From 1707848d3105cde28e3478005f46cc5a5334f216 Mon Sep 17 00:00:00 2001 From: Dorian Wouters Date: Sun, 24 Jan 2016 14:59:14 +0100 Subject: [PATCH] Replace lodepng by stb_image; add Vorbis banner sound import --- CMakeLists.txt | 2 +- source/3ds/util.cpp | 25 +- source/cmd.cpp | 62 +- source/pc/lodepng.cpp | 6104 ------------------------------------ source/pc/lodepng.h | 1702 ----------- source/pc/stb_image.c | 3 + source/pc/stb_image.h | 6614 ++++++++++++++++++++++++++++++++++++++++ source/pc/stb_vorbis.c | 5462 +++++++++++++++++++++++++++++++++ source/pc/stb_vorbis.h | 2 + source/pc/wav.cpp | 6 +- source/pc/wav.h | 4 +- 11 files changed, 12154 insertions(+), 7832 deletions(-) delete mode 100644 source/pc/lodepng.cpp delete mode 100644 source/pc/lodepng.h create mode 100644 source/pc/stb_image.c create mode 100644 source/pc/stb_image.h create mode 100644 source/pc/stb_vorbis.c create mode 100644 source/pc/stb_vorbis.h diff --git a/CMakeLists.txt b/CMakeLists.txt index 053b3d6..9dc190b 100644 --- a/CMakeLists.txt +++ b/CMakeLists.txt @@ -3,5 +3,5 @@ project(bannertool) set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11") -set(SOURCE_FILES source/main.cpp source/cmd.cpp source/pc/wav.cpp source/3ds/cbmd.cpp source/3ds/cwav.cpp source/3ds/lz11.cpp source/3ds/util.cpp source/pc/lodepng.cpp) +set(SOURCE_FILES source/main.cpp source/cmd.cpp source/pc/wav.cpp source/3ds/cbmd.cpp source/3ds/cwav.cpp source/3ds/lz11.cpp source/3ds/util.cpp source/pc/stb_image.c source/pc/stb_vorbis.c) add_executable(bannertool ${SOURCE_FILES}) \ No newline at end of file diff --git a/source/3ds/util.cpp b/source/3ds/util.cpp index 941c1a0..28a3494 100644 --- a/source/3ds/util.cpp +++ b/source/3ds/util.cpp @@ -1,6 +1,6 @@ #include "util.h" -#include "../pc/lodepng.h" +#include "../pc/stb_image.h" #include @@ -39,9 +39,12 @@ u16 pack_color(u8 r, u8 g, u8 b, u8 a, PixelFormat format) { u8* load_image(const char* image, u32 width, u32 height) { unsigned char *img; - unsigned int imgWidth, imgHeight; - if(lodepng_decode32_file(&img, &imgWidth, &imgHeight, image)) { - printf("ERROR: Could not load png file.\n"); + int imgWidth, imgHeight, imgDepth; + + img = stbi_load(image, &imgWidth, &imgHeight, &imgDepth, STBI_rgb_alpha); + + if(img == NULL) { + printf("ERROR: Could not load image file: %s.\n", stbi_failure_reason()); return NULL; } @@ -58,9 +61,19 @@ u8* load_image(const char* image, u32 width, u32 height) { return NULL; } + if(imgDepth != STBI_rgb_alpha) { + printf("ERROR: Decoded image does't match expected format (%d, wanted %d).\n", + imgDepth, STBI_rgb_alpha); + return NULL; + } + return img; } +void free_image(u8* img) { + stbi_image_free(img); +} + u16* image_data_to_tiles(u8* img, u32 width, u32 height, PixelFormat format, u32* size) { u16* converted = (u16*) malloc(width * height * sizeof(u16)); u32 n = 0; @@ -89,5 +102,7 @@ u16* image_to_tiles(const char* image, u32 width, u32 height, PixelFormat format return NULL; } - return image_data_to_tiles(img, width, height, format, size); + u16* tiles = image_data_to_tiles(img, width, height, format, size); + free_image(img); + return tiles; } \ No newline at end of file diff --git a/source/cmd.cpp b/source/cmd.cpp index 3c90b47..1354296 100644 --- a/source/cmd.cpp +++ b/source/cmd.cpp @@ -2,6 +2,7 @@ #include "3ds/3ds.h" #include "pc/wav.h" +#include "pc/stb_vorbis.h" #include "types.h" #include @@ -24,26 +25,59 @@ u8* convert_to_cgfx(const std::string image, u32 width, u32 height, u32* size) { memcpy(ret + BANNER_CGFX_HEADER_LENGTH, converted, convertedSize); *size = BANNER_CGFX_HEADER_LENGTH + convertedSize; + free(converted); return ret; } u8* convert_to_cwav(const std::string file, u32* size) { - WAV* wav = wav_read(file.c_str()); - if(wav == NULL) { - return NULL; + u8* ret = NULL; + // Determine what file type we have + FILE* fd = fopen(file.c_str(), "rb"); + char magic[4]; + fread(magic, sizeof(magic), 1, fd); + rewind(fd); // equivalent to SEEK_SET to pos 0 + + if (magic[0] == 'R' && magic[1] == 'I' && magic[2] == 'F' && magic[3] == 'F') { + WAV* wav = wav_read(fd); + if(wav != NULL) { + CWAV cwav; + cwav.channels = wav->format.numChannels; + cwav.sampleRate = wav->format.sampleRate; + cwav.bitsPerSample = wav->format.bitsPerSample; + cwav.dataSize = wav->data.chunkSize; + cwav.data = wav->data.data; + + ret = cwav_build(cwav, size); + + wav_free(wav); + } + } else if (magic[0] == 'O' && magic[1] == 'g' && magic[2] == 'g' && magic[3] == 'S') { + int error; + stb_vorbis* vorb = stb_vorbis_open_file(fd, false, &error, NULL); + if(vorb != NULL) { + stb_vorbis_info info = stb_vorbis_get_info(vorb); + + CWAV cwav; + cwav.channels = info.channels; + cwav.sampleRate = info.sample_rate; + cwav.bitsPerSample = 16; // stb_vorbis always outputs 16 bit samples + int sampleCount = stb_vorbis_stream_length_in_samples(vorb) * info.channels; + cwav.dataSize = sampleCount * 2; + cwav.data = (u8*) calloc(sampleCount, 2); + stb_vorbis_get_samples_short_interleaved(vorb, info.channels, (short*) cwav.data, sampleCount); + + ret = cwav_build(cwav, size); + + free(cwav.data); + stb_vorbis_close(vorb); + } else { + printf("ERROR: Vorbis open failed, error %d.\n", error); + } + } else { + printf("ERROR: Audio file header '%c%c%c%c' unrecognized.\n", magic[0], magic[1], magic[2], magic[3]); } - CWAV cwav; - cwav.channels = wav->format.numChannels; - cwav.sampleRate = wav->format.sampleRate; - cwav.bitsPerSample = wav->format.bitsPerSample; - cwav.dataSize = wav->data.chunkSize; - cwav.data = wav->data.data; - - u8* ret = cwav_build(cwav, size); - - wav_free(wav); - + fclose(fd); return ret; } diff --git a/source/pc/lodepng.cpp b/source/pc/lodepng.cpp deleted file mode 100644 index 168503d..0000000 --- a/source/pc/lodepng.cpp +++ /dev/null @@ -1,6104 +0,0 @@ -/* -LodePNG version 20140823 - -Copyright (c) 2005-2014 Lode Vandevenne - -This software is provided 'as-is', without any express or implied -warranty. In no event will the authors be held liable for any damages -arising from the use of this software. - -Permission is granted to anyone to use this software for any purpose, -including commercial applications, and to alter it and redistribute it -freely, subject to the following restrictions: - - 1. The origin of this software must not be misrepresented; you must not - claim that you wrote the original software. If you use this software - in a product, an acknowledgment in the product documentation would be - appreciated but is not required. - - 2. Altered source versions must be plainly marked as such, and must not be - misrepresented as being the original software. - - 3. This notice may not be removed or altered from any source - distribution. -*/ - -/* -The manual and changelog are in the header file "lodepng.h" -Rename this file to lodepng.cpp to use it for C++, or to lodepng.c to use it for C. -*/ - -#include "lodepng.h" - -#include -#include - -#ifdef LODEPNG_COMPILE_CPP -#include -#endif /*LODEPNG_COMPILE_CPP*/ - -#define VERSION_STRING "20140823" - -#if defined(_MSC_VER) && (_MSC_VER >= 1310) /*Visual Studio: A few warning types are not desired here.*/ -#pragma warning( disable : 4244 ) /*implicit conversions: not warned by gcc -Wall -Wextra and requires too much casts*/ -#pragma warning( disable : 4996 ) /*VS does not like fopen, but fopen_s is not standard C so unusable here*/ -#endif /*_MSC_VER */ - -/* -This source file is built up in the following large parts. The code sections -with the "LODEPNG_COMPILE_" #defines divide this up further in an intermixed way. --Tools for C and common code for PNG and Zlib --C Code for Zlib (huffman, deflate, ...) --C Code for PNG (file format chunks, adam7, PNG filters, color conversions, ...) --The C++ wrapper around all of the above -*/ - -/*The malloc, realloc and free functions defined here with "lodepng_" in front -of the name, so that you can easily change them to others related to your -platform if needed. Everything else in the code calls these. Pass --DLODEPNG_NO_COMPILE_ALLOCATORS to the compiler, or comment out -#define LODEPNG_COMPILE_ALLOCATORS in the header, to disable the ones here and -define them in your own project's source files without needing to change -lodepng source code. Don't forget to remove "static" if you copypaste them -from here.*/ - -#ifdef LODEPNG_COMPILE_ALLOCATORS -static void* lodepng_malloc(size_t size) -{ - return malloc(size); -} - -static void* lodepng_realloc(void* ptr, size_t new_size) -{ - return realloc(ptr, new_size); -} - -static void lodepng_free(void* ptr) -{ - free(ptr); -} -#else /*LODEPNG_COMPILE_ALLOCATORS*/ -void* lodepng_malloc(size_t size); -void* lodepng_realloc(void* ptr, size_t new_size); -void lodepng_free(void* ptr); -#endif /*LODEPNG_COMPILE_ALLOCATORS*/ - -/* ////////////////////////////////////////////////////////////////////////// */ -/* ////////////////////////////////////////////////////////////////////////// */ -/* // Tools for C, and common code for PNG and Zlib. // */ -/* ////////////////////////////////////////////////////////////////////////// */ -/* ////////////////////////////////////////////////////////////////////////// */ - -/* -Often in case of an error a value is assigned to a variable and then it breaks -out of a loop (to go to the cleanup phase of a function). This macro does that. -It makes the error handling code shorter and more readable. - -Example: if(!uivector_resizev(&frequencies_ll, 286, 0)) ERROR_BREAK(83); -*/ -#define CERROR_BREAK(errorvar, code)\ -{\ - errorvar = code;\ - break;\ -} - -/*version of CERROR_BREAK that assumes the common case where the error variable is named "error"*/ -#define ERROR_BREAK(code) CERROR_BREAK(error, code) - -/*Set error var to the error code, and return it.*/ -#define CERROR_RETURN_ERROR(errorvar, code)\ -{\ - errorvar = code;\ - return code;\ -} - -/*Try the code, if it returns error, also return the error.*/ -#define CERROR_TRY_RETURN(call)\ -{\ - unsigned error = call;\ - if(error) return error;\ -} - -/* -About uivector, ucvector and string: --All of them wrap dynamic arrays or text strings in a similar way. --LodePNG was originally written in C++. The vectors replace the std::vectors that were used in the C++ version. --The string tools are made to avoid problems with compilers that declare things like strncat as deprecated. --They're not used in the interface, only internally in this file as static functions. --As with many other structs in this file, the init and cleanup functions serve as ctor and dtor. -*/ - -#ifdef LODEPNG_COMPILE_ZLIB -/*dynamic vector of unsigned ints*/ -typedef struct uivector -{ - unsigned* data; - size_t size; /*size in number of unsigned longs*/ - size_t allocsize; /*allocated size in bytes*/ -} uivector; - -static void uivector_cleanup(void* p) -{ - ((uivector*)p)->size = ((uivector*)p)->allocsize = 0; - lodepng_free(((uivector*)p)->data); - ((uivector*)p)->data = NULL; -} - -/*returns 1 if success, 0 if failure ==> nothing done*/ -static unsigned uivector_reserve(uivector* p, size_t allocsize) -{ - if(allocsize > p->allocsize) - { - size_t newsize = (allocsize > p->allocsize * 2) ? allocsize : (allocsize * 3 / 2); - void* data = lodepng_realloc(p->data, newsize); - if(data) - { - p->allocsize = newsize; - p->data = (unsigned*)data; - } - else return 0; /*error: not enough memory*/ - } - return 1; -} - -/*returns 1 if success, 0 if failure ==> nothing done*/ -static unsigned uivector_resize(uivector* p, size_t size) -{ - if(!uivector_reserve(p, size * sizeof(unsigned))) return 0; - p->size = size; - return 1; /*success*/ -} - -/*resize and give all new elements the value*/ -static unsigned uivector_resizev(uivector* p, size_t size, unsigned value) -{ - size_t oldsize = p->size, i; - if(!uivector_resize(p, size)) return 0; - for(i = oldsize; i < size; i++) p->data[i] = value; - return 1; -} - -static void uivector_init(uivector* p) -{ - p->data = NULL; - p->size = p->allocsize = 0; -} - -#ifdef LODEPNG_COMPILE_ENCODER -/*returns 1 if success, 0 if failure ==> nothing done*/ -static unsigned uivector_push_back(uivector* p, unsigned c) -{ - if(!uivector_resize(p, p->size + 1)) return 0; - p->data[p->size - 1] = c; - return 1; -} - -/*copy q to p, returns 1 if success, 0 if failure ==> nothing done*/ -static unsigned uivector_copy(uivector* p, const uivector* q) -{ - size_t i; - if(!uivector_resize(p, q->size)) return 0; - for(i = 0; i < q->size; i++) p->data[i] = q->data[i]; - return 1; -} -#endif /*LODEPNG_COMPILE_ENCODER*/ -#endif /*LODEPNG_COMPILE_ZLIB*/ - -/* /////////////////////////////////////////////////////////////////////////// */ - -/*dynamic vector of unsigned chars*/ -typedef struct ucvector -{ - unsigned char* data; - size_t size; /*used size*/ - size_t allocsize; /*allocated size*/ -} ucvector; - -/*returns 1 if success, 0 if failure ==> nothing done*/ -static unsigned ucvector_reserve(ucvector* p, size_t allocsize) -{ - if(allocsize > p->allocsize) - { - size_t newsize = (allocsize > p->allocsize * 2) ? allocsize : (allocsize * 3 / 2); - void* data = lodepng_realloc(p->data, newsize); - if(data) - { - p->allocsize = newsize; - p->data = (unsigned char*)data; - } - else return 0; /*error: not enough memory*/ - } - return 1; -} - -/*returns 1 if success, 0 if failure ==> nothing done*/ -static unsigned ucvector_resize(ucvector* p, size_t size) -{ - if(!ucvector_reserve(p, size * sizeof(unsigned char))) return 0; - p->size = size; - return 1; /*success*/ -} - -#ifdef LODEPNG_COMPILE_PNG - -static void ucvector_cleanup(void* p) -{ - ((ucvector*)p)->size = ((ucvector*)p)->allocsize = 0; - lodepng_free(((ucvector*)p)->data); - ((ucvector*)p)->data = NULL; -} - -static void ucvector_init(ucvector* p) -{ - p->data = NULL; - p->size = p->allocsize = 0; -} - -#ifdef LODEPNG_COMPILE_DECODER -/*resize and give all new elements the value*/ -static unsigned ucvector_resizev(ucvector* p, size_t size, unsigned char value) -{ - size_t oldsize = p->size, i; - if(!ucvector_resize(p, size)) return 0; - for(i = oldsize; i < size; i++) p->data[i] = value; - return 1; -} -#endif /*LODEPNG_COMPILE_DECODER*/ -#endif /*LODEPNG_COMPILE_PNG*/ - -#ifdef LODEPNG_COMPILE_ZLIB -/*you can both convert from vector to buffer&size and vica versa. If you use -init_buffer to take over a buffer and size, it is not needed to use cleanup*/ -static void ucvector_init_buffer(ucvector* p, unsigned char* buffer, size_t size) -{ - p->data = buffer; - p->allocsize = p->size = size; -} -#endif /*LODEPNG_COMPILE_ZLIB*/ - -#if (defined(LODEPNG_COMPILE_PNG) && defined(LODEPNG_COMPILE_ANCILLARY_CHUNKS)) || defined(LODEPNG_COMPILE_ENCODER) -/*returns 1 if success, 0 if failure ==> nothing done*/ -static unsigned ucvector_push_back(ucvector* p, unsigned char c) -{ - if(!ucvector_resize(p, p->size + 1)) return 0; - p->data[p->size - 1] = c; - return 1; -} -#endif /*defined(LODEPNG_COMPILE_PNG) || defined(LODEPNG_COMPILE_ENCODER)*/ - - -/* ////////////////////////////////////////////////////////////////////////// */ - -#ifdef LODEPNG_COMPILE_PNG -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS -/*returns 1 if success, 0 if failure ==> nothing done*/ -static unsigned string_resize(char** out, size_t size) -{ - char* data = (char*)lodepng_realloc(*out, size + 1); - if(data) - { - data[size] = 0; /*null termination char*/ - *out = data; - } - return data != 0; -} - -/*init a {char*, size_t} pair for use as string*/ -static void string_init(char** out) -{ - *out = NULL; - string_resize(out, 0); -} - -/*free the above pair again*/ -static void string_cleanup(char** out) -{ - lodepng_free(*out); - *out = NULL; -} - -static void string_set(char** out, const char* in) -{ - size_t insize = strlen(in), i = 0; - if(string_resize(out, insize)) - { - for(i = 0; i < insize; i++) - { - (*out)[i] = in[i]; - } - } -} -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ -#endif /*LODEPNG_COMPILE_PNG*/ - -/* ////////////////////////////////////////////////////////////////////////// */ - -unsigned lodepng_read32bitInt(const unsigned char* buffer) -{ - return (unsigned)((buffer[0] << 24) | (buffer[1] << 16) | (buffer[2] << 8) | buffer[3]); -} - -#if defined(LODEPNG_COMPILE_PNG) || defined(LODEPNG_COMPILE_ENCODER) -/*buffer must have at least 4 allocated bytes available*/ -static void lodepng_set32bitInt(unsigned char* buffer, unsigned value) -{ - buffer[0] = (unsigned char)((value >> 24) & 0xff); - buffer[1] = (unsigned char)((value >> 16) & 0xff); - buffer[2] = (unsigned char)((value >> 8) & 0xff); - buffer[3] = (unsigned char)((value ) & 0xff); -} -#endif /*defined(LODEPNG_COMPILE_PNG) || defined(LODEPNG_COMPILE_ENCODER)*/ - -#ifdef LODEPNG_COMPILE_ENCODER -static void lodepng_add32bitInt(ucvector* buffer, unsigned value) -{ - ucvector_resize(buffer, buffer->size + 4); /*todo: give error if resize failed*/ - lodepng_set32bitInt(&buffer->data[buffer->size - 4], value); -} -#endif /*LODEPNG_COMPILE_ENCODER*/ - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / File IO / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -#ifdef LODEPNG_COMPILE_DISK - -unsigned lodepng_load_file(unsigned char** out, size_t* outsize, const char* filename) -{ - FILE* file; - long size; - - /*provide some proper output values if error will happen*/ - *out = 0; - *outsize = 0; - - file = fopen(filename, "rb"); - if(!file) return 78; - - /*get filesize:*/ - fseek(file , 0 , SEEK_END); - size = ftell(file); - rewind(file); - - /*read contents of the file into the vector*/ - *outsize = 0; - *out = (unsigned char*)lodepng_malloc((size_t)size); - if(size && (*out)) (*outsize) = fread(*out, 1, (size_t)size, file); - - fclose(file); - if(!(*out) && size) return 83; /*the above malloc failed*/ - return 0; -} - -/*write given buffer to the file, overwriting the file, it doesn't append to it.*/ -unsigned lodepng_save_file(const unsigned char* buffer, size_t buffersize, const char* filename) -{ - FILE* file; - file = fopen(filename, "wb" ); - if(!file) return 79; - fwrite((char*)buffer , 1 , buffersize, file); - fclose(file); - return 0; -} - -#endif /*LODEPNG_COMPILE_DISK*/ - -/* ////////////////////////////////////////////////////////////////////////// */ -/* ////////////////////////////////////////////////////////////////////////// */ -/* // End of common code and tools. Begin of Zlib related code. // */ -/* ////////////////////////////////////////////////////////////////////////// */ -/* ////////////////////////////////////////////////////////////////////////// */ - -#ifdef LODEPNG_COMPILE_ZLIB -#ifdef LODEPNG_COMPILE_ENCODER -/*TODO: this ignores potential out of memory errors*/ -#define addBitToStream(/*size_t**/ bitpointer, /*ucvector**/ bitstream, /*unsigned char*/ bit)\ -{\ - /*add a new byte at the end*/\ - if(((*bitpointer) & 7) == 0) ucvector_push_back(bitstream, (unsigned char)0);\ - /*earlier bit of huffman code is in a lesser significant bit of an earlier byte*/\ - (bitstream->data[bitstream->size - 1]) |= (bit << ((*bitpointer) & 0x7));\ - (*bitpointer)++;\ -} - -static void addBitsToStream(size_t* bitpointer, ucvector* bitstream, unsigned value, size_t nbits) -{ - size_t i; - for(i = 0; i < nbits; i++) addBitToStream(bitpointer, bitstream, (unsigned char)((value >> i) & 1)); -} - -static void addBitsToStreamReversed(size_t* bitpointer, ucvector* bitstream, unsigned value, size_t nbits) -{ - size_t i; - for(i = 0; i < nbits; i++) addBitToStream(bitpointer, bitstream, (unsigned char)((value >> (nbits - 1 - i)) & 1)); -} -#endif /*LODEPNG_COMPILE_ENCODER*/ - -#ifdef LODEPNG_COMPILE_DECODER - -#define READBIT(bitpointer, bitstream) ((bitstream[bitpointer >> 3] >> (bitpointer & 0x7)) & (unsigned char)1) - -static unsigned char readBitFromStream(size_t* bitpointer, const unsigned char* bitstream) -{ - unsigned char result = (unsigned char)(READBIT(*bitpointer, bitstream)); - (*bitpointer)++; - return result; -} - -static unsigned readBitsFromStream(size_t* bitpointer, const unsigned char* bitstream, size_t nbits) -{ - unsigned result = 0, i; - for(i = 0; i < nbits; i++) - { - result += ((unsigned)READBIT(*bitpointer, bitstream)) << i; - (*bitpointer)++; - } - return result; -} -#endif /*LODEPNG_COMPILE_DECODER*/ - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / Deflate - Huffman / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -#define FIRST_LENGTH_CODE_INDEX 257 -#define LAST_LENGTH_CODE_INDEX 285 -/*256 literals, the end code, some length codes, and 2 unused codes*/ -#define NUM_DEFLATE_CODE_SYMBOLS 288 -/*the distance codes have their own symbols, 30 used, 2 unused*/ -#define NUM_DISTANCE_SYMBOLS 32 -/*the code length codes. 0-15: code lengths, 16: copy previous 3-6 times, 17: 3-10 zeros, 18: 11-138 zeros*/ -#define NUM_CODE_LENGTH_CODES 19 - -/*the base lengths represented by codes 257-285*/ -static const unsigned LENGTHBASE[29] - = {3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31, 35, 43, 51, 59, - 67, 83, 99, 115, 131, 163, 195, 227, 258}; - -/*the extra bits used by codes 257-285 (added to base length)*/ -static const unsigned LENGTHEXTRA[29] - = {0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, - 4, 4, 4, 4, 5, 5, 5, 5, 0}; - -/*the base backwards distances (the bits of distance codes appear after length codes and use their own huffman tree)*/ -static const unsigned DISTANCEBASE[30] - = {1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193, 257, 385, 513, - 769, 1025, 1537, 2049, 3073, 4097, 6145, 8193, 12289, 16385, 24577}; - -/*the extra bits of backwards distances (added to base)*/ -static const unsigned DISTANCEEXTRA[30] - = {0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, - 8, 9, 9, 10, 10, 11, 11, 12, 12, 13, 13}; - -/*the order in which "code length alphabet code lengths" are stored, out of this -the huffman tree of the dynamic huffman tree lengths is generated*/ -static const unsigned CLCL_ORDER[NUM_CODE_LENGTH_CODES] - = {16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15}; - -/* ////////////////////////////////////////////////////////////////////////// */ - -/* -Huffman tree struct, containing multiple representations of the tree -*/ -typedef struct HuffmanTree -{ - unsigned* tree2d; - unsigned* tree1d; - unsigned* lengths; /*the lengths of the codes of the 1d-tree*/ - unsigned maxbitlen; /*maximum number of bits a single code can get*/ - unsigned numcodes; /*number of symbols in the alphabet = number of codes*/ -} HuffmanTree; - -/*function used for debug purposes to draw the tree in ascii art with C++*/ -/* -static void HuffmanTree_draw(HuffmanTree* tree) -{ - std::cout << "tree. length: " << tree->numcodes << " maxbitlen: " << tree->maxbitlen << std::endl; - for(size_t i = 0; i < tree->tree1d.size; i++) - { - if(tree->lengths.data[i]) - std::cout << i << " " << tree->tree1d.data[i] << " " << tree->lengths.data[i] << std::endl; - } - std::cout << std::endl; -}*/ - -static void HuffmanTree_init(HuffmanTree* tree) -{ - tree->tree2d = 0; - tree->tree1d = 0; - tree->lengths = 0; -} - -static void HuffmanTree_cleanup(HuffmanTree* tree) -{ - lodepng_free(tree->tree2d); - lodepng_free(tree->tree1d); - lodepng_free(tree->lengths); -} - -/*the tree representation used by the decoder. return value is error*/ -static unsigned HuffmanTree_make2DTree(HuffmanTree* tree) -{ - unsigned nodefilled = 0; /*up to which node it is filled*/ - unsigned treepos = 0; /*position in the tree (1 of the numcodes columns)*/ - unsigned n, i; - - tree->tree2d = (unsigned*)lodepng_malloc(tree->numcodes * 2 * sizeof(unsigned)); - if(!tree->tree2d) return 83; /*alloc fail*/ - - /* - convert tree1d[] to tree2d[][]. In the 2D array, a value of 32767 means - uninited, a value >= numcodes is an address to another bit, a value < numcodes - is a code. The 2 rows are the 2 possible bit values (0 or 1), there are as - many columns as codes - 1. - A good huffmann tree has N * 2 - 1 nodes, of which N - 1 are internal nodes. - Here, the internal nodes are stored (what their 0 and 1 option point to). - There is only memory for such good tree currently, if there are more nodes - (due to too long length codes), error 55 will happen - */ - for(n = 0; n < tree->numcodes * 2; n++) - { - tree->tree2d[n] = 32767; /*32767 here means the tree2d isn't filled there yet*/ - } - - for(n = 0; n < tree->numcodes; n++) /*the codes*/ - { - for(i = 0; i < tree->lengths[n]; i++) /*the bits for this code*/ - { - unsigned char bit = (unsigned char)((tree->tree1d[n] >> (tree->lengths[n] - i - 1)) & 1); - if(treepos > tree->numcodes - 2) return 55; /*oversubscribed, see comment in lodepng_error_text*/ - if(tree->tree2d[2 * treepos + bit] == 32767) /*not yet filled in*/ - { - if(i + 1 == tree->lengths[n]) /*last bit*/ - { - tree->tree2d[2 * treepos + bit] = n; /*put the current code in it*/ - treepos = 0; - } - else - { - /*put address of the next step in here, first that address has to be found of course - (it's just nodefilled + 1)...*/ - nodefilled++; - /*addresses encoded with numcodes added to it*/ - tree->tree2d[2 * treepos + bit] = nodefilled + tree->numcodes; - treepos = nodefilled; - } - } - else treepos = tree->tree2d[2 * treepos + bit] - tree->numcodes; - } - } - - for(n = 0; n < tree->numcodes * 2; n++) - { - if(tree->tree2d[n] == 32767) tree->tree2d[n] = 0; /*remove possible remaining 32767's*/ - } - - return 0; -} - -/* -Second step for the ...makeFromLengths and ...makeFromFrequencies functions. -numcodes, lengths and maxbitlen must already be filled in correctly. return -value is error. -*/ -static unsigned HuffmanTree_makeFromLengths2(HuffmanTree* tree) -{ - uivector blcount; - uivector nextcode; - unsigned bits, n, error = 0; - - uivector_init(&blcount); - uivector_init(&nextcode); - - tree->tree1d = (unsigned*)lodepng_malloc(tree->numcodes * sizeof(unsigned)); - if(!tree->tree1d) error = 83; /*alloc fail*/ - - if(!uivector_resizev(&blcount, tree->maxbitlen + 1, 0) - || !uivector_resizev(&nextcode, tree->maxbitlen + 1, 0)) - error = 83; /*alloc fail*/ - - if(!error) - { - /*step 1: count number of instances of each code length*/ - for(bits = 0; bits < tree->numcodes; bits++) blcount.data[tree->lengths[bits]]++; - /*step 2: generate the nextcode values*/ - for(bits = 1; bits <= tree->maxbitlen; bits++) - { - nextcode.data[bits] = (nextcode.data[bits - 1] + blcount.data[bits - 1]) << 1; - } - /*step 3: generate all the codes*/ - for(n = 0; n < tree->numcodes; n++) - { - if(tree->lengths[n] != 0) tree->tree1d[n] = nextcode.data[tree->lengths[n]]++; - } - } - - uivector_cleanup(&blcount); - uivector_cleanup(&nextcode); - - if(!error) return HuffmanTree_make2DTree(tree); - else return error; -} - -/* -given the code lengths (as stored in the PNG file), generate the tree as defined -by Deflate. maxbitlen is the maximum bits that a code in the tree can have. -return value is error. -*/ -static unsigned HuffmanTree_makeFromLengths(HuffmanTree* tree, const unsigned* bitlen, - size_t numcodes, unsigned maxbitlen) -{ - unsigned i; - tree->lengths = (unsigned*)lodepng_malloc(numcodes * sizeof(unsigned)); - if(!tree->lengths) return 83; /*alloc fail*/ - for(i = 0; i < numcodes; i++) tree->lengths[i] = bitlen[i]; - tree->numcodes = (unsigned)numcodes; /*number of symbols*/ - tree->maxbitlen = maxbitlen; - return HuffmanTree_makeFromLengths2(tree); -} - -#ifdef LODEPNG_COMPILE_ENCODER - -/* -A coin, this is the terminology used for the package-merge algorithm and the -coin collector's problem. This is used to generate the huffman tree. -A coin can be multiple coins (when they're merged) -*/ -typedef struct Coin -{ - uivector symbols; - float weight; /*the sum of all weights in this coin*/ -} Coin; - -static void coin_init(Coin* c) -{ - uivector_init(&c->symbols); -} - -/*argument c is void* so that this dtor can be given as function pointer to the vector resize function*/ -static void coin_cleanup(void* c) -{ - uivector_cleanup(&((Coin*)c)->symbols); -} - -static void coin_copy(Coin* c1, const Coin* c2) -{ - c1->weight = c2->weight; - uivector_copy(&c1->symbols, &c2->symbols); -} - -static void add_coins(Coin* c1, const Coin* c2) -{ - size_t i; - for(i = 0; i < c2->symbols.size; i++) uivector_push_back(&c1->symbols, c2->symbols.data[i]); - c1->weight += c2->weight; -} - -static void init_coins(Coin* coins, size_t num) -{ - size_t i; - for(i = 0; i < num; i++) coin_init(&coins[i]); -} - -static void cleanup_coins(Coin* coins, size_t num) -{ - size_t i; - for(i = 0; i < num; i++) coin_cleanup(&coins[i]); -} - -static int coin_compare(const void* a, const void* b) { - float wa = ((const Coin*)a)->weight; - float wb = ((const Coin*)b)->weight; - return wa > wb ? 1 : wa < wb ? -1 : 0; -} - -static unsigned append_symbol_coins(Coin* coins, const unsigned* frequencies, unsigned numcodes, size_t sum) -{ - unsigned i; - unsigned j = 0; /*index of present symbols*/ - for(i = 0; i < numcodes; i++) - { - if(frequencies[i] != 0) /*only include symbols that are present*/ - { - coins[j].weight = frequencies[i] / (float)sum; - uivector_push_back(&coins[j].symbols, i); - j++; - } - } - return 0; -} - -unsigned lodepng_huffman_code_lengths(unsigned* lengths, const unsigned* frequencies, - size_t numcodes, unsigned maxbitlen) -{ - unsigned i, j; - size_t sum = 0, numpresent = 0; - unsigned error = 0; - Coin* coins; /*the coins of the currently calculated row*/ - Coin* prev_row; /*the previous row of coins*/ - size_t numcoins; - size_t coinmem; - - if(numcodes == 0) return 80; /*error: a tree of 0 symbols is not supposed to be made*/ - - for(i = 0; i < numcodes; i++) - { - if(frequencies[i] > 0) - { - numpresent++; - sum += frequencies[i]; - } - } - - for(i = 0; i < numcodes; i++) lengths[i] = 0; - - /*ensure at least two present symbols. There should be at least one symbol - according to RFC 1951 section 3.2.7. To decoders incorrectly require two. To - make these work as well ensure there are at least two symbols. The - Package-Merge code below also doesn't work correctly if there's only one - symbol, it'd give it the theoritical 0 bits but in practice zlib wants 1 bit*/ - if(numpresent == 0) - { - lengths[0] = lengths[1] = 1; /*note that for RFC 1951 section 3.2.7, only lengths[0] = 1 is needed*/ - } - else if(numpresent == 1) - { - for(i = 0; i < numcodes; i++) - { - if(frequencies[i]) - { - lengths[i] = 1; - lengths[i == 0 ? 1 : 0] = 1; - break; - } - } - } - else - { - /*Package-Merge algorithm represented by coin collector's problem - For every symbol, maxbitlen coins will be created*/ - - coinmem = numpresent * 2; /*max amount of coins needed with the current algo*/ - coins = (Coin*)lodepng_malloc(sizeof(Coin) * coinmem); - prev_row = (Coin*)lodepng_malloc(sizeof(Coin) * coinmem); - if(!coins || !prev_row) - { - lodepng_free(coins); - lodepng_free(prev_row); - return 83; /*alloc fail*/ - } - init_coins(coins, coinmem); - init_coins(prev_row, coinmem); - - /*first row, lowest denominator*/ - error = append_symbol_coins(coins, frequencies, numcodes, sum); - numcoins = numpresent; - qsort(coins, numcoins, sizeof(Coin), coin_compare); - if(!error) - { - unsigned numprev = 0; - for(j = 1; j <= maxbitlen && !error; j++) /*each of the remaining rows*/ - { - unsigned tempnum; - Coin* tempcoins; - /*swap prev_row and coins, and their amounts*/ - tempcoins = prev_row; prev_row = coins; coins = tempcoins; - tempnum = numprev; numprev = numcoins; numcoins = tempnum; - - cleanup_coins(coins, numcoins); - init_coins(coins, numcoins); - - numcoins = 0; - - /*fill in the merged coins of the previous row*/ - for(i = 0; i + 1 < numprev; i += 2) - { - /*merge prev_row[i] and prev_row[i + 1] into new coin*/ - Coin* coin = &coins[numcoins++]; - coin_copy(coin, &prev_row[i]); - add_coins(coin, &prev_row[i + 1]); - } - /*fill in all the original symbols again*/ - if(j < maxbitlen) - { - error = append_symbol_coins(coins + numcoins, frequencies, numcodes, sum); - numcoins += numpresent; - } - qsort(coins, numcoins, sizeof(Coin), coin_compare); - } - } - - if(!error) - { - /*calculate the lenghts of each symbol, as the amount of times a coin of each symbol is used*/ - for(i = 0; i < numpresent - 1; i++) - { - Coin* coin = &coins[i]; - for(j = 0; j < coin->symbols.size; j++) lengths[coin->symbols.data[j]]++; - } - } - - cleanup_coins(coins, coinmem); - lodepng_free(coins); - cleanup_coins(prev_row, coinmem); - lodepng_free(prev_row); - } - - return error; -} - -/*Create the Huffman tree given the symbol frequencies*/ -static unsigned HuffmanTree_makeFromFrequencies(HuffmanTree* tree, const unsigned* frequencies, - size_t mincodes, size_t numcodes, unsigned maxbitlen) -{ - unsigned error = 0; - while(!frequencies[numcodes - 1] && numcodes > mincodes) numcodes--; /*trim zeroes*/ - tree->maxbitlen = maxbitlen; - tree->numcodes = (unsigned)numcodes; /*number of symbols*/ - tree->lengths = (unsigned*)lodepng_realloc(tree->lengths, numcodes * sizeof(unsigned)); - if(!tree->lengths) return 83; /*alloc fail*/ - /*initialize all lengths to 0*/ - memset(tree->lengths, 0, numcodes * sizeof(unsigned)); - - error = lodepng_huffman_code_lengths(tree->lengths, frequencies, numcodes, maxbitlen); - if(!error) error = HuffmanTree_makeFromLengths2(tree); - return error; -} - -static unsigned HuffmanTree_getCode(const HuffmanTree* tree, unsigned index) -{ - return tree->tree1d[index]; -} - -static unsigned HuffmanTree_getLength(const HuffmanTree* tree, unsigned index) -{ - return tree->lengths[index]; -} -#endif /*LODEPNG_COMPILE_ENCODER*/ - -/*get the literal and length code tree of a deflated block with fixed tree, as per the deflate specification*/ -static unsigned generateFixedLitLenTree(HuffmanTree* tree) -{ - unsigned i, error = 0; - unsigned* bitlen = (unsigned*)lodepng_malloc(NUM_DEFLATE_CODE_SYMBOLS * sizeof(unsigned)); - if(!bitlen) return 83; /*alloc fail*/ - - /*288 possible codes: 0-255=literals, 256=endcode, 257-285=lengthcodes, 286-287=unused*/ - for(i = 0; i <= 143; i++) bitlen[i] = 8; - for(i = 144; i <= 255; i++) bitlen[i] = 9; - for(i = 256; i <= 279; i++) bitlen[i] = 7; - for(i = 280; i <= 287; i++) bitlen[i] = 8; - - error = HuffmanTree_makeFromLengths(tree, bitlen, NUM_DEFLATE_CODE_SYMBOLS, 15); - - lodepng_free(bitlen); - return error; -} - -/*get the distance code tree of a deflated block with fixed tree, as specified in the deflate specification*/ -static unsigned generateFixedDistanceTree(HuffmanTree* tree) -{ - unsigned i, error = 0; - unsigned* bitlen = (unsigned*)lodepng_malloc(NUM_DISTANCE_SYMBOLS * sizeof(unsigned)); - if(!bitlen) return 83; /*alloc fail*/ - - /*there are 32 distance codes, but 30-31 are unused*/ - for(i = 0; i < NUM_DISTANCE_SYMBOLS; i++) bitlen[i] = 5; - error = HuffmanTree_makeFromLengths(tree, bitlen, NUM_DISTANCE_SYMBOLS, 15); - - lodepng_free(bitlen); - return error; -} - -#ifdef LODEPNG_COMPILE_DECODER - -/* -returns the code, or (unsigned)(-1) if error happened -inbitlength is the length of the complete buffer, in bits (so its byte length times 8) -*/ -static unsigned huffmanDecodeSymbol(const unsigned char* in, size_t* bp, - const HuffmanTree* codetree, size_t inbitlength) -{ - unsigned treepos = 0, ct; - for(;;) - { - if(*bp >= inbitlength) return (unsigned)(-1); /*error: end of input memory reached without endcode*/ - /* - decode the symbol from the tree. The "readBitFromStream" code is inlined in - the expression below because this is the biggest bottleneck while decoding - */ - ct = codetree->tree2d[(treepos << 1) + READBIT(*bp, in)]; - (*bp)++; - if(ct < codetree->numcodes) return ct; /*the symbol is decoded, return it*/ - else treepos = ct - codetree->numcodes; /*symbol not yet decoded, instead move tree position*/ - - if(treepos >= codetree->numcodes) return (unsigned)(-1); /*error: it appeared outside the codetree*/ - } -} -#endif /*LODEPNG_COMPILE_DECODER*/ - -#ifdef LODEPNG_COMPILE_DECODER - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / Inflator (Decompressor) / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -/*get the tree of a deflated block with fixed tree, as specified in the deflate specification*/ -static void getTreeInflateFixed(HuffmanTree* tree_ll, HuffmanTree* tree_d) -{ - /*TODO: check for out of memory errors*/ - generateFixedLitLenTree(tree_ll); - generateFixedDistanceTree(tree_d); -} - -/*get the tree of a deflated block with dynamic tree, the tree itself is also Huffman compressed with a known tree*/ -static unsigned getTreeInflateDynamic(HuffmanTree* tree_ll, HuffmanTree* tree_d, - const unsigned char* in, size_t* bp, size_t inlength) -{ - /*make sure that length values that aren't filled in will be 0, or a wrong tree will be generated*/ - unsigned error = 0; - unsigned n, HLIT, HDIST, HCLEN, i; - size_t inbitlength = inlength * 8; - - /*see comments in deflateDynamic for explanation of the context and these variables, it is analogous*/ - unsigned* bitlen_ll = 0; /*lit,len code lengths*/ - unsigned* bitlen_d = 0; /*dist code lengths*/ - /*code length code lengths ("clcl"), the bit lengths of the huffman tree used to compress bitlen_ll and bitlen_d*/ - unsigned* bitlen_cl = 0; - HuffmanTree tree_cl; /*the code tree for code length codes (the huffman tree for compressed huffman trees)*/ - - if((*bp) >> 3 >= inlength - 2) return 49; /*error: the bit pointer is or will go past the memory*/ - - /*number of literal/length codes + 257. Unlike the spec, the value 257 is added to it here already*/ - HLIT = readBitsFromStream(bp, in, 5) + 257; - /*number of distance codes. Unlike the spec, the value 1 is added to it here already*/ - HDIST = readBitsFromStream(bp, in, 5) + 1; - /*number of code length codes. Unlike the spec, the value 4 is added to it here already*/ - HCLEN = readBitsFromStream(bp, in, 4) + 4; - - HuffmanTree_init(&tree_cl); - - while(!error) - { - /*read the code length codes out of 3 * (amount of code length codes) bits*/ - - bitlen_cl = (unsigned*)lodepng_malloc(NUM_CODE_LENGTH_CODES * sizeof(unsigned)); - if(!bitlen_cl) ERROR_BREAK(83 /*alloc fail*/); - - for(i = 0; i < NUM_CODE_LENGTH_CODES; i++) - { - if(i < HCLEN) bitlen_cl[CLCL_ORDER[i]] = readBitsFromStream(bp, in, 3); - else bitlen_cl[CLCL_ORDER[i]] = 0; /*if not, it must stay 0*/ - } - - error = HuffmanTree_makeFromLengths(&tree_cl, bitlen_cl, NUM_CODE_LENGTH_CODES, 7); - if(error) break; - - /*now we can use this tree to read the lengths for the tree that this function will return*/ - bitlen_ll = (unsigned*)lodepng_malloc(NUM_DEFLATE_CODE_SYMBOLS * sizeof(unsigned)); - bitlen_d = (unsigned*)lodepng_malloc(NUM_DISTANCE_SYMBOLS * sizeof(unsigned)); - if(!bitlen_ll || !bitlen_d) ERROR_BREAK(83 /*alloc fail*/); - for(i = 0; i < NUM_DEFLATE_CODE_SYMBOLS; i++) bitlen_ll[i] = 0; - for(i = 0; i < NUM_DISTANCE_SYMBOLS; i++) bitlen_d[i] = 0; - - /*i is the current symbol we're reading in the part that contains the code lengths of lit/len and dist codes*/ - i = 0; - while(i < HLIT + HDIST) - { - unsigned code = huffmanDecodeSymbol(in, bp, &tree_cl, inbitlength); - if(code <= 15) /*a length code*/ - { - if(i < HLIT) bitlen_ll[i] = code; - else bitlen_d[i - HLIT] = code; - i++; - } - else if(code == 16) /*repeat previous*/ - { - unsigned replength = 3; /*read in the 2 bits that indicate repeat length (3-6)*/ - unsigned value; /*set value to the previous code*/ - - if(*bp >= inbitlength) ERROR_BREAK(50); /*error, bit pointer jumps past memory*/ - if (i == 0) ERROR_BREAK(54); /*can't repeat previous if i is 0*/ - - replength += readBitsFromStream(bp, in, 2); - - if(i < HLIT + 1) value = bitlen_ll[i - 1]; - else value = bitlen_d[i - HLIT - 1]; - /*repeat this value in the next lengths*/ - for(n = 0; n < replength; n++) - { - if(i >= HLIT + HDIST) ERROR_BREAK(13); /*error: i is larger than the amount of codes*/ - if(i < HLIT) bitlen_ll[i] = value; - else bitlen_d[i - HLIT] = value; - i++; - } - } - else if(code == 17) /*repeat "0" 3-10 times*/ - { - unsigned replength = 3; /*read in the bits that indicate repeat length*/ - if(*bp >= inbitlength) ERROR_BREAK(50); /*error, bit pointer jumps past memory*/ - - replength += readBitsFromStream(bp, in, 3); - - /*repeat this value in the next lengths*/ - for(n = 0; n < replength; n++) - { - if(i >= HLIT + HDIST) ERROR_BREAK(14); /*error: i is larger than the amount of codes*/ - - if(i < HLIT) bitlen_ll[i] = 0; - else bitlen_d[i - HLIT] = 0; - i++; - } - } - else if(code == 18) /*repeat "0" 11-138 times*/ - { - unsigned replength = 11; /*read in the bits that indicate repeat length*/ - if(*bp >= inbitlength) ERROR_BREAK(50); /*error, bit pointer jumps past memory*/ - - replength += readBitsFromStream(bp, in, 7); - - /*repeat this value in the next lengths*/ - for(n = 0; n < replength; n++) - { - if(i >= HLIT + HDIST) ERROR_BREAK(15); /*error: i is larger than the amount of codes*/ - - if(i < HLIT) bitlen_ll[i] = 0; - else bitlen_d[i - HLIT] = 0; - i++; - } - } - else /*if(code == (unsigned)(-1))*/ /*huffmanDecodeSymbol returns (unsigned)(-1) in case of error*/ - { - if(code == (unsigned)(-1)) - { - /*return error code 10 or 11 depending on the situation that happened in huffmanDecodeSymbol - (10=no endcode, 11=wrong jump outside of tree)*/ - error = (*bp) > inbitlength ? 