10c16b537SWarner Losh /* 237f1f268SConrad Meyer * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc. 30c16b537SWarner Losh * All rights reserved. 40c16b537SWarner Losh * 50c16b537SWarner Losh * This source code is licensed under both the BSD-style license (found in the 60c16b537SWarner Losh * LICENSE file in the root directory of this source tree) and the GPLv2 (found 70c16b537SWarner Losh * in the COPYING file in the root directory of this source tree). 80c16b537SWarner Losh * You may select, at your option, one of the above-listed licenses. 90c16b537SWarner Losh */ 100c16b537SWarner Losh 110c16b537SWarner Losh 120c16b537SWarner Losh /*- Dependencies -*/ 130c16b537SWarner Losh #include "zstd_v05.h" 1437f1f268SConrad Meyer #include "../common/error_private.h" 150c16b537SWarner Losh 160c16b537SWarner Losh 170c16b537SWarner Losh /* ****************************************************************** 180c16b537SWarner Losh mem.h 190c16b537SWarner Losh low-level memory access routines 200c16b537SWarner Losh Copyright (C) 2013-2015, Yann Collet. 210c16b537SWarner Losh 220c16b537SWarner Losh BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php) 230c16b537SWarner Losh 240c16b537SWarner Losh Redistribution and use in source and binary forms, with or without 250c16b537SWarner Losh modification, are permitted provided that the following conditions are 260c16b537SWarner Losh met: 270c16b537SWarner Losh 280c16b537SWarner Losh * Redistributions of source code must retain the above copyright 290c16b537SWarner Losh notice, this list of conditions and the following disclaimer. 300c16b537SWarner Losh * Redistributions in binary form must reproduce the above 310c16b537SWarner Losh copyright notice, this list of conditions and the following disclaimer 320c16b537SWarner Losh in the documentation and/or other materials provided with the 330c16b537SWarner Losh distribution. 340c16b537SWarner Losh 350c16b537SWarner Losh THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 360c16b537SWarner Losh "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 370c16b537SWarner Losh LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 380c16b537SWarner Losh A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 390c16b537SWarner Losh OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 400c16b537SWarner Losh SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 410c16b537SWarner Losh LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 420c16b537SWarner Losh DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 430c16b537SWarner Losh THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 440c16b537SWarner Losh (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 450c16b537SWarner Losh OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 460c16b537SWarner Losh 470c16b537SWarner Losh You can contact the author at : 480c16b537SWarner Losh - FSEv05 source repository : https://github.com/Cyan4973/FiniteStateEntropy 490c16b537SWarner Losh - Public forum : https://groups.google.com/forum/#!forum/lz4c 500c16b537SWarner Losh ****************************************************************** */ 510c16b537SWarner Losh #ifndef MEM_H_MODULE 520c16b537SWarner Losh #define MEM_H_MODULE 530c16b537SWarner Losh 540c16b537SWarner Losh #if defined (__cplusplus) 550c16b537SWarner Losh extern "C" { 560c16b537SWarner Losh #endif 570c16b537SWarner Losh 580c16b537SWarner Losh /*-**************************************** 590c16b537SWarner Losh * Dependencies 600c16b537SWarner Losh ******************************************/ 610c16b537SWarner Losh #include <stddef.h> /* size_t, ptrdiff_t */ 620c16b537SWarner Losh #include <string.h> /* memcpy */ 630c16b537SWarner Losh 640c16b537SWarner Losh 650c16b537SWarner Losh /*-**************************************** 660c16b537SWarner Losh * Compiler specifics 670c16b537SWarner Losh ******************************************/ 680c16b537SWarner Losh #if defined(__GNUC__) 690c16b537SWarner Losh # define MEM_STATIC static __attribute__((unused)) 700c16b537SWarner Losh #elif defined (__cplusplus) || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */) 710c16b537SWarner Losh # define MEM_STATIC static inline 720c16b537SWarner Losh #elif defined(_MSC_VER) 730c16b537SWarner Losh # define MEM_STATIC static __inline 740c16b537SWarner Losh #else 750c16b537SWarner Losh # define MEM_STATIC static /* this version may generate warnings for unused static functions; disable the relevant warning */ 760c16b537SWarner Losh #endif 770c16b537SWarner Losh 780c16b537SWarner Losh 790c16b537SWarner Losh /*-************************************************************** 800c16b537SWarner Losh * Basic Types 810c16b537SWarner Losh *****************************************************************/ 820c16b537SWarner Losh #if defined (__cplusplus) || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */) 83*f7cd7fe5SConrad Meyer # if defined(_AIX) 84*f7cd7fe5SConrad Meyer # include <inttypes.h> 85*f7cd7fe5SConrad Meyer # else 86*f7cd7fe5SConrad Meyer # include <stdint.h> /* intptr_t */ 87*f7cd7fe5SConrad Meyer # endif 880c16b537SWarner Losh typedef uint8_t BYTE; 890c16b537SWarner Losh typedef uint16_t U16; 900c16b537SWarner Losh typedef int16_t S16; 910c16b537SWarner Losh typedef uint32_t U32; 920c16b537SWarner Losh typedef int32_t S32; 930c16b537SWarner Losh typedef uint64_t U64; 940c16b537SWarner Losh typedef int64_t S64; 950c16b537SWarner Losh #else 960c16b537SWarner Losh typedef unsigned char BYTE; 970c16b537SWarner Losh typedef unsigned short U16; 980c16b537SWarner Losh typedef signed short S16; 990c16b537SWarner Losh typedef unsigned int U32; 1000c16b537SWarner Losh typedef signed int S32; 1010c16b537SWarner Losh typedef unsigned long long U64; 1020c16b537SWarner Losh typedef signed long long S64; 1030c16b537SWarner Losh #endif 1040c16b537SWarner Losh 1050c16b537SWarner Losh 1060c16b537SWarner Losh /*-************************************************************** 1070c16b537SWarner Losh * Memory I/O 1080c16b537SWarner Losh *****************************************************************/ 1090c16b537SWarner Losh /* MEM_FORCE_MEMORY_ACCESS : 1100c16b537SWarner Losh * By default, access to unaligned memory is controlled by `memcpy()`, which is safe and portable. 1110c16b537SWarner Losh * Unfortunately, on some target/compiler combinations, the generated assembly is sub-optimal. 1120c16b537SWarner Losh * The below switch allow to select different access method for improved performance. 1130c16b537SWarner Losh * Method 0 (default) : use `memcpy()`. Safe and portable. 1140c16b537SWarner Losh * Method 1 : `__packed` statement. It depends on compiler extension (ie, not portable). 1150c16b537SWarner Losh * This method is safe if your compiler supports it, and *generally* as fast or faster than `memcpy`. 1160c16b537SWarner Losh * Method 2 : direct access. This method is portable but violate C standard. 1170c16b537SWarner Losh * It can generate buggy code on targets depending on alignment. 1180c16b537SWarner Losh * In some circumstances, it's the only known way to get the most performance (ie GCC + ARMv6) 1190c16b537SWarner Losh * See http://fastcompression.blogspot.fr/2015/08/accessing-unaligned-memory.html for details. 1200c16b537SWarner Losh * Prefer these methods in priority order (0 > 1 > 2) 1210c16b537SWarner Losh */ 1220c16b537SWarner Losh #ifndef MEM_FORCE_MEMORY_ACCESS /* can be defined externally, on command line for example */ 1230c16b537SWarner Losh # if defined(__GNUC__) && ( defined(__ARM_ARCH_6__) || defined(__ARM_ARCH_6J__) || defined(__ARM_ARCH_6K__) || defined(__ARM_ARCH_6Z__) || defined(__ARM_ARCH_6ZK__) || defined(__ARM_ARCH_6T2__) ) 1240c16b537SWarner Losh # define MEM_FORCE_MEMORY_ACCESS 2 1250c16b537SWarner Losh # elif (defined(__INTEL_COMPILER) && !defined(WIN32)) || \ 1260c16b537SWarner Losh (defined(__GNUC__) && ( defined(__ARM_ARCH_7__) || defined(__ARM_ARCH_7A__) || defined(__ARM_ARCH_7R__) || defined(__ARM_ARCH_7M__) || defined(__ARM_ARCH_7S__) )) 1270c16b537SWarner Losh # define MEM_FORCE_MEMORY_ACCESS 1 1280c16b537SWarner Losh # endif 1290c16b537SWarner Losh #endif 1300c16b537SWarner Losh 1310c16b537SWarner Losh MEM_STATIC unsigned MEM_32bits(void) { return sizeof(void*)==4; } 1320c16b537SWarner Losh MEM_STATIC unsigned MEM_64bits(void) { return sizeof(void*)==8; } 1330c16b537SWarner Losh 1340c16b537SWarner Losh MEM_STATIC unsigned MEM_isLittleEndian(void) 1350c16b537SWarner Losh { 1360c16b537SWarner Losh const union { U32 u; BYTE c[4]; } one = { 1 }; /* don't use static : performance detrimental */ 1370c16b537SWarner Losh return one.c[0]; 1380c16b537SWarner Losh } 1390c16b537SWarner Losh 1400c16b537SWarner Losh #if defined(MEM_FORCE_MEMORY_ACCESS) && (MEM_FORCE_MEMORY_ACCESS==2) 1410c16b537SWarner Losh 1420c16b537SWarner Losh /* violates C standard, by lying on structure alignment. 1430c16b537SWarner Losh Only use if no other choice to achieve best performance on target platform */ 1440c16b537SWarner Losh MEM_STATIC U16 MEM_read16(const void* memPtr) { return *(const U16*) memPtr; } 1450c16b537SWarner Losh MEM_STATIC U32 MEM_read32(const void* memPtr) { return *(const U32*) memPtr; } 1460c16b537SWarner Losh MEM_STATIC U64 MEM_read64(const void* memPtr) { return *(const U64*) memPtr; } 1470c16b537SWarner Losh 1480c16b537SWarner Losh MEM_STATIC void MEM_write16(void* memPtr, U16 value) { *(U16*)memPtr = value; } 1490c16b537SWarner Losh MEM_STATIC void MEM_write32(void* memPtr, U32 value) { *(U32*)memPtr = value; } 1500c16b537SWarner Losh MEM_STATIC void MEM_write64(void* memPtr, U64 value) { *(U64*)memPtr = value; } 1510c16b537SWarner Losh 1520c16b537SWarner Losh #elif defined(MEM_FORCE_MEMORY_ACCESS) && (MEM_FORCE_MEMORY_ACCESS==1) 1530c16b537SWarner Losh 1540c16b537SWarner Losh /* __pack instructions are safer, but compiler specific, hence potentially problematic for some compilers */ 1550c16b537SWarner Losh /* currently only defined for gcc and icc */ 1560c16b537SWarner Losh typedef union { U16 u16; U32 u32; U64 u64; size_t st; } __attribute__((packed)) unalign; 1570c16b537SWarner Losh 1580c16b537SWarner Losh MEM_STATIC U16 MEM_read16(const void* ptr) { return ((const unalign*)ptr)->u16; } 1590c16b537SWarner Losh MEM_STATIC U32 MEM_read32(const void* ptr) { return ((const unalign*)ptr)->u32; } 1600c16b537SWarner Losh MEM_STATIC U64 MEM_read64(const void* ptr) { return ((const unalign*)ptr)->u64; } 1610c16b537SWarner Losh 1620c16b537SWarner Losh MEM_STATIC void MEM_write16(void* memPtr, U16 value) { ((unalign*)memPtr)->u16 = value; } 1630c16b537SWarner Losh MEM_STATIC void MEM_write32(void* memPtr, U32 value) { ((unalign*)memPtr)->u32 = value; } 1640c16b537SWarner Losh MEM_STATIC void MEM_write64(void* memPtr, U64 value) { ((unalign*)memPtr)->u64 = value; } 1650c16b537SWarner Losh 1660c16b537SWarner Losh #else 1670c16b537SWarner Losh 1680c16b537SWarner Losh /* default method, safe and standard. 1690c16b537SWarner Losh can sometimes prove slower */ 1700c16b537SWarner Losh 1710c16b537SWarner Losh MEM_STATIC U16 MEM_read16(const void* memPtr) 1720c16b537SWarner Losh { 1730c16b537SWarner Losh U16 val; memcpy(&val, memPtr, sizeof(val)); return val; 1740c16b537SWarner Losh } 1750c16b537SWarner Losh 1760c16b537SWarner Losh MEM_STATIC U32 MEM_read32(const void* memPtr) 1770c16b537SWarner Losh { 1780c16b537SWarner Losh U32 val; memcpy(&val, memPtr, sizeof(val)); return val; 1790c16b537SWarner Losh } 1800c16b537SWarner Losh 1810c16b537SWarner Losh MEM_STATIC U64 MEM_read64(const void* memPtr) 1820c16b537SWarner Losh { 1830c16b537SWarner Losh U64 val; memcpy(&val, memPtr, sizeof(val)); return val; 1840c16b537SWarner Losh } 1850c16b537SWarner Losh 1860c16b537SWarner Losh MEM_STATIC void MEM_write16(void* memPtr, U16 value) 1870c16b537SWarner Losh { 1880c16b537SWarner Losh memcpy(memPtr, &value, sizeof(value)); 1890c16b537SWarner Losh } 1900c16b537SWarner Losh 1910c16b537SWarner Losh MEM_STATIC void MEM_write32(void* memPtr, U32 value) 1920c16b537SWarner Losh { 1930c16b537SWarner Losh memcpy(memPtr, &value, sizeof(value)); 1940c16b537SWarner Losh } 1950c16b537SWarner Losh 1960c16b537SWarner Losh MEM_STATIC void MEM_write64(void* memPtr, U64 value) 1970c16b537SWarner Losh { 1980c16b537SWarner Losh memcpy(memPtr, &value, sizeof(value)); 1990c16b537SWarner Losh } 2000c16b537SWarner Losh 2010c16b537SWarner Losh #endif /* MEM_FORCE_MEMORY_ACCESS */ 2020c16b537SWarner Losh 2030c16b537SWarner Losh 2040c16b537SWarner Losh MEM_STATIC U16 MEM_readLE16(const void* memPtr) 2050c16b537SWarner Losh { 2060c16b537SWarner Losh if (MEM_isLittleEndian()) 2070c16b537SWarner Losh return MEM_read16(memPtr); 2080c16b537SWarner Losh else { 2090c16b537SWarner Losh const BYTE* p = (const BYTE*)memPtr; 2100c16b537SWarner Losh return (U16)(p[0] + (p[1]<<8)); 2110c16b537SWarner Losh } 2120c16b537SWarner Losh } 2130c16b537SWarner Losh 2140c16b537SWarner Losh MEM_STATIC void MEM_writeLE16(void* memPtr, U16 val) 2150c16b537SWarner Losh { 2160c16b537SWarner Losh if (MEM_isLittleEndian()) { 2170c16b537SWarner Losh MEM_write16(memPtr, val); 2180c16b537SWarner Losh } else { 2190c16b537SWarner Losh BYTE* p = (BYTE*)memPtr; 2200c16b537SWarner Losh p[0] = (BYTE)val; 2210c16b537SWarner Losh p[1] = (BYTE)(val>>8); 2220c16b537SWarner Losh } 2230c16b537SWarner Losh } 2240c16b537SWarner Losh 2250c16b537SWarner Losh MEM_STATIC U32 MEM_readLE32(const void* memPtr) 2260c16b537SWarner Losh { 2270c16b537SWarner Losh if (MEM_isLittleEndian()) 2280c16b537SWarner Losh return MEM_read32(memPtr); 2290c16b537SWarner Losh else { 2300c16b537SWarner Losh const BYTE* p = (const BYTE*)memPtr; 2310c16b537SWarner Losh return (U32)((U32)p[0] + ((U32)p[1]<<8) + ((U32)p[2]<<16) + ((U32)p[3]<<24)); 2320c16b537SWarner Losh } 2330c16b537SWarner Losh } 2340c16b537SWarner Losh 2350c16b537SWarner Losh 2360c16b537SWarner Losh MEM_STATIC U64 MEM_readLE64(const void* memPtr) 2370c16b537SWarner Losh { 2380c16b537SWarner Losh if (MEM_isLittleEndian()) 2390c16b537SWarner Losh return MEM_read64(memPtr); 2400c16b537SWarner Losh else { 2410c16b537SWarner Losh const BYTE* p = (const BYTE*)memPtr; 2420c16b537SWarner Losh return (U64)((U64)p[0] + ((U64)p[1]<<8) + ((U64)p[2]<<16) + ((U64)p[3]<<24) 2430c16b537SWarner Losh + ((U64)p[4]<<32) + ((U64)p[5]<<40) + ((U64)p[6]<<48) + ((U64)p[7]<<56)); 2440c16b537SWarner Losh } 2450c16b537SWarner Losh } 2460c16b537SWarner Losh 2470c16b537SWarner Losh 2480c16b537SWarner Losh MEM_STATIC size_t MEM_readLEST(const void* memPtr) 2490c16b537SWarner Losh { 2500c16b537SWarner Losh if (MEM_32bits()) 2510c16b537SWarner Losh return (size_t)MEM_readLE32(memPtr); 2520c16b537SWarner Losh else 2530c16b537SWarner Losh return (size_t)MEM_readLE64(memPtr); 2540c16b537SWarner Losh } 2550c16b537SWarner Losh 2560c16b537SWarner Losh 2570c16b537SWarner Losh #if defined (__cplusplus) 2580c16b537SWarner Losh } 2590c16b537SWarner Losh #endif 2600c16b537SWarner Losh 2610c16b537SWarner Losh #endif /* MEM_H_MODULE */ 2620c16b537SWarner Losh 2630c16b537SWarner Losh /* 2640c16b537SWarner Losh zstd - standard compression library 2650c16b537SWarner Losh Header File for static linking only 2660c16b537SWarner Losh Copyright (C) 2014-2016, Yann Collet. 2670c16b537SWarner Losh 2680c16b537SWarner Losh BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php) 2690c16b537SWarner Losh 2700c16b537SWarner Losh Redistribution and use in source and binary forms, with or without 2710c16b537SWarner Losh modification, are permitted provided that the following conditions are 2720c16b537SWarner Losh met: 2730c16b537SWarner Losh * Redistributions of source code must retain the above copyright 2740c16b537SWarner Losh notice, this list of conditions and the following disclaimer. 2750c16b537SWarner Losh * Redistributions in binary form must reproduce the above 2760c16b537SWarner Losh copyright notice, this list of conditions and the following disclaimer 2770c16b537SWarner Losh in the documentation and/or other materials provided with the 2780c16b537SWarner Losh distribution. 2790c16b537SWarner Losh THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 2800c16b537SWarner Losh "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 2810c16b537SWarner Losh LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 2820c16b537SWarner Losh A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 2830c16b537SWarner Losh OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 2840c16b537SWarner Losh SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 2850c16b537SWarner Losh LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 2860c16b537SWarner Losh DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 2870c16b537SWarner Losh THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 2880c16b537SWarner Losh (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 2890c16b537SWarner Losh OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 2900c16b537SWarner Losh 2910c16b537SWarner Losh You can contact the author at : 2920c16b537SWarner Losh - zstd homepage : http://www.zstd.net 2930c16b537SWarner Losh */ 2940c16b537SWarner Losh #ifndef ZSTD_STATIC_H 2950c16b537SWarner Losh #define ZSTD_STATIC_H 2960c16b537SWarner Losh 2970c16b537SWarner Losh /* The prototypes defined within this file are considered experimental. 2980c16b537SWarner Losh * They should not be used in the context DLL as they may change in the future. 2990c16b537SWarner Losh * Prefer static linking if you need them, to control breaking version changes issues. 3000c16b537SWarner Losh */ 3010c16b537SWarner Losh 3020c16b537SWarner Losh #if defined (__cplusplus) 3030c16b537SWarner Losh extern "C" { 3040c16b537SWarner Losh #endif 3050c16b537SWarner Losh 3060c16b537SWarner Losh 3070c16b537SWarner Losh 3080c16b537SWarner Losh /*-************************************* 3090c16b537SWarner Losh * Types 3100c16b537SWarner Losh ***************************************/ 3110c16b537SWarner Losh #define ZSTDv05_WINDOWLOG_ABSOLUTEMIN 11 3120c16b537SWarner Losh 3130c16b537SWarner Losh 3140c16b537SWarner Losh /*-************************************* 3150c16b537SWarner Losh * Advanced functions 3160c16b537SWarner Losh ***************************************/ 3170c16b537SWarner Losh /*- Advanced Decompression functions -*/ 3180c16b537SWarner Losh 3190c16b537SWarner Losh /*! ZSTDv05_decompress_usingPreparedDCtx() : 3200c16b537SWarner Losh * Same as ZSTDv05_decompress_usingDict, but using a reference context `preparedDCtx`, where dictionary has been loaded. 3210c16b537SWarner Losh * It avoids reloading the dictionary each time. 3220c16b537SWarner Losh * `preparedDCtx` must have been properly initialized using ZSTDv05_decompressBegin_usingDict(). 3230c16b537SWarner Losh * Requires 2 contexts : 1 for reference, which will not be modified, and 1 to run the decompression operation */ 3240c16b537SWarner Losh size_t ZSTDv05_decompress_usingPreparedDCtx( 3250c16b537SWarner Losh ZSTDv05_DCtx* dctx, const ZSTDv05_DCtx* preparedDCtx, 3260c16b537SWarner Losh void* dst, size_t dstCapacity, 3270c16b537SWarner Losh const void* src, size_t srcSize); 3280c16b537SWarner Losh 3290c16b537SWarner Losh 3300c16b537SWarner Losh /* ************************************** 3310c16b537SWarner Losh * Streaming functions (direct mode) 3320c16b537SWarner Losh ****************************************/ 3330c16b537SWarner Losh size_t ZSTDv05_decompressBegin(ZSTDv05_DCtx* dctx); 3340c16b537SWarner Losh 3350c16b537SWarner Losh /* 3360c16b537SWarner Losh Streaming decompression, direct mode (bufferless) 3370c16b537SWarner Losh 3380c16b537SWarner Losh A ZSTDv05_DCtx object is required to track streaming operations. 3390c16b537SWarner Losh Use ZSTDv05_createDCtx() / ZSTDv05_freeDCtx() to manage it. 3400c16b537SWarner Losh A ZSTDv05_DCtx object can be re-used multiple times. 3410c16b537SWarner Losh 3420c16b537SWarner Losh First typical operation is to retrieve frame parameters, using ZSTDv05_getFrameParams(). 3430c16b537SWarner Losh This operation is independent, and just needs enough input data to properly decode the frame header. 3440c16b537SWarner Losh Objective is to retrieve *params.windowlog, to know minimum amount of memory required during decoding. 3450c16b537SWarner Losh Result : 0 when successful, it means the ZSTDv05_parameters structure has been filled. 3460c16b537SWarner Losh >0 : means there is not enough data into src. Provides the expected size to successfully decode header. 3470c16b537SWarner Losh errorCode, which can be tested using ZSTDv05_isError() 3480c16b537SWarner Losh 3490c16b537SWarner Losh Start decompression, with ZSTDv05_decompressBegin() or ZSTDv05_decompressBegin_usingDict() 3500c16b537SWarner Losh Alternatively, you can copy a prepared context, using ZSTDv05_copyDCtx() 3510c16b537SWarner Losh 3520c16b537SWarner Losh Then use ZSTDv05_nextSrcSizeToDecompress() and ZSTDv05_decompressContinue() alternatively. 3530c16b537SWarner Losh ZSTDv05_nextSrcSizeToDecompress() tells how much bytes to provide as 'srcSize' to ZSTDv05_decompressContinue(). 3540c16b537SWarner Losh ZSTDv05_decompressContinue() requires this exact amount of bytes, or it will fail. 3550c16b537SWarner Losh ZSTDv05_decompressContinue() needs previous data blocks during decompression, up to (1 << windowlog). 3560c16b537SWarner Losh They should preferably be located contiguously, prior to current block. Alternatively, a round buffer is also possible. 3570c16b537SWarner Losh 3580c16b537SWarner Losh @result of ZSTDv05_decompressContinue() is the number of bytes regenerated within 'dst'. 3590c16b537SWarner Losh It can be zero, which is not an error; it just means ZSTDv05_decompressContinue() has decoded some header. 3600c16b537SWarner Losh 3610c16b537SWarner Losh A frame is fully decoded when ZSTDv05_nextSrcSizeToDecompress() returns zero. 3620c16b537SWarner Losh Context can then be reset to start a new decompression. 3630c16b537SWarner Losh */ 3640c16b537SWarner Losh 3650c16b537SWarner Losh 3660c16b537SWarner Losh /* ************************************** 3670c16b537SWarner Losh * Block functions 3680c16b537SWarner Losh ****************************************/ 3690c16b537SWarner Losh /*! Block functions produce and decode raw zstd blocks, without frame metadata. 3700c16b537SWarner Losh User will have to take in charge required information to regenerate data, such as block sizes. 3710c16b537SWarner Losh 3720c16b537SWarner Losh A few rules to respect : 3730c16b537SWarner Losh - Uncompressed block size must be <= 128 KB 3740c16b537SWarner Losh - Compressing or decompressing requires a context structure 3750c16b537SWarner Losh + Use ZSTDv05_createCCtx() and ZSTDv05_createDCtx() 3760c16b537SWarner Losh - It is necessary to init context before starting 3770c16b537SWarner Losh + compression : ZSTDv05_compressBegin() 3780c16b537SWarner Losh + decompression : ZSTDv05_decompressBegin() 3790c16b537SWarner Losh + variants _usingDict() are also allowed 3800c16b537SWarner Losh + copyCCtx() and copyDCtx() work too 3810c16b537SWarner Losh - When a block is considered not compressible enough, ZSTDv05_compressBlock() result will be zero. 3820c16b537SWarner Losh In which case, nothing is produced into `dst`. 3830c16b537SWarner Losh + User must test for such outcome and deal directly with uncompressed data 3840c16b537SWarner Losh + ZSTDv05_decompressBlock() doesn't accept uncompressed data as input !! 3850c16b537SWarner Losh */ 3860c16b537SWarner Losh 3870c16b537SWarner Losh size_t ZSTDv05_decompressBlock(ZSTDv05_DCtx* dctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize); 3880c16b537SWarner Losh 3890c16b537SWarner Losh 3900c16b537SWarner Losh 3910c16b537SWarner Losh 3920c16b537SWarner Losh #if defined (__cplusplus) 3930c16b537SWarner Losh } 3940c16b537SWarner Losh #endif 3950c16b537SWarner Losh 3960c16b537SWarner Losh #endif /* ZSTDv05_STATIC_H */ 3970c16b537SWarner Losh 3980c16b537SWarner Losh 3990c16b537SWarner Losh /* 4000c16b537SWarner Losh zstd_internal - common functions to include 4010c16b537SWarner Losh Header File for include 4020c16b537SWarner Losh Copyright (C) 2014-2016, Yann Collet. 4030c16b537SWarner Losh 4040c16b537SWarner Losh BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php) 4050c16b537SWarner Losh 4060c16b537SWarner Losh Redistribution and use in source and binary forms, with or without 4070c16b537SWarner Losh modification, are permitted provided that the following conditions are 4080c16b537SWarner Losh met: 4090c16b537SWarner Losh * Redistributions of source code must retain the above copyright 4100c16b537SWarner Losh notice, this list of conditions and the following disclaimer. 4110c16b537SWarner Losh * Redistributions in binary form must reproduce the above 4120c16b537SWarner Losh copyright notice, this list of conditions and the following disclaimer 4130c16b537SWarner Losh in the documentation and/or other materials provided with the 4140c16b537SWarner Losh distribution. 4150c16b537SWarner Losh THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 4160c16b537SWarner Losh "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 4170c16b537SWarner Losh LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 4180c16b537SWarner Losh A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 4190c16b537SWarner Losh OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 4200c16b537SWarner Losh SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 4210c16b537SWarner Losh LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 4220c16b537SWarner Losh DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 4230c16b537SWarner Losh THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 4240c16b537SWarner Losh (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 4250c16b537SWarner Losh OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 4260c16b537SWarner Losh 4270c16b537SWarner Losh You can contact the author at : 4280c16b537SWarner Losh - zstd source repository : https://github.com/Cyan4973/zstd 4290c16b537SWarner Losh */ 4300c16b537SWarner Losh #ifndef ZSTD_CCOMMON_H_MODULE 4310c16b537SWarner Losh #define ZSTD_CCOMMON_H_MODULE 4320c16b537SWarner Losh 4330c16b537SWarner Losh 4340c16b537SWarner Losh 4350c16b537SWarner Losh /*-************************************* 4360c16b537SWarner Losh * Common macros 4370c16b537SWarner Losh ***************************************/ 4380c16b537SWarner Losh #define MIN(a,b) ((a)<(b) ? (a) : (b)) 4390c16b537SWarner Losh #define MAX(a,b) ((a)>(b) ? (a) : (b)) 4400c16b537SWarner Losh 4410c16b537SWarner Losh 4420c16b537SWarner Losh /*-************************************* 4430c16b537SWarner Losh * Common constants 4440c16b537SWarner Losh ***************************************/ 4450c16b537SWarner Losh #define ZSTDv05_DICT_MAGIC 0xEC30A435 4460c16b537SWarner Losh 4470c16b537SWarner Losh #define KB *(1 <<10) 4480c16b537SWarner Losh #define MB *(1 <<20) 4490c16b537SWarner Losh #define GB *(1U<<30) 4500c16b537SWarner Losh 4510c16b537SWarner Losh #define BLOCKSIZE (128 KB) /* define, for static allocation */ 4520c16b537SWarner Losh 4530c16b537SWarner Losh static const size_t ZSTDv05_blockHeaderSize = 3; 4540c16b537SWarner Losh static const size_t ZSTDv05_frameHeaderSize_min = 5; 4550c16b537SWarner Losh #define ZSTDv05_frameHeaderSize_max 5 /* define, for static allocation */ 4560c16b537SWarner Losh 4570c16b537SWarner Losh #define BITv057 128 4580c16b537SWarner Losh #define BITv056 64 4590c16b537SWarner Losh #define BITv055 32 4600c16b537SWarner Losh #define BITv054 16 4610c16b537SWarner Losh #define BITv051 2 4620c16b537SWarner Losh #define BITv050 1 4630c16b537SWarner Losh 4640c16b537SWarner Losh #define IS_HUFv05 0 4650c16b537SWarner Losh #define IS_PCH 1 4660c16b537SWarner Losh #define IS_RAW 2 4670c16b537SWarner Losh #define IS_RLE 3 4680c16b537SWarner Losh 4690c16b537SWarner Losh #define MINMATCH 4 4700c16b537SWarner Losh #define REPCODE_STARTVALUE 1 4710c16b537SWarner Losh 4720c16b537SWarner Losh #define Litbits 8 4730c16b537SWarner Losh #define MLbits 7 4740c16b537SWarner Losh #define LLbits 6 4750c16b537SWarner Losh #define Offbits 5 4760c16b537SWarner Losh #define MaxLit ((1<<Litbits) - 1) 4770c16b537SWarner Losh #define MaxML ((1<<MLbits) - 1) 4780c16b537SWarner Losh #define MaxLL ((1<<LLbits) - 1) 4790c16b537SWarner Losh #define MaxOff ((1<<Offbits)- 1) 4800c16b537SWarner Losh #define MLFSEv05Log 10 4810c16b537SWarner Losh #define LLFSEv05Log 10 4820c16b537SWarner Losh #define OffFSEv05Log 9 4830c16b537SWarner Losh #define MaxSeq MAX(MaxLL, MaxML) 4840c16b537SWarner Losh 4850c16b537SWarner Losh #define FSEv05_ENCODING_RAW 0 4860c16b537SWarner Losh #define FSEv05_ENCODING_RLE 1 4870c16b537SWarner Losh #define FSEv05_ENCODING_STATIC 2 4880c16b537SWarner Losh #define FSEv05_ENCODING_DYNAMIC 3 4890c16b537SWarner Losh 4900c16b537SWarner Losh 4910c16b537SWarner Losh #define HufLog 12 4920c16b537SWarner Losh 4930c16b537SWarner Losh #define MIN_SEQUENCES_SIZE 1 /* nbSeq==0 */ 4940c16b537SWarner Losh #define MIN_CBLOCK_SIZE (1 /*litCSize*/ + 1 /* RLE or RAW */ + MIN_SEQUENCES_SIZE /* nbSeq==0 */) /* for a non-null block */ 4950c16b537SWarner Losh 4960c16b537SWarner Losh #define WILDCOPY_OVERLENGTH 8 4970c16b537SWarner Losh 4982b9c00cbSConrad Meyer #define ZSTD_CONTENTSIZE_ERROR (0ULL - 2) 4992b9c00cbSConrad Meyer 5000c16b537SWarner Losh typedef enum { bt_compressed, bt_raw, bt_rle, bt_end } blockType_t; 5010c16b537SWarner Losh 5020c16b537SWarner Losh 5030c16b537SWarner Losh /*-******************************************* 5040c16b537SWarner Losh * Shared functions to include for inlining 5050c16b537SWarner Losh *********************************************/ 5060c16b537SWarner Losh static void ZSTDv05_copy8(void* dst, const void* src) { memcpy(dst, src, 8); } 5070c16b537SWarner Losh 5080c16b537SWarner Losh #define COPY8(d,s) { ZSTDv05_copy8(d,s); d+=8; s+=8; } 5090c16b537SWarner Losh 5100c16b537SWarner Losh /*! ZSTDv05_wildcopy() : 5110c16b537SWarner Losh * custom version of memcpy(), can copy up to 7 bytes too many (8 bytes if length==0) */ 5120c16b537SWarner Losh MEM_STATIC void ZSTDv05_wildcopy(void* dst, const void* src, ptrdiff_t length) 5130c16b537SWarner Losh { 5140c16b537SWarner Losh const BYTE* ip = (const BYTE*)src; 5150c16b537SWarner Losh BYTE* op = (BYTE*)dst; 5160c16b537SWarner Losh BYTE* const oend = op + length; 5170c16b537SWarner Losh do 5180c16b537SWarner Losh COPY8(op, ip) 5190c16b537SWarner Losh while (op < oend); 5200c16b537SWarner Losh } 5210c16b537SWarner Losh 5220c16b537SWarner Losh 5230c16b537SWarner Losh /*-******************************************* 5240c16b537SWarner Losh * Private interfaces 5250c16b537SWarner Losh *********************************************/ 5260c16b537SWarner Losh typedef struct { 5270c16b537SWarner Losh void* buffer; 5280c16b537SWarner Losh U32* offsetStart; 5290c16b537SWarner Losh U32* offset; 5300c16b537SWarner Losh BYTE* offCodeStart; 5310c16b537SWarner Losh BYTE* offCode; 5320c16b537SWarner Losh BYTE* litStart; 5330c16b537SWarner Losh BYTE* lit; 5340c16b537SWarner Losh BYTE* litLengthStart; 5350c16b537SWarner Losh BYTE* litLength; 5360c16b537SWarner Losh BYTE* matchLengthStart; 5370c16b537SWarner Losh BYTE* matchLength; 5380c16b537SWarner Losh BYTE* dumpsStart; 5390c16b537SWarner Losh BYTE* dumps; 5400c16b537SWarner Losh /* opt */ 5410c16b537SWarner Losh U32* matchLengthFreq; 5420c16b537SWarner Losh U32* litLengthFreq; 5430c16b537SWarner Losh U32* litFreq; 5440c16b537SWarner Losh U32* offCodeFreq; 5450c16b537SWarner Losh U32 matchLengthSum; 5460c16b537SWarner Losh U32 litLengthSum; 5470c16b537SWarner Losh U32 litSum; 5480c16b537SWarner Losh U32 offCodeSum; 5490c16b537SWarner Losh } seqStore_t; 5500c16b537SWarner Losh 5510c16b537SWarner Losh 5520c16b537SWarner Losh 5530c16b537SWarner Losh #endif /* ZSTDv05_CCOMMON_H_MODULE */ 5540c16b537SWarner Losh /* ****************************************************************** 5550c16b537SWarner Losh FSEv05 : Finite State Entropy coder 5560c16b537SWarner Losh header file 5570c16b537SWarner Losh Copyright (C) 2013-2015, Yann Collet. 