU32FromCharChecked(stringPtr, &result)) { static const char errorMsg[] = "Error: numeric value too large"; errorOut(errorMsg); } return result; } static int XXH_main(int argc, const char* const* argv) { int i, filenamesStart = 0; const char* const exename = argv[0]; U32 benchmarkMode = 0; U32 fileCheckMode = 0; U32 strictMode = 0; U32 statusOnly = 0; U32 warn = 0; U32 selectBenchIDs= 0; /* 0 == use default k_testIDs_default, kBenchAll == bench all */ static const U32 kBenchAll = 99; size_t keySize = XXH_DEFAULT_SAMPLE_SIZE; AlgoSelected algo = g_defaultAlgo; Display_endianess displayEndianess = big_endian; Display_convention convention = display_gnu; /* special case: xxhNNsum default to NN bits checksum */ if (strstr(exename, "xxh32sum") != NULL) algo = algo_xxh32; if (strstr(exename, "xxh64sum") != NULL) algo = algo_xxh64; if (strstr(exename, "xxh128sum") != NULL) algo = algo_xxh128; for(i=1; i