10 : 11; - } - else error = 16; /*unexisting code, this can never happen*/ - break; - } - } - if(error) break; - - if(bitlen_ll[256] == 0) ERROR_BREAK(64); /*the length of the end code 256 must be larger than 0*/ - - /*now we've finally got HLIT and HDIST, so generate the code trees, and the function is done*/ - error = HuffmanTree_makeFromLengths(tree_ll, bitlen_ll, NUM_DEFLATE_CODE_SYMBOLS, 15); - if(error) break; - error = HuffmanTree_makeFromLengths(tree_d, bitlen_d, NUM_DISTANCE_SYMBOLS, 15); - - break; /*end of error-while*/ - } - - lodepng_free(bitlen_cl); - lodepng_free(bitlen_ll); - lodepng_free(bitlen_d); - HuffmanTree_cleanup(&tree_cl); - - return error; -} - -/*inflate a block with dynamic of fixed Huffman tree*/ -static unsigned inflateHuffmanBlock(ucvector* out, const unsigned char* in, size_t* bp, - size_t* pos, size_t inlength, unsigned btype) -{ - unsigned error = 0; - HuffmanTree tree_ll; /*the huffman tree for literal and length codes*/ - HuffmanTree tree_d; /*the huffman tree for distance codes*/ - size_t inbitlength = inlength * 8; - - HuffmanTree_init(&tree_ll); - HuffmanTree_init(&tree_d); - - if(btype == 1) getTreeInflateFixed(&tree_ll, &tree_d); - else if(btype == 2) error = getTreeInflateDynamic(&tree_ll, &tree_d, in, bp, inlength); - - while(!error) /*decode all symbols until end reached, breaks at end code*/ - { - /*code_ll is literal, length or end code*/ - unsigned code_ll = huffmanDecodeSymbol(in, bp, &tree_ll, inbitlength); - if(code_ll <= 255) /*literal symbol*/ - { - /*ucvector_push_back would do the same, but for some reason the two lines below run 10% faster*/ - if(!ucvector_resize(out, (*pos) + 1)) ERROR_BREAK(83 /*alloc fail*/); - out->data[*pos] = (unsigned char)code_ll; - (*pos)++; - } - else if(code_ll >= FIRST_LENGTH_CODE_INDEX && code_ll <= LAST_LENGTH_CODE_INDEX) /*length code*/ - { - unsigned code_d, distance; - unsigned numextrabits_l, numextrabits_d; /*extra bits for length and distance*/ - size_t start, forward, backward, length; - - /*part 1: get length base*/ - length = LENGTHBASE[code_ll - FIRST_LENGTH_CODE_INDEX]; - - /*part 2: get extra bits and add the value of that to length*/ - numextrabits_l = LENGTHEXTRA[code_ll - FIRST_LENGTH_CODE_INDEX]; - if(*bp >= inbitlength) ERROR_BREAK(51); /*error, bit pointer will jump past memory*/ - length += readBitsFromStream(bp, in, numextrabits_l); - - /*part 3: get distance code*/ - code_d = huffmanDecodeSymbol(in, bp, &tree_d, inbitlength); - if(code_d > 29) - { - if(code_ll == (unsigned)(-1)) /*huffmanDecodeSymbol returns (unsigned)(-1) in case of error*/ - { - /*return error code 10 or 11 depending on the situation that happened in huffmanDecodeSymbol - (10=no endcode, 11=wrong jump outside of tree)*/ - error = (*bp) > inlength * 8 ? 10 : 11; - } - else error = 18; /*error: invalid distance code (30-31 are never used)*/ - break; - } - distance = DISTANCEBASE[code_d]; - - /*part 4: get extra bits from distance*/ - numextrabits_d = DISTANCEEXTRA[code_d]; - if(*bp >= inbitlength) ERROR_BREAK(51); /*error, bit pointer will jump past memory*/ - - distance += readBitsFromStream(bp, in, numextrabits_d); - - /*part 5: fill in all the out[n] values based on the length and dist*/ - start = (*pos); - if(distance > start) ERROR_BREAK(52); /*too long backward distance*/ - backward = start - distance; - - if(!ucvector_resize(out, (*pos) + length)) ERROR_BREAK(83 /*alloc fail*/); - for(forward = 0; forward < length; forward++) - { - out->data[(*pos)] = out->data[backward]; - (*pos)++; - backward++; - if(backward >= start) backward = start - distance; - } - } - else if(code_ll == 256) - { - break; /*end code, break the loop*/ - } - else /*if(code == (unsigned)(-1))*/ /*huffmanDecodeSymbol returns (unsigned)(-1) in case of error*/ - { - /*return error code 10 or 11 depending on the situation that happened in huffmanDecodeSymbol - (10=no endcode, 11=wrong jump outside of tree)*/ - error = (*bp) > inlength * 8 ? 10 : 11; - break; - } - } - - HuffmanTree_cleanup(&tree_ll); - HuffmanTree_cleanup(&tree_d); - - return error; -} - -static unsigned inflateNoCompression(ucvector* out, const unsigned char* in, size_t* bp, size_t* pos, size_t inlength) -{ - /*go to first boundary of byte*/ - size_t p; - unsigned LEN, NLEN, n, error = 0; - while(((*bp) & 0x7) != 0) (*bp)++; - p = (*bp) / 8; /*byte position*/ - - /*read LEN (2 bytes) and NLEN (2 bytes)*/ - if(p >= inlength - 4) return 52; /*error, bit pointer will jump past memory*/ - LEN = in[p] + 256u * in[p + 1]; p += 2; - NLEN = in[p] + 256u * in[p + 1]; p += 2; - - /*check if 16-bit NLEN is really the one's complement of LEN*/ - if(LEN + NLEN != 65535) return 21; /*error: NLEN is not one's complement of LEN*/ - - if(!ucvector_resize(out, (*pos) + LEN)) return 83; /*alloc fail*/ - - /*read the literal data: LEN bytes are now stored in the out buffer*/ - if(p + LEN > inlength) return 23; /*error: reading outside of in buffer*/ - for(n = 0; n < LEN; n++) out->data[(*pos)++] = in[p++]; - - (*bp) = p * 8; - - return error; -} - -static unsigned lodepng_inflatev(ucvector* out, - const unsigned char* in, size_t insize, - const LodePNGDecompressSettings* settings) -{ - /*bit pointer in the "in" data, current byte is bp >> 3, current bit is bp & 0x7 (from lsb to msb of the byte)*/ - size_t bp = 0; - unsigned BFINAL = 0; - size_t pos = 0; /*byte position in the out buffer*/ - unsigned error = 0; - - (void)settings; - - while(!BFINAL) - { - unsigned BTYPE; - if(bp + 2 >= insize * 8) return 52; /*error, bit pointer will jump past memory*/ - BFINAL = readBitFromStream(&bp, in); - BTYPE = 1u * readBitFromStream(&bp, in); - BTYPE += 2u * readBitFromStream(&bp, in); - - if(BTYPE == 3) return 20; /*error: invalid BTYPE*/ - else if(BTYPE == 0) error = inflateNoCompression(out, in, &bp, &pos, insize); /*no compression*/ - else error = inflateHuffmanBlock(out, in, &bp, &pos, insize, BTYPE); /*compression, BTYPE 01 or 10*/ - - if(error) return error; - } - - return error; -} - -unsigned lodepng_inflate(unsigned char** out, size_t* outsize, - const unsigned char* in, size_t insize, - const LodePNGDecompressSettings* settings) -{ - unsigned error; - ucvector v; - ucvector_init_buffer(&v, *out, *outsize); - error = lodepng_inflatev(&v, in, insize, settings); - *out = v.data; - *outsize = v.size; - return error; -} - -static unsigned inflate(unsigned char** out, size_t* outsize, - const unsigned char* in, size_t insize, - const LodePNGDecompressSettings* settings) -{ - if(settings->custom_inflate) - { - return settings->custom_inflate(out, outsize, in, insize, settings); - } - else - { - return lodepng_inflate(out, outsize, in, insize, settings); - } -} - -#endif /*LODEPNG_COMPILE_DECODER*/ - -#ifdef LODEPNG_COMPILE_ENCODER - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / Deflator (Compressor) / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -static const size_t MAX_SUPPORTED_DEFLATE_LENGTH = 258; - -/*bitlen is the size in bits of the code*/ -static void addHuffmanSymbol(size_t* bp, ucvector* compressed, unsigned code, unsigned bitlen) -{ - addBitsToStreamReversed(bp, compressed, code, bitlen); -} - -/*search the index in the array, that has the largest value smaller than or equal to the given value, -given array must be sorted (if no value is smaller, it returns the size of the given array)*/ -static size_t searchCodeIndex(const unsigned* array, size_t array_size, size_t value) -{ - /*linear search implementation*/ - /*for(size_t i = 1; i < array_size; i++) if(array[i] > value) return i - 1; - return array_size - 1;*/ - - /*binary search implementation (not that much faster) (precondition: array_size > 0)*/ - size_t left = 1; - size_t right = array_size - 1; - while(left <= right) - { - size_t mid = (left + right) / 2; - if(array[mid] <= value) left = mid + 1; /*the value to find is more to the right*/ - else if(array[mid - 1] > value) right = mid - 1; /*the value to find is more to the left*/ - else return mid - 1; - } - return array_size - 1; -} - -static void addLengthDistance(uivector* values, size_t length, size_t distance) -{ - /*values in encoded vector are those used by deflate: - 0-255: literal bytes - 256: end - 257-285: length/distance pair (length code, followed by extra length bits, distance code, extra distance bits) - 286-287: invalid*/ - - unsigned length_code = (unsigned)searchCodeIndex(LENGTHBASE, 29, length); - unsigned extra_length = (unsigned)(length - LENGTHBASE[length_code]); - unsigned dist_code = (unsigned)searchCodeIndex(DISTANCEBASE, 30, distance); - unsigned extra_distance = (unsigned)(distance - DISTANCEBASE[dist_code]); - - uivector_push_back(values, length_code + FIRST_LENGTH_CODE_INDEX); - uivector_push_back(values, extra_length); - uivector_push_back(values, dist_code); - uivector_push_back(values, extra_distance); -} - -/*3 bytes of data get encoded into two bytes. The hash cannot use more than 3 -bytes as input because 3 is the minimum match length for deflate*/ -static const unsigned HASH_NUM_VALUES = 65536; -static const unsigned HASH_BIT_MASK = 65535; /*HASH_NUM_VALUES - 1, but C90 does not like that as initializer*/ - -typedef struct Hash -{ - int* head; /*hash value to head circular pos - can be outdated if went around window*/ - /*circular pos to prev circular pos*/ - unsigned short* chain; - int* val; /*circular pos to hash value*/ - - /*TODO: do this not only for zeros but for any repeated byte. However for PNG - it's always going to be the zeros that dominate, so not important for PNG*/ - int* headz; /*similar to head, but for chainz*/ - unsigned short* chainz; /*those with same amount of zeros*/ - unsigned short* zeros; /*length of zeros streak, used as a second hash chain*/ -} Hash; - -static unsigned hash_init(Hash* hash, unsigned windowsize) -{ - unsigned i; - hash->head = (int*)lodepng_malloc(sizeof(int) * HASH_NUM_VALUES); - hash->val = (int*)lodepng_malloc(sizeof(int) * windowsize); - hash->chain = (unsigned short*)lodepng_malloc(sizeof(unsigned short) * windowsize); - - hash->zeros = (unsigned short*)lodepng_malloc(sizeof(unsigned short) * windowsize); - hash->headz = (int*)lodepng_malloc(sizeof(int) * (MAX_SUPPORTED_DEFLATE_LENGTH + 1)); - hash->chainz = (unsigned short*)lodepng_malloc(sizeof(unsigned short) * windowsize); - - if(!hash->head || !hash->chain || !hash->val || !hash->headz|| !hash->chainz || !hash->zeros) - { - return 83; /*alloc fail*/ - } - - /*initialize hash table*/ - for(i = 0; i < HASH_NUM_VALUES; i++) hash->head[i] = -1; - for(i = 0; i < windowsize; i++) hash->val[i] = -1; - for(i = 0; i < windowsize; i++) hash->chain[i] = i; /*same value as index indicates uninitialized*/ - - for(i = 0; i <= MAX_SUPPORTED_DEFLATE_LENGTH; i++) hash->headz[i] = -1; - for(i = 0; i < windowsize; i++) hash->chainz[i] = i; /*same value as index indicates uninitialized*/ - - return 0; -} - -static void hash_cleanup(Hash* hash) -{ - lodepng_free(hash->head); - lodepng_free(hash->val); - lodepng_free(hash->chain); - - lodepng_free(hash->zeros); - lodepng_free(hash->headz); - lodepng_free(hash->chainz); -} - - - -static unsigned getHash(const unsigned char* data, size_t size, size_t pos) -{ - unsigned result = 0; - if (pos + 2 < size) - { - /*A simple shift and xor hash is used. Since the data of PNGs is dominated - by zeroes due to the filters, a better hash does not have a significant - effect on speed in traversing the chain, and causes more time spend on - calculating the hash.*/ - result ^= (unsigned)(data[pos + 0] << 0u); - result ^= (unsigned)(data[pos + 1] << 4u); - result ^= (unsigned)(data[pos + 2] << 8u); - } else { - size_t amount, i; - if(pos >= size) return 0; - amount = size - pos; - for(i = 0; i < amount; i++) result ^= (unsigned)(data[pos + i] << (i * 8u)); - } - return result & HASH_BIT_MASK; -} - -static unsigned countZeros(const unsigned char* data, size_t size, size_t pos) -{ - const unsigned char* start = data + pos; - const unsigned char* end = start + MAX_SUPPORTED_DEFLATE_LENGTH; - if(end > data + size) end = data + size; - data = start; - while (data != end && *data == 0) data++; - /*subtracting two addresses returned as 32-bit number (max value is MAX_SUPPORTED_DEFLATE_LENGTH)*/ - return (unsigned)(data - start); -} - -/*wpos = pos & (windowsize - 1)*/ -static void updateHashChain(Hash* hash, size_t wpos, unsigned hashval, unsigned short numzeros) -{ - hash->val[wpos] = (int)hashval; - if(hash->head[hashval] != -1) hash->chain[wpos] = hash->head[hashval]; - hash->head[hashval] = wpos; - - hash->zeros[wpos] = numzeros; - if(hash->headz[numzeros] != -1) hash->chainz[wpos] = hash->headz[numzeros]; - hash->headz[numzeros] = wpos; -} - -/* -LZ77-encode the data. Return value is error code. The input are raw bytes, the output -is in the form of unsigned integers with codes representing for example literal bytes, or -length/distance pairs. -It uses a hash table technique to let it encode faster. When doing LZ77 encoding, a -sliding window (of windowsize) is used, and all past bytes in that window can be used as -the "dictionary". A brute force search through all possible distances would be slow, and -this hash technique is one out of several ways to speed this up. -*/ -static unsigned encodeLZ77(uivector* out, Hash* hash, - const unsigned char* in, size_t inpos, size_t insize, unsigned windowsize, - unsigned minmatch, unsigned nicematch, unsigned lazymatching) -{ - size_t pos; - unsigned i, error = 0; - /*for large window lengths, assume the user wants no compression loss. Otherwise, max hash chain length speedup.*/ - unsigned maxchainlength = windowsize >= 8192 ? windowsize : windowsize / 8; - unsigned maxlazymatch = windowsize >= 8192 ? MAX_SUPPORTED_DEFLATE_LENGTH : 64; - - unsigned usezeros = 1; /*not sure if setting it to false for windowsize < 8192 is better or worse*/ - unsigned numzeros = 0; - - unsigned offset; /*the offset represents the distance in LZ77 terminology*/ - unsigned length; - unsigned lazy = 0; - unsigned lazylength = 0, lazyoffset = 0; - unsigned hashval; - unsigned current_offset, current_length; - unsigned prev_offset; - const unsigned char *lastptr, *foreptr, *backptr; - unsigned hashpos; - - if(windowsize <= 0 || windowsize > 32768) return 60; /*error: windowsize smaller/larger than allowed*/ - if((windowsize & (windowsize - 1)) != 0) return 90; /*error: must be power of two*/ - - if(nicematch > MAX_SUPPORTED_DEFLATE_LENGTH) nicematch = MAX_SUPPORTED_DEFLATE_LENGTH; - - for(pos = inpos; pos < insize; pos++) - { - size_t wpos = pos & (windowsize - 1); /*position for in 'circular' hash buffers*/ - unsigned chainlength = 0; - - hashval = getHash(in, insize, pos); - - if(usezeros && hashval == 0) - { - if (numzeros == 0) numzeros = countZeros(in, insize, pos); - else if (pos + numzeros > insize || in[pos + numzeros - 1] != 0) numzeros--; - } - else - { - numzeros = 0; - } - - updateHashChain(hash, wpos, hashval, numzeros); - - /*the length and offset found for the current position*/ - length = 0; - offset = 0; - - hashpos = hash->chain[wpos]; - - lastptr = &in[insize < pos + MAX_SUPPORTED_DEFLATE_LENGTH ? insize : pos + MAX_SUPPORTED_DEFLATE_LENGTH]; - - /*search for the longest string*/ - prev_offset = 0; - for(;;) - { - if(chainlength++ >= maxchainlength) break; - current_offset = hashpos <= wpos ? wpos - hashpos : wpos - hashpos + windowsize; - - if(current_offset < prev_offset) break; /*stop when went completely around the circular buffer*/ - prev_offset = current_offset; - if(current_offset > 0) - { - /*test the next characters*/ - foreptr = &in[pos]; - backptr = &in[pos - current_offset]; - - /*common case in PNGs is lots of zeros. Quickly skip over them as a speedup*/ - if(numzeros >= 3) - { - unsigned skip = hash->zeros[hashpos]; - if(skip > numzeros) skip = numzeros; - backptr += skip; - foreptr += skip; - } - - while(foreptr != lastptr && *backptr == *foreptr) /*maximum supported length by deflate is max length*/ - { - ++backptr; - ++foreptr; - } - current_length = (unsigned)(foreptr - &in[pos]); - - if(current_length > length) - { - length = current_length; /*the longest length*/ - offset = current_offset; /*the offset that is related to this longest length*/ - /*jump out once a length of max length is found (speed gain). This also jumps - out if length is MAX_SUPPORTED_DEFLATE_LENGTH*/ - if(current_length >= nicematch) break; - } - } - - if(hashpos == hash->chain[hashpos]) break; - - if(numzeros >= 3 && length > numzeros) { - hashpos = hash->chainz[hashpos]; - if(hash->zeros[hashpos] != numzeros) break; - } else { - hashpos = hash->chain[hashpos]; - /*outdated hash value, happens if particular value was not encountered in whole last window*/ - if(hash->val[hashpos] != (int)hashval) break; - } - } - - if(lazymatching) - { - if(!lazy && length >= 3 && length <= maxlazymatch && length < MAX_SUPPORTED_DEFLATE_LENGTH) - { - lazy = 1; - lazylength = length; - lazyoffset = offset; - continue; /*try the next byte*/ - } - if(lazy) - { - lazy = 0; - if(pos == 0) ERROR_BREAK(81); - if(length > lazylength + 1) - { - /*push the previous character as literal*/ - if(!uivector_push_back(out, in[pos - 1])) ERROR_BREAK(83 /*alloc fail*/); - } - else - { - length = lazylength; - offset = lazyoffset; - hash->head[hashval] = -1; /*the same hashchain update will be done, this ensures no wrong alteration*/ - hash->headz[numzeros] = -1; /*idem*/ - pos--; - } - } - } - if(length >= 3 && offset > windowsize) ERROR_BREAK(86 /*too big (or overflown negative) offset*/); - - /*encode it as length/distance pair or literal value*/ - if(length < 3) /*only lengths of 3 or higher are supported as length/distance pair*/ - { - if(!uivector_push_back(out, in[pos])) ERROR_BREAK(83 /*alloc fail*/); - } - else if(length < minmatch || (length == 3 && offset > 4096)) - { - /*compensate for the fact that longer offsets have more extra bits, a - length of only 3 may be not worth it then*/ - if(!uivector_push_back(out, in[pos])) ERROR_BREAK(83 /*alloc fail*/); - } - else - { - addLengthDistance(out, length, offset); - for(i = 1; i < length; i++) - { - pos++; - wpos = pos & (windowsize - 1); - hashval = getHash(in, insize, pos); - if(usezeros && hashval == 0) - { - if (numzeros == 0) numzeros = countZeros(in, insize, pos); - else if (pos + numzeros > insize || in[pos + numzeros - 1] != 0) numzeros--; - } - else - { - numzeros = 0; - } - updateHashChain(hash, wpos, hashval, numzeros); - } - } - } /*end of the loop through each character of input*/ - - return error; -} - -/* /////////////////////////////////////////////////////////////////////////// */ - -static unsigned deflateNoCompression(ucvector* out, const unsigned char* data, size_t datasize) -{ - /*non compressed deflate block data: 1 bit BFINAL,2 bits BTYPE,(5 bits): it jumps to start of next byte, - 2 bytes LEN, 2 bytes NLEN, LEN bytes literal DATA*/ - - size_t i, j, numdeflateblocks = (datasize + 65534) / 65535; - unsigned datapos = 0; - for(i = 0; i < numdeflateblocks; i++) - { - unsigned BFINAL, BTYPE, LEN, NLEN; - unsigned char firstbyte; - - BFINAL = (i == numdeflateblocks - 1); - BTYPE = 0; - - firstbyte = (unsigned char)(BFINAL + ((BTYPE & 1) << 1) + ((BTYPE & 2) << 1)); - ucvector_push_back(out, firstbyte); - - LEN = 65535; - if(datasize - datapos < 65535) LEN = (unsigned)datasize - datapos; - NLEN = 65535 - LEN; - - ucvector_push_back(out, (unsigned char)(LEN % 256)); - ucvector_push_back(out, (unsigned char)(LEN / 256)); - ucvector_push_back(out, (unsigned char)(NLEN % 256)); - ucvector_push_back(out, (unsigned char)(NLEN / 256)); - - /*Decompressed data*/ - for(j = 0; j < 65535 && datapos < datasize; j++) - { - ucvector_push_back(out, data[datapos++]); - } - } - - return 0; -} - -/* -write the lz77-encoded data, which has lit, len and dist codes, to compressed stream using huffman trees. -tree_ll: the tree for lit and len codes. -tree_d: the tree for distance codes. -*/ -static void writeLZ77data(size_t* bp, ucvector* out, const uivector* lz77_encoded, - const HuffmanTree* tree_ll, const HuffmanTree* tree_d) -{ - size_t i = 0; - for(i = 0; i < lz77_encoded->size; i++) - { - unsigned val = lz77_encoded->data[i]; - addHuffmanSymbol(bp, out, HuffmanTree_getCode(tree_ll, val), HuffmanTree_getLength(tree_ll, val)); - if(val > 256) /*for a length code, 3 more things have to be added*/ - { - unsigned length_index = val - FIRST_LENGTH_CODE_INDEX; - unsigned n_length_extra_bits = LENGTHEXTRA[length_index]; - unsigned length_extra_bits = lz77_encoded->data[++i]; - - unsigned distance_code = lz77_encoded->data[++i]; - - unsigned distance_index = distance_code; - unsigned n_distance_extra_bits = DISTANCEEXTRA[distance_index]; - unsigned distance_extra_bits = lz77_encoded->data[++i]; - - addBitsToStream(bp, out, length_extra_bits, n_length_extra_bits); - addHuffmanSymbol(bp, out, HuffmanTree_getCode(tree_d, distance_code), - HuffmanTree_getLength(tree_d, distance_code)); - addBitsToStream(bp, out, distance_extra_bits, n_distance_extra_bits); - } - } -} - -/*Deflate for a block of type "dynamic", that is, with freely, optimally, created huffman trees*/ -static unsigned deflateDynamic(ucvector* out, size_t* bp, Hash* hash, - const unsigned char* data, size_t datapos, size_t dataend, - const LodePNGCompressSettings* settings, unsigned final) -{ - unsigned error = 0; - - /* - A block is compressed as follows: The PNG data is lz77 encoded, resulting in - literal bytes and length/distance pairs. This is then huffman compressed with - two huffman trees. One huffman tree is used for the lit and len values ("ll"), - another huffman tree is used for the dist values ("d"). These two trees are - stored using their code lengths, and to compress even more these code lengths - are also run-length encoded and huffman compressed. This gives a huffman tree - of code lengths "cl". The code lenghts used to describe this third tree are - the code length code lengths ("clcl"). - */ - - /*The lz77 encoded data, represented with integers since there will also be length and distance codes in it*/ - uivector lz77_encoded; - HuffmanTree tree_ll; /*tree for lit,len values*/ - HuffmanTree tree_d; /*tree for distance codes*/ - HuffmanTree tree_cl; /*tree for encoding the code lengths representing tree_ll and tree_d*/ - uivector frequencies_ll; /*frequency of lit,len codes*/ - uivector frequencies_d; /*frequency of dist codes*/ - uivector frequencies_cl; /*frequency of code length codes*/ - uivector bitlen_lld; /*lit,len,dist code lenghts (int bits), literally (without repeat codes).*/ - uivector bitlen_lld_e; /*bitlen_lld encoded with repeat codes (this is a rudemtary run length compression)*/ - /*bitlen_cl is the code length code lengths ("clcl"). The bit lengths of codes to represent tree_cl - (these are written as is in the file, it would be crazy to compress these using yet another huffman - tree that needs to be represented by yet another set of code lengths)*/ - uivector bitlen_cl; - size_t datasize = dataend - datapos; - - /* - Due to the huffman compression of huffman tree representations ("two levels"), there are some anologies: - bitlen_lld is to tree_cl what data is to tree_ll and tree_d. - bitlen_lld_e is to bitlen_lld what lz77_encoded is to data. - bitlen_cl is to bitlen_lld_e what bitlen_lld is to lz77_encoded. - */ - - unsigned BFINAL = final; - size_t numcodes_ll, numcodes_d, i; - unsigned HLIT, HDIST, HCLEN; - - uivector_init(&lz77_encoded); - HuffmanTree_init(&tree_ll); - HuffmanTree_init(&tree_d); - HuffmanTree_init(&tree_cl); - uivector_init(&frequencies_ll); - uivector_init(&frequencies_d); - uivector_init(&frequencies_cl); - uivector_init(&bitlen_lld); - uivector_init(&bitlen_lld_e); - uivector_init(&bitlen_cl); - - /*This while loop never loops due to a break at the end, it is here to - allow breaking out of it to the cleanup phase on error conditions.*/ - while(!error) - { - if(settings->use_lz77) - { - error = encodeLZ77(&lz77_encoded, hash, data, datapos, dataend, settings->windowsize, - settings->minmatch, settings->nicematch, settings->lazymatching); - if(error) break; - } - else - { - if(!uivector_resize(&lz77_encoded, datasize)) ERROR_BREAK(83 /*alloc fail*/); - for(i = datapos; i < dataend; i++) lz77_encoded.data[i] = data[i]; /*no LZ77, but still will be Huffman compressed*/ - } - - if(!uivector_resizev(&frequencies_ll, 286, 0)) ERROR_BREAK(83 /*alloc fail*/); - if(!uivector_resizev(&frequencies_d, 30, 0)) ERROR_BREAK(83 /*alloc fail*/); - - /*Count the frequencies of lit, len and dist codes*/ - for(i = 0; i < lz77_encoded.size; i++) - { - unsigned symbol = lz77_encoded.data[i]; - frequencies_ll.data[symbol]++; - if(symbol > 256) - { - unsigned dist = lz77_encoded.data[i + 2]; - frequencies_d.data[dist]++; - i += 3; - } - } - frequencies_ll.data[256] = 1; /*there will be exactly 1 end code, at the end of the block*/ - - /*Make both huffman trees, one for the lit and len codes, one for the dist codes*/ - error = HuffmanTree_makeFromFrequencies(&tree_ll, frequencies_ll.data, 257, frequencies_ll.size, 15); - if(error) break; - /*2, not 1, is chosen for mincodes: some buggy PNG decoders require at least 2 symbols in the dist tree*/ - error = HuffmanTree_makeFromFrequencies(&tree_d, frequencies_d.data, 2, frequencies_d.size, 15); - if(error) break; - - numcodes_ll = tree_ll.numcodes; if(numcodes_ll > 286) numcodes_ll = 286; - numcodes_d = tree_d.numcodes; if(numcodes_d > 30) numcodes_d = 30; - /*store the code lengths of both generated trees in bitlen_lld*/ - for(i = 0; i < numcodes_ll; i++) uivector_push_back(&bitlen_lld, HuffmanTree_getLength(&tree_ll, (unsigned)i)); - for(i = 0; i < numcodes_d; i++) uivector_push_back(&bitlen_lld, HuffmanTree_getLength(&tree_d, (unsigned)i)); - - /*run-length compress bitlen_ldd into bitlen_lld_e by using repeat codes 16 (copy length 3-6 times), - 17 (3-10 zeroes), 18 (11-138 zeroes)*/ - for(i = 0; i < (unsigned)bitlen_lld.size; i++) - { - unsigned j = 0; /*amount of repititions*/ - while(i + j + 1 < (unsigned)bitlen_lld.size && bitlen_lld.data[i + j + 1] == bitlen_lld.data[i]) j++; - - if(bitlen_lld.data[i] == 0 && j >= 2) /*repeat code for zeroes*/ - { - j++; /*include the first zero*/ - if(j <= 10) /*repeat code 17 supports max 10 zeroes*/ - { - uivector_push_back(&bitlen_lld_e, 17); - uivector_push_back(&bitlen_lld_e, j - 3); - } - else /*repeat code 18 supports max 138 zeroes*/ - { - if(j > 138) j = 138; - uivector_push_back(&bitlen_lld_e, 18); - uivector_push_back(&bitlen_lld_e, j - 11); - } - i += (j - 1); - } - else if(j >= 3) /*repeat code for value other than zero*/ - { - size_t k; - unsigned num = j / 6, rest = j % 6; - uivector_push_back(&bitlen_lld_e, bitlen_lld.data[i]); - for(k = 0; k < num; k++) - { - uivector_push_back(&bitlen_lld_e, 16); - uivector_push_back(&bitlen_lld_e, 6 - 3); - } - if(rest >= 3) - { - uivector_push_back(&bitlen_lld_e, 16); - uivector_push_back(&bitlen_lld_e, rest - 3); - } - else j -= rest; - i += j; - } - else /*too short to benefit from repeat code*/ - { - uivector_push_back(&bitlen_lld_e, bitlen_lld.data[i]); - } - } - - /*generate tree_cl, the huffmantree of huffmantrees*/ - - if(!uivector_resizev(&frequencies_cl, NUM_CODE_LENGTH_CODES, 0)) ERROR_BREAK(83 /*alloc fail*/); - for(i = 0; i < bitlen_lld_e.size; i++) - { - frequencies_cl.data[bitlen_lld_e.data[i]]++; - /*after a repeat code come the bits that specify the number of repetitions, - those don't need to be in the frequencies_cl calculation*/ - if(bitlen_lld_e.data[i] >= 16) i++; - } - - error = HuffmanTree_makeFromFrequencies(&tree_cl, frequencies_cl.data, - frequencies_cl.size, frequencies_cl.size, 7); - if(error) break; - - if(!uivector_resize(&bitlen_cl, tree_cl.numcodes)) ERROR_BREAK(83 /*alloc fail*/); - for(i = 0; i < tree_cl.numcodes; i++) - { - /*lenghts of code length tree is in the order as specified by deflate*/ - bitlen_cl.data[i] = HuffmanTree_getLength(&tree_cl, CLCL_ORDER[i]); - } - while(bitlen_cl.data[bitlen_cl.size - 1] == 0 && bitlen_cl.size > 4) - { - /*remove zeros at the end, but minimum size must be 4*/ - if(!uivector_resize(&bitlen_cl, bitlen_cl.size - 1)) ERROR_BREAK(83 /*alloc fail*/); - } - if(error) break; - - /* - Write everything into the output - - After the BFINAL and BTYPE, the dynamic block consists out of the following: - - 5 bits HLIT, 5 bits HDIST, 4 bits HCLEN - - (HCLEN+4)*3 bits code lengths of code length alphabet - - HLIT + 257 code lenghts of lit/length alphabet (encoded using the code length - alphabet, + possible repetition codes 16, 17, 18) - - HDIST + 1 code lengths of distance alphabet (encoded using the code length - alphabet, + possible repetition codes 16, 17, 18) - - compressed data - - 256 (end code) - */ - - /*Write block type*/ - addBitToStream(bp, out, BFINAL); - addBitToStream(bp, out, 0); /*first bit of BTYPE "dynamic"*/ - addBitToStream(bp, out, 1); /*second bit of BTYPE "dynamic"*/ - - /*write the HLIT, HDIST and HCLEN values*/ - HLIT = (unsigned)(numcodes_ll - 257); - HDIST = (unsigned)(numcodes_d - 1); - HCLEN = (unsigned)bitlen_cl.size - 4; - /*trim zeroes for HCLEN. HLIT and HDIST were already trimmed at tree creation*/ - while(!bitlen_cl.data[HCLEN + 4 - 1] && HCLEN > 0) HCLEN--; - addBitsToStream(bp, out, HLIT, 5); - addBitsToStream(bp, out, HDIST, 5); - addBitsToStream(bp, out, HCLEN, 4); - - /*write the code lenghts of the code length alphabet*/ - for(i = 0; i < HCLEN + 4; i++) addBitsToStream(bp, out, bitlen_cl.data[i], 3); - - /*write the lenghts of the lit/len AND the dist alphabet*/ - for(i = 0; i < bitlen_lld_e.size; i++) - { - addHuffmanSymbol(bp, out, HuffmanTree_getCode(&tree_cl, bitlen_lld_e.data[i]), - HuffmanTree_getLength(&tree_cl, bitlen_lld_e.data[i])); - /*extra bits of repeat codes*/ - if(bitlen_lld_e.data[i] == 16) addBitsToStream(bp, out, bitlen_lld_e.data[++i], 2); - else if(bitlen_lld_e.data[i] == 17) addBitsToStream(bp, out, bitlen_lld_e.data[++i], 3); - else if(bitlen_lld_e.data[i] == 18) addBitsToStream(bp, out, bitlen_lld_e.data[++i], 7); - } - - /*write the compressed data symbols*/ - writeLZ77data(bp, out, &lz77_encoded, &tree_ll, &tree_d); - /*error: the length of the end code 256 must be larger than 0*/ - if(HuffmanTree_getLength(&tree_ll, 256) == 0) ERROR_BREAK(64); - - /*write the end code*/ - addHuffmanSymbol(bp, out, HuffmanTree_getCode(&tree_ll, 256), HuffmanTree_getLength(&tree_ll, 256)); - - break; /*end of error-while*/ - } - - /*cleanup*/ - uivector_cleanup(&lz77_encoded); - HuffmanTree_cleanup(&tree_ll); - HuffmanTree_cleanup(&tree_d); - HuffmanTree_cleanup(&tree_cl); - uivector_cleanup(&frequencies_ll); - uivector_cleanup(&frequencies_d); - uivector_cleanup(&frequencies_cl); - uivector_cleanup(&bitlen_lld_e); - uivector_cleanup(&bitlen_lld); - uivector_cleanup(&bitlen_cl); - - return error; -} - -static unsigned deflateFixed(ucvector* out, size_t* bp, Hash* hash, - const unsigned char* data, - size_t datapos, size_t dataend, - const LodePNGCompressSettings* settings, unsigned final) -{ - HuffmanTree tree_ll; /*tree for literal values and length codes*/ - HuffmanTree tree_d; /*tree for distance codes*/ - - unsigned BFINAL = final; - unsigned error = 0; - size_t i; - - HuffmanTree_init(&tree_ll); - HuffmanTree_init(&tree_d); - - generateFixedLitLenTree(&tree_ll); - generateFixedDistanceTree(&tree_d); - - addBitToStream(bp, out, BFINAL); - addBitToStream(bp, out, 1); /*first bit of BTYPE*/ - addBitToStream(bp, out, 0); /*second bit of BTYPE*/ - - if(settings->use_lz77) /*LZ77 encoded*/ - { - uivector lz77_encoded; - uivector_init(&lz77_encoded); - error = encodeLZ77(&lz77_encoded, hash, data, datapos, dataend, settings->windowsize, - settings->minmatch, settings->nicematch, settings->lazymatching); - if(!error) writeLZ77data(bp, out, &lz77_encoded, &tree_ll, &tree_d); - uivector_cleanup(&lz77_encoded); - } - else /*no LZ77, but still will be Huffman compressed*/ - { - for(i = datapos; i < dataend; i++) - { - addHuffmanSymbol(bp, out, HuffmanTree_getCode(&tree_ll, data[i]), HuffmanTree_getLength(&tree_ll, data[i])); - } - } - /*add END code*/ - if(!error) addHuffmanSymbol(bp, out, HuffmanTree_getCode(&tree_ll, 256), HuffmanTree_getLength(&tree_ll, 256)); - - /*cleanup*/ - HuffmanTree_cleanup(&tree_ll); - HuffmanTree_cleanup(&tree_d); - - return error; -} - -static unsigned lodepng_deflatev(ucvector* out, const unsigned char* in, size_t insize, - const LodePNGCompressSettings* settings) -{ - unsigned error = 0; - size_t i, blocksize, numdeflateblocks; - size_t bp = 0; /*the bit pointer*/ - Hash hash; - - if(settings->btype > 2) return 61; - else if(settings->btype == 0) return deflateNoCompression(out, in, insize); - else if(settings->btype == 1) blocksize = insize; - else /*if(settings->btype == 2)*/ - { - blocksize = insize / 8 + 8; - if(blocksize < 65535) blocksize = 65535; - } - - numdeflateblocks = (insize + blocksize - 1) / blocksize; - if(numdeflateblocks == 0) numdeflateblocks = 1; - - error = hash_init(&hash, settings->windowsize); - if(error) return error; - - for(i = 0; i < numdeflateblocks && !error; i++) - { - unsigned final = (i == numdeflateblocks - 1); - size_t start = i * blocksize; - size_t end = start + blocksize; - if(end > insize) end = insize; - - if(settings->btype == 1) error = deflateFixed(out, &bp, &hash, in, start, end, settings, final); - else if(settings->btype == 2) error = deflateDynamic(out, &bp, &hash, in, start, end, settings, final); - } - - hash_cleanup(&hash); - - return error; -} - -unsigned lodepng_deflate(unsigned char** out, size_t* outsize, - const unsigned char* in, size_t insize, - const LodePNGCompressSettings* settings) -{ - unsigned error; - ucvector v; - ucvector_init_buffer(&v, *out, *outsize); - error = lodepng_deflatev(&v, in, insize, settings); - *out = v.data; - *outsize = v.size; - return error; -} - -static unsigned deflate(unsigned char** out, size_t* outsize, - const unsigned char* in, size_t insize, - const LodePNGCompressSettings* settings) -{ - if(settings->custom_deflate) - { - return settings->custom_deflate(out, outsize, in, insize, settings); - } - else - { - return lodepng_deflate(out, outsize, in, insize, settings); - } -} - -#endif /*LODEPNG_COMPILE_DECODER*/ - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / Adler32 */ -/* ////////////////////////////////////////////////////////////////////////// */ - -static unsigned update_adler32(unsigned adler, const unsigned char* data, unsigned len) -{ - unsigned s1 = adler & 0xffff; - unsigned s2 = (adler >> 16) & 0xffff; - - while(len > 0) - { - /*at least 5550 sums can be done before the sums overflow, saving a lot of module divisions*/ - unsigned amount = len > 5550 ? 5550 : len; - len -= amount; - while(amount > 0) - { - s1 += (*data++); - s2 += s1; - amount--; - } - s1 %= 65521; - s2 %= 65521; - } - - return (s2 << 16) | s1; -} - -/*Return the adler32 of the bytes data[0..len-1]*/ -static unsigned adler32(const unsigned char* data, unsigned len) -{ - return update_adler32(1L, data, len); -} - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / Zlib / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -#ifdef LODEPNG_COMPILE_DECODER - -unsigned lodepng_zlib_decompress(unsigned char** out, size_t* outsize, const unsigned char* in, - size_t insize, const LodePNGDecompressSettings* settings) -{ - unsigned error = 0; - unsigned CM, CINFO, FDICT; - - if(insize < 2) return 53; /*error, size of zlib data too small*/ - /*read information from zlib header*/ - if((in[0] * 256 + in[1]) % 31 != 0) - { - /*error: 256 * in[0] + in[1] must be a multiple of 31, the FCHECK value is supposed to be made that way*/ - return 24; - } - - CM = in[0] & 15; - CINFO = (in[0] >> 4) & 15; - /*FCHECK = in[1] & 31;*/ /*FCHECK is already tested above*/ - FDICT = (in[1] >> 5) & 1; - /*FLEVEL = (in[1] >> 6) & 3;*/ /*FLEVEL is not used here*/ - - if(CM != 8 || CINFO > 7) - { - /*error: only compression method 8: inflate with sliding window of 32k is supported by the PNG spec*/ - return 25; - } - if(FDICT != 0) - { - /*error: the specification of PNG says about the zlib stream: - "The additional flags shall not specify a preset dictionary."*/ - return 26; - } - - error = inflate(out, outsize, in + 2, insize - 2, settings); - if(error) return error; - - if(!settings->ignore_adler32) - { - unsigned ADLER32 = lodepng_read32bitInt(&in[insize - 4]); - unsigned checksum = adler32(*out, (unsigned)(*outsize)); - if(checksum != ADLER32) return 58; /*error, adler checksum not correct, data must be corrupted*/ - } - - return 0; /*no error*/ -} - -static unsigned zlib_decompress(unsigned char** out, size_t* outsize, const unsigned char* in, - size_t insize, const LodePNGDecompressSettings* settings) -{ - if(settings->custom_zlib) - { - return settings->custom_zlib(out, outsize, in, insize, settings); - } - else - { - return lodepng_zlib_decompress(out, outsize, in, insize, settings); - } -} - -#endif /*LODEPNG_COMPILE_DECODER*/ - -#ifdef LODEPNG_COMPILE_ENCODER - -unsigned lodepng_zlib_compress(unsigned char** out, size_t* outsize, const unsigned char* in, - size_t insize, const LodePNGCompressSettings* settings) -{ - /*initially, *out must be NULL and outsize 0, if you just give some random *out - that's pointing to a non allocated buffer, this'll crash*/ - ucvector outv; - size_t i; - unsigned error; - unsigned char* deflatedata = 0; - size_t deflatesize = 0; - - unsigned ADLER32; - /*zlib data: 1 byte CMF (CM+CINFO), 1 byte FLG, deflate data, 4 byte ADLER32 checksum of the Decompressed data*/ - unsigned CMF = 120; /*0b01111000: CM 8, CINFO 7. With CINFO 7, any window size up to 32768 can be used.