5580c16b537SWarner Losh 5590c16b537SWarner Losh BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php) 5600c16b537SWarner Losh 5610c16b537SWarner Losh Redistribution and use in source and binary forms, with or without 5620c16b537SWarner Losh modification, are permitted provided that the following conditions are 5630c16b537SWarner Losh met: 5640c16b537SWarner Losh 5650c16b537SWarner Losh * Redistributions of source code must retain the above copyright 5660c16b537SWarner Losh notice, this list of conditions and the following disclaimer. 5670c16b537SWarner Losh * Redistributions in binary form must reproduce the above 5680c16b537SWarner Losh copyright notice, this list of conditions and the following disclaimer 5690c16b537SWarner Losh in the documentation and/or other materials provided with the 5700c16b537SWarner Losh distribution. 5710c16b537SWarner Losh 5720c16b537SWarner Losh THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 5730c16b537SWarner Losh "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 5740c16b537SWarner Losh LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 5750c16b537SWarner Losh A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 5760c16b537SWarner Losh OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 5770c16b537SWarner Losh SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 5780c16b537SWarner Losh LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 5790c16b537SWarner Losh DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 5800c16b537SWarner Losh THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 5810c16b537SWarner Losh (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 5820c16b537SWarner Losh OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 5830c16b537SWarner Losh 5840c16b537SWarner Losh You can contact the author at : 5850c16b537SWarner Losh - Source repository : https://github.com/Cyan4973/FiniteStateEntropy 5860c16b537SWarner Losh - Public forum : https://groups.google.com/forum/#!forum/lz4c 5870c16b537SWarner Losh ****************************************************************** */ 5880c16b537SWarner Losh #ifndef FSEv05_H 5890c16b537SWarner Losh #define FSEv05_H 5900c16b537SWarner Losh 5910c16b537SWarner Losh #if defined (__cplusplus) 5920c16b537SWarner Losh extern "C" { 5930c16b537SWarner Losh #endif 5940c16b537SWarner Losh 5950c16b537SWarner Losh 5960c16b537SWarner Losh /* ***************************************** 5970c16b537SWarner Losh * Includes 5980c16b537SWarner Losh ******************************************/ 5990c16b537SWarner Losh #include <stddef.h> /* size_t, ptrdiff_t */ 6000c16b537SWarner Losh 6010c16b537SWarner Losh 6020c16b537SWarner Losh /*-**************************************** 6030c16b537SWarner Losh * FSEv05 simple functions 6040c16b537SWarner Losh ******************************************/ 6050c16b537SWarner Losh size_t FSEv05_decompress(void* dst, size_t maxDstSize, 6060c16b537SWarner Losh const void* cSrc, size_t cSrcSize); 6070c16b537SWarner Losh /*! 6080c16b537SWarner Losh FSEv05_decompress(): 6090c16b537SWarner Losh Decompress FSEv05 data from buffer 'cSrc', of size 'cSrcSize', 6100c16b537SWarner Losh into already allocated destination buffer 'dst', of size 'maxDstSize'. 6110c16b537SWarner Losh return : size of regenerated data (<= maxDstSize) 6120c16b537SWarner Losh or an error code, which can be tested using FSEv05_isError() 6130c16b537SWarner Losh 6140c16b537SWarner Losh ** Important ** : FSEv05_decompress() doesn't decompress non-compressible nor RLE data !!! 6150c16b537SWarner Losh Why ? : making this distinction requires a header. 6160c16b537SWarner Losh Header management is intentionally delegated to the user layer, which can better manage special cases. 6170c16b537SWarner Losh */ 6180c16b537SWarner Losh 6190c16b537SWarner Losh 6200c16b537SWarner Losh /* ***************************************** 6210c16b537SWarner Losh * Tool functions 6220c16b537SWarner Losh ******************************************/ 6230c16b537SWarner Losh /* Error Management */ 6240c16b537SWarner Losh unsigned FSEv05_isError(size_t code); /* tells if a return value is an error code */ 6250c16b537SWarner Losh const char* FSEv05_getErrorName(size_t code); /* provides error code string (useful for debugging) */ 6260c16b537SWarner Losh 6270c16b537SWarner Losh 6280c16b537SWarner Losh 6290c16b537SWarner Losh 6300c16b537SWarner Losh /* ***************************************** 6310c16b537SWarner Losh * FSEv05 detailed API 6320c16b537SWarner Losh ******************************************/ 6330c16b537SWarner Losh /* *** DECOMPRESSION *** */ 6340c16b537SWarner Losh 6350c16b537SWarner Losh /*! 6360c16b537SWarner Losh FSEv05_readNCount(): 6370c16b537SWarner Losh Read compactly saved 'normalizedCounter' from 'rBuffer'. 6380c16b537SWarner Losh return : size read from 'rBuffer' 6390c16b537SWarner Losh or an errorCode, which can be tested using FSEv05_isError() 6400c16b537SWarner Losh maxSymbolValuePtr[0] and tableLogPtr[0] will also be updated with their respective values */ 6410c16b537SWarner Losh size_t FSEv05_readNCount (short* normalizedCounter, unsigned* maxSymbolValuePtr, unsigned* tableLogPtr, const void* rBuffer, size_t rBuffSize); 6420c16b537SWarner Losh 6430c16b537SWarner Losh /*! 6440c16b537SWarner Losh Constructor and Destructor of type FSEv05_DTable 6450c16b537SWarner Losh Note that its size depends on 'tableLog' */ 6460c16b537SWarner Losh typedef unsigned FSEv05_DTable; /* don't allocate that. It's just a way to be more restrictive than void* */ 6470c16b537SWarner Losh FSEv05_DTable* FSEv05_createDTable(unsigned tableLog); 6480c16b537SWarner Losh void FSEv05_freeDTable(FSEv05_DTable* dt); 6490c16b537SWarner Losh 6500c16b537SWarner Losh /*! 6510c16b537SWarner Losh FSEv05_buildDTable(): 6520c16b537SWarner Losh Builds 'dt', which must be already allocated, using FSEv05_createDTable() 6530c16b537SWarner Losh @return : 0, 6540c16b537SWarner Losh or an errorCode, which can be tested using FSEv05_isError() */ 6550c16b537SWarner Losh size_t FSEv05_buildDTable (FSEv05_DTable* dt, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog); 6560c16b537SWarner Losh 6570c16b537SWarner Losh /*! 6580c16b537SWarner Losh FSEv05_decompress_usingDTable(): 6590c16b537SWarner Losh Decompress compressed source @cSrc of size @cSrcSize using `dt` 6600c16b537SWarner Losh into `dst` which must be already allocated. 6610c16b537SWarner Losh @return : size of regenerated data (necessarily <= @dstCapacity) 6620c16b537SWarner Losh or an errorCode, which can be tested using FSEv05_isError() */ 6630c16b537SWarner Losh size_t FSEv05_decompress_usingDTable(void* dst, size_t dstCapacity, const void* cSrc, size_t cSrcSize, const FSEv05_DTable* dt); 6640c16b537SWarner Losh 6650c16b537SWarner Losh 6660c16b537SWarner Losh 6670c16b537SWarner Losh #if defined (__cplusplus) 6680c16b537SWarner Losh } 6690c16b537SWarner Losh #endif 6700c16b537SWarner Losh 6710c16b537SWarner Losh #endif /* FSEv05_H */ 6720c16b537SWarner Losh /* ****************************************************************** 6730c16b537SWarner Losh bitstream 6740c16b537SWarner Losh Part of FSEv05 library 6750c16b537SWarner Losh header file (to include) 6760c16b537SWarner Losh Copyright (C) 2013-2016, Yann Collet. 6770c16b537SWarner Losh 6780c16b537SWarner Losh BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php) 6790c16b537SWarner Losh 6800c16b537SWarner Losh Redistribution and use in source and binary forms, with or without 6810c16b537SWarner Losh modification, are permitted provided that the following conditions are 6820c16b537SWarner Losh met: 6830c16b537SWarner Losh 6840c16b537SWarner Losh * Redistributions of source code must retain the above copyright 6850c16b537SWarner Losh notice, this list of conditions and the following disclaimer. 6860c16b537SWarner Losh * Redistributions in binary form must reproduce the above 6870c16b537SWarner Losh copyright notice, this list of conditions and the following disclaimer 6880c16b537SWarner Losh in the documentation and/or other materials provided with the 6890c16b537SWarner Losh distribution. 6900c16b537SWarner Losh 6910c16b537SWarner Losh THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 6920c16b537SWarner Losh "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 6930c16b537SWarner Losh LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 6940c16b537SWarner Losh A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 6950c16b537SWarner Losh OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 6960c16b537SWarner Losh SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 6970c16b537SWarner Losh LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 6980c16b537SWarner Losh DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 6990c16b537SWarner Losh THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 7000c16b537SWarner Losh (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 7010c16b537SWarner Losh OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 7020c16b537SWarner Losh 7030c16b537SWarner Losh You can contact the author at : 7040c16b537SWarner Losh - Source repository : https://github.com/Cyan4973/FiniteStateEntropy 7050c16b537SWarner Losh ****************************************************************** */ 7060c16b537SWarner Losh #ifndef BITv05STREAM_H_MODULE 7070c16b537SWarner Losh #define BITv05STREAM_H_MODULE 7080c16b537SWarner Losh 7090c16b537SWarner Losh #if defined (__cplusplus) 7100c16b537SWarner Losh extern "C" { 7110c16b537SWarner Losh #endif 7120c16b537SWarner Losh 7130c16b537SWarner Losh 7140c16b537SWarner Losh /* 7150c16b537SWarner Losh * This API consists of small unitary functions, which highly benefit from being inlined. 7160c16b537SWarner Losh * Since link-time-optimization is not available for all compilers, 7170c16b537SWarner Losh * these functions are defined into a .h to be included. 7180c16b537SWarner Losh */ 7190c16b537SWarner Losh 7200c16b537SWarner Losh 7210c16b537SWarner Losh 7220c16b537SWarner Losh /*-******************************************** 7230c16b537SWarner Losh * bitStream decoding API (read backward) 7240c16b537SWarner Losh **********************************************/ 7250c16b537SWarner Losh typedef struct 7260c16b537SWarner Losh { 7270c16b537SWarner Losh size_t bitContainer; 7280c16b537SWarner Losh unsigned bitsConsumed; 7290c16b537SWarner Losh const char* ptr; 7300c16b537SWarner Losh const char* start; 7310c16b537SWarner Losh } BITv05_DStream_t; 7320c16b537SWarner Losh 7330c16b537SWarner Losh typedef enum { BITv05_DStream_unfinished = 0, 7340c16b537SWarner Losh BITv05_DStream_endOfBuffer = 1, 7350c16b537SWarner Losh BITv05_DStream_completed = 2, 7360c16b537SWarner Losh BITv05_DStream_overflow = 3 } BITv05_DStream_status; /* result of BITv05_reloadDStream() */ 7370c16b537SWarner Losh /* 1,2,4,8 would be better for bitmap combinations, but slows down performance a bit ... :( */ 7380c16b537SWarner Losh 7390c16b537SWarner Losh MEM_STATIC size_t BITv05_initDStream(BITv05_DStream_t* bitD, const void* srcBuffer, size_t srcSize); 7400c16b537SWarner Losh MEM_STATIC size_t BITv05_readBits(BITv05_DStream_t* bitD, unsigned nbBits); 7410c16b537SWarner Losh MEM_STATIC BITv05_DStream_status BITv05_reloadDStream(BITv05_DStream_t* bitD); 7420c16b537SWarner Losh MEM_STATIC unsigned BITv05_endOfDStream(const BITv05_DStream_t* bitD); 7430c16b537SWarner Losh 7440c16b537SWarner Losh 7450c16b537SWarner Losh /*-**************************************** 7460c16b537SWarner Losh * unsafe API 7470c16b537SWarner Losh ******************************************/ 7480c16b537SWarner Losh MEM_STATIC size_t BITv05_readBitsFast(BITv05_DStream_t* bitD, unsigned nbBits); 7490c16b537SWarner Losh /* faster, but works only if nbBits >= 1 */ 7500c16b537SWarner Losh 7510c16b537SWarner Losh 7520c16b537SWarner Losh 7530c16b537SWarner Losh /*-************************************************************** 7540c16b537SWarner Losh * Helper functions 7550c16b537SWarner Losh ****************************************************************/ 756052d3c12SConrad Meyer MEM_STATIC unsigned BITv05_highbit32 (U32 val) 7570c16b537SWarner Losh { 7580c16b537SWarner Losh # if defined(_MSC_VER) /* Visual */ 7590c16b537SWarner Losh unsigned long r=0; 7600c16b537SWarner Losh _BitScanReverse ( &r, val ); 7610c16b537SWarner Losh return (unsigned) r; 7620c16b537SWarner Losh # elif defined(__GNUC__) && (__GNUC__ >= 3) /* Use GCC Intrinsic */ 7639cbefe25SConrad Meyer return __builtin_clz (val) ^ 31; 7640c16b537SWarner Losh # else /* Software version */ 7650c16b537SWarner Losh static const unsigned DeBruijnClz[32] = { 0, 9, 1, 10, 13, 21, 2, 29, 11, 14, 16, 18, 22, 25, 3, 30, 8, 12, 20, 28, 15, 17, 24, 7, 19, 27, 23, 6, 26, 5, 4, 31 }; 7660c16b537SWarner Losh U32 v = val; 7670c16b537SWarner Losh unsigned r; 7680c16b537SWarner Losh v |= v >> 1; 7690c16b537SWarner Losh v |= v >> 2; 7700c16b537SWarner Losh v |= v >> 4; 7710c16b537SWarner Losh v |= v >> 8; 7720c16b537SWarner Losh v |= v >> 16; 7730c16b537SWarner Losh r = DeBruijnClz[ (U32) (v * 0x07C4ACDDU) >> 27]; 7740c16b537SWarner Losh return r; 7750c16b537SWarner Losh # endif 7760c16b537SWarner Losh } 7770c16b537SWarner Losh 7780c16b537SWarner Losh 7790c16b537SWarner Losh 7800c16b537SWarner Losh /*-******************************************************** 7810c16b537SWarner Losh * bitStream decoding 7820c16b537SWarner Losh **********************************************************/ 7830c16b537SWarner Losh /*!BITv05_initDStream 7840c16b537SWarner Losh * Initialize a BITv05_DStream_t. 7850c16b537SWarner Losh * @bitD : a pointer to an already allocated BITv05_DStream_t structure 7860c16b537SWarner Losh * @srcBuffer must point at the beginning of a bitStream 7870c16b537SWarner Losh * @srcSize must be the exact size of the bitStream 7880c16b537SWarner Losh * @result : size of stream (== srcSize) or an errorCode if a problem is detected 7890c16b537SWarner Losh */ 7900c16b537SWarner Losh MEM_STATIC size_t BITv05_initDStream(BITv05_DStream_t* bitD, const void* srcBuffer, size_t srcSize) 7910c16b537SWarner Losh { 7920c16b537SWarner Losh if (srcSize < 1) { memset(bitD, 0, sizeof(*bitD)); return ERROR(srcSize_wrong); } 7930c16b537SWarner Losh 7940c16b537SWarner Losh if (srcSize >= sizeof(size_t)) { /* normal case */ 7950c16b537SWarner Losh U32 contain32; 7960c16b537SWarner Losh bitD->start = (const char*)srcBuffer; 7970c16b537SWarner Losh bitD->ptr = (const char*)srcBuffer + srcSize - sizeof(size_t); 7980c16b537SWarner Losh bitD->bitContainer = MEM_readLEST(bitD->ptr); 7990c16b537SWarner Losh contain32 = ((const BYTE*)srcBuffer)[srcSize-1]; 8000c16b537SWarner Losh if (contain32 == 0) return ERROR(GENERIC); /* endMark not present */ 8010c16b537SWarner Losh bitD->bitsConsumed = 8 - BITv05_highbit32(contain32); 8020c16b537SWarner Losh } else { 8030c16b537SWarner Losh U32 contain32; 8040c16b537SWarner Losh bitD->start = (const char*)srcBuffer; 8050c16b537SWarner Losh bitD->ptr = bitD->start; 8060c16b537SWarner Losh bitD->bitContainer = *(const BYTE*)(bitD->start); 8070c16b537SWarner Losh switch(srcSize) 8080c16b537SWarner Losh { 8090c16b537SWarner Losh case 7: bitD->bitContainer += (size_t)(((const BYTE*)(bitD->start))[6]) << (sizeof(size_t)*8 - 16);/* fall-through */ 8100c16b537SWarner Losh case 6: bitD->bitContainer += (size_t)(((const BYTE*)(bitD->start))[5]) << (sizeof(size_t)*8 - 24);/* fall-through */ 8110c16b537SWarner Losh case 5: bitD->bitContainer += (size_t)(((const BYTE*)(bitD->start))[4]) << (sizeof(size_t)*8 - 32);/* fall-through */ 8120c16b537SWarner Losh case 4: bitD->bitContainer += (size_t)(((const BYTE*)(bitD->start))[3]) << 24; /* fall-through */ 8130c16b537SWarner Losh case 3: bitD->bitContainer += (size_t)(((const BYTE*)(bitD->start))[2]) << 16; /* fall-through */ 8140c16b537SWarner Losh case 2: bitD->bitContainer += (size_t)(((const BYTE*)(bitD->start))[1]) << 8; /* fall-through */ 8150c16b537SWarner Losh default: break; 8160c16b537SWarner Losh } 8170c16b537SWarner Losh contain32 = ((const BYTE*)srcBuffer)[srcSize-1]; 8180c16b537SWarner Losh if (contain32 == 0) return ERROR(GENERIC); /* endMark not present */ 8190c16b537SWarner Losh bitD->bitsConsumed = 8 - BITv05_highbit32(contain32); 8200c16b537SWarner Losh bitD->bitsConsumed += (U32)(sizeof(size_t) - srcSize)*8; 8210c16b537SWarner Losh } 8220c16b537SWarner Losh 8230c16b537SWarner Losh return srcSize; 8240c16b537SWarner Losh } 8250c16b537SWarner Losh 8260c16b537SWarner Losh MEM_STATIC size_t BITv05_lookBits(BITv05_DStream_t* bitD, U32 nbBits) 8270c16b537SWarner Losh { 8280c16b537SWarner Losh const U32 bitMask = sizeof(bitD->bitContainer)*8 - 1; 8290c16b537SWarner Losh return ((bitD->bitContainer << (bitD->bitsConsumed & bitMask)) >> 1) >> ((bitMask-nbBits) & bitMask); 8300c16b537SWarner Losh } 8310c16b537SWarner Losh 8320c16b537SWarner Losh /*! BITv05_lookBitsFast : 8330c16b537SWarner Losh * unsafe version; only works only if nbBits >= 1 */ 8340c16b537SWarner Losh MEM_STATIC size_t BITv05_lookBitsFast(BITv05_DStream_t* bitD, U32 nbBits) 8350c16b537SWarner Losh { 8360c16b537SWarner Losh const U32 bitMask = sizeof(bitD->bitContainer)*8 - 1; 8370c16b537SWarner Losh return (bitD->bitContainer << (bitD->bitsConsumed & bitMask)) >> (((bitMask+1)-nbBits) & bitMask); 8380c16b537SWarner Losh } 8390c16b537SWarner Losh 8400c16b537SWarner Losh MEM_STATIC void BITv05_skipBits(BITv05_DStream_t* bitD, U32 nbBits) 8410c16b537SWarner Losh { 8420c16b537SWarner Losh bitD->bitsConsumed += nbBits; 8430c16b537SWarner Losh } 8440c16b537SWarner Losh 845a0483764SConrad Meyer MEM_STATIC size_t BITv05_readBits(BITv05_DStream_t* bitD, unsigned nbBits) 8460c16b537SWarner Losh { 8470c16b537SWarner Losh size_t value = BITv05_lookBits(bitD, nbBits); 8480c16b537SWarner Losh BITv05_skipBits(bitD, nbBits); 8490c16b537SWarner Losh return value; 8500c16b537SWarner Losh } 8510c16b537SWarner Losh 8520c16b537SWarner Losh /*!BITv05_readBitsFast : 8530c16b537SWarner Losh * unsafe version; only works only if nbBits >= 1 */ 854a0483764SConrad Meyer MEM_STATIC size_t BITv05_readBitsFast(BITv05_DStream_t* bitD, unsigned nbBits) 8550c16b537SWarner Losh { 8560c16b537SWarner Losh size_t value = BITv05_lookBitsFast(bitD, nbBits); 8570c16b537SWarner Losh BITv05_skipBits(bitD, nbBits); 8580c16b537SWarner Losh return value; 8590c16b537SWarner Losh } 8600c16b537SWarner Losh 8610c16b537SWarner Losh MEM_STATIC BITv05_DStream_status BITv05_reloadDStream(BITv05_DStream_t* bitD) 8620c16b537SWarner Losh { 8630c16b537SWarner Losh if (bitD->bitsConsumed > (sizeof(bitD->bitContainer)*8)) /* should never happen */ 8640c16b537SWarner Losh return BITv05_DStream_overflow; 8650c16b537SWarner Losh 8660c16b537SWarner Losh if (bitD->ptr >= bitD->start + sizeof(bitD->bitContainer)) { 8670c16b537SWarner Losh bitD->ptr -= bitD->bitsConsumed >> 3; 8680c16b537SWarner Losh bitD->bitsConsumed &= 7; 8690c16b537SWarner Losh bitD->bitContainer = MEM_readLEST(bitD->ptr); 8700c16b537SWarner Losh return BITv05_DStream_unfinished; 8710c16b537SWarner Losh } 8720c16b537SWarner Losh if (bitD->ptr == bitD->start) { 8730c16b537SWarner Losh if (bitD->bitsConsumed < sizeof(bitD->bitContainer)*8) return BITv05_DStream_endOfBuffer; 8740c16b537SWarner Losh return BITv05_DStream_completed; 8750c16b537SWarner Losh } 8760c16b537SWarner Losh { 8770c16b537SWarner Losh U32 nbBytes = bitD->bitsConsumed >> 3; 8780c16b537SWarner Losh BITv05_DStream_status result = BITv05_DStream_unfinished; 8790c16b537SWarner Losh if (bitD->ptr - nbBytes < bitD->start) { 8800c16b537SWarner Losh nbBytes = (U32)(bitD->ptr - bitD->start); /* ptr > start */ 8810c16b537SWarner Losh result = BITv05_DStream_endOfBuffer; 8820c16b537SWarner Losh } 8830c16b537SWarner Losh bitD->ptr -= nbBytes; 8840c16b537SWarner Losh bitD->bitsConsumed -= nbBytes*8; 8850c16b537SWarner Losh bitD->bitContainer = MEM_readLEST(bitD->ptr); /* reminder : srcSize > sizeof(bitD) */ 8860c16b537SWarner Losh return result; 8870c16b537SWarner Losh } 8880c16b537SWarner Losh } 8890c16b537SWarner Losh 8900c16b537SWarner Losh /*! BITv05_endOfDStream 8910c16b537SWarner Losh * @return Tells if DStream has reached its exact end 8920c16b537SWarner Losh */ 8930c16b537SWarner Losh MEM_STATIC unsigned BITv05_endOfDStream(const BITv05_DStream_t* DStream) 8940c16b537SWarner Losh { 8950c16b537SWarner Losh return ((DStream->ptr == DStream->start) && (DStream->bitsConsumed == sizeof(DStream->bitContainer)*8)); 8960c16b537SWarner Losh } 8970c16b537SWarner Losh 8980c16b537SWarner Losh #if defined (__cplusplus) 8990c16b537SWarner Losh } 9000c16b537SWarner Losh #endif 9010c16b537SWarner Losh 9020c16b537SWarner Losh #endif /* BITv05STREAM_H_MODULE */ 9030c16b537SWarner Losh /* ****************************************************************** 9040c16b537SWarner Losh FSEv05 : Finite State Entropy coder 9050c16b537SWarner Losh header file for static linking (only) 9060c16b537SWarner Losh Copyright (C) 2013-2015, Yann Collet 9070c16b537SWarner Losh 9080c16b537SWarner Losh BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php) 9090c16b537SWarner Losh 9100c16b537SWarner Losh Redistribution and use in source and binary forms, with or without 9110c16b537SWarner Losh modification, are permitted provided that the following conditions are 9120c16b537SWarner Losh met: 9130c16b537SWarner Losh 9140c16b537SWarner Losh * Redistributions of source code must retain the above copyright 9150c16b537SWarner Losh notice, this list of conditions and the following disclaimer. 9160c16b537SWarner Losh * Redistributions in binary form must reproduce the above 9170c16b537SWarner Losh copyright notice, this list of conditions and the following disclaimer 9180c16b537SWarner Losh in the documentation and/or other materials provided with the 9190c16b537SWarner Losh distribution. 9200c16b537SWarner Losh 9210c16b537SWarner Losh THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 9220c16b537SWarner Losh "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 9230c16b537SWarner Losh LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 9240c16b537SWarner Losh A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 9250c16b537SWarner Losh OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 9260c16b537SWarner Losh SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 9270c16b537SWarner Losh LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 9280c16b537SWarner Losh DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 9290c16b537SWarner Losh THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 9300c16b537SWarner Losh (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 9310c16b537SWarner Losh OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 9320c16b537SWarner Losh 9330c16b537SWarner Losh You can contact the author at : 9340c16b537SWarner Losh - Source repository : https://github.com/Cyan4973/FiniteStateEntropy 9350c16b537SWarner Losh - Public forum : https://groups.google.com/forum/#!forum/lz4c 9360c16b537SWarner Losh ****************************************************************** */ 9370c16b537SWarner Losh #ifndef FSEv05_STATIC_H 9380c16b537SWarner Losh #define FSEv05_STATIC_H 9390c16b537SWarner Losh 9400c16b537SWarner Losh #if defined (__cplusplus) 9410c16b537SWarner Losh extern "C" { 9420c16b537SWarner Losh #endif 9430c16b537SWarner Losh 9440c16b537SWarner Losh 9450c16b537SWarner Losh 9460c16b537SWarner Losh /* ***************************************** 9470c16b537SWarner Losh * Static allocation 9480c16b537SWarner Losh *******************************************/ 9490c16b537SWarner Losh /* It is possible to statically allocate FSEv05 CTable/DTable as a table of unsigned using below macros */ 9500c16b537SWarner Losh #define FSEv05_DTABLE_SIZE_U32(maxTableLog) (1 + (1<<maxTableLog)) 9510c16b537SWarner Losh 9520c16b537SWarner Losh 9530c16b537SWarner Losh /* ***************************************** 9540c16b537SWarner Losh * FSEv05 advanced API 9550c16b537SWarner Losh *******************************************/ 9560c16b537SWarner Losh size_t FSEv05_buildDTable_raw (FSEv05_DTable* dt, unsigned nbBits); 9570c16b537SWarner Losh /* build a fake FSEv05_DTable, designed to read an uncompressed bitstream where each symbol uses nbBits */ 9580c16b537SWarner Losh 9590c16b537SWarner Losh size_t FSEv05_buildDTable_rle (FSEv05_DTable* dt, unsigned char symbolValue); 9600c16b537SWarner Losh /* build a fake FSEv05_DTable, designed to always generate the same symbolValue */ 9610c16b537SWarner Losh 9620c16b537SWarner Losh 9630c16b537SWarner Losh 9640c16b537SWarner Losh /* ***************************************** 9650c16b537SWarner Losh * FSEv05 symbol decompression API 9660c16b537SWarner Losh *******************************************/ 9670c16b537SWarner Losh typedef struct 9680c16b537SWarner Losh { 9690c16b537SWarner Losh size_t state; 9700c16b537SWarner Losh const void* table; /* precise table may vary, depending on U16 */ 9710c16b537SWarner Losh } FSEv05_DState_t; 9720c16b537SWarner Losh 9730c16b537SWarner Losh 9740c16b537SWarner Losh static void FSEv05_initDState(FSEv05_DState_t* DStatePtr, BITv05_DStream_t* bitD, const FSEv05_DTable* dt); 9750c16b537SWarner Losh 9760c16b537SWarner Losh static unsigned char FSEv05_decodeSymbol(FSEv05_DState_t* DStatePtr, BITv05_DStream_t* bitD); 9770c16b537SWarner Losh 9780c16b537SWarner Losh static unsigned FSEv05_endOfDState(const FSEv05_DState_t* DStatePtr); 9790c16b537SWarner Losh 9800c16b537SWarner Losh 9810c16b537SWarner Losh 9820c16b537SWarner Losh /* ***************************************** 9830c16b537SWarner Losh * FSEv05 unsafe API 9840c16b537SWarner Losh *******************************************/ 9850c16b537SWarner Losh static unsigned char FSEv05_decodeSymbolFast(FSEv05_DState_t* DStatePtr, BITv05_DStream_t* bitD); 9860c16b537SWarner Losh /* faster, but works only if nbBits is always >= 1 (otherwise, result will be corrupted) */ 9870c16b537SWarner Losh 9880c16b537SWarner Losh 9890c16b537SWarner Losh /* ***************************************** 9900c16b537SWarner Losh * Implementation of inlined functions 9910c16b537SWarner Losh *******************************************/ 9920c16b537SWarner Losh /* decompression */ 9930c16b537SWarner Losh 9940c16b537SWarner Losh typedef struct { 9950c16b537SWarner Losh U16 tableLog; 9960c16b537SWarner Losh U16 fastMode; 9970c16b537SWarner Losh } FSEv05_DTableHeader; /* sizeof U32 */ 9980c16b537SWarner Losh 9990c16b537SWarner Losh typedef struct 10000c16b537SWarner Losh { 10010c16b537SWarner Losh unsigned short newState; 10020c16b537SWarner Losh unsigned char symbol; 10030c16b537SWarner Losh unsigned char nbBits; 10040c16b537SWarner Losh } FSEv05_decode_t; /* size == U32 */ 10050c16b537SWarner Losh 10060c16b537SWarner Losh MEM_STATIC void FSEv05_initDState(FSEv05_DState_t* DStatePtr, BITv05_DStream_t* bitD, const FSEv05_DTable* dt) 10070c16b537SWarner Losh { 10080c16b537SWarner Losh const void* ptr = dt; 10090c16b537SWarner Losh const FSEv05_DTableHeader* const DTableH = (const FSEv05_DTableHeader*)ptr; 10100c16b537SWarner Losh DStatePtr->state = BITv05_readBits(bitD, DTableH->tableLog); 10110c16b537SWarner Losh BITv05_reloadDStream(bitD); 10120c16b537SWarner Losh DStatePtr->table = dt + 1; 10130c16b537SWarner Losh } 10140c16b537SWarner Losh 10150c16b537SWarner Losh MEM_STATIC BYTE FSEv05_peakSymbol(FSEv05_DState_t* DStatePtr) 10160c16b537SWarner Losh { 10170c16b537SWarner Losh const FSEv05_decode_t DInfo = ((const FSEv05_decode_t*)(DStatePtr->table))[DStatePtr->state]; 10180c16b537SWarner Losh return DInfo.symbol; 10190c16b537SWarner Losh } 10200c16b537SWarner Losh 10210c16b537SWarner Losh MEM_STATIC BYTE FSEv05_decodeSymbol(FSEv05_DState_t* DStatePtr, BITv05_DStream_t* bitD) 10220c16b537SWarner Losh { 10230c16b537SWarner Losh const FSEv05_decode_t DInfo = ((const FSEv05_decode_t*)(DStatePtr->table))[DStatePtr->state]; 10240c16b537SWarner Losh const U32 nbBits = DInfo.nbBits; 10250c16b537SWarner Losh BYTE symbol = DInfo.symbol; 10260c16b537SWarner Losh size_t lowBits = BITv05_readBits(bitD, nbBits); 10270c16b537SWarner Losh 10280c16b537SWarner Losh DStatePtr->state = DInfo.newState + lowBits; 10290c16b537SWarner Losh return symbol; 10300c16b537SWarner Losh } 10310c16b537SWarner Losh 10320c16b537SWarner Losh MEM_STATIC BYTE FSEv05_decodeSymbolFast(FSEv05_DState_t* DStatePtr, BITv05_DStream_t* bitD) 10330c16b537SWarner Losh { 10340c16b537SWarner Losh const FSEv05_decode_t DInfo = ((const FSEv05_decode_t*)(DStatePtr->table))[DStatePtr->state]; 10350c16b537SWarner Losh const U32 nbBits = DInfo.nbBits; 10360c16b537SWarner Losh BYTE symbol = DInfo.symbol; 10370c16b537SWarner Losh size_t lowBits = BITv05_readBitsFast(bitD, nbBits); 10380c16b537SWarner Losh 10390c16b537SWarner Losh DStatePtr->state = DInfo.newState + lowBits; 10400c16b537SWarner Losh return symbol; 10410c16b537SWarner Losh } 10420c16b537SWarner Losh 10430c16b537SWarner Losh MEM_STATIC unsigned FSEv05_endOfDState(const FSEv05_DState_t* DStatePtr) 10440c16b537SWarner Losh { 10450c16b537SWarner Losh return DStatePtr->state == 0; 10460c16b537SWarner Losh } 10470c16b537SWarner Losh 10480c16b537SWarner Losh 10490c16b537SWarner Losh #if defined (__cplusplus) 10500c16b537SWarner Losh } 10510c16b537SWarner Losh #endif 10520c16b537SWarner Losh 10530c16b537SWarner Losh #endif /* FSEv05_STATIC_H */ 10540c16b537SWarner Losh /* ****************************************************************** 10550c16b537SWarner Losh FSEv05 : Finite State Entropy coder 10560c16b537SWarner Losh Copyright (C) 2013-2015, Yann Collet. 