*/ - unsigned FLEVEL = 0; - unsigned FDICT = 0; - unsigned CMFFLG = 256 * CMF + FDICT * 32 + FLEVEL * 64; - unsigned FCHECK = 31 - CMFFLG % 31; - CMFFLG += FCHECK; - - /*ucvector-controlled version of the output buffer, for dynamic array*/ - ucvector_init_buffer(&outv, *out, *outsize); - - ucvector_push_back(&outv, (unsigned char)(CMFFLG / 256)); - ucvector_push_back(&outv, (unsigned char)(CMFFLG % 256)); - - error = deflate(&deflatedata, &deflatesize, in, insize, settings); - - if(!error) - { - ADLER32 = adler32(in, (unsigned)insize); - for(i = 0; i < deflatesize; i++) ucvector_push_back(&outv, deflatedata[i]); - lodepng_free(deflatedata); - lodepng_add32bitInt(&outv, ADLER32); - } - - *out = outv.data; - *outsize = outv.size; - - return error; -} - -/* compress using the default or custom zlib function */ -static unsigned zlib_compress(unsigned char** out, size_t* outsize, const unsigned char* in, - size_t insize, const LodePNGCompressSettings* settings) -{ - if(settings->custom_zlib) - { - return settings->custom_zlib(out, outsize, in, insize, settings); - } - else - { - return lodepng_zlib_compress(out, outsize, in, insize, settings); - } -} - -#endif /*LODEPNG_COMPILE_ENCODER*/ - -#else /*no LODEPNG_COMPILE_ZLIB*/ - -#ifdef LODEPNG_COMPILE_DECODER -static unsigned zlib_decompress(unsigned char** out, size_t* outsize, const unsigned char* in, - size_t insize, const LodePNGDecompressSettings* settings) -{ - if (!settings->custom_zlib) return 87; /*no custom zlib function provided */ - return settings->custom_zlib(out, outsize, in, insize, settings); -} -#endif /*LODEPNG_COMPILE_DECODER*/ -#ifdef LODEPNG_COMPILE_ENCODER -static unsigned zlib_compress(unsigned char** out, size_t* outsize, const unsigned char* in, - size_t insize, const LodePNGCompressSettings* settings) -{ - if (!settings->custom_zlib) return 87; /*no custom zlib function provided */ - return settings->custom_zlib(out, outsize, in, insize, settings); -} -#endif /*LODEPNG_COMPILE_ENCODER*/ - -#endif /*LODEPNG_COMPILE_ZLIB*/ - -/* ////////////////////////////////////////////////////////////////////////// */ - -#ifdef LODEPNG_COMPILE_ENCODER - -/*this is a good tradeoff between speed and compression ratio*/ -#define DEFAULT_WINDOWSIZE 2048 - -void lodepng_compress_settings_init(LodePNGCompressSettings* settings) -{ - /*compress with dynamic huffman tree (not in the mathematical sense, just not the predefined one)*/ - settings->btype = 2; - settings->use_lz77 = 1; - settings->windowsize = DEFAULT_WINDOWSIZE; - settings->minmatch = 3; - settings->nicematch = 128; - settings->lazymatching = 1; - - settings->custom_zlib = 0; - settings->custom_deflate = 0; - settings->custom_context = 0; -} - -const LodePNGCompressSettings lodepng_default_compress_settings = {2, 1, DEFAULT_WINDOWSIZE, 3, 128, 1, 0, 0, 0}; - - -#endif /*LODEPNG_COMPILE_ENCODER*/ - -#ifdef LODEPNG_COMPILE_DECODER - -void lodepng_decompress_settings_init(LodePNGDecompressSettings* settings) -{ - settings->ignore_adler32 = 0; - - settings->custom_zlib = 0; - settings->custom_inflate = 0; - settings->custom_context = 0; -} - -const LodePNGDecompressSettings lodepng_default_decompress_settings = {0, 0, 0, 0}; - -#endif /*LODEPNG_COMPILE_DECODER*/ - -/* ////////////////////////////////////////////////////////////////////////// */ -/* ////////////////////////////////////////////////////////////////////////// */ -/* // End of Zlib related code. Begin of PNG related code. // */ -/* ////////////////////////////////////////////////////////////////////////// */ -/* ////////////////////////////////////////////////////////////////////////// */ - -#ifdef LODEPNG_COMPILE_PNG - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / CRC32 / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -/* CRC polynomial: 0xedb88320 */ -static unsigned lodepng_crc32_table[256] = { - 0u, 1996959894u, 3993919788u, 2567524794u, 124634137u, 1886057615u, 3915621685u, 2657392035u, - 249268274u, 2044508324u, 3772115230u, 2547177864u, 162941995u, 2125561021u, 3887607047u, 2428444049u, - 498536548u, 1789927666u, 4089016648u, 2227061214u, 450548861u, 1843258603u, 4107580753u, 2211677639u, - 325883990u, 1684777152u, 4251122042u, 2321926636u, 335633487u, 1661365465u, 4195302755u, 2366115317u, - 997073096u, 1281953886u, 3579855332u, 2724688242u, 1006888145u, 1258607687u, 3524101629u, 2768942443u, - 901097722u, 1119000684u, 3686517206u, 2898065728u, 853044451u, 1172266101u, 3705015759u, 2882616665u, - 651767980u, 1373503546u, 3369554304u, 3218104598u, 565507253u, 1454621731u, 3485111705u, 3099436303u, - 671266974u, 1594198024u, 3322730930u, 2970347812u, 795835527u, 1483230225u, 3244367275u, 3060149565u, - 1994146192u, 31158534u, 2563907772u, 4023717930u, 1907459465u, 112637215u, 2680153253u, 3904427059u, - 2013776290u, 251722036u, 2517215374u, 3775830040u, 2137656763u, 141376813u, 2439277719u, 3865271297u, - 1802195444u, 476864866u, 2238001368u, 4066508878u, 1812370925u, 453092731u, 2181625025u, 4111451223u, - 1706088902u, 314042704u, 2344532202u, 4240017532u, 1658658271u, 366619977u, 2362670323u, 4224994405u, - 1303535960u, 984961486u, 2747007092u, 3569037538u, 1256170817u, 1037604311u, 2765210733u, 3554079995u, - 1131014506u, 879679996u, 2909243462u, 3663771856u, 1141124467u, 855842277u, 2852801631u, 3708648649u, - 1342533948u, 654459306u, 3188396048u, 3373015174u, 1466479909u, 544179635u, 3110523913u, 3462522015u, - 1591671054u, 702138776u, 2966460450u, 3352799412u, 1504918807u, 783551873u, 3082640443u, 3233442989u, - 3988292384u, 2596254646u, 62317068u, 1957810842u, 3939845945u, 2647816111u, 81470997u, 1943803523u, - 3814918930u, 2489596804u, 225274430u, 2053790376u, 3826175755u, 2466906013u, 167816743u, 2097651377u, - 4027552580u, 2265490386u, 503444072u, 1762050814u, 4150417245u, 2154129355u, 426522225u, 1852507879u, - 4275313526u, 2312317920u, 282753626u, 1742555852u, 4189708143u, 2394877945u, 397917763u, 1622183637u, - 3604390888u, 2714866558u, 953729732u, 1340076626u, 3518719985u, 2797360999u, 1068828381u, 1219638859u, - 3624741850u, 2936675148u, 906185462u, 1090812512u, 3747672003u, 2825379669u, 829329135u, 1181335161u, - 3412177804u, 3160834842u, 628085408u, 1382605366u, 3423369109u, 3138078467u, 570562233u, 1426400815u, - 3317316542u, 2998733608u, 733239954u, 1555261956u, 3268935591u, 3050360625u, 752459403u, 1541320221u, - 2607071920u, 3965973030u, 1969922972u, 40735498u, 2617837225u, 3943577151u, 1913087877u, 83908371u, - 2512341634u, 3803740692u, 2075208622u, 213261112u, 2463272603u, 3855990285u, 2094854071u, 198958881u, - 2262029012u, 4057260610u, 1759359992u, 534414190u, 2176718541u, 4139329115u, 1873836001u, 414664567u, - 2282248934u, 4279200368u, 1711684554u, 285281116u, 2405801727u, 4167216745u, 1634467795u, 376229701u, - 2685067896u, 3608007406u, 1308918612u, 956543938u, 2808555105u, 3495958263u, 1231636301u, 1047427035u, - 2932959818u, 3654703836u, 1088359270u, 936918000u, 2847714899u, 3736837829u, 1202900863u, 817233897u, - 3183342108u, 3401237130u, 1404277552u, 615818150u, 3134207493u, 3453421203u, 1423857449u, 601450431u, - 3009837614u, 3294710456u, 1567103746u, 711928724u, 3020668471u, 3272380065u, 1510334235u, 755167117u -}; - -/*Return the CRC of the bytes buf[0..len-1].*/ -unsigned lodepng_crc32(const unsigned char* buf, size_t len) -{ - unsigned c = 0xffffffffL; - size_t n; - - for(n = 0; n < len; n++) - { - c = lodepng_crc32_table[(c ^ buf[n]) & 0xff] ^ (c >> 8); - } - return c ^ 0xffffffffL; -} - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / Reading and writing single bits and bytes from/to stream for LodePNG / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -static unsigned char readBitFromReversedStream(size_t* bitpointer, const unsigned char* bitstream) -{ - unsigned char result = (unsigned char)((bitstream[(*bitpointer) >> 3] >> (7 - ((*bitpointer) & 0x7))) & 1); - (*bitpointer)++; - return result; -} - -static unsigned readBitsFromReversedStream(size_t* bitpointer, const unsigned char* bitstream, size_t nbits) -{ - unsigned result = 0; - size_t i; - for(i = nbits - 1; i < nbits; i--) - { - result += (unsigned)readBitFromReversedStream(bitpointer, bitstream) << i; - } - return result; -} - -#ifdef LODEPNG_COMPILE_DECODER -static void setBitOfReversedStream0(size_t* bitpointer, unsigned char* bitstream, unsigned char bit) -{ - /*the current bit in bitstream must be 0 for this to work*/ - if(bit) - { - /*earlier bit of huffman code is in a lesser significant bit of an earlier byte*/ - bitstream[(*bitpointer) >> 3] |= (bit << (7 - ((*bitpointer) & 0x7))); - } - (*bitpointer)++; -} -#endif /*LODEPNG_COMPILE_DECODER*/ - -static void setBitOfReversedStream(size_t* bitpointer, unsigned char* bitstream, unsigned char bit) -{ - /*the current bit in bitstream may be 0 or 1 for this to work*/ - if(bit == 0) bitstream[(*bitpointer) >> 3] &= (unsigned char)(~(1 << (7 - ((*bitpointer) & 0x7)))); - else bitstream[(*bitpointer) >> 3] |= (1 << (7 - ((*bitpointer) & 0x7))); - (*bitpointer)++; -} - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / PNG chunks / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -unsigned lodepng_chunk_length(const unsigned char* chunk) -{ - return lodepng_read32bitInt(&chunk[0]); -} - -void lodepng_chunk_type(char type[5], const unsigned char* chunk) -{ - unsigned i; - for(i = 0; i < 4; i++) type[i] = (char)chunk[4 + i]; - type[4] = 0; /*null termination char*/ -} - -unsigned char lodepng_chunk_type_equals(const unsigned char* chunk, const char* type) -{ - if(strlen(type) != 4) return 0; - return (chunk[4] == type[0] && chunk[5] == type[1] && chunk[6] == type[2] && chunk[7] == type[3]); -} - -unsigned char lodepng_chunk_ancillary(const unsigned char* chunk) -{ - return((chunk[4] & 32) != 0); -} - -unsigned char lodepng_chunk_private(const unsigned char* chunk) -{ - return((chunk[6] & 32) != 0); -} - -unsigned char lodepng_chunk_safetocopy(const unsigned char* chunk) -{ - return((chunk[7] & 32) != 0); -} - -unsigned char* lodepng_chunk_data(unsigned char* chunk) -{ - return &chunk[8]; -} - -const unsigned char* lodepng_chunk_data_const(const unsigned char* chunk) -{ - return &chunk[8]; -} - -unsigned lodepng_chunk_check_crc(const unsigned char* chunk) -{ - unsigned length = lodepng_chunk_length(chunk); - unsigned CRC = lodepng_read32bitInt(&chunk[length + 8]); - /*the CRC is taken of the data and the 4 chunk type letters, not the length*/ - unsigned checksum = lodepng_crc32(&chunk[4], length + 4); - if(CRC != checksum) return 1; - else return 0; -} - -void lodepng_chunk_generate_crc(unsigned char* chunk) -{ - unsigned length = lodepng_chunk_length(chunk); - unsigned CRC = lodepng_crc32(&chunk[4], length + 4); - lodepng_set32bitInt(chunk + 8 + length, CRC); -} - -unsigned char* lodepng_chunk_next(unsigned char* chunk) -{ - unsigned total_chunk_length = lodepng_chunk_length(chunk) + 12; - return &chunk[total_chunk_length]; -} - -const unsigned char* lodepng_chunk_next_const(const unsigned char* chunk) -{ - unsigned total_chunk_length = lodepng_chunk_length(chunk) + 12; - return &chunk[total_chunk_length]; -} - -unsigned lodepng_chunk_append(unsigned char** out, size_t* outlength, const unsigned char* chunk) -{ - unsigned i; - unsigned total_chunk_length = lodepng_chunk_length(chunk) + 12; - unsigned char *chunk_start, *new_buffer; - size_t new_length = (*outlength) + total_chunk_length; - if(new_length < total_chunk_length || new_length < (*outlength)) return 77; /*integer overflow happened*/ - - new_buffer = (unsigned char*)lodepng_realloc(*out, new_length); - if(!new_buffer) return 83; /*alloc fail*/ - (*out) = new_buffer; - (*outlength) = new_length; - chunk_start = &(*out)[new_length - total_chunk_length]; - - for(i = 0; i < total_chunk_length; i++) chunk_start[i] = chunk[i]; - - return 0; -} - -unsigned lodepng_chunk_create(unsigned char** out, size_t* outlength, unsigned length, - const char* type, const unsigned char* data) -{ - unsigned i; - unsigned char *chunk, *new_buffer; - size_t new_length = (*outlength) + length + 12; - if(new_length < length + 12 || new_length < (*outlength)) return 77; /*integer overflow happened*/ - new_buffer = (unsigned char*)lodepng_realloc(*out, new_length); - if(!new_buffer) return 83; /*alloc fail*/ - (*out) = new_buffer; - (*outlength) = new_length; - chunk = &(*out)[(*outlength) - length - 12]; - - /*1: length*/ - lodepng_set32bitInt(chunk, (unsigned)length); - - /*2: chunk name (4 letters)*/ - chunk[4] = (unsigned char)type[0]; - chunk[5] = (unsigned char)type[1]; - chunk[6] = (unsigned char)type[2]; - chunk[7] = (unsigned char)type[3]; - - /*3: the data*/ - for(i = 0; i < length; i++) chunk[8 + i] = data[i]; - - /*4: CRC (of the chunkname characters and the data)*/ - lodepng_chunk_generate_crc(chunk); - - return 0; -} - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / Color types and such / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -/*return type is a LodePNG error code*/ -static unsigned checkColorValidity(LodePNGColorType colortype, unsigned bd) /*bd = bitdepth*/ -{ - switch(colortype) - { - case 0: if(!(bd == 1 || bd == 2 || bd == 4 || bd == 8 || bd == 16)) return 37; break; /*grey*/ - case 2: if(!( bd == 8 || bd == 16)) return 37; break; /*RGB*/ - case 3: if(!(bd == 1 || bd == 2 || bd == 4 || bd == 8 )) return 37; break; /*palette*/ - case 4: if(!( bd == 8 || bd == 16)) return 37; break; /*grey + alpha*/ - case 6: if(!( bd == 8 || bd == 16)) return 37; break; /*RGBA*/ - default: return 31; - } - return 0; /*allowed color type / bits combination*/ -} - -static unsigned getNumColorChannels(LodePNGColorType colortype) -{ - switch(colortype) - { - case 0: return 1; /*grey*/ - case 2: return 3; /*RGB*/ - case 3: return 1; /*palette*/ - case 4: return 2; /*grey + alpha*/ - case 6: return 4; /*RGBA*/ - } - return 0; /*unexisting color type*/ -} - -static unsigned lodepng_get_bpp_lct(LodePNGColorType colortype, unsigned bitdepth) -{ - /*bits per pixel is amount of channels * bits per channel*/ - return getNumColorChannels(colortype) * bitdepth; -} - -/* ////////////////////////////////////////////////////////////////////////// */ - -void lodepng_color_mode_init(LodePNGColorMode* info) -{ - info->key_defined = 0; - info->key_r = info->key_g = info->key_b = 0; - info->colortype = LCT_RGBA; - info->bitdepth = 8; - info->palette = 0; - info->palettesize = 0; -} - -void lodepng_color_mode_cleanup(LodePNGColorMode* info) -{ - lodepng_palette_clear(info); -} - -unsigned lodepng_color_mode_copy(LodePNGColorMode* dest, const LodePNGColorMode* source) -{ - size_t i; - lodepng_color_mode_cleanup(dest); - *dest = *source; - if(source->palette) - { - dest->palette = (unsigned char*)lodepng_malloc(1024); - if(!dest->palette && source->palettesize) return 83; /*alloc fail*/ - for(i = 0; i < source->palettesize * 4; i++) dest->palette[i] = source->palette[i]; - } - return 0; -} - -static int lodepng_color_mode_equal(const LodePNGColorMode* a, const LodePNGColorMode* b) -{ - size_t i; - if(a->colortype != b->colortype) return 0; - if(a->bitdepth != b->bitdepth) return 0; - if(a->key_defined != b->key_defined) return 0; - if(a->key_defined) - { - if(a->key_r != b->key_r) return 0; - if(a->key_g != b->key_g) return 0; - if(a->key_b != b->key_b) return 0; - } - if(a->palettesize != b->palettesize) return 0; - for(i = 0; i < a->palettesize * 4; i++) - { - if(a->palette[i] != b->palette[i]) return 0; - } - return 1; -} - -void lodepng_palette_clear(LodePNGColorMode* info) -{ - if(info->palette) lodepng_free(info->palette); - info->palette = 0; - info->palettesize = 0; -} - -unsigned lodepng_palette_add(LodePNGColorMode* info, - unsigned char r, unsigned char g, unsigned char b, unsigned char a) -{ - unsigned char* data; - /*the same resize technique as C++ std::vectors is used, and here it's made so that for a palette with - the max of 256 colors, it'll have the exact alloc size*/ - if(!info->palette) /*allocate palette if empty*/ - { - /*room for 256 colors with 4 bytes each*/ - data = (unsigned char*)lodepng_realloc(info->palette, 1024); - if(!data) return 83; /*alloc fail*/ - else info->palette = data; - } - info->palette[4 * info->palettesize + 0] = r; - info->palette[4 * info->palettesize + 1] = g; - info->palette[4 * info->palettesize + 2] = b; - info->palette[4 * info->palettesize + 3] = a; - info->palettesize++; - return 0; -} - -unsigned lodepng_get_bpp(const LodePNGColorMode* info) -{ - /*calculate bits per pixel out of colortype and bitdepth*/ - return lodepng_get_bpp_lct(info->colortype, info->bitdepth); -} - -unsigned lodepng_get_channels(const LodePNGColorMode* info) -{ - return getNumColorChannels(info->colortype); -} - -unsigned lodepng_is_greyscale_type(const LodePNGColorMode* info) -{ - return info->colortype == LCT_GREY || info->colortype == LCT_GREY_ALPHA; -} - -unsigned lodepng_is_alpha_type(const LodePNGColorMode* info) -{ - return (info->colortype & 4) != 0; /*4 or 6*/ -} - -unsigned lodepng_is_palette_type(const LodePNGColorMode* info) -{ - return info->colortype == LCT_PALETTE; -} - -unsigned lodepng_has_palette_alpha(const LodePNGColorMode* info) -{ - size_t i; - for(i = 0; i < info->palettesize; i++) - { - if(info->palette[i * 4 + 3] < 255) return 1; - } - return 0; -} - -unsigned lodepng_can_have_alpha(const LodePNGColorMode* info) -{ - return info->key_defined - || lodepng_is_alpha_type(info) - || lodepng_has_palette_alpha(info); -} - -size_t lodepng_get_raw_size(unsigned w, unsigned h, const LodePNGColorMode* color) -{ - return (w * h * lodepng_get_bpp(color) + 7) / 8; -} - -size_t lodepng_get_raw_size_lct(unsigned w, unsigned h, LodePNGColorType colortype, unsigned bitdepth) -{ - return (w * h * lodepng_get_bpp_lct(colortype, bitdepth) + 7) / 8; -} - - -#ifdef LODEPNG_COMPILE_PNG -#ifdef LODEPNG_COMPILE_DECODER -/*in an idat chunk, each scanline is a multiple of 8 bits, unlike the lodepng output buffer*/ -static size_t lodepng_get_raw_size_idat(unsigned w, unsigned h, const LodePNGColorMode* color) -{ - return h * ((w * lodepng_get_bpp(color) + 7) / 8); -} -#endif /*LODEPNG_COMPILE_DECODER*/ -#endif /*LODEPNG_COMPILE_PNG*/ - -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - -static void LodePNGUnknownChunks_init(LodePNGInfo* info) -{ - unsigned i; - for(i = 0; i < 3; i++) info->unknown_chunks_data[i] = 0; - for(i = 0; i < 3; i++) info->unknown_chunks_size[i] = 0; -} - -static void LodePNGUnknownChunks_cleanup(LodePNGInfo* info) -{ - unsigned i; - for(i = 0; i < 3; i++) lodepng_free(info->unknown_chunks_data[i]); -} - -static unsigned LodePNGUnknownChunks_copy(LodePNGInfo* dest, const LodePNGInfo* src) -{ - unsigned i; - - LodePNGUnknownChunks_cleanup(dest); - - for(i = 0; i < 3; i++) - { - size_t j; - dest->unknown_chunks_size[i] = src->unknown_chunks_size[i]; - dest->unknown_chunks_data[i] = (unsigned char*)lodepng_malloc(src->unknown_chunks_size[i]); - if(!dest->unknown_chunks_data[i] && dest->unknown_chunks_size[i]) return 83; /*alloc fail*/ - for(j = 0; j < src->unknown_chunks_size[i]; j++) - { - dest->unknown_chunks_data[i][j] = src->unknown_chunks_data[i][j]; - } - } - - return 0; -} - -/******************************************************************************/ - -static void LodePNGText_init(LodePNGInfo* info) -{ - info->text_num = 0; - info->text_keys = NULL; - info->text_strings = NULL; -} - -static void LodePNGText_cleanup(LodePNGInfo* info) -{ - size_t i; - for(i = 0; i < info->text_num; i++) - { - string_cleanup(&info->text_keys[i]); - string_cleanup(&info->text_strings[i]); - } - lodepng_free(info->text_keys); - lodepng_free(info->text_strings); -} - -static unsigned LodePNGText_copy(LodePNGInfo* dest, const LodePNGInfo* source) -{ - size_t i = 0; - dest->text_keys = 0; - dest->text_strings = 0; - dest->text_num = 0; - for(i = 0; i < source->text_num; i++) - { - CERROR_TRY_RETURN(lodepng_add_text(dest, source->text_keys[i], source->text_strings[i])); - } - return 0; -} - -void lodepng_clear_text(LodePNGInfo* info) -{ - LodePNGText_cleanup(info); -} - -unsigned lodepng_add_text(LodePNGInfo* info, const char* key, const char* str) -{ - char** new_keys = (char**)(lodepng_realloc(info->text_keys, sizeof(char*) * (info->text_num + 1))); - char** new_strings = (char**)(lodepng_realloc(info->text_strings, sizeof(char*) * (info->text_num + 1))); - if(!new_keys || !new_strings) - { - lodepng_free(new_keys); - lodepng_free(new_strings); - return 83; /*alloc fail*/ - } - - info->text_num++; - info->text_keys = new_keys; - info->text_strings = new_strings; - - string_init(&info->text_keys[info->text_num - 1]); - string_set(&info->text_keys[info->text_num - 1], key); - - string_init(&info->text_strings[info->text_num - 1]); - string_set(&info->text_strings[info->text_num - 1], str); - - return 0; -} - -/******************************************************************************/ - -static void LodePNGIText_init(LodePNGInfo* info) -{ - info->itext_num = 0; - info->itext_keys = NULL; - info->itext_langtags = NULL; - info->itext_transkeys = NULL; - info->itext_strings = NULL; -} - -static void LodePNGIText_cleanup(LodePNGInfo* info) -{ - size_t i; - for(i = 0; i < info->itext_num; i++) - { - string_cleanup(&info->itext_keys[i]); - string_cleanup(&info->itext_langtags[i]); - string_cleanup(&info->itext_transkeys[i]); - string_cleanup(&info->itext_strings[i]); - } - lodepng_free(info->itext_keys); - lodepng_free(info->itext_langtags); - lodepng_free(info->itext_transkeys); - lodepng_free(info->itext_strings); -} - -static unsigned LodePNGIText_copy(LodePNGInfo* dest, const LodePNGInfo* source) -{ - size_t i = 0; - dest->itext_keys = 0; - dest->itext_langtags = 0; - dest->itext_transkeys = 0; - dest->itext_strings = 0; - dest->itext_num = 0; - for(i = 0; i < source->itext_num; i++) - { - CERROR_TRY_RETURN(lodepng_add_itext(dest, source->itext_keys[i], source->itext_langtags[i], - source->itext_transkeys[i], source->itext_strings[i])); - } - return 0; -} - -void lodepng_clear_itext(LodePNGInfo* info) -{ - LodePNGIText_cleanup(info); -} - -unsigned lodepng_add_itext(LodePNGInfo* info, const char* key, const char* langtag, - const char* transkey, const char* str) -{ - char** new_keys = (char**)(lodepng_realloc(info->itext_keys, sizeof(char*) * (info->itext_num + 1))); - char** new_langtags = (char**)(lodepng_realloc(info->itext_langtags, sizeof(char*) * (info->itext_num + 1))); - char** new_transkeys = (char**)(lodepng_realloc(info->itext_transkeys, sizeof(char*) * (info->itext_num + 1))); - char** new_strings = (char**)(lodepng_realloc(info->itext_strings, sizeof(char*) * (info->itext_num + 1))); - if(!new_keys || !new_langtags || !new_transkeys || !new_strings) - { - lodepng_free(new_keys); - lodepng_free(new_langtags); - lodepng_free(new_transkeys); - lodepng_free(new_strings); - return 83; /*alloc fail*/ - } - - info->itext_num++; - info->itext_keys = new_keys; - info->itext_langtags = new_langtags; - info->itext_transkeys = new_transkeys; - info->itext_strings = new_strings; - - string_init(&info->itext_keys[info->itext_num - 1]); - string_set(&info->itext_keys[info->itext_num - 1], key); - - string_init(&info->itext_langtags[info->itext_num - 1]); - string_set(&info->itext_langtags[info->itext_num - 1], langtag); - - string_init(&info->itext_transkeys[info->itext_num - 1]); - string_set(&info->itext_transkeys[info->itext_num - 1], transkey); - - string_init(&info->itext_strings[info->itext_num - 1]); - string_set(&info->itext_strings[info->itext_num - 1], str); - - return 0; -} -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - -void lodepng_info_init(LodePNGInfo* info) -{ - lodepng_color_mode_init(&info->color); - info->interlace_method = 0; - info->compression_method = 0; - info->filter_method = 0; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - info->background_defined = 0; - info->background_r = info->background_g = info->background_b = 0; - - LodePNGText_init(info); - LodePNGIText_init(info); - - info->time_defined = 0; - info->phys_defined = 0; - - LodePNGUnknownChunks_init(info); -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ -} - -void lodepng_info_cleanup(LodePNGInfo* info) -{ - lodepng_color_mode_cleanup(&info->color); -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - LodePNGText_cleanup(info); - LodePNGIText_cleanup(info); - - LodePNGUnknownChunks_cleanup(info); -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ -} - -unsigned lodepng_info_copy(LodePNGInfo* dest, const LodePNGInfo* source) -{ - lodepng_info_cleanup(dest); - *dest = *source; - lodepng_color_mode_init(&dest->color); - CERROR_TRY_RETURN(lodepng_color_mode_copy(&dest->color, &source->color)); - -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - CERROR_TRY_RETURN(LodePNGText_copy(dest, source)); - CERROR_TRY_RETURN(LodePNGIText_copy(dest, source)); - - LodePNGUnknownChunks_init(dest); - CERROR_TRY_RETURN(LodePNGUnknownChunks_copy(dest, source)); -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - return 0; -} - -void lodepng_info_swap(LodePNGInfo* a, LodePNGInfo* b) -{ - LodePNGInfo temp = *a; - *a = *b; - *b = temp; -} - -/* ////////////////////////////////////////////////////////////////////////// */ - -/*index: bitgroup index, bits: bitgroup size(1, 2 or 4), in: bitgroup value, out: octet array to add bits to*/ -static void addColorBits(unsigned char* out, size_t index, unsigned bits, unsigned in) -{ - unsigned m = bits == 1 ? 7 : bits == 2 ? 3 : 1; /*8 / bits - 1*/ - /*p = the partial index in the byte, e.g. with 4 palettebits it is 0 for first half or 1 for second half*/ - unsigned p = index & m; - in &= (1u << bits) - 1u; /*filter out any other bits of the input value*/ - in = in << (bits * (m - p)); - if(p == 0) out[index * bits / 8] = in; - else out[index * bits / 8] |= in; -} - -typedef struct ColorTree ColorTree; - -/* -One node of a color tree -This is the data structure used to count the number of unique colors and to get a palette -index for a color. It's like an octree, but because the alpha channel is used too, each -node has 16 instead of 8 children. -*/ -struct ColorTree -{ - ColorTree* children[16]; /*up to 16 pointers to ColorTree of next level*/ - int index; /*the payload. Only has a meaningful value if this is in the last level*/ -}; - -static void color_tree_init(ColorTree* tree) -{ - int i; - for(i = 0; i < 16; i++) tree->children[i] = 0; - tree->index = -1; -} - -static void color_tree_cleanup(ColorTree* tree) -{ - int i; - for(i = 0; i < 16; i++) - { - if(tree->children[i]) - { - color_tree_cleanup(tree->children[i]); - lodepng_free(tree->children[i]); - } - } -} - -/*returns -1 if color not present, its index otherwise*/ -static int color_tree_get(ColorTree* tree, unsigned char r, unsigned char g, unsigned char b, unsigned char a) -{ - int bit = 0; - for(bit = 0; bit < 8; bit++) - { - int i = 8 * ((r >> bit) & 1) + 4 * ((g >> bit) & 1) + 2 * ((b >> bit) & 1) + 1 * ((a >> bit) & 1); - if(!tree->children[i]) return -1; - else tree = tree->children[i]; - } - return tree ? tree->index : -1; -} - -#ifdef LODEPNG_COMPILE_ENCODER -static int color_tree_has(ColorTree* tree, unsigned char r, unsigned char g, unsigned char b, unsigned char a) -{ - return color_tree_get(tree, r, g, b, a) >= 0; -} -#endif /*LODEPNG_COMPILE_ENCODER*/ - -/*color is not allowed to already exist. -Index should be >= 0 (it's signed to be compatible with using -1 for "doesn't exist")*/ -static void color_tree_add(ColorTree* tree, - unsigned char r, unsigned char g, unsigned char b, unsigned char a, unsigned index) -{ - int bit; - for(bit = 0; bit < 8; bit++) - { - int i = 8 * ((r >> bit) & 1) + 4 * ((g >> bit) & 1) + 2 * ((b >> bit) & 1) + 1 * ((a >> bit) & 1); - if(!tree->children[i]) - { - tree->children[i] = (ColorTree*)lodepng_malloc(sizeof(ColorTree)); - color_tree_init(tree->children[i]); - } - tree = tree->children[i]; - } - tree->index = (int)index; -} - -/*put a pixel, given its RGBA color, into image of any color type*/ -static unsigned rgba8ToPixel(unsigned char* out, size_t i, - const LodePNGColorMode* mode, ColorTree* tree /*for palette*/, - unsigned char r, unsigned char g, unsigned char b, unsigned char a) -{ - if(mode->colortype == LCT_GREY) - { - unsigned char grey = r; /*((unsigned short)r + g + b) / 3*/; - if(mode->bitdepth == 8) out[i] = grey; - else if(mode->bitdepth == 16) out[i * 2 + 0] = out[i * 2 + 1] = grey; - else - { - /*take the most significant bits of grey*/ - grey = (grey >> (8 - mode->bitdepth)) & ((1 << mode->bitdepth) - 1); - addColorBits(out, i, mode->bitdepth, grey); - } - } - else if(mode->colortype == LCT_RGB) - { - if(mode->bitdepth == 8) - { - out[i * 3 + 0] = r; - out[i * 3 + 1] = g; - out[i * 3 + 2] = b; - } - else - { - out[i * 6 + 0] = out[i * 6 + 1] = r; - out[i * 6 + 2] = out[i * 6 + 3] = g; - out[i * 6 + 4] = out[i * 6 + 5] = b; - } - } - else if(mode->colortype == LCT_PALETTE) - { - int index = color_tree_get(tree, r, g, b, a); - if(index < 0) return 82; /*color not in palette*/ - if(mode->bitdepth == 8) out[i] = index; - else addColorBits(out, i, mode->bitdepth, (unsigned)index); - } - else if(mode->colortype == LCT_GREY_ALPHA) - { - unsigned char grey = r; /*((unsigned short)r + g + b) / 3*/; - if(mode->bitdepth == 8) - { - out[i * 2 + 0] = grey; - out[i * 2 + 1] = a; - } - else if(mode->bitdepth == 16) - { - out[i * 4 + 0] = out[i * 4 + 1] = grey; - out[i * 4 + 2] = out[i * 4 + 3] = a; - } - } - else if(mode->colortype == LCT_RGBA) - { - if(mode->bitdepth == 8) - { - out[i * 4 + 0] = r; - out[i * 4 + 1] = g; - out[i * 4 + 2] = b; - out[i * 4 + 3] = a; - } - else - { - out[i * 8 + 0] = out[i * 8 + 1] = r; - out[i * 8 + 2] = out[i * 8 + 3] = g; - out[i * 8 + 4] = out[i * 8 + 5] = b; - out[i * 8 + 6] = out[i * 8 + 7] = a; - } - } - - return 0; /*no error*/ -} - -/*put a pixel, given its RGBA16 color, into image of any color 16-bitdepth type*/ -static void rgba16ToPixel(unsigned char* out, size_t i, - const LodePNGColorMode* mode, - unsigned short r, unsigned short g, unsigned short b, unsigned short a) -{ - if(mode->colortype == LCT_GREY) - { - unsigned short grey = r; /*((unsigned)r + g + b) / 3*/; - out[i * 2 + 0] = (grey >> 8) & 255; - out[i * 2 + 1] = grey & 255; - } - else if(mode->colortype == LCT_RGB) - { - out[i * 6 + 0] = (r >> 8) & 255; - out[i * 6 + 1] = r & 255; - out[i * 6 + 2] = (g >> 8) & 255; - out[i * 6 + 3] = g & 255; - out[i * 6 + 4] = (b >> 8) & 255; - out[i * 6 + 5] = b & 255; - } - else if(mode->colortype == LCT_GREY_ALPHA) - { - unsigned short grey = r; /*((unsigned)r + g + b) / 3*/; - out[i * 4 + 0] = (grey >> 8) & 255; - out[i * 4 + 1] = grey & 255; - out[i * 4 + 2] = (a >> 8) & 255; - out[i * 4 + 3] = a & 255; - } - else if(mode->colortype == LCT_RGBA) - { - out[i * 8 + 0] = (r >> 8) & 255; - out[i * 8 + 1] = r & 255; - out[i * 8 + 2] = (g >> 8) & 255; - out[i * 8 + 3] = g & 255; - out[i * 8 + 4] = (b >> 8) & 255; - out[i * 8 + 5] = b & 255; - out[i * 8 + 6] = (a >> 8) & 255; - out[i * 8 + 7] = a & 255; - } -} - -/*Get RGBA8 color of pixel with index i (y * width + x) from the raw image with given color type.*/ -static void getPixelColorRGBA8(unsigned char* r, unsigned char* g, - unsigned char* b, unsigned char* a, - const unsigned char* in, size_t i, - const LodePNGColorMode* mode) -{ - if(mode->colortype == LCT_GREY) - { - if(mode->bitdepth == 8) - { - *r = *g = *b = in[i]; - if(mode->key_defined && *r == mode->key_r) *a = 0; - else *a = 255; - } - else if(mode->bitdepth == 16) - { - *r = *g = *b = in[i * 2 + 0]; - if(mode->key_defined && 256U * in[i * 2 + 0] + in[i * 2 + 1] == mode->key_r) *a = 0; - else *a = 255; - } - else - { - unsigned highest = ((1U << mode->bitdepth) - 1U); /*highest possible value for this bit depth*/ - size_t j = i * mode->bitdepth; - unsigned value = readBitsFromReversedStream(&j, in, mode->bitdepth); - *r = *g = *b = (value * 255) / highest; - if(mode->key_defined && value == mode->key_r) *a = 0; - else *a = 255; - } - } - else if(mode->colortype == LCT_RGB) - { - if(mode->bitdepth == 8) - { - *r = in[i * 3 + 0]; *g = in[i * 3 + 1]; *b = in[i * 3 + 2]; - if(mode->key_defined && *r == mode->key_r && *g == mode->key_g && *b == mode->key_b) *a = 0; - else *a = 255; - } - else - { - *r = in[i * 6 + 0]; - *g = in[i * 6 + 2]; - *b = in[i * 6 + 4]; - if(mode->key_defined && 256U * in[i * 6 + 0] + in[i * 6 + 1] == mode->key_r - && 256U * in[i * 6 + 2] + in[i * 6 + 3] == mode->key_g - && 256U * in[i * 6 + 4] + in[i * 6 + 5] == mode->key_b) *a = 0; - else *a = 255; - } - } - else if(mode->colortype == LCT_PALETTE) - { - unsigned index; - if(mode->bitdepth == 8) index = in[i]; - else - { - size_t j = i * mode->bitdepth; - index = readBitsFromReversedStream(&j, in, mode->bitdepth); - } - - if(index >= mode->palettesize) - { - /*This is an error according to the PNG spec, but common PNG decoders make it black instead. - Done here too, slightly faster due to no error handling needed.*/ - *r = *g = *b = 0; - *a = 255; - } - else - { - *r = mode->palette[index * 4 + 0]; - *g = mode->palette[index * 4 + 1]; - *b = mode->palette[index * 4 + 2]; - *a = mode->palette[index * 4 + 3]; - } - } - else if(mode->colortype == LCT_GREY_ALPHA) - { - if(mode->bitdepth == 8) - { - *r = *g = *b = in[i * 2 + 0]; - *a = in[i * 2 + 1]; - } - else - { - *r = *g = *b = in[i * 4 + 0]; - *a = in[i * 4 + 2]; - } - } - else if(mode->colortype == LCT_RGBA) - { - if(mode->bitdepth == 8) - { - *r = in[i * 4 + 0]; - *g = in[i * 4 + 1]; - *b = in[i * 4 + 2]; - *a = in[i * 4 + 3]; - } - else - { - *r = in[i * 8 + 0]; - *g = in[i * 8 + 2]; - *b = in[i * 8 + 4]; - *a = in[i * 8 + 6]; - } - } -} - -/*Similar to getPixelColorRGBA8, but with all the for loops inside of the color -mode test cases, optimized to convert the colors much faster, when converting -to RGBA or RGB with 8 bit per cannel. buffer must be RGBA or RGB output with -enough memory, if has_alpha is true the output is RGBA. mode has the color mode -of the input buffer.*/ -static void getPixelColorsRGBA8(unsigned char* buffer, size_t numpixels, - unsigned has_alpha, const unsigned char* in, - const LodePNGColorMode* mode) -{ - unsigned num_channels = has_alpha ? 4 : 3; - size_t i; - if(mode->colortype == LCT_GREY) - { - if(mode->bitdepth == 8) - { - for(i = 0; i < numpixels; i++, buffer += num_channels) - { - buffer[0] = buffer[1] = buffer[2] = in[i]; - if(has_alpha) buffer[3] = mode->key_defined && in[i] == mode->key_r ? 0 : 255; - } - } - else if(mode->bitdepth == 16) - { - for(i = 0; i < numpixels; i++, buffer += num_channels) - { - buffer[0] = buffer[1] = buffer[2] = in[i * 2]; - if(has_alpha) buffer[3] = mode->key_defined && 256U * in[i * 2 + 0] + in[i * 2 + 1] == mode->key_r ? 0 : 255; - } - } - else - { - unsigned highest = ((1U << mode->bitdepth) - 1U); /*highest possible value for this bit depth*/ - size_t j = 0; - for(i = 0; i < numpixels; i++, buffer += num_channels) - { - unsigned value = readBitsFromReversedStream(&j, in, mode->bitdepth); - buffer[0] = buffer[1] = buffer[2] = (value * 255) / highest; - if(has_alpha) buffer[3] = mode->key_defined && value == mode->key_r ? 0 : 255; - } - } - } - else if(mode->colortype == LCT_RGB) - { - if(mode->bitdepth == 8) - { - for(i = 0; i < numpixels; i++, buffer += num_channels) - { - buffer[0] = in[i * 3 + 0]; - buffer[1] = in[i * 3 + 1]; - buffer[2] = in[i * 3 + 2]; - if(has_alpha) buffer[3] = mode->key_defined && buffer[0] == mode->key_r - && buffer[1]== mode->key_g && buffer[2] == mode->key_b ? 0 : 255; - } - } - else - { - for(i = 0; i < numpixels; i++, buffer += num_channels) - { - buffer[0] = in[i * 6 + 0]; - buffer[1] = in[i * 6 + 2]; - buffer[2] = in[i * 6 + 4]; - if(has_alpha) buffer[3] = mode->key_defined - && 256U * in[i * 6 + 0] + in[i * 6 + 1] == mode->key_r - && 256U * in[i * 6 + 2] + in[i * 6 + 3] == mode->key_g - && 256U * in[i * 6 + 4] + in[i * 6 + 5] == mode->key_b ? 0 : 255; - } - } - } - else if(mode->colortype == LCT_PALETTE) - { - unsigned index; - size_t j = 0; - for(i = 0; i < numpixels; i++, buffer += num_channels) - { - if(mode->bitdepth == 8) index = in[i]; - else index = readBitsFromReversedStream(&j, in, mode->bitdepth); - - if(index >= mode->palettesize) - { - /*This is an error according to the PNG spec, but most PNG decoders make it black instead. - Done here too, slightly faster due to no error handling needed.*/ - buffer[0] = buffer[1] = buffer[2] = 0; - if(has_alpha) buffer[3] = 255; - } - else - { - buffer[0] = mode->palette[index * 4 + 0]; - buffer[1] = mode->palette[index * 4 + 1]; - buffer[2] = mode->palette[index * 4 + 2]; - if(has_alpha) buffer[3] = mode->palette[index * 4 + 3]; - } - } - } - else if(mode->colortype == LCT_GREY_ALPHA) - { - if(mode->bitdepth == 8) - { - for(i = 0; i < numpixels; i++, buffer += num_channels) - { - buffer[0] = buffer[1] = buffer[2] = in[i * 2 + 0]; - if(has_alpha) buffer[3] = in[i * 2 + 1]; - } - } - else - { - for(i = 0; i < numpixels; i++, buffer += num_channels) - { - buffer[0] = buffer[1] = buffer[2] = in[i * 4 + 0]; - if(has_alpha) buffer[3] = in[i * 4 + 2]; - } - } - } - else if(mode->colortype == LCT_RGBA) - { - if(mode->bitdepth == 8) - { - for(i = 0; i < numpixels; i++, buffer += num_channels) - { - buffer[0] = in[i * 4 + 0]; - buffer[1] = in[i * 4 + 1]; - buffer[2] = in[i * 4 + 2]; - if(has_alpha) buffer[3] = in[i * 4 + 3]; - } - } - else - { - for(i = 0; i < numpixels; i++, buffer += num_channels) - { - buffer[0] = in[i * 8 + 0]; - buffer[1] = in[i * 8 + 2]; - buffer[2] = in[i * 8 + 4]; - if(has_alpha) buffer[3] = in[i * 8 + 6]; - } - } - } -} - -/*Get RGBA16 color of pixel with index i (y * width + x) from the raw image with -given color type, but the given color type must be 16-bit itself.*/ -static void getPixelColorRGBA16(unsigned short* r, unsigned short* g, unsigned short* b, unsigned short* a, - const unsigned char* in, size_t i, const LodePNGColorMode* mode) -{ - if(mode->colortype == LCT_GREY) - { - *r = *g = *b = 256 * in[i * 2 + 0] + in[i * 2 + 1]; - if(mode->key_defined && 256U * in[i * 2 + 0] + in[i * 2 + 1] == mode->key_r) *a = 0; - else *a = 65535; - } - else if(mode->colortype == LCT_RGB) - { - *r = 256 * in[i * 6 + 0] + in[i * 6 + 1]; - *g = 256 * in[i * 6 + 2] + in[i * 6 + 3]; - *b = 256 * in[i * 6 + 4] + in[i * 6 + 5]; - if(mode->key_defined && 256U * in[i * 6 + 0] + in[i * 6 + 1] == mode->key_r - && 256U * in[i * 6 + 2] + in[i * 6 + 3] == mode->key_g - && 256U * in[i * 6 + 4] + in[i * 6 + 5] == mode->key_b) *a = 0; - else *a = 65535; - } - else if(mode->colortype == LCT_GREY_ALPHA) - { - *r = *g = *b = 256 * in[i * 4 + 0] + in[i * 4 + 1]; - *a = 256 * in[i * 4 + 2] + in[i * 4 + 3]; - } - else if(mode->colortype == LCT_RGBA) - { - *r = 256 * in[i * 8 + 0] + in[i * 8 + 1]; - *g = 256 * in[i * 8 + 2] + in[i * 8 + 3]; - *b = 256 * in[i * 8 + 4] + in[i * 8 + 5]; - *a = 256 * in[i * 8 + 6] + in[i * 8 + 7]; - } -} - -unsigned lodepng_convert(unsigned char* out, const unsigned char* in, - LodePNGColorMode* mode_out, const LodePNGColorMode* mode_in, - unsigned w, unsigned h) -{ - size_t i; - ColorTree tree; - size_t numpixels = w * h; - - if(lodepng_color_mode_equal(mode_out, mode_in)) - { - size_t numbytes = lodepng_get_raw_size(w, h, mode_in); - for(i = 0; i < numbytes; i++) out[i] = in[i]; - return 0; - } - - if(mode_out->colortype == LCT_PALETTE) - { - size_t palsize = 1u << mode_out->bitdepth; - if(mode_out->palettesize < palsize) palsize = mode_out->palettesize; - color_tree_init(&tree); - for(i = 0; i < palsize; i++) - { - unsigned char* p = &mode_out->palette[i * 4]; - color_tree_add(&tree, p[0], p[1], p[2], p[3], i); - } - } - - if(mode_in->bitdepth == 16 && mode_out->bitdepth == 16) - { - for(i = 0; i < numpixels; i++) - { - unsigned short r = 0, g = 0, b = 0, a = 0; - getPixelColorRGBA16(&r, &g, &b, &a, in, i, mode_in); - rgba16ToPixel(out, i, mode_out, r, g, b, a); - } - } - else if(mode_out->bitdepth == 8 && mode_out->colortype == LCT_RGBA) - { - getPixelColorsRGBA8(out, numpixels, 1, in, mode_in); - } - else if(mode_out->bitdepth == 8 && mode_out->colortype == LCT_RGB) - { - getPixelColorsRGBA8(out, numpixels, 0, in, mode_in); - } - else - { - unsigned char r = 0, g = 0, b = 0, a = 0; - for(i = 0; i < numpixels; i++) - { - getPixelColorRGBA8(&r, &g, &b, &a, in, i, mode_in); - rgba8ToPixel(out, i, mode_out, &tree, r, g, b, a); - } - } - - if(mode_out->colortype == LCT_PALETTE) - { - color_tree_cleanup(&tree); - } - - return 0; /*no error (this function currently never has one, but maybe OOM detection added later.)*/ -} - -#ifdef LODEPNG_COMPILE_ENCODER - -void lodepng_color_profile_init(LodePNGColorProfile* profile) -{ - profile->colored = 0; - profile->key = 0; - profile->alpha = 0; - profile->key_r = profile->key_g = profile->key_b = 0; - profile->numcolors = 0; - profile->bits = 1; -} - -/*function used for debug purposes with C++*/ -/*void printColorProfile(LodePNGColorProfile* p) -{ - std::cout << "colored: " << (int)p->colored << ", "; - std::cout << "key: " << (int)p->key << ", "; - std::cout << "key_r: " << (int)p->key_r << ", "; - std::cout << "key_g: " << (int)p->key_g << ", "; - std::cout << "key_b: " << (int)p->key_b << ", "; - std::cout << "alpha: " << (int)p->alpha << ", "; - std::cout << "numcolors: " << (int)p->numcolors << ", "; - std::cout << "bits: " << (int)p->bits << std::endl; -}*/ - -/*Returns how many bits needed to represent given value (max 8 bit)*/ -unsigned getValueRequiredBits(unsigned char value) -{ - if(value == 0 || value == 255) return 1; - /*The scaling of 2-bit and 4-bit values uses multiples of 85 and 17*/ - if(value % 17 == 0) return value % 85 == 0 ? 2 : 4; - return 8; -} - -/*profile must already have been inited with mode. -It's ok to set some parameters of profile to done already.*/ -unsigned get_color_profile(LodePNGColorProfile* profile, - const unsigned char* in, unsigned w, unsigned h, - const LodePNGColorMode* mode) -{ - unsigned error = 0; - size_t i; - ColorTree tree; - size_t numpixels = w * h; - - unsigned colored_done = lodepng_is_greyscale_type(mode) ? 1 : 0; - unsigned alpha_done = lodepng_can_have_alpha(mode) ? 0 : 1; - unsigned numcolors_done = 0; - unsigned bpp = lodepng_get_bpp(mode); - unsigned bits_done = bpp == 1 ? 1 : 0; - unsigned maxnumcolors = 257; - unsigned sixteen = 0; - if(bpp <= 8) maxnumcolors = bpp == 1 ? 2 : (bpp == 2 ? 4 : (bpp == 4 ? 16 : 256)); - - color_tree_init(&tree); - - /*Check if the 16-bit input is truly 16-bit*/ - if(mode->bitdepth == 16) - { - unsigned short r, g, b, a; - for(i = 0; i < numpixels; i++) - { - getPixelColorRGBA16(&r, &g, &b, &a, in, i, mode); - if(r % 257u != 0 || g % 257u != 0 || b % 257u != 0 || a % 257u != 0) /*first and second byte differ*/ - { - sixteen = 1; - break; - } - } - } - - if(sixteen) - { - unsigned short r = 0, g = 0, b = 0, a = 0; - profile->bits = 16; - bits_done = numcolors_done = 1; /*counting colors no longer useful, palette doesn't support 16-bit*/ - - for(i = 0; i < numpixels; i++) - { - getPixelColorRGBA16(&r, &g, &b, &a, in, i, mode); - - if(!colored_done && (r != g || r != b)) - { - profile->colored = 1; - colored_done = 1; - } - - if(!alpha_done) - { - unsigned matchkey = (r == profile->key_r && g == profile->key_g && b == profile->key_b); - if(a != 65535 && (a != 0 || (profile->key && !matchkey))) - { - profile->alpha = 1; - alpha_done = 1; - if(profile->bits < 8) profile->bits = 8; /*PNG has no alphachannel modes with less than 8-bit per channel*/ - } - else if(a == 0 && !profile->alpha && !profile->key) - { - profile->key = 1; - profile->key_r = r; - profile->key_g = g; - profile->key_b = b; - } - else if(a == 65535 && profile->key && matchkey) - { - /* Color key cannot be used if an opaque pixel also has that RGB color. */ - profile->alpha = 1; - alpha_done = 1; - } - } - - if(alpha_done && numcolors_done && colored_done && bits_done) break; - } - } - else /* < 16-bit */ - { - for(i = 0; i < numpixels; i++) - { - unsigned char r = 0, g = 0, b = 0, a = 0; - getPixelColorRGBA8(&r, &g, &b, &a, in, i, mode); - - if(!bits_done && profile->bits < 8) - { - /*only r is checked, < 8 bits is only relevant for greyscale*/ - unsigned bits = getValueRequiredBits(r); - if(bits > profile->bits) profile->bits = bits; - } - bits_done = (profile->bits >= bpp); - - if(!colored_done && (r != g || r != b)) - { - profile->colored = 1; - colored_done = 1; - if(profile->bits < 8) profile->bits = 8; /*PNG has no colored modes with less than 8-bit per channel*/ - } - - if(!alpha_done) - { - unsigned matchkey = (r == profile->key_r && g == profile->key_g && b == profile->key_b); - if(a != 255 && (a != 0 || (profile->key && !matchkey))) - { - profile->alpha = 1; - alpha_done = 1; - if(profile->bits < 8) profile->bits = 8; /*PNG has no alphachannel modes with less than 8-bit per channel*/ - } - else if(a == 0 && !profile->alpha && !profile->key) - { - profile->key = 1; - profile->key_r = r; - profile->key_g = g; - profile->key_b = b; - } - else if(a == 255 && profile->key && matchkey) - { - /* Color key cannot be used if an opaque pixel also has that RGB color. */ - profile->alpha = 1; - alpha_done = 1; - if(profile->bits < 8) profile->bits = 8; /*PNG has no alphachannel modes with less than 8-bit per channel*/ - } - } - - if(!numcolors_done) - { - if(!color_tree_has(&tree, r, g, b, a)) - { - color_tree_add(&tree, r, g, b, a, profile->numcolors); - if(profile->numcolors < 256) - { - unsigned char* p = profile->palette; - unsigned n = profile->numcolors; - p[n * 4 + 0] = r; - p[n * 4 + 1] = g; - p[n * 4 + 2] = b; - p[n * 4 + 3] = a; - } - profile->numcolors++; - numcolors_done = profile->numcolors >= maxnumcolors; - } - } - - if(alpha_done && numcolors_done && colored_done && bits_done) break; - } - - /*make the profile's key always 16-bit for consistency - repeat each byte twice*/ - profile->key_r *= 257; - profile->key_g *= 257; - profile->key_b *= 257; - } - - color_tree_cleanup(&tree); - return error; -} - -/*Automatically chooses color type that gives smallest amount of bits in the -output image, e.g. grey if there are only greyscale pixels, palette if there -are less than 256 colors, ... -Updates values of mode with a potentially smaller color model. mode_out should -contain the user chosen color model, but will be overwritten with the new chosen one.*/ -unsigned lodepng_auto_choose_color(LodePNGColorMode* mode_out, - const unsigned char* image, unsigned w, unsigned h, - const LodePNGColorMode* mode_in) -{ - LodePNGColorProfile prof; - unsigned error = 0; - unsigned i, n, palettebits, grey_ok, palette_ok; - - lodepng_color_profile_init(&prof); - error = get_color_profile(&prof, image, w, h, mode_in); - if(error) return error; - mode_out->key_defined = 0; - - if(prof.key && w * h <= 16) prof.alpha = 1; /*too few pixels to justify tRNS chunk overhead*/ - grey_ok = !prof.colored && !prof.alpha; /*grey without alpha, with potentially low bits*/ - n = prof.numcolors; - palettebits = n <= 2 ? 1 : (n <= 4 ? 2 : (n <= 16 ? 4 : 8)); - palette_ok = n <= 256 && (n * 2 < w * h) && prof.bits <= 8; - if(w * h < n * 2) palette_ok = 0; /*don't add palette overhead if image has only a few pixels*/ - if(grey_ok && prof.bits <= palettebits) palette_ok = 0; /*grey is less overhead*/ - - if(palette_ok) - { - unsigned char* p = prof.palette; - lodepng_palette_clear(mode_out); /*remove potential earlier palette*/ - for(i = 0; i < prof.numcolors; i++) - { - error = lodepng_palette_add(mode_out, p[i * 4 + 0], p[i * 4 + 1], p[i * 4 + 2], p[i * 4 + 3]); - if(error) break; - } - - mode_out->colortype = LCT_PALETTE; - mode_out->bitdepth = palettebits; - - if(mode_in->colortype == LCT_PALETTE && mode_in->palettesize >= mode_out->palettesize - && mode_in->bitdepth == mode_out->bitdepth) - { - /*If input should have same palette colors, keep original to preserve its order and prevent conversion*/ - lodepng_color_mode_cleanup(mode_out); - lodepng_color_mode_copy(mode_out, mode_in); - } - } - else /*8-bit or 16-bit per channel*/ - { - mode_out->bitdepth = prof.bits; - mode_out->colortype = prof.alpha ? (prof.colored ? LCT_RGBA : LCT_GREY_ALPHA) - : (prof.colored ? LCT_RGB : LCT_GREY); - - if(prof.key && !prof.alpha) - { - unsigned mask = (1u << mode_out->bitdepth) - 1u; /*profile always uses 16-bit, mask converts it*/ - mode_out->key_r = prof.key_r & mask; - mode_out->key_g = prof.key_g & mask; - mode_out->key_b = prof.key_b & mask; - mode_out->key_defined = 1; - } - } - - return error; -} - -#endif /* #ifdef LODEPNG_COMPILE_ENCODER */ - -/* -Paeth predicter, used by PNG filter type 4 -The parameters are of type short, but should come from unsigned chars, the shorts -are only needed to make the paeth calculation correct. -*/ -static unsigned char paethPredictor(short a, short b, short c) -{ - short pa = abs(b - c); - short pb = abs(a - c); - short pc = abs(a + b - c - c); - - if(pc < pa && pc < pb) return (unsigned char)c; - else if(pb < pa) return (unsigned char)b; - else return (unsigned char)a; -} - -/*shared values used by multiple Adam7 related functions*/ - -static const unsigned ADAM7_IX[7] = { 0, 4, 0, 2, 0, 1, 0 }; /*x start values*/ -static const unsigned ADAM7_IY[7] = { 0, 0, 4, 0, 2, 0, 1 }; /*y start values*/ -static const unsigned ADAM7_DX[7] = { 8, 8, 4, 4, 2, 2, 1 }; /*x delta values*/ -static const unsigned ADAM7_DY[7] = { 8, 8, 8, 4, 4, 2, 2 }; /*y delta values*/ - -/* -Outputs various dimensions and positions in the image related to the Adam7 reduced images. -passw: output containing the width of the 7 passes -passh: output containing the height of the 7 passes -filter_passstart: output containing the index of the start and end of each - reduced image with filter bytes -padded_passstart output containing the index of the start and end of each - reduced image when without filter bytes but with padded scanlines -passstart: output containing the index of the start and end of each reduced - image without padding between scanlines, but still padding between the images -w, h: width and height of non-interlaced image -bpp: bits per pixel -"padded" is only relevant if bpp is less than 8 and a scanline or image does not - end at a full byte -*/ -static void Adam7_getpassvalues(unsigned passw[7], unsigned passh[7], size_t filter_passstart[8], - size_t padded_passstart[8], size_t passstart[8], unsigned w, unsigned h, unsigned bpp) -{ - /*the passstart values have 8 values: the 8th one indicates the byte after the end of the 7th (= last) pass*/ - unsigned i; - - /*calculate width and height in pixels of each pass*/ - for(i = 0; i < 7; i++) - { - passw[i] = (w + ADAM7_DX[i] - ADAM7_IX[i] - 1) / ADAM7_DX[i]; - passh[i] = (h + ADAM7_DY[i] - ADAM7_IY[i] - 1) / ADAM7_DY[i]; - if(passw[i] == 0) passh[i] = 0; - if(passh[i] == 0) passw[i] = 0; - } - - filter_passstart[0] = padded_passstart[0] = passstart[0] = 0; - for(i = 0; i < 7; i++) - { - /*if passw[i] is 0, it's 0 bytes, not 1 (no filtertype-byte)*/ - filter_passstart[i + 1] = filter_passstart[i] - + ((passw[i] && passh[i]) ? passh[i] * (1 + (passw[i] * bpp + 7) / 8) : 0); - /*bits padded if needed to fill full byte at end of each scanline*/ - padded_passstart[i + 1] = padded_passstart[i] + passh[i] * ((passw[i] * bpp + 7) / 8); - /*only padded at end of reduced image*/ - passstart[i + 1] = passstart[i] + (passh[i] * passw[i] * bpp + 7) / 8; - } -} - -#ifdef LODEPNG_COMPILE_DECODER - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / PNG Decoder / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -/*read the information from the header and store it in the LodePNGInfo. return value is error*/ -unsigned lodepng_inspect(unsigned* w, unsigned* h, LodePNGState* state, - const unsigned char* in, size_t insize) -{ - LodePNGInfo* info = &state->info_png; - if(insize == 0 || in == 0) - { - CERROR_RETURN_ERROR(state->error, 48); /*error: the given data is empty*/ - } - if(insize < 29) - { - CERROR_RETURN_ERROR(state->error, 27); /*error: the data length is smaller than the length of a PNG header*/ - } - - /*when decoding a new PNG image, make sure all parameters created after previous decoding are reset*/ - lodepng_info_cleanup(info); - lodepng_info_init(info); - - if(in[0] != 137 || in[1] != 80 || in[2] != 78 || in[3] != 71 - || in[4] != 13 || in[5] != 10 || in[6] != 26 || in[7] != 10) - { - CERROR_RETURN_ERROR(state->error, 28); /*error: the first 8 bytes are not the correct PNG signature*/ - } - if(in[12] != 'I' || in[13] != 'H' || in[14] != 'D' || in[15] != 'R') - { - CERROR_RETURN_ERROR(state->error, 29); /*error: it doesn't start with a IHDR chunk!*/ - } - - /*read the values given in the header*/ - *w = lodepng_read32bitInt(&in[16]); - *h = lodepng_read32bitInt(&in[20]); - info->color.bitdepth = in[24]; - info->color.colortype = (LodePNGColorType)in[25]; - info->compression_method = in[26]; - info->filter_method = in[27]; - info->interlace_method = in[28]; - - if(!state->decoder.ignore_crc) - { - unsigned CRC = lodepng_read32bitInt(&in[29]); - unsigned checksum = lodepng_crc32(&in[12], 17); - if(CRC != checksum) - { - CERROR_RETURN_ERROR(state->error, 57); /*invalid CRC*/ - } - } - - /*error: only compression method 0 is allowed in the specification*/ - if(info->compression_method != 0) CERROR_RETURN_ERROR(state->error, 32); - /*error: only filter method 0 is allowed in the specification*/ - if(info->filter_method != 0) CERROR_RETURN_ERROR(state->error, 33); - /*error: only interlace methods 0 and 1 exist in the specification*/ - if(info->interlace_method > 1) CERROR_RETURN_ERROR(state->error, 34); - - state->error = checkColorValidity(info->color.colortype, info->color.bitdepth); - return state->error; -} - -static unsigned unfilterScanline(unsigned char* recon, const unsigned char* scanline, const unsigned char* precon, - size_t bytewidth, unsigned char filterType, size_t length) -{ - /* - For PNG filter method 0 - unfilter a PNG image scanline by scanline. when the pixels are smaller than 1 byte, - the filter works byte per byte (bytewidth = 1) - precon is the previous unfiltered scanline, recon the result, scanline the current one - the incoming scanlines do NOT include the filtertype byte, that one is given in the parameter filterType instead - recon and scanline MAY be the same memory address! precon must be disjoint. - */ - - size_t i; - switch(filterType) - { - case 0: - for(i = 0; i < length; i++) recon[i] = scanline[i]; - break; - case 1: - for(i = 0; i < bytewidth; i++) recon[i] = scanline[i]; - for(i = bytewidth; i < length; i++) recon[i] = scanline[i] + recon[i - bytewidth]; - break; - case 2: - if(precon) - { - for(i = 0; i < length; i++) recon[i] = scanline[i] + precon[i]; - } - else - { - for(i = 0; i < length; i++) recon[i] = scanline[i]; - } - break; - case 3: - if(precon) - { - for(i = 0; i < bytewidth; i++) recon[i] = scanline[i] + precon[i] / 2; - for(i = bytewidth; i < length; i++) recon[i] = scanline[i] + ((recon[i - bytewidth] + precon[i]) / 2); - } - else - { - for(i = 0; i < bytewidth; i++) recon[i] = scanline[i]; - for(i = bytewidth; i < length; i++) recon[i] = scanline[i] + recon[i - bytewidth] / 2; - } - break; - case 4: - if(precon) - { - for(i = 0; i < bytewidth; i++) - { - recon[i] = (scanline[i] + precon[i]); /*paethPredictor(0, precon[i], 0) is always precon[i]*/ - } - for(i = bytewidth; i < length; i++) - { - recon[i] = (scanline[i] + paethPredictor(recon[i - bytewidth], precon[i], precon[i - bytewidth])); - } - } - else - { - for(i = 0; i < bytewidth; i++) - { - recon[i] = scanline[i]; - } - for(i = bytewidth; i < length; i++) - { - /*paethPredictor(recon[i - bytewidth], 0, 0) is always recon[i - bytewidth]*/ - recon[i] = (scanline[i] + recon[i - bytewidth]); - } - } - break; - default: return 36; /*error: unexisting filter type given*/ - } - return 0; -} - -static unsigned unfilter(unsigned char* out, const unsigned char* in, unsigned w, unsigned h, unsigned bpp) -{ - /* - For PNG filter method 0 - this function unfilters a single image (e.g. without interlacing this is called once, with Adam7 seven times) - out must have enough bytes allocated already, in must have the scanlines + 1 filtertype byte per scanline - w and h are image dimensions or dimensions of reduced image, bpp is bits per pixel - in and out are allowed to be the same memory address (but aren't the same size since in has the extra filter bytes) - */ - - unsigned y; - unsigned char* prevline = 0; - - /*bytewidth is used for filtering, is 1 when bpp < 8, number of bytes per pixel otherwise*/ - size_t bytewidth = (bpp + 7) / 8; - size_t linebytes = (w * bpp + 7) / 8; - - for(y = 0; y < h; y++) - { - size_t outindex = linebytes * y; - size_t inindex = (1 + linebytes) * y; /*the extra filterbyte added to each row*/ - unsigned char filterType = in[inindex]; - - CERROR_TRY_RETURN(unfilterScanline(&out[outindex], &in[inindex + 1], prevline, bytewidth, filterType, linebytes)); - - prevline = &out[outindex]; - } - - return 0; -} - -/* -in: Adam7 interlaced image, with no padding bits between scanlines, but between - reduced images so that each reduced image starts at a byte. -out: the same pixels, but re-ordered so that they're now a non-interlaced image with size w*h -bpp: bits per pixel -out has the following size in bits: w * h * bpp. -in is possibly bigger due to padding bits between reduced images. -out must be big enough AND must be 0 everywhere if bpp < 8 in the current implementation -(because that's likely a little bit faster) -NOTE: comments about padding bits are only relevant if bpp < 8 -*/ -static void Adam7_deinterlace(unsigned char* out, const unsigned char* in, unsigned w, unsigned h, unsigned bpp) -{ - unsigned passw[7], passh[7]; - size_t filter_passstart[8], padded_passstart[8], passstart[8]; - unsigned i; - - Adam7_getpassvalues(passw, passh, filter_passstart, padded_passstart, passstart, w, h, bpp); - - if(bpp >= 8) - { - for(i = 0; i < 7; i++) - { - unsigned x, y, b; - size_t bytewidth = bpp / 8; - for(y = 0; y < passh[i]; y++) - for(x = 0; x < passw[i]; x++) - { - size_t pixelinstart = passstart[i] + (y * passw[i] + x) * bytewidth; - size_t pixeloutstart = ((ADAM7_IY[i] + y * ADAM7_DY[i]) * w + ADAM7_IX[i] + x * ADAM7_DX[i]) * bytewidth; - for(b = 0; b < bytewidth; b++) - { - out[pixeloutstart + b] = in[pixelinstart + b]; - } - } - } - } - else /*bpp < 8: Adam7 with pixels < 8 bit is a bit trickier: with bit pointers*/ - { - for(i = 0; i < 7; i++) - { - unsigned x, y, b; - unsigned ilinebits = bpp * passw[i]; - unsigned olinebits = bpp * w; - size_t obp, ibp; /*bit pointers (for out and in buffer)*/ - for(y = 0; y < passh[i]; y++) - for(x = 0; x < passw[i]; x++) - { - ibp = (8 * passstart[i]) + (y * ilinebits + x * bpp); - obp = (ADAM7_IY[i] + y * ADAM7_DY[i]) * olinebits + (ADAM7_IX[i] + x * ADAM7_DX[i]) * bpp; - for(b = 0; b < bpp; b++) - { - unsigned char bit = readBitFromReversedStream(&ibp, in); - /*note that this function assumes the out buffer is completely 0, use setBitOfReversedStream otherwise*/ - setBitOfReversedStream0(&obp, out, bit); - } - } - } - } -} - -static void removePaddingBits(unsigned char* out, const unsigned char* in, - size_t olinebits, size_t ilinebits, unsigned h) -{ - /* - After filtering there are still padding bits if scanlines have non multiple of 8 bit amounts. They need - to be removed (except at last scanline of (Adam7-reduced) image) before working with pure image buffers - for the Adam7 code, the color convert code and the output to the user. - in and out are allowed to be the same buffer, in may also be higher but still overlapping; in must - have >= ilinebits*h bits, out must have >= olinebits*h bits, olinebits must be <= ilinebits - also used to move bits after earlier such operations happened, e.g. in a sequence of reduced images from Adam7 - only useful if (ilinebits - olinebits) is a value in the range 1..7 - */ - unsigned y; - size_t diff = ilinebits - olinebits; - size_t ibp = 0, obp = 0; /*input and output bit pointers*/ - for(y = 0; y < h; y++) - { - size_t x; - for(x = 0; x < olinebits; x++) - { - unsigned char bit = readBitFromReversedStream(&ibp, in); - setBitOfReversedStream(&obp, out, bit); - } - ibp += diff; - } -} - -/*out must be buffer big enough to contain full image, and in must contain the full decompressed data from -the IDAT chunks (with filter index bytes and possible padding bits) -return value is error*/ -static unsigned postProcessScanlines(unsigned char* out, unsigned char* in, - unsigned w, unsigned h, const LodePNGInfo* info_png) -{ - /* - This function converts the filtered-padded-interlaced data into pure 2D image buffer with the PNG's colortype. - Steps: - *) if no Adam7: 1) unfilter 2) remove padding bits (= posible extra bits per scanline if bpp < 8) - *) if adam7: 1) 7x unfilter 2) 7x remove padding bits 3) Adam7_deinterlace - NOTE: the in buffer will be overwritten with intermediate data! - */ - unsigned bpp = lodepng_get_bpp(&info_png->color); - if(bpp == 0) return 31; /*error: invalid colortype*/ - - if(info_png->interlace_method == 0) - { - if(bpp < 8 && w * bpp != ((w * bpp + 7) / 8) * 8) - { - CERROR_TRY_RETURN(unfilter(in, in, w, h, bpp)); - removePaddingBits(out, in, w * bpp, ((w * bpp + 7) / 8) * 8, h); - } - /*we can immediatly filter into the out buffer, no other steps needed*/ - else CERROR_TRY_RETURN(unfilter(out, in, w, h, bpp)); - } - else /*interlace_method is 1 (Adam7)*/ - { - unsigned passw[7], passh[7]; size_t filter_passstart[8], padded_passstart[8], passstart[8]; - unsigned i; - - Adam7_getpassvalues(passw, passh, filter_passstart, padded_passstart, passstart, w, h, bpp); - - for(i = 0; i < 7; i++) - { - CERROR_TRY_RETURN(unfilter(&in[padded_passstart[i]], &in[filter_passstart[i]], passw[i], passh[i], bpp)); - /*TODO: possible efficiency improvement: if in this reduced image the bits fit nicely in 1 scanline, - move bytes instead of bits or move not at all*/ - if(bpp < 8) - { - /*remove padding bits in scanlines; after this there still may be padding - bits between the different reduced images: each reduced image still starts nicely at a byte*/ - removePaddingBits(&in[passstart[i]], &in[padded_passstart[i]], passw[i] * bpp, - ((passw[i] * bpp + 7) / 8) * 8, passh[i]); - } - } - - Adam7_deinterlace(out, in, w, h, bpp); - } - - return 0; -} - -static unsigned readChunk_PLTE(LodePNGColorMode* color, const unsigned char* data, size_t chunkLength) -{ - unsigned pos = 0, i; - if(color->palette) lodepng_free(color->palette); - color->palettesize = chunkLength / 3; - color->palette = (unsigned char*)lodepng_malloc(4 * color->palettesize); - if(!color->palette && color->palettesize) - { - color->palettesize = 0; - return 83; /*alloc fail*/ - } - if(color->palettesize > 256) return 38; /*error: palette too big*/ - - for(i = 0; i < color->palettesize; i++) - { - color->palette[4 * i + 0] = data[pos++]; /*R*/ - color->palette[4 * i + 1] = data[pos++]; /*G*/ - color->palette[4 * i + 2] = data[pos++]; /*B*/ - color->palette[4 * i + 3] = 255; /*alpha*/ - } - - return 0; /* OK */ -} - -static unsigned readChunk_tRNS(LodePNGColorMode* color, const unsigned char* data, size_t chunkLength) -{ - unsigned i; - if(color->colortype == LCT_PALETTE) - { - /*error: more alpha values given than there are palette entries*/ - if(chunkLength > color->palettesize) return 38; - - for(i = 0; i < chunkLength; i++) color->palette[4 * i + 3] = data[i]; - } - else if(color->colortype == LCT_GREY) - { - /*error: this chunk must be 2 bytes for greyscale image*/ - if(chunkLength != 2) return 30; - - color->key_defined = 1; - color->key_r = color->key_g = color->key_b = 256u * data[0] + data[1]; - } - else if(color->colortype == LCT_RGB) - { - /*error: this chunk must be 6 bytes for RGB image*/ - if(chunkLength != 6) return 41; - - color->key_defined = 1; - color->key_r = 256u * data[0] + data[1]; - color->key_g = 256u * data[2] + data[3]; - color->key_b = 256u * data[4] + data[5]; - } - else return 42; /*error: tRNS chunk not allowed for other color models*/ - - return 0; /* OK */ -} - - -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS -/*background color chunk (bKGD)*/ -static unsigned readChunk_bKGD(LodePNGInfo* info, const unsigned char* data, size_t chunkLength) -{ - if(info->color.colortype == LCT_PALETTE) - { - /*error: this chunk must be 1 byte for indexed color image*/ - if(chunkLength != 1) return 43; - - info->background_defined = 1; - info->background_r = info->background_g = info->background_b = data[0]; - } - else if(info->color.colortype == LCT_GREY || info->color.colortype == LCT_GREY_ALPHA) - { - /*error: this chunk must be 2 bytes for greyscale image*/ - if(chunkLength != 2) return 44; - - info->background_defined = 1; - info->background_r = info->background_g = info->background_b = 256u * data[0] + data[1]; - } - else if(info->color.colortype == LCT_RGB || info->color.colortype == LCT_RGBA) - { - /*error: this chunk must be 6 bytes for greyscale image*/ - if(chunkLength != 6) return 45; - - info->background_defined = 1; - info->background_r = 256u * data[0] + data[1]; - info->background_g = 256u * data[2] + data[3]; - info->background_b = 256u * data[4] + data[5]; - } - - return 0; /* OK */ -} - -/*text chunk (tEXt)*/ -static unsigned readChunk_tEXt(LodePNGInfo* info, const unsigned char* data, size_t chunkLength) -{ - unsigned error = 0; - char *key = 0, *str = 0; - unsigned i; - - while(!error) /*not really a while loop, only used to break on error*/ - { - unsigned length, string2_begin; - - length = 0; - while(length < chunkLength && data[length] != 0) length++; - /*even though it's not allowed by the standard, no error is thrown if - there's no null termination char, if the text is empty*/ - if(length < 1 || length > 79) CERROR_BREAK(error, 89); /*keyword too short or long*/ - - key = (char*)lodepng_malloc(length + 1); - if(!key) CERROR_BREAK(error, 83); /*alloc fail*/ - - key[length] = 0; - for(i = 0; i < length; i++) key[i] = (char)data[i]; - - string2_begin = length + 1; /*skip keyword null terminator*/ - - length = chunkLength < string2_begin ? 0 : chunkLength - string2_begin; - str = (char*)lodepng_malloc(length + 1); - if(!str) CERROR_BREAK(error, 83); /*alloc fail*/ - - str[length] = 0; - for(i = 0; i < length; i++) str[i] = (char)data[string2_begin + i]; - - error = lodepng_add_text(info, key, str); - - break; - } - - lodepng_free(key); - lodepng_free(str); - - return error; -} - -/*compressed text chunk (zTXt)*/ -static unsigned readChunk_zTXt(LodePNGInfo* info, const LodePNGDecompressSettings* zlibsettings, - const unsigned char* data, size_t chunkLength) -{ - unsigned error = 0; - unsigned i; - - unsigned length, string2_begin; - char *key = 0; - ucvector decoded; - - ucvector_init(&decoded); - - while(!error) /*not really a while loop, only used to break on error*/ - { - for(length = 0; length < chunkLength && data[length] != 0; length++) ; - if(length + 2 >= chunkLength) CERROR_BREAK(error, 75); /*no null termination, corrupt?*/ - if(length < 1 || length > 79) CERROR_BREAK(error, 89); /*keyword too short or long*/ - - key = (char*)lodepng_malloc(length + 1); - if(!key) CERROR_BREAK(error, 83); /*alloc fail*/ - - key[length] = 0; - for(i = 0; i < length; i++) key[i] = (char)data[i]; - - if(data[length + 1] != 0) CERROR_BREAK(error, 72); /*the 0 byte indicating compression must be 0*/ - - string2_begin = length + 2; - if(string2_begin > chunkLength) CERROR_BREAK(error, 75); /*no null termination, corrupt?*/ - - length = chunkLength - string2_begin; - /*will fail if zlib error, e.g. if length is too small*/ - error = zlib_decompress(&decoded.data, &decoded.size, - (unsigned char*)(&data[string2_begin]), - length, zlibsettings); - if(error) break; - ucvector_push_back(&decoded, 0); - - error = lodepng_add_text(info, key, (char*)decoded.data); - - break; - } - - lodepng_free(key); - ucvector_cleanup(&decoded); - - return error; -} - -/*international text chunk (iTXt)*/ -static unsigned readChunk_iTXt(LodePNGInfo* info, const LodePNGDecompressSettings* zlibsettings, - const unsigned char* data, size_t chunkLength) -{ - unsigned error = 0; - unsigned i; - - unsigned length, begin, compressed; - char *key = 0, *langtag = 0, *transkey = 0; - ucvector decoded; - ucvector_init(&decoded); - - while(!error) /*not really a while loop, only used to break on error*/ - { - /*Quick check if the chunk length isn't too small. Even without check - it'd still fail with other error checks below if it's too short. This just gives a different error code.*/ - if(chunkLength < 5) CERROR_BREAK(error, 30); /*iTXt chunk too short*/ - - /*read the key*/ - for(length = 0; length < chunkLength && data[length] != 0; length++) ; - if(length + 3 >= chunkLength) CERROR_BREAK(error, 75); /*no null termination char, corrupt?*/ - if(length < 1 || length > 79) CERROR_BREAK(error, 89); /*keyword too short or long*/ - - key = (char*)lodepng_malloc(length + 1); - if(!key) CERROR_BREAK(error, 83); /*alloc fail*/ - - key[length] = 0; - for(i = 0; i < length; i++) key[i] = (char)data[i]; - - /*read the compression method*/ - compressed = data[length + 1]; - if(data[length + 2] != 0) CERROR_BREAK(error, 72); /*the 0 byte indicating compression must be 0*/ - - /*even though it's not allowed by the standard, no error is thrown if - there's no null termination char, if the text is empty for the next 3 texts*/ - - /*read the langtag*/ - begin = length + 3; - length = 0; - for(i = begin; i < chunkLength && data[i] != 0; i++) length++; - - langtag = (char*)lodepng_malloc(length + 1); - if(!langtag) CERROR_BREAK(error, 83); /*alloc fail*/ - - langtag[length] = 0; - for(i = 0; i < length; i++) langtag[i] = (char)data[begin + i]; - - /*read the transkey*/ - begin += length + 1; - length = 0; - for(i = begin; i < chunkLength && data[i] != 0; i++) length++; - - transkey = (char*)lodepng_malloc(length + 1); - if(!transkey) CERROR_BREAK(error, 83); /*alloc fail*/ - - transkey[length] = 0; - for(i = 0; i < length; i++) transkey[i] = (char)data[begin + i]; - - /*read the actual text*/ - begin += length + 1; - - length = chunkLength < begin ? 0 : chunkLength - begin; - - if(compressed) - { - /*will fail if zlib error, e.g. if length is too small*/ - error = zlib_decompress(&decoded.data, &decoded.size, - (unsigned char*)(&data[begin]), - length, zlibsettings); - if(error) break; - if(decoded.allocsize < decoded.size) decoded.allocsize = decoded.size; - ucvector_push_back(&decoded, 0); - } - else - { - if(!ucvector_resize(&decoded, length + 1)) CERROR_BREAK(error, 83 /*alloc fail*/); - - decoded.data[length] = 0; - for(i = 0; i < length; i++) decoded.data[i] = data[begin + i]; - } - - error = lodepng_add_itext(info, key, langtag, transkey, (char*)decoded.data); - - break; - } - - lodepng_free(key); - lodepng_free(langtag); - lodepng_free(transkey); - ucvector_cleanup(&decoded); - - return error; -} - -static unsigned readChunk_tIME(LodePNGInfo* info, const unsigned char* data, size_t chunkLength) -{ - if(chunkLength != 7) return 73; /*invalid tIME chunk size*/ - - info->time_defined = 1; - info->time.year = 256u * data[0] + data[1]; - info->time.month = data[2]; - info->time.day = data[3]; - info->time.hour = data[4]; - info->time.minute = data[5]; - info->time.second = data[6]; - - return 0; /* OK */ -} - -static unsigned readChunk_pHYs(LodePNGInfo* info, const unsigned char* data, size_t chunkLength) -{ - if(chunkLength != 9) return 74; /*invalid pHYs chunk size*/ - - info->phys_defined = 1; - info->phys_x = 16777216u * data[0] + 65536u * data[1] + 256u * data[2] + data[3]; - info->phys_y = 16777216u * data[4] + 65536u * data[5] + 256u * data[6] + data[7]; - info->phys_unit = data[8]; - - return 0; /* OK */ -} -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - -/*read a PNG, the result will be in the same color type as the PNG (hence "generic")*/ -static void decodeGeneric(unsigned char** out, unsigned* w, unsigned* h, - LodePNGState* state, - const unsigned char* in, size_t insize) -{ - unsigned char IEND = 0; - const unsigned char* chunk; - size_t i; - ucvector idat; /*the data from idat chunks*/ - ucvector scanlines; - size_t predict; - - /*for unknown chunk order*/ - unsigned unknown = 0; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - unsigned critical_pos = 1; /*1 = after IHDR, 2 = after PLTE, 3 = after IDAT*/ -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - - /*provide some proper output values if error will happen*/ - *out = 0; - - state->error = lodepng_inspect(w, h, state, in, insize); /*reads header and resets other parameters in state->info_png*/ - if(state->error) return; - - ucvector_init(&idat); - chunk = &in[33]; /*first byte of the first chunk after the header*/ - - /*loop through the chunks, ignoring unknown chunks and stopping at IEND chunk. - IDAT data is put at the start of the in buffer*/ - while(!IEND && !state->error) - { - unsigned chunkLength; - const unsigned char* data; /*the data in the chunk*/ - - /*error: size of the in buffer too small to contain next chunk*/ - if((size_t)((chunk - in) + 12) > insize || chunk < in) CERROR_BREAK(state->error, 30); - - /*length of the data of the chunk, excluding the length bytes, chunk type and CRC bytes*/ - chunkLength = lodepng_chunk_length(chunk); - /*error: chunk length larger than the max PNG chunk size*/ - if(chunkLength > 2147483647) CERROR_BREAK(state->error, 63); - - if((size_t)((chunk - in) + chunkLength + 12) > insize || (chunk + chunkLength + 12) < in) - { - CERROR_BREAK(state->error, 64); /*error: size of the in buffer too small to contain next chunk*/ - } - - data = lodepng_chunk_data_const(chunk); - - /*IDAT chunk, containing compressed image data*/ - if(lodepng_chunk_type_equals(chunk, "IDAT")) - { - size_t oldsize = idat.size; - if(!ucvector_resize(&idat, oldsize + chunkLength)) CERROR_BREAK(state->error, 83 /*alloc fail*/); - for(i = 0; i < chunkLength; i++) idat.data[oldsize + i] = data[i]; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - critical_pos = 3; -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - } - /*IEND chunk*/ - else if(lodepng_chunk_type_equals(chunk, "IEND")) - { - IEND = 1; - } - /*palette chunk (PLTE)*/ - else if(lodepng_chunk_type_equals(chunk, "PLTE")) - { - state->error = readChunk_PLTE(&state->info_png.color, data, chunkLength); - if(state->error) break; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - critical_pos = 2; -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - } - /*palette transparency chunk (tRNS)*/ - else if(lodepng_chunk_type_equals(chunk, "tRNS")) - { - state->error = readChunk_tRNS(&state->info_png.color, data, chunkLength); - if(state->error) break; - } -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - /*background color chunk (bKGD)*/ - else if(lodepng_chunk_type_equals(chunk, "bKGD")) - { - state->error = readChunk_bKGD(&state->info_png, data, chunkLength); - if(state->error) break; - } - /*text chunk (tEXt)*/ - else if(lodepng_chunk_type_equals(chunk, "tEXt")) - { - if(state->decoder.read_text_chunks) - { - state->error = readChunk_tEXt(&state->info_png, data, chunkLength); - if(state->error) break; - } - } - /*compressed text chunk (zTXt)*/ - else if(lodepng_chunk_type_equals(chunk, "zTXt")) - { - if(state->decoder.read_text_chunks) - { - state->error = readChunk_zTXt(&state->info_png, &state->decoder.zlibsettings, data, chunkLength); - if(state->error) break; - } - } - /*international text chunk (iTXt)*/ - else if(lodepng_chunk_type_equals(chunk, "iTXt")) - { - if(state->decoder.read_text_chunks) - { - state->error = readChunk_iTXt(&state->info_png, &state->decoder.zlibsettings, data, chunkLength); - if(state->error) break; - } - } - else if(lodepng_chunk_type_equals(chunk, "tIME")) - { - state->error = readChunk_tIME(&state->info_png, data, chunkLength); - if(state->error) break; - } - else if(lodepng_chunk_type_equals(chunk, "pHYs")) - { - state->error = readChunk_pHYs(&state->info_png, data, chunkLength); - if(state->error) break; - } -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - else /*it's not an implemented chunk type, so ignore it: skip over the data*/ - { - /*error: unknown critical chunk (5th bit of first byte of chunk type is 0)*/ - if(!lodepng_chunk_ancillary(chunk)) CERROR_BREAK(state->error, 69); - - unknown = 1; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - if(state->decoder.remember_unknown_chunks) - { - state->error = lodepng_chunk_append(&state->info_png.unknown_chunks_data[critical_pos - 1], - &state->info_png.unknown_chunks_size[critical_pos - 1], chunk); - if(state->error) break; - } -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - } - - if(!state->decoder.ignore_crc && !unknown) /*check CRC if wanted, only on known chunk types*/ - { - if(lodepng_chunk_check_crc(chunk)) CERROR_BREAK(state->error, 57); /*invalid CRC*/ - } - - if(!IEND) chunk = lodepng_chunk_next_const(chunk); - } - - ucvector_init(&scanlines); - /*predict output size, to allocate exact size for output buffer to avoid more dynamic allocation. - The prediction is currently not correct for interlaced PNG images.*/ - predict = lodepng_get_raw_size_idat(*w, *h, &state->info_png.color) + *h; - if(!state->error && !ucvector_reserve(&scanlines, predict)) state->error = 83; /*alloc fail*/ - if(!state->error) - { - state->error = zlib_decompress(&scanlines.data, &scanlines.size, idat.data, - idat.size, &state->decoder.zlibsettings); - } - ucvector_cleanup(&idat); - - if(!state->error) - { - ucvector outv; - ucvector_init(&outv); - if(!ucvector_resizev(&outv, - lodepng_get_raw_size(*w, *h, &state->info_png.color), 0)) state->error = 83; /*alloc fail*/ - if(!state->error) state->error = postProcessScanlines(outv.data, scanlines.data, *w, *h, &state->info_png); - *out = outv.data; - } - ucvector_cleanup(&scanlines); -} - -unsigned lodepng_decode(unsigned char** out, unsigned* w, unsigned* h, - LodePNGState* state, - const unsigned char* in, size_t insize) -{ - *out = 0; - decodeGeneric(out, w, h, state, in, insize); - if(state->error) return state->error; - if(!state->decoder.color_convert || lodepng_color_mode_equal(&state->info_raw, &state->info_png.color)) - { - /*same color type, no copying or converting of data needed*/ - /*store the info_png color settings on the info_raw so that the info_raw still reflects what colortype - the raw image has to the end user*/ - if(!state->decoder.color_convert) - { - state->error = lodepng_color_mode_copy(&state->info_raw, &state->info_png.color); - if(state->error) return state->error; - } - } - else - { - /*color conversion needed; sort of copy of the data*/ - unsigned char* data = *out; - size_t outsize; - - /*TODO: check if this works according to the statement in the documentation: "The converter can convert - from greyscale input color type, to 8-bit greyscale or greyscale with alpha"*/ - if(!(state->info_raw.colortype == LCT_RGB || state->info_raw.colortype == LCT_RGBA) - && !(state->info_raw.bitdepth == 8)) - { - return 56; /*unsupported color mode conversion*/ - } - - outsize = lodepng_get_raw_size(*w, *h, &state->info_raw); - *out = (unsigned char*)lodepng_malloc(outsize); - if(!