10570c16b537SWarner Losh 10580c16b537SWarner Losh BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php) 10590c16b537SWarner Losh 10600c16b537SWarner Losh Redistribution and use in source and binary forms, with or without 10610c16b537SWarner Losh modification, are permitted provided that the following conditions are 10620c16b537SWarner Losh met: 10630c16b537SWarner Losh 10640c16b537SWarner Losh * Redistributions of source code must retain the above copyright 10650c16b537SWarner Losh notice, this list of conditions and the following disclaimer. 10660c16b537SWarner Losh * Redistributions in binary form must reproduce the above 10670c16b537SWarner Losh copyright notice, this list of conditions and the following disclaimer 10680c16b537SWarner Losh in the documentation and/or other materials provided with the 10690c16b537SWarner Losh distribution. 10700c16b537SWarner Losh 10710c16b537SWarner Losh THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 10720c16b537SWarner Losh "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 10730c16b537SWarner Losh LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 10740c16b537SWarner Losh A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 10750c16b537SWarner Losh OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 10760c16b537SWarner Losh SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 10770c16b537SWarner Losh LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 10780c16b537SWarner Losh DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 10790c16b537SWarner Losh THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 10800c16b537SWarner Losh (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 10810c16b537SWarner Losh OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 10820c16b537SWarner Losh 10830c16b537SWarner Losh You can contact the author at : 10840c16b537SWarner Losh - FSEv05 source repository : https://github.com/Cyan4973/FiniteStateEntropy 10850c16b537SWarner Losh - Public forum : https://groups.google.com/forum/#!forum/lz4c 10860c16b537SWarner Losh ****************************************************************** */ 10870c16b537SWarner Losh 10880c16b537SWarner Losh #ifndef FSEv05_COMMONDEFS_ONLY 10890c16b537SWarner Losh 10900c16b537SWarner Losh /* ************************************************************** 10910c16b537SWarner Losh * Tuning parameters 10920c16b537SWarner Losh ****************************************************************/ 10930c16b537SWarner Losh /*!MEMORY_USAGE : 10940c16b537SWarner Losh * Memory usage formula : N->2^N Bytes (examples : 10 -> 1KB; 12 -> 4KB ; 16 -> 64KB; 20 -> 1MB; etc.) 10950c16b537SWarner Losh * Increasing memory usage improves compression ratio 10960c16b537SWarner Losh * Reduced memory usage can improve speed, due to cache effect 10970c16b537SWarner Losh * Recommended max value is 14, for 16KB, which nicely fits into Intel x86 L1 cache */ 10980c16b537SWarner Losh #define FSEv05_MAX_MEMORY_USAGE 14 10990c16b537SWarner Losh #define FSEv05_DEFAULT_MEMORY_USAGE 13 11000c16b537SWarner Losh 11010c16b537SWarner Losh /*!FSEv05_MAX_SYMBOL_VALUE : 11020c16b537SWarner Losh * Maximum symbol value authorized. 11030c16b537SWarner Losh * Required for proper stack allocation */ 11040c16b537SWarner Losh #define FSEv05_MAX_SYMBOL_VALUE 255 11050c16b537SWarner Losh 11060c16b537SWarner Losh 11070c16b537SWarner Losh /* ************************************************************** 11080c16b537SWarner Losh * template functions type & suffix 11090c16b537SWarner Losh ****************************************************************/ 11100c16b537SWarner Losh #define FSEv05_FUNCTION_TYPE BYTE 11110c16b537SWarner Losh #define FSEv05_FUNCTION_EXTENSION 11120c16b537SWarner Losh #define FSEv05_DECODE_TYPE FSEv05_decode_t 11130c16b537SWarner Losh 11140c16b537SWarner Losh 11150c16b537SWarner Losh #endif /* !FSEv05_COMMONDEFS_ONLY */ 11160c16b537SWarner Losh 11170c16b537SWarner Losh /* ************************************************************** 11180c16b537SWarner Losh * Compiler specifics 11190c16b537SWarner Losh ****************************************************************/ 11200c16b537SWarner Losh #ifdef _MSC_VER /* Visual Studio */ 11210c16b537SWarner Losh # define FORCE_INLINE static __forceinline 11220c16b537SWarner Losh # include <intrin.h> /* For Visual 2005 */ 11230c16b537SWarner Losh # pragma warning(disable : 4127) /* disable: C4127: conditional expression is constant */ 11240c16b537SWarner Losh # pragma warning(disable : 4214) /* disable: C4214: non-int bitfields */ 11250c16b537SWarner Losh #else 11260c16b537SWarner Losh # if defined (__cplusplus) || defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L /* C99 */ 11270c16b537SWarner Losh # ifdef __GNUC__ 11280c16b537SWarner Losh # define FORCE_INLINE static inline __attribute__((always_inline)) 11290c16b537SWarner Losh # else 11300c16b537SWarner Losh # define FORCE_INLINE static inline 11310c16b537SWarner Losh # endif 11320c16b537SWarner Losh # else 11330c16b537SWarner Losh # define FORCE_INLINE static 11340c16b537SWarner Losh # endif /* __STDC_VERSION__ */ 11350c16b537SWarner Losh #endif 11360c16b537SWarner Losh 11370c16b537SWarner Losh 11380c16b537SWarner Losh /* ************************************************************** 11390c16b537SWarner Losh * Includes 11400c16b537SWarner Losh ****************************************************************/ 11410c16b537SWarner Losh #include <stdlib.h> /* malloc, free, qsort */ 11420c16b537SWarner Losh #include <string.h> /* memcpy, memset */ 11430c16b537SWarner Losh #include <stdio.h> /* printf (debug) */ 11440c16b537SWarner Losh 11450c16b537SWarner Losh 11460c16b537SWarner Losh 11470c16b537SWarner Losh /* *************************************************************** 11480c16b537SWarner Losh * Constants 11490c16b537SWarner Losh *****************************************************************/ 11500c16b537SWarner Losh #define FSEv05_MAX_TABLELOG (FSEv05_MAX_MEMORY_USAGE-2) 11510c16b537SWarner Losh #define FSEv05_MAX_TABLESIZE (1U<<FSEv05_MAX_TABLELOG) 11520c16b537SWarner Losh #define FSEv05_MAXTABLESIZE_MASK (FSEv05_MAX_TABLESIZE-1) 11530c16b537SWarner Losh #define FSEv05_DEFAULT_TABLELOG (FSEv05_DEFAULT_MEMORY_USAGE-2) 11540c16b537SWarner Losh #define FSEv05_MIN_TABLELOG 5 11550c16b537SWarner Losh 11560c16b537SWarner Losh #define FSEv05_TABLELOG_ABSOLUTE_MAX 15 11570c16b537SWarner Losh #if FSEv05_MAX_TABLELOG > FSEv05_TABLELOG_ABSOLUTE_MAX 11580c16b537SWarner Losh #error "FSEv05_MAX_TABLELOG > FSEv05_TABLELOG_ABSOLUTE_MAX is not supported" 11590c16b537SWarner Losh #endif 11600c16b537SWarner Losh 11610c16b537SWarner Losh 11620c16b537SWarner Losh /* ************************************************************** 11630c16b537SWarner Losh * Error Management 11640c16b537SWarner Losh ****************************************************************/ 11650c16b537SWarner Losh #define FSEv05_STATIC_ASSERT(c) { enum { FSEv05_static_assert = 1/(int)(!!(c)) }; } /* use only *after* variable declarations */ 11660c16b537SWarner Losh 11670c16b537SWarner Losh 11680c16b537SWarner Losh /* ************************************************************** 11690c16b537SWarner Losh * Complex types 11700c16b537SWarner Losh ****************************************************************/ 1171a0483764SConrad Meyer typedef unsigned DTable_max_t[FSEv05_DTABLE_SIZE_U32(FSEv05_MAX_TABLELOG)]; 11720c16b537SWarner Losh 11730c16b537SWarner Losh 11740c16b537SWarner Losh /* ************************************************************** 11750c16b537SWarner Losh * Templates 11760c16b537SWarner Losh ****************************************************************/ 11770c16b537SWarner Losh /* 11780c16b537SWarner Losh designed to be included 11790c16b537SWarner Losh for type-specific functions (template emulation in C) 11800c16b537SWarner Losh Objective is to write these functions only once, for improved maintenance 11810c16b537SWarner Losh */ 11820c16b537SWarner Losh 11830c16b537SWarner Losh /* safety checks */ 11840c16b537SWarner Losh #ifndef FSEv05_FUNCTION_EXTENSION 11850c16b537SWarner Losh # error "FSEv05_FUNCTION_EXTENSION must be defined" 11860c16b537SWarner Losh #endif 11870c16b537SWarner Losh #ifndef FSEv05_FUNCTION_TYPE 11880c16b537SWarner Losh # error "FSEv05_FUNCTION_TYPE must be defined" 11890c16b537SWarner Losh #endif 11900c16b537SWarner Losh 11910c16b537SWarner Losh /* Function names */ 11920c16b537SWarner Losh #define FSEv05_CAT(X,Y) X##Y 11930c16b537SWarner Losh #define FSEv05_FUNCTION_NAME(X,Y) FSEv05_CAT(X,Y) 11940c16b537SWarner Losh #define FSEv05_TYPE_NAME(X,Y) FSEv05_CAT(X,Y) 11950c16b537SWarner Losh 11960c16b537SWarner Losh 11970c16b537SWarner Losh /* Function templates */ 11980c16b537SWarner Losh static U32 FSEv05_tableStep(U32 tableSize) { return (tableSize>>1) + (tableSize>>3) + 3; } 11990c16b537SWarner Losh 12000c16b537SWarner Losh 12010c16b537SWarner Losh 12020c16b537SWarner Losh FSEv05_DTable* FSEv05_createDTable (unsigned tableLog) 12030c16b537SWarner Losh { 12040c16b537SWarner Losh if (tableLog > FSEv05_TABLELOG_ABSOLUTE_MAX) tableLog = FSEv05_TABLELOG_ABSOLUTE_MAX; 12050c16b537SWarner Losh return (FSEv05_DTable*)malloc( FSEv05_DTABLE_SIZE_U32(tableLog) * sizeof (U32) ); 12060c16b537SWarner Losh } 12070c16b537SWarner Losh 12080c16b537SWarner Losh void FSEv05_freeDTable (FSEv05_DTable* dt) 12090c16b537SWarner Losh { 12100c16b537SWarner Losh free(dt); 12110c16b537SWarner Losh } 12120c16b537SWarner Losh 12130c16b537SWarner Losh size_t FSEv05_buildDTable(FSEv05_DTable* dt, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog) 12140c16b537SWarner Losh { 12150c16b537SWarner Losh FSEv05_DTableHeader DTableH; 12160c16b537SWarner Losh void* const tdPtr = dt+1; /* because dt is unsigned, 32-bits aligned on 32-bits */ 12170c16b537SWarner Losh FSEv05_DECODE_TYPE* const tableDecode = (FSEv05_DECODE_TYPE*) (tdPtr); 12180c16b537SWarner Losh const U32 tableSize = 1 << tableLog; 12190c16b537SWarner Losh const U32 tableMask = tableSize-1; 12200c16b537SWarner Losh const U32 step = FSEv05_tableStep(tableSize); 12210c16b537SWarner Losh U16 symbolNext[FSEv05_MAX_SYMBOL_VALUE+1]; 12220c16b537SWarner Losh U32 position = 0; 12230c16b537SWarner Losh U32 highThreshold = tableSize-1; 12240c16b537SWarner Losh const S16 largeLimit= (S16)(1 << (tableLog-1)); 12250c16b537SWarner Losh U32 noLarge = 1; 12260c16b537SWarner Losh U32 s; 12270c16b537SWarner Losh 12280c16b537SWarner Losh /* Sanity Checks */ 12290c16b537SWarner Losh if (maxSymbolValue > FSEv05_MAX_SYMBOL_VALUE) return ERROR(maxSymbolValue_tooLarge); 12300c16b537SWarner Losh if (tableLog > FSEv05_MAX_TABLELOG) return ERROR(tableLog_tooLarge); 12310c16b537SWarner Losh 12320c16b537SWarner Losh /* Init, lay down lowprob symbols */ 12330f743729SConrad Meyer memset(tableDecode, 0, sizeof(FSEv05_FUNCTION_TYPE) * (maxSymbolValue+1) ); /* useless init, but keep static analyzer happy, and we don't need to performance optimize legacy decoders */ 12340c16b537SWarner Losh DTableH.tableLog = (U16)tableLog; 12350c16b537SWarner Losh for (s=0; s<=maxSymbolValue; s++) { 12360c16b537SWarner Losh if (normalizedCounter[s]==-1) { 12370c16b537SWarner Losh tableDecode[highThreshold--].symbol = (FSEv05_FUNCTION_TYPE)s; 12380c16b537SWarner Losh symbolNext[s] = 1; 12390c16b537SWarner Losh } else { 12400c16b537SWarner Losh if (normalizedCounter[s] >= largeLimit) noLarge=0; 12410c16b537SWarner Losh symbolNext[s] = normalizedCounter[s]; 12420c16b537SWarner Losh } } 12430c16b537SWarner Losh 12440c16b537SWarner Losh /* Spread symbols */ 12450c16b537SWarner Losh for (s=0; s<=maxSymbolValue; s++) { 12460c16b537SWarner Losh int i; 12470c16b537SWarner Losh for (i=0; i<normalizedCounter[s]; i++) { 12480c16b537SWarner Losh tableDecode[position].symbol = (FSEv05_FUNCTION_TYPE)s; 12490c16b537SWarner Losh position = (position + step) & tableMask; 12500c16b537SWarner Losh while (position > highThreshold) position = (position + step) & tableMask; /* lowprob area */ 12510c16b537SWarner Losh } } 12520c16b537SWarner Losh 12530c16b537SWarner Losh if (position!=0) return ERROR(GENERIC); /* position must reach all cells once, otherwise normalizedCounter is incorrect */ 12540c16b537SWarner Losh 12550c16b537SWarner Losh /* Build Decoding table */ 12560c16b537SWarner Losh { 12570c16b537SWarner Losh U32 i; 12580c16b537SWarner Losh for (i=0; i<tableSize; i++) { 12590c16b537SWarner Losh FSEv05_FUNCTION_TYPE symbol = (FSEv05_FUNCTION_TYPE)(tableDecode[i].symbol); 12600c16b537SWarner Losh U16 nextState = symbolNext[symbol]++; 12610c16b537SWarner Losh tableDecode[i].nbBits = (BYTE) (tableLog - BITv05_highbit32 ((U32)nextState) ); 12620c16b537SWarner Losh tableDecode[i].newState = (U16) ( (nextState << tableDecode[i].nbBits) - tableSize); 12630c16b537SWarner Losh } } 12640c16b537SWarner Losh 12650c16b537SWarner Losh DTableH.fastMode = (U16)noLarge; 12660c16b537SWarner Losh memcpy(dt, &DTableH, sizeof(DTableH)); 12670c16b537SWarner Losh return 0; 12680c16b537SWarner Losh } 12690c16b537SWarner Losh 12700c16b537SWarner Losh 12710c16b537SWarner Losh #ifndef FSEv05_COMMONDEFS_ONLY 12720c16b537SWarner Losh /*-**************************************** 12730c16b537SWarner Losh * FSEv05 helper functions 12740c16b537SWarner Losh ******************************************/ 12750c16b537SWarner Losh unsigned FSEv05_isError(size_t code) { return ERR_isError(code); } 12760c16b537SWarner Losh 12770c16b537SWarner Losh const char* FSEv05_getErrorName(size_t code) { return ERR_getErrorName(code); } 12780c16b537SWarner Losh 12790c16b537SWarner Losh 12800c16b537SWarner Losh /*-************************************************************** 12810c16b537SWarner Losh * FSEv05 NCount encoding-decoding 12820c16b537SWarner Losh ****************************************************************/ 12830c16b537SWarner Losh static short FSEv05_abs(short a) { return a<0 ? -a : a; } 12840c16b537SWarner Losh 12850c16b537SWarner Losh 12860c16b537SWarner Losh size_t FSEv05_readNCount (short* normalizedCounter, unsigned* maxSVPtr, unsigned* tableLogPtr, 12870c16b537SWarner Losh const void* headerBuffer, size_t hbSize) 12880c16b537SWarner Losh { 12890c16b537SWarner Losh const BYTE* const istart = (const BYTE*) headerBuffer; 12900c16b537SWarner Losh const BYTE* const iend = istart + hbSize; 12910c16b537SWarner Losh const BYTE* ip = istart; 12920c16b537SWarner Losh int nbBits; 12930c16b537SWarner Losh int remaining; 12940c16b537SWarner Losh int threshold; 12950c16b537SWarner Losh U32 bitStream; 12960c16b537SWarner Losh int bitCount; 12970c16b537SWarner Losh unsigned charnum = 0; 12980c16b537SWarner Losh int previous0 = 0; 12990c16b537SWarner Losh 13000c16b537SWarner Losh if (hbSize < 4) return ERROR(srcSize_wrong); 13010c16b537SWarner Losh bitStream = MEM_readLE32(ip); 13020c16b537SWarner Losh nbBits = (bitStream & 0xF) + FSEv05_MIN_TABLELOG; /* extract tableLog */ 13030c16b537SWarner Losh if (nbBits > FSEv05_TABLELOG_ABSOLUTE_MAX) return ERROR(tableLog_tooLarge); 13040c16b537SWarner Losh bitStream >>= 4; 13050c16b537SWarner Losh bitCount = 4; 13060c16b537SWarner Losh *tableLogPtr = nbBits; 13070c16b537SWarner Losh remaining = (1<<nbBits)+1; 13080c16b537SWarner Losh threshold = 1<<nbBits; 13090c16b537SWarner Losh nbBits++; 13100c16b537SWarner Losh 13110c16b537SWarner Losh while ((remaining>1) && (charnum<=*maxSVPtr)) { 13120c16b537SWarner Losh if (previous0) { 13130c16b537SWarner Losh unsigned n0 = charnum; 13140c16b537SWarner Losh while ((bitStream & 0xFFFF) == 0xFFFF) { 13150c16b537SWarner Losh n0+=24; 13160c16b537SWarner Losh if (ip < iend-5) { 13170c16b537SWarner Losh ip+=2; 13180c16b537SWarner Losh bitStream = MEM_readLE32(ip) >> bitCount; 13190c16b537SWarner Losh } else { 13200c16b537SWarner Losh bitStream >>= 16; 13210c16b537SWarner Losh bitCount+=16; 13220c16b537SWarner Losh } } 13230c16b537SWarner Losh while ((bitStream & 3) == 3) { 13240c16b537SWarner Losh n0+=3; 13250c16b537SWarner Losh bitStream>>=2; 13260c16b537SWarner Losh bitCount+=2; 13270c16b537SWarner Losh } 13280c16b537SWarner Losh n0 += bitStream & 3; 13290c16b537SWarner Losh bitCount += 2; 13300c16b537SWarner Losh if (n0 > *maxSVPtr) return ERROR(maxSymbolValue_tooSmall); 13310c16b537SWarner Losh while (charnum < n0) normalizedCounter[charnum++] = 0; 13320c16b537SWarner Losh if ((ip <= iend-7) || (ip + (bitCount>>3) <= iend-4)) { 13330c16b537SWarner Losh ip += bitCount>>3; 13340c16b537SWarner Losh bitCount &= 7; 13350c16b537SWarner Losh bitStream = MEM_readLE32(ip) >> bitCount; 13360c16b537SWarner Losh } 13370c16b537SWarner Losh else 13380c16b537SWarner Losh bitStream >>= 2; 13390c16b537SWarner Losh } 13400c16b537SWarner Losh { 13410c16b537SWarner Losh const short max = (short)((2*threshold-1)-remaining); 13420c16b537SWarner Losh short count; 13430c16b537SWarner Losh 13440c16b537SWarner Losh if ((bitStream & (threshold-1)) < (U32)max) { 13450c16b537SWarner Losh count = (short)(bitStream & (threshold-1)); 13460c16b537SWarner Losh bitCount += nbBits-1; 13470c16b537SWarner Losh } else { 13480c16b537SWarner Losh count = (short)(bitStream & (2*threshold-1)); 13490c16b537SWarner Losh if (count >= threshold) count -= max; 13500c16b537SWarner Losh bitCount += nbBits; 13510c16b537SWarner Losh } 13520c16b537SWarner Losh 13530c16b537SWarner Losh count--; /* extra accuracy */ 13540c16b537SWarner Losh remaining -= FSEv05_abs(count); 13550c16b537SWarner Losh normalizedCounter[charnum++] = count; 13560c16b537SWarner Losh previous0 = !count; 13570c16b537SWarner Losh while (remaining < threshold) { 13580c16b537SWarner Losh nbBits--; 13590c16b537SWarner Losh threshold >>= 1; 13600c16b537SWarner Losh } 13610c16b537SWarner Losh 13620c16b537SWarner Losh if ((ip <= iend-7) || (ip + (bitCount>>3) <= iend-4)) { 13630c16b537SWarner Losh ip += bitCount>>3; 13640c16b537SWarner Losh bitCount &= 7; 13650c16b537SWarner Losh } else { 13660c16b537SWarner Losh bitCount -= (int)(8 * (iend - 4 - ip)); 13670c16b537SWarner Losh ip = iend - 4; 13680c16b537SWarner Losh } 13690c16b537SWarner Losh bitStream = MEM_readLE32(ip) >> (bitCount & 31); 13700c16b537SWarner Losh } } 13710c16b537SWarner Losh if (remaining != 1) return ERROR(GENERIC); 13720c16b537SWarner Losh *maxSVPtr = charnum-1; 13730c16b537SWarner Losh 13740c16b537SWarner Losh ip += (bitCount+7)>>3; 13750c16b537SWarner Losh if ((size_t)(ip-istart) > hbSize) return ERROR(srcSize_wrong); 13760c16b537SWarner Losh return ip-istart; 13770c16b537SWarner Losh } 13780c16b537SWarner Losh 13790c16b537SWarner Losh 13800c16b537SWarner Losh 13810c16b537SWarner Losh /*-******************************************************* 13820c16b537SWarner Losh * Decompression (Byte symbols) 13830c16b537SWarner Losh *********************************************************/ 13840c16b537SWarner Losh size_t FSEv05_buildDTable_rle (FSEv05_DTable* dt, BYTE symbolValue) 13850c16b537SWarner Losh { 13860c16b537SWarner Losh void* ptr = dt; 13870c16b537SWarner Losh FSEv05_DTableHeader* const DTableH = (FSEv05_DTableHeader*)ptr; 13880c16b537SWarner Losh void* dPtr = dt + 1; 13890c16b537SWarner Losh FSEv05_decode_t* const cell = (FSEv05_decode_t*)dPtr; 13900c16b537SWarner Losh 13910c16b537SWarner Losh DTableH->tableLog = 0; 13920c16b537SWarner Losh DTableH->fastMode = 0; 13930c16b537SWarner Losh 13940c16b537SWarner Losh cell->newState = 0; 13950c16b537SWarner Losh cell->symbol = symbolValue; 13960c16b537SWarner Losh cell->nbBits = 0; 13970c16b537SWarner Losh 13980c16b537SWarner Losh return 0; 13990c16b537SWarner Losh } 14000c16b537SWarner Losh 14010c16b537SWarner Losh 14020c16b537SWarner Losh size_t FSEv05_buildDTable_raw (FSEv05_DTable* dt, unsigned nbBits) 14030c16b537SWarner Losh { 14040c16b537SWarner Losh void* ptr = dt; 14050c16b537SWarner Losh FSEv05_DTableHeader* const DTableH = (FSEv05_DTableHeader*)ptr; 14060c16b537SWarner Losh void* dPtr = dt + 1; 14070c16b537SWarner Losh FSEv05_decode_t* const dinfo = (FSEv05_decode_t*)dPtr; 14080c16b537SWarner Losh const unsigned tableSize = 1 << nbBits; 14090c16b537SWarner Losh const unsigned tableMask = tableSize - 1; 14100c16b537SWarner Losh const unsigned maxSymbolValue = tableMask; 14110c16b537SWarner Losh unsigned s; 14120c16b537SWarner Losh 14130c16b537SWarner Losh /* Sanity checks */ 14140c16b537SWarner Losh if (nbBits < 1) return ERROR(GENERIC); /* min size */ 14150c16b537SWarner Losh 14160c16b537SWarner Losh /* Build Decoding Table */ 14170c16b537SWarner Losh DTableH->tableLog = (U16)nbBits; 14180c16b537SWarner Losh DTableH->fastMode = 1; 14190c16b537SWarner Losh for (s=0; s<=maxSymbolValue; s++) { 14200c16b537SWarner Losh dinfo[s].newState = 0; 14210c16b537SWarner Losh dinfo[s].symbol = (BYTE)s; 14220c16b537SWarner Losh dinfo[s].nbBits = (BYTE)nbBits; 14230c16b537SWarner Losh } 14240c16b537SWarner Losh 14250c16b537SWarner Losh return 0; 14260c16b537SWarner Losh } 14270c16b537SWarner Losh 14280c16b537SWarner Losh FORCE_INLINE size_t FSEv05_decompress_usingDTable_generic( 14290c16b537SWarner Losh void* dst, size_t maxDstSize, 14300c16b537SWarner Losh const void* cSrc, size_t cSrcSize, 14310c16b537SWarner Losh const FSEv05_DTable* dt, const unsigned fast) 14320c16b537SWarner Losh { 14330c16b537SWarner Losh BYTE* const ostart = (BYTE*) dst; 14340c16b537SWarner Losh BYTE* op = ostart; 14350c16b537SWarner Losh BYTE* const omax = op + maxDstSize; 14360c16b537SWarner Losh BYTE* const olimit = omax-3; 14370c16b537SWarner Losh 14380c16b537SWarner Losh BITv05_DStream_t bitD; 14390c16b537SWarner Losh FSEv05_DState_t state1; 14400c16b537SWarner Losh FSEv05_DState_t state2; 14410c16b537SWarner Losh size_t errorCode; 14420c16b537SWarner Losh 14430c16b537SWarner Losh /* Init */ 14440c16b537SWarner Losh errorCode = BITv05_initDStream(&bitD, cSrc, cSrcSize); /* replaced last arg by maxCompressed Size */ 14450c16b537SWarner Losh if (FSEv05_isError(errorCode)) return errorCode; 14460c16b537SWarner Losh 14470c16b537SWarner Losh FSEv05_initDState(&state1, &bitD, dt); 14480c16b537SWarner Losh FSEv05_initDState(&state2, &bitD, dt); 14490c16b537SWarner Losh 14500c16b537SWarner Losh #define FSEv05_GETSYMBOL(statePtr) fast ? FSEv05_decodeSymbolFast(statePtr, &bitD) : FSEv05_decodeSymbol(statePtr, &bitD) 14510c16b537SWarner Losh 14520c16b537SWarner Losh /* 4 symbols per loop */ 14530c16b537SWarner Losh for ( ; (BITv05_reloadDStream(&bitD)==BITv05_DStream_unfinished) && (op<olimit) ; op+=4) { 14540c16b537SWarner Losh op[0] = FSEv05_GETSYMBOL(&state1); 14550c16b537SWarner Losh 14560c16b537SWarner Losh if (FSEv05_MAX_TABLELOG*2+7 > sizeof(bitD.bitContainer)*8) /* This test must be static */ 14570c16b537SWarner Losh BITv05_reloadDStream(&bitD); 14580c16b537SWarner Losh 14590c16b537SWarner Losh op[1] = FSEv05_GETSYMBOL(&state2); 14600c16b537SWarner Losh 14610c16b537SWarner Losh if (FSEv05_MAX_TABLELOG*4+7 > sizeof(bitD.bitContainer)*8) /* This test must be static */ 14620c16b537SWarner Losh { if (BITv05_reloadDStream(&bitD) > BITv05_DStream_unfinished) { op+=2; break; } } 14630c16b537SWarner Losh 14640c16b537SWarner Losh op[2] = FSEv05_GETSYMBOL(&state1); 14650c16b537SWarner Losh 14660c16b537SWarner Losh if (FSEv05_MAX_TABLELOG*2+7 > sizeof(bitD.bitContainer)*8) /* This test must be static */ 14670c16b537SWarner Losh BITv05_reloadDStream(&bitD); 14680c16b537SWarner Losh 14690c16b537SWarner Losh op[3] = FSEv05_GETSYMBOL(&state2); 14700c16b537SWarner Losh } 14710c16b537SWarner Losh 14720c16b537SWarner Losh /* tail */ 14730c16b537SWarner Losh /* note : BITv05_reloadDStream(&bitD) >= FSEv05_DStream_partiallyFilled; Ends at exactly BITv05_DStream_completed */ 14740c16b537SWarner Losh while (1) { 14750c16b537SWarner Losh if ( (BITv05_reloadDStream(&bitD)>BITv05_DStream_completed) || (op==omax) || (BITv05_endOfDStream(&bitD) && (fast || FSEv05_endOfDState(&state1))) ) 14760c16b537SWarner Losh break; 14770c16b537SWarner Losh 14780c16b537SWarner Losh *op++ = FSEv05_GETSYMBOL(&state1); 14790c16b537SWarner Losh 14800c16b537SWarner Losh if ( (BITv05_reloadDStream(&bitD)>BITv05_DStream_completed) || (op==omax) || (BITv05_endOfDStream(&bitD) && (fast || FSEv05_endOfDState(&state2))) ) 14810c16b537SWarner Losh break; 14820c16b537SWarner Losh 14830c16b537SWarner Losh *op++ = FSEv05_GETSYMBOL(&state2); 14840c16b537SWarner Losh } 14850c16b537SWarner Losh 14860c16b537SWarner Losh /* end ? */ 14870c16b537SWarner Losh if (BITv05_endOfDStream(&bitD) && FSEv05_endOfDState(&state1) && FSEv05_endOfDState(&state2)) 14880c16b537SWarner Losh return op-ostart; 14890c16b537SWarner Losh 14900c16b537SWarner Losh if (op==omax) return ERROR(dstSize_tooSmall); /* dst buffer is full, but cSrc unfinished */ 14910c16b537SWarner Losh 14920c16b537SWarner Losh return ERROR(corruption_detected); 14930c16b537SWarner Losh } 14940c16b537SWarner Losh 14950c16b537SWarner Losh 14960c16b537SWarner Losh size_t FSEv05_decompress_usingDTable(void* dst, size_t originalSize, 14970c16b537SWarner Losh const void* cSrc, size_t cSrcSize, 14980c16b537SWarner Losh const FSEv05_DTable* dt) 14990c16b537SWarner Losh { 15000c16b537SWarner Losh const void* ptr = dt; 15010c16b537SWarner Losh const FSEv05_DTableHeader* DTableH = (const FSEv05_DTableHeader*)ptr; 15020c16b537SWarner Losh const U32 fastMode = DTableH->fastMode; 15030c16b537SWarner Losh 15040c16b537SWarner Losh /* select fast mode (static) */ 15050c16b537SWarner Losh if (fastMode) return FSEv05_decompress_usingDTable_generic(dst, originalSize, cSrc, cSrcSize, dt, 1); 15060c16b537SWarner Losh return FSEv05_decompress_usingDTable_generic(dst, originalSize, cSrc, cSrcSize, dt, 0); 15070c16b537SWarner Losh } 15080c16b537SWarner Losh 15090c16b537SWarner Losh 15100c16b537SWarner Losh size_t FSEv05_decompress(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize) 15110c16b537SWarner Losh { 15120c16b537SWarner Losh const BYTE* const istart = (const BYTE*)cSrc; 15130c16b537SWarner Losh const BYTE* ip = istart; 15140c16b537SWarner Losh short counting[FSEv05_MAX_SYMBOL_VALUE+1]; 15150c16b537SWarner Losh DTable_max_t dt; /* Static analyzer seems unable to understand this table will be properly initialized later */ 15160c16b537SWarner Losh unsigned tableLog; 15170c16b537SWarner Losh unsigned maxSymbolValue = FSEv05_MAX_SYMBOL_VALUE; 15180c16b537SWarner Losh size_t errorCode; 15190c16b537SWarner Losh 15200c16b537SWarner Losh if (cSrcSize<2) return ERROR(srcSize_wrong); /* too small input size */ 15210c16b537SWarner Losh 15220c16b537SWarner Losh /* normal FSEv05 decoding mode */ 15230c16b537SWarner Losh errorCode = FSEv05_readNCount (counting, &maxSymbolValue, &tableLog, istart, cSrcSize); 15240c16b537SWarner Losh if (FSEv05_isError(errorCode)) return errorCode; 15250c16b537SWarner Losh if (errorCode >= cSrcSize) return ERROR(srcSize_wrong); /* too small input size */ 15260c16b537SWarner Losh ip += errorCode; 15270c16b537SWarner Losh cSrcSize -= errorCode; 15280c16b537SWarner Losh 15290c16b537SWarner Losh errorCode = FSEv05_buildDTable (dt, counting, maxSymbolValue, tableLog); 15300c16b537SWarner Losh if (FSEv05_isError(errorCode)) return errorCode; 15310c16b537SWarner Losh 15320c16b537SWarner Losh /* always return, even if it is an error code */ 15330c16b537SWarner Losh return FSEv05_decompress_usingDTable (dst, maxDstSize, ip, cSrcSize, dt); 15340c16b537SWarner Losh } 15350c16b537SWarner Losh 15360c16b537SWarner Losh 15370c16b537SWarner Losh 15380c16b537SWarner Losh #endif /* FSEv05_COMMONDEFS_ONLY */ 15390c16b537SWarner Losh /* ****************************************************************** 15400c16b537SWarner Losh Huff0 : Huffman coder, part of New Generation Entropy library 15410c16b537SWarner Losh header file 15420c16b537SWarner Losh Copyright (C) 2013-2016, Yann Collet. 