(*out)) - { - state->error = 83; /*alloc fail*/ - } - else state->error = lodepng_convert(*out, data, &state->info_raw, - &state->info_png.color, *w, *h); - lodepng_free(data); - } - return state->error; -} - -unsigned lodepng_decode_memory(unsigned char** out, unsigned* w, unsigned* h, const unsigned char* in, - size_t insize, LodePNGColorType colortype, unsigned bitdepth) -{ - unsigned error; - LodePNGState state; - lodepng_state_init(&state); - state.info_raw.colortype = colortype; - state.info_raw.bitdepth = bitdepth; - error = lodepng_decode(out, w, h, &state, in, insize); - lodepng_state_cleanup(&state); - return error; -} - -unsigned lodepng_decode32(unsigned char** out, unsigned* w, unsigned* h, const unsigned char* in, size_t insize) -{ - return lodepng_decode_memory(out, w, h, in, insize, LCT_RGBA, 8); -} - -unsigned lodepng_decode24(unsigned char** out, unsigned* w, unsigned* h, const unsigned char* in, size_t insize) -{ - return lodepng_decode_memory(out, w, h, in, insize, LCT_RGB, 8); -} - -#ifdef LODEPNG_COMPILE_DISK -unsigned lodepng_decode_file(unsigned char** out, unsigned* w, unsigned* h, const char* filename, - LodePNGColorType colortype, unsigned bitdepth) -{ - unsigned char* buffer; - size_t buffersize; - unsigned error; - error = lodepng_load_file(&buffer, &buffersize, filename); - if(!error) error = lodepng_decode_memory(out, w, h, buffer, buffersize, colortype, bitdepth); - lodepng_free(buffer); - return error; -} - -unsigned lodepng_decode32_file(unsigned char** out, unsigned* w, unsigned* h, const char* filename) -{ - return lodepng_decode_file(out, w, h, filename, LCT_RGBA, 8); -} - -unsigned lodepng_decode24_file(unsigned char** out, unsigned* w, unsigned* h, const char* filename) -{ - return lodepng_decode_file(out, w, h, filename, LCT_RGB, 8); -} -#endif /*LODEPNG_COMPILE_DISK*/ - -void lodepng_decoder_settings_init(LodePNGDecoderSettings* settings) -{ - settings->color_convert = 1; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - settings->read_text_chunks = 1; - settings->remember_unknown_chunks = 0; -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - settings->ignore_crc = 0; - lodepng_decompress_settings_init(&settings->zlibsettings); -} - -#endif /*LODEPNG_COMPILE_DECODER*/ - -#if defined(LODEPNG_COMPILE_DECODER) || defined(LODEPNG_COMPILE_ENCODER) - -void lodepng_state_init(LodePNGState* state) -{ -#ifdef LODEPNG_COMPILE_DECODER - lodepng_decoder_settings_init(&state->decoder); -#endif /*LODEPNG_COMPILE_DECODER*/ -#ifdef LODEPNG_COMPILE_ENCODER - lodepng_encoder_settings_init(&state->encoder); -#endif /*LODEPNG_COMPILE_ENCODER*/ - lodepng_color_mode_init(&state->info_raw); - lodepng_info_init(&state->info_png); - state->error = 1; -} - -void lodepng_state_cleanup(LodePNGState* state) -{ - lodepng_color_mode_cleanup(&state->info_raw); - lodepng_info_cleanup(&state->info_png); -} - -void lodepng_state_copy(LodePNGState* dest, const LodePNGState* source) -{ - lodepng_state_cleanup(dest); - *dest = *source; - lodepng_color_mode_init(&dest->info_raw); - lodepng_info_init(&dest->info_png); - dest->error = lodepng_color_mode_copy(&dest->info_raw, &source->info_raw); if(dest->error) return; - dest->error = lodepng_info_copy(&dest->info_png, &source->info_png); if(dest->error) return; -} - -#endif /* defined(LODEPNG_COMPILE_DECODER) || defined(LODEPNG_COMPILE_ENCODER) */ - -#ifdef LODEPNG_COMPILE_ENCODER - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / PNG Encoder / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -/*chunkName must be string of 4 characters*/ -static unsigned addChunk(ucvector* out, const char* chunkName, const unsigned char* data, size_t length) -{ - CERROR_TRY_RETURN(lodepng_chunk_create(&out->data, &out->size, (unsigned)length, chunkName, data)); - out->allocsize = out->size; /*fix the allocsize again*/ - return 0; -} - -static void writeSignature(ucvector* out) -{ - /*8 bytes PNG signature, aka the magic bytes*/ - ucvector_push_back(out, 137); - ucvector_push_back(out, 80); - ucvector_push_back(out, 78); - ucvector_push_back(out, 71); - ucvector_push_back(out, 13); - ucvector_push_back(out, 10); - ucvector_push_back(out, 26); - ucvector_push_back(out, 10); -} - -static unsigned addChunk_IHDR(ucvector* out, unsigned w, unsigned h, - LodePNGColorType colortype, unsigned bitdepth, unsigned interlace_method) -{ - unsigned error = 0; - ucvector header; - ucvector_init(&header); - - lodepng_add32bitInt(&header, w); /*width*/ - lodepng_add32bitInt(&header, h); /*height*/ - ucvector_push_back(&header, (unsigned char)bitdepth); /*bit depth*/ - ucvector_push_back(&header, (unsigned char)colortype); /*color type*/ - ucvector_push_back(&header, 0); /*compression method*/ - ucvector_push_back(&header, 0); /*filter method*/ - ucvector_push_back(&header, interlace_method); /*interlace method*/ - - error = addChunk(out, "IHDR", header.data, header.size); - ucvector_cleanup(&header); - - return error; -} - -static unsigned addChunk_PLTE(ucvector* out, const LodePNGColorMode* info) -{ - unsigned error = 0; - size_t i; - ucvector PLTE; - ucvector_init(&PLTE); - for(i = 0; i < info->palettesize * 4; i++) - { - /*add all channels except alpha channel*/ - if(i % 4 != 3) ucvector_push_back(&PLTE, info->palette[i]); - } - error = addChunk(out, "PLTE", PLTE.data, PLTE.size); - ucvector_cleanup(&PLTE); - - return error; -} - -static unsigned addChunk_tRNS(ucvector* out, const LodePNGColorMode* info) -{ - unsigned error = 0; - size_t i; - ucvector tRNS; - ucvector_init(&tRNS); - if(info->colortype == LCT_PALETTE) - { - size_t amount = info->palettesize; - /*the tail of palette values that all have 255 as alpha, does not have to be encoded*/ - for(i = info->palettesize; i > 0; i--) - { - if(info->palette[4 * (i - 1) + 3] == 255) amount--; - else break; - } - /*add only alpha channel*/ - for(i = 0; i < amount; i++) ucvector_push_back(&tRNS, info->palette[4 * i + 3]); - } - else if(info->colortype == LCT_GREY) - { - if(info->key_defined) - { - ucvector_push_back(&tRNS, (unsigned char)(info->key_r / 256)); - ucvector_push_back(&tRNS, (unsigned char)(info->key_r % 256)); - } - } - else if(info->colortype == LCT_RGB) - { - if(info->key_defined) - { - ucvector_push_back(&tRNS, (unsigned char)(info->key_r / 256)); - ucvector_push_back(&tRNS, (unsigned char)(info->key_r % 256)); - ucvector_push_back(&tRNS, (unsigned char)(info->key_g / 256)); - ucvector_push_back(&tRNS, (unsigned char)(info->key_g % 256)); - ucvector_push_back(&tRNS, (unsigned char)(info->key_b / 256)); - ucvector_push_back(&tRNS, (unsigned char)(info->key_b % 256)); - } - } - - error = addChunk(out, "tRNS", tRNS.data, tRNS.size); - ucvector_cleanup(&tRNS); - - return error; -} - -static unsigned addChunk_IDAT(ucvector* out, const unsigned char* data, size_t datasize, - LodePNGCompressSettings* zlibsettings) -{ - ucvector zlibdata; - unsigned error = 0; - - /*compress with the Zlib compressor*/ - ucvector_init(&zlibdata); - error = zlib_compress(&zlibdata.data, &zlibdata.size, data, datasize, zlibsettings); - if(!error) error = addChunk(out, "IDAT", zlibdata.data, zlibdata.size); - ucvector_cleanup(&zlibdata); - - return error; -} - -static unsigned addChunk_IEND(ucvector* out) -{ - unsigned error = 0; - error = addChunk(out, "IEND", 0, 0); - return error; -} - -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - -static unsigned addChunk_tEXt(ucvector* out, const char* keyword, const char* textstring) -{ - unsigned error = 0; - size_t i; - ucvector text; - ucvector_init(&text); - for(i = 0; keyword[i] != 0; i++) ucvector_push_back(&text, (unsigned char)keyword[i]); - if(i < 1 || i > 79) return 89; /*error: invalid keyword size*/ - ucvector_push_back(&text, 0); /*0 termination char*/ - for(i = 0; textstring[i] != 0; i++) ucvector_push_back(&text, (unsigned char)textstring[i]); - error = addChunk(out, "tEXt", text.data, text.size); - ucvector_cleanup(&text); - - return error; -} - -static unsigned addChunk_zTXt(ucvector* out, const char* keyword, const char* textstring, - LodePNGCompressSettings* zlibsettings) -{ - unsigned error = 0; - ucvector data, compressed; - size_t i, textsize = strlen(textstring); - - ucvector_init(&data); - ucvector_init(&compressed); - for(i = 0; keyword[i] != 0; i++) ucvector_push_back(&data, (unsigned char)keyword[i]); - if(i < 1 || i > 79) return 89; /*error: invalid keyword size*/ - ucvector_push_back(&data, 0); /*0 termination char*/ - ucvector_push_back(&data, 0); /*compression method: 0*/ - - error = zlib_compress(&compressed.data, &compressed.size, - (unsigned char*)textstring, textsize, zlibsettings); - if(!error) - { - for(i = 0; i < compressed.size; i++) ucvector_push_back(&data, compressed.data[i]); - error = addChunk(out, "zTXt", data.data, data.size); - } - - ucvector_cleanup(&compressed); - ucvector_cleanup(&data); - return error; -} - -static unsigned addChunk_iTXt(ucvector* out, unsigned compressed, const char* keyword, const char* langtag, - const char* transkey, const char* textstring, LodePNGCompressSettings* zlibsettings) -{ - unsigned error = 0; - ucvector data; - size_t i, textsize = strlen(textstring); - - ucvector_init(&data); - - for(i = 0; keyword[i] != 0; i++) ucvector_push_back(&data, (unsigned char)keyword[i]); - if(i < 1 || i > 79) return 89; /*error: invalid keyword size*/ - ucvector_push_back(&data, 0); /*null termination char*/ - ucvector_push_back(&data, compressed ? 1 : 0); /*compression flag*/ - ucvector_push_back(&data, 0); /*compression method*/ - for(i = 0; langtag[i] != 0; i++) ucvector_push_back(&data, (unsigned char)langtag[i]); - ucvector_push_back(&data, 0); /*null termination char*/ - for(i = 0; transkey[i] != 0; i++) ucvector_push_back(&data, (unsigned char)transkey[i]); - ucvector_push_back(&data, 0); /*null termination char*/ - - if(compressed) - { - ucvector compressed_data; - ucvector_init(&compressed_data); - error = zlib_compress(&compressed_data.data, &compressed_data.size, - (unsigned char*)textstring, textsize, zlibsettings); - if(!error) - { - for(i = 0; i < compressed_data.size; i++) ucvector_push_back(&data, compressed_data.data[i]); - } - ucvector_cleanup(&compressed_data); - } - else /*not compressed*/ - { - for(i = 0; textstring[i] != 0; i++) ucvector_push_back(&data, (unsigned char)textstring[i]); - } - - if(!error) error = addChunk(out, "iTXt", data.data, data.size); - ucvector_cleanup(&data); - return error; -} - -static unsigned addChunk_bKGD(ucvector* out, const LodePNGInfo* info) -{ - unsigned error = 0; - ucvector bKGD; - ucvector_init(&bKGD); - if(info->color.colortype == LCT_GREY || info->color.colortype == LCT_GREY_ALPHA) - { - ucvector_push_back(&bKGD, (unsigned char)(info->background_r / 256)); - ucvector_push_back(&bKGD, (unsigned char)(info->background_r % 256)); - } - else if(info->color.colortype == LCT_RGB || info->color.colortype == LCT_RGBA) - { - ucvector_push_back(&bKGD, (unsigned char)(info->background_r / 256)); - ucvector_push_back(&bKGD, (unsigned char)(info->background_r % 256)); - ucvector_push_back(&bKGD, (unsigned char)(info->background_g / 256)); - ucvector_push_back(&bKGD, (unsigned char)(info->background_g % 256)); - ucvector_push_back(&bKGD, (unsigned char)(info->background_b / 256)); - ucvector_push_back(&bKGD, (unsigned char)(info->background_b % 256)); - } - else if(info->color.colortype == LCT_PALETTE) - { - ucvector_push_back(&bKGD, (unsigned char)(info->background_r % 256)); /*palette index*/ - } - - error = addChunk(out, "bKGD", bKGD.data, bKGD.size); - ucvector_cleanup(&bKGD); - - return error; -} - -static unsigned addChunk_tIME(ucvector* out, const LodePNGTime* time) -{ - unsigned error = 0; - unsigned char* data = (unsigned char*)lodepng_malloc(7); - if(!data) return 83; /*alloc fail*/ - data[0] = (unsigned char)(time->year / 256); - data[1] = (unsigned char)(time->year % 256); - data[2] = (unsigned char)time->month; - data[3] = (unsigned char)time->day; - data[4] = (unsigned char)time->hour; - data[5] = (unsigned char)time->minute; - data[6] = (unsigned char)time->second; - error = addChunk(out, "tIME", data, 7); - lodepng_free(data); - return error; -} - -static unsigned addChunk_pHYs(ucvector* out, const LodePNGInfo* info) -{ - unsigned error = 0; - ucvector data; - ucvector_init(&data); - - lodepng_add32bitInt(&data, info->phys_x); - lodepng_add32bitInt(&data, info->phys_y); - ucvector_push_back(&data, info->phys_unit); - - error = addChunk(out, "pHYs", data.data, data.size); - ucvector_cleanup(&data); - - return error; -} - -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - -static void filterScanline(unsigned char* out, const unsigned char* scanline, const unsigned char* prevline, - size_t length, size_t bytewidth, unsigned char filterType) -{ - size_t i; - switch(filterType) - { - case 0: /*None*/ - for(i = 0; i < length; i++) out[i] = scanline[i]; - break; - case 1: /*Sub*/ - if(prevline) - { - for(i = 0; i < bytewidth; i++) out[i] = scanline[i]; - for(i = bytewidth; i < length; i++) out[i] = scanline[i] - scanline[i - bytewidth]; - } - else - { - for(i = 0; i < bytewidth; i++) out[i] = scanline[i]; - for(i = bytewidth; i < length; i++) out[i] = scanline[i] - scanline[i - bytewidth]; - } - break; - case 2: /*Up*/ - if(prevline) - { - for(i = 0; i < length; i++) out[i] = scanline[i] - prevline[i]; - } - else - { - for(i = 0; i < length; i++) out[i] = scanline[i]; - } - break; - case 3: /*Average*/ - if(prevline) - { - for(i = 0; i < bytewidth; i++) out[i] = scanline[i] - prevline[i] / 2; - for(i = bytewidth; i < length; i++) out[i] = scanline[i] - ((scanline[i - bytewidth] + prevline[i]) / 2); - } - else - { - for(i = 0; i < bytewidth; i++) out[i] = scanline[i]; - for(i = bytewidth; i < length; i++) out[i] = scanline[i] - scanline[i - bytewidth] / 2; - } - break; - case 4: /*Paeth*/ - if(prevline) - { - /*paethPredictor(0, prevline[i], 0) is always prevline[i]*/ - for(i = 0; i < bytewidth; i++) out[i] = (scanline[i] - prevline[i]); - for(i = bytewidth; i < length; i++) - { - out[i] = (scanline[i] - paethPredictor(scanline[i - bytewidth], prevline[i], prevline[i - bytewidth])); - } - } - else - { - for(i = 0; i < bytewidth; i++) out[i] = scanline[i]; - /*paethPredictor(scanline[i - bytewidth], 0, 0) is always scanline[i - bytewidth]*/ - for(i = bytewidth; i < length; i++) out[i] = (scanline[i] - scanline[i - bytewidth]); - } - break; - default: return; /*unexisting filter type given*/ - } -} - -/* log2 approximation. A slight bit faster than std::log. */ -static float flog2(float f) -{ - float result = 0; - while(f > 32) { result += 4; f /= 16; } - while(f > 2) { result++; f /= 2; } - return result + 1.442695f * (f * f * f / 3 - 3 * f * f / 2 + 3 * f - 1.83333f); -} - -static unsigned filter(unsigned char* out, const unsigned char* in, unsigned w, unsigned h, - const LodePNGColorMode* info, const LodePNGEncoderSettings* settings) -{ - /* - For PNG filter method 0 - out must be a buffer with as size: h + (w * h * bpp + 7) / 8, because there are - the scanlines with 1 extra byte per scanline - */ - - unsigned bpp = lodepng_get_bpp(info); - /*the width of a scanline in bytes, not including the filter type*/ - size_t linebytes = (w * bpp + 7) / 8; - /*bytewidth is used for filtering, is 1 when bpp < 8, number of bytes per pixel otherwise*/ - size_t bytewidth = (bpp + 7) / 8; - const unsigned char* prevline = 0; - unsigned x, y; - unsigned error = 0; - LodePNGFilterStrategy strategy = settings->filter_strategy; - - /* - There is a heuristic called the minimum sum of absolute differences heuristic, suggested by the PNG standard: - * If the image type is Palette, or the bit depth is smaller than 8, then do not filter the image (i.e. - use fixed filtering, with the filter None). - * (The other case) If the image type is Grayscale or RGB (with or without Alpha), and the bit depth is - not smaller than 8, then use adaptive filtering heuristic as follows: independently for each row, apply - all five filters and select the filter that produces the smallest sum of absolute values per row. - This heuristic is used if filter strategy is LFS_MINSUM and filter_palette_zero is true. - - If filter_palette_zero is true and filter_strategy is not LFS_MINSUM, the above heuristic is followed, - but for "the other case", whatever strategy filter_strategy is set to instead of the minimum sum - heuristic is used. - */ - if(settings->filter_palette_zero && - (info->colortype == LCT_PALETTE || info->bitdepth < 8)) strategy = LFS_ZERO; - - if(bpp == 0) return 31; /*error: invalid color type*/ - - if(strategy == LFS_ZERO) - { - for(y = 0; y < h; y++) - { - size_t outindex = (1 + linebytes) * y; /*the extra filterbyte added to each row*/ - size_t inindex = linebytes * y; - out[outindex] = 0; /*filter type byte*/ - filterScanline(&out[outindex + 1], &in[inindex], prevline, linebytes, bytewidth, 0); - prevline = &in[inindex]; - } - } - else if(strategy == LFS_MINSUM) - { - /*adaptive filtering*/ - size_t sum[5]; - ucvector attempt[5]; /*five filtering attempts, one for each filter type*/ - size_t smallest = 0; - unsigned char type, bestType = 0; - - for(type = 0; type < 5; type++) - { - ucvector_init(&attempt[type]); - if(!ucvector_resize(&attempt[type], linebytes)) return 83; /*alloc fail*/ - } - - if(!error) - { - for(y = 0; y < h; y++) - { - /*try the 5 filter types*/ - for(type = 0; type < 5; type++) - { - filterScanline(attempt[type].data, &in[y * linebytes], prevline, linebytes, bytewidth, type); - - /*calculate the sum of the result*/ - sum[type] = 0; - if(type == 0) - { - for(x = 0; x < linebytes; x++) sum[type] += (unsigned char)(attempt[type].data[x]); - } - else - { - for(x = 0; x < linebytes; x++) - { - /*For differences, each byte should be treated as signed, values above 127 are negative - (converted to signed char). Filtertype 0 isn't a difference though, so use unsigned there. - This means filtertype 0 is almost never chosen, but that is justified.*/ - unsigned char s = attempt[type].data[x]; - sum[type] += s < 128 ? s : (255U - s); - } - } - - /*check if this is smallest sum (or if type == 0 it's the first case so always store the values)*/ - if(type == 0 || sum[type] < smallest) - { - bestType = type; - smallest = sum[type]; - } - } - - prevline = &in[y * linebytes]; - - /*now fill the out values*/ - out[y * (linebytes + 1)] = bestType; /*the first byte of a scanline will be the filter type*/ - for(x = 0; x < linebytes; x++) out[y * (linebytes + 1) + 1 + x] = attempt[bestType].data[x]; - } - } - - for(type = 0; type < 5; type++) ucvector_cleanup(&attempt[type]); - } - else if(strategy == LFS_ENTROPY) - { - float sum[5]; - ucvector attempt[5]; /*five filtering attempts, one for each filter type*/ - float smallest = 0; - unsigned type, bestType = 0; - unsigned count[256]; - - for(type = 0; type < 5; type++) - { - ucvector_init(&attempt[type]); - if(!ucvector_resize(&attempt[type], linebytes)) return 83; /*alloc fail*/ - } - - for(y = 0; y < h; y++) - { - /*try the 5 filter types*/ - for(type = 0; type < 5; type++) - { - filterScanline(attempt[type].data, &in[y * linebytes], prevline, linebytes, bytewidth, type); - for(x = 0; x < 256; x++) count[x] = 0; - for(x = 0; x < linebytes; x++) count[attempt[type].data[x]]++; - count[type]++; /*the filter type itself is part of the scanline*/ - sum[type] = 0; - for(x = 0; x < 256; x++) - { - float p = count[x] / (float)(linebytes + 1); - sum[type] += count[x] == 0 ? 0 : flog2(1 / p) * p; - } - /*check if this is smallest sum (or if type == 0 it's the first case so always store the values)*/ - if(type == 0 || sum[type] < smallest) - { - bestType = type; - smallest = sum[type]; - } - } - - prevline = &in[y * linebytes]; - - /*now fill the out values*/ - out[y * (linebytes + 1)] = bestType; /*the first byte of a scanline will be the filter type*/ - for(x = 0; x < linebytes; x++) out[y * (linebytes + 1) + 1 + x] = attempt[bestType].data[x]; - } - - for(type = 0; type < 5; type++) ucvector_cleanup(&attempt[type]); - } - else if(strategy == LFS_PREDEFINED) - { - for(y = 0; y < h; y++) - { - size_t outindex = (1 + linebytes) * y; /*the extra filterbyte added to each row*/ - size_t inindex = linebytes * y; - unsigned char type = settings->predefined_filters[y]; - out[outindex] = type; /*filter type byte*/ - filterScanline(&out[outindex + 1], &in[inindex], prevline, linebytes, bytewidth, type); - prevline = &in[inindex]; - } - } - else if(strategy == LFS_BRUTE_FORCE) - { - /*brute force filter chooser. - deflate the scanline after every filter attempt to see which one deflates best. - This is very slow and gives only slightly smaller, sometimes even larger, result*/ - size_t size[5]; - ucvector attempt[5]; /*five filtering attempts, one for each filter type*/ - size_t smallest = 0; - unsigned type = 0, bestType = 0; - unsigned char* dummy; - LodePNGCompressSettings zlibsettings = settings->zlibsettings; - /*use fixed tree on the attempts so that the tree is not adapted to the filtertype on purpose, - to simulate the true case where the tree is the same for the whole image. Sometimes it gives - better result with dynamic tree anyway. Using the fixed tree sometimes gives worse, but in rare - cases better compression. It does make this a bit less slow, so it's worth doing this.*/ - zlibsettings.btype = 1; - /*a custom encoder likely doesn't read the btype setting and is optimized for complete PNG - images only, so disable it*/ - zlibsettings.custom_zlib = 0; - zlibsettings.custom_deflate = 0; - for(type = 0; type < 5; type++) - { - ucvector_init(&attempt[type]); - ucvector_resize(&attempt[type], linebytes); /*todo: give error if resize failed*/ - } - for(y = 0; y < h; y++) /*try the 5 filter types*/ - { - for(type = 0; type < 5; type++) - { - unsigned testsize = attempt[type].size; - /*if(testsize > 8) testsize /= 8;*/ /*it already works good enough by testing a part of the row*/ - - filterScanline(attempt[type].data, &in[y * linebytes], prevline, linebytes, bytewidth, type); - size[type] = 0; - dummy = 0; - zlib_compress(&dummy, &size[type], attempt[type].data, testsize, &zlibsettings); - lodepng_free(dummy); - /*check if this is smallest size (or if type == 0 it's the first case so always store the values)*/ - if(type == 0 || size[type] < smallest) - { - bestType = type; - smallest = size[type]; - } - } - prevline = &in[y * linebytes]; - out[y * (linebytes + 1)] = bestType; /*the first byte of a scanline will be the filter type*/ - for(x = 0; x < linebytes; x++) out[y * (linebytes + 1) + 1 + x] = attempt[bestType].data[x]; - } - for(type = 0; type < 5; type++) ucvector_cleanup(&attempt[type]); - } - else return 88; /* unknown filter strategy */ - - return error; -} - -static void addPaddingBits(unsigned char* out, const unsigned char* in, - size_t olinebits, size_t ilinebits, unsigned h) -{ - /*The opposite of the removePaddingBits function - olinebits must be >= ilinebits*/ - unsigned y; - size_t diff = olinebits - ilinebits; - size_t obp = 0, ibp = 0; /*bit pointers*/ - for(y = 0; y < h; y++) - { - size_t x; - for(x = 0; x < ilinebits; x++) - { - unsigned char bit = readBitFromReversedStream(&ibp, in); - setBitOfReversedStream(&obp, out, bit); - } - /*obp += diff; --> no, fill in some value in the padding bits too, to avoid - "Use of uninitialised value of size ###" warning from valgrind*/ - for(x = 0; x < diff; x++) setBitOfReversedStream(&obp, out, 0); - } -} - -/* -in: non-interlaced image with size w*h -out: the same pixels, but re-ordered according to PNG's Adam7 interlacing, with - no padding bits between scanlines, but between reduced images so that each - reduced image starts at a byte. -bpp: bits per pixel -there are no padding bits, not between scanlines, not between reduced images -in has the following size in bits: w * h * bpp. -out is possibly bigger due to padding bits between reduced images -NOTE: comments about padding bits are only relevant if bpp < 8 -*/ -static void Adam7_interlace(unsigned char* out, const unsigned char* in, unsigned w, unsigned h, unsigned bpp) -{ - unsigned passw[7], passh[7]; - size_t filter_passstart[8], padded_passstart[8], passstart[8]; - unsigned i; - - Adam7_getpassvalues(passw, passh, filter_passstart, padded_passstart, passstart, w, h, bpp); - - if(bpp >= 8) - { - for(i = 0; i < 7; i++) - { - unsigned x, y, b; - size_t bytewidth = bpp / 8; - for(y = 0; y < passh[i]; y++) - for(x = 0; x < passw[i]; x++) - { - size_t pixelinstart = ((ADAM7_IY[i] + y * ADAM7_DY[i]) * w + ADAM7_IX[i] + x * ADAM7_DX[i]) * bytewidth; - size_t pixeloutstart = passstart[i] + (y * passw[i] + x) * bytewidth; - for(b = 0; b < bytewidth; b++) - { - out[pixeloutstart + b] = in[pixelinstart + b]; - } - } - } - } - else /*bpp < 8: Adam7 with pixels < 8 bit is a bit trickier: with bit pointers*/ - { - for(i = 0; i < 7; i++) - { - unsigned x, y, b; - unsigned ilinebits = bpp * passw[i]; - unsigned olinebits = bpp * w; - size_t obp, ibp; /*bit pointers (for out and in buffer)*/ - for(y = 0; y < passh[i]; y++) - for(x = 0; x < passw[i]; x++) - { - ibp = (ADAM7_IY[i] + y * ADAM7_DY[i]) * olinebits + (ADAM7_IX[i] + x * ADAM7_DX[i]) * bpp; - obp = (8 * passstart[i]) + (y * ilinebits + x * bpp); - for(b = 0; b < bpp; b++) - { - unsigned char bit = readBitFromReversedStream(&ibp, in); - setBitOfReversedStream(&obp, out, bit); - } - } - } - } -} - -/*out must be buffer big enough to contain uncompressed IDAT chunk data, and in must contain the full image. -return value is error**/ -static unsigned preProcessScanlines(unsigned char** out, size_t* outsize, const unsigned char* in, - unsigned w, unsigned h, - const LodePNGInfo* info_png, const LodePNGEncoderSettings* settings) -{ - /* - This function converts the pure 2D image with the PNG's colortype, into filtered-padded-interlaced data. Steps: - *) if no Adam7: 1) add padding bits (= posible extra bits per scanline if bpp < 8) 2) filter - *) if adam7: 1) Adam7_interlace 2) 7x add padding bits 3) 7x filter - */ - unsigned bpp = lodepng_get_bpp(&info_png->color); - unsigned error = 0; - - if(info_png->interlace_method == 0) - { - *outsize = h + (h * ((w * bpp + 7) / 8)); /*image size plus an extra byte per scanline + possible padding bits*/ - *out = (unsigned char*)lodepng_malloc(*outsize); - if(!(*out) && (*outsize)) error = 83; /*alloc fail*/ - - if(!error) - { - /*non multiple of 8 bits per scanline, padding bits needed per scanline*/ - if(bpp < 8 && w * bpp != ((w * bpp + 7) / 8) * 8) - { - unsigned char* padded = (unsigned char*)lodepng_malloc(h * ((w * bpp + 7) / 8)); - if(!padded) error = 83; /*alloc fail*/ - if(!error) - { - addPaddingBits(padded, in, ((w * bpp + 7) / 8) * 8, w * bpp, h); - error = filter(*out, padded, w, h, &info_png->color, settings); - } - lodepng_free(padded); - } - else - { - /*we can immediatly filter into the out buffer, no other steps needed*/ - error = filter(*out, in, w, h, &info_png->color, settings); - } - } - } - else /*interlace_method is 1 (Adam7)*/ - { - unsigned passw[7], passh[7]; - size_t filter_passstart[8], padded_passstart[8], passstart[8]; - unsigned char* adam7; - - Adam7_getpassvalues(passw, passh, filter_passstart, padded_passstart, passstart, w, h, bpp); - - *outsize = filter_passstart[7]; /*image size plus an extra byte per scanline + possible padding bits*/ - *out = (unsigned char*)lodepng_malloc(*outsize); - if(!(*out)) error = 83; /*alloc fail*/ - - adam7 = (unsigned char*)lodepng_malloc(passstart[7]); - if(!adam7 && passstart[7]) error = 83; /*alloc fail*/ - - if(!error) - { - unsigned i; - - Adam7_interlace(adam7, in, w, h, bpp); - for(i = 0; i < 7; i++) - { - if(bpp < 8) - { - unsigned char* padded = (unsigned char*)lodepng_malloc(padded_passstart[i + 1] - padded_passstart[i]); - if(!padded) ERROR_BREAK(83); /*alloc fail*/ - addPaddingBits(padded, &adam7[passstart[i]], - ((passw[i] * bpp + 7) / 8) * 8, passw[i] * bpp, passh[i]); - error = filter(&(*out)[filter_passstart[i]], padded, - passw[i], passh[i], &info_png->color, settings); - lodepng_free(padded); - } - else - { - error = filter(&(*out)[filter_passstart[i]], &adam7[padded_passstart[i]], - passw[i], passh[i], &info_png->color, settings); - } - - if(error) break; - } - } - - lodepng_free(adam7); - } - - return error; -} - -/* -palette must have 4 * palettesize bytes allocated, and given in format RGBARGBARGBARGBA... -returns 0 if the palette is opaque, -returns 1 if the palette has a single color with alpha 0 ==> color key -returns 2 if the palette is semi-translucent. -*/ -static unsigned getPaletteTranslucency(const unsigned char* palette, size_t palettesize) -{ - size_t i; - unsigned key = 0; - unsigned r = 0, g = 0, b = 0; /*the value of the color with alpha 0, so long as color keying is possible*/ - for(i = 0; i < palettesize; i++) - { - if(!key && palette[4 * i + 3] == 0) - { - r = palette[4 * i + 0]; g = palette[4 * i + 1]; b = palette[4 * i + 2]; - key = 1; - i = (size_t)(-1); /*restart from beginning, to detect earlier opaque colors with key's value*/ - } - else if(palette[4 * i + 3] != 255) return 2; - /*when key, no opaque RGB may have key's RGB*/ - else if(key && r == palette[i * 4 + 0] && g == palette[i * 4 + 1] && b == palette[i * 4 + 2]) return 2; - } - return key; -} - -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS -static unsigned addUnknownChunks(ucvector* out, unsigned char* data, size_t datasize) -{ - unsigned char* inchunk = data; - while((size_t)(inchunk - data) < datasize) - { - CERROR_TRY_RETURN(lodepng_chunk_append(&out->data, &out->size, inchunk)); - out->allocsize = out->size; /*fix the allocsize again*/ - inchunk = lodepng_chunk_next(inchunk); - } - return 0; -} -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - -unsigned lodepng_encode(unsigned char** out, size_t* outsize, - const unsigned char* image, unsigned w, unsigned h, - LodePNGState* state) -{ - LodePNGInfo info; - ucvector outv; - unsigned char* data = 0; /*uncompressed version of the IDAT chunk data*/ - size_t datasize = 0; - - /*provide some proper output values if error will happen*/ - *out = 0; - *outsize = 0; - state->error = 0; - - lodepng_info_init(&info); - lodepng_info_copy(&info, &state->info_png); - - if((info.color.colortype == LCT_PALETTE || state->encoder.force_palette) - && (info.color.palettesize == 0 || info.color.palettesize > 256)) - { - state->error = 68; /*invalid palette size, it is only allowed to be 1-256*/ - return state->error; - } - - if(state->encoder.auto_convert) - { - state->error = lodepng_auto_choose_color(&info.color, image, w, h, &state->info_raw); - } - if(state->error) return state->error; - - if(state->encoder.zlibsettings.btype > 2) - { - CERROR_RETURN_ERROR(state->error, 61); /*error: unexisting btype*/ - } - if(state->info_png.interlace_method > 1) - { - CERROR_RETURN_ERROR(state->error, 71); /*error: unexisting interlace mode*/ - } - - state->error = checkColorValidity(info.color.colortype, info.color.bitdepth); - if(state->error) return state->error; /*error: unexisting color type given*/ - state->error = checkColorValidity(state->info_raw.colortype, state->info_raw.bitdepth); - if(state->error) return state->error; /*error: unexisting color type given*/ - - if(!lodepng_color_mode_equal(&state->info_raw, &info.color)) - { - unsigned char* converted; - size_t size = (w * h * lodepng_get_bpp(&info.color) + 7) / 8; - - converted = (unsigned char*)lodepng_malloc(size); - if(!converted && size) state->error = 83; /*alloc fail*/ - if(!state->error) - { - state->error = lodepng_convert(converted, image, &info.color, &state->info_raw, w, h); - } - if(!state->error) preProcessScanlines(&data, &datasize, converted, w, h, &info, &state->encoder); - lodepng_free(converted); - } - else preProcessScanlines(&data, &datasize, image, w, h, &info, &state->encoder); - - ucvector_init(&outv); - while(!state->error) /*while only executed once, to break on error*/ - { -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - size_t i; -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - /*write signature and chunks*/ - writeSignature(&outv); - /*IHDR*/ - addChunk_IHDR(&outv, w, h, info.color.colortype, info.color.bitdepth, info.interlace_method); -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - /*unknown chunks between IHDR and PLTE*/ - if(info.unknown_chunks_data[0]) - { - state->error = addUnknownChunks(&outv, info.unknown_chunks_data[0], info.unknown_chunks_size[0]); - if(state->error) break; - } -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - /*PLTE*/ - if(info.color.colortype == LCT_PALETTE) - { - addChunk_PLTE(&outv, &info.color); - } - if(state->encoder.force_palette && (info.color.colortype == LCT_RGB || info.color.colortype == LCT_RGBA)) - { - addChunk_PLTE(&outv, &info.color); - } - /*tRNS*/ - if(info.color.colortype == LCT_PALETTE && getPaletteTranslucency(info.color.palette, info.color.palettesize) != 0) - { - addChunk_tRNS(&outv, &info.color); - } - if((info.color.colortype == LCT_GREY || info.color.colortype == LCT_RGB) && info.color.key_defined) - { - addChunk_tRNS(&outv, &info.color); - } -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - /*bKGD (must come between PLTE and the IDAt chunks*/ - if(info.background_defined) addChunk_bKGD(&outv, &info); - /*pHYs (must come before the IDAT chunks)*/ - if(info.phys_defined) addChunk_pHYs(&outv, &info); - - /*unknown chunks between PLTE and IDAT*/ - if(info.unknown_chunks_data[1]) - { - state->error = addUnknownChunks(&outv, info.unknown_chunks_data[1], info.unknown_chunks_size[1]); - if(state->error) break; - } -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - /*IDAT (multiple IDAT chunks must be consecutive)*/ - state->error = addChunk_IDAT(&outv, data, datasize, &state->encoder.zlibsettings); - if(state->error) break; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - /*tIME*/ - if(info.time_defined) addChunk_tIME(&outv, &info.time); - /*tEXt and/or zTXt*/ - for(i = 0; i < info.text_num; i++) - { - if(strlen(info.text_keys[i]) > 79) - { - state->error = 66; /*text chunk too large*/ - break; - } - if(strlen(info.text_keys[i]) < 1) - { - state->error = 67; /*text chunk too small*/ - break; - } - if(state->encoder.text_compression) - { - addChunk_zTXt(&outv, info.text_keys[i], info.text_strings[i], &state->encoder.zlibsettings); - } - else - { - addChunk_tEXt(&outv, info.text_keys[i], info.text_strings[i]); - } - } - /*LodePNG version id in text chunk*/ - if(state->encoder.add_id) - { - unsigned alread_added_id_text = 0; - for(i = 0; i < info.text_num; i++) - { - if(!strcmp(info.text_keys[i], "LodePNG")) - { - alread_added_id_text = 1; - break; - } - } - if(alread_added_id_text == 0) - { - addChunk_tEXt(&outv, "LodePNG", VERSION_STRING); /*it's shorter as tEXt than as zTXt chunk*/ - } - } - /*iTXt*/ - for(i = 0; i < info.itext_num; i++) - { - if(strlen(info.itext_keys[i]) > 79) - { - state->error = 66; /*text chunk too large*/ - break; - } - if(strlen(info.itext_keys[i]) < 1) - { - state->error = 67; /*text chunk too small*/ - break; - } - addChunk_iTXt(&outv, state->encoder.text_compression, - info.itext_keys[i], info.itext_langtags[i], info.itext_transkeys[i], info.itext_strings[i], - &state->encoder.zlibsettings); - } - - /*unknown chunks between IDAT and IEND*/ - if(info.unknown_chunks_data[2]) - { - state->error = addUnknownChunks(&outv, info.unknown_chunks_data[2], info.unknown_chunks_size[2]); - if(state->error) break; - } -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - addChunk_IEND(&outv); - - break; /*this isn't really a while loop; no error happened so break out now!*/ - } - - lodepng_info_cleanup(&info); - lodepng_free(data); - /*instead of cleaning the vector up, give it to the output*/ - *out = outv.data; - *outsize = outv.size; - - return state->error; -} - -unsigned lodepng_encode_memory(unsigned char** out, size_t* outsize, const unsigned char* image, - unsigned w, unsigned h, LodePNGColorType colortype, unsigned bitdepth) -{ - unsigned error; - LodePNGState state; - lodepng_state_init(&state); - state.info_raw.colortype = colortype; - state.info_raw.bitdepth = bitdepth; - state.info_png.color.colortype = colortype; - state.info_png.color.bitdepth = bitdepth; - lodepng_encode(out, outsize, image, w, h, &state); - error = state.error; - lodepng_state_cleanup(&state); - return error; -} - -unsigned lodepng_encode32(unsigned char** out, size_t* outsize, const unsigned char* image, unsigned w, unsigned h) -{ - return lodepng_encode_memory(out, outsize, image, w, h, LCT_RGBA, 8); -} - -unsigned lodepng_encode24(unsigned char** out, size_t* outsize, const unsigned char* image, unsigned w, unsigned h) -{ - return lodepng_encode_memory(out, outsize, image, w, h, LCT_RGB, 8); -} - -#ifdef LODEPNG_COMPILE_DISK -unsigned lodepng_encode_file(const char* filename, const unsigned char* image, unsigned w, unsigned h, - LodePNGColorType colortype, unsigned bitdepth) -{ - unsigned char* buffer; - size_t buffersize; - unsigned error = lodepng_encode_memory(&buffer, &buffersize, image, w, h, colortype, bitdepth); - if(!error) error = lodepng_save_file(buffer, buffersize, filename); - lodepng_free(buffer); - return error; -} - -unsigned lodepng_encode32_file(const char* filename, const unsigned char* image, unsigned w, unsigned h) -{ - return lodepng_encode_file(filename, image, w, h, LCT_RGBA, 8); -} - -unsigned lodepng_encode24_file(const char* filename, const unsigned char* image, unsigned w, unsigned h) -{ - return lodepng_encode_file(filename, image, w, h, LCT_RGB, 8); -} -#endif /*LODEPNG_COMPILE_DISK*/ - -void lodepng_encoder_settings_init(LodePNGEncoderSettings* settings) -{ - lodepng_compress_settings_init(&settings->zlibsettings); - settings->filter_palette_zero = 1; - settings->filter_strategy = LFS_MINSUM; - settings->auto_convert = 1; - settings->force_palette = 0; - settings->predefined_filters = 0; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - settings->add_id = 0; - settings->text_compression = 1; -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ -} - -#endif /*LODEPNG_COMPILE_ENCODER*/ -#endif /*LODEPNG_COMPILE_PNG*/ - -#ifdef LODEPNG_COMPILE_ERROR_TEXT -/* -This returns the description of a numerical error code in English. This is also -the documentation of all the error codes. -*/ -const char* lodepng_error_text(unsigned code) -{ - switch(code) - { - case 0: return "no error, everything went ok"; - case 1: return "nothing done yet"; /*the Encoder/Decoder has done nothing yet, error checking makes no sense yet*/ - case 10: return "end of input memory reached without huffman end code"; /*while huffman decoding*/ - case 11: return "error in code tree made it jump outside of huffman tree"; /*while huffman decoding*/ - case 13: return "problem while processing dynamic deflate block"; - case 14: return "problem while processing dynamic deflate block"; - case 15: return "problem while processing dynamic deflate block"; - case 16: return "unexisting code while processing dynamic deflate block"; - case 17: return "end of out buffer memory reached while inflating"; - case 18: return "invalid distance code while inflating"; - case 19: return "end of out buffer memory reached while inflating"; - case 20: return "invalid deflate block BTYPE encountered while decoding"; - case 21: return "NLEN is not ones complement of LEN in a deflate block"; - /*end of out buffer memory reached while inflating: - This can happen if the inflated deflate data is longer than the amount of bytes required to fill up - all the pixels of the image, given the color depth and image dimensions. Something that doesn't - happen in a normal, well encoded, PNG image.*/ - case 22: return "end of out buffer memory reached while inflating"; - case 23: return "end of in buffer memory reached while inflating"; - case 24: return "invalid FCHECK in zlib header"; - case 25: return "invalid compression method in zlib header"; - case 26: return "FDICT encountered in zlib header while it's not used for PNG"; - case 27: return "PNG file is smaller than a PNG header"; - /*Checks the magic file header, the first 8 bytes of the PNG file*/ - case 28: return "incorrect PNG signature, it's no PNG or corrupted"; - case 29: return "first chunk is not the header chunk"; - case 30: return "chunk length too large, chunk broken off at end of file"; - case 31: return "illegal PNG color type or bpp"; - case 32: return "illegal PNG compression method"; - case 33: return "illegal PNG filter method"; - case 34: return "illegal PNG interlace method"; - case 35: return "chunk length of a chunk is too large or the chunk too small"; - case 36: return "illegal PNG filter type encountered"; - case 37: return "illegal bit depth for this color type given"; - case 38: return "the palette is too big"; /*more than 256 colors*/ - case 39: return "more palette alpha values given in tRNS chunk than there are colors in the palette"; - case 40: return "tRNS chunk has wrong size for greyscale image"; - case 41: return "tRNS chunk has wrong size for RGB image"; - case 42: return "tRNS chunk appeared while it was not allowed for this color type"; - case 43: return "bKGD chunk has wrong size for palette image"; - case 44: return "bKGD chunk has wrong size for greyscale image"; - case 45: return "bKGD chunk has wrong size for RGB image"; - /*the input data is empty, maybe a PNG file doesn't exist or is in the wrong path*/ - case 48: return "empty input or file doesn't exist"; - case 49: return "jumped past memory while generating dynamic huffman tree"; - case 50: return "jumped past memory while generating dynamic huffman tree"; - case 51: return "jumped past memory while inflating huffman block"; - case 52: return "jumped past memory while inflating"; - case 53: return "size of zlib data too small"; - case 54: return "repeat symbol in tree while there was no value symbol yet"; - /*jumped past tree while generating huffman tree, this could be when the - tree will have more leaves than symbols after generating it out of the - given lenghts. They call this an oversubscribed dynamic bit lengths tree in zlib.*/ - case 55: return "jumped past tree while generating huffman tree"; - case 56: return "given output image colortype or bitdepth not supported for color conversion"; - case 57: return "invalid CRC encountered (checking CRC can be disabled)"; - case 58: return "invalid ADLER32 encountered (checking ADLER32 can be disabled)"; - case 59: return "requested color conversion not supported"; - case 60: return "invalid window size given in the settings of the encoder (must be 0-32768)"; - case 61: return "invalid BTYPE given in the settings of the encoder (only 0, 1 and 2 are allowed)"; - /*LodePNG leaves the choice of RGB to greyscale conversion formula to the user.*/ - case 62: return "conversion from color to greyscale not supported"; - case 63: return "length of a chunk too long, max allowed for PNG is 2147483647 bytes per chunk"; /*(2^31-1)*/ - /*this would result in the inability of a deflated block to ever contain an end code. It must be at least 1.*/ - case 64: return "the length of the END symbol 256 in the Huffman tree is 0"; - case 66: return "the length of a text chunk keyword given to the encoder is longer than the maximum of 79 bytes"; - case 67: return "the length of a text chunk keyword given to the encoder is smaller than the minimum of 1 byte"; - case 68: return "tried to encode a PLTE chunk with a palette that has less than 1 or more than 256 colors"; - case 69: return "unknown chunk type with 'critical' flag encountered by the decoder"; - case 71: return "unexisting interlace mode given to encoder (must be 0 or 1)"; - case 72: return "while decoding, unexisting compression method encountering in zTXt or iTXt chunk (it must be 0)"; - case 73: return "invalid tIME chunk size"; - case 74: return "invalid pHYs chunk size"; - /*length could be wrong, or data chopped off*/ - case 75: return "no null termination char found while decoding text chunk"; - case 76: return "iTXt chunk too short to contain required bytes"; - case 77: return "integer overflow in buffer size"; - case 78: return "failed to open file for reading"; /*file doesn't exist or couldn't be opened for reading*/ - case 79: return "failed to open file for writing"; - case 80: return "tried creating a tree of 0 symbols"; - case 81: return "lazy matching at pos 0 is impossible"; - case 82: return "color conversion to palette requested while a color isn't in palette"; - case 83: return "memory allocation failed"; - case 84: return "given image too small to contain all pixels to be encoded"; - case 86: return "impossible offset in lz77 encoding (internal bug)"; - case 87: return "must provide custom zlib function pointer if LODEPNG_COMPILE_ZLIB is not defined"; - case 88: return "invalid filter strategy given for LodePNGEncoderSettings.filter_strategy"; - case 89: return "text chunk keyword too short or long: must have size 1-79"; - /*the windowsize in the LodePNGCompressSettings. Requiring POT(==> & instead of %) makes encoding 12% faster.*/ - case 90: return "windowsize must be a power of two"; - } - return "unknown error code"; -} -#endif /*LODEPNG_COMPILE_ERROR_TEXT*/ - -/* ////////////////////////////////////////////////////////////////////////// */ -/* ////////////////////////////////////////////////////////////////////////// */ -/* // C++ Wrapper // */ -/* ////////////////////////////////////////////////////////////////////////// */ -/* ////////////////////////////////////////////////////////////////////////// */ - -#ifdef LODEPNG_COMPILE_CPP -namespace lodepng -{ - -#ifdef LODEPNG_COMPILE_DISK - void load_file(std::vector& buffer, const std::string& filename) - { - std::ifstream file(filename.c_str(), std::ios::in|std::ios::binary|std::ios::ate); - - /*get filesize*/ - std::streamsize size = 0; - if(file.seekg(0, std::ios::end).good()) size = file.tellg(); - if(file.seekg(0, std::ios::beg).good()) size -= file.tellg(); - - /*read contents of the file into the vector*/ - buffer.resize(size_t(size)); - if(size > 0) file.read((char*)(&buffer[0]), size); - } - -/*write given buffer to the file, overwriting the file, it doesn't append to it.*/ - void save_file(const std::vector& buffer, const std::string& filename) - { - std::ofstream file(filename.c_str(), std::ios::out|std::ios::binary); - file.write(buffer.empty() ? 0 : (char*)&buffer[0], std::streamsize(buffer.size())); - } -#endif //LODEPNG_COMPILE_DISK - -#ifdef LODEPNG_COMPILE_ZLIB -#ifdef LODEPNG_COMPILE_DECODER - unsigned decompress(std::vector& out, const unsigned char* in, size_t insize, - const LodePNGDecompressSettings& settings) - { - unsigned char* buffer = 0; - size_t buffersize = 0; - unsigned error = zlib_decompress(&buffer, &buffersize, in, insize, &settings); - if(buffer) - { - out.insert(out.end(), &buffer[0], &buffer[buffersize]); - lodepng_free(buffer); - } - return error; - } - - unsigned decompress(std::vector& out, const std::vector& in, - const LodePNGDecompressSettings& settings) - { - return decompress(out, in.empty() ? 0 : &in[0], in.size(), settings); - } -#endif //LODEPNG_COMPILE_DECODER - -#ifdef LODEPNG_COMPILE_ENCODER - unsigned compress(std::vector& out, const unsigned char* in, size_t insize, - const LodePNGCompressSettings& settings) - { - unsigned char* buffer = 0; - size_t buffersize = 0; - unsigned error = zlib_compress(&buffer, &buffersize, in, insize, &settings); - if(buffer) - { - out.insert(out.end(), &buffer[0], &buffer[buffersize]); - lodepng_free(buffer); - } - return error; - } - - unsigned compress(std::vector& out, const std::vector& in, - const LodePNGCompressSettings& settings) - { - return compress(out, in.empty() ? 0 : &in[0], in.size(), settings); - } -#endif //LODEPNG_COMPILE_ENCODER -#endif //LODEPNG_COMPILE_ZLIB - - -#ifdef LODEPNG_COMPILE_PNG - - State::State() - { - lodepng_state_init(this); - } - - State::State(const State& other) - { - lodepng_state_init(this); - lodepng_state_copy(this, &other); - } - - State::~State() - { - lodepng_state_cleanup(this); - } - - State& State::operator=(const State& other) - { - lodepng_state_copy(this, &other); - return *this; - } - -#ifdef LODEPNG_COMPILE_DECODER - - unsigned decode(std::vector& out, unsigned& w, unsigned& h, const unsigned char* in, - size_t insize, LodePNGColorType colortype, unsigned bitdepth) - { - unsigned char* buffer; - unsigned error = lodepng_decode_memory(&buffer, &w, &h, in, insize, colortype, bitdepth); - if(buffer && !error) - { - State state; - state.info_raw.colortype = colortype; - state.info_raw.bitdepth = bitdepth; - size_t buffersize = lodepng_get_raw_size(w, h, &state.info_raw); - out.insert(out.end(), &buffer[0], &buffer[buffersize]); - lodepng_free(buffer); - } - return error; - } - - unsigned decode(std::vector& out, unsigned& w, unsigned& h, - const std::vector& in, LodePNGColorType colortype, unsigned bitdepth) - { - return decode(out, w, h, in.empty() ? 0 : &in[0], (unsigned)in.size(), colortype, bitdepth); - } - - unsigned decode(std::vector& out, unsigned& w, unsigned& h, - State& state, - const unsigned char* in, size_t insize) - { - unsigned char* buffer = NULL; - unsigned error = lodepng_decode(&buffer, &w, &h, &state, in, insize); - if(buffer && !error) - { - size_t buffersize = lodepng_get_raw_size(w, h, &state.info_raw); - out.insert(out.end(), &buffer[0], &buffer[buffersize]); - } - lodepng_free(buffer); - return error; - } - - unsigned decode(std::vector& out, unsigned& w, unsigned& h, - State& state, - const std::vector& in) - { - return decode(out, w, h, state, in.empty() ? 0 : &in[0], in.size()); - } - -#ifdef LODEPNG_COMPILE_DISK - unsigned decode(std::vector& out, unsigned& w, unsigned& h, const std::string& filename, - LodePNGColorType colortype, unsigned bitdepth) - { - std::vector buffer; - load_file(buffer, filename); - return decode(out, w, h, buffer, colortype, bitdepth); - } -#endif //LODEPNG_COMPILE_DECODER -#endif //LODEPNG_COMPILE_DISK - -#ifdef LODEPNG_COMPILE_ENCODER - unsigned encode(std::vector& out, const unsigned char* in, unsigned w, unsigned h, - LodePNGColorType colortype, unsigned bitdepth) - { - unsigned char* buffer; - size_t buffersize; - unsigned error = lodepng_encode_memory(&buffer, &buffersize, in, w, h, colortype, bitdepth); - if(buffer) - { - out.insert(out.end(), &buffer[0], &buffer[buffersize]); - lodepng_free(buffer); - } - return error; - } - - unsigned encode(std::vector& out, - const std::vector& in, unsigned w, unsigned h, - LodePNGColorType colortype, unsigned bitdepth) - { - if(lodepng_get_raw_size_lct(w, h, colortype, bitdepth) > in.size()) return 84; - return encode(out, in.empty() ? 0 : &in[0], w, h, colortype, bitdepth); - } - - unsigned encode(std::vector& out, - const unsigned char* in, unsigned w, unsigned h, - State& state) - { - unsigned char* buffer; - size_t buffersize; - unsigned error = lodepng_encode(&buffer, &buffersize, in, w, h, &state); - if(buffer) - { - out.insert(out.end(), &buffer[0], &buffer[buffersize]); - lodepng_free(buffer); - } - return error; - } - - unsigned encode(std::vector& out, - const std::vector& in, unsigned w, unsigned h, - State& state) - { - if(lodepng_get_raw_size(w, h, &state.info_raw) > in.size()) return 84; - return encode(out, in.empty() ? 0 : &in[0], w, h, state); - } - -#ifdef LODEPNG_COMPILE_DISK - unsigned encode(const std::string& filename, - const unsigned char* in, unsigned w, unsigned h, - LodePNGColorType colortype, unsigned bitdepth) - { - std::vector buffer; - unsigned error = encode(buffer, in, w, h, colortype, bitdepth); - if(!error) save_file(buffer, filename); - return error; - } - - unsigned encode(const std::string& filename, - const std::vector& in, unsigned w, unsigned h, - LodePNGColorType colortype, unsigned bitdepth) - { - if(lodepng_get_raw_size_lct(w, h, colortype, bitdepth) > in.size()) return 84; - return encode(filename, in.empty() ? 0 : &in[0], w, h, colortype, bitdepth); - } -#endif //LODEPNG_COMPILE_DISK -#endif //LODEPNG_COMPILE_ENCODER -#endif //LODEPNG_COMPILE_PNG -} //namespace lodepng -#endif /*LODEPNG_COMPILE_CPP*/ \ No newline at end of file diff --git a/source/pc/lodepng.h b/source/pc/lodepng.h deleted file mode 100644 index bad1445..0000000 --- a/source/pc/lodepng.h +++ /dev/null @@ -1,1702 +0,0 @@ -/* -LodePNG version 20140823 - -Copyright (c) 2005-2014 Lode Vandevenne - -This software is provided 'as-is', without any express or implied -warranty. In no event will the authors be held liable for any damages -arising from the use of this software. - -Permission is granted to anyone to use this software for any purpose, -including commercial applications, and to alter it and redistribute it -freely, subject to the following restrictions: - - 1. The origin of this software must not be misrepresented; you must not - claim that you wrote the original software. If you use this software - in a product, an acknowledgment in the product documentation would be - appreciated but is not required. - - 2. Altered source versions must be plainly marked as such, and must not be - misrepresented as being the original software. - - 3. This notice may not be removed or altered from any source - distribution. -*/ - -#ifndef LODEPNG_H -#define LODEPNG_H - -#include /*for size_t*/ - -#ifdef __cplusplus -#include -#include -#endif /*__cplusplus*/ - -/* -The following #defines are used to create code sections. They can be disabled -to disable code sections, which can give faster compile time and smaller binary. -The "NO_COMPILE" defines are designed to be used to pass as defines to the -compiler command to disable them without modifying this header, e.g. --DLODEPNG_NO_COMPILE_ZLIB for gcc. -*/ -/*deflate & zlib. If disabled, you must specify alternative zlib functions in -the custom_zlib field of the compress and decompress settings*/ -#ifndef LODEPNG_NO_COMPILE_ZLIB -#define LODEPNG_COMPILE_ZLIB -#endif -/*png encoder and png decoder*/ -#ifndef LODEPNG_NO_COMPILE_PNG -#define LODEPNG_COMPILE_PNG -#endif -/*deflate&zlib decoder and png decoder*/ -#ifndef LODEPNG_NO_COMPILE_DECODER -#define LODEPNG_COMPILE_DECODER -#endif -/*deflate&zlib encoder and png encoder*/ -#ifndef LODEPNG_NO_COMPILE_ENCODER -#define LODEPNG_COMPILE_ENCODER -#endif -/*the optional built in harddisk file loading and saving functions*/ -#ifndef LODEPNG_NO_COMPILE_DISK -#define LODEPNG_COMPILE_DISK -#endif -/*support for chunks other than IHDR, IDAT, PLTE, tRNS, IEND: ancillary and unknown chunks*/ -#ifndef LODEPNG_NO_COMPILE_ANCILLARY_CHUNKS -#define LODEPNG_COMPILE_ANCILLARY_CHUNKS -#endif -/*ability to convert error numerical codes to English text string*/ -#ifndef LODEPNG_NO_COMPILE_ERROR_TEXT -#define LODEPNG_COMPILE_ERROR_TEXT -#endif -/*Compile the default allocators (C's free, malloc and realloc). If you disable this, -you can define the functions lodepng_free, lodepng_malloc and lodepng_realloc in your -source files with custom allocators.*/ -#ifndef LODEPNG_NO_COMPILE_ALLOCATORS -#define LODEPNG_COMPILE_ALLOCATORS -#endif -/*compile the C++ version (you can disable the C++ wrapper here even when compiling for C++)*/ -#ifdef __cplusplus -#ifndef LODEPNG_NO_COMPILE_CPP -#define LODEPNG_COMPILE_CPP -#endif -#endif - -#ifdef LODEPNG_COMPILE_PNG -/*The PNG color types (also used for raw).*/ -typedef enum LodePNGColorType -{ - LCT_GREY = 0, /*greyscale: 1,2,4,8,16 bit*/ - LCT_RGB = 2, /*RGB: 8,16 bit*/ - LCT_PALETTE = 3, /*palette: 1,2,4,8 bit*/ - LCT_GREY_ALPHA = 4, /*greyscale with alpha: 8,16 bit*/ - LCT_RGBA = 6 /*RGB with alpha: 8,16 bit*/ -} LodePNGColorType; - -#ifdef LODEPNG_COMPILE_DECODER -/* -Converts PNG data in memory to raw pixel data. -out: Output parameter. Pointer to buffer that will contain the raw pixel data. - After decoding, its size is w * h * (bytes per pixel) bytes larger than - initially. Bytes per pixel depends on colortype and bitdepth. - Must be freed after usage with free(*out). - Note: for 16-bit per channel colors, uses big endian format like PNG does. -w: Output parameter. Pointer to width of pixel data. -h: Output parameter. Pointer to height of pixel data. -in: Memory buffer with the PNG file. -insize: size of the in buffer. -colortype: the desired color type for the raw output image. See explanation on PNG color types. -bitdepth: the desired bit depth for the raw output image. See explanation on PNG color types. -Return value: LodePNG error code (0 means no error). -*/ -unsigned lodepng_decode_memory(unsigned char** out, unsigned* w, unsigned* h, - const unsigned char* in, size_t insize, - LodePNGColorType colortype, unsigned bitdepth); - -/*Same as lodepng_decode_memory, but always decodes to 32-bit RGBA raw image*/ -unsigned lodepng_decode32(unsigned char** out, unsigned* w, unsigned* h, - const unsigned char* in, size_t insize); - -/*Same as lodepng_decode_memory, but always decodes to 24-bit RGB raw image*/ -unsigned lodepng_decode24(unsigned char** out, unsigned* w, unsigned* h, - const unsigned char* in, size_t insize); - -#ifdef LODEPNG_COMPILE_DISK -/* -Load PNG from disk, from file with given name. -Same as the other decode functions, but instead takes a filename as input. -*/ -unsigned lodepng_decode_file(unsigned char** out, unsigned* w, unsigned* h, - const char* filename, - LodePNGColorType colortype, unsigned bitdepth); - -/*Same as lodepng_decode_file, but always decodes to 32-bit RGBA raw image.*/ -unsigned lodepng_decode32_file(unsigned char** out, unsigned* w, unsigned* h, - const char* filename); - -/*Same as lodepng_decode_file, but always decodes to 24-bit RGB raw image.*/ -unsigned lodepng_decode24_file(unsigned char** out, unsigned* w, unsigned* h, - const char* filename); -#endif /*LODEPNG_COMPILE_DISK*/ -#endif /*LODEPNG_COMPILE_DECODER*/ - - -#ifdef LODEPNG_COMPILE_ENCODER -/* -Converts raw pixel data into a PNG image in memory. The colortype and bitdepth - of the output PNG image cannot be chosen, they are automatically determined - by the colortype, bitdepth and content of the input pixel data. - Note: for 16-bit per channel colors, needs big endian format like PNG does. -out: Output parameter. Pointer to buffer that will contain the PNG image data. - Must be freed after usage with free(*out). -outsize: Output parameter. Pointer to the size in bytes of the out buffer. -image: The raw pixel data to encode. The size of this buffer should be - w * h * (bytes per pixel), bytes per pixel depends on colortype and bitdepth. -w: width of the raw pixel data in pixels. -h: height of the raw pixel data in pixels. -colortype: the color type of the raw input image. See explanation on PNG color types. -bitdepth: the bit depth of the raw input image. See explanation on PNG color types. -Return value: LodePNG error code (0 means no error). -*/ -unsigned lodepng_encode_memory(unsigned char** out, size_t* outsize, - const unsigned char* image, unsigned w, unsigned h, - LodePNGColorType colortype, unsigned bitdepth); - -/*Same as lodepng_encode_memory, but always encodes from 32-bit RGBA raw image.*/ -unsigned lodepng_encode32(unsigned char** out, size_t* outsize, - const unsigned char* image, unsigned w, unsigned h); - -/*Same as lodepng_encode_memory, but always encodes from 24-bit RGB raw image.*/ -unsigned lodepng_encode24(unsigned char** out, size_t* outsize, - const unsigned char* image, unsigned w, unsigned h); - -#ifdef LODEPNG_COMPILE_DISK -/* -Converts raw pixel data into a PNG file on disk. -Same as the other encode functions, but instead takes a filename as output. -NOTE: This overwrites existing files without warning! -*/ -unsigned lodepng_encode_file(const char* filename, - const unsigned char* image, unsigned w, unsigned h, - LodePNGColorType colortype, unsigned bitdepth); - -/*Same as lodepng_encode_file, but always encodes from 32-bit RGBA raw image.*/ -unsigned lodepng_encode32_file(const char* filename, - const unsigned char* image, unsigned w, unsigned h); - -/*Same as lodepng_encode_file, but always encodes from 24-bit RGB raw image.*/ -unsigned lodepng_encode24_file(const char* filename, - const unsigned char* image, unsigned w, unsigned h); -#endif /*LODEPNG_COMPILE_DISK*/ -#endif /*LODEPNG_COMPILE_ENCODER*/ - - -#ifdef LODEPNG_COMPILE_CPP -namespace lodepng -{ -#ifdef LODEPNG_COMPILE_DECODER -/*Same as lodepng_decode_memory, but decodes to an std::vector. The colortype -is the format to output the pixels to. Default is RGBA 8-bit per channel.*/ - unsigned decode(std::vector& out, unsigned& w, unsigned& h, - const unsigned char* in, size_t insize, - LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); - unsigned decode(std::vector& out, unsigned& w, unsigned& h, - const std::vector& in, - LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); -#ifdef LODEPNG_COMPILE_DISK -/* -Converts PNG file from disk to raw pixel data in memory. -Same as the other decode functions, but instead takes a filename as input. -*/ - unsigned decode(std::vector& out, unsigned& w, unsigned& h, - const std::string& filename, - LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); -#endif //LODEPNG_COMPILE_DISK -#endif //LODEPNG_COMPILE_DECODER - -#ifdef LODEPNG_COMPILE_ENCODER -/*Same as lodepng_encode_memory, but encodes to an std::vector. colortype -is that of the raw input data. The output PNG color type will be auto chosen.*/ - unsigned encode(std::vector& out, - const unsigned char* in, unsigned w, unsigned h, - LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); - unsigned encode(std::vector& out, - const std::vector& in, unsigned w, unsigned h, - LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); -#ifdef LODEPNG_COMPILE_DISK -/* -Converts 32-bit RGBA raw pixel data into a PNG file on disk. -Same as the other encode functions, but instead takes a filename as output. -NOTE: This overwrites existing files without warning! -*/ - unsigned encode(const std::string& filename, - const unsigned char* in, unsigned w, unsigned h, - LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); - unsigned encode(const std::string& filename, - const std::vector& in, unsigned w, unsigned h, - LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); -#endif //LODEPNG_COMPILE_DISK -#endif //LODEPNG_COMPILE_ENCODER -} //namespace lodepng -#endif /*LODEPNG_COMPILE_CPP*/ -#endif /*LODEPNG_COMPILE_PNG*/ - -#ifdef LODEPNG_COMPILE_ERROR_TEXT -/*Returns an English description of the numerical error code.*/ -const char* lodepng_error_text(unsigned code); -#endif /*LODEPNG_COMPILE_ERROR_TEXT*/ - -#ifdef LODEPNG_COMPILE_DECODER -/*Settings for zlib decompression*/ -typedef struct LodePNGDecompressSettings LodePNGDecompressSettings; -struct LodePNGDecompressSettings -{ - unsigned ignore_adler32; /*if 1, continue and don't give an error message if the Adler32 checksum is corrupted*/ - - /*use custom zlib decoder instead of built in one (default: null)*/ - unsigned (*custom_zlib)(unsigned char**, size_t*, - const unsigned char*, size_t, - const LodePNGDecompressSettings*); - /*use custom deflate decoder instead of built in one (default: null) - if custom_zlib is used, custom_deflate is ignored since only the built in - zlib function will call custom_deflate*/ - unsigned (*custom_inflate)(unsigned char**, size_t*, - const unsigned char*, size_t, - const LodePNGDecompressSettings*); - - const void* custom_context; /*optional custom settings for custom functions*/ -}; - -extern const LodePNGDecompressSettings lodepng_default_decompress_settings; -void lodepng_decompress_settings_init(LodePNGDecompressSettings* settings); -#endif /*LODEPNG_COMPILE_DECODER*/ - -#ifdef LODEPNG_COMPILE_ENCODER -/* -Settings for zlib compression. Tweaking these settings tweaks the balance -between speed and compression ratio. -*/ -typedef struct LodePNGCompressSettings LodePNGCompressSettings; -struct LodePNGCompressSettings /*deflate = compress*/ -{ - /*LZ77 related settings*/ - unsigned btype; /*the block type for LZ (0, 1, 2 or 3, see zlib standard). Should be 2 for proper compression.*/ - unsigned use_lz77; /*whether or not to use LZ77. Should be 1 for proper compression.*/ - unsigned windowsize; /*must be a power of two <= 32768. higher compresses more but is slower. Default value: 2048.*/ - unsigned minmatch; /*mininum lz77 length. 3 is normally best, 6 can be better for some PNGs. Default: 0*/ - unsigned nicematch; /*stop searching if >= this length found. Set to 258 for best compression. Default: 128*/ - unsigned lazymatching; /*use lazy matching: better compression but a bit slower. Default: true*/ - - /*use custom zlib encoder instead of built in one (default: null)*/ - unsigned (*custom_zlib)(unsigned char**, size_t*, - const unsigned char*, size_t, - const LodePNGCompressSettings*); - /*use custom deflate encoder instead of built in one (default: null) - if custom_zlib is used, custom_deflate is ignored since only the built in - zlib function will call custom_deflate*/ - unsigned (*custom_deflate)(unsigned char**, size_t*, - const unsigned char*, size_t, - const LodePNGCompressSettings*); - - const void* custom_context; /*optional custom settings for custom functions*/ -}; - -extern const LodePNGCompressSettings lodepng_default_compress_settings; -void lodepng_compress_settings_init(LodePNGCompressSettings* settings); -#endif /*LODEPNG_COMPILE_ENCODER*/ - -#ifdef LODEPNG_COMPILE_PNG -/* -Color mode of an image. Contains all information required to decode the pixel -bits to RGBA colors. This information is the same as used in the PNG file -format, and is used both for PNG and raw image data in LodePNG. -*/ -typedef struct LodePNGColorMode -{ - /*header (IHDR)*/ - LodePNGColorType colortype; /*color type, see PNG standard or documentation further in this header file*/ - unsigned bitdepth; /*bits per sample, see PNG standard or documentation further in this header file*/ - - /* - palette (PLTE and tRNS) - - Dynamically allocated with the colors of the palette, including alpha. - When encoding a PNG, to store your colors in the palette of the LodePNGColorMode, first use - lodepng_palette_clear, then for each color use lodepng_palette_add. - If you encode an image without alpha with palette, don't forget to put value 255 in each A byte of the palette. - - When decoding, by default you can ignore this palette, since LodePNG already - fills the palette colors in the pixels of the raw RGBA output. - - The palette is only supported for color type 3. - */ - unsigned char* palette; /*palette in RGBARGBA... order. When allocated, must be either 0, or have size 1024*/ - size_t palettesize; /*palette size in number of colors (amount of bytes is 4 * palettesize)*/ - - /* - transparent color key (tRNS) - - This color uses the same bit depth as the bitdepth value in this struct, which can be 1-bit to 16-bit. - For greyscale PNGs, r, g and b will all 3 be set to the same. - - When decoding, by default you can ignore this information, since LodePNG sets - pixels with this key to transparent already in the raw RGBA output. - - The color key is only supported for color types 0 and 2. - */ - unsigned key_defined; /*is a transparent color key given? 0 = false, 1 = true*/ - unsigned key_r; /*red/greyscale component of color key*/ - unsigned key_g; /*green component of color key*/ - unsigned key_b; /*blue component of color key*/ -} LodePNGColorMode; - -/*init, cleanup and copy functions to use with this struct*/ -void lodepng_color_mode_init(LodePNGColorMode* info); -void lodepng_color_mode_cleanup(LodePNGColorMode* info); -/*return value is error code (0 means no error)*/ -unsigned lodepng_color_mode_copy(LodePNGColorMode* dest, const LodePNGColorMode* source); - -void lodepng_palette_clear(LodePNGColorMode* info); -/*add 1 color to the palette*/ -unsigned lodepng_palette_add(LodePNGColorMode* info, - unsigned char r, unsigned char g, unsigned char b, unsigned char a); - -/*get the total amount of bits per pixel, based on colortype and bitdepth in the struct*/ -unsigned lodepng_get_bpp(const LodePNGColorMode* info); -/*get the amount of color channels used, based on colortype in the struct. -If a palette is used, it counts as 1 channel.*/ -unsigned lodepng_get_channels(const LodePNGColorMode* info); -/*is it a greyscale type? (only colortype 0 or 4)*/ -unsigned lodepng_is_greyscale_type(const LodePNGColorMode* info); -/*has it got an alpha channel? (only colortype 2 or 6)*/ -unsigned lodepng_is_alpha_type(const LodePNGColorMode* info); -/*has it got a palette? (only colortype 3)*/ -unsigned lodepng_is_palette_type(const LodePNGColorMode* info); -/*only returns true if there is a palette and there is a value in the palette with alpha < 255. -Loops through the palette to check this.*/ -unsigned lodepng_has_palette_alpha(const LodePNGColorMode* info); -/* -Check if the given color info indicates the possibility of having non-opaque pixels in the PNG image. -Returns true if the image can have translucent or invisible pixels (it still be opaque if it doesn't use such pixels). -Returns false if the image can only have opaque pixels. -In detail, it returns true only if it's a color type with alpha, or has a palette with non-opaque values, -or if "key_defined" is true. -*/ -unsigned lodepng_can_have_alpha(const LodePNGColorMode* info); -/*Returns the byte size of a raw image buffer with given width, height and color mode*/ -size_t lodepng_get_raw_size(unsigned w, unsigned h, const LodePNGColorMode* color); - -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS -/*The information of a Time chunk in PNG.*/ -typedef struct LodePNGTime -{ - unsigned year; /*2 bytes used (0-65535)*/ - unsigned month; /*1-12*/ - unsigned day; /*1-31*/ - unsigned hour; /*0-23*/ - unsigned minute; /*0-59*/ - unsigned second; /*0-60 (to allow for leap seconds)*/ -} LodePNGTime; -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - -/*Information about the PNG image, except pixels, width and height.*/ -typedef struct LodePNGInfo -{ - /*header (IHDR), palette (PLTE) and transparency (tRNS) chunks*/ - unsigned compression_method;/*compression method of the original file. Always 0.*/ - unsigned filter_method; /*filter method of the original file*/ - unsigned interlace_method; /*interlace method of the original file*/ - LodePNGColorMode color; /*color type and bits, palette and transparency of the PNG file*/ - -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - /* - suggested background color chunk (bKGD) - This color uses the same color mode as the PNG (except alpha channel), which can be 1-bit to 16-bit. - - For greyscale PNGs, r, g and b will all 3 be set to the same. When encoding - the encoder writes the red one. For palette PNGs: When decoding, the RGB value - will be stored, not a palette index. But when encoding, specify the index of - the palette in background_r, the other two are then ignored. - - The decoder does not use this background color to edit the color of pixels. - */ - unsigned background_defined; /*is a suggested background color given?*/ - unsigned background_r; /*red component of suggested background color*/ - unsigned background_g; /*green component of suggested background color*/ - unsigned background_b; /*blue component of suggested background color*/ - - /* - non-international text chunks (tEXt and zTXt) - - The char** arrays each contain num strings. The actual messages are in - text_strings, while text_keys are keywords that give a short description what - the actual text represents, e.g. Title, Author, Description, or anything else. - - A keyword is minimum 1 character and maximum 79 characters long. It's - discouraged to use a single line length longer than 79 characters for texts. - - Don't allocate these text buffers yourself. Use the init/cleanup functions - correctly and use lodepng_add_text and lodepng_clear_text. - */ - size_t text_num; /*the amount of texts in these char** buffers (there may be more texts in itext)*/ - char** text_keys; /*the keyword of a text chunk (e.g. "Comment")*/ - char** text_strings; /*the actual text*/ - - /* - international text chunks (iTXt) - Similar to the non-international text chunks, but with additional strings - "langtags" and "transkeys". - */ - size_t itext_num; /*the amount of international texts in this PNG*/ - char** itext_keys; /*the English keyword of the text chunk (e.g. "Comment")*/ - char** itext_langtags; /*language tag for this text's language, ISO/IEC 646 string, e.g. ISO 639 language tag*/ - char** itext_transkeys; /*keyword translated to the international language - UTF-8 string*/ - char** itext_strings; /*the actual international text - UTF-8 string*/ - - /*time chunk (tIME)*/ - unsigned time_defined; /*set to 1 to make the encoder generate a tIME chunk*/ - LodePNGTime time; - - /*phys chunk (pHYs)*/ - unsigned phys_defined; /*if 0, there is no pHYs chunk and the values below are undefined, if 1 else there is one*/ - unsigned phys_x; /*pixels per unit in x direction*/ - unsigned phys_y; /*pixels per unit in y direction*/ - unsigned phys_unit; /*may be 0 (unknown unit) or 1 (metre)*/ - - /* - unknown chunks - There are 3 buffers, one for each position in the PNG where unknown chunks can appear - each buffer contains all unknown chunks for that position consecutively - The 3 buffers are the unknown chunks between certain critical chunks: - 0: IHDR-PLTE, 1: PLTE-IDAT, 2: IDAT-IEND - Do not allocate or traverse this data yourself. Use the chunk traversing functions declared - later, such as lodepng_chunk_next and lodepng_chunk_append, to read/write this struct. - */ - unsigned char* unknown_chunks_data[3]; - size_t unknown_chunks_size[3]; /*size in bytes of the unknown chunks, given for protection*/ -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ -} LodePNGInfo; - -/*init, cleanup and copy functions to use with this struct*/ -void lodepng_info_init(LodePNGInfo* info); -void lodepng_info_cleanup(LodePNGInfo* info); -/*return value is error code (0 means no error)*/ -unsigned lodepng_info_copy(LodePNGInfo* dest, const LodePNGInfo* source); - -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS -void lodepng_clear_text(LodePNGInfo* info); /*use this to clear the texts again after you filled them in*/ -unsigned lodepng_add_text(LodePNGInfo* info, const char* key, const char* str); /*push back both texts at once*/ - -void lodepng_clear_itext(LodePNGInfo* info); /*use this to clear the itexts again after you filled them in*/ -unsigned lodepng_add_itext(LodePNGInfo* info, const char* key, const char* langtag, - const char* transkey, const char* str); /*push back the 4 texts of 1 chunk at once*/ -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - -/* -Converts raw buffer from one color type to another color type, based on -LodePNGColorMode structs to describe the input and output color type. -See the reference manual at the end of this header file to see which color conversions are supported. -return value = LodePNG error code (0 if all went ok, an error if the conversion isn't supported) -The out buffer must have size (w * h * bpp + 7) / 8, where bpp is the bits per pixel -of the output color type (lodepng_get_bpp). -For < 8 bpp images, there should not be padding bits at the end of scanlines. -For 16-bit per channel colors, uses big endian format like PNG does. -Return value is LodePNG error code -*/ -unsigned lodepng_convert(unsigned char* out, const unsigned char* in, - LodePNGColorMode* mode_out, const LodePNGColorMode* mode_in, - unsigned w, unsigned h); - -#ifdef LODEPNG_COMPILE_DECODER -/* -Settings for the decoder. This contains settings for the PNG and the Zlib -decoder, but not the Info settings from the Info structs. -*/ -typedef struct LodePNGDecoderSettings -{ - LodePNGDecompressSettings zlibsettings; /*in here is the setting to ignore Adler32 checksums*/ - - unsigned ignore_crc; /*ignore CRC checksums*/ - - unsigned color_convert; /*whether to convert the PNG to the color type you want. Default: yes*/ - -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - unsigned read_text_chunks; /*if false but remember_unknown_chunks is true, they're stored in the unknown chunks*/ - /*store all bytes from unknown chunks in the LodePNGInfo (off by default, useful for a png editor)*/ - unsigned remember_unknown_chunks; -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ -} LodePNGDecoderSettings; - -void lodepng_decoder_settings_init(LodePNGDecoderSettings* settings); -#endif /*LODEPNG_COMPILE_DECODER*/ - -#ifdef LODEPNG_COMPILE_ENCODER -/*automatically use color type with less bits per pixel if losslessly possible. Default: AUTO*/ -typedef enum LodePNGFilterStrategy -{ - /*every filter at zero*/ - LFS_ZERO, - /*Use filter that gives minumum sum, as described in the official PNG filter heuristic.*/ - LFS_MINSUM, - /*Use the filter type that gives smallest Shannon entropy for this scanline. Depending - on the image, this is better or worse than minsum.*/ - LFS_ENTROPY, - /* - Brute-force-search PNG filters by compressing each filter for each scanline. - Experimental, very slow, and only rarely gives better compression than MINSUM. - */ - LFS_BRUTE_FORCE, - /*use predefined_filters buffer: you specify the filter type for each scanline*/ - LFS_PREDEFINED -} LodePNGFilterStrategy; - -/*Gives characteristics about the colors of the image, which helps decide which color model to use for encoding. -Used internally by default if "auto_convert" is enabled. Public because it's useful for custom algorithms.*/ -typedef struct LodePNGColorProfile -{ - unsigned colored; /*not greyscale*/ - unsigned key; /*if true, image is not opaque. Only if true and alpha is false, color key is possible.*/ - unsigned short key_r; /*these values are always in 16-bit bitdepth in the profile*/ - unsigned short key_g; - unsigned short key_b; - unsigned alpha; /*alpha channel or alpha palette required*/ - unsigned numcolors; /*amount of colors, up to 257. Not valid if bits == 16.*/ - unsigned char palette[1024]; /*Remembers up to the first 256 RGBA colors, in no particular order*/ - unsigned bits; /*bits per channel (not for palette). 1,2 or 4 for greyscale only. 16 if 16-bit per channel required.*/ -} LodePNGColorProfile; - -void lodepng_color_profile_init(LodePNGColorProfile* profile); - -/*Get a LodePNGColorProfile of the image.*/ -unsigned get_color_profile(LodePNGColorProfile* profile, - const unsigned char* image, unsigned w, unsigned h, - const LodePNGColorMode* mode_in); -/*The function LodePNG uses internally to decide the PNG color with auto_convert. -Chooses an optimal color model, e.g. grey if only grey pixels, palette if < 256 colors, ...*/ -unsigned lodepng_auto_choose_color(LodePNGColorMode* mode_out, - const unsigned char* image, unsigned w, unsigned h, - const LodePNGColorMode* mode_in); - -/*Settings for the encoder.*/ -typedef struct LodePNGEncoderSettings -{ - LodePNGCompressSettings zlibsettings; /*settings for the zlib encoder, such as window size, ...*/ - - unsigned auto_convert; /*automatically choose output PNG color type. Default: true*/ - - /*If true, follows the official PNG heuristic: if the PNG uses a palette or lower than - 8 bit depth, set all filters to zero. Otherwise use the filter_strategy. Note that to - completely follow the official PNG heuristic, filter_palette_zero must be true and - filter_strategy must be LFS_MINSUM*/ - unsigned filter_palette_zero; - /*Which filter strategy to use when not using zeroes due to filter_palette_zero. - Set filter_palette_zero to 0 to ensure always using your chosen strategy. Default: LFS_MINSUM*/ - LodePNGFilterStrategy filter_strategy; - /*used if filter_strategy is LFS_PREDEFINED. In that case, this must point to a buffer with - the same length as the amount of scanlines in the image, and each value must <= 5. You - have to cleanup this buffer, LodePNG will never free it. Don't forget that filter_palette_zero - must be set to 0 to ensure this is also used on palette or low bitdepth images.*/ - const unsigned char* predefined_filters; - - /*force creating a PLTE chunk if colortype is 2 or 6 (= a suggested palette). - If colortype is 3, PLTE is _always_ created.*/ - unsigned force_palette; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - /*add LodePNG identifier and version as a text chunk, for debugging*/ - unsigned add_id; - /*encode text chunks as zTXt chunks instead of tEXt chunks, and use compression in iTXt chunks*/ - unsigned text_compression; -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ -} LodePNGEncoderSettings; - -void lodepng_encoder_settings_init(LodePNGEncoderSettings* settings); -#endif /*LODEPNG_COMPILE_ENCODER*/ - - -#if defined(LODEPNG_COMPILE_DECODER) || defined(LODEPNG_COMPILE_ENCODER) -/*The settings, state and information for extended encoding and decoding.*/ -typedef struct LodePNGState -{ -#ifdef LODEPNG_COMPILE_DECODER - LodePNGDecoderSettings decoder; /*the decoding settings*/ -#endif /*LODEPNG_COMPILE_DECODER*/ -#ifdef LODEPNG_COMPILE_ENCODER - LodePNGEncoderSettings encoder; /*the encoding settings*/ -#endif /*LODEPNG_COMPILE_ENCODER*/ - LodePNGColorMode info_raw; /*specifies the format in which you would like to get the raw pixel buffer*/ - LodePNGInfo info_png; /*info of the PNG image obtained after decoding*/ - unsigned error; -#ifdef LODEPNG_COMPILE_CPP - //For the lodepng::State subclass. - virtual ~LodePNGState(){} -#endif -} LodePNGState; - -/*init, cleanup and copy functions to use with this struct*/ -void lodepng_state_init(LodePNGState* state); -void lodepng_state_cleanup(LodePNGState* state); -void lodepng_state_copy(LodePNGState* dest, const LodePNGState* source); -#endif /* defined(LODEPNG_COMPILE_DECODER) || defined(LODEPNG_COMPILE_ENCODER) */ - -#ifdef LODEPNG_COMPILE_DECODER -/* -Same as lodepng_decode_memory, but uses a LodePNGState to allow custom settings and -getting much more information about the PNG image and color mode. -*/ -unsigned lodepng_decode(unsigned char** out, unsigned* w, unsigned* h, - LodePNGState* state, - const unsigned char* in, size_t insize); - -/* -Read the PNG header, but not the actual data. This returns only the information -that is in the header chunk of the PNG, such as width, height and color type. The -information is placed in the info_png field of the LodePNGState. -*/ -unsigned lodepng_inspect(unsigned* w, unsigned* h, - LodePNGState* state, - const unsigned char* in, size_t insize); -#endif /*LODEPNG_COMPILE_DECODER*/ - - -#ifdef LODEPNG_COMPILE_ENCODER -/*This function allocates the out buffer with standard malloc and stores the size in *outsize.*/ -unsigned lodepng_encode(unsigned char** out, size_t* outsize, - const unsigned char* image, unsigned w, unsigned h, - LodePNGState* state); -#endif /*LODEPNG_COMPILE_ENCODER*/ - -/* -The lodepng_chunk functions are normally not needed, except to traverse the -unknown chunks stored in the LodePNGInfo struct, or add new ones to it. -It also allows traversing the chunks of an encoded PNG file yourself. - -PNG standard chunk naming conventions: -First byte: uppercase = critical, lowercase = ancillary -Second byte: uppercase = public, lowercase = private -Third byte: must be uppercase -Fourth byte: uppercase = unsafe to copy, lowercase = safe to copy -*/ - -/*get the length of the data of the chunk. Total chunk length has 12 bytes more.*/ -unsigned lodepng_chunk_length(const unsigned char* chunk); - -/*puts the 4-byte type in null terminated string*/ -void lodepng_chunk_type(char type[5], const unsigned char* chunk); - -/*check if the type is the given type*/ -unsigned char lodepng_chunk_type_equals(const unsigned char* chunk, const char* type); - -/*0: it's one of the critical chunk types, 1: it's an ancillary chunk (see PNG standard)*/ -unsigned char lodepng_chunk_ancillary(const unsigned char* chunk); - -/*0: public, 1: private (see PNG standard)*/ -unsigned char lodepng_chunk_private(const unsigned char* chunk); - -/*0: the chunk is unsafe to copy, 1: the chunk is safe to copy (see PNG standard)*/ -unsigned char lodepng_chunk_safetocopy(const unsigned char* chunk); - -/*get pointer to the data of the chunk, where the input points to the header of the chunk*/ -unsigned char* lodepng_chunk_data(unsigned char* chunk); -const unsigned char* lodepng_chunk_data_const(const unsigned char* chunk); - -/*returns 0 if the crc is correct, 1 if it's incorrect (0 for OK as usual!)*/ -unsigned lodepng_chunk_check_crc(const unsigned char* chunk); - -/*generates the correct CRC from the data and puts it in the last 4 bytes of the chunk*/ -void lodepng_chunk_generate_crc(unsigned char* chunk); - -/*iterate to next chunks. don't use on IEND chunk, as there is no next chunk then*/ -unsigned char* lodepng_chunk_next(unsigned char* chunk); -const unsigned char* lodepng_chunk_next_const(const unsigned char* chunk); - -/* -Appends chunk to the data in out. The given chunk should already have its chunk header. -The out variable and outlength are updated to reflect the new reallocated buffer. -Returns error code (0 if it went ok) -*/ -unsigned lodepng_chunk_append(unsigned char** out, size_t* outlength, const unsigned char* chunk); - -/* -Appends new chunk to out. The chunk to append is given by giving its length, type -and data separately. The type is a 4-letter string. -The out variable and outlength are updated to reflect the new reallocated buffer. -Returne error code (0 if it went ok) -*/ -unsigned lodepng_chunk_create(unsigned char** out, size_t* outlength, unsigned length, - const char* type, const unsigned char* data); - - -/*Calculate CRC32 of buffer*/ -unsigned lodepng_crc32(const unsigned char* buf, size_t len); -#endif /*LODEPNG_COMPILE_PNG*/ - - -#ifdef LODEPNG_COMPILE_ZLIB -/* -This zlib part can be used independently to zlib compress and decompress a -buffer. It cannot be used to create gzip files however, and it only supports the -part of zlib that is required for PNG, it does not support dictionaries. -*/ - -#ifdef LODEPNG_COMPILE_DECODER -/*Inflate a buffer. Inflate is the decompression step of deflate. Out buffer must be freed after use.