15430c16b537SWarner Losh 15440c16b537SWarner Losh BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php) 15450c16b537SWarner Losh 15460c16b537SWarner Losh Redistribution and use in source and binary forms, with or without 15470c16b537SWarner Losh modification, are permitted provided that the following conditions are 15480c16b537SWarner Losh met: 15490c16b537SWarner Losh 15500c16b537SWarner Losh * Redistributions of source code must retain the above copyright 15510c16b537SWarner Losh notice, this list of conditions and the following disclaimer. 15520c16b537SWarner Losh * Redistributions in binary form must reproduce the above 15530c16b537SWarner Losh copyright notice, this list of conditions and the following disclaimer 15540c16b537SWarner Losh in the documentation and/or other materials provided with the 15550c16b537SWarner Losh distribution. 15560c16b537SWarner Losh 15570c16b537SWarner Losh THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 15580c16b537SWarner Losh "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 15590c16b537SWarner Losh LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 15600c16b537SWarner Losh A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 15610c16b537SWarner Losh OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 15620c16b537SWarner Losh SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 15630c16b537SWarner Losh LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 15640c16b537SWarner Losh DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 15650c16b537SWarner Losh THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 15660c16b537SWarner Losh (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 15670c16b537SWarner Losh OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 15680c16b537SWarner Losh 15690c16b537SWarner Losh You can contact the author at : 15700c16b537SWarner Losh - Source repository : https://github.com/Cyan4973/FiniteStateEntropy 15710c16b537SWarner Losh ****************************************************************** */ 15720c16b537SWarner Losh #ifndef HUFF0_H 15730c16b537SWarner Losh #define HUFF0_H 15740c16b537SWarner Losh 15750c16b537SWarner Losh #if defined (__cplusplus) 15760c16b537SWarner Losh extern "C" { 15770c16b537SWarner Losh #endif 15780c16b537SWarner Losh 15790c16b537SWarner Losh 15800c16b537SWarner Losh 15810c16b537SWarner Losh /* **************************************** 15820c16b537SWarner Losh * Huff0 simple functions 15830c16b537SWarner Losh ******************************************/ 15840c16b537SWarner Losh size_t HUFv05_decompress(void* dst, size_t dstSize, 15850c16b537SWarner Losh const void* cSrc, size_t cSrcSize); 15860c16b537SWarner Losh /*! 15870c16b537SWarner Losh HUFv05_decompress(): 15880c16b537SWarner Losh Decompress Huff0 data from buffer 'cSrc', of size 'cSrcSize', 15890c16b537SWarner Losh into already allocated destination buffer 'dst', of size 'dstSize'. 15900c16b537SWarner Losh @dstSize : must be the **exact** size of original (uncompressed) data. 15910c16b537SWarner Losh Note : in contrast with FSEv05, HUFv05_decompress can regenerate 15920c16b537SWarner Losh RLE (cSrcSize==1) and uncompressed (cSrcSize==dstSize) data, 15930c16b537SWarner Losh because it knows size to regenerate. 15940c16b537SWarner Losh @return : size of regenerated data (== dstSize) 15950c16b537SWarner Losh or an error code, which can be tested using HUFv05_isError() 15960c16b537SWarner Losh */ 15970c16b537SWarner Losh 15980c16b537SWarner Losh 15990c16b537SWarner Losh /* **************************************** 16000c16b537SWarner Losh * Tool functions 16010c16b537SWarner Losh ******************************************/ 16020c16b537SWarner Losh /* Error Management */ 16030c16b537SWarner Losh unsigned HUFv05_isError(size_t code); /* tells if a return value is an error code */ 16040c16b537SWarner Losh const char* HUFv05_getErrorName(size_t code); /* provides error code string (useful for debugging) */ 16050c16b537SWarner Losh 16060c16b537SWarner Losh 16070c16b537SWarner Losh #if defined (__cplusplus) 16080c16b537SWarner Losh } 16090c16b537SWarner Losh #endif 16100c16b537SWarner Losh 16110c16b537SWarner Losh #endif /* HUF0_H */ 16120c16b537SWarner Losh /* ****************************************************************** 16130c16b537SWarner Losh Huff0 : Huffman codec, part of New Generation Entropy library 16140c16b537SWarner Losh header file, for static linking only 16150c16b537SWarner Losh Copyright (C) 2013-2016, Yann Collet 16160c16b537SWarner Losh 16170c16b537SWarner Losh BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php) 16180c16b537SWarner Losh 16190c16b537SWarner Losh Redistribution and use in source and binary forms, with or without 16200c16b537SWarner Losh modification, are permitted provided that the following conditions are 16210c16b537SWarner Losh met: 16220c16b537SWarner Losh 16230c16b537SWarner Losh * Redistributions of source code must retain the above copyright 16240c16b537SWarner Losh notice, this list of conditions and the following disclaimer. 16250c16b537SWarner Losh * Redistributions in binary form must reproduce the above 16260c16b537SWarner Losh copyright notice, this list of conditions and the following disclaimer 16270c16b537SWarner Losh in the documentation and/or other materials provided with the 16280c16b537SWarner Losh distribution. 16290c16b537SWarner Losh 16300c16b537SWarner Losh THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 16310c16b537SWarner Losh "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 16320c16b537SWarner Losh LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 16330c16b537SWarner Losh A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 16340c16b537SWarner Losh OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 16350c16b537SWarner Losh SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 16360c16b537SWarner Losh LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 16370c16b537SWarner Losh DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 16380c16b537SWarner Losh THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 16390c16b537SWarner Losh (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 16400c16b537SWarner Losh OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 16410c16b537SWarner Losh 16420c16b537SWarner Losh You can contact the author at : 16430c16b537SWarner Losh - Source repository : https://github.com/Cyan4973/FiniteStateEntropy 16440c16b537SWarner Losh ****************************************************************** */ 16450c16b537SWarner Losh #ifndef HUF0_STATIC_H 16460c16b537SWarner Losh #define HUF0_STATIC_H 16470c16b537SWarner Losh 16480c16b537SWarner Losh #if defined (__cplusplus) 16490c16b537SWarner Losh extern "C" { 16500c16b537SWarner Losh #endif 16510c16b537SWarner Losh 16520c16b537SWarner Losh 16530c16b537SWarner Losh 16540c16b537SWarner Losh /* **************************************** 16550c16b537SWarner Losh * Static allocation 16560c16b537SWarner Losh ******************************************/ 16570c16b537SWarner Losh /* static allocation of Huff0's DTable */ 16580c16b537SWarner Losh #define HUFv05_DTABLE_SIZE(maxTableLog) (1 + (1<<maxTableLog)) 16590c16b537SWarner Losh #define HUFv05_CREATE_STATIC_DTABLEX2(DTable, maxTableLog) \ 16600c16b537SWarner Losh unsigned short DTable[HUFv05_DTABLE_SIZE(maxTableLog)] = { maxTableLog } 16610c16b537SWarner Losh #define HUFv05_CREATE_STATIC_DTABLEX4(DTable, maxTableLog) \ 16620c16b537SWarner Losh unsigned int DTable[HUFv05_DTABLE_SIZE(maxTableLog)] = { maxTableLog } 16630c16b537SWarner Losh #define HUFv05_CREATE_STATIC_DTABLEX6(DTable, maxTableLog) \ 16640c16b537SWarner Losh unsigned int DTable[HUFv05_DTABLE_SIZE(maxTableLog) * 3 / 2] = { maxTableLog } 16650c16b537SWarner Losh 16660c16b537SWarner Losh 16670c16b537SWarner Losh /* **************************************** 16680c16b537SWarner Losh * Advanced decompression functions 16690c16b537SWarner Losh ******************************************/ 16700c16b537SWarner Losh size_t HUFv05_decompress4X2 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize); /* single-symbol decoder */ 16710c16b537SWarner Losh size_t HUFv05_decompress4X4 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize); /* double-symbols decoder */ 16720c16b537SWarner Losh 16730c16b537SWarner Losh 16740c16b537SWarner Losh /* **************************************** 16750c16b537SWarner Losh * Huff0 detailed API 16760c16b537SWarner Losh ******************************************/ 16770c16b537SWarner Losh /*! 16780c16b537SWarner Losh HUFv05_decompress() does the following: 16790c16b537SWarner Losh 1. select the decompression algorithm (X2, X4, X6) based on pre-computed heuristics 16800c16b537SWarner Losh 2. build Huffman table from save, using HUFv05_readDTableXn() 16810c16b537SWarner Losh 3. decode 1 or 4 segments in parallel using HUFv05_decompressSXn_usingDTable 16820c16b537SWarner Losh */ 16830c16b537SWarner Losh size_t HUFv05_readDTableX2 (unsigned short* DTable, const void* src, size_t srcSize); 16840c16b537SWarner Losh size_t HUFv05_readDTableX4 (unsigned* DTable, const void* src, size_t srcSize); 16850c16b537SWarner Losh 16860c16b537SWarner Losh size_t HUFv05_decompress4X2_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const unsigned short* DTable); 16870c16b537SWarner Losh size_t HUFv05_decompress4X4_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const unsigned* DTable); 16880c16b537SWarner Losh 16890c16b537SWarner Losh 16900c16b537SWarner Losh /* single stream variants */ 16910c16b537SWarner Losh 16920c16b537SWarner Losh size_t HUFv05_decompress1X2 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize); /* single-symbol decoder */ 16930c16b537SWarner Losh size_t HUFv05_decompress1X4 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize); /* double-symbol decoder */ 16940c16b537SWarner Losh 16950c16b537SWarner Losh size_t HUFv05_decompress1X2_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const unsigned short* DTable); 16960c16b537SWarner Losh size_t HUFv05_decompress1X4_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const unsigned* DTable); 16970c16b537SWarner Losh 16980c16b537SWarner Losh 16990c16b537SWarner Losh 17000c16b537SWarner Losh #if defined (__cplusplus) 17010c16b537SWarner Losh } 17020c16b537SWarner Losh #endif 17030c16b537SWarner Losh 17040c16b537SWarner Losh #endif /* HUF0_STATIC_H */ 17050c16b537SWarner Losh /* ****************************************************************** 17060c16b537SWarner Losh Huff0 : Huffman coder, part of New Generation Entropy library 17070c16b537SWarner Losh Copyright (C) 2013-2015, Yann Collet. 17080c16b537SWarner Losh 17090c16b537SWarner Losh BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php) 17100c16b537SWarner Losh 17110c16b537SWarner Losh Redistribution and use in source and binary forms, with or without 17120c16b537SWarner Losh modification, are permitted provided that the following conditions are 17130c16b537SWarner Losh met: 17140c16b537SWarner Losh 17150c16b537SWarner Losh * Redistributions of source code must retain the above copyright 17160c16b537SWarner Losh notice, this list of conditions and the following disclaimer. 17170c16b537SWarner Losh * Redistributions in binary form must reproduce the above 17180c16b537SWarner Losh copyright notice, this list of conditions and the following disclaimer 17190c16b537SWarner Losh in the documentation and/or other materials provided with the 17200c16b537SWarner Losh distribution. 17210c16b537SWarner Losh 17220c16b537SWarner Losh THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 17230c16b537SWarner Losh "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 17240c16b537SWarner Losh LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 17250c16b537SWarner Losh A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 17260c16b537SWarner Losh OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 17270c16b537SWarner Losh SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 17280c16b537SWarner Losh LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 17290c16b537SWarner Losh DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 17300c16b537SWarner Losh THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 17310c16b537SWarner Losh (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 17320c16b537SWarner Losh OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 17330c16b537SWarner Losh 17340c16b537SWarner Losh You can contact the author at : 17350c16b537SWarner Losh - FSEv05+Huff0 source repository : https://github.com/Cyan4973/FiniteStateEntropy 17360c16b537SWarner Losh - Public forum : https://groups.google.com/forum/#!forum/lz4c 17370c16b537SWarner Losh ****************************************************************** */ 17380c16b537SWarner Losh 17390c16b537SWarner Losh /* ************************************************************** 17400c16b537SWarner Losh * Compiler specifics 17410c16b537SWarner Losh ****************************************************************/ 17420c16b537SWarner Losh #if defined (__cplusplus) || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */) 17430c16b537SWarner Losh /* inline is defined */ 17440c16b537SWarner Losh #elif defined(_MSC_VER) 17450c16b537SWarner Losh # define inline __inline 17460c16b537SWarner Losh #else 17470c16b537SWarner Losh # define inline /* disable inline */ 17480c16b537SWarner Losh #endif 17490c16b537SWarner Losh 17500c16b537SWarner Losh 17510c16b537SWarner Losh #ifdef _MSC_VER /* Visual Studio */ 17520c16b537SWarner Losh # pragma warning(disable : 4127) /* disable: C4127: conditional expression is constant */ 17530c16b537SWarner Losh #endif 17540c16b537SWarner Losh 17550c16b537SWarner Losh 17560c16b537SWarner Losh /* ************************************************************** 17570c16b537SWarner Losh * Includes 17580c16b537SWarner Losh ****************************************************************/ 17590c16b537SWarner Losh #include <stdlib.h> /* malloc, free, qsort */ 17600c16b537SWarner Losh #include <string.h> /* memcpy, memset */ 17610c16b537SWarner Losh #include <stdio.h> /* printf (debug) */ 17620c16b537SWarner Losh 17630c16b537SWarner Losh 17640c16b537SWarner Losh /* ************************************************************** 17650c16b537SWarner Losh * Constants 17660c16b537SWarner Losh ****************************************************************/ 17670c16b537SWarner Losh #define HUFv05_ABSOLUTEMAX_TABLELOG 16 /* absolute limit of HUFv05_MAX_TABLELOG. Beyond that value, code does not work */ 17680c16b537SWarner Losh #define HUFv05_MAX_TABLELOG 12 /* max configured tableLog (for static allocation); can be modified up to HUFv05_ABSOLUTEMAX_TABLELOG */ 17690c16b537SWarner Losh #define HUFv05_DEFAULT_TABLELOG HUFv05_MAX_TABLELOG /* tableLog by default, when not specified */ 17700c16b537SWarner Losh #define HUFv05_MAX_SYMBOL_VALUE 255 17710c16b537SWarner Losh #if (HUFv05_MAX_TABLELOG > HUFv05_ABSOLUTEMAX_TABLELOG) 17720c16b537SWarner Losh # error "HUFv05_MAX_TABLELOG is too large !" 17730c16b537SWarner Losh #endif 17740c16b537SWarner Losh 17750c16b537SWarner Losh 17760c16b537SWarner Losh /* ************************************************************** 17770c16b537SWarner Losh * Error Management 17780c16b537SWarner Losh ****************************************************************/ 17790c16b537SWarner Losh unsigned HUFv05_isError(size_t code) { return ERR_isError(code); } 17800c16b537SWarner Losh const char* HUFv05_getErrorName(size_t code) { return ERR_getErrorName(code); } 17810c16b537SWarner Losh #define HUFv05_STATIC_ASSERT(c) { enum { HUFv05_static_assert = 1/(int)(!!(c)) }; } /* use only *after* variable declarations */ 17820c16b537SWarner Losh 17830c16b537SWarner Losh 17840c16b537SWarner Losh /* ******************************************************* 17850c16b537SWarner Losh * Huff0 : Huffman block decompression 17860c16b537SWarner Losh *********************************************************/ 17870c16b537SWarner Losh typedef struct { BYTE byte; BYTE nbBits; } HUFv05_DEltX2; /* single-symbol decoding */ 17880c16b537SWarner Losh 17890c16b537SWarner Losh typedef struct { U16 sequence; BYTE nbBits; BYTE length; } HUFv05_DEltX4; /* double-symbols decoding */ 17900c16b537SWarner Losh 17910c16b537SWarner Losh typedef struct { BYTE symbol; BYTE weight; } sortedSymbol_t; 17920c16b537SWarner Losh 17930c16b537SWarner Losh /*! HUFv05_readStats 17940c16b537SWarner Losh Read compact Huffman tree, saved by HUFv05_writeCTable 17950c16b537SWarner Losh @huffWeight : destination buffer 17960c16b537SWarner Losh @return : size read from `src` 17970c16b537SWarner Losh */ 17980c16b537SWarner Losh static size_t HUFv05_readStats(BYTE* huffWeight, size_t hwSize, U32* rankStats, 17990c16b537SWarner Losh U32* nbSymbolsPtr, U32* tableLogPtr, 18000c16b537SWarner Losh const void* src, size_t srcSize) 18010c16b537SWarner Losh { 18020c16b537SWarner Losh U32 weightTotal; 18030c16b537SWarner Losh U32 tableLog; 18040c16b537SWarner Losh const BYTE* ip = (const BYTE*) src; 18050c16b537SWarner Losh size_t iSize; 18060c16b537SWarner Losh size_t oSize; 18070c16b537SWarner Losh U32 n; 18080c16b537SWarner Losh 18090c16b537SWarner Losh if (!srcSize) return ERROR(srcSize_wrong); 18100c16b537SWarner Losh iSize = ip[0]; 181137f1f268SConrad Meyer /* memset(huffWeight, 0, hwSize); */ /* is not necessary, even though some analyzer complain ... */ 18120c16b537SWarner Losh 18130c16b537SWarner Losh if (iSize >= 128) { /* special header */ 18140c16b537SWarner Losh if (iSize >= (242)) { /* RLE */ 18150c16b537SWarner Losh static int l[14] = { 1, 2, 3, 4, 7, 8, 15, 16, 31, 32, 63, 64, 127, 128 }; 18160c16b537SWarner Losh oSize = l[iSize-242]; 18170c16b537SWarner Losh memset(huffWeight, 1, hwSize); 18180c16b537SWarner Losh iSize = 0; 18190c16b537SWarner Losh } 18200c16b537SWarner Losh else { /* Incompressible */ 18210c16b537SWarner Losh oSize = iSize - 127; 18220c16b537SWarner Losh iSize = ((oSize+1)/2); 18230c16b537SWarner Losh if (iSize+1 > srcSize) return ERROR(srcSize_wrong); 18240c16b537SWarner Losh if (oSize >= hwSize) return ERROR(corruption_detected); 18250c16b537SWarner Losh ip += 1; 18260c16b537SWarner Losh for (n=0; n<oSize; n+=2) { 18270c16b537SWarner Losh huffWeight[n] = ip[n/2] >> 4; 18280c16b537SWarner Losh huffWeight[n+1] = ip[n/2] & 15; 18290c16b537SWarner Losh } } } 18300c16b537SWarner Losh else { /* header compressed with FSEv05 (normal case) */ 18310c16b537SWarner Losh if (iSize+1 > srcSize) return ERROR(srcSize_wrong); 18320c16b537SWarner Losh oSize = FSEv05_decompress(huffWeight, hwSize-1, ip+1, iSize); /* max (hwSize-1) values decoded, as last one is implied */ 18330c16b537SWarner Losh if (FSEv05_isError(oSize)) return oSize; 18340c16b537SWarner Losh } 18350c16b537SWarner Losh 18360c16b537SWarner Losh /* collect weight stats */ 18370c16b537SWarner Losh memset(rankStats, 0, (HUFv05_ABSOLUTEMAX_TABLELOG + 1) * sizeof(U32)); 18380c16b537SWarner Losh weightTotal = 0; 18390c16b537SWarner Losh for (n=0; n<oSize; n++) { 18400c16b537SWarner Losh if (huffWeight[n] >= HUFv05_ABSOLUTEMAX_TABLELOG) return ERROR(corruption_detected); 18410c16b537SWarner Losh rankStats[huffWeight[n]]++; 18420c16b537SWarner Losh weightTotal += (1 << huffWeight[n]) >> 1; 18430c16b537SWarner Losh } 18440c16b537SWarner Losh if (weightTotal == 0) return ERROR(corruption_detected); 18450c16b537SWarner Losh 18460c16b537SWarner Losh /* get last non-null symbol weight (implied, total must be 2^n) */ 18470c16b537SWarner Losh tableLog = BITv05_highbit32(weightTotal) + 1; 18480c16b537SWarner Losh if (tableLog > HUFv05_ABSOLUTEMAX_TABLELOG) return ERROR(corruption_detected); 18490c16b537SWarner Losh { /* determine last weight */ 18500c16b537SWarner Losh U32 total = 1 << tableLog; 18510c16b537SWarner Losh U32 rest = total - weightTotal; 18520c16b537SWarner Losh U32 verif = 1 << BITv05_highbit32(rest); 18530c16b537SWarner Losh U32 lastWeight = BITv05_highbit32(rest) + 1; 18540c16b537SWarner Losh if (verif != rest) return ERROR(corruption_detected); /* last value must be a clean power of 2 */ 18550c16b537SWarner Losh huffWeight[oSize] = (BYTE)lastWeight; 18560c16b537SWarner Losh rankStats[lastWeight]++; 18570c16b537SWarner Losh } 18580c16b537SWarner Losh 18590c16b537SWarner Losh /* check tree construction validity */ 18600c16b537SWarner Losh if ((rankStats[1] < 2) || (rankStats[1] & 1)) return ERROR(corruption_detected); /* by construction : at least 2 elts of rank 1, must be even */ 18610c16b537SWarner Losh 18620c16b537SWarner Losh /* results */ 18630c16b537SWarner Losh *nbSymbolsPtr = (U32)(oSize+1); 18640c16b537SWarner Losh *tableLogPtr = tableLog; 18650c16b537SWarner Losh return iSize+1; 18660c16b537SWarner Losh } 18670c16b537SWarner Losh 18680c16b537SWarner Losh 18690c16b537SWarner Losh /*-***************************/ 18700c16b537SWarner Losh /* single-symbol decoding */ 18710c16b537SWarner Losh /*-***************************/ 18720c16b537SWarner Losh 18730c16b537SWarner Losh size_t HUFv05_readDTableX2 (U16* DTable, const void* src, size_t srcSize) 18740c16b537SWarner Losh { 18750c16b537SWarner Losh BYTE huffWeight[HUFv05_MAX_SYMBOL_VALUE + 1]; 18760c16b537SWarner Losh U32 rankVal[HUFv05_ABSOLUTEMAX_TABLELOG + 1]; /* large enough for values from 0 to 16 */ 18770c16b537SWarner Losh U32 tableLog = 0; 18780c16b537SWarner Losh size_t iSize; 18790c16b537SWarner Losh U32 nbSymbols = 0; 18800c16b537SWarner Losh U32 n; 18810c16b537SWarner Losh U32 nextRankStart; 18820c16b537SWarner Losh void* const dtPtr = DTable + 1; 18830c16b537SWarner Losh HUFv05_DEltX2* const dt = (HUFv05_DEltX2*)dtPtr; 18840c16b537SWarner Losh 18850c16b537SWarner Losh HUFv05_STATIC_ASSERT(sizeof(HUFv05_DEltX2) == sizeof(U16)); /* if compilation fails here, assertion is false */ 188637f1f268SConrad Meyer /* memset(huffWeight, 0, sizeof(huffWeight)); */ /* is not necessary, even though some analyzer complain ... */ 18870c16b537SWarner Losh 18880c16b537SWarner Losh iSize = HUFv05_readStats(huffWeight, HUFv05_MAX_SYMBOL_VALUE + 1, rankVal, &nbSymbols, &tableLog, src, srcSize); 18890c16b537SWarner Losh if (HUFv05_isError(iSize)) return iSize; 18900c16b537SWarner Losh 18910c16b537SWarner Losh /* check result */ 18920c16b537SWarner Losh if (tableLog > DTable[0]) return ERROR(tableLog_tooLarge); /* DTable is too small */ 18930c16b537SWarner Losh DTable[0] = (U16)tableLog; /* maybe should separate sizeof allocated DTable, from used size of DTable, in case of re-use */ 18940c16b537SWarner Losh 18950c16b537SWarner Losh /* Prepare ranks */ 18960c16b537SWarner Losh nextRankStart = 0; 18970c16b537SWarner Losh for (n=1; n<=tableLog; n++) { 18980c16b537SWarner Losh U32 current = nextRankStart; 18990c16b537SWarner Losh nextRankStart += (rankVal[n] << (n-1)); 19000c16b537SWarner Losh rankVal[n] = current; 19010c16b537SWarner Losh } 19020c16b537SWarner Losh 19030c16b537SWarner Losh /* fill DTable */ 19040c16b537SWarner Losh for (n=0; n<nbSymbols; n++) { 19050c16b537SWarner Losh const U32 w = huffWeight[n]; 19060c16b537SWarner Losh const U32 length = (1 << w) >> 1; 19070c16b537SWarner Losh U32 i; 19080c16b537SWarner Losh HUFv05_DEltX2 D; 19090c16b537SWarner Losh D.byte = (BYTE)n; D.nbBits = (BYTE)(tableLog + 1 - w); 19100c16b537SWarner Losh for (i = rankVal[w]; i < rankVal[w] + length; i++) 19110c16b537SWarner Losh dt[i] = D; 19120c16b537SWarner Losh rankVal[w] += length; 19130c16b537SWarner Losh } 19140c16b537SWarner Losh 19150c16b537SWarner Losh return iSize; 19160c16b537SWarner Losh } 19170c16b537SWarner Losh 19180c16b537SWarner Losh static BYTE HUFv05_decodeSymbolX2(BITv05_DStream_t* Dstream, const HUFv05_DEltX2* dt, const U32 dtLog) 19190c16b537SWarner Losh { 19200c16b537SWarner Losh const size_t val = BITv05_lookBitsFast(Dstream, dtLog); /* note : dtLog >= 1 */ 19210c16b537SWarner Losh const BYTE c = dt[val].byte; 19220c16b537SWarner Losh BITv05_skipBits(Dstream, dt[val].nbBits); 19230c16b537SWarner Losh return c; 19240c16b537SWarner Losh } 19250c16b537SWarner Losh 19260c16b537SWarner Losh #define HUFv05_DECODE_SYMBOLX2_0(ptr, DStreamPtr) \ 19270c16b537SWarner Losh *ptr++ = HUFv05_decodeSymbolX2(DStreamPtr, dt, dtLog) 19280c16b537SWarner Losh 19290c16b537SWarner Losh #define HUFv05_DECODE_SYMBOLX2_1(ptr, DStreamPtr) \ 19300c16b537SWarner Losh if (MEM_64bits() || (HUFv05_MAX_TABLELOG<=12)) \ 19310c16b537SWarner Losh HUFv05_DECODE_SYMBOLX2_0(ptr, DStreamPtr) 19320c16b537SWarner Losh 19330c16b537SWarner Losh #define HUFv05_DECODE_SYMBOLX2_2(ptr, DStreamPtr) \ 19340c16b537SWarner Losh if (MEM_64bits()) \ 19350c16b537SWarner Losh HUFv05_DECODE_SYMBOLX2_0(ptr, DStreamPtr) 19360c16b537SWarner Losh 19370c16b537SWarner Losh static inline size_t HUFv05_decodeStreamX2(BYTE* p, BITv05_DStream_t* const bitDPtr, BYTE* const pEnd, const HUFv05_DEltX2* const dt, const U32 dtLog) 19380c16b537SWarner Losh { 19390c16b537SWarner Losh BYTE* const pStart = p; 19400c16b537SWarner Losh 19410c16b537SWarner Losh /* up to 4 symbols at a time */ 19420c16b537SWarner Losh while ((BITv05_reloadDStream(bitDPtr) == BITv05_DStream_unfinished) && (p <= pEnd-4)) { 19430c16b537SWarner Losh HUFv05_DECODE_SYMBOLX2_2(p, bitDPtr); 19440c16b537SWarner Losh HUFv05_DECODE_SYMBOLX2_1(p, bitDPtr); 19450c16b537SWarner Losh HUFv05_DECODE_SYMBOLX2_2(p, bitDPtr); 19460c16b537SWarner Losh HUFv05_DECODE_SYMBOLX2_0(p, bitDPtr); 19470c16b537SWarner Losh } 19480c16b537SWarner Losh 19490c16b537SWarner Losh /* closer to the end */ 19500c16b537SWarner Losh while ((BITv05_reloadDStream(bitDPtr) == BITv05_DStream_unfinished) && (p < pEnd)) 19510c16b537SWarner Losh HUFv05_DECODE_SYMBOLX2_0(p, bitDPtr); 19520c16b537SWarner Losh 19530c16b537SWarner Losh /* no more data to retrieve from bitstream, hence no need to reload */ 19540c16b537SWarner Losh while (p < pEnd) 19550c16b537SWarner Losh HUFv05_DECODE_SYMBOLX2_0(p, bitDPtr); 19560c16b537SWarner Losh 19570c16b537SWarner Losh return pEnd-pStart; 19580c16b537SWarner Losh } 19590c16b537SWarner Losh 19600c16b537SWarner Losh size_t HUFv05_decompress1X2_usingDTable( 19610c16b537SWarner Losh void* dst, size_t dstSize, 19620c16b537SWarner Losh const void* cSrc, size_t cSrcSize, 19630c16b537SWarner Losh const U16* DTable) 19640c16b537SWarner Losh { 19650c16b537SWarner Losh BYTE* op = (BYTE*)dst; 19660c16b537SWarner Losh BYTE* const oend = op + dstSize; 19670c16b537SWarner Losh const U32 dtLog = DTable[0]; 19680c16b537SWarner Losh const void* dtPtr = DTable; 19690c16b537SWarner Losh const HUFv05_DEltX2* const dt = ((const HUFv05_DEltX2*)dtPtr)+1; 19700c16b537SWarner Losh BITv05_DStream_t bitD; 19710c16b537SWarner Losh 19720c16b537SWarner Losh if (dstSize <= cSrcSize) return ERROR(dstSize_tooSmall); 19730c16b537SWarner Losh { size_t const errorCode = BITv05_initDStream(&bitD, cSrc, cSrcSize); 19740c16b537SWarner Losh if (HUFv05_isError(errorCode)) return errorCode; } 19750c16b537SWarner Losh 19760c16b537SWarner Losh HUFv05_decodeStreamX2(op, &bitD, oend, dt, dtLog); 19770c16b537SWarner Losh 19780c16b537SWarner Losh /* check */ 19790c16b537SWarner Losh if (!BITv05_endOfDStream(&bitD)) return ERROR(corruption_detected); 19800c16b537SWarner Losh 19810c16b537SWarner Losh return dstSize; 19820c16b537SWarner Losh } 19830c16b537SWarner Losh 19840c16b537SWarner Losh size_t HUFv05_decompress1X2 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize) 19850c16b537SWarner Losh { 19860c16b537SWarner Losh HUFv05_CREATE_STATIC_DTABLEX2(DTable, HUFv05_MAX_TABLELOG); 19870c16b537SWarner Losh const BYTE* ip = (const BYTE*) cSrc; 19880c16b537SWarner Losh size_t errorCode; 19890c16b537SWarner Losh 19900c16b537SWarner Losh errorCode = HUFv05_readDTableX2 (DTable, cSrc, cSrcSize); 19910c16b537SWarner Losh if (HUFv05_isError(errorCode)) return errorCode; 19920c16b537SWarner Losh if (errorCode >= cSrcSize) return ERROR(srcSize_wrong); 19930c16b537SWarner Losh ip += errorCode; 19940c16b537SWarner Losh cSrcSize -= errorCode; 19950c16b537SWarner Losh 19960c16b537SWarner Losh return HUFv05_decompress1X2_usingDTable (dst, dstSize, ip, cSrcSize, DTable); 19970c16b537SWarner Losh } 19980c16b537SWarner Losh 19990c16b537SWarner Losh 20000c16b537SWarner Losh size_t HUFv05_decompress4X2_usingDTable( 20010c16b537SWarner Losh void* dst, size_t dstSize, 20020c16b537SWarner Losh const void* cSrc, size_t cSrcSize, 20030c16b537SWarner Losh const U16* DTable) 20040c16b537SWarner Losh { 20054d3f1eafSConrad Meyer /* Check */ 20064d3f1eafSConrad Meyer if (cSrcSize < 10) return ERROR(corruption_detected); /* strict minimum : jump table + 1 byte per stream */ 20074d3f1eafSConrad Meyer { 20080c16b537SWarner Losh const BYTE* const istart = (const BYTE*) cSrc; 20090c16b537SWarner Losh BYTE* const ostart = (BYTE*) dst; 20100c16b537SWarner Losh BYTE* const oend = ostart + dstSize; 20110c16b537SWarner Losh const void* const dtPtr = DTable; 20120c16b537SWarner Losh const HUFv05_DEltX2* const dt = ((const HUFv05_DEltX2*)dtPtr) +1; 20130c16b537SWarner Losh const U32 dtLog = DTable[0]; 20140c16b537SWarner Losh size_t errorCode; 20150c16b537SWarner Losh 20160c16b537SWarner Losh /* Init */ 20170c16b537SWarner Losh BITv05_DStream_t bitD1; 20180c16b537SWarner Losh BITv05_DStream_t bitD2; 20190c16b537SWarner Losh BITv05_DStream_t bitD3; 20200c16b537SWarner Losh BITv05_DStream_t bitD4; 20210c16b537SWarner Losh const size_t length1 = MEM_readLE16(istart); 20220c16b537SWarner Losh const size_t length2 = MEM_readLE16(istart+2); 20230c16b537SWarner Losh const size_t length3 = MEM_readLE16(istart+4); 20240c16b537SWarner Losh size_t length4; 20250c16b537SWarner Losh const BYTE* const istart1 = istart + 6; /* jumpTable */ 20260c16b537SWarner Losh const BYTE* const istart2 = istart1 + length1; 20270c16b537SWarner Losh const BYTE* const istart3 = istart2 + length2; 20280c16b537SWarner Losh const BYTE* const istart4 = istart3 + length3; 20290c16b537SWarner Losh const size_t segmentSize = (dstSize+3) / 4; 20300c16b537SWarner Losh BYTE* const opStart2 = ostart + segmentSize; 20310c16b537SWarner Losh BYTE* const opStart3 = opStart2 + segmentSize; 20320c16b537SWarner Losh BYTE* const opStart4 = opStart3 + segmentSize; 20330c16b537SWarner Losh BYTE* op1 = ostart; 20340c16b537SWarner Losh BYTE* op2 = opStart2; 20350c16b537SWarner Losh BYTE* op3 = opStart3; 20360c16b537SWarner Losh BYTE* op4 = opStart4; 20370c16b537SWarner Losh U32 endSignal; 20380c16b537SWarner Losh 20390c16b537SWarner Losh length4 = cSrcSize - (length1 + length2 + length3 + 6); 20400c16b537SWarner Losh if (length4 > cSrcSize) return ERROR(corruption_detected); /* overflow */ 20410c16b537SWarner Losh errorCode = BITv05_initDStream(&bitD1, istart1, length1); 20420c16b537SWarner Losh if (HUFv05_isError(errorCode)) return errorCode; 20430c16b537SWarner Losh errorCode = BITv05_initDStream(&bitD2, istart2, length2); 20440c16b537SWarner Losh if (HUFv05_isError(errorCode)) return errorCode; 20450c16b537SWarner Losh errorCode = BITv05_initDStream(&bitD3, istart3, length3); 20460c16b537SWarner Losh if (HUFv05_isError(errorCode)) return errorCode; 20470c16b537SWarner Losh errorCode = BITv05_initDStream(&bitD4, istart4, length4); 20480c16b537SWarner Losh if (HUFv05_isError(errorCode)) return errorCode; 20490c16b537SWarner Losh 20500c16b537SWarner Losh /* 16-32 symbols per loop (4-8 symbols per stream) */ 20510c16b537SWarner Losh endSignal = BITv05_reloadDStream(&bitD1) | BITv05_reloadDStream(&bitD2) | BITv05_reloadDStream(&bitD3) | BITv05_reloadDStream(&bitD4); 20520c16b537SWarner Losh for ( ; (endSignal==BITv05_DStream_unfinished) && (op4<(oend-7)) ; ) { 20530c16b537SWarner Losh HUFv05_DECODE_SYMBOLX2_2(op1, &bitD1); 20540c16b537SWarner Losh HUFv05_DECODE_SYMBOLX2_2(op2, &bitD2); 20550c16b537SWarner Losh HUFv05_DECODE_SYMBOLX2_2(op3, &bitD3); 20560c16b537SWarner Losh HUFv05_DECODE_SYMBOLX2_2(op4, &bitD4); 20570c16b537SWarner Losh HUFv05_DECODE_SYMBOLX2_1(op1, &bitD1); 20580c16b537SWarner Losh HUFv05_DECODE_SYMBOLX2_1(op2, &bitD2); 20590c16b537SWarner Losh HUFv05_DECODE_SYMBOLX2_1(op3, &bitD3); 20600c16b537SWarner Losh HUFv05_DECODE_SYMBOLX2_1(op4, &bitD4); 20610c16b537SWarner Losh HUFv05_DECODE_SYMBOLX2_2(op1, &bitD1); 20620c16b537SWarner Losh HUFv05_DECODE_SYMBOLX2_2(op2, &bitD2); 20630c16b537SWarner Losh HUFv05_DECODE_SYMBOLX2_2(op3, &bitD3); 20640c16b537SWarner Losh HUFv05_DECODE_SYMBOLX2_2(op4, &bitD4); 20650c16b537SWarner Losh HUFv05_DECODE_SYMBOLX2_0(op1, &bitD1); 20660c16b537SWarner Losh HUFv05_DECODE_SYMBOLX2_0(op2, &bitD2); 20670c16b537SWarner Losh HUFv05_DECODE_SYMBOLX2_0(op3, &bitD3); 20680c16b537SWarner Losh HUFv05_DECODE_SYMBOLX2_0(op4, &bitD4); 20690c16b537SWarner Losh endSignal = BITv05_reloadDStream(&bitD1) | BITv05_reloadDStream(&bitD2) | BITv05_reloadDStream(&bitD3) | BITv05_reloadDStream(&bitD4); 20700c16b537SWarner Losh } 20710c16b537SWarner Losh 20720c16b537SWarner Losh /* check corruption */ 20730c16b537SWarner Losh if (op1 > opStart2) return ERROR(corruption_detected); 20740c16b537SWarner Losh if (op2 > opStart3) return ERROR(corruption_detected); 20750c16b537SWarner Losh if (op3 > opStart4) return ERROR(corruption_detected); 20760c16b537SWarner Losh /* note : op4 supposed already verified within main loop */ 20770c16b537SWarner Losh 20780c16b537SWarner Losh /* finish bitStreams one by one */ 20790c16b537SWarner Losh HUFv05_decodeStreamX2(op1, &bitD1, opStart2, dt, dtLog); 20800c16b537SWarner Losh HUFv05_decodeStreamX2(op2, &bitD2, opStart3, dt, dtLog); 20810c16b537SWarner Losh HUFv05_decodeStreamX2(op3, &bitD3, opStart4, dt, dtLog); 20820c16b537SWarner Losh HUFv05_decodeStreamX2(op4, &bitD4, oend, dt, dtLog); 20830c16b537SWarner Losh 20840c16b537SWarner Losh /* check */ 20850c16b537SWarner Losh endSignal = BITv05_endOfDStream(&bitD1) & BITv05_endOfDStream(&bitD2) & BITv05_endOfDStream(&bitD3) & BITv05_endOfDStream(&bitD4); 20860c16b537SWarner Losh if (!endSignal) return ERROR(corruption_detected); 20870c16b537SWarner Losh 20880c16b537SWarner Losh /* decoded size */ 20890c16b537SWarner Losh return dstSize; 20900c16b537SWarner Losh } 20914d3f1eafSConrad Meyer } 20920c16b537SWarner Losh 20930c16b537SWarner Losh 20940c16b537SWarner Losh size_t HUFv05_decompress4X2 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize) 20950c16b537SWarner Losh { 20960c16b537SWarner Losh HUFv05_CREATE_STATIC_DTABLEX2(DTable, HUFv05_MAX_TABLELOG); 20970c16b537SWarner Losh const BYTE* ip = (const BYTE*) cSrc; 20980c16b537SWarner Losh size_t errorCode; 20990c16b537SWarner Losh 21000c16b537SWarner Losh errorCode = HUFv05_readDTableX2 (DTable, cSrc, cSrcSize); 21010c16b537SWarner Losh if (HUFv05_isError(errorCode)) return errorCode; 21020c16b537SWarner Losh if (errorCode >= cSrcSize) return ERROR(srcSize_wrong); 21030c16b537SWarner Losh ip += errorCode; 21040c16b537SWarner Losh cSrcSize -= errorCode; 21050c16b537SWarner Losh 21060c16b537SWarner Losh return HUFv05_decompress4X2_usingDTable (dst, dstSize, ip, cSrcSize, DTable); 21070c16b537SWarner Losh } 21080c16b537SWarner Losh 21090c16b537SWarner Losh 21100c16b537SWarner Losh /* *************************/ 21110c16b537SWarner Losh /* double-symbols decoding */ 21120c16b537SWarner Losh /* *************************/ 21130c16b537SWarner Losh 21140c16b537SWarner Losh static void HUFv05_fillDTableX4Level2(HUFv05_DEltX4* DTable, U32 sizeLog, const U32 consumed, 21150c16b537SWarner Losh const U32* rankValOrigin, const int minWeight, 21160c16b537SWarner Losh const sortedSymbol_t* sortedSymbols, const U32 sortedListSize, 21170c16b537SWarner Losh U32 nbBitsBaseline, U16 baseSeq) 21180c16b537SWarner Losh { 21190c16b537SWarner Losh HUFv05_DEltX4 DElt; 21200c16b537SWarner Losh U32 rankVal[HUFv05_ABSOLUTEMAX_TABLELOG + 1]; 21210c16b537SWarner Losh U32 s; 21220c16b537SWarner Losh 21230c16b537SWarner Losh /* get pre-calculated rankVal */ 21240c16b537SWarner Losh memcpy(rankVal, rankValOrigin, sizeof(rankVal)); 21250c16b537SWarner Losh 21260c16b537SWarner Losh /* fill skipped values */ 21270c16b537SWarner Losh if (minWeight>1) { 21280c16b537SWarner Losh U32 i, skipSize = rankVal[minWeight]; 21290c16b537SWarner Losh MEM_writeLE16(&(DElt.sequence), baseSeq); 21300c16b537SWarner Losh DElt.nbBits = (BYTE)(consumed); 21310c16b537SWarner Losh DElt.length = 1; 21320c16b537SWarner Losh for (i = 0; i < skipSize; i++) 21330c16b537SWarner Losh DTable[i] = DElt; 21340c16b537SWarner Losh } 21350c16b537SWarner Losh 21360c16b537SWarner Losh /* fill DTable */ 21370c16b537SWarner Losh for (s=0; s<sortedListSize; s++) { /* note : sortedSymbols already skipped */ 21380c16b537SWarner Losh const U32 symbol = sortedSymbols[s].symbol; 21390c16b537SWarner Losh const U32 weight = sortedSymbols[s].weight; 21400c16b537SWarner Losh const U32 nbBits = nbBitsBaseline - weight; 21410c16b537SWarner Losh const U32 length = 1 << (sizeLog-nbBits); 21420c16b537SWarner Losh const U32 start = rankVal[weight]; 21430c16b537SWarner Losh U32 i = start; 21440c16b537SWarner Losh const U32 end = start + length; 21450c16b537SWarner Losh 21460c16b537SWarner Losh MEM_writeLE16(&(DElt.sequence), (U16)(baseSeq + (symbol << 8))); 21470c16b537SWarner Losh DElt.nbBits = (BYTE)(nbBits + consumed); 21480c16b537SWarner Losh DElt.length = 2; 21490c16b537SWarner Losh do { DTable[i++] = DElt; } while (i<end); /* since length >= 1 */ 21500c16b537SWarner Losh 21510c16b537SWarner Losh rankVal[weight] += length; 21520c16b537SWarner Losh } 21530c16b537SWarner Losh } 21540c16b537SWarner Losh 21550c16b537SWarner Losh typedef U32 rankVal_t[HUFv05_ABSOLUTEMAX_TABLELOG][HUFv05_ABSOLUTEMAX_TABLELOG + 1]; 21560c16b537SWarner Losh 21570c16b537SWarner Losh static void HUFv05_fillDTableX4(HUFv05_DEltX4* DTable, const U32 targetLog, 21580c16b537SWarner Losh const sortedSymbol_t* sortedList, const U32 sortedListSize, 21590c16b537SWarner Losh const U32* rankStart, rankVal_t rankValOrigin, const U32 maxWeight, 21600c16b537SWarner Losh const U32 nbBitsBaseline) 21610c16b537SWarner Losh { 21620c16b537SWarner Losh U32 rankVal[HUFv05_ABSOLUTEMAX_TABLELOG + 1]; 21630c16b537SWarner Losh const int scaleLog = nbBitsBaseline - targetLog; /* note : targetLog >= srcLog, hence scaleLog <= 1 */ 21640c16b537SWarner Losh const U32 minBits = nbBitsBaseline - maxWeight; 21650c16b537SWarner Losh U32 s; 21660c16b537SWarner Losh 21670c16b537SWarner Losh memcpy(rankVal, rankValOrigin, sizeof(rankVal)); 21680c16b537SWarner Losh 21690c16b537SWarner Losh /* fill DTable */ 21700c16b537SWarner Losh for (s=0; s<sortedListSize; s++) { 21710c16b537SWarner Losh const U16 symbol = sortedList[s].symbol; 21720c16b537SWarner Losh const U32 weight = sortedList[s].weight; 21730c16b537SWarner Losh const U32 nbBits = nbBitsBaseline - weight; 21740c16b537SWarner Losh const U32 start = rankVal[weight]; 21750c16b537SWarner Losh const U32 length = 1 << (targetLog-nbBits); 21760c16b537SWarner Losh 21770c16b537SWarner Losh if (targetLog-nbBits >= minBits) { /* enough room for a second symbol */ 21780c16b537SWarner Losh U32 sortedRank; 21790c16b537SWarner Losh int minWeight = nbBits + scaleLog; 21800c16b537SWarner Losh if (minWeight < 1) minWeight = 1; 21810c16b537SWarner Losh sortedRank = rankStart[minWeight]; 21820c16b537SWarner Losh HUFv05_fillDTableX4Level2(DTable+start, targetLog-nbBits, nbBits, 21830c16b537SWarner Losh rankValOrigin[nbBits], minWeight, 21840c16b537SWarner Losh sortedList+sortedRank, sortedListSize-sortedRank, 21850c16b537SWarner Losh nbBitsBaseline, symbol); 21860c16b537SWarner Losh } else { 21870c16b537SWarner Losh U32 i; 21880c16b537SWarner Losh const U32 end = start + length; 21890c16b537SWarner Losh HUFv05_DEltX4 DElt; 21900c16b537SWarner Losh 21910c16b537SWarner Losh MEM_writeLE16(&(DElt.sequence), symbol); 21920c16b537SWarner Losh DElt.nbBits = (BYTE)(nbBits); 21930c16b537SWarner Losh DElt.length = 1; 21940c16b537SWarner Losh for (i = start; i < end; i++) 21950c16b537SWarner Losh DTable[i] = DElt; 21960c16b537SWarner Losh } 21970c16b537SWarner Losh rankVal[weight] += length; 21980c16b537SWarner Losh } 21990c16b537SWarner Losh } 22000c16b537SWarner Losh 2201a0483764SConrad Meyer size_t HUFv05_readDTableX4 (unsigned* DTable, const void* src, size_t srcSize) 22020c16b537SWarner Losh { 22030c16b537SWarner Losh BYTE weightList[HUFv05_MAX_SYMBOL_VALUE + 1]; 22040c16b537SWarner Losh sortedSymbol_t sortedSymbol[HUFv05_MAX_SYMBOL_VALUE + 1]; 22050c16b537SWarner Losh U32 rankStats[HUFv05_ABSOLUTEMAX_TABLELOG + 1] = { 0 }; 22060c16b537SWarner Losh U32 rankStart0[HUFv05_ABSOLUTEMAX_TABLELOG + 2] = { 0 }; 22070c16b537SWarner Losh U32* const rankStart = rankStart0+1; 22080c16b537SWarner Losh rankVal_t rankVal; 22090c16b537SWarner Losh U32 tableLog, maxW, sizeOfSort, nbSymbols; 22100c16b537SWarner Losh const U32 memLog = DTable[0]; 22110c16b537SWarner Losh size_t iSize; 22120c16b537SWarner Losh void* dtPtr = DTable; 22130c16b537SWarner Losh HUFv05_DEltX4* const dt = ((HUFv05_DEltX4*)dtPtr) + 1; 22140c16b537SWarner Losh 2215a0483764SConrad Meyer HUFv05_STATIC_ASSERT(sizeof(HUFv05_DEltX4) == sizeof(unsigned)); /* if compilation fails here, assertion is false */ 22160c16b537SWarner Losh if (memLog > HUFv05_ABSOLUTEMAX_TABLELOG) return ERROR(tableLog_tooLarge); 221737f1f268SConrad Meyer /* memset(weightList, 0, sizeof(weightList)); */ /* is not necessary, even though some analyzer complain ... */ 22180c16b537SWarner Losh 22190c16b537SWarner Losh iSize = HUFv05_readStats(weightList, HUFv05_MAX_SYMBOL_VALUE + 1, rankStats, &nbSymbols, &tableLog, src, srcSize); 22200c16b537SWarner Losh if (HUFv05_isError(iSize)) return iSize; 22210c16b537SWarner Losh 22220c16b537SWarner Losh /* check result */ 22230c16b537SWarner Losh if (tableLog > memLog) return ERROR(tableLog_tooLarge); /* DTable can't fit code depth */ 22240c16b537SWarner Losh 22250c16b537SWarner Losh /* find maxWeight */ 22260c16b537SWarner Losh for (maxW = tableLog; rankStats[maxW]==0; maxW--) {} /* necessarily finds a solution before 0 */ 22270c16b537SWarner Losh 22280c16b537SWarner Losh /* Get start index of each weight */ 22290c16b537SWarner Losh { 22300c16b537SWarner Losh U32 w, nextRankStart = 0; 22310c16b537SWarner Losh for (w=1; w<=maxW; w++) { 22320c16b537SWarner Losh U32 current = nextRankStart; 22330c16b537SWarner Losh nextRankStart += rankStats[w]; 22340c16b537SWarner Losh rankStart[w] = current; 22350c16b537SWarner Losh } 22360c16b537SWarner Losh rankStart[0] = nextRankStart; /* put all 0w symbols at the end of sorted list*/ 22370c16b537SWarner Losh sizeOfSort = nextRankStart; 22380c16b537SWarner Losh } 22390c16b537SWarner Losh 22400c16b537SWarner Losh /* sort symbols by weight */ 22410c16b537SWarner Losh { 22420c16b537SWarner Losh U32 s; 22430c16b537SWarner Losh for (s=0; s<nbSymbols; s++) { 22440c16b537SWarner Losh U32 w = weightList[s]; 22450c16b537SWarner Losh U32 r = rankStart[w]++; 22460c16b537SWarner Losh sortedSymbol[r].symbol = (BYTE)s; 22470c16b537SWarner Losh sortedSymbol[r].weight = (BYTE)w; 22480c16b537SWarner Losh } 22490c16b537SWarner Losh rankStart[0] = 0; /* forget 0w symbols; this is beginning of weight(1) */ 22500c16b537SWarner Losh } 22510c16b537SWarner Losh 22520c16b537SWarner Losh /* Build rankVal */ 22530c16b537SWarner Losh { 22540c16b537SWarner Losh const U32 minBits = tableLog+1 - maxW; 22550c16b537SWarner Losh U32 nextRankVal = 0; 22560c16b537SWarner Losh U32 w, consumed; 22570c16b537SWarner Losh const int rescale = (memLog-tableLog) - 1; /* tableLog <= memLog */ 22580c16b537SWarner Losh U32* rankVal0 = rankVal[0]; 22590c16b537SWarner Losh for (w=1; w<=maxW; w++) { 22600c16b537SWarner Losh U32 current = nextRankVal; 22610c16b537SWarner Losh nextRankVal += rankStats[w] << (w+rescale); 22620c16b537SWarner Losh rankVal0[w] = current; 22630c16b537SWarner Losh } 22640c16b537SWarner Losh for (consumed = minBits; consumed <= memLog - minBits; consumed++) { 22650c16b537SWarner Losh U32* rankValPtr = rankVal[consumed]; 22660c16b537SWarner Losh for (w = 1; w <= maxW; w++) { 22670c16b537SWarner Losh rankValPtr[w] = rankVal0[w] >> consumed; 22680c16b537SWarner Losh } } } 22690c16b537SWarner Losh 22700c16b537SWarner Losh HUFv05_fillDTableX4(dt, memLog, 22710c16b537SWarner Losh sortedSymbol, sizeOfSort, 22720c16b537SWarner Losh rankStart0, rankVal, maxW, 22730c16b537SWarner Losh tableLog+1); 22740c16b537SWarner Losh 22750c16b537SWarner Losh return iSize; 22760c16b537SWarner Losh } 22770c16b537SWarner Losh 22780c16b537SWarner Losh 22790c16b537SWarner Losh static U32 HUFv05_decodeSymbolX4(void* op, BITv05_DStream_t* DStream, const HUFv05_DEltX4* dt, const U32 dtLog) 22800c16b537SWarner Losh { 22810c16b537SWarner Losh const size_t val = BITv05_lookBitsFast(DStream, dtLog); /* note : dtLog >= 1 */ 22820c16b537SWarner Losh memcpy(op, dt+val, 2); 22830c16b537SWarner Losh BITv05_skipBits(DStream, dt[val].nbBits); 22840c16b537SWarner Losh return dt[val].length; 22850c16b537SWarner Losh } 22860c16b537SWarner Losh 22870c16b537SWarner Losh static U32 HUFv05_decodeLastSymbolX4(void* op, BITv05_DStream_t* DStream, const HUFv05_DEltX4* dt, const U32 dtLog) 22880c16b537SWarner Losh { 22890c16b537SWarner Losh const size_t val = BITv05_lookBitsFast(DStream, dtLog); /* note : dtLog >= 1 */ 22900c16b537SWarner Losh memcpy(op, dt+val, 1); 22910c16b537SWarner Losh if (dt[val].length==1) BITv05_skipBits(DStream, dt[val].nbBits); 22920c16b537SWarner Losh else { 22930c16b537SWarner Losh if (DStream->bitsConsumed < (sizeof(DStream->bitContainer)*8)) { 22940c16b537SWarner Losh BITv05_skipBits(DStream, dt[val].nbBits); 22950c16b537SWarner Losh if (DStream->bitsConsumed > (sizeof(DStream->bitContainer)*8)) 22960c16b537SWarner Losh DStream->bitsConsumed = (sizeof(DStream->bitContainer)*8); /* ugly hack; works only because it's the last symbol. Note : can't easily extract nbBits from just this symbol */ 22970c16b537SWarner Losh } } 22980c16b537SWarner Losh return 1; 22990c16b537SWarner Losh } 23000c16b537SWarner Losh 23010c16b537SWarner Losh 23020c16b537SWarner Losh #define HUFv05_DECODE_SYMBOLX4_0(ptr, DStreamPtr) \ 23030c16b537SWarner Losh ptr += HUFv05_decodeSymbolX4(ptr, DStreamPtr, dt, dtLog) 23040c16b537SWarner Losh 23050c16b537SWarner Losh #define HUFv05_DECODE_SYMBOLX4_1(ptr, DStreamPtr) \ 23060c16b537SWarner Losh if (MEM_64bits() || (HUFv05_MAX_TABLELOG<=12)) \ 23070c16b537SWarner Losh ptr += HUFv05_decodeSymbolX4(ptr, DStreamPtr, dt, dtLog) 23080c16b537SWarner Losh 23090c16b537SWarner Losh #define HUFv05_DECODE_SYMBOLX4_2(ptr, DStreamPtr) \ 23100c16b537SWarner Losh if (MEM_64bits()) \ 23110c16b537SWarner Losh ptr += HUFv05_decodeSymbolX4(ptr, DStreamPtr, dt, dtLog) 23120c16b537SWarner Losh 23130c16b537SWarner Losh static inline size_t HUFv05_decodeStreamX4(BYTE* p, BITv05_DStream_t* bitDPtr, BYTE* const pEnd, const HUFv05_DEltX4* const dt, const U32 dtLog) 23140c16b537SWarner Losh { 23150c16b537SWarner Losh BYTE* const pStart = p; 23160c16b537SWarner Losh 23170c16b537SWarner Losh /* up to 8 symbols at a time */ 23180c16b537SWarner Losh while ((BITv05_reloadDStream(bitDPtr) == BITv05_DStream_unfinished) && (p < pEnd-7)) { 23190c16b537SWarner Losh HUFv05_DECODE_SYMBOLX4_2(p, bitDPtr); 23200c16b537SWarner Losh HUFv05_DECODE_SYMBOLX4_1(p, bitDPtr); 23210c16b537SWarner Losh HUFv05_DECODE_SYMBOLX4_2(p, bitDPtr); 23220c16b537SWarner Losh HUFv05_DECODE_SYMBOLX4_0(p, bitDPtr); 23230c16b537SWarner Losh } 23240c16b537SWarner Losh 23250c16b537SWarner Losh /* closer to the end */ 23260c16b537SWarner Losh while ((BITv05_reloadDStream(bitDPtr) == BITv05_DStream_unfinished) && (p <= pEnd-2)) 23270c16b537SWarner Losh HUFv05_DECODE_SYMBOLX4_0(p, bitDPtr); 23280c16b537SWarner Losh 23290c16b537SWarner Losh while (p <= pEnd-2) 23300c16b537SWarner Losh HUFv05_DECODE_SYMBOLX4_0(p, bitDPtr); /* no need to reload : reached the end of DStream */ 23310c16b537SWarner Losh 23320c16b537SWarner Losh if (p < pEnd) 23330c16b537SWarner Losh p += HUFv05_decodeLastSymbolX4(p, bitDPtr, dt, dtLog); 23340c16b537SWarner Losh 23350c16b537SWarner Losh return p-pStart; 23360c16b537SWarner Losh } 23370c16b537SWarner Losh 23380c16b537SWarner Losh 23390c16b537SWarner Losh size_t HUFv05_decompress1X4_usingDTable( 23400c16b537SWarner Losh void* dst, size_t dstSize, 23410c16b537SWarner Losh const void* cSrc, size_t cSrcSize, 2342a0483764SConrad Meyer const unsigned* DTable) 23430c16b537SWarner Losh { 23440c16b537SWarner Losh const BYTE* const istart = (const BYTE*) cSrc; 23450c16b537SWarner Losh BYTE* const ostart = (BYTE*) dst; 23460c16b537SWarner Losh BYTE* const oend = ostart + dstSize; 23470c16b537SWarner Losh 23480c16b537SWarner Losh const U32 dtLog = DTable[0]; 23490c16b537SWarner Losh const void* const dtPtr = DTable; 23500c16b537SWarner Losh const HUFv05_DEltX4* const dt = ((const HUFv05_DEltX4*)dtPtr) +1; 23510c16b537SWarner Losh size_t errorCode; 23520c16b537SWarner Losh 23530c16b537SWarner Losh /* Init */ 23540c16b537SWarner Losh BITv05_DStream_t bitD; 23550c16b537SWarner Losh errorCode = BITv05_initDStream(&bitD, istart, cSrcSize); 23560c16b537SWarner Losh if (HUFv05_isError(errorCode)) return errorCode; 23570c16b537SWarner Losh 23580c16b537SWarner Losh /* finish bitStreams one by one */ 23590c16b537SWarner Losh HUFv05_decodeStreamX4(ostart, &bitD, oend, dt, dtLog); 23600c16b537SWarner Losh 23610c16b537SWarner Losh /* check */ 23620c16b537SWarner Losh if (!BITv05_endOfDStream(&bitD)) return ERROR(corruption_detected); 23630c16b537SWarner Losh 23640c16b537SWarner Losh /* decoded size */ 23650c16b537SWarner Losh return dstSize; 23660c16b537SWarner Losh } 23670c16b537SWarner Losh 23680c16b537SWarner Losh size_t HUFv05_decompress1X4 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize) 23690c16b537SWarner Losh { 23700c16b537SWarner Losh HUFv05_CREATE_STATIC_DTABLEX4(DTable, HUFv05_MAX_TABLELOG); 23710c16b537SWarner Losh const BYTE* ip = (const BYTE*) cSrc; 23720c16b537SWarner Losh 23730c16b537SWarner Losh size_t hSize = HUFv05_readDTableX4 (DTable, cSrc, cSrcSize); 23740c16b537SWarner Losh if (HUFv05_isError(hSize)) return hSize; 23750c16b537SWarner Losh if (hSize >= cSrcSize) return ERROR(srcSize_wrong); 23760c16b537SWarner Losh ip += hSize; 23770c16b537SWarner Losh cSrcSize -= hSize; 23780c16b537SWarner Losh 23790c16b537SWarner Losh return HUFv05_decompress1X4_usingDTable (dst, dstSize, ip, cSrcSize, DTable); 23800c16b537SWarner Losh } 23810c16b537SWarner Losh 23820c16b537SWarner Losh size_t HUFv05_decompress4X4_usingDTable( 23830c16b537SWarner Losh void* dst, size_t dstSize, 23840c16b537SWarner Losh const void* cSrc, size_t cSrcSize, 2385a0483764SConrad Meyer const unsigned* DTable) 23860c16b537SWarner Losh { 23870c16b537SWarner Losh if (cSrcSize < 10) return ERROR(corruption_detected); /* strict minimum : jump table + 1 byte per stream */ 23880c16b537SWarner Losh 23890c16b537SWarner Losh { 23900c16b537SWarner Losh const BYTE* const istart = (const BYTE*) cSrc; 23910c16b537SWarner Losh BYTE* const ostart = (BYTE*) dst; 23920c16b537SWarner Losh BYTE* const oend = ostart + dstSize; 23930c16b537SWarner Losh const void* const dtPtr = DTable; 23940c16b537SWarner Losh const HUFv05_DEltX4* const dt = ((const HUFv05_DEltX4*)dtPtr) +1; 23950c16b537SWarner Losh const U32 dtLog = DTable[0]; 23960c16b537SWarner Losh size_t errorCode; 23970c16b537SWarner Losh 23980c16b537SWarner Losh /* Init */ 23990c16b537SWarner Losh BITv05_DStream_t bitD1; 24000c16b537SWarner Losh BITv05_DStream_t bitD2; 24010c16b537SWarner Losh BITv05_DStream_t bitD3; 24020c16b537SWarner Losh BITv05_DStream_t bitD4; 24030c16b537SWarner Losh const size_t length1 = MEM_readLE16(istart); 24040c16b537SWarner Losh const size_t length2 = MEM_readLE16(istart+2); 24050c16b537SWarner Losh const size_t length3 = MEM_readLE16(istart+4); 24060c16b537SWarner Losh size_t length4; 24070c16b537SWarner Losh const BYTE* const istart1 = istart + 6; /* jumpTable */ 24080c16b537SWarner Losh const BYTE* const istart2 = istart1 + length1; 24090c16b537SWarner Losh const BYTE* const istart3 = istart2 + length2; 24100c16b537SWarner Losh const BYTE* const istart4 = istart3 + length3; 24110c16b537SWarner Losh const size_t segmentSize = (dstSize+3) / 4; 24120c16b537SWarner Losh BYTE* const opStart2 = ostart + segmentSize; 24130c16b537SWarner Losh BYTE* const opStart3 = opStart2 + segmentSize; 24140c16b537SWarner Losh BYTE* const opStart4 = opStart3 + segmentSize; 24150c16b537SWarner Losh BYTE* op1 = ostart; 24160c16b537SWarner Losh BYTE* op2 = opStart2; 24170c16b537SWarner Losh BYTE* op3 = opStart3; 24180c16b537SWarner Losh BYTE* op4 = opStart4; 24190c16b537SWarner Losh U32 endSignal; 24200c16b537SWarner Losh 24210c16b537SWarner Losh length4 = cSrcSize - (length1 + length2 + length3 + 6); 24220c16b537SWarner Losh if (length4 > cSrcSize) return ERROR(corruption_detected); /* overflow */ 24230c16b537SWarner Losh errorCode = BITv05_initDStream(&bitD1, istart1, length1); 24240c16b537SWarner Losh if (HUFv05_isError(errorCode)) return errorCode; 24250c16b537SWarner Losh errorCode = BITv05_initDStream(&bitD2, istart2, length2); 24260c16b537SWarner Losh if (HUFv05_isError(errorCode)) return errorCode; 24270c16b537SWarner Losh errorCode = BITv05_initDStream(&bitD3, istart3, length3); 24280c16b537SWarner Losh if (HUFv05_isError(errorCode)) return errorCode; 24290c16b537SWarner Losh errorCode = BITv05_initDStream(&bitD4, istart4, length4); 24300c16b537SWarner Losh if (HUFv05_isError(errorCode)) return errorCode; 24310c16b537SWarner Losh 24320c16b537SWarner Losh /* 16-32 symbols per loop (4-8 symbols per stream) */ 24330c16b537SWarner Losh endSignal = BITv05_reloadDStream(&bitD1) | BITv05_reloadDStream(&bitD2) | BITv05_reloadDStream(&bitD3) | BITv05_reloadDStream(&bitD4); 24340c16b537SWarner Losh for ( ; (endSignal==BITv05_DStream_unfinished) && (op4<(oend-7)) ; ) { 24350c16b537SWarner Losh HUFv05_DECODE_SYMBOLX4_2(op1, &bitD1); 24360c16b537SWarner Losh HUFv05_DECODE_SYMBOLX4_2(op2, &bitD2); 24370c16b537SWarner Losh HUFv05_DECODE_SYMBOLX4_2(op3, &bitD3); 24380c16b537SWarner Losh HUFv05_DECODE_SYMBOLX4_2(op4, &bitD4); 24390c16b537SWarner Losh HUFv05_DECODE_SYMBOLX4_1(op1, &bitD1); 24400c16b537SWarner Losh HUFv05_DECODE_SYMBOLX4_1(op2, &bitD2); 24410c16b537SWarner Losh HUFv05_DECODE_SYMBOLX4_1(op3, &bitD3); 24420c16b537SWarner Losh HUFv05_DECODE_SYMBOLX4_1(op4, &bitD4); 24430c16b537SWarner Losh HUFv05_DECODE_SYMBOLX4_2(op1, &bitD1); 24440c16b537SWarner Losh HUFv05_DECODE_SYMBOLX4_2(op2, &bitD2); 24450c16b537SWarner Losh HUFv05_DECODE_SYMBOLX4_2(op3, &bitD3); 24460c16b537SWarner Losh HUFv05_DECODE_SYMBOLX4_2(op4, &bitD4); 24470c16b537SWarner Losh HUFv05_DECODE_SYMBOLX4_0(op1, &bitD1); 24480c16b537SWarner Losh HUFv05_DECODE_SYMBOLX4_0(op2, &bitD2); 24490c16b537SWarner Losh HUFv05_DECODE_SYMBOLX4_0(op3, &bitD3); 24500c16b537SWarner Losh HUFv05_DECODE_SYMBOLX4_0(op4, &bitD4); 24510c16b537SWarner Losh 24520c16b537SWarner Losh endSignal = BITv05_reloadDStream(&bitD1) | BITv05_reloadDStream(&bitD2) | BITv05_reloadDStream(&bitD3) | BITv05_reloadDStream(&bitD4); 24530c16b537SWarner Losh } 24540c16b537SWarner Losh 24550c16b537SWarner Losh /* check corruption */ 24560c16b537SWarner Losh if (op1 > opStart2) return ERROR(corruption_detected); 24570c16b537SWarner Losh if (op2 > opStart3) return ERROR(corruption_detected); 24580c16b537SWarner Losh if (op3 > opStart4) return ERROR(corruption_detected); 24590c16b537SWarner Losh /* note : op4 supposed already verified within main loop */ 24600c16b537SWarner Losh 24610c16b537SWarner Losh /* finish bitStreams one by one */ 24620c16b537SWarner Losh HUFv05_decodeStreamX4(op1, &bitD1, opStart2, dt, dtLog); 24630c16b537SWarner Losh HUFv05_decodeStreamX4(op2, &bitD2, opStart3, dt, dtLog); 24640c16b537SWarner Losh HUFv05_decodeStreamX4(op3, &bitD3, opStart4, dt, dtLog); 24650c16b537SWarner Losh HUFv05_decodeStreamX4(op4, &bitD4, oend, dt, dtLog); 24660c16b537SWarner Losh 24670c16b537SWarner Losh /* check */ 24680c16b537SWarner Losh endSignal = BITv05_endOfDStream(&bitD1) & BITv05_endOfDStream(&bitD2) & BITv05_endOfDStream(&bitD3) & BITv05_endOfDStream(&bitD4); 24690c16b537SWarner Losh if (!endSignal) return ERROR(corruption_detected); 24700c16b537SWarner Losh 24710c16b537SWarner Losh /* decoded size */ 24720c16b537SWarner Losh return dstSize; 24730c16b537SWarner Losh } 24740c16b537SWarner Losh } 24750c16b537SWarner Losh 24760c16b537SWarner Losh 24770c16b537SWarner Losh size_t HUFv05_decompress4X4 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize) 24780c16b537SWarner Losh { 24790c16b537SWarner Losh HUFv05_CREATE_STATIC_DTABLEX4(DTable, HUFv05_MAX_TABLELOG); 24800c16b537SWarner Losh const BYTE* ip = (const BYTE*) cSrc; 24810c16b537SWarner Losh 24820c16b537SWarner Losh size_t hSize = HUFv05_readDTableX4 (DTable, cSrc, cSrcSize); 24830c16b537SWarner Losh if (HUFv05_isError(hSize)) return hSize; 24840c16b537SWarner Losh if (hSize >= cSrcSize) return ERROR(srcSize_wrong); 24850c16b537SWarner Losh ip += hSize; 24860c16b537SWarner Losh cSrcSize -= hSize; 24870c16b537SWarner Losh 24880c16b537SWarner Losh return HUFv05_decompress4X4_usingDTable (dst, dstSize, ip, cSrcSize, DTable); 24890c16b537SWarner Losh } 24900c16b537SWarner Losh 24910c16b537SWarner Losh 24920c16b537SWarner Losh /* ********************************/ 24930c16b537SWarner Losh /* Generic decompression selector */ 24940c16b537SWarner Losh /* ********************************/ 24950c16b537SWarner Losh 24960c16b537SWarner Losh typedef struct { U32 tableTime; U32 decode256Time; } algo_time_t; 24970c16b537SWarner Losh static const algo_time_t algoTime[16 /* Quantization */][3 /* single, double, quad */] = 24980c16b537SWarner Losh { 24990c16b537SWarner Losh /* single, double, quad */ 25000c16b537SWarner Losh {{0,0}, {1,1}, {2,2}}, /* Q==0 : impossible */ 25010c16b537SWarner Losh {{0,0}, {1,1}, {2,2}}, /* Q==1 : impossible */ 25020c16b537SWarner Losh {{ 38,130}, {1313, 74}, {2151, 38}}, /* Q == 2 : 12-18% */ 25030c16b537SWarner Losh {{ 448,128}, {1353, 74}, {2238, 41}}, /* Q == 3 : 18-25% */ 25040c16b537SWarner Losh {{ 556,128}, {1353, 74}, {2238, 47}}, /* Q == 4 : 25-32% */ 25050c16b537SWarner Losh {{ 714,128}, {1418, 74}, {2436, 53}}, /* Q == 5 : 32-38% */ 25060c16b537SWarner Losh {{ 883,128}, {1437, 74}, {2464, 61}}, /* Q == 6 : 38-44% */ 25070c16b537SWarner Losh {{ 897,128}, {1515, 75}, {2622, 68}}, /* Q == 7 : 44-50% */ 25080c16b537SWarner Losh {{ 926,128}, {1613, 75}, {2730, 75}}, /* Q == 8 : 50-56% */ 25090c16b537SWarner Losh {{ 947,128}, {1729, 77}, {3359, 77}}, /* Q == 9 : 56-62% */ 25100c16b537SWarner Losh {{1107,128}, {2083, 81}, {4006, 84}}, /* Q ==10 : 62-69% */ 25110c16b537SWarner Losh {{1177,128}, {2379, 87}, {4785, 88}}, /* Q ==11 : 69-75% */ 25120c16b537SWarner Losh {{1242,128}, {2415, 93}, {5155, 84}}, /* Q ==12 : 75-81% */ 25130c16b537SWarner Losh {{1349,128}, {2644,106}, {5260,106}}, /* Q ==13 : 81-87% */ 25140c16b537SWarner Losh {{1455,128}, {2422,124}, {4174,124}}, /* Q ==14 : 87-93% */ 25150c16b537SWarner Losh {{ 722,128}, {1891,145}, {1936,146}}, /* Q ==15 : 93-99% */ 25160c16b537SWarner Losh }; 25170c16b537SWarner Losh 25180c16b537SWarner Losh typedef size_t (*decompressionAlgo)(void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize); 25190c16b537SWarner Losh 25200c16b537SWarner Losh size_t HUFv05_decompress (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize) 25210c16b537SWarner Losh { 25220c16b537SWarner Losh static const decompressionAlgo decompress[3] = { HUFv05_decompress4X2, HUFv05_decompress4X4, NULL }; 25230c16b537SWarner Losh /* estimate decompression time */ 25240c16b537SWarner Losh U32 Q; 25250c16b537SWarner Losh const U32 D256 = (U32)(dstSize >> 8); 25260c16b537SWarner Losh U32 Dtime[3]; 25270c16b537SWarner Losh U32 algoNb = 0; 25280c16b537SWarner Losh int n; 25290c16b537SWarner Losh 25300c16b537SWarner Losh /* validation checks */ 25310c16b537SWarner Losh if (dstSize == 0) return ERROR(dstSize_tooSmall); 25320c16b537SWarner Losh if (cSrcSize >= dstSize) return ERROR(corruption_detected); /* invalid, or not compressed, but not compressed already dealt with */ 25330c16b537SWarner Losh if (cSrcSize == 1) { memset(dst, *(const BYTE*)cSrc, dstSize); return dstSize; } /* RLE */ 25340c16b537SWarner Losh 25350c16b537SWarner Losh /* decoder timing evaluation */ 25360c16b537SWarner Losh Q = (U32)(cSrcSize * 16 / dstSize); /* Q < 16 since dstSize > cSrcSize */ 25370c16b537SWarner Losh for (n=0; n<3; n++) 25380c16b537SWarner Losh Dtime[n] = algoTime[Q][n].tableTime + (algoTime[Q][n].decode256Time * D256); 25390c16b537SWarner Losh 25400c16b537SWarner Losh Dtime[1] += Dtime[1] >> 4; Dtime[2] += Dtime[2] >> 3; /* advantage to algorithms using less memory, for cache eviction */ 25410c16b537SWarner Losh 25420c16b537SWarner Losh if (Dtime[1] < Dtime[0]) algoNb = 1; 25430c16b537SWarner Losh 25440c16b537SWarner Losh return decompress[algoNb](dst, dstSize, cSrc, cSrcSize); 25450c16b537SWarner Losh 254637f1f268SConrad Meyer /* return HUFv05_decompress4X2(dst, dstSize, cSrc, cSrcSize); */ /* multi-streams single-symbol decoding */ 254737f1f268SConrad Meyer /* return HUFv05_decompress4X4(dst, dstSize, cSrc, cSrcSize); */ /* multi-streams double-symbols decoding */ 254837f1f268SConrad Meyer /* return HUFv05_decompress4X6(dst, dstSize, cSrc, cSrcSize); */ /* multi-streams quad-symbols decoding */ 25490c16b537SWarner Losh } 25500c16b537SWarner Losh /* 25510c16b537SWarner Losh zstd - standard compression library 25520c16b537SWarner Losh Copyright (C) 2014-2016, Yann Collet. 25530c16b537SWarner Losh 25540c16b537SWarner Losh BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php) 25550c16b537SWarner Losh 25560c16b537SWarner Losh Redistribution and use in source and binary forms, with or without 25570c16b537SWarner Losh modification, are permitted provided that the following conditions are 25580c16b537SWarner Losh met: 25590c16b537SWarner Losh * Redistributions of source code must retain the above copyright 25600c16b537SWarner Losh notice, this list of conditions and the following disclaimer. 25610c16b537SWarner Losh * Redistributions in binary form must reproduce the above 25620c16b537SWarner Losh copyright notice, this list of conditions and the following disclaimer 25630c16b537SWarner Losh in the documentation and/or other materials provided with the 25640c16b537SWarner Losh distribution. 25650c16b537SWarner Losh THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 25660c16b537SWarner Losh "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 25670c16b537SWarner Losh LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 25680c16b537SWarner Losh A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 25690c16b537SWarner Losh OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 25700c16b537SWarner Losh SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 25710c16b537SWarner Losh LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 25720c16b537SWarner Losh DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 25730c16b537SWarner Losh THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 25740c16b537SWarner Losh (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 25750c16b537SWarner Losh OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 25760c16b537SWarner Losh 25770c16b537SWarner Losh You can contact the author at : 25780c16b537SWarner Losh - zstd source repository : https://github.com/Cyan4973/zstd 25790c16b537SWarner Losh */ 25800c16b537SWarner Losh 25810c16b537SWarner Losh /* *************************************************************** 25820c16b537SWarner Losh * Tuning parameters 25830c16b537SWarner Losh *****************************************************************/ 25840c16b537SWarner Losh /*! 25850c16b537SWarner Losh * HEAPMODE : 25860c16b537SWarner Losh * Select how default decompression function ZSTDv05_decompress() will allocate memory, 25870c16b537SWarner Losh * in memory stack (0), or in memory heap (1, requires malloc()) 25880c16b537SWarner Losh */ 25890c16b537SWarner Losh #ifndef ZSTDv05_HEAPMODE 25900c16b537SWarner Losh # define ZSTDv05_HEAPMODE 1 25910c16b537SWarner Losh #endif 25920c16b537SWarner Losh 25930c16b537SWarner Losh 25940c16b537SWarner Losh /*-******************************************************* 25950c16b537SWarner Losh * Dependencies 25960c16b537SWarner Losh *********************************************************/ 25970c16b537SWarner Losh #include <stdlib.h> /* calloc */ 25980c16b537SWarner Losh #include <string.h> /* memcpy, memmove */ 25990c16b537SWarner Losh #include <stdio.h> /* debug only : printf */ 26000c16b537SWarner Losh 26010c16b537SWarner Losh 26020c16b537SWarner Losh /*-******************************************************* 26030c16b537SWarner Losh * Compiler specifics 26040c16b537SWarner Losh *********************************************************/ 26050c16b537SWarner Losh #ifdef _MSC_VER /* Visual Studio */ 26060c16b537SWarner Losh # include <intrin.h> /* For Visual 2005 */ 26070c16b537SWarner Losh # pragma warning(disable : 4127) /* disable: C4127: conditional expression is constant */ 26080c16b537SWarner Losh # pragma warning(disable : 4324) /* disable: C4324: padded structure */ 26090c16b537SWarner Losh #endif 26100c16b537SWarner Losh 26110c16b537SWarner Losh 26120c16b537SWarner Losh /*-************************************* 26130c16b537SWarner Losh * Local types 26140c16b537SWarner Losh ***************************************/ 26150c16b537SWarner Losh typedef struct 26160c16b537SWarner Losh { 26170c16b537SWarner Losh blockType_t blockType; 26180c16b537SWarner Losh U32 origSize; 26190c16b537SWarner Losh } blockProperties_t; 26200c16b537SWarner Losh 26210c16b537SWarner Losh 26220c16b537SWarner Losh /* ******************************************************* 26230c16b537SWarner Losh * Memory operations 26240c16b537SWarner Losh **********************************************************/ 26250c16b537SWarner Losh static void ZSTDv05_copy4(void* dst, const void* src) { memcpy(dst, src, 4); } 26260c16b537SWarner Losh 26270c16b537SWarner Losh 26280c16b537SWarner Losh /* ************************************* 26290c16b537SWarner Losh * Error Management 26300c16b537SWarner Losh ***************************************/ 26310c16b537SWarner Losh /*! ZSTDv05_isError() : 26320c16b537SWarner Losh * tells if a return value is an error code */ 26330c16b537SWarner Losh unsigned ZSTDv05_isError(size_t code) { return ERR_isError(code); } 26340c16b537SWarner Losh 26350c16b537SWarner Losh 26360c16b537SWarner Losh /*! ZSTDv05_getErrorName() : 26370c16b537SWarner Losh * provides error code string (useful for debugging) */ 26380c16b537SWarner Losh const char* ZSTDv05_getErrorName(size_t code) { return ERR_getErrorName(code); } 26390c16b537SWarner Losh 26400c16b537SWarner Losh 26410c16b537SWarner Losh /* ************************************************************* 26420c16b537SWarner Losh * Context management 26430c16b537SWarner Losh ***************************************************************/ 26440c16b537SWarner Losh typedef enum { ZSTDv05ds_getFrameHeaderSize, ZSTDv05ds_decodeFrameHeader, 26450c16b537SWarner Losh ZSTDv05ds_decodeBlockHeader, ZSTDv05ds_decompressBlock } ZSTDv05_dStage; 26460c16b537SWarner Losh 26470c16b537SWarner Losh struct ZSTDv05_DCtx_s 26480c16b537SWarner Losh { 26490c16b537SWarner Losh FSEv05_DTable LLTable[FSEv05_DTABLE_SIZE_U32(LLFSEv05Log)]; 26500c16b537SWarner Losh FSEv05_DTable OffTable[FSEv05_DTABLE_SIZE_U32(OffFSEv05Log)]; 26510c16b537SWarner Losh FSEv05_DTable MLTable[FSEv05_DTABLE_SIZE_U32(MLFSEv05Log)]; 26520c16b537SWarner Losh unsigned hufTableX4[HUFv05_DTABLE_SIZE(HufLog)]; 26530c16b537SWarner Losh const void* previousDstEnd; 26540c16b537SWarner Losh const void* base; 26550c16b537SWarner Losh const void* vBase; 26560c16b537SWarner Losh const void* dictEnd; 26570c16b537SWarner Losh size_t expected; 26580c16b537SWarner Losh size_t headerSize; 26590c16b537SWarner Losh ZSTDv05_parameters params; 26600c16b537SWarner Losh blockType_t bType; /* used in ZSTDv05_decompressContinue(), to transfer blockType between header decoding and block decoding stages */ 26610c16b537SWarner Losh ZSTDv05_dStage stage; 26620c16b537SWarner Losh U32 flagStaticTables; 26630c16b537SWarner Losh const BYTE* litPtr; 26640c16b537SWarner Losh size_t litSize; 26650c16b537SWarner Losh BYTE litBuffer[BLOCKSIZE + WILDCOPY_OVERLENGTH]; 26660c16b537SWarner Losh BYTE headerBuffer[ZSTDv05_frameHeaderSize_max]; 26670c16b537SWarner Losh }; /* typedef'd to ZSTDv05_DCtx within "zstd_static.h" */ 26680c16b537SWarner Losh 26690f743729SConrad Meyer size_t ZSTDv05_sizeofDCtx (void); /* Hidden declaration */ 26700c16b537SWarner Losh size_t ZSTDv05_sizeofDCtx (void) { return sizeof(ZSTDv05_DCtx); } 26710c16b537SWarner Losh 26720c16b537SWarner Losh size_t ZSTDv05_decompressBegin(ZSTDv05_DCtx* dctx) 26730c16b537SWarner Losh { 26740c16b537SWarner Losh dctx->expected = ZSTDv05_frameHeaderSize_min; 26750c16b537SWarner Losh dctx->stage = ZSTDv05ds_getFrameHeaderSize; 26760c16b537SWarner Losh dctx->previousDstEnd = NULL; 26770c16b537SWarner Losh dctx->base = NULL; 26780c16b537SWarner Losh dctx->vBase = NULL; 26790c16b537SWarner Losh dctx->dictEnd = NULL; 26800c16b537SWarner Losh dctx->hufTableX4[0] = HufLog; 26810c16b537SWarner Losh dctx->flagStaticTables = 0; 26820c16b537SWarner Losh return 0; 26830c16b537SWarner Losh } 26840c16b537SWarner Losh 26850c16b537SWarner Losh ZSTDv05_DCtx* ZSTDv05_createDCtx(void) 26860c16b537SWarner Losh { 26870c16b537SWarner Losh ZSTDv05_DCtx* dctx = (ZSTDv05_DCtx*)malloc(sizeof(ZSTDv05_DCtx)); 26880c16b537SWarner Losh if (dctx==NULL) return NULL; 26890c16b537SWarner Losh ZSTDv05_decompressBegin(dctx); 26900c16b537SWarner Losh return dctx; 26910c16b537SWarner Losh } 26920c16b537SWarner Losh 26930c16b537SWarner Losh size_t ZSTDv05_freeDCtx(ZSTDv05_DCtx* dctx) 26940c16b537SWarner Losh { 26950c16b537SWarner Losh free(dctx); 26960c16b537SWarner Losh return 0; /* reserved as a potential error code in the future */ 26970c16b537SWarner Losh } 26980c16b537SWarner Losh 26990c16b537SWarner Losh void ZSTDv05_copyDCtx(ZSTDv05_DCtx* dstDCtx, const ZSTDv05_DCtx* srcDCtx) 27000c16b537SWarner Losh { 27010c16b537SWarner Losh memcpy(dstDCtx, srcDCtx, 27020c16b537SWarner Losh sizeof(ZSTDv05_DCtx) - (BLOCKSIZE+WILDCOPY_OVERLENGTH + ZSTDv05_frameHeaderSize_max)); /* no need to copy workspace */ 27030c16b537SWarner Losh } 27040c16b537SWarner Losh 27050c16b537SWarner Losh 27060c16b537SWarner Losh /* ************************************************************* 27070c16b537SWarner Losh * Decompression section 27080c16b537SWarner Losh ***************************************************************/ 27090c16b537SWarner Losh 27100c16b537SWarner Losh /* Frame format description 27110c16b537SWarner Losh Frame Header - [ Block Header - Block ] - Frame End 27120c16b537SWarner Losh 1) Frame Header 27130c16b537SWarner Losh - 4 bytes - Magic Number : ZSTDv05_MAGICNUMBER (defined within zstd_internal.h) 27140c16b537SWarner Losh - 1 byte - Window Descriptor 27150c16b537SWarner Losh 2) Block Header 27160c16b537SWarner Losh - 3 bytes, starting with a 2-bits descriptor 27170c16b537SWarner Losh Uncompressed, Compressed, Frame End, unused 27180c16b537SWarner Losh 3) Block 27190c16b537SWarner Losh See Block Format Description 27200c16b537SWarner Losh 4) Frame End 27210c16b537SWarner Losh - 3 bytes, compatible with Block Header 27220c16b537SWarner Losh */ 27230c16b537SWarner Losh 27240c16b537SWarner Losh /* Block format description 27250c16b537SWarner Losh 27260c16b537SWarner Losh Block = Literal Section - Sequences Section 27270c16b537SWarner Losh Prerequisite : size of (compressed) block, maximum size of regenerated data 27280c16b537SWarner Losh 27290c16b537SWarner Losh 1) Literal Section 27300c16b537SWarner Losh 27310c16b537SWarner Losh 1.1) Header : 1-5 bytes 27320c16b537SWarner Losh flags: 2 bits 27330c16b537SWarner Losh 00 compressed by Huff0 27340c16b537SWarner Losh 01 unused 27350c16b537SWarner Losh 10 is Raw (uncompressed) 27360c16b537SWarner Losh 11 is Rle 27370c16b537SWarner Losh Note : using 01 => Huff0 with precomputed table ? 27380c16b537SWarner Losh Note : delta map ? => compressed ? 27390c16b537SWarner Losh 27400c16b537SWarner Losh 1.1.1) Huff0-compressed literal block : 3-5 bytes 27410c16b537SWarner Losh srcSize < 1 KB => 3 bytes (2-2-10-10) => single stream 27420c16b537SWarner Losh srcSize < 1 KB => 3 bytes (2-2-10-10) 27430c16b537SWarner Losh srcSize < 16KB => 4 bytes (2-2-14-14) 27440c16b537SWarner Losh else => 5 bytes (2-2-18-18) 27450c16b537SWarner Losh big endian convention 27460c16b537SWarner Losh 27470c16b537SWarner Losh 1.1.2) Raw (uncompressed) literal block header : 1-3 bytes 27480c16b537SWarner Losh size : 5 bits: (IS_RAW<<6) + (0<<4) + size 27490c16b537SWarner Losh 12 bits: (IS_RAW<<6) + (2<<4) + (size>>8) 27500c16b537SWarner Losh size&255 27510c16b537SWarner Losh 20 bits: (IS_RAW<<6) + (3<<4) + (size>>16) 27520c16b537SWarner Losh size>>8&255 27530c16b537SWarner Losh size&255 27540c16b537SWarner Losh 27550c16b537SWarner Losh 1.1.3) Rle (repeated single byte) literal block header : 1-3 bytes 27560c16b537SWarner Losh size : 5 bits: (IS_RLE<<6) + (0<<4) + size 27570c16b537SWarner Losh 12 bits: (IS_RLE<<6) + (2<<4) + (size>>8) 27580c16b537SWarner Losh size&255 27590c16b537SWarner Losh 20 bits: (IS_RLE<<6) + (3<<4) + (size>>16) 27600c16b537SWarner Losh size>>8&255 27610c16b537SWarner Losh size&255 27620c16b537SWarner Losh 27630c16b537SWarner Losh 1.1.4) Huff0-compressed literal block, using precomputed CTables : 3-5 bytes 27640c16b537SWarner Losh srcSize < 1 KB => 3 bytes (2-2-10-10) => single stream 27650c16b537SWarner Losh srcSize < 1 KB => 3 bytes (2-2-10-10) 27660c16b537SWarner Losh srcSize < 16KB => 4 bytes (2-2-14-14) 27670c16b537SWarner Losh else => 5 bytes (2-2-18-18) 27680c16b537SWarner Losh big endian convention 27690c16b537SWarner Losh 27700c16b537SWarner Losh 1- CTable available (stored into workspace ?) 27710c16b537SWarner Losh 2- Small input (fast heuristic ? Full comparison ? depend on clevel ?) 27720c16b537SWarner Losh 27730c16b537SWarner Losh 27740c16b537SWarner Losh 1.2) Literal block content 27750c16b537SWarner Losh 27760c16b537SWarner Losh 1.2.1) Huff0 block, using sizes from header 27770c16b537SWarner Losh See Huff0 format 27780c16b537SWarner Losh 27790c16b537SWarner Losh 1.2.2) Huff0 block, using prepared table 27800c16b537SWarner Losh 27810c16b537SWarner Losh 1.2.3) Raw content 27820c16b537SWarner Losh 27830c16b537SWarner Losh 1.2.4) single byte 27840c16b537SWarner Losh 27850c16b537SWarner Losh 27860c16b537SWarner Losh 2) Sequences section 27870c16b537SWarner Losh TO DO 27880c16b537SWarner Losh */ 27890c16b537SWarner Losh 27900c16b537SWarner Losh 27910c16b537SWarner Losh /** ZSTDv05_decodeFrameHeader_Part1() : 27920c16b537SWarner Losh * decode the 1st part of the Frame Header, which tells Frame Header size. 27930c16b537SWarner Losh * srcSize must be == ZSTDv05_frameHeaderSize_min. 27940c16b537SWarner Losh * @return : the full size of the Frame Header */ 27950c16b537SWarner Losh static size_t ZSTDv05_decodeFrameHeader_Part1(ZSTDv05_DCtx* zc, const void* src, size_t srcSize) 27960c16b537SWarner Losh { 27970c16b537SWarner Losh U32 magicNumber; 27980c16b537SWarner Losh if (srcSize != ZSTDv05_frameHeaderSize_min) 27990c16b537SWarner Losh return ERROR(srcSize_wrong); 28000c16b537SWarner Losh magicNumber = MEM_readLE32(src); 28010c16b537SWarner Losh if (magicNumber != ZSTDv05_MAGICNUMBER) return ERROR(prefix_unknown); 28020c16b537SWarner Losh zc->headerSize = ZSTDv05_frameHeaderSize_min; 28030c16b537SWarner Losh return zc->headerSize; 28040c16b537SWarner Losh } 28050c16b537SWarner Losh 28060c16b537SWarner Losh 28070c16b537SWarner Losh size_t ZSTDv05_getFrameParams(ZSTDv05_parameters* params, const void* src, size_t srcSize) 28080c16b537SWarner Losh { 28090c16b537SWarner Losh U32 magicNumber; 28100c16b537SWarner Losh if (srcSize < ZSTDv05_frameHeaderSize_min) return ZSTDv05_frameHeaderSize_max; 28110c16b537SWarner Losh magicNumber = MEM_readLE32(src); 28120c16b537SWarner Losh if (magicNumber != ZSTDv05_MAGICNUMBER) return ERROR(prefix_unknown); 28130c16b537SWarner Losh memset(params, 0, sizeof(*params)); 28140c16b537SWarner Losh params->windowLog = (((const BYTE*)src)[4] & 15) + ZSTDv05_WINDOWLOG_ABSOLUTEMIN; 28150c16b537SWarner Losh if ((((const BYTE*)src)[4] >> 4) != 0) return ERROR(frameParameter_unsupported); /* reserved bits */ 28160c16b537SWarner Losh return 0; 28170c16b537SWarner Losh } 28180c16b537SWarner Losh 28190c16b537SWarner Losh /** ZSTDv05_decodeFrameHeader_Part2() : 28200c16b537SWarner Losh * decode the full Frame Header. 28210c16b537SWarner Losh * srcSize must be the size provided by ZSTDv05_decodeFrameHeader_Part1(). 28220c16b537SWarner Losh * @return : 0, or an error code, which can be tested using ZSTDv05_isError() */ 28230c16b537SWarner Losh static size_t ZSTDv05_decodeFrameHeader_Part2(ZSTDv05_DCtx* zc, const void* src, size_t srcSize) 28240c16b537SWarner Losh { 28250c16b537SWarner Losh size_t result; 28260c16b537SWarner Losh if (srcSize != zc->headerSize) 28270c16b537SWarner Losh return ERROR(srcSize_wrong); 28280c16b537SWarner Losh result = ZSTDv05_getFrameParams(&(zc->params), src, srcSize); 28290c16b537SWarner Losh if ((MEM_32bits()) && (zc->params.windowLog > 25)) return ERROR(frameParameter_unsupported); 28300c16b537SWarner Losh return result; 28310c16b537SWarner Losh } 28320c16b537SWarner Losh 28330c16b537SWarner Losh 28340f743729SConrad Meyer static size_t ZSTDv05_getcBlockSize(const void* src, size_t srcSize, blockProperties_t* bpPtr) 28350c16b537SWarner Losh { 28360c16b537SWarner Losh const BYTE* const in = (const BYTE* const)src; 28370c16b537SWarner Losh BYTE headerFlags; 28380c16b537SWarner Losh U32 cSize; 28390c16b537SWarner Losh 28400c16b537SWarner Losh if (srcSize < 3) 28410c16b537SWarner Losh return ERROR(srcSize_wrong); 28420c16b537SWarner Losh 28430c16b537SWarner Losh headerFlags = *in; 28440c16b537SWarner Losh cSize = in[2] + (in[1]<<8) + ((in[0] & 7)<<16); 28450c16b537SWarner Losh 28460c16b537SWarner Losh bpPtr->blockType = (blockType_t)(headerFlags >> 6); 28470c16b537SWarner Losh bpPtr->origSize = (bpPtr->blockType == bt_rle) ? cSize : 0; 28480c16b537SWarner Losh 28490c16b537SWarner Losh if (bpPtr->blockType == bt_end) return 0; 28500c16b537SWarner Losh if (bpPtr->blockType == bt_rle) return 1; 28510c16b537SWarner Losh return cSize; 28520c16b537SWarner Losh } 28530c16b537SWarner Losh 28540c16b537SWarner Losh 28550c16b537SWarner Losh static size_t ZSTDv05_copyRawBlock(void* dst, size_t maxDstSize, const void* src, size_t srcSize) 28560c16b537SWarner Losh { 28570f743729SConrad Meyer if (dst==NULL) return ERROR(dstSize_tooSmall); 28580c16b537SWarner Losh if (srcSize > maxDstSize) return ERROR(dstSize_tooSmall); 28590c16b537SWarner Losh memcpy(dst, src, srcSize); 28600c16b537SWarner Losh return srcSize; 28610c16b537SWarner Losh } 28620c16b537SWarner Losh 28630c16b537SWarner Losh 28640c16b537SWarner Losh /*! ZSTDv05_decodeLiteralsBlock() : 28650c16b537SWarner Losh @return : nb of bytes read from src (< srcSize ) */ 28660f743729SConrad Meyer static size_t ZSTDv05_decodeLiteralsBlock(ZSTDv05_DCtx* dctx, 28670c16b537SWarner Losh const void* src, size_t srcSize) /* note : srcSize < BLOCKSIZE */ 28680c16b537SWarner Losh { 28690c16b537SWarner Losh const BYTE* const istart = (const BYTE*) src; 28700c16b537SWarner Losh 28710c16b537SWarner Losh /* any compressed block with literals segment must be at least this size */ 28720c16b537SWarner Losh if (srcSize < MIN_CBLOCK_SIZE) return ERROR(corruption_detected); 28730c16b537SWarner Losh 28740c16b537SWarner Losh switch(istart[0]>> 6) 28750c16b537SWarner Losh { 28760c16b537SWarner Losh case IS_HUFv05: 28770c16b537SWarner Losh { 28780c16b537SWarner Losh size_t litSize, litCSize, singleStream=0; 28790c16b537SWarner Losh U32 lhSize = ((istart[0]) >> 4) & 3; 28800c16b537SWarner Losh if (srcSize < 5) return ERROR(corruption_detected); /* srcSize >= MIN_CBLOCK_SIZE == 3; here we need up to 5 for case 3 */ 28810c16b537SWarner Losh switch(lhSize) 28820c16b537SWarner Losh { 28830c16b537SWarner Losh case 0: case 1: default: /* note : default is impossible, since lhSize into [0..3] */ 28840c16b537SWarner Losh /* 2 - 2 - 10 - 10 */ 28850c16b537SWarner Losh lhSize=3; 28860c16b537SWarner Losh singleStream = istart[0] & 16; 28870c16b537SWarner Losh litSize = ((istart[0] & 15) << 6) + (istart[1] >> 2); 28880c16b537SWarner Losh litCSize = ((istart[1] & 3) << 8) + istart[2]; 28890c16b537SWarner Losh break; 28900c16b537SWarner Losh case 2: 28910c16b537SWarner Losh /* 2 - 2 - 14 - 14 */ 28920c16b537SWarner Losh lhSize=4; 28930c16b537SWarner Losh litSize = ((istart[0] & 15) << 10) + (istart[1] << 2) + (istart[2] >> 6); 28940c16b537SWarner Losh litCSize = ((istart[2] & 63) << 8) + istart[3]; 28950c16b537SWarner Losh break; 28960c16b537SWarner Losh case 3: 28970c16b537SWarner Losh /* 2 - 2 - 18 - 18 */ 28980c16b537SWarner Losh lhSize=5; 28990c16b537SWarner Losh litSize = ((istart[0] & 15) << 14) + (istart[1] << 6) + (istart[2] >> 2); 29000c16b537SWarner Losh litCSize = ((istart[2] & 3) << 16) + (istart[3] << 8) + istart[4]; 29010c16b537SWarner Losh break; 29020c16b537SWarner Losh } 29030c16b537SWarner Losh if (litSize > BLOCKSIZE) return ERROR(corruption_detected); 29040c16b537SWarner Losh if (litCSize + lhSize > srcSize) return ERROR(corruption_detected); 29050c16b537SWarner Losh 29060c16b537SWarner Losh if (HUFv05_isError(singleStream ? 29070c16b537SWarner Losh HUFv05_decompress1X2(dctx->litBuffer, litSize, istart+lhSize, litCSize) : 29080c16b537SWarner Losh HUFv05_decompress (dctx->litBuffer, litSize, istart+lhSize, litCSize) )) 29090c16b537SWarner Losh return ERROR(corruption_detected); 29100c16b537SWarner Losh 29110c16b537SWarner Losh dctx->litPtr = dctx->litBuffer; 29120c16b537SWarner Losh dctx->litSize = litSize; 29130c16b537SWarner Losh memset(dctx->litBuffer + dctx->litSize, 0, WILDCOPY_OVERLENGTH); 29140c16b537SWarner Losh return litCSize + lhSize; 29150c16b537SWarner Losh } 29160c16b537SWarner Losh case IS_PCH: 29170c16b537SWarner Losh { 29180c16b537SWarner Losh size_t errorCode; 29190c16b537SWarner Losh size_t litSize, litCSize; 29200c16b537SWarner Losh U32 lhSize = ((istart[0]) >> 4) & 3; 29210c16b537SWarner Losh if (lhSize != 1) /* only case supported for now : small litSize, single stream */ 29220c16b537SWarner Losh return ERROR(corruption_detected); 29230c16b537SWarner Losh if (!dctx->flagStaticTables) 29240c16b537SWarner Losh return ERROR(dictionary_corrupted); 29250c16b537SWarner Losh 29260c16b537SWarner Losh /* 2 - 2 - 10 - 10 */ 29270c16b537SWarner Losh lhSize=3; 29280c16b537SWarner Losh litSize = ((istart[0] & 15) << 6) + (istart[1] >> 2); 29290c16b537SWarner Losh litCSize = ((istart[1] & 3) << 8) + istart[2]; 29300c16b537SWarner Losh if (litCSize + lhSize > srcSize) return ERROR(corruption_detected); 29310c16b537SWarner Losh 29320c16b537SWarner Losh errorCode = HUFv05_decompress1X4_usingDTable(dctx->litBuffer, litSize, istart+lhSize, litCSize, dctx->hufTableX4); 29330c16b537SWarner Losh if (HUFv05_isError(errorCode)) return ERROR(corruption_detected); 29340c16b537SWarner Losh 29350c16b537SWarner Losh dctx->litPtr = dctx->litBuffer; 29360c16b537SWarner Losh dctx->litSize = litSize; 29370c16b537SWarner Losh memset(dctx->litBuffer + dctx->litSize, 0, WILDCOPY_OVERLENGTH); 29380c16b537SWarner Losh return litCSize + lhSize; 29390c16b537SWarner Losh } 29400c16b537SWarner Losh case IS_RAW: 29410c16b537SWarner Losh { 29420c16b537SWarner Losh size_t litSize; 29430c16b537SWarner Losh U32 lhSize = ((istart[0]) >> 4) & 3; 29440c16b537SWarner Losh switch(lhSize) 29450c16b537SWarner Losh { 29460c16b537SWarner Losh case 0: case 1: default: /* note : default is impossible, since lhSize into [0..3] */ 29470c16b537SWarner Losh lhSize=1; 29480c16b537SWarner Losh litSize = istart[0] & 31; 29490c16b537SWarner Losh break; 29500c16b537SWarner Losh case 2: 29510c16b537SWarner Losh litSize = ((istart[0] & 15) << 8) + istart[1]; 29520c16b537SWarner Losh break; 29530c16b537SWarner Losh case 3: 29540c16b537SWarner Losh litSize = ((istart[0] & 15) << 16) + (istart[1] << 8) + istart[2]; 29550c16b537SWarner Losh break; 29560c16b537SWarner Losh } 29570c16b537SWarner Losh 29580c16b537SWarner Losh if (lhSize+litSize+WILDCOPY_OVERLENGTH > srcSize) { /* risk reading beyond src buffer with wildcopy */ 29590c16b537SWarner Losh if (litSize+lhSize > srcSize) return ERROR(corruption_detected); 29600c16b537SWarner Losh memcpy(dctx->litBuffer, istart+lhSize, litSize); 29610c16b537SWarner Losh dctx->litPtr = dctx->litBuffer; 29620c16b537SWarner Losh dctx->litSize = litSize; 29630c16b537SWarner Losh memset(dctx->litBuffer + dctx->litSize, 0, WILDCOPY_OVERLENGTH); 29640c16b537SWarner Losh return lhSize+litSize; 29650c16b537SWarner Losh } 29660c16b537SWarner Losh /* direct reference into compressed stream */ 29670c16b537SWarner Losh dctx->litPtr = istart+lhSize; 29680c16b537SWarner Losh dctx->litSize = litSize; 29690c16b537SWarner Losh return lhSize+litSize; 29700c16b537SWarner Losh } 29710c16b537SWarner Losh case IS_RLE: 29720c16b537SWarner Losh { 29730c16b537SWarner Losh size_t litSize; 29740c16b537SWarner Losh U32 lhSize = ((istart[0]) >> 4) & 3; 29750c16b537SWarner Losh switch(lhSize) 29760c16b537SWarner Losh { 29770c16b537SWarner Losh case 0: case 1: default: /* note : default is impossible, since lhSize into [0..3] */ 29780c16b537SWarner Losh lhSize = 1; 29790c16b537SWarner Losh litSize = istart[0] & 31; 29800c16b537SWarner Losh break; 29810c16b537SWarner Losh case 2: 29820c16b537SWarner Losh litSize = ((istart[0] & 15) << 8) + istart[1]; 29830c16b537SWarner Losh break; 29840c16b537SWarner Losh case 3: 29850c16b537SWarner Losh litSize = ((istart[0] & 15) << 16) + (istart[1] << 8) + istart[2]; 29860c16b537SWarner Losh if (srcSize<4) return ERROR(corruption_detected); /* srcSize >= MIN_CBLOCK_SIZE == 3; here we need lhSize+1 = 4 */ 29870c16b537SWarner Losh break; 29880c16b537SWarner Losh } 29890c16b537SWarner Losh if (litSize > BLOCKSIZE) return ERROR(corruption_detected); 29900c16b537SWarner Losh memset(dctx->litBuffer, istart[lhSize], litSize + WILDCOPY_OVERLENGTH); 29910c16b537SWarner Losh dctx->litPtr = dctx->litBuffer; 29920c16b537SWarner Losh dctx->litSize = litSize; 29930c16b537SWarner Losh return lhSize+1; 29940c16b537SWarner Losh } 29950c16b537SWarner Losh default: 29960c16b537SWarner Losh return ERROR(corruption_detected); /* impossible */ 29970c16b537SWarner Losh } 29980c16b537SWarner Losh } 29990c16b537SWarner Losh 30000c16b537SWarner Losh 30010f743729SConrad Meyer static size_t ZSTDv05_decodeSeqHeaders(int* nbSeq, const BYTE** dumpsPtr, size_t* dumpsLengthPtr, 30020c16b537SWarner Losh FSEv05_DTable* DTableLL, FSEv05_DTable* DTableML, FSEv05_DTable* DTableOffb, 30030c16b537SWarner Losh const void* src, size_t srcSize, U32 flagStaticTable) 30040c16b537SWarner Losh { 30050c16b537SWarner Losh const BYTE* const istart = (const BYTE* const)src; 30060c16b537SWarner Losh const BYTE* ip = istart; 30070c16b537SWarner Losh const BYTE* const iend = istart + srcSize; 30080c16b537SWarner Losh U32 LLtype, Offtype, MLtype; 3009a0483764SConrad Meyer unsigned LLlog, Offlog, MLlog; 30100c16b537SWarner Losh size_t dumpsLength; 30110c16b537SWarner Losh 30120c16b537SWarner Losh /* check */ 30130c16b537SWarner Losh if (srcSize < MIN_SEQUENCES_SIZE) 30140c16b537SWarner Losh return ERROR(srcSize_wrong); 30150c16b537SWarner Losh 30160c16b537SWarner Losh /* SeqHead */ 