*/ -unsigned lodepng_inflate(unsigned char** out, size_t* outsize, - const unsigned char* in, size_t insize, - const LodePNGDecompressSettings* settings); - -/* -Decompresses Zlib data. Reallocates the out buffer and appends the data. The -data must be according to the zlib specification. -Either, *out must be NULL and *outsize must be 0, or, *out must be a valid -buffer and *outsize its size in bytes. out must be freed by user after usage. -*/ -unsigned lodepng_zlib_decompress(unsigned char** out, size_t* outsize, - const unsigned char* in, size_t insize, - const LodePNGDecompressSettings* settings); -#endif /*LODEPNG_COMPILE_DECODER*/ - -#ifdef LODEPNG_COMPILE_ENCODER -/* -Compresses data with Zlib. Reallocates the out buffer and appends the data. -Zlib adds a small header and trailer around the deflate data. -The data is output in the format of the zlib specification. -Either, *out must be NULL and *outsize must be 0, or, *out must be a valid -buffer and *outsize its size in bytes. out must be freed by user after usage. -*/ -unsigned lodepng_zlib_compress(unsigned char** out, size_t* outsize, - const unsigned char* in, size_t insize, - const LodePNGCompressSettings* settings); - -/* -Find length-limited Huffman code for given frequencies. This function is in the -public interface only for tests, it's used internally by lodepng_deflate. -*/ -unsigned lodepng_huffman_code_lengths(unsigned* lengths, const unsigned* frequencies, - size_t numcodes, unsigned maxbitlen); - -/*Compress a buffer with deflate. See RFC 1951. Out buffer must be freed after use.*/ -unsigned lodepng_deflate(unsigned char** out, size_t* outsize, - const unsigned char* in, size_t insize, - const LodePNGCompressSettings* settings); - -#endif /*LODEPNG_COMPILE_ENCODER*/ -#endif /*LODEPNG_COMPILE_ZLIB*/ - -#ifdef LODEPNG_COMPILE_DISK -/* -Load a file from disk into buffer. The function allocates the out buffer, and -after usage you should free it. -out: output parameter, contains pointer to loaded buffer. -outsize: output parameter, size of the allocated out buffer -filename: the path to the file to load -return value: error code (0 means ok) -*/ -unsigned lodepng_load_file(unsigned char** out, size_t* outsize, const char* filename); - -/* -Save a file from buffer to disk. Warning, if it exists, this function overwrites -the file without warning! -buffer: the buffer to write -buffersize: size of the buffer to write -filename: the path to the file to save to -return value: error code (0 means ok) -*/ -unsigned lodepng_save_file(const unsigned char* buffer, size_t buffersize, const char* filename); -#endif /*LODEPNG_COMPILE_DISK*/ - -#ifdef LODEPNG_COMPILE_CPP -//The LodePNG C++ wrapper uses std::vectors instead of manually allocated memory buffers. -namespace lodepng -{ -#ifdef LODEPNG_COMPILE_PNG - class State : public LodePNGState - { - public: - State(); - State(const State& other); - virtual ~State(); - State& operator=(const State& other); - }; - -#ifdef LODEPNG_COMPILE_DECODER -//Same as other lodepng::decode, but using a State for more settings and information. - unsigned decode(std::vector& out, unsigned& w, unsigned& h, - State& state, - const unsigned char* in, size_t insize); - unsigned decode(std::vector& out, unsigned& w, unsigned& h, - State& state, - const std::vector& in); -#endif /*LODEPNG_COMPILE_DECODER*/ - -#ifdef LODEPNG_COMPILE_ENCODER -//Same as other lodepng::encode, but using a State for more settings and information. - unsigned encode(std::vector& out, - const unsigned char* in, unsigned w, unsigned h, - State& state); - unsigned encode(std::vector& out, - const std::vector& in, unsigned w, unsigned h, - State& state); -#endif /*LODEPNG_COMPILE_ENCODER*/ - -#ifdef LODEPNG_COMPILE_DISK -/* -Load a file from disk into an std::vector. If the vector is empty, then either -the file doesn't exist or is an empty file. -*/ - void load_file(std::vector& buffer, const std::string& filename); - -/* -Save the binary data in an std::vector to a file on disk. The file is overwritten -without warning. -*/ - void save_file(const std::vector& buffer, const std::string& filename); -#endif //LODEPNG_COMPILE_DISK -#endif //LODEPNG_COMPILE_PNG - -#ifdef LODEPNG_COMPILE_ZLIB -#ifdef LODEPNG_COMPILE_DECODER -//Zlib-decompress an unsigned char buffer - unsigned decompress(std::vector& out, const unsigned char* in, size_t insize, - const LodePNGDecompressSettings& settings = lodepng_default_decompress_settings); - -//Zlib-decompress an std::vector - unsigned decompress(std::vector& out, const std::vector& in, - const LodePNGDecompressSettings& settings = lodepng_default_decompress_settings); -#endif //LODEPNG_COMPILE_DECODER - -#ifdef LODEPNG_COMPILE_ENCODER -//Zlib-compress an unsigned char buffer - unsigned compress(std::vector& out, const unsigned char* in, size_t insize, - const LodePNGCompressSettings& settings = lodepng_default_compress_settings); - -//Zlib-compress an std::vector - unsigned compress(std::vector& out, const std::vector& in, - const LodePNGCompressSettings& settings = lodepng_default_compress_settings); -#endif //LODEPNG_COMPILE_ENCODER -#endif //LODEPNG_COMPILE_ZLIB -} //namespace lodepng -#endif /*LODEPNG_COMPILE_CPP*/ - -/* -TODO: -[.] test if there are no memory leaks or security exploits - done a lot but needs to be checked often -[.] check compatibility with vareous compilers - done but needs to be redone for every newer version -[X] converting color to 16-bit per channel types -[ ] read all public PNG chunk types (but never let the color profile and gamma ones touch RGB values) -[ ] make sure encoder generates no chunks with size > (2^31)-1 -[ ] partial decoding (stream processing) -[X] let the "isFullyOpaque" function check color keys and transparent palettes too -[X] better name for the variables "codes", "codesD", "codelengthcodes", "clcl" and "lldl" -[ ] don't stop decoding on errors like 69, 57, 58 (make warnings) -[ ] make option to choose if the raw image with non multiple of 8 bits per scanline should have padding bits or not -[ ] let the C++ wrapper catch exceptions coming from the standard library and return LodePNG error codes -*/ - -#endif /*LODEPNG_H inclusion guard*/ - -/* -LodePNG Documentation ---------------------- - -0. table of contents --------------------- - - 1. about - 1.1. supported features - 1.2. features not supported - 2. C and C++ version - 3. security - 4. decoding - 5. encoding - 6. color conversions - 6.1. PNG color types - 6.2. color conversions - 6.3. padding bits - 6.4. A note about 16-bits per channel and endianness - 7. error values - 8. chunks and PNG editing - 9. compiler support - 10. examples - 10.1. decoder C++ example - 10.2. decoder C example - 11. changes - 12. contact information - - -1. about --------- - -PNG is a file format to store raster images losslessly with good compression, -supporting different color types and alpha channel. - -LodePNG is a PNG codec according to the Portable Network Graphics (PNG) -Specification (Second Edition) - W3C Recommendation 10 November 2003. - -The specifications used are: - -*) Portable Network Graphics (PNG) Specification (Second Edition): - http://www.w3.org/TR/2003/REC-PNG-20031110 -*) RFC 1950 ZLIB Compressed Data Format version 3.3: - http://www.gzip.org/zlib/rfc-zlib.html -*) RFC 1951 DEFLATE Compressed Data Format Specification ver 1.3: - http://www.gzip.org/zlib/rfc-deflate.html - -The most recent version of LodePNG can currently be found at -http://lodev.org/lodepng/ - -LodePNG works both in C (ISO C90) and C++, with a C++ wrapper that adds -extra functionality. - -LodePNG exists out of two files: --lodepng.h: the header file for both C and C++ --lodepng.c(pp): give it the name lodepng.c or lodepng.cpp (or .cc) depending on your usage - -If you want to start using LodePNG right away without reading this doc, get the -examples from the LodePNG website to see how to use it in code, or check the -smaller examples in chapter 13 here. - -LodePNG is simple but only supports the basic requirements. To achieve -simplicity, the following design choices were made: There are no dependencies -on any external library. There are functions to decode and encode a PNG with -a single function call, and extended versions of these functions taking a -LodePNGState struct allowing to specify or get more information. By default -the colors of the raw image are always RGB or RGBA, no matter what color type -the PNG file uses. To read and write files, there are simple functions to -convert the files to/from buffers in memory. - -This all makes LodePNG suitable for loading textures in games, demos and small -programs, ... It's less suitable for full fledged image editors, loading PNGs -over network (it requires all the image data to be available before decoding can -begin), life-critical systems, ... - -1.1. supported features ------------------------ - -The following features are supported by the decoder: - -*) decoding of PNGs with any color type, bit depth and interlace mode, to a 24- or 32-bit color raw image, - or the same color type as the PNG -*) encoding of PNGs, from any raw image to 24- or 32-bit color, or the same color type as the raw image -*) Adam7 interlace and deinterlace for any color type -*) loading the image from harddisk or decoding it from a buffer from other sources than harddisk -*) support for alpha channels, including RGBA color model, translucent palettes and color keying -*) zlib decompression (inflate) -*) zlib compression (deflate) -*) CRC32 and ADLER32 checksums -*) handling of unknown chunks, allowing making a PNG editor that stores custom and unknown chunks. -*) the following chunks are supported (generated/interpreted) by both encoder and decoder: - IHDR: header information - PLTE: color palette - IDAT: pixel data - IEND: the final chunk - tRNS: transparency for palettized images - tEXt: textual information - zTXt: compressed textual information - iTXt: international textual information - bKGD: suggested background color - pHYs: physical dimensions - tIME: modification time - -1.2. features not supported ---------------------------- - -The following features are _not_ supported: - -*) some features needed to make a conformant PNG-Editor might be still missing. -*) partial loading/stream processing. All data must be available and is processed in one call. -*) The following public chunks are not supported but treated as unknown chunks by LodePNG - cHRM, gAMA, iCCP, sRGB, sBIT, hIST, sPLT - Some of these are not supported on purpose: LodePNG wants to provide the RGB values - stored in the pixels, not values modified by system dependent gamma or color models. - - -2. C and C++ version --------------------- - -The C version uses buffers allocated with alloc that you need to free() -yourself. You need to use init and cleanup functions for each struct whenever -using a struct from the C version to avoid exploits and memory leaks. - -The C++ version has extra functions with std::vectors in the interface and the -lodepng::State class which is a LodePNGState with constructor and destructor. - -These files work without modification for both C and C++ compilers because all -the additional C++ code is in "#ifdef __cplusplus" blocks that make C-compilers -ignore it, and the C code is made to compile both with strict ISO C90 and C++. - -To use the C++ version, you need to rename the source file to lodepng.cpp -(instead of lodepng.c), and compile it with a C++ compiler. - -To use the C version, you need to rename the source file to lodepng.c (instead -of lodepng.cpp), and compile it with a C compiler. - - -3. Security ------------ - -Even if carefully designed, it's always possible that LodePNG contains possible -exploits. If you discover one, please let me know, and it will be fixed. - -When using LodePNG, care has to be taken with the C version of LodePNG, as well -as the C-style structs when working with C++. The following conventions are used -for all C-style structs: - --if a struct has a corresponding init function, always call the init function when making a new one --if a struct has a corresponding cleanup function, call it before the struct disappears to avoid memory leaks --if a struct has a corresponding copy function, use the copy function instead of "=". - The destination must also be inited already. - - -4. Decoding ------------ - -Decoding converts a PNG compressed image to a raw pixel buffer. - -Most documentation on using the decoder is at its declarations in the header -above. For C, simple decoding can be done with functions such as -lodepng_decode32, and more advanced decoding can be done with the struct -LodePNGState and lodepng_decode. For C++, all decoding can be done with the -various lodepng::decode functions, and lodepng::State can be used for advanced -features. - -When using the LodePNGState, it uses the following fields for decoding: -*) LodePNGInfo info_png: it stores extra information about the PNG (the input) in here -*) LodePNGColorMode info_raw: here you can say what color mode of the raw image (the output) you want to get -*) LodePNGDecoderSettings decoder: you can specify a few extra settings for the decoder to use - -LodePNGInfo info_png --------------------- - -After decoding, this contains extra information of the PNG image, except the actual -pixels, width and height because these are already gotten directly from the decoder -functions. - -It contains for example the original color type of the PNG image, text comments, -suggested background color, etc... More details about the LodePNGInfo struct are -at its declaration documentation. - -LodePNGColorMode info_raw -------------------------- - -When decoding, here you can specify which color type you want -the resulting raw image to be. If this is different from the colortype of the -PNG, then the decoder will automatically convert the result. This conversion -always works, except if you want it to convert a color PNG to greyscale or to -a palette with missing colors. - -By default, 32-bit color is used for the result. - -LodePNGDecoderSettings decoder ------------------------------- - -The settings can be used to ignore the errors created by invalid CRC and Adler32 -chunks, and to disable the decoding of tEXt chunks. - -There's also a setting color_convert, true by default. If false, no conversion -is done, the resulting data will be as it was in the PNG (after decompression) -and you'll have to puzzle the colors of the pixels together yourself using the -color type information in the LodePNGInfo. - - -5. Encoding ------------ - -Encoding converts a raw pixel buffer to a PNG compressed image. - -Most documentation on using the encoder is at its declarations in the header -above. For C, simple encoding can be done with functions such as -lodepng_encode32, and more advanced decoding can be done with the struct -LodePNGState and lodepng_encode. For C++, all encoding can be done with the -various lodepng::encode functions, and lodepng::State can be used for advanced -features. - -Like the decoder, the encoder can also give errors. However it gives less errors -since the encoder input is trusted, the decoder input (a PNG image that could -be forged by anyone) is not trusted. - -When using the LodePNGState, it uses the following fields for encoding: -*) LodePNGInfo info_png: here you specify how you want the PNG (the output) to be. -*) LodePNGColorMode info_raw: here you say what color type of the raw image (the input) has -*) LodePNGEncoderSettings encoder: you can specify a few settings for the encoder to use - -LodePNGInfo info_png --------------------- - -When encoding, you use this the opposite way as when decoding: for encoding, -you fill in the values you want the PNG to have before encoding. By default it's -not needed to specify a color type for the PNG since it's automatically chosen, -but it's possible to choose it yourself given the right settings. - -The encoder will not always exactly match the LodePNGInfo struct you give, -it tries as close as possible. Some things are ignored by the encoder. The -encoder uses, for example, the following settings from it when applicable: -colortype and bitdepth, text chunks, time chunk, the color key, the palette, the -background color, the interlace method, unknown chunks, ... - -When encoding to a PNG with colortype 3, the encoder will generate a PLTE chunk. -If the palette contains any colors for which the alpha channel is not 255 (so -there are translucent colors in the palette), it'll add a tRNS chunk. - -LodePNGColorMode info_raw -------------------------- - -You specify the color type of the raw image that you give to the input here, -including a possible transparent color key and palette you happen to be using in -your raw image data. - -By default, 32-bit color is assumed, meaning your input has to be in RGBA -format with 4 bytes (unsigned chars) per pixel. - -LodePNGEncoderSettings encoder ------------------------------- - -The following settings are supported (some are in sub-structs): -*) auto_convert: when this option is enabled, the encoder will -automatically choose the smallest possible color mode (including color key) that -can encode the colors of all pixels without information loss. -*) btype: the block type for LZ77. 0 = uncompressed, 1 = fixed huffman tree, - 2 = dynamic huffman tree (best compression). Should be 2 for proper - compression. -*) use_lz77: whether or not to use LZ77 for compressed block types. Should be - true for proper compression. -*) windowsize: the window size used by the LZ77 encoder (1 - 32768). Has value - 2048 by default, but can be set to 32768 for better, but slow, compression. -*) force_palette: if colortype is 2 or 6, you can make the encoder write a PLTE - chunk if force_palette is true. This can used as suggested palette to convert - to by viewers that don't support more than 256 colors (if those still exist) -*) add_id: add text chunk "Encoder: LodePNG " to the image. -*) text_compression: default 1. If 1, it'll store texts as zTXt instead of tEXt chunks. - zTXt chunks use zlib compression on the text. This gives a smaller result on - large texts but a larger result on small texts (such as a single program name). - It's all tEXt or all zTXt though, there's no separate setting per text yet. - - -6. color conversions --------------------- - -An important thing to note about LodePNG, is that the color type of the PNG, and -the color type of the raw image, are completely independent. By default, when -you decode a PNG, you get the result as a raw image in the color type you want, -no matter whether the PNG was encoded with a palette, greyscale or RGBA color. -And if you encode an image, by default LodePNG will automatically choose the PNG -color type that gives good compression based on the values of colors and amount -of colors in the image. It can be configured to let you control it instead as -well, though. - -To be able to do this, LodePNG does conversions from one color mode to another. -It can convert from almost any color type to any other color type, except the -following conversions: RGB to greyscale is not supported, and converting to a -palette when the palette doesn't have a required color is not supported. This is -not supported on purpose: this is information loss which requires a color -reduction algorithm that is beyong the scope of a PNG encoder (yes, RGB to grey -is easy, but there are multiple ways if you want to give some channels more -weight). - -By default, when decoding, you get the raw image in 32-bit RGBA or 24-bit RGB -color, no matter what color type the PNG has. And by default when encoding, -LodePNG automatically picks the best color model for the output PNG, and expects -the input image to be 32-bit RGBA or 24-bit RGB. So, unless you want to control -the color format of the images yourself, you can skip this chapter. - -6.1. PNG color types --------------------- - -A PNG image can have many color types, ranging from 1-bit color to 64-bit color, -as well as palettized color modes. After the zlib decompression and unfiltering -in the PNG image is done, the raw pixel data will have that color type and thus -a certain amount of bits per pixel. If you want the output raw image after -decoding to have another color type, a conversion is done by LodePNG. - -The PNG specification gives the following color types: - -0: greyscale, bit depths 1, 2, 4, 8, 16 -2: RGB, bit depths 8 and 16 -3: palette, bit depths 1, 2, 4 and 8 -4: greyscale with alpha, bit depths 8 and 16 -6: RGBA, bit depths 8 and 16 - -Bit depth is the amount of bits per pixel per color channel. So the total amount -of bits per pixel is: amount of channels * bitdepth. - -6.2. color conversions ----------------------- - -As explained in the sections about the encoder and decoder, you can specify -color types and bit depths in info_png and info_raw to change the default -behaviour. - -If, when decoding, you want the raw image to be something else than the default, -you need to set the color type and bit depth you want in the LodePNGColorMode, -or the parameters colortype and bitdepth of the simple decoding function. - -If, when encoding, you use another color type than the default in the raw input -image, you need to specify its color type and bit depth in the LodePNGColorMode -of the raw image, or use the parameters colortype and bitdepth of the simple -encoding function. - -If, when encoding, you don't want LodePNG to choose the output PNG color type -but control it yourself, you need to set auto_convert in the encoder settings -to false, and specify the color type you want in the LodePNGInfo of the -encoder (including palette: it can generate a palette if auto_convert is true, -otherwise not). - -If the input and output color type differ (whether user chosen or auto chosen), -LodePNG will do a color conversion, which follows the rules below, and may -sometimes result in an error. - -To avoid some confusion: --the decoder converts from PNG to raw image --the encoder converts from raw image to PNG --the colortype and bitdepth in LodePNGColorMode info_raw, are those of the raw image --the colortype and bitdepth in the color field of LodePNGInfo info_png, are those of the PNG --when encoding, the color type in LodePNGInfo is ignored if auto_convert - is enabled, it is automatically generated instead --when decoding, the color type in LodePNGInfo is set by the decoder to that of the original - PNG image, but it can be ignored since the raw image has the color type you requested instead --if the color type of the LodePNGColorMode and PNG image aren't the same, a conversion - between the color types is done if the color types are supported. If it is not - supported, an error is returned. If the types are the same, no conversion is done. --even though some conversions aren't supported, LodePNG supports loading PNGs from any - colortype and saving PNGs to any colortype, sometimes it just requires preparing - the raw image correctly before encoding. --both encoder and decoder use the same color converter. - -Non supported color conversions: --color to greyscale: no error is thrown, but the result will look ugly because -only the red channel is taken --anything to palette when that palette does not have that color in it: in this -case an error is thrown - -Supported color conversions: --anything to 8-bit RGB, 8-bit RGBA, 16-bit RGB, 16-bit RGBA --any grey or grey+alpha, to grey or grey+alpha --anything to a palette, as long as the palette has the requested colors in it --removing alpha channel --higher to smaller bitdepth, and vice versa - -If you want no color conversion to be done (e.g. for speed or control): --In the encoder, you can make it save a PNG with any color type by giving the -raw color mode and LodePNGInfo the same color mode, and setting auto_convert to -false. --In the decoder, you can make it store the pixel data in the same color type -as the PNG has, by setting the color_convert setting to false. Settings in -info_raw are then ignored. - -The function lodepng_convert does the color conversion. It is available in the -interface but normally isn't needed since the encoder and decoder already call -it. - -6.3. padding bits ------------------ - -In the PNG file format, if a less than 8-bit per pixel color type is used and the scanlines -have a bit amount that isn't a multiple of 8, then padding bits are used so that each -scanline starts at a fresh byte. But that is NOT true for the LodePNG raw input and output. -The raw input image you give to the encoder, and the raw output image you get from the decoder -will NOT have these padding bits, e.g. in the case of a 1-bit image with a width -of 7 pixels, the first pixel of the second scanline will the the 8th bit of the first byte, -not the first bit of a new byte. - -6.4. A note about 16-bits per channel and endianness ----------------------------------------------------- - -LodePNG uses unsigned char arrays for 16-bit per channel colors too, just like -for any other color format. The 16-bit values are stored in big endian (most -significant byte first) in these arrays. This is the opposite order of the -little endian used by x86 CPU's. - -LodePNG always uses big endian because the PNG file format does so internally. -Conversions to other formats than PNG uses internally are not supported by -LodePNG on purpose, there are myriads of formats, including endianness of 16-bit -colors, the order in which you store R, G, B and A, and so on. Supporting and -converting to/from all that is outside the scope of LodePNG. - -This may mean that, depending on your use case, you may want to convert the big -endian output of LodePNG to little endian with a for loop. This is certainly not -always needed, many applications and libraries support big endian 16-bit colors -anyway, but it means you cannot simply cast the unsigned char* buffer to an -unsigned short* buffer on x86 CPUs. - - -7. error values ---------------- - -All functions in LodePNG that return an error code, return 0 if everything went -OK, or a non-zero code if there was an error. - -The meaning of the LodePNG error values can be retrieved with the function -lodepng_error_text: given the numerical error code, it returns a description -of the error in English as a string. - -Check the implementation of lodepng_error_text to see the meaning of each code. - - -8. chunks and PNG editing -------------------------- - -If you want to add extra chunks to a PNG you encode, or use LodePNG for a PNG -editor that should follow the rules about handling of unknown chunks, or if your -program is able to read other types of chunks than the ones handled by LodePNG, -then that's possible with the chunk functions of LodePNG. - -A PNG chunk has the following layout: - -4 bytes length -4 bytes type name -length bytes data -4 bytes CRC - -8.1. iterating through chunks ------------------------------ - -If you have a buffer containing the PNG image data, then the first chunk (the -IHDR chunk) starts at byte number 8 of that buffer. The first 8 bytes are the -signature of the PNG and are not part of a chunk. But if you start at byte 8 -then you have a chunk, and can check the following things of it. - -NOTE: none of these functions check for memory buffer boundaries. To avoid -exploits, always make sure the buffer contains all the data of the chunks. -When using lodepng_chunk_next, make sure the returned value is within the -allocated memory. - -unsigned lodepng_chunk_length(const unsigned char* chunk): - -Get the length of the chunk's data. The total chunk length is this length + 12. - -void lodepng_chunk_type(char type[5], const unsigned char* chunk): -unsigned char lodepng_chunk_type_equals(const unsigned char* chunk, const char* type): - -Get the type of the chunk or compare if it's a certain type - -unsigned char lodepng_chunk_critical(const unsigned char* chunk): -unsigned char lodepng_chunk_private(const unsigned char* chunk): -unsigned char lodepng_chunk_safetocopy(const unsigned char* chunk): - -Check if the chunk is critical in the PNG standard (only IHDR, PLTE, IDAT and IEND are). -Check if the chunk is private (public chunks are part of the standard, private ones not). -Check if the chunk is safe to copy. If it's not, then, when modifying data in a critical -chunk, unsafe to copy chunks of the old image may NOT be saved in the new one if your -program doesn't handle that type of unknown chunk. - -unsigned char* lodepng_chunk_data(unsigned char* chunk): -const unsigned char* lodepng_chunk_data_const(const unsigned char* chunk): - -Get a pointer to the start of the data of the chunk. - -unsigned lodepng_chunk_check_crc(const unsigned char* chunk): -void lodepng_chunk_generate_crc(unsigned char* chunk): - -Check if the crc is correct or generate a correct one. - -unsigned char* lodepng_chunk_next(unsigned char* chunk): -const unsigned char* lodepng_chunk_next_const(const unsigned char* chunk): - -Iterate to the next chunk. This works if you have a buffer with consecutive chunks. Note that these -functions do no boundary checking of the allocated data whatsoever, so make sure there is enough -data available in the buffer to be able to go to the next chunk. - -unsigned lodepng_chunk_append(unsigned char** out, size_t* outlength, const unsigned char* chunk): -unsigned lodepng_chunk_create(unsigned char** out, size_t* outlength, unsigned length, - const char* type, const unsigned char* data): - -These functions are used to create new chunks that are appended to the data in *out that has -length *outlength. The append function appends an existing chunk to the new data. The create -function creates a new chunk with the given parameters and appends it. Type is the 4-letter -name of the chunk. - -8.2. chunks in info_png ------------------------ - -The LodePNGInfo struct contains fields with the unknown chunk in it. It has 3 -buffers (each with size) to contain 3 types of unknown chunks: -the ones that come before the PLTE chunk, the ones that come between the PLTE -and the IDAT chunks, and the ones that come after the IDAT chunks. -It's necessary to make the distionction between these 3 cases because the PNG -standard forces to keep the ordering of unknown chunks compared to the critical -chunks, but does not force any other ordering rules. - -info_png.unknown_chunks_data[0] is the chunks before PLTE -info_png.unknown_chunks_data[1] is the chunks after PLTE, before IDAT -info_png.unknown_chunks_data[2] is the chunks after IDAT - -The chunks in these 3 buffers can be iterated through and read by using the same -way described in the previous subchapter. - -When using the decoder to decode a PNG, you can make it store all unknown chunks -if you set the option settings.remember_unknown_chunks to 1. By default, this -option is off (0). - -The encoder will always encode unknown chunks that are stored in the info_png. -If you need it to add a particular chunk that isn't known by LodePNG, you can -use lodepng_chunk_append or lodepng_chunk_create to the chunk data in -info_png.unknown_chunks_data[x]. - -Chunks that are known by LodePNG should not be added in that way. E.g. to make -LodePNG add a bKGD chunk, set background_defined to true and add the correct -parameters there instead. - - -9. compiler support -------------------- - -No libraries other than the current standard C library are needed to compile -LodePNG. For the C++ version, only the standard C++ library is needed on top. -Add the files lodepng.c(pp) and lodepng.h to your project, include -lodepng.h where needed, and your program can read/write PNG files. - -It is compatible with C90 and up, and C++03 and up. - -If performance is important, use optimization when compiling! For both the -encoder and decoder, this makes a large difference. - -Make sure that LodePNG is compiled with the same compiler of the same version -and with the same settings as the rest of the program, or the interfaces with -std::vectors and std::strings in C++ can be incompatible. - -CHAR_BITS must be 8 or higher, because LodePNG uses unsigned chars for octets. - -*) gcc and g++ - -LodePNG is developed in gcc so this compiler is natively supported. It gives no -warnings with compiler options "-Wall -Wextra -pedantic -ansi", with gcc and g++ -version 4.7.1 on Linux, 32-bit and 64-bit. - -*) Clang - -Fully supported and warning-free. - -*) Mingw - -The Mingw compiler (a port of gcc for Windows) should be fully supported by -LodePNG. - -*) Visual Studio and Visual C++ Express Edition - -LodePNG should be warning-free with warning level W4. Two warnings were disabled -with pragmas though: warning 4244 about implicit conversions, and warning 4996 -where it wants to use a non-standard function fopen_s instead of the standard C -fopen. - -Visual Studio may want "stdafx.h" files to be included in each source file and -give an error "unexpected end of file while looking for precompiled header". -This is not standard C++ and will not be added to the stock LodePNG. You can -disable it for lodepng.cpp only by right clicking it, Properties, C/C++, -Precompiled Headers, and set it to Not Using Precompiled Headers there. - -NOTE: Modern versions of VS should be fully supported, but old versions, e.g. -VS6, are not guaranteed to work. - -*) Compilers on Macintosh - -LodePNG has been reported to work both with gcc and LLVM for Macintosh, both for -C and C++. - -*) Other Compilers - -If you encounter problems on any compilers, feel free to let me know and I may -try to fix it if the compiler is modern and standards complient. - - -10. examples ------------- - -This decoder example shows the most basic usage of LodePNG. More complex -examples can be found on the LodePNG website. - -10.1. decoder C++ example -------------------------- - -#include "lodepng.h" -#include - -int main(int argc, char *argv[]) -{ - const char* filename = argc > 1 ? argv[1] : "test.png"; - - //load and decode - std::vector image; - unsigned width, height; - unsigned error = lodepng::decode(image, width, height, filename); - - //if there's an error, display it - if(error) std::cout << "decoder error " << error << ": " << lodepng_error_text(error) << std::endl; - - //the pixels are now in the vector "image", 4 bytes per pixel, ordered RGBARGBA..., use it as texture, draw it, ... -} - -10.2. decoder C example ------------------------ - -#include "lodepng.h" - -int main(int argc, char *argv[]) -{ - unsigned error; - unsigned char* image; - size_t width, height; - const char* filename = argc > 1 ? argv[1] : "test.png"; - - error = lodepng_decode32_file(&image, &width, &height, filename); - - if(error) printf("decoder error %u: %s\n", error, lodepng_error_text(error)); - - / * use image here * / - - free(image); - return 0; -} - - -11. changes ------------ - -The version number of LodePNG is the date of the change given in the format -yyyymmdd. - -Some changes aren't backwards compatible. Those are indicated with a (!) -symbol. - -*) 23 aug 2014: Reduced needless memory usage of decoder. -*) 28 jun 2014: Removed fix_png setting, always support palette OOB for - simplicity. Made ColorProfile public. -*) 09 jun 2014: Faster encoder by fixing hash bug and more zeros optimization. -*) 22 dec 2013: Power of two windowsize required for optimization. -*) 15 apr 2013: Fixed bug with LAC_ALPHA and color key. -*) 25 mar 2013: Added an optional feature to ignore some PNG errors (fix_png). -*) 11 mar 2013 (!): Bugfix with custom free. Changed from "my" to "lodepng_" - prefix for the custom allocators and made it possible with a new #define to - use custom ones in your project without needing to change lodepng's code. -*) 28 jan 2013: Bugfix with color key. -*) 27 okt 2012: Tweaks in text chunk keyword length error handling. -*) 8 okt 2012 (!): Added new filter strategy (entropy) and new auto color mode. - (no palette). Better deflate tree encoding. New compression tweak settings. - Faster color conversions while decoding. Some internal cleanups. -*) 23 sep 2012: Reduced warnings in Visual Studio a little bit. -*) 1 sep 2012 (!): Removed #define's for giving custom (de)compression functions - and made it work with function pointers instead. -*) 23 jun 2012: Added more filter strategies. Made it easier to use custom alloc - and free functions and toggle #defines from compiler flags. Small fixes. -*) 6 may 2012 (!): Made plugging in custom zlib/deflate functions more flexible. -*) 22 apr 2012 (!): Made interface more consistent, renaming a lot. Removed - redundant C++ codec classes. Reduced amount of structs. Everything changed, - but it is cleaner now imho and functionality remains the same. Also fixed - several bugs and shrinked the implementation code. Made new samples. -*) 6 nov 2011 (!): By default, the encoder now automatically chooses the best - PNG color model and bit depth, based on the amount and type of colors of the - raw image. For this, autoLeaveOutAlphaChannel replaced by auto_choose_color. -*) 9 okt 2011: simpler hash chain implementation for the encoder. -*) 8 sep 2011: lz77 encoder lazy matching instead of greedy matching. -*) 23 aug 2011: tweaked the zlib compression parameters after benchmarking. - A bug with the PNG filtertype heuristic was fixed, so that it chooses much - better ones (it's quite significant). A setting to do an experimental, slow, - brute force search for PNG filter types is added. -*) 17 aug 2011 (!): changed some C zlib related function names. -*) 16 aug 2011: made the code less wide (max 120 characters per line). -*) 17 apr 2011: code cleanup. Bugfixes. Convert low to 16-bit per sample colors. -*) 21 feb 2011: fixed compiling for C90. Fixed compiling with sections disabled. -*) 11 dec 2010: encoding is made faster, based on suggestion by Peter Eastman - to optimize long sequences of zeros. -*) 13 nov 2010: added LodePNG_InfoColor_hasPaletteAlpha and - LodePNG_InfoColor_canHaveAlpha functions for convenience. -*) 7 nov 2010: added LodePNG_error_text function to get error code description. -*) 30 okt 2010: made decoding slightly faster -*) 26 okt 2010: (!) changed some C function and struct names (more consistent). - Reorganized the documentation and the declaration order in the header. -*) 08 aug 2010: only changed some comments and external samples. -*) 05 jul 2010: fixed bug thanks to warnings in the new gcc version. -*) 14 mar 2010: fixed bug where too much memory was allocated for char buffers. -*) 02 sep 2008: fixed bug where it could create empty tree that linux apps could - read by ignoring the problem but windows apps couldn't. -*) 06 jun 2008: added more error checks for out of memory cases. -*) 26 apr 2008: added a few more checks here and there to ensure more safety. -*) 06 mar 2008: crash with encoding of strings fixed -*) 02 feb 2008: support for international text chunks added (iTXt) -*) 23 jan 2008: small cleanups, and #defines to divide code in sections -*) 20 jan 2008: support for unknown chunks allowing using LodePNG for an editor. -*) 18 jan 2008: support for tIME and pHYs chunks added to encoder and decoder. -*) 17 jan 2008: ability to encode and decode compressed zTXt chunks added - Also vareous fixes, such as in the deflate and the padding bits code. -*) 13 jan 2008: Added ability to encode Adam7-interlaced images. Improved - filtering code of encoder. -*) 07 jan 2008: (!) changed LodePNG to use ISO C90 instead of C++. A - C++ wrapper around this provides an interface almost identical to before. - Having LodePNG be pure ISO C90 makes it more portable. The C and C++ code - are together in these files but it works both for C and C++ compilers. -*) 29 dec 2007: (!) changed most integer types to unsigned int + other tweaks -*) 30 aug 2007: bug fixed which makes this Borland C++ compatible -*) 09 aug 2007: some VS2005 warnings removed again -*) 21 jul 2007: deflate code placed in new namespace separate from zlib code -*) 08 jun 2007: fixed bug with 2- and 4-bit color, and small interlaced images -*) 04 jun 2007: improved support for Visual Studio 2005: crash with accessing - invalid std::vector element [0] fixed, and level 3 and 4 warnings removed -*) 02 jun 2007: made the encoder add a tag with version by default -*) 27 may 2007: zlib and png code separated (but still in the same file), - simple encoder/decoder functions added for more simple usage cases -*) 19 may 2007: minor fixes, some code cleaning, new error added (error 