30170c16b537SWarner Losh *nbSeq = *ip++; 30180c16b537SWarner Losh if (*nbSeq==0) return 1; 30190c16b537SWarner Losh if (*nbSeq >= 128) { 30200c16b537SWarner Losh if (ip >= iend) return ERROR(srcSize_wrong); 30210c16b537SWarner Losh *nbSeq = ((nbSeq[0]-128)<<8) + *ip++; 30220c16b537SWarner Losh } 30230c16b537SWarner Losh 30240c16b537SWarner Losh if (ip >= iend) return ERROR(srcSize_wrong); 30250c16b537SWarner Losh LLtype = *ip >> 6; 30260c16b537SWarner Losh Offtype = (*ip >> 4) & 3; 30270c16b537SWarner Losh MLtype = (*ip >> 2) & 3; 30280c16b537SWarner Losh if (*ip & 2) { 30290c16b537SWarner Losh if (ip+3 > iend) return ERROR(srcSize_wrong); 30300c16b537SWarner Losh dumpsLength = ip[2]; 30310c16b537SWarner Losh dumpsLength += ip[1] << 8; 30320c16b537SWarner Losh ip += 3; 30330c16b537SWarner Losh } else { 30340c16b537SWarner Losh if (ip+2 > iend) return ERROR(srcSize_wrong); 30350c16b537SWarner Losh dumpsLength = ip[1]; 30360c16b537SWarner Losh dumpsLength += (ip[0] & 1) << 8; 30370c16b537SWarner Losh ip += 2; 30380c16b537SWarner Losh } 30390c16b537SWarner Losh *dumpsPtr = ip; 30400c16b537SWarner Losh ip += dumpsLength; 30410c16b537SWarner Losh *dumpsLengthPtr = dumpsLength; 30420c16b537SWarner Losh 30430c16b537SWarner Losh /* check */ 30440c16b537SWarner Losh if (ip > iend-3) return ERROR(srcSize_wrong); /* min : all 3 are "raw", hence no header, but at least xxLog bits per type */ 30450c16b537SWarner Losh 30460c16b537SWarner Losh /* sequences */ 30470c16b537SWarner Losh { 30480c16b537SWarner Losh S16 norm[MaxML+1]; /* assumption : MaxML >= MaxLL >= MaxOff */ 30490c16b537SWarner Losh size_t headerSize; 30500c16b537SWarner Losh 30510c16b537SWarner Losh /* Build DTables */ 30520c16b537SWarner Losh switch(LLtype) 30530c16b537SWarner Losh { 30540c16b537SWarner Losh case FSEv05_ENCODING_RLE : 30550c16b537SWarner Losh LLlog = 0; 30560c16b537SWarner Losh FSEv05_buildDTable_rle(DTableLL, *ip++); 30570c16b537SWarner Losh break; 30580c16b537SWarner Losh case FSEv05_ENCODING_RAW : 30590c16b537SWarner Losh LLlog = LLbits; 30600c16b537SWarner Losh FSEv05_buildDTable_raw(DTableLL, LLbits); 30610c16b537SWarner Losh break; 30620c16b537SWarner Losh case FSEv05_ENCODING_STATIC: 30630c16b537SWarner Losh if (!flagStaticTable) return ERROR(corruption_detected); 30640c16b537SWarner Losh break; 30650c16b537SWarner Losh case FSEv05_ENCODING_DYNAMIC : 30660c16b537SWarner Losh default : /* impossible */ 3067a0483764SConrad Meyer { unsigned max = MaxLL; 30680c16b537SWarner Losh headerSize = FSEv05_readNCount(norm, &max, &LLlog, ip, iend-ip); 30690c16b537SWarner Losh if (FSEv05_isError(headerSize)) return ERROR(GENERIC); 30700c16b537SWarner Losh if (LLlog > LLFSEv05Log) return ERROR(corruption_detected); 30710c16b537SWarner Losh ip += headerSize; 30720c16b537SWarner Losh FSEv05_buildDTable(DTableLL, norm, max, LLlog); 30730c16b537SWarner Losh } } 30740c16b537SWarner Losh 30750c16b537SWarner Losh switch(Offtype) 30760c16b537SWarner Losh { 30770c16b537SWarner Losh case FSEv05_ENCODING_RLE : 30780c16b537SWarner Losh Offlog = 0; 30790c16b537SWarner Losh if (ip > iend-2) return ERROR(srcSize_wrong); /* min : "raw", hence no header, but at least xxLog bits */ 30800c16b537SWarner Losh FSEv05_buildDTable_rle(DTableOffb, *ip++ & MaxOff); /* if *ip > MaxOff, data is corrupted */ 30810c16b537SWarner Losh break; 30820c16b537SWarner Losh case FSEv05_ENCODING_RAW : 30830c16b537SWarner Losh Offlog = Offbits; 30840c16b537SWarner Losh FSEv05_buildDTable_raw(DTableOffb, Offbits); 30850c16b537SWarner Losh break; 30860c16b537SWarner Losh case FSEv05_ENCODING_STATIC: 30870c16b537SWarner Losh if (!flagStaticTable) return ERROR(corruption_detected); 30880c16b537SWarner Losh break; 30890c16b537SWarner Losh case FSEv05_ENCODING_DYNAMIC : 30900c16b537SWarner Losh default : /* impossible */ 3091a0483764SConrad Meyer { unsigned max = MaxOff; 30920c16b537SWarner Losh headerSize = FSEv05_readNCount(norm, &max, &Offlog, ip, iend-ip); 30930c16b537SWarner Losh if (FSEv05_isError(headerSize)) return ERROR(GENERIC); 30940c16b537SWarner Losh if (Offlog > OffFSEv05Log) return ERROR(corruption_detected); 30950c16b537SWarner Losh ip += headerSize; 30960c16b537SWarner Losh FSEv05_buildDTable(DTableOffb, norm, max, Offlog); 30970c16b537SWarner Losh } } 30980c16b537SWarner Losh 30990c16b537SWarner Losh switch(MLtype) 31000c16b537SWarner Losh { 31010c16b537SWarner Losh case FSEv05_ENCODING_RLE : 31020c16b537SWarner Losh MLlog = 0; 31030c16b537SWarner Losh if (ip > iend-2) return ERROR(srcSize_wrong); /* min : "raw", hence no header, but at least xxLog bits */ 31040c16b537SWarner Losh FSEv05_buildDTable_rle(DTableML, *ip++); 31050c16b537SWarner Losh break; 31060c16b537SWarner Losh case FSEv05_ENCODING_RAW : 31070c16b537SWarner Losh MLlog = MLbits; 31080c16b537SWarner Losh FSEv05_buildDTable_raw(DTableML, MLbits); 31090c16b537SWarner Losh break; 31100c16b537SWarner Losh case FSEv05_ENCODING_STATIC: 31110c16b537SWarner Losh if (!flagStaticTable) return ERROR(corruption_detected); 31120c16b537SWarner Losh break; 31130c16b537SWarner Losh case FSEv05_ENCODING_DYNAMIC : 31140c16b537SWarner Losh default : /* impossible */ 3115a0483764SConrad Meyer { unsigned max = MaxML; 31160c16b537SWarner Losh headerSize = FSEv05_readNCount(norm, &max, &MLlog, ip, iend-ip); 31170c16b537SWarner Losh if (FSEv05_isError(headerSize)) return ERROR(GENERIC); 31180c16b537SWarner Losh if (MLlog > MLFSEv05Log) return ERROR(corruption_detected); 31190c16b537SWarner Losh ip += headerSize; 31200c16b537SWarner Losh FSEv05_buildDTable(DTableML, norm, max, MLlog); 31210c16b537SWarner Losh } } } 31220c16b537SWarner Losh 31230c16b537SWarner Losh return ip-istart; 31240c16b537SWarner Losh } 31250c16b537SWarner Losh 31260c16b537SWarner Losh 31270c16b537SWarner Losh typedef struct { 31280c16b537SWarner Losh size_t litLength; 31290c16b537SWarner Losh size_t matchLength; 31300c16b537SWarner Losh size_t offset; 31310c16b537SWarner Losh } seq_t; 31320c16b537SWarner Losh 31330c16b537SWarner Losh typedef struct { 31340c16b537SWarner Losh BITv05_DStream_t DStream; 31350c16b537SWarner Losh FSEv05_DState_t stateLL; 31360c16b537SWarner Losh FSEv05_DState_t stateOffb; 31370c16b537SWarner Losh FSEv05_DState_t stateML; 31380c16b537SWarner Losh size_t prevOffset; 31390c16b537SWarner Losh const BYTE* dumps; 31400c16b537SWarner Losh const BYTE* dumpsEnd; 31410c16b537SWarner Losh } seqState_t; 31420c16b537SWarner Losh 31430c16b537SWarner Losh 31440c16b537SWarner Losh 31450c16b537SWarner Losh static void ZSTDv05_decodeSequence(seq_t* seq, seqState_t* seqState) 31460c16b537SWarner Losh { 31470c16b537SWarner Losh size_t litLength; 31480c16b537SWarner Losh size_t prevOffset; 31490c16b537SWarner Losh size_t offset; 31500c16b537SWarner Losh size_t matchLength; 31510c16b537SWarner Losh const BYTE* dumps = seqState->dumps; 31520c16b537SWarner Losh const BYTE* const de = seqState->dumpsEnd; 31530c16b537SWarner Losh 31540c16b537SWarner Losh /* Literal length */ 31550c16b537SWarner Losh litLength = FSEv05_peakSymbol(&(seqState->stateLL)); 31560c16b537SWarner Losh prevOffset = litLength ? seq->offset : seqState->prevOffset; 31570c16b537SWarner Losh if (litLength == MaxLL) { 31584d3f1eafSConrad Meyer const U32 add = *dumps++; 31590c16b537SWarner Losh if (add < 255) litLength += add; 31604d3f1eafSConrad Meyer else if (dumps + 2 <= de) { 31614d3f1eafSConrad Meyer litLength = MEM_readLE16(dumps); 31624d3f1eafSConrad Meyer dumps += 2; 31634d3f1eafSConrad Meyer if ((litLength & 1) && dumps < de) { 31644d3f1eafSConrad Meyer litLength += *dumps << 16; 31654d3f1eafSConrad Meyer dumps += 1; 31660c16b537SWarner Losh } 31674d3f1eafSConrad Meyer litLength>>=1; 31684d3f1eafSConrad Meyer } 31690c16b537SWarner Losh if (dumps >= de) { dumps = de-1; } /* late correction, to avoid read overflow (data is now corrupted anyway) */ 31700c16b537SWarner Losh } 31710c16b537SWarner Losh 31720c16b537SWarner Losh /* Offset */ 31730c16b537SWarner Losh { 31740c16b537SWarner Losh static const U32 offsetPrefix[MaxOff+1] = { 31750c16b537SWarner Losh 1 /*fake*/, 1, 2, 4, 8, 16, 32, 64, 128, 256, 31760c16b537SWarner Losh 512, 1024, 2048, 4096, 8192, 16384, 32768, 65536, 131072, 262144, 31770c16b537SWarner Losh 524288, 1048576, 2097152, 4194304, 8388608, 16777216, 33554432, /*fake*/ 1, 1, 1, 1, 1 }; 31780c16b537SWarner Losh U32 offsetCode = FSEv05_peakSymbol(&(seqState->stateOffb)); /* <= maxOff, by table construction */ 31790c16b537SWarner Losh U32 nbBits = offsetCode - 1; 31800c16b537SWarner Losh if (offsetCode==0) nbBits = 0; /* cmove */ 31810c16b537SWarner Losh offset = offsetPrefix[offsetCode] + BITv05_readBits(&(seqState->DStream), nbBits); 31820c16b537SWarner Losh if (MEM_32bits()) BITv05_reloadDStream(&(seqState->DStream)); 31830c16b537SWarner Losh if (offsetCode==0) offset = prevOffset; /* repcode, cmove */ 31840c16b537SWarner Losh if (offsetCode | !litLength) seqState->prevOffset = seq->offset; /* cmove */ 31850c16b537SWarner Losh FSEv05_decodeSymbol(&(seqState->stateOffb), &(seqState->DStream)); /* update */ 31860c16b537SWarner Losh } 31870c16b537SWarner Losh 31880c16b537SWarner Losh /* Literal length update */ 31890c16b537SWarner Losh FSEv05_decodeSymbol(&(seqState->stateLL), &(seqState->DStream)); /* update */ 31900c16b537SWarner Losh if (MEM_32bits()) BITv05_reloadDStream(&(seqState->DStream)); 31910c16b537SWarner Losh 31920c16b537SWarner Losh /* MatchLength */ 31930c16b537SWarner Losh matchLength = FSEv05_decodeSymbol(&(seqState->stateML), &(seqState->DStream)); 31940c16b537SWarner Losh if (matchLength == MaxML) { 31954d3f1eafSConrad Meyer const U32 add = dumps<de ? *dumps++ : 0; 31960c16b537SWarner Losh if (add < 255) matchLength += add; 31974d3f1eafSConrad Meyer else if (dumps + 2 <= de) { 31984d3f1eafSConrad Meyer matchLength = MEM_readLE16(dumps); 31994d3f1eafSConrad Meyer dumps += 2; 32004d3f1eafSConrad Meyer if ((matchLength & 1) && dumps < de) { 32014d3f1eafSConrad Meyer matchLength += *dumps << 16; 32024d3f1eafSConrad Meyer dumps += 1; 32030c16b537SWarner Losh } 32044d3f1eafSConrad Meyer matchLength >>= 1; 32054d3f1eafSConrad Meyer } 32060c16b537SWarner Losh if (dumps >= de) { dumps = de-1; } /* late correction, to avoid read overflow (data is now corrupted anyway) */ 32070c16b537SWarner Losh } 32080c16b537SWarner Losh matchLength += MINMATCH; 32090c16b537SWarner Losh 32100c16b537SWarner Losh /* save result */ 32110c16b537SWarner Losh seq->litLength = litLength; 32120c16b537SWarner Losh seq->offset = offset; 32130c16b537SWarner Losh seq->matchLength = matchLength; 32140c16b537SWarner Losh seqState->dumps = dumps; 32150c16b537SWarner Losh 32160c16b537SWarner Losh #if 0 /* debug */ 32170c16b537SWarner Losh { 32180c16b537SWarner Losh static U64 totalDecoded = 0; 32190c16b537SWarner Losh printf("pos %6u : %3u literals & match %3u bytes at distance %6u \n", 32200c16b537SWarner Losh (U32)(totalDecoded), (U32)litLength, (U32)matchLength, (U32)offset); 32210c16b537SWarner Losh totalDecoded += litLength + matchLength; 32220c16b537SWarner Losh } 32230c16b537SWarner Losh #endif 32240c16b537SWarner Losh } 32250c16b537SWarner Losh 32260c16b537SWarner Losh 32270c16b537SWarner Losh static size_t ZSTDv05_execSequence(BYTE* op, 32280c16b537SWarner Losh BYTE* const oend, seq_t sequence, 32290c16b537SWarner Losh const BYTE** litPtr, const BYTE* const litLimit, 32300c16b537SWarner Losh const BYTE* const base, const BYTE* const vBase, const BYTE* const dictEnd) 32310c16b537SWarner Losh { 32320c16b537SWarner Losh static const int dec32table[] = { 0, 1, 2, 1, 4, 4, 4, 4 }; /* added */ 32332b9c00cbSConrad Meyer static const int dec64table[] = { 8, 8, 8, 7, 8, 9,10,11 }; /* subtracted */ 32340c16b537SWarner Losh BYTE* const oLitEnd = op + sequence.litLength; 32350c16b537SWarner Losh const size_t sequenceLength = sequence.litLength + sequence.matchLength; 32360c16b537SWarner Losh BYTE* const oMatchEnd = op + sequenceLength; /* risk : address space overflow (32-bits) */ 32370c16b537SWarner Losh BYTE* const oend_8 = oend-8; 32380c16b537SWarner Losh const BYTE* const litEnd = *litPtr + sequence.litLength; 32390c16b537SWarner Losh const BYTE* match = oLitEnd - sequence.offset; 32400c16b537SWarner Losh 32410c16b537SWarner Losh /* check */ 32420c16b537SWarner Losh if (oLitEnd > oend_8) return ERROR(dstSize_tooSmall); /* last match must start at a minimum distance of 8 from oend */ 32430c16b537SWarner Losh if (oMatchEnd > oend) return ERROR(dstSize_tooSmall); /* overwrite beyond dst buffer */ 32440c16b537SWarner Losh if (litEnd > litLimit) return ERROR(corruption_detected); /* risk read beyond lit buffer */ 32450c16b537SWarner Losh 32460c16b537SWarner Losh /* copy Literals */ 32470c16b537SWarner Losh ZSTDv05_wildcopy(op, *litPtr, sequence.litLength); /* note : oLitEnd <= oend-8 : no risk of overwrite beyond oend */ 32480c16b537SWarner Losh op = oLitEnd; 32490c16b537SWarner Losh *litPtr = litEnd; /* update for next sequence */ 32500c16b537SWarner Losh 32510c16b537SWarner Losh /* copy Match */ 32520c16b537SWarner Losh if (sequence.offset > (size_t)(oLitEnd - base)) { 32530c16b537SWarner Losh /* offset beyond prefix */ 32540c16b537SWarner Losh if (sequence.offset > (size_t)(oLitEnd - vBase)) 32550c16b537SWarner Losh return ERROR(corruption_detected); 32560c16b537SWarner Losh match = dictEnd - (base-match); 32570c16b537SWarner Losh if (match + sequence.matchLength <= dictEnd) { 32580c16b537SWarner Losh memmove(oLitEnd, match, sequence.matchLength); 32590c16b537SWarner Losh return sequenceLength; 32600c16b537SWarner Losh } 32610c16b537SWarner Losh /* span extDict & currentPrefixSegment */ 32620c16b537SWarner Losh { 32630c16b537SWarner Losh size_t length1 = dictEnd - match; 32640c16b537SWarner Losh memmove(oLitEnd, match, length1); 32650c16b537SWarner Losh op = oLitEnd + length1; 32660c16b537SWarner Losh sequence.matchLength -= length1; 32670c16b537SWarner Losh match = base; 32680c16b537SWarner Losh if (op > oend_8 || sequence.matchLength < MINMATCH) { 32690c16b537SWarner Losh while (op < oMatchEnd) *op++ = *match++; 32700c16b537SWarner Losh return sequenceLength; 32710c16b537SWarner Losh } 32720c16b537SWarner Losh } } 32730c16b537SWarner Losh /* Requirement: op <= oend_8 */ 32740c16b537SWarner Losh 32750c16b537SWarner Losh /* match within prefix */ 32760c16b537SWarner Losh if (sequence.offset < 8) { 32770c16b537SWarner Losh /* close range match, overlap */ 32780c16b537SWarner Losh const int sub2 = dec64table[sequence.offset]; 32790c16b537SWarner Losh op[0] = match[0]; 32800c16b537SWarner Losh op[1] = match[1]; 32810c16b537SWarner Losh op[2] = match[2]; 32820c16b537SWarner Losh op[3] = match[3]; 32830c16b537SWarner Losh match += dec32table[sequence.offset]; 32840c16b537SWarner Losh ZSTDv05_copy4(op+4, match); 32850c16b537SWarner Losh match -= sub2; 32860c16b537SWarner Losh } else { 32870c16b537SWarner Losh ZSTDv05_copy8(op, match); 32880c16b537SWarner Losh } 32890c16b537SWarner Losh op += 8; match += 8; 32900c16b537SWarner Losh 32910c16b537SWarner Losh if (oMatchEnd > oend-(16-MINMATCH)) { 32920c16b537SWarner Losh if (op < oend_8) { 32930c16b537SWarner Losh ZSTDv05_wildcopy(op, match, oend_8 - op); 32940c16b537SWarner Losh match += oend_8 - op; 32950c16b537SWarner Losh op = oend_8; 32960c16b537SWarner Losh } 32970c16b537SWarner Losh while (op < oMatchEnd) 32980c16b537SWarner Losh *op++ = *match++; 32990c16b537SWarner Losh } else { 33000c16b537SWarner Losh ZSTDv05_wildcopy(op, match, (ptrdiff_t)sequence.matchLength-8); /* works even if matchLength < 8 */ 33010c16b537SWarner Losh } 33020c16b537SWarner Losh return sequenceLength; 33030c16b537SWarner Losh } 33040c16b537SWarner Losh 33050c16b537SWarner Losh 33060c16b537SWarner Losh static size_t ZSTDv05_decompressSequences( 33070c16b537SWarner Losh ZSTDv05_DCtx* dctx, 33080c16b537SWarner Losh void* dst, size_t maxDstSize, 33090c16b537SWarner Losh const void* seqStart, size_t seqSize) 33100c16b537SWarner Losh { 33110c16b537SWarner Losh const BYTE* ip = (const BYTE*)seqStart; 33120c16b537SWarner Losh const BYTE* const iend = ip + seqSize; 33130c16b537SWarner Losh BYTE* const ostart = (BYTE* const)dst; 33140c16b537SWarner Losh BYTE* op = ostart; 33150c16b537SWarner Losh BYTE* const oend = ostart + maxDstSize; 33160f743729SConrad Meyer size_t errorCode, dumpsLength=0; 33170c16b537SWarner Losh const BYTE* litPtr = dctx->litPtr; 33180c16b537SWarner Losh const BYTE* const litEnd = litPtr + dctx->litSize; 33190f743729SConrad Meyer int nbSeq=0; 33200f743729SConrad Meyer const BYTE* dumps = NULL; 3321a0483764SConrad Meyer unsigned* DTableLL = dctx->LLTable; 3322a0483764SConrad Meyer unsigned* DTableML = dctx->MLTable; 3323a0483764SConrad Meyer unsigned* DTableOffb = dctx->OffTable; 33240c16b537SWarner Losh const BYTE* const base = (const BYTE*) (dctx->base); 33250c16b537SWarner Losh const BYTE* const vBase = (const BYTE*) (dctx->vBase); 33260c16b537SWarner Losh const BYTE* const dictEnd = (const BYTE*) (dctx->dictEnd); 33270c16b537SWarner Losh 33280c16b537SWarner Losh /* Build Decoding Tables */ 33290c16b537SWarner Losh errorCode = ZSTDv05_decodeSeqHeaders(&nbSeq, &dumps, &dumpsLength, 33300c16b537SWarner Losh DTableLL, DTableML, DTableOffb, 33310c16b537SWarner Losh ip, seqSize, dctx->flagStaticTables); 33320c16b537SWarner Losh if (ZSTDv05_isError(errorCode)) return errorCode; 33330c16b537SWarner Losh ip += errorCode; 33340c16b537SWarner Losh 33350c16b537SWarner Losh /* Regen sequences */ 33360c16b537SWarner Losh if (nbSeq) { 33370c16b537SWarner Losh seq_t sequence; 33380c16b537SWarner Losh seqState_t seqState; 33390c16b537SWarner Losh 33400c16b537SWarner Losh memset(&sequence, 0, sizeof(sequence)); 33410c16b537SWarner Losh sequence.offset = REPCODE_STARTVALUE; 33420c16b537SWarner Losh seqState.dumps = dumps; 33430c16b537SWarner Losh seqState.dumpsEnd = dumps + dumpsLength; 33440c16b537SWarner Losh seqState.prevOffset = REPCODE_STARTVALUE; 33450c16b537SWarner Losh errorCode = BITv05_initDStream(&(seqState.DStream), ip, iend-ip); 33460c16b537SWarner Losh if (ERR_isError(errorCode)) return ERROR(corruption_detected); 33470c16b537SWarner Losh FSEv05_initDState(&(seqState.stateLL), &(seqState.DStream), DTableLL); 33480c16b537SWarner Losh FSEv05_initDState(&(seqState.stateOffb), &(seqState.DStream), DTableOffb); 33490c16b537SWarner Losh FSEv05_initDState(&(seqState.stateML), &(seqState.DStream), DTableML); 33500c16b537SWarner Losh 33510c16b537SWarner Losh for ( ; (BITv05_reloadDStream(&(seqState.DStream)) <= BITv05_DStream_completed) && nbSeq ; ) { 33520c16b537SWarner Losh size_t oneSeqSize; 33530c16b537SWarner Losh nbSeq--; 33540c16b537SWarner Losh ZSTDv05_decodeSequence(&sequence, &seqState); 33550c16b537SWarner Losh oneSeqSize = ZSTDv05_execSequence(op, oend, sequence, &litPtr, litEnd, base, vBase, dictEnd); 33560c16b537SWarner Losh if (ZSTDv05_isError(oneSeqSize)) return oneSeqSize; 33570c16b537SWarner Losh op += oneSeqSize; 33580c16b537SWarner Losh } 33590c16b537SWarner Losh 33600c16b537SWarner Losh /* check if reached exact end */ 33610c16b537SWarner Losh if (nbSeq) return ERROR(corruption_detected); 33620c16b537SWarner Losh } 33630c16b537SWarner Losh 33640c16b537SWarner Losh /* last literal segment */ 33650c16b537SWarner Losh { 33660c16b537SWarner Losh size_t lastLLSize = litEnd - litPtr; 33670c16b537SWarner Losh if (litPtr > litEnd) return ERROR(corruption_detected); /* too many literals already used */ 33680c16b537SWarner Losh if (op+lastLLSize > oend) return ERROR(dstSize_tooSmall); 336937f1f268SConrad Meyer if (lastLLSize > 0) { 33700c16b537SWarner Losh memcpy(op, litPtr, lastLLSize); 33710c16b537SWarner Losh op += lastLLSize; 33720c16b537SWarner Losh } 337337f1f268SConrad Meyer } 33740c16b537SWarner Losh 33750c16b537SWarner Losh return op-ostart; 33760c16b537SWarner Losh } 33770c16b537SWarner Losh 33780c16b537SWarner Losh 33790c16b537SWarner Losh static void ZSTDv05_checkContinuity(ZSTDv05_DCtx* dctx, const void* dst) 33800c16b537SWarner Losh { 33810c16b537SWarner Losh if (dst != dctx->previousDstEnd) { /* not contiguous */ 33820c16b537SWarner Losh dctx->dictEnd = dctx->previousDstEnd; 33830c16b537SWarner Losh dctx->vBase = (const char*)dst - ((const char*)(dctx->previousDstEnd) - (const char*)(dctx->base)); 33840c16b537SWarner Losh dctx->base = dst; 33850c16b537SWarner Losh dctx->previousDstEnd = dst; 33860c16b537SWarner Losh } 33870c16b537SWarner Losh } 33880c16b537SWarner Losh 33890c16b537SWarner Losh 33900c16b537SWarner Losh static size_t ZSTDv05_decompressBlock_internal(ZSTDv05_DCtx* dctx, 33910c16b537SWarner Losh void* dst, size_t dstCapacity, 33920c16b537SWarner Losh const void* src, size_t srcSize) 33930c16b537SWarner Losh { /* blockType == blockCompressed */ 33940c16b537SWarner Losh const BYTE* ip = (const BYTE*)src; 33950c16b537SWarner Losh size_t litCSize; 33960c16b537SWarner Losh 33970c16b537SWarner Losh if (srcSize >= BLOCKSIZE) return ERROR(srcSize_wrong); 33980c16b537SWarner Losh 33990c16b537SWarner Losh /* Decode literals sub-block */ 34000c16b537SWarner Losh litCSize = ZSTDv05_decodeLiteralsBlock(dctx, src, srcSize); 34010c16b537SWarner Losh if (ZSTDv05_isError(litCSize)) return litCSize; 34020c16b537SWarner Losh ip += litCSize; 34030c16b537SWarner Losh srcSize -= litCSize; 34040c16b537SWarner Losh 34050c16b537SWarner Losh return ZSTDv05_decompressSequences(dctx, dst, dstCapacity, ip, srcSize); 34060c16b537SWarner Losh } 34070c16b537SWarner Losh 34080c16b537SWarner Losh 34090c16b537SWarner Losh size_t ZSTDv05_decompressBlock(ZSTDv05_DCtx* dctx, 34100c16b537SWarner Losh void* dst, size_t dstCapacity, 34110c16b537SWarner Losh const void* src, size_t srcSize) 34120c16b537SWarner Losh { 34130c16b537SWarner Losh ZSTDv05_checkContinuity(dctx, dst); 34140c16b537SWarner Losh return ZSTDv05_decompressBlock_internal(dctx, dst, dstCapacity, src, srcSize); 34150c16b537SWarner Losh } 34160c16b537SWarner Losh 34170c16b537SWarner Losh 34180c16b537SWarner Losh /*! ZSTDv05_decompress_continueDCtx 34190c16b537SWarner Losh * dctx must have been properly initialized */ 34200c16b537SWarner Losh static size_t ZSTDv05_decompress_continueDCtx(ZSTDv05_DCtx* dctx, 34210c16b537SWarner Losh void* dst, size_t maxDstSize, 34220c16b537SWarner Losh const void* src, size_t srcSize) 34230c16b537SWarner Losh { 34240c16b537SWarner Losh const BYTE* ip = (const BYTE*)src; 34250c16b537SWarner Losh const BYTE* iend = ip + srcSize; 34260c16b537SWarner Losh BYTE* const ostart = (BYTE* const)dst; 34270c16b537SWarner Losh BYTE* op = ostart; 34280c16b537SWarner Losh BYTE* const oend = ostart + maxDstSize; 34290c16b537SWarner Losh size_t remainingSize = srcSize; 34300c16b537SWarner Losh blockProperties_t blockProperties; 34310f743729SConrad Meyer memset(&blockProperties, 0, sizeof(blockProperties)); 34320c16b537SWarner Losh 34330c16b537SWarner Losh /* Frame Header */ 34340f743729SConrad Meyer { size_t frameHeaderSize; 34350c16b537SWarner Losh if (srcSize < ZSTDv05_frameHeaderSize_min+ZSTDv05_blockHeaderSize) return ERROR(srcSize_wrong); 34360c16b537SWarner Losh frameHeaderSize = ZSTDv05_decodeFrameHeader_Part1(dctx, src, ZSTDv05_frameHeaderSize_min); 34370c16b537SWarner Losh if (ZSTDv05_isError(frameHeaderSize)) return frameHeaderSize; 34380c16b537SWarner Losh if (srcSize < frameHeaderSize+ZSTDv05_blockHeaderSize) return ERROR(srcSize_wrong); 34390c16b537SWarner Losh ip += frameHeaderSize; remainingSize -= frameHeaderSize; 34400c16b537SWarner Losh frameHeaderSize = ZSTDv05_decodeFrameHeader_Part2(dctx, src, frameHeaderSize); 34410c16b537SWarner Losh if (ZSTDv05_isError(frameHeaderSize)) return frameHeaderSize; 34420c16b537SWarner Losh } 34430c16b537SWarner Losh 34440c16b537SWarner Losh /* Loop on each block */ 34450c16b537SWarner Losh while (1) 34460c16b537SWarner Losh { 34470c16b537SWarner Losh size_t decodedSize=0; 34480c16b537SWarner Losh size_t cBlockSize = ZSTDv05_getcBlockSize(ip, iend-ip, &blockProperties); 34490c16b537SWarner Losh if (ZSTDv05_isError(cBlockSize)) return cBlockSize; 34500c16b537SWarner Losh 34510c16b537SWarner Losh ip += ZSTDv05_blockHeaderSize; 34520c16b537SWarner Losh remainingSize -= ZSTDv05_blockHeaderSize; 34530c16b537SWarner Losh if (cBlockSize > remainingSize) return ERROR(srcSize_wrong); 34540c16b537SWarner Losh 34550c16b537SWarner Losh switch(blockProperties.blockType) 34560c16b537SWarner Losh { 34570c16b537SWarner Losh case bt_compressed: 34580c16b537SWarner Losh decodedSize = ZSTDv05_decompressBlock_internal(dctx, op, oend-op, ip, cBlockSize); 34590c16b537SWarner Losh break; 34600c16b537SWarner Losh case bt_raw : 34610c16b537SWarner Losh decodedSize = ZSTDv05_copyRawBlock(op, oend-op, ip, cBlockSize); 34620c16b537SWarner Losh break; 34630c16b537SWarner Losh case bt_rle : 34640c16b537SWarner Losh return ERROR(GENERIC); /* not yet supported */ 34650c16b537SWarner Losh break; 34660c16b537SWarner Losh case bt_end : 34670c16b537SWarner Losh /* end of frame */ 34680c16b537SWarner Losh if (remainingSize) return ERROR(srcSize_wrong); 34690c16b537SWarner Losh break; 34700c16b537SWarner Losh default: 34710c16b537SWarner Losh return ERROR(GENERIC); /* impossible */ 34720c16b537SWarner Losh } 34730c16b537SWarner Losh if (cBlockSize == 0) break; /* bt_end */ 34740c16b537SWarner Losh 34750c16b537SWarner Losh if (ZSTDv05_isError(decodedSize)) return decodedSize; 34760c16b537SWarner Losh op += decodedSize; 34770c16b537SWarner Losh ip += cBlockSize; 34780c16b537SWarner Losh remainingSize -= cBlockSize; 34790c16b537SWarner Losh } 34800c16b537SWarner Losh 34810c16b537SWarner Losh return op-ostart; 34820c16b537SWarner Losh } 34830c16b537SWarner Losh 34840c16b537SWarner Losh 34850c16b537SWarner Losh size_t ZSTDv05_decompress_usingPreparedDCtx(ZSTDv05_DCtx* dctx, const ZSTDv05_DCtx* refDCtx, 34860c16b537SWarner Losh void* dst, size_t maxDstSize, 34870c16b537SWarner Losh const void* src, size_t srcSize) 34880c16b537SWarner Losh { 34890c16b537SWarner Losh ZSTDv05_copyDCtx(dctx, refDCtx); 34900c16b537SWarner Losh ZSTDv05_checkContinuity(dctx, dst); 34910c16b537SWarner Losh return ZSTDv05_decompress_continueDCtx(dctx, dst, maxDstSize, src, srcSize); 34920c16b537SWarner Losh } 34930c16b537SWarner Losh 34940c16b537SWarner Losh 34950c16b537SWarner Losh size_t ZSTDv05_decompress_usingDict(ZSTDv05_DCtx* dctx, 34960c16b537SWarner Losh void* dst, size_t maxDstSize, 34970c16b537SWarner Losh const void* src, size_t srcSize, 34980c16b537SWarner Losh const void* dict, size_t dictSize) 34990c16b537SWarner Losh { 35000c16b537SWarner Losh ZSTDv05_decompressBegin_usingDict(dctx, dict, dictSize); 35010c16b537SWarner Losh ZSTDv05_checkContinuity(dctx, dst); 35020c16b537SWarner Losh return ZSTDv05_decompress_continueDCtx(dctx, dst, maxDstSize, src, srcSize); 35030c16b537SWarner Losh } 35040c16b537SWarner Losh 35050c16b537SWarner Losh 35060c16b537SWarner Losh size_t ZSTDv05_decompressDCtx(ZSTDv05_DCtx* dctx, void* dst, size_t maxDstSize, const void* src, size_t srcSize) 35070c16b537SWarner Losh { 35080c16b537SWarner Losh return ZSTDv05_decompress_usingDict(dctx, dst, maxDstSize, src, srcSize, NULL, 0); 35090c16b537SWarner Losh } 35100c16b537SWarner Losh 35110c16b537SWarner Losh size_t ZSTDv05_decompress(void* dst, size_t maxDstSize, const void* src, size_t srcSize) 35120c16b537SWarner Losh { 35130c16b537SWarner Losh #if defined(ZSTDv05_HEAPMODE) && (ZSTDv05_HEAPMODE==1) 35140c16b537SWarner Losh size_t regenSize; 35150c16b537SWarner Losh ZSTDv05_DCtx* dctx = ZSTDv05_createDCtx(); 35160c16b537SWarner Losh if (dctx==NULL) return ERROR(memory_allocation); 35170c16b537SWarner Losh regenSize = ZSTDv05_decompressDCtx(dctx, dst, maxDstSize, src, srcSize); 35180c16b537SWarner Losh ZSTDv05_freeDCtx(dctx); 35190c16b537SWarner Losh return regenSize; 35200c16b537SWarner Losh #else 35210c16b537SWarner Losh ZSTDv05_DCtx dctx; 35220c16b537SWarner Losh return ZSTDv05_decompressDCtx(&dctx, dst, maxDstSize, src, srcSize); 35230c16b537SWarner Losh #endif 35240c16b537SWarner Losh } 35250c16b537SWarner Losh 35262b9c00cbSConrad Meyer /* ZSTD_errorFrameSizeInfoLegacy() : 35272b9c00cbSConrad Meyer assumes `cSize` and `dBound` are _not_ NULL */ 35282b9c00cbSConrad Meyer static void ZSTD_errorFrameSizeInfoLegacy(size_t* cSize, unsigned long long* dBound, size_t ret) 35292b9c00cbSConrad Meyer { 35302b9c00cbSConrad Meyer *cSize = ret; 35312b9c00cbSConrad Meyer *dBound = ZSTD_CONTENTSIZE_ERROR; 35322b9c00cbSConrad Meyer } 35332b9c00cbSConrad Meyer 35342b9c00cbSConrad Meyer void ZSTDv05_findFrameSizeInfoLegacy(const void *src, size_t srcSize, size_t* cSize, unsigned long long* dBound) 35350c16b537SWarner Losh { 35360c16b537SWarner Losh const BYTE* ip = (const BYTE*)src; 35370c16b537SWarner Losh size_t remainingSize = srcSize; 35382b9c00cbSConrad Meyer size_t nbBlocks = 0; 35390c16b537SWarner Losh blockProperties_t blockProperties; 35400c16b537SWarner Losh 35410c16b537SWarner Losh /* Frame Header */ 35422b9c00cbSConrad Meyer if (srcSize < ZSTDv05_frameHeaderSize_min) { 35432b9c00cbSConrad Meyer ZSTD_errorFrameSizeInfoLegacy(cSize, dBound, ERROR(srcSize_wrong)); 35442b9c00cbSConrad Meyer return; 35452b9c00cbSConrad Meyer } 35462b9c00cbSConrad Meyer if (MEM_readLE32(src) != ZSTDv05_MAGICNUMBER) { 35472b9c00cbSConrad Meyer ZSTD_errorFrameSizeInfoLegacy(cSize, dBound, ERROR(prefix_unknown)); 35482b9c00cbSConrad Meyer return; 35492b9c00cbSConrad Meyer } 35500c16b537SWarner Losh ip += ZSTDv05_frameHeaderSize_min; remainingSize -= ZSTDv05_frameHeaderSize_min; 35510c16b537SWarner Losh 35520c16b537SWarner Losh /* Loop on each block */ 35530c16b537SWarner Losh while (1) 35540c16b537SWarner Losh { 35550c16b537SWarner Losh size_t cBlockSize = ZSTDv05_getcBlockSize(ip, remainingSize, &blockProperties); 35562b9c00cbSConrad Meyer if (ZSTDv05_isError(cBlockSize)) { 35572b9c00cbSConrad Meyer ZSTD_errorFrameSizeInfoLegacy(cSize, dBound, cBlockSize); 35582b9c00cbSConrad Meyer return; 35592b9c00cbSConrad Meyer } 35600c16b537SWarner Losh 35610c16b537SWarner Losh ip += ZSTDv05_blockHeaderSize; 35620c16b537SWarner Losh remainingSize -= ZSTDv05_blockHeaderSize; 35632b9c00cbSConrad Meyer if (cBlockSize > remainingSize) { 35642b9c00cbSConrad Meyer ZSTD_errorFrameSizeInfoLegacy(cSize, dBound, ERROR(srcSize_wrong)); 35652b9c00cbSConrad Meyer return; 35662b9c00cbSConrad Meyer } 35670c16b537SWarner Losh 35680c16b537SWarner Losh if (cBlockSize == 0) break; /* bt_end */ 35690c16b537SWarner Losh 35700c16b537SWarner Losh ip += cBlockSize; 35710c16b537SWarner Losh remainingSize -= cBlockSize; 35722b9c00cbSConrad Meyer nbBlocks++; 35730c16b537SWarner Losh } 35740c16b537SWarner Losh 35752b9c00cbSConrad Meyer *cSize = ip - (const BYTE*)src; 35762b9c00cbSConrad Meyer *dBound = nbBlocks * BLOCKSIZE; 35770c16b537SWarner Losh } 35780c16b537SWarner Losh 35790c16b537SWarner Losh /* ****************************** 35800c16b537SWarner Losh * Streaming Decompression API 35810c16b537SWarner Losh ********************************/ 35820c16b537SWarner Losh size_t ZSTDv05_nextSrcSizeToDecompress(ZSTDv05_DCtx* dctx) 35830c16b537SWarner Losh { 35840c16b537SWarner Losh return dctx->expected; 35850c16b537SWarner Losh } 35860c16b537SWarner Losh 35870c16b537SWarner Losh size_t ZSTDv05_decompressContinue(ZSTDv05_DCtx* dctx, void* dst, size_t maxDstSize, const void* src, size_t srcSize) 35880c16b537SWarner Losh { 35890c16b537SWarner Losh /* Sanity check */ 35900c16b537SWarner Losh if (srcSize != dctx->expected) return ERROR(srcSize_wrong); 35910c16b537SWarner Losh ZSTDv05_checkContinuity(dctx, dst); 35920c16b537SWarner Losh 35930c16b537SWarner Losh /* Decompress : frame header; part 1 */ 35940c16b537SWarner Losh switch (dctx->stage) 35950c16b537SWarner Losh { 35960c16b537SWarner Losh case ZSTDv05ds_getFrameHeaderSize : 35970c16b537SWarner Losh /* get frame header size */ 35980c16b537SWarner Losh if (srcSize != ZSTDv05_frameHeaderSize_min) return ERROR(srcSize_wrong); /* impossible */ 35990c16b537SWarner Losh dctx->headerSize = ZSTDv05_decodeFrameHeader_Part1(dctx, src, ZSTDv05_frameHeaderSize_min); 36000c16b537SWarner Losh if (ZSTDv05_isError(dctx->headerSize)) return dctx->headerSize; 36010c16b537SWarner Losh memcpy(dctx->headerBuffer, src, ZSTDv05_frameHeaderSize_min); 36020c16b537SWarner Losh if (dctx->headerSize > ZSTDv05_frameHeaderSize_min) return ERROR(GENERIC); /* should never happen */ 36030c16b537SWarner Losh dctx->expected = 0; /* not necessary to copy more */ 36040c16b537SWarner Losh /* fallthrough */ 36050c16b537SWarner Losh case ZSTDv05ds_decodeFrameHeader: 36060c16b537SWarner Losh /* get frame header */ 36070c16b537SWarner Losh { size_t const result = ZSTDv05_decodeFrameHeader_Part2(dctx, dctx->headerBuffer, dctx->headerSize); 36080c16b537SWarner Losh if (ZSTDv05_isError(result)) return result; 36090c16b537SWarner Losh dctx->expected = ZSTDv05_blockHeaderSize; 36100c16b537SWarner Losh dctx->stage = ZSTDv05ds_decodeBlockHeader; 36110c16b537SWarner Losh return 0; 36120c16b537SWarner Losh } 36130c16b537SWarner Losh case ZSTDv05ds_decodeBlockHeader: 36140c16b537SWarner Losh { 36150c16b537SWarner Losh /* Decode block header */ 36160c16b537SWarner Losh blockProperties_t bp; 36170c16b537SWarner Losh size_t blockSize = ZSTDv05_getcBlockSize(src, ZSTDv05_blockHeaderSize, &bp); 36180c16b537SWarner Losh if (ZSTDv05_isError(blockSize)) return blockSize; 36190c16b537SWarner Losh if (bp.blockType == bt_end) { 36200c16b537SWarner Losh dctx->expected = 0; 36210c16b537SWarner Losh dctx->stage = ZSTDv05ds_getFrameHeaderSize; 36220c16b537SWarner Losh } 36230c16b537SWarner Losh else { 36240c16b537SWarner Losh dctx->expected = blockSize; 36250c16b537SWarner Losh dctx->bType = bp.blockType; 36260c16b537SWarner Losh dctx->stage = ZSTDv05ds_decompressBlock; 36270c16b537SWarner Losh } 36280c16b537SWarner Losh return 0; 36290c16b537SWarner Losh } 36300c16b537SWarner Losh case ZSTDv05ds_decompressBlock: 36310c16b537SWarner Losh { 36320c16b537SWarner Losh /* Decompress : block content */ 36330c16b537SWarner Losh size_t rSize; 36340c16b537SWarner Losh switch(dctx->bType) 36350c16b537SWarner Losh { 36360c16b537SWarner Losh case bt_compressed: 36370c16b537SWarner Losh rSize = ZSTDv05_decompressBlock_internal(dctx, dst, maxDstSize, src, srcSize); 36380c16b537SWarner Losh break; 36390c16b537SWarner Losh case bt_raw : 36400c16b537SWarner Losh rSize = ZSTDv05_copyRawBlock(dst, maxDstSize, src, srcSize); 36410c16b537SWarner Losh break; 36420c16b537SWarner Losh case bt_rle : 36430c16b537SWarner Losh return ERROR(GENERIC); /* not yet handled */ 36440c16b537SWarner Losh break; 36450c16b537SWarner Losh case bt_end : /* should never happen (filtered at phase 1) */ 36460c16b537SWarner Losh rSize = 0; 36470c16b537SWarner Losh break; 36480c16b537SWarner Losh default: 36490c16b537SWarner Losh return ERROR(GENERIC); /* impossible */ 36500c16b537SWarner Losh } 36510c16b537SWarner Losh dctx->stage = ZSTDv05ds_decodeBlockHeader; 36520c16b537SWarner Losh dctx->expected = ZSTDv05_blockHeaderSize; 36530c16b537SWarner Losh dctx->previousDstEnd = (char*)dst + rSize; 36540c16b537SWarner Losh return rSize; 36550c16b537SWarner Losh } 36560c16b537SWarner Losh default: 36570c16b537SWarner Losh return ERROR(GENERIC); /* impossible */ 36580c16b537SWarner Losh } 36590c16b537SWarner Losh } 36600c16b537SWarner Losh 36610c16b537SWarner Losh 36620c16b537SWarner Losh static void ZSTDv05_refDictContent(ZSTDv05_DCtx* dctx, const void* dict, size_t dictSize) 36630c16b537SWarner Losh { 36640c16b537SWarner Losh dctx->dictEnd = dctx->previousDstEnd; 36650c16b537SWarner Losh dctx->vBase = (const char*)dict - ((const char*)(dctx->previousDstEnd) - (const char*)(dctx->base)); 36660c16b537SWarner Losh dctx->base = dict; 36670c16b537SWarner Losh dctx->previousDstEnd = (const char*)dict + dictSize; 36680c16b537SWarner Losh } 36690c16b537SWarner Losh 36700c16b537SWarner Losh static size_t ZSTDv05_loadEntropy(ZSTDv05_DCtx* dctx, const void* dict, size_t dictSize) 36710c16b537SWarner Losh { 36720c16b537SWarner Losh size_t hSize, offcodeHeaderSize, matchlengthHeaderSize, errorCode, litlengthHeaderSize; 36730c16b537SWarner Losh short offcodeNCount[MaxOff+1]; 3674a0483764SConrad Meyer unsigned offcodeMaxValue=MaxOff, offcodeLog; 36750c16b537SWarner Losh short matchlengthNCount[MaxML+1]; 36760c16b537SWarner Losh unsigned matchlengthMaxValue = MaxML, matchlengthLog; 36770c16b537SWarner Losh short litlengthNCount[MaxLL+1]; 36780c16b537SWarner Losh unsigned litlengthMaxValue = MaxLL, litlengthLog; 36790c16b537SWarner Losh 36800c16b537SWarner Losh hSize = HUFv05_readDTableX4(dctx->hufTableX4, dict, dictSize); 36810c16b537SWarner Losh if (HUFv05_isError(hSize)) return ERROR(dictionary_corrupted); 36820c16b537SWarner Losh dict = (const char*)dict + hSize; 36830c16b537SWarner Losh dictSize -= hSize; 36840c16b537SWarner Losh 36850c16b537SWarner Losh offcodeHeaderSize = FSEv05_readNCount(offcodeNCount, &offcodeMaxValue, &offcodeLog, dict, dictSize); 36860c16b537SWarner Losh if (FSEv05_isError(offcodeHeaderSize)) return ERROR(dictionary_corrupted); 36870c16b537SWarner Losh if (offcodeLog > OffFSEv05Log) return ERROR(dictionary_corrupted); 36880c16b537SWarner Losh errorCode = FSEv05_buildDTable(dctx->OffTable, offcodeNCount, offcodeMaxValue, offcodeLog); 36890c16b537SWarner Losh if (FSEv05_isError(errorCode)) return ERROR(dictionary_corrupted); 36900c16b537SWarner Losh dict = (const char*)dict + offcodeHeaderSize; 36910c16b537SWarner Losh dictSize -= offcodeHeaderSize; 36920c16b537SWarner Losh 36930c16b537SWarner Losh matchlengthHeaderSize = FSEv05_readNCount(matchlengthNCount, &matchlengthMaxValue, &matchlengthLog, dict, dictSize); 36940c16b537SWarner Losh if (FSEv05_isError(matchlengthHeaderSize)) return ERROR(dictionary_corrupted); 36950c16b537SWarner Losh if (matchlengthLog > MLFSEv05Log) return ERROR(dictionary_corrupted); 36960c16b537SWarner Losh errorCode = FSEv05_buildDTable(dctx->MLTable, matchlengthNCount, matchlengthMaxValue, matchlengthLog); 36970c16b537SWarner Losh if (FSEv05_isError(errorCode)) return ERROR(dictionary_corrupted); 36980c16b537SWarner Losh dict = (const char*)dict + matchlengthHeaderSize; 36990c16b537SWarner Losh dictSize -= matchlengthHeaderSize; 37000c16b537SWarner Losh 37010c16b537SWarner Losh litlengthHeaderSize = FSEv05_readNCount(litlengthNCount, &litlengthMaxValue, &litlengthLog, dict, dictSize); 37020c16b537SWarner Losh if (litlengthLog > LLFSEv05Log) return ERROR(dictionary_corrupted); 37030c16b537SWarner Losh if (FSEv05_isError(litlengthHeaderSize)) return ERROR(dictionary_corrupted); 37040c16b537SWarner Losh errorCode = FSEv05_buildDTable(dctx->LLTable, litlengthNCount, litlengthMaxValue, litlengthLog); 37050c16b537SWarner Losh if (FSEv05_isError(errorCode)) return ERROR(dictionary_corrupted); 37060c16b537SWarner Losh 37070c16b537SWarner Losh dctx->flagStaticTables = 1; 37080c16b537SWarner Losh return hSize + offcodeHeaderSize + matchlengthHeaderSize + litlengthHeaderSize; 37090c16b537SWarner Losh } 37100c16b537SWarner Losh 37110c16b537SWarner Losh static size_t ZSTDv05_decompress_insertDictionary(ZSTDv05_DCtx* dctx, const void* dict, size_t dictSize) 37120c16b537SWarner Losh { 37130c16b537SWarner Losh size_t eSize; 37140c16b537SWarner Losh U32 magic = MEM_readLE32(dict); 37150c16b537SWarner Losh if (magic != ZSTDv05_DICT_MAGIC) { 37160c16b537SWarner Losh /* pure content mode */ 37170c16b537SWarner Losh ZSTDv05_refDictContent(dctx, dict, dictSize); 37180c16b537SWarner Losh return 0; 37190c16b537SWarner Losh } 37200c16b537SWarner Losh /* load entropy tables */ 37210c16b537SWarner Losh dict = (const char*)dict + 4; 37220c16b537SWarner Losh dictSize -= 4; 37230c16b537SWarner Losh eSize = ZSTDv05_loadEntropy(dctx, dict, dictSize); 37240c16b537SWarner Losh if (ZSTDv05_isError(eSize)) return ERROR(dictionary_corrupted); 37250c16b537SWarner Losh 37260c16b537SWarner Losh /* reference dictionary content */ 37270c16b537SWarner Losh dict = (const char*)dict + eSize; 37280c16b537SWarner Losh dictSize -= eSize; 37290c16b537SWarner Losh ZSTDv05_refDictContent(dctx, dict, dictSize); 37300c16b537SWarner Losh 37310c16b537SWarner Losh return 0; 37320c16b537SWarner Losh } 37330c16b537SWarner Losh 37340c16b537SWarner Losh 37350c16b537SWarner Losh size_t ZSTDv05_decompressBegin_usingDict(ZSTDv05_DCtx* dctx, const void* dict, size_t dictSize) 37360c16b537SWarner Losh { 37370c16b537SWarner Losh size_t errorCode; 37380c16b537SWarner Losh errorCode = ZSTDv05_decompressBegin(dctx); 37390c16b537SWarner Losh if (ZSTDv05_isError(errorCode)) return errorCode; 37400c16b537SWarner Losh 37410c16b537SWarner Losh if (dict && dictSize) { 37420c16b537SWarner Losh errorCode = ZSTDv05_decompress_insertDictionary(dctx, dict, dictSize); 37430c16b537SWarner Losh if (ZSTDv05_isError(errorCode)) return ERROR(dictionary_corrupted); 37440c16b537SWarner Losh } 37450c16b537SWarner Losh 37460c16b537SWarner Losh return 0; 37470c16b537SWarner Losh } 37480c16b537SWarner Losh 37490c16b537SWarner Losh /* 37500c16b537SWarner Losh Buffered version of Zstd compression library 37510c16b537SWarner Losh Copyright (C) 2015-2016, Yann Collet. 37520c16b537SWarner Losh 37530c16b537SWarner Losh BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php) 37540c16b537SWarner Losh 37550c16b537SWarner Losh Redistribution and use in source and binary forms, with or without 37560c16b537SWarner Losh modification, are permitted provided that the following conditions are 37570c16b537SWarner Losh met: 37580c16b537SWarner Losh * Redistributions of source code must retain the above copyright 37590c16b537SWarner Losh notice, this list of conditions and the following disclaimer. 37600c16b537SWarner Losh * Redistributions in binary form must reproduce the above 37610c16b537SWarner Losh copyright notice, this list of conditions and the following disclaimer 37620c16b537SWarner Losh in the documentation and/or other materials provided with the 37630c16b537SWarner Losh distribution. 37640c16b537SWarner Losh THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 37650c16b537SWarner Losh "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 37660c16b537SWarner Losh LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 37670c16b537SWarner Losh A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 37680c16b537SWarner Losh OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 37690c16b537SWarner Losh SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 37700c16b537SWarner Losh LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 37710c16b537SWarner Losh DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 37720c16b537SWarner Losh THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 37730c16b537SWarner Losh (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 37740c16b537SWarner Losh OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 37750c16b537SWarner Losh 37760c16b537SWarner Losh You can contact the author at : 37770c16b537SWarner Losh - zstd source repository : https://github.com/Cyan4973/zstd 37780c16b537SWarner Losh - ztsd public forum : https://groups.google.com/forum/#!forum/lz4c 37790c16b537SWarner Losh */ 37800c16b537SWarner Losh 37810c16b537SWarner Losh /* The objects defined into this file should be considered experimental. 37820c16b537SWarner Losh * They are not labelled stable, as their prototype may change in the future. 37830c16b537SWarner Losh * You can use them for tests, provide feedback, or if you can endure risk of future changes. 37840c16b537SWarner Losh */ 37850c16b537SWarner Losh 37860c16b537SWarner Losh 37870c16b537SWarner Losh 37880c16b537SWarner Losh /* ************************************* 37890c16b537SWarner Losh * Constants 37900c16b537SWarner Losh ***************************************/ 37910c16b537SWarner Losh static size_t ZBUFFv05_blockHeaderSize = 3; 37920c16b537SWarner Losh 37930c16b537SWarner Losh 37940c16b537SWarner Losh 37950c16b537SWarner Losh /* *** Compression *** */ 37960c16b537SWarner Losh 37970c16b537SWarner Losh static size_t ZBUFFv05_limitCopy(void* dst, size_t maxDstSize, const void* src, size_t srcSize) 37980c16b537SWarner Losh { 37990c16b537SWarner Losh size_t length = MIN(maxDstSize, srcSize); 380037f1f268SConrad Meyer if (length > 0) { 38010c16b537SWarner Losh memcpy(dst, src, length); 380237f1f268SConrad Meyer } 38030c16b537SWarner Losh return length; 38040c16b537SWarner Losh } 38050c16b537SWarner Losh 38060c16b537SWarner Losh 38070c16b537SWarner Losh 38080c16b537SWarner Losh 38090c16b537SWarner Losh /** ************************************************ 38100c16b537SWarner Losh * Streaming decompression 38110c16b537SWarner Losh * 38120c16b537SWarner Losh * A ZBUFFv05_DCtx object is required to track streaming operation. 38130c16b537SWarner Losh * Use ZBUFFv05_createDCtx() and ZBUFFv05_freeDCtx() to create/release resources. 38140c16b537SWarner Losh * Use ZBUFFv05_decompressInit() to start a new decompression operation. 38150c16b537SWarner Losh * ZBUFFv05_DCtx objects can be reused multiple times. 38160c16b537SWarner Losh * 38170c16b537SWarner Losh * Use ZBUFFv05_decompressContinue() repetitively to consume your input. 38180c16b537SWarner Losh * *srcSizePtr and *maxDstSizePtr can be any size. 38190c16b537SWarner Losh * The function will report how many bytes were read or written by modifying *srcSizePtr and *maxDstSizePtr. 38200c16b537SWarner Losh * Note that it may not consume the entire input, in which case it's up to the caller to call again the function with remaining input. 38210c16b537SWarner Losh * The content of dst will be overwritten (up to *maxDstSizePtr) at each function call, so save its content if it matters or change dst . 38220c16b537SWarner Losh * return : a hint to preferred nb of bytes to use as input for next function call (it's only a hint, to improve latency) 38230c16b537SWarner Losh * or 0 when a frame is completely decoded 38240c16b537SWarner Losh * or an error code, which can be tested using ZBUFFv05_isError(). 38250c16b537SWarner Losh * 38260c16b537SWarner Losh * Hint : recommended buffer sizes (not compulsory) 38270c16b537SWarner Losh * output : 128 KB block size is the internal unit, it ensures it's always possible to write a full block when it's decoded. 38280c16b537SWarner Losh * input : just follow indications from ZBUFFv05_decompressContinue() to minimize latency. It should always be <= 128 KB + 3 . 38290c16b537SWarner Losh * **************************************************/ 38300c16b537SWarner Losh 38310c16b537SWarner Losh typedef enum { ZBUFFv05ds_init, ZBUFFv05ds_readHeader, ZBUFFv05ds_loadHeader, ZBUFFv05ds_decodeHeader, 38320c16b537SWarner Losh ZBUFFv05ds_read, ZBUFFv05ds_load, ZBUFFv05ds_flush } ZBUFFv05_dStage; 38330c16b537SWarner Losh 38340c16b537SWarner Losh /* *** Resource management *** */ 38350c16b537SWarner Losh 38360c16b537SWarner Losh #define ZSTDv05_frameHeaderSize_max 5 /* too magical, should come from reference */ 38370c16b537SWarner Losh struct ZBUFFv05_DCtx_s { 38380c16b537SWarner Losh ZSTDv05_DCtx* zc; 38390c16b537SWarner Losh ZSTDv05_parameters params; 38400c16b537SWarner Losh char* inBuff; 38410c16b537SWarner Losh size_t inBuffSize; 38420c16b537SWarner Losh size_t inPos; 38430c16b537SWarner Losh char* outBuff; 38440c16b537SWarner Losh size_t outBuffSize; 38450c16b537SWarner Losh size_t outStart; 38460c16b537SWarner Losh size_t outEnd; 38470c16b537SWarner Losh size_t hPos; 38480c16b537SWarner Losh ZBUFFv05_dStage stage; 38490c16b537SWarner Losh unsigned char headerBuffer[ZSTDv05_frameHeaderSize_max]; 38500c16b537SWarner Losh }; /* typedef'd to ZBUFFv05_DCtx within "zstd_buffered.h" */ 38510c16b537SWarner Losh 38520c16b537SWarner Losh 38530c16b537SWarner Losh ZBUFFv05_DCtx* ZBUFFv05_createDCtx(void) 38540c16b537SWarner Losh { 38550c16b537SWarner Losh ZBUFFv05_DCtx* zbc = (ZBUFFv05_DCtx*)malloc(sizeof(ZBUFFv05_DCtx)); 38560c16b537SWarner Losh if (zbc==NULL) return NULL; 38570c16b537SWarner Losh memset(zbc, 0, sizeof(*zbc)); 38580c16b537SWarner Losh zbc->zc = ZSTDv05_createDCtx(); 38590c16b537SWarner Losh zbc->stage = ZBUFFv05ds_init; 38600c16b537SWarner Losh return zbc; 38610c16b537SWarner Losh } 38620c16b537SWarner Losh 38630c16b537SWarner Losh size_t ZBUFFv05_freeDCtx(ZBUFFv05_DCtx* zbc) 38640c16b537SWarner Losh { 38650c16b537SWarner Losh if (zbc==NULL) return 0; /* support free on null */ 38660c16b537SWarner Losh ZSTDv05_freeDCtx(zbc->zc); 38670c16b537SWarner Losh free(zbc->inBuff); 38680c16b537SWarner Losh free(zbc->outBuff); 38690c16b537SWarner Losh free(zbc); 38700c16b537SWarner Losh return 0; 38710c16b537SWarner Losh } 38720c16b537SWarner Losh 38730c16b537SWarner Losh 38740c16b537SWarner Losh /* *** Initialization *** */ 38750c16b537SWarner Losh 38760c16b537SWarner Losh size_t ZBUFFv05_decompressInitDictionary(ZBUFFv05_DCtx* zbc, const void* dict, size_t dictSize) 38770c16b537SWarner Losh { 38780c16b537SWarner Losh zbc->stage = ZBUFFv05ds_readHeader; 38790c16b537SWarner Losh zbc->hPos = zbc->inPos = zbc->outStart = zbc->outEnd = 0; 38800c16b537SWarner Losh return ZSTDv05_decompressBegin_usingDict(zbc->zc, dict, dictSize); 38810c16b537SWarner Losh } 38820c16b537SWarner Losh 38830c16b537SWarner Losh size_t ZBUFFv05_decompressInit(ZBUFFv05_DCtx* zbc) 38840c16b537SWarner Losh { 38850c16b537SWarner Losh return ZBUFFv05_decompressInitDictionary(zbc, NULL, 0); 38860c16b537SWarner Losh } 38870c16b537SWarner Losh 38880c16b537SWarner Losh 38890c16b537SWarner Losh /* *** Decompression *** */ 38900c16b537SWarner Losh 38910c16b537SWarner Losh size_t ZBUFFv05_decompressContinue(ZBUFFv05_DCtx* zbc, void* dst, size_t* maxDstSizePtr, const void* src, size_t* srcSizePtr) 38920c16b537SWarner Losh { 38930c16b537SWarner Losh const char* const istart = (const char*)src; 38940c16b537SWarner Losh const char* ip = istart; 38950c16b537SWarner Losh const char* const iend = istart + *srcSizePtr; 38960c16b537SWarner Losh char* const ostart = (char*)dst; 38970c16b537SWarner Losh char* op = ostart; 38980c16b537SWarner Losh char* const oend = ostart + *maxDstSizePtr; 38990c16b537SWarner Losh U32 notDone = 1; 39000c16b537SWarner Losh 39010c16b537SWarner Losh while (notDone) { 39020c16b537SWarner Losh switch(zbc->stage) 39030c16b537SWarner Losh { 39040c16b537SWarner Losh case ZBUFFv05ds_init : 39050c16b537SWarner Losh return ERROR(init_missing); 39060c16b537SWarner Losh 39070c16b537SWarner Losh case ZBUFFv05ds_readHeader : 39080c16b537SWarner Losh /* read header from src */ 39090c16b537SWarner Losh { 39100c16b537SWarner Losh size_t headerSize = ZSTDv05_getFrameParams(&(zbc->params), src, *srcSizePtr); 39110c16b537SWarner Losh if (ZSTDv05_isError(headerSize)) return headerSize; 39120c16b537SWarner Losh if (headerSize) { 39130c16b537SWarner Losh /* not enough input to decode header : tell how many bytes would be necessary */ 39140c16b537SWarner Losh memcpy(zbc->headerBuffer+zbc->hPos, src, *srcSizePtr); 39150c16b537SWarner Losh zbc->hPos += *srcSizePtr; 39160c16b537SWarner Losh *maxDstSizePtr = 0; 39170c16b537SWarner Losh zbc->stage = ZBUFFv05ds_loadHeader; 39180c16b537SWarner Losh return headerSize - zbc->hPos; 39190c16b537SWarner Losh } 39200c16b537SWarner Losh zbc->stage = ZBUFFv05ds_decodeHeader; 39210c16b537SWarner Losh break; 39220c16b537SWarner Losh } 39230c16b537SWarner Losh /* fall-through */ 39240c16b537SWarner Losh case ZBUFFv05ds_loadHeader: 39250c16b537SWarner Losh /* complete header from src */ 39260c16b537SWarner Losh { 39270c16b537SWarner Losh size_t headerSize = ZBUFFv05_limitCopy( 39280c16b537SWarner Losh zbc->headerBuffer + zbc->hPos, ZSTDv05_frameHeaderSize_max - zbc->hPos, 39290c16b537SWarner Losh src, *srcSizePtr); 39300c16b537SWarner Losh zbc->hPos += headerSize; 39310c16b537SWarner Losh ip += headerSize; 39320c16b537SWarner Losh headerSize = ZSTDv05_getFrameParams(&(zbc->params), zbc->headerBuffer, zbc->hPos); 39330c16b537SWarner Losh if (ZSTDv05_isError(headerSize)) return headerSize; 39340c16b537SWarner Losh if (headerSize) { 39350c16b537SWarner Losh /* not enough input to decode header : tell how many bytes would be necessary */ 39360c16b537SWarner Losh *maxDstSizePtr = 0; 39370c16b537SWarner Losh return headerSize - zbc->hPos; 39380c16b537SWarner Losh } 393937f1f268SConrad Meyer /* zbc->stage = ZBUFFv05ds_decodeHeader; break; */ /* useless : stage follows */ 39400c16b537SWarner Losh } 39410c16b537SWarner Losh /* fall-through */ 39420c16b537SWarner Losh case ZBUFFv05ds_decodeHeader: 39430c16b537SWarner Losh /* apply header to create / resize buffers */ 39440c16b537SWarner Losh { 39450c16b537SWarner Losh size_t neededOutSize = (size_t)1 << zbc->params.windowLog; 39460c16b537SWarner Losh size_t neededInSize = BLOCKSIZE; /* a block is never > BLOCKSIZE */ 39470c16b537SWarner Losh if (zbc->inBuffSize < neededInSize) { 39480c16b537SWarner Losh free(zbc->inBuff); 39490c16b537SWarner Losh zbc->inBuffSize = neededInSize; 39500c16b537SWarner Losh zbc->inBuff = (char*)malloc(neededInSize); 39510c16b537SWarner Losh if (zbc->inBuff == NULL) return ERROR(memory_allocation); 39520c16b537SWarner Losh } 39530c16b537SWarner Losh if (zbc->outBuffSize < neededOutSize) { 39540c16b537SWarner Losh free(zbc->outBuff); 39550c16b537SWarner Losh zbc->outBuffSize = neededOutSize; 39560c16b537SWarner Losh zbc->outBuff = (char*)malloc(neededOutSize); 39570c16b537SWarner Losh if (zbc->outBuff == NULL) return ERROR(memory_allocation); 39580c16b537SWarner Losh } } 39590c16b537SWarner Losh if (zbc->hPos) { 39600c16b537SWarner Losh /* some data already loaded into headerBuffer : transfer into inBuff */ 39610c16b537SWarner Losh memcpy(zbc->inBuff, zbc->headerBuffer, zbc->hPos); 39620c16b537SWarner Losh zbc->inPos = zbc->hPos; 39630c16b537SWarner Losh zbc->hPos = 0; 39640c16b537SWarner Losh zbc->stage = ZBUFFv05ds_load; 39650c16b537SWarner Losh break; 39660c16b537SWarner Losh } 39670c16b537SWarner Losh zbc->stage = ZBUFFv05ds_read; 39680c16b537SWarner Losh /* fall-through */ 39690c16b537SWarner Losh case ZBUFFv05ds_read: 39700c16b537SWarner Losh { 39710c16b537SWarner Losh size_t neededInSize = ZSTDv05_nextSrcSizeToDecompress(zbc->zc); 39720c16b537SWarner Losh if (neededInSize==0) { /* end of frame */ 39730c16b537SWarner Losh zbc->stage = ZBUFFv05ds_init; 39740c16b537SWarner Losh notDone = 0; 39750c16b537SWarner Losh break; 39760c16b537SWarner Losh } 39770c16b537SWarner Losh if ((size_t)(iend-ip) >= neededInSize) { 39780c16b537SWarner Losh /* directly decode from src */ 39790c16b537SWarner Losh size_t decodedSize = ZSTDv05_decompressContinue(zbc->zc, 39800c16b537SWarner Losh zbc->outBuff + zbc->outStart, zbc->outBuffSize - zbc->outStart, 39810c16b537SWarner Losh ip, neededInSize); 39820c16b537SWarner Losh if (ZSTDv05_isError(decodedSize)) return decodedSize; 39830c16b537SWarner Losh ip += neededInSize; 39840c16b537SWarner Losh if (!decodedSize) break; /* this was just a header */ 39850c16b537SWarner Losh zbc->outEnd = zbc->outStart + decodedSize; 39860c16b537SWarner Losh zbc->stage = ZBUFFv05ds_flush; 39870c16b537SWarner Losh break; 39880c16b537SWarner Losh } 39890c16b537SWarner Losh if (ip==iend) { notDone = 0; break; } /* no more input */ 39900c16b537SWarner Losh zbc->stage = ZBUFFv05ds_load; 39910c16b537SWarner Losh } 39920c16b537SWarner Losh /* fall-through */ 39930c16b537SWarner Losh case ZBUFFv05ds_load: 39940c16b537SWarner Losh { 39950c16b537SWarner Losh size_t neededInSize = ZSTDv05_nextSrcSizeToDecompress(zbc->zc); 39960c16b537SWarner Losh size_t toLoad = neededInSize - zbc->inPos; /* should always be <= remaining space within inBuff */ 39970c16b537SWarner Losh size_t loadedSize; 39980c16b537SWarner Losh if (toLoad > zbc->inBuffSize - zbc->inPos) return ERROR(corruption_detected); /* should never happen */ 39990c16b537SWarner Losh loadedSize = ZBUFFv05_limitCopy(zbc->inBuff + zbc->inPos, toLoad, ip, iend-ip); 40000c16b537SWarner Losh ip += loadedSize; 40010c16b537SWarner Losh zbc->inPos += loadedSize; 40020c16b537SWarner Losh if (loadedSize < toLoad) { notDone = 0; break; } /* not enough input, wait for more */ 40030c16b537SWarner Losh { 40040c16b537SWarner Losh size_t decodedSize = ZSTDv05_decompressContinue(zbc->zc, 40050c16b537SWarner Losh zbc->outBuff + zbc->outStart, zbc->outBuffSize - zbc->outStart, 40060c16b537SWarner Losh zbc->inBuff, neededInSize); 40070c16b537SWarner Losh if (ZSTDv05_isError(decodedSize)) return decodedSize; 40080c16b537SWarner Losh zbc->inPos = 0; /* input is consumed */ 40090c16b537SWarner Losh if (!decodedSize) { zbc->stage = ZBUFFv05ds_read; break; } /* this was just a header */ 40100c16b537SWarner Losh zbc->outEnd = zbc->outStart + decodedSize; 40110c16b537SWarner Losh zbc->stage = ZBUFFv05ds_flush; 401237f1f268SConrad Meyer /* break; */ /* ZBUFFv05ds_flush follows */ 40130c16b537SWarner Losh } 40140c16b537SWarner Losh } 40150c16b537SWarner Losh /* fall-through */ 40160c16b537SWarner Losh case ZBUFFv05ds_flush: 40170c16b537SWarner Losh { 40180c16b537SWarner Losh size_t toFlushSize = zbc->outEnd - zbc->outStart; 40190c16b537SWarner Losh size_t flushedSize = ZBUFFv05_limitCopy(op, oend-op, zbc->outBuff + zbc->outStart, toFlushSize); 40200c16b537SWarner Losh op += flushedSize; 40210c16b537SWarner Losh zbc->outStart += flushedSize; 40220c16b537SWarner Losh if (flushedSize == toFlushSize) { 40230c16b537SWarner Losh zbc->stage = ZBUFFv05ds_read; 40240c16b537SWarner Losh if (zbc->outStart + BLOCKSIZE > zbc->outBuffSize) 40250c16b537SWarner Losh zbc->outStart = zbc->outEnd = 0; 40260c16b537SWarner Losh break; 40270c16b537SWarner Losh } 40280c16b537SWarner Losh /* cannot flush everything */ 40290c16b537SWarner Losh notDone = 0; 40300c16b537SWarner Losh break; 40310c16b537SWarner Losh } 40320c16b537SWarner Losh default: return ERROR(GENERIC); /* impossible */ 40330c16b537SWarner Losh } } 40340c16b537SWarner Losh 40350c16b537SWarner Losh *srcSizePtr = ip-istart; 40360c16b537SWarner Losh *maxDstSizePtr = op-ostart; 40370c16b537SWarner Losh 40380c16b537SWarner Losh { size_t nextSrcSizeHint = ZSTDv05_nextSrcSizeToDecompress(zbc->zc); 40390c16b537SWarner Losh if (nextSrcSizeHint > ZBUFFv05_blockHeaderSize) nextSrcSizeHint+= ZBUFFv05_blockHeaderSize; /* get next block header too */ 40400c16b537SWarner Losh nextSrcSizeHint -= zbc->inPos; /* already loaded*/ 40410c16b537SWarner Losh return nextSrcSizeHint; 40420c16b537SWarner Losh } 40430c16b537SWarner Losh } 40440c16b537SWarner Losh 40450c16b537SWarner Losh 40460c16b537SWarner Losh 40470c16b537SWarner Losh /* ************************************* 40480c16b537SWarner Losh * Tool functions 40490c16b537SWarner Losh ***************************************/ 40500c16b537SWarner Losh unsigned ZBUFFv05_isError(size_t errorCode) { return ERR_isError(errorCode); } 40510c16b537SWarner Losh const char* ZBUFFv05_getErrorName(size_t errorCode) { return ERR_getErrorName(errorCode); } 40520c16b537SWarner Losh 40530c16b537SWarner Losh size_t ZBUFFv05_recommendedDInSize(void) { return BLOCKSIZE + ZBUFFv05_blockHeaderSize /* block header size*/ ; } 40540c16b537SWarner Losh size_t ZBUFFv05_recommendedDOutSize(void) { return BLOCKSIZE; } 4055