69), - moved some examples from here to lodepng_examples.cpp -*) 12 may 2007: palette decoding bug fixed -*) 24 apr 2007: changed the license from BSD to the zlib license -*) 11 mar 2007: very simple addition: ability to encode bKGD chunks. -*) 04 mar 2007: (!) tEXt chunk related fixes, and support for encoding - palettized PNG images. Plus little interface change with palette and texts. -*) 03 mar 2007: Made it encode dynamic Huffman shorter with repeat codes. - Fixed a bug where the end code of a block had length 0 in the Huffman tree. -*) 26 feb 2007: Huffman compression with dynamic trees (BTYPE 2) now implemented - and supported by the encoder, resulting in smaller PNGs at the output. -*) 27 jan 2007: Made the Adler-32 test faster so that a timewaste is gone. -*) 24 jan 2007: gave encoder an error interface. Added color conversion from any - greyscale type to 8-bit greyscale with or without alpha. -*) 21 jan 2007: (!) Totally changed the interface. It allows more color types - to convert to and is more uniform. See the manual for how it works now. -*) 07 jan 2007: Some cleanup & fixes, and a few changes over the last days: - encode/decode custom tEXt chunks, separate classes for zlib & deflate, and - at last made the decoder give errors for incorrect Adler32 or Crc. -*) 01 jan 2007: Fixed bug with encoding PNGs with less than 8 bits per channel. -*) 29 dec 2006: Added support for encoding images without alpha channel, and - cleaned out code as well as making certain parts faster. -*) 28 dec 2006: Added "Settings" to the encoder. -*) 26 dec 2006: The encoder now does LZ77 encoding and produces much smaller files now. - Removed some code duplication in the decoder. Fixed little bug in an example. -*) 09 dec 2006: (!) Placed output parameters of public functions as first parameter. - Fixed a bug of the decoder with 16-bit per color. -*) 15 okt 2006: Changed documentation structure -*) 09 okt 2006: Encoder class added. It encodes a valid PNG image from the - given image buffer, however for now it's not compressed. -*) 08 sep 2006: (!) Changed to interface with a Decoder class -*) 30 jul 2006: (!) LodePNG_InfoPng , width and height are now retrieved in different - way. Renamed decodePNG to decodePNGGeneric. -*) 29 jul 2006: (!) Changed the interface: image info is now returned as a - struct of type LodePNG::LodePNG_Info, instead of a vector, which was a bit clumsy. -*) 28 jul 2006: Cleaned the code and added new error checks. - Corrected terminology "deflate" into "inflate". -*) 23 jun 2006: Added SDL example in the documentation in the header, this - example allows easy debugging by displaying the PNG and its transparency. -*) 22 jun 2006: (!) Changed way to obtain error value. Added - loadFile function for convenience. Made decodePNG32 faster. -*) 21 jun 2006: (!) Changed type of info vector to unsigned. - Changed position of palette in info vector. Fixed an important bug that - happened on PNGs with an uncompressed block. -*) 16 jun 2006: Internally changed unsigned into unsigned where - needed, and performed some optimizations. -*) 07 jun 2006: (!) Renamed functions to decodePNG and placed them - in LodePNG namespace. Changed the order of the parameters. Rewrote the - documentation in the header. Renamed files to lodepng.cpp and lodepng.h -*) 22 apr 2006: Optimized and improved some code -*) 07 sep 2005: (!) Changed to std::vector interface -*) 12 aug 2005: Initial release (C++, decoder only) - - -12. contact information ------------------------ - -Feel free to contact me with suggestions, problems, comments, ... concerning -LodePNG. If you encounter a PNG image that doesn't work properly with this -decoder, feel free to send it and I'll use it to find and fix the problem. - -My email address is (puzzle the account and domain together with an @ symbol): -Domain: gmail dot com. -Account: lode dot vandevenne. - - -Copyright (c) 2005-2014 Lode Vandevenne -*/ \ No newline at end of file diff --git a/source/pc/stb_image.c b/source/pc/stb_image.c new file mode 100644 index 0000000..8d45f85 --- /dev/null +++ b/source/pc/stb_image.c @@ -0,0 +1,3 @@ +#define STB_IMAGE_IMPLEMENTATION +#define STBI_ONLY_PNG +#include "stb_image.h" \ No newline at end of file diff --git a/source/pc/stb_image.h b/source/pc/stb_image.h new file mode 100644 index 0000000..e06f7a1 --- /dev/null +++ b/source/pc/stb_image.h @@ -0,0 +1,6614 @@ +/* stb_image - v2.10 - public domain image loader - http://nothings.org/stb_image.h + no warranty implied; use at your own risk + + Do this: + #define STB_IMAGE_IMPLEMENTATION + before you include this file in *one* C or C++ file to create the implementation. + + // i.e. it should look like this: + #include ... + #include ... + #include ... + #define STB_IMAGE_IMPLEMENTATION + #include "stb_image.h" + + You can #define STBI_ASSERT(x) before the #include to avoid using assert.h. + And #define STBI_MALLOC, STBI_REALLOC, and STBI_FREE to avoid using malloc,realloc,free + + + QUICK NOTES: + Primarily of interest to game developers and other people who can + avoid problematic images and only need the trivial interface + + JPEG baseline & progressive (12 bpc/arithmetic not supported, same as stock IJG lib) + PNG 1/2/4/8-bit-per-channel (16 bpc not supported) + + TGA (not sure what subset, if a subset) + BMP non-1bpp, non-RLE + PSD (composited view only, no extra channels, 8/16 bit-per-channel) + + GIF (*comp always reports as 4-channel) + HDR (radiance rgbE format) + PIC (Softimage PIC) + PNM (PPM and PGM binary only) + + Animated GIF still needs a proper API, but here's one way to do it: + http://gist.github.com/urraka/685d9a6340b26b830d49 + + - decode from memory or through FILE (define STBI_NO_STDIO to remove code) + - decode from arbitrary I/O callbacks + - SIMD acceleration on x86/x64 (SSE2) and ARM (NEON) + + Full documentation under "DOCUMENTATION" below. + + + Revision 2.00 release notes: + + - Progressive JPEG is now supported. + + - PPM and PGM binary formats are now supported, thanks to Ken Miller. + + - x86 platforms now make use of SSE2 SIMD instructions for + JPEG decoding, and ARM platforms can use NEON SIMD if requested. + This work was done by Fabian "ryg" Giesen. SSE2 is used by + default, but NEON must be enabled explicitly; see docs. + + With other JPEG optimizations included in this version, we see + 2x speedup on a JPEG on an x86 machine, and a 1.5x speedup + on a JPEG on an ARM machine, relative to previous versions of this + library. The same results will not obtain for all JPGs and for all + x86/ARM machines. (Note that progressive JPEGs are significantly + slower to decode than regular JPEGs.) This doesn't mean that this + is the fastest JPEG decoder in the land; rather, it brings it + closer to parity with standard libraries. If you want the fastest + decode, look elsewhere. (See "Philosophy" section of docs below.) + + See final bullet items below for more info on SIMD. + + - Added STBI_MALLOC, STBI_REALLOC, and STBI_FREE macros for replacing + the memory allocator. Unlike other STBI libraries, these macros don't + support a context parameter, so if you need to pass a context in to + the allocator, you'll have to store it in a global or a thread-local + variable. + + - Split existing STBI_NO_HDR flag into two flags, STBI_NO_HDR and + STBI_NO_LINEAR. + STBI_NO_HDR: suppress implementation of .hdr reader format + STBI_NO_LINEAR: suppress high-dynamic-range light-linear float API + + - You can suppress implementation of any of the decoders to reduce + your code footprint by #defining one or more of the following + symbols before creating the implementation. + + STBI_NO_JPEG + STBI_NO_PNG + STBI_NO_BMP + STBI_NO_PSD + STBI_NO_TGA + STBI_NO_GIF + STBI_NO_HDR + STBI_NO_PIC + STBI_NO_PNM (.ppm and .pgm) + + - You can request *only* certain decoders and suppress all other ones + (this will be more forward-compatible, as addition of new decoders + doesn't require you to disable them explicitly): + + STBI_ONLY_JPEG + STBI_ONLY_PNG + STBI_ONLY_BMP + STBI_ONLY_PSD + STBI_ONLY_TGA + STBI_ONLY_GIF + STBI_ONLY_HDR + STBI_ONLY_PIC + STBI_ONLY_PNM (.ppm and .pgm) + + Note that you can define multiples of these, and you will get all + of them ("only x" and "only y" is interpreted to mean "only x&y"). + + - If you use STBI_NO_PNG (or _ONLY_ without PNG), and you still + want the zlib decoder to be available, #define STBI_SUPPORT_ZLIB + + - Compilation of all SIMD code can be suppressed with + #define STBI_NO_SIMD + It should not be necessary to disable SIMD unless you have issues + compiling (e.g. using an x86 compiler which doesn't support SSE + intrinsics or that doesn't support the method used to detect + SSE2 support at run-time), and even those can be reported as + bugs so I can refine the built-in compile-time checking to be + smarter. + + - The old STBI_SIMD system which allowed installing a user-defined + IDCT etc. has been removed. If you need this, don't upgrade. My + assumption is that almost nobody was doing this, and those who + were will find the built-in SIMD more satisfactory anyway. + + - RGB values computed for JPEG images are slightly different from + previous versions of stb_image. (This is due to using less + integer precision in SIMD.) The C code has been adjusted so + that the same RGB values will be computed regardless of whether + SIMD support is available, so your app should always produce + consistent results. But these results are slightly different from + previous versions. (Specifically, about 3% of available YCbCr values + will compute different RGB results from pre-1.49 versions by +-1; + most of the deviating values are one smaller in the G channel.) + + - If you must produce consistent results with previous versions of + stb_image, #define STBI_JPEG_OLD and you will get the same results + you used to; however, you will not get the SIMD speedups for + the YCbCr-to-RGB conversion step (although you should still see + significant JPEG speedup from the other changes). + + Please note that STBI_JPEG_OLD is a temporary feature; it will be + removed in future versions of the library. It is only intended for + near-term back-compatibility use. + + + Latest revision history: + 2.10 (2016-01-22) avoid warning introduced in 2.09 + 2.09 (2016-01-16) 16-bit TGA; comments in PNM files; STBI_REALLOC_SIZED + 2.08 (2015-09-13) fix to 2.07 cleanup, reading RGB PSD as RGBA + 2.07 (2015-09-13) partial animated GIF support + limited 16-bit PSD support + minor bugs, code cleanup, and compiler warnings + 2.06 (2015-04-19) fix bug where PSD returns wrong '*comp' value + 2.05 (2015-04-19) fix bug in progressive JPEG handling, fix warning + 2.04 (2015-04-15) try to re-enable SIMD on MinGW 64-bit + 2.03 (2015-04-12) additional corruption checking + stbi_set_flip_vertically_on_load + fix NEON support; fix mingw support + 2.02 (2015-01-19) fix incorrect assert, fix warning + 2.01 (2015-01-17) fix various warnings + 2.00b (2014-12-25) fix STBI_MALLOC in progressive JPEG + 2.00 (2014-12-25) optimize JPEG, including x86 SSE2 & ARM NEON SIMD + progressive JPEG + PGM/PPM support + STBI_MALLOC,STBI_REALLOC,STBI_FREE + STBI_NO_*, STBI_ONLY_* + GIF bugfix + 1.48 (2014-12-14) fix incorrectly-named assert() + 1.47 (2014-12-14) 1/2/4-bit PNG support (both grayscale and paletted) + optimize PNG + fix bug in interlaced PNG with user-specified channel count + + See end of file for full revision history. + + + ============================ Contributors ========================= + + Image formats Extensions, features + Sean Barrett (jpeg, png, bmp) Jetro Lauha (stbi_info) + Nicolas Schulz (hdr, psd) Martin "SpartanJ" Golini (stbi_info) + Jonathan Dummer (tga) James "moose2000" Brown (iPhone PNG) + Jean-Marc Lienher (gif) Ben "Disch" Wenger (io callbacks) + Tom Seddon (pic) Omar Cornut (1/2/4-bit PNG) + Thatcher Ulrich (psd) Nicolas Guillemot (vertical flip) + Ken Miller (pgm, ppm) Richard Mitton (16-bit PSD) + urraka@github (animated gif) Junggon Kim (PNM comments) + Daniel Gibson (16-bit TGA) + + Optimizations & bugfixes + Fabian "ryg" Giesen + Arseny Kapoulkine + + Bug & warning fixes + Marc LeBlanc David Woo Guillaume George Martins Mozeiko + Christpher Lloyd Martin Golini Jerry Jansson Joseph Thomson + Dave Moore Roy Eltham Hayaki Saito Phil Jordan + Won Chun Luke Graham Johan Duparc Nathan Reed + the Horde3D community Thomas Ruf Ronny Chevalier Nick Verigakis + Janez Zemva John Bartholomew Michal Cichon svdijk@github + Jonathan Blow Ken Hamada Tero Hanninen Baldur Karlsson + Laurent Gomila Cort Stratton Sergio Gonzalez romigrou@github + Aruelien Pocheville Thibault Reuille Cass Everitt + Ryamond Barbiero Paul Du Bois Engin Manap + Blazej Dariusz Roszkowski + Michaelangel007@github + + +LICENSE + +This software is in the public domain. Where that dedication is not +recognized, you are granted a perpetual, irrevocable license to copy, +distribute, and modify this file as you see fit. + +*/ + +#ifndef STBI_INCLUDE_STB_IMAGE_H +#define STBI_INCLUDE_STB_IMAGE_H + +// DOCUMENTATION +// +// Limitations: +// - no 16-bit-per-channel PNG +// - no 12-bit-per-channel JPEG +// - no JPEGs with arithmetic coding +// - no 1-bit BMP +// - GIF always returns *comp=4 +// +// Basic usage (see HDR discussion below for HDR usage): +// int x,y,n; +// unsigned char *data = stbi_load(filename, &x, &y, &n, 0); +// // ... process data if not NULL ... +// // ... x = width, y = height, n = # 8-bit components per pixel ... +// // ... replace '0' with '1'..'4' to force that many components per pixel +// // ... but 'n' will always be the number that it would have been if you said 0 +// stbi_image_free(data) +// +// Standard parameters: +// int *x -- outputs image width in pixels +// int *y -- outputs image height in pixels +// int *comp -- outputs # of image components in image file +// int req_comp -- if non-zero, # of image components requested in result +// +// The return value from an image loader is an 'unsigned char *' which points +// to the pixel data, or NULL on an allocation failure or if the image is +// corrupt or invalid. The pixel data consists of *y scanlines of *x pixels, +// with each pixel consisting of N interleaved 8-bit components; the first +// pixel pointed to is top-left-most in the image. There is no padding between +// image scanlines or between pixels, regardless of format. The number of +// components N is 'req_comp' if req_comp is non-zero, or *comp otherwise. +// If req_comp is non-zero, *comp has the number of components that _would_ +// have been output otherwise. E.g. if you set req_comp to 4, you will always +// get RGBA output, but you can check *comp to see if it's trivially opaque +// because e.g. there were only 3 channels in the source image. +// +// An output image with N components has the following components interleaved +// in this order in each pixel: +// +// N=#comp components +// 1 grey +// 2 grey, alpha +// 3 red, green, blue +// 4 red, green, blue, alpha +// +// If image loading fails for any reason, the return value will be NULL, +// and *x, *y, *comp will be unchanged. The function stbi_failure_reason() +// can be queried for an extremely brief, end-user unfriendly explanation +// of why the load failed. Define STBI_NO_FAILURE_STRINGS to avoid +// compiling these strings at all, and STBI_FAILURE_USERMSG to get slightly +// more user-friendly ones. +// +// Paletted PNG, BMP, GIF, and PIC images are automatically depalettized. +// +// =========================================================================== +// +// Philosophy +// +// stb libraries are designed with the following priorities: +// +// 1. easy to use +// 2. easy to maintain +// 3. good performance +// +// Sometimes I let "good performance" creep up in priority over "easy to maintain", +// and for best performance I may provide less-easy-to-use APIs that give higher +// performance, in addition to the easy to use ones. Nevertheless, it's important +// to keep in mind that from the standpoint of you, a client of this library, +// all you care about is #1 and #3, and stb libraries do not emphasize #3 above all. +// +// Some secondary priorities arise directly from the first two, some of which +// make more explicit reasons why performance can't be emphasized. +// +// - Portable ("ease of use") +// - Small footprint ("easy to maintain") +// - No dependencies ("ease of use") +// +// =========================================================================== +// +// I/O callbacks +// +// I/O callbacks allow you to read from arbitrary sources, like packaged +// files or some other source. Data read from callbacks are processed +// through a small internal buffer (currently 128 bytes) to try to reduce +// overhead. +// +// The three functions you must define are "read" (reads some bytes of data), +// "skip" (skips some bytes of data), "eof" (reports if the stream is at the end). +// +// =========================================================================== +// +// SIMD support +// +// The JPEG decoder will try to automatically use SIMD kernels on x86 when +// supported by the compiler. For ARM Neon support, you must explicitly +// request it. +// +// (The old do-it-yourself SIMD API is no longer supported in the current +// code.) +// +// On x86, SSE2 will automatically be used when available based on a run-time +// test; if not, the generic C versions are used as a fall-back. On ARM targets, +// the typical path is to have separate builds for NEON and non-NEON devices +// (at least this is true for iOS and Android). Therefore, the NEON support is +// toggled by a build flag: define STBI_NEON to get NEON loops. +// +// The output of the JPEG decoder is slightly different from versions where +// SIMD support was introduced (that is, for versions before 1.49). The +// difference is only +-1 in the 8-bit RGB channels, and only on a small +// fraction of pixels. You can force the pre-1.49 behavior by defining +// STBI_JPEG_OLD, but this will disable some of the SIMD decoding path +// and hence cost some performance. +// +// If for some reason you do not want to use any of SIMD code, or if +// you have issues compiling it, you can disable it entirely by +// defining STBI_NO_SIMD. +// +// =========================================================================== +// +// HDR image support (disable by defining STBI_NO_HDR) +// +// stb_image now supports loading HDR images in general, and currently +// the Radiance .HDR file format, although the support is provided +// generically. You can still load any file through the existing interface; +// if you attempt to load an HDR file, it will be automatically remapped to +// LDR, assuming gamma 2.2 and an arbitrary scale factor defaulting to 1; +// both of these constants can be reconfigured through this interface: +// +// stbi_hdr_to_ldr_gamma(2.2f); +// stbi_hdr_to_ldr_scale(1.0f); +// +// (note, do not use _inverse_ constants; stbi_image will invert them +// appropriately). +// +// Additionally, there is a new, parallel interface for loading files as +// (linear) floats to preserve the full dynamic range: +// +// float *data = stbi_loadf(filename, &x, &y, &n, 0); +// +// If you load LDR images through this interface, those images will +// be promoted to floating point values, run through the inverse of +// constants corresponding to the above: +// +// stbi_ldr_to_hdr_scale(1.0f); +// stbi_ldr_to_hdr_gamma(2.2f); +// +// Finally, given a filename (or an open file or memory block--see header +// file for details) containing image data, you can query for the "most +// appropriate" interface to use (that is, whether the image is HDR or +// not), using: +// +// stbi_is_hdr(char *filename); +// +// =========================================================================== +// +// iPhone PNG support: +// +// By default we convert iphone-formatted PNGs back to RGB, even though +// they are internally encoded differently. You can disable this conversion +// by by calling stbi_convert_iphone_png_to_rgb(0), in which case +// you will always just get the native iphone "format" through (which +// is BGR stored in RGB). +// +// Call stbi_set_unpremultiply_on_load(1) as well to force a divide per +// pixel to remove any premultiplied alpha *only* if the image file explicitly +// says there's premultiplied data (currently only happens in iPhone images, +// and only if iPhone convert-to-rgb processing is on). +// + + +#ifndef STBI_NO_STDIO +#include +#endif // STBI_NO_STDIO + +#define STBI_VERSION 1 + +enum +{ + STBI_default = 0, // only used for req_comp + + STBI_grey = 1, + STBI_grey_alpha = 2, + STBI_rgb = 3, + STBI_rgb_alpha = 4 +}; + +typedef unsigned char stbi_uc; + +#ifdef __cplusplus +extern "C" { +#endif + +#ifdef STB_IMAGE_STATIC +#define STBIDEF static +#else +#define STBIDEF extern +#endif + +////////////////////////////////////////////////////////////////////////////// +// +// PRIMARY API - works on images of any type +// + +// +// load image by filename, open file, or memory buffer +// + +typedef struct +{ + int (*read) (void *user,char *data,int size); // fill 'data' with 'size' bytes. return number of bytes actually read + void (*skip) (void *user,int n); // skip the next 'n' bytes, or 'unget' the last -n bytes if negative + int (*eof) (void *user); // returns nonzero if we are at end of file/data +} stbi_io_callbacks; + +STBIDEF stbi_uc *stbi_load (char const *filename, int *x, int *y, int *comp, int req_comp); +STBIDEF stbi_uc *stbi_load_from_memory (stbi_uc const *buffer, int len , int *x, int *y, int *comp, int req_comp); +STBIDEF stbi_uc *stbi_load_from_callbacks(stbi_io_callbacks const *clbk , void *user, int *x, int *y, int *comp, int req_comp); + +#ifndef STBI_NO_STDIO +STBIDEF stbi_uc *stbi_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); +// for stbi_load_from_file, file pointer is left pointing immediately after image +#endif + +#ifndef STBI_NO_LINEAR + STBIDEF float *stbi_loadf (char const *filename, int *x, int *y, int *comp, int req_comp); + STBIDEF float *stbi_loadf_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp); + STBIDEF float *stbi_loadf_from_callbacks (stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp); + + #ifndef STBI_NO_STDIO + STBIDEF float *stbi_loadf_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); + #endif +#endif + +#ifndef STBI_NO_HDR + STBIDEF void stbi_hdr_to_ldr_gamma(float gamma); + STBIDEF void stbi_hdr_to_ldr_scale(float scale); +#endif // STBI_NO_HDR + +#ifndef STBI_NO_LINEAR + STBIDEF void stbi_ldr_to_hdr_gamma(float gamma); + STBIDEF void stbi_ldr_to_hdr_scale(float scale); +#endif // STBI_NO_LINEAR + +// stbi_is_hdr is always defined, but always returns false if STBI_NO_HDR +STBIDEF int stbi_is_hdr_from_callbacks(stbi_io_callbacks const *clbk, void *user); +STBIDEF int stbi_is_hdr_from_memory(stbi_uc const *buffer, int len); +#ifndef STBI_NO_STDIO +STBIDEF int stbi_is_hdr (char const *filename); +STBIDEF int stbi_is_hdr_from_file(FILE *f); +#endif // STBI_NO_STDIO + + +// get a VERY brief reason for failure +// NOT THREADSAFE +STBIDEF const char *stbi_failure_reason (void); + +// free the loaded image -- this is just free() +STBIDEF void stbi_image_free (void *retval_from_stbi_load); + +// get image dimensions & components without fully decoding +STBIDEF int stbi_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp); +STBIDEF int stbi_info_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp); + +#ifndef STBI_NO_STDIO +STBIDEF int stbi_info (char const *filename, int *x, int *y, int *comp); +STBIDEF int stbi_info_from_file (FILE *f, int *x, int *y, int *comp); + +#endif + + + +// for image formats that explicitly notate that they have premultiplied alpha, +// we just return the colors as stored in the file. set this flag to force +// unpremultiplication. results are undefined if the unpremultiply overflow. +STBIDEF void stbi_set_unpremultiply_on_load(int flag_true_if_should_unpremultiply); + +// indicate whether we should process iphone images back to canonical format, +// or just pass them through "as-is" +STBIDEF void stbi_convert_iphone_png_to_rgb(int flag_true_if_should_convert); + +// flip the image vertically, so the first pixel in the output array is the bottom left +STBIDEF void stbi_set_flip_vertically_on_load(int flag_true_if_should_flip); + +// ZLIB client - used by PNG, available for other purposes + +STBIDEF char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen); +STBIDEF char *stbi_zlib_decode_malloc_guesssize_headerflag(const char *buffer, int len, int initial_size, int *outlen, int parse_header); +STBIDEF char *stbi_zlib_decode_malloc(const char *buffer, int len, int *outlen); +STBIDEF int stbi_zlib_decode_buffer(char *obuffer, int olen, const char *ibuffer, int ilen); + +STBIDEF char *stbi_zlib_decode_noheader_malloc(const char *buffer, int len, int *outlen); +STBIDEF int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen); + + +#ifdef __cplusplus +} +#endif + +// +// +//// end header file ///////////////////////////////////////////////////// +#endif // STBI_INCLUDE_STB_IMAGE_H + +#ifdef STB_IMAGE_IMPLEMENTATION + +#if defined(STBI_ONLY_JPEG) || defined(STBI_ONLY_PNG) || defined(STBI_ONLY_BMP) \ + || defined(STBI_ONLY_TGA) || defined(STBI_ONLY_GIF) || defined(STBI_ONLY_PSD) \ + || defined(STBI_ONLY_HDR) || defined(STBI_ONLY_PIC) || defined(STBI_ONLY_PNM) \ + || defined(STBI_ONLY_ZLIB) + #ifndef STBI_ONLY_JPEG + #define STBI_NO_JPEG + #endif + #ifndef STBI_ONLY_PNG + #define STBI_NO_PNG + #endif + #ifndef STBI_ONLY_BMP + #define STBI_NO_BMP + #endif + #ifndef STBI_ONLY_PSD + #define STBI_NO_PSD + #endif + #ifndef STBI_ONLY_TGA + #define STBI_NO_TGA + #endif + #ifndef STBI_ONLY_GIF + #define STBI_NO_GIF + #endif + #ifndef STBI_ONLY_HDR + #define STBI_NO_HDR + #endif + #ifndef STBI_ONLY_PIC + #define STBI_NO_PIC + #endif + #ifndef STBI_ONLY_PNM + #define STBI_NO_PNM + #endif +#endif + +#if defined(STBI_NO_PNG) && !defined(STBI_SUPPORT_ZLIB) && !defined(STBI_NO_ZLIB) +#define STBI_NO_ZLIB +#endif + + +#include +#include // ptrdiff_t on osx +#include +#include + +#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR) +#include // ldexp +#endif + +#ifndef STBI_NO_STDIO +#include +#endif + +#ifndef STBI_ASSERT +#include +#define STBI_ASSERT(x) assert(x) +#endif + + +#ifndef _MSC_VER + #ifdef __cplusplus + #define stbi_inline inline + #else + #define stbi_inline + #endif +#else + #define stbi_inline __forceinline +#endif + + +#ifdef _MSC_VER +typedef unsigned short stbi__uint16; +typedef signed short stbi__int16; +typedef unsigned int stbi__uint32; +typedef signed int stbi__int32; +#else +#include +typedef uint16_t stbi__uint16; +typedef int16_t stbi__int16; +typedef uint32_t stbi__uint32; +typedef int32_t stbi__int32; +#endif + +// should produce compiler error if size is wrong +typedef unsigned char validate_uint32[sizeof(stbi__uint32)==4 ? 1 : -1]; + +#ifdef _MSC_VER +#define STBI_NOTUSED(v) (void)(v) +#else +#define STBI_NOTUSED(v) (void)sizeof(v) +#endif + +#ifdef _MSC_VER +#define STBI_HAS_LROTL +#endif + +#ifdef STBI_HAS_LROTL + #define stbi_lrot(x,y) _lrotl(x,y) +#else + #define stbi_lrot(x,y) (((x) << (y)) | ((x) >> (32 - (y)))) +#endif + +#if defined(STBI_MALLOC) && defined(STBI_FREE) && (defined(STBI_REALLOC) || defined(STBI_REALLOC_SIZED)) +// ok +#elif !defined(STBI_MALLOC) && !defined(STBI_FREE) && !defined(STBI_REALLOC) && !defined(STBI_REALLOC_SIZED) +// ok +#else +#error "Must define all or none of STBI_MALLOC, STBI_FREE, and STBI_REALLOC (or STBI_REALLOC_SIZED)." +#endif + +#ifndef STBI_MALLOC +#define STBI_MALLOC(sz) malloc(sz) +#define STBI_REALLOC(p,newsz) realloc(p,newsz) +#define STBI_FREE(p) free(p) +#endif + +#ifndef STBI_REALLOC_SIZED +#define STBI_REALLOC_SIZED(p,oldsz,newsz) STBI_REALLOC(p,newsz) +#endif + +// x86/x64 detection +#if defined(__x86_64__) || defined(_M_X64) +#define STBI__X64_TARGET +#elif defined(__i386) || defined(_M_IX86) +#define STBI__X86_TARGET +#endif + +#if defined(__GNUC__) && (defined(STBI__X86_TARGET) || defined(STBI__X64_TARGET)) && !defined(__SSE2__) && !defined(STBI_NO_SIMD) +// NOTE: not clear do we actually need this for the 64-bit path? +// gcc doesn't support sse2 intrinsics unless you compile with -msse2, +// (but compiling with -msse2 allows the compiler to use SSE2 everywhere; +// this is just broken and gcc are jerks for not fixing it properly +// http://www.virtualdub.org/blog/pivot/entry.php?id=363 ) +#define STBI_NO_SIMD +#endif + +#if defined(__MINGW32__) && defined(STBI__X86_TARGET) && !defined(STBI_MINGW_ENABLE_SSE2) && !defined(STBI_NO_SIMD) +// Note that __MINGW32__ doesn't actually mean 32-bit, so we have to avoid STBI__X64_TARGET +// +// 32-bit MinGW wants ESP to be 16-byte aligned, but this is not in the +// Windows ABI and VC++ as well as Windows DLLs don't maintain that invariant. +// As a result, enabling SSE2 on 32-bit MinGW is dangerous when not +// simultaneously enabling "-mstackrealign". +// +// See https://github.com/nothings/stb/issues/81 for more information. +// +// So default to no SSE2 on 32-bit MinGW. If you've read this far and added +// -mstackrealign to your build settings, feel free to #define STBI_MINGW_ENABLE_SSE2. +#define STBI_NO_SIMD +#endif + +#if !defined(STBI_NO_SIMD) && defined(STBI__X86_TARGET) +#define STBI_SSE2 +#include + +#ifdef _MSC_VER + +#if _MSC_VER >= 1400 // not VC6 +#include // __cpuid +static int stbi__cpuid3(void) +{ + int info[4]; + __cpuid(info,1); + return info[3]; +} +#else +static int stbi__cpuid3(void) +{ + int res; + __asm { + mov eax,1 + cpuid + mov res,edx + } + return res; +} +#endif + +#define STBI_SIMD_ALIGN(type, name) __declspec(align(16)) type name + +static int stbi__sse2_available() +{ + int info3 = stbi__cpuid3(); + return ((info3 >> 26) & 1) != 0; +} +#else // assume GCC-style if not VC++ +#define STBI_SIMD_ALIGN(type, name) type name __attribute__((aligned(16))) + +static int stbi__sse2_available() +{ +#if defined(__GNUC__) && (__GNUC__ * 100 + __GNUC_MINOR__) >= 408 // GCC 4.8 or later + // GCC 4.8+ has a nice way to do this + return __builtin_cpu_supports("sse2"); +#else + // portable way to do this, preferably without using GCC inline ASM? + // just bail for now. + return 0; +#endif +} +#endif +#endif + +// ARM NEON +#if defined(STBI_NO_SIMD) && defined(STBI_NEON) +#undef STBI_NEON +#endif + +#ifdef STBI_NEON +#include +// assume GCC or Clang on ARM targets +#define STBI_SIMD_ALIGN(type, name) type name __attribute__((aligned(16))) +#endif + +#ifndef STBI_SIMD_ALIGN +#define STBI_SIMD_ALIGN(type, name) type name +#endif + +/////////////////////////////////////////////// +// +// stbi__context struct and start_xxx functions + +// stbi__context structure is our basic context used by all images, so it +// contains all the IO context, plus some basic image information +typedef struct +{ + stbi__uint32 img_x, img_y; + int img_n, img_out_n; + + stbi_io_callbacks io; + void *io_user_data; + + int read_from_callbacks; + int buflen; + stbi_uc buffer_start[128]; + + stbi_uc *img_buffer, *img_buffer_end; + stbi_uc *img_buffer_original, *img_buffer_original_end; +} stbi__context; + + +static void stbi__refill_buffer(stbi__context *s); + +// initialize a memory-decode context +static void stbi__start_mem(stbi__context *s, stbi_uc const *buffer, int len) +{ + s->io.read = NULL; + s->read_from_callbacks = 0; + s->img_buffer = s->img_buffer_original = (stbi_uc *) buffer; + s->img_buffer_end = s->img_buffer_original_end = (stbi_uc *) buffer+len; +} + +// initialize a callback-based context +static void stbi__start_callbacks(stbi__context *s, stbi_io_callbacks *c, void *user) +{ + s->io = *c; + s->io_user_data = user; + s->buflen = sizeof(s->buffer_start); + s->read_from_callbacks = 1; + s->img_buffer_original = s->buffer_start; + stbi__refill_buffer(s); + s->img_buffer_original_end = s->img_buffer_end; +} + +#ifndef STBI_NO_STDIO + +static int stbi__stdio_read(void *user, char *data, int size) +{ + return (int) fread(data,1,size,(FILE*) user); +} + +static void stbi__stdio_skip(void *user, int n) +{ + fseek((FILE*) user, n, SEEK_CUR); +} + +static int stbi__stdio_eof(void *user) +{ + return feof((FILE*) user); +} + +static stbi_io_callbacks stbi__stdio_callbacks = +{ + stbi__stdio_read, + stbi__stdio_skip, + stbi__stdio_eof, +}; + +static void stbi__start_file(stbi__context *s, FILE *f) +{ + stbi__start_callbacks(s, &stbi__stdio_callbacks, (void *) f); +} + +//static void stop_file(stbi__context *s) { } + +#endif // !STBI_NO_STDIO + +static void stbi__rewind(stbi__context *s) +{ + // conceptually rewind SHOULD rewind to the beginning of the stream, + // but we just rewind to the beginning of the initial buffer, because + // we only use it after doing 'test', which only ever looks at at most 92 bytes + s->img_buffer = s->img_buffer_original; + s->img_buffer_end = s->img_buffer_original_end; +} + +#ifndef STBI_NO_JPEG +static int stbi__jpeg_test(stbi__context *s); +static stbi_uc *stbi__jpeg_load(stbi__context *s, int *x, int *y, int *comp, int req_comp); +static int stbi__jpeg_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_PNG +static int stbi__png_test(stbi__context *s); +static stbi_uc *stbi__png_load(stbi__context *s, int *x, int *y, int *comp, int req_comp); +static int stbi__png_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_BMP +static int stbi__bmp_test(stbi__context *s); +static stbi_uc *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int req_comp); +static int stbi__bmp_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_TGA +static int stbi__tga_test(stbi__context *s); +static stbi_uc *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int req_comp); +static int stbi__tga_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_PSD +static int stbi__psd_test(stbi__context *s); +static stbi_uc *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int req_comp); +static int stbi__psd_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_HDR +static int stbi__hdr_test(stbi__context *s); +static float *stbi__hdr_load(stbi__context *s, int *x, int *y, int *comp, int req_comp); +static int stbi__hdr_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_PIC +static int stbi__pic_test(stbi__context *s); +static stbi_uc *stbi__pic_load(stbi__context *s, int *x, int *y, int *comp, int req_comp); +static int stbi__pic_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_GIF +static int stbi__gif_test(stbi__context *s); +static stbi_uc *stbi__gif_load(stbi__context *s, int *x, int *y, int *comp, int req_comp); +static int stbi__gif_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_PNM +static int stbi__pnm_test(stbi__context *s); +static stbi_uc *stbi__pnm_load(stbi__context *s, int *x, int *y, int *comp, int req_comp); +static int stbi__pnm_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +// this is not threadsafe +static const char *stbi__g_failure_reason; + +STBIDEF const char *stbi_failure_reason(void) +{ + return stbi__g_failure_reason; +} + +static int stbi__err(const char *str) +{ + stbi__g_failure_reason = str; + return 0; +} + +static void *stbi__malloc(size_t size) +{ + return STBI_MALLOC(size); +} + +// stbi__err - error +// stbi__errpf - error returning pointer to float +// stbi__errpuc - error returning pointer to unsigned char + +#ifdef STBI_NO_FAILURE_STRINGS + #define stbi__err(x,y) 0 +#elif defined(STBI_FAILURE_USERMSG) + #define stbi__err(x,y) stbi__err(y) +#else + #define stbi__err(x,y) stbi__err(x) +#endif + +#define stbi__errpf(x,y) ((float *)(size_t) (stbi__err(x,y)?NULL:NULL)) +#define stbi__errpuc(x,y) ((unsigned char *)(size_t) (stbi__err(x,y)?NULL:NULL)) + +STBIDEF void stbi_image_free(void *retval_from_stbi_load) +{ + STBI_FREE(retval_from_stbi_load); +} + +#ifndef STBI_NO_LINEAR +static float *stbi__ldr_to_hdr(stbi_uc *data, int x, int y, int comp); +#endif + +#ifndef STBI_NO_HDR +static stbi_uc *stbi__hdr_to_ldr(float *data, int x, int y, int comp); +#endif + +static int stbi__vertically_flip_on_load = 0; + +STBIDEF void stbi_set_flip_vertically_on_load(int flag_true_if_should_flip) +{ + stbi__vertically_flip_on_load = flag_true_if_should_flip; +} + +static unsigned char *stbi__load_main(stbi__context *s, int *x, int *y, int *comp, int req_comp) +{ + #ifndef STBI_NO_JPEG + if (stbi__jpeg_test(s)) return stbi__jpeg_load(s,x,y,comp,req_comp); + #endif + #ifndef STBI_NO_PNG + if (stbi__png_test(s)) return stbi__png_load(s,x,y,comp,req_comp); + #endif + #ifndef STBI_NO_BMP + if (stbi__bmp_test(s)) return stbi__bmp_load(s,x,y,comp,req_comp); + #endif + #ifndef STBI_NO_GIF + if (stbi__gif_test(s)) return stbi__gif_load(s,x,y,comp,req_comp); + #endif + #ifndef STBI_NO_PSD + if (stbi__psd_test(s)) return stbi__psd_load(s,x,y,comp,req_comp); + #endif + #ifndef STBI_NO_PIC + if (stbi__pic_test(s)) return stbi__pic_load(s,x,y,comp,req_comp); + #endif + #ifndef STBI_NO_PNM + if (stbi__pnm_test(s)) return stbi__pnm_load(s,x,y,comp,req_comp); + #endif + + #ifndef STBI_NO_HDR + if (stbi__hdr_test(s)) { + float *hdr = stbi__hdr_load(s, x,y,comp,req_comp); + return stbi__hdr_to_ldr(hdr, *x, *y, req_comp ? req_comp : *comp); + } + #endif + + #ifndef STBI_NO_TGA + // test tga last because it's a crappy test! + if (stbi__tga_test(s)) + return stbi__tga_load(s,x,y,comp,req_comp); + #endif + + return stbi__errpuc("unknown image type", "Image not of any known type, or corrupt"); +} + +static unsigned char *stbi__load_flip(stbi__context *s, int *x, int *y, int *comp, int req_comp) +{ + unsigned char *result = stbi__load_main(s, x, y, comp, req_comp); + + if (stbi__vertically_flip_on_load && result != NULL) { + int w = *x, h = *y; + int depth = req_comp ? req_comp : *comp; + int row,col,z; + stbi_uc temp; + + // @OPTIMIZE: use a bigger temp buffer and memcpy multiple pixels at once + for (row = 0; row < (h>>1); row++) { + for (col = 0; col < w; col++) { + for (z = 0; z < depth; z++) { + temp = result[(row * w + col) * depth + z]; + result[(row * w + col) * depth + z] = result[((h - row - 1) * w + col) * depth + z]; + result[((h - row - 1) * w + col) * depth + z] = temp; + } + } + } + } + + return result; +} + +#ifndef STBI_NO_HDR +static void stbi__float_postprocess(float *result, int *x, int *y, int *comp, int req_comp) +{ + if (stbi__vertically_flip_on_load && result != NULL) { + int w = *x, h = *y; + int depth = req_comp ? req_comp : *comp; + int row,col,z; + float temp; + + // @OPTIMIZE: use a bigger temp buffer and memcpy multiple pixels at once + for (row = 0; row < (h>>1); row++) { + for (col = 0; col < w; col++) { + for (z = 0; z < depth; z++) { + temp = result[(row * w + col) * depth + z]; + result[(row * w + col) * depth + z] = result[((h - row - 1) * w + col) * depth + z]; + result[((h - row - 1) * w + col) * depth + z] = temp; + } + } + } + } +} +#endif + +#ifndef STBI_NO_STDIO + +static FILE *stbi__fopen(char const *filename, char const *mode) +{ + FILE *f; +#if defined(_MSC_VER) && _MSC_VER >= 1400 + if (0 != fopen_s(&f, filename, mode)) + f=0; +#else + f = fopen(filename, mode); +#endif + return f; +} + + +STBIDEF stbi_uc *stbi_load(char const *filename, int *x, int *y, int *comp, int req_comp) +{ + FILE *f = stbi__fopen(filename, "rb"); + unsigned char *result; + if (!f) return stbi__errpuc("can't fopen", "Unable to open file"); + result = stbi_load_from_file(f,x,y,comp,req_comp); + fclose(f); + return result; +} + +STBIDEF stbi_uc *stbi_load_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + unsigned char *result; + stbi__context s; + stbi__start_file(&s,f); + result = stbi__load_flip(&s,x,y,comp,req_comp); + if (result) { + // need to 'unget' all the characters in the IO buffer + fseek(f, - (int) (s.img_buffer_end - s.img_buffer), SEEK_CUR); + } + return result; +} +#endif //!STBI_NO_STDIO + +STBIDEF stbi_uc *stbi_load_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__load_flip(&s,x,y,comp,req_comp); +} + +STBIDEF stbi_uc *stbi_load_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp) +{ + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user); + return stbi__load_flip(&s,x,y,comp,req_comp); +} + +#ifndef STBI_NO_LINEAR +static float *stbi__loadf_main(stbi__context *s, int *x, int *y, int *comp, int req_comp) +{ + unsigned char *data; + #ifndef STBI_NO_HDR + if (stbi__hdr_test(s)) { + float *hdr_data = stbi__hdr_load(s,x,y,comp,req_comp); + if (hdr_data) + stbi__float_postprocess(hdr_data,x,y,comp,req_comp); + return hdr_data; + } + #endif + data = stbi__load_flip(s, x, y, comp, req_comp); + if (data) + return stbi__ldr_to_hdr(data, *x, *y, req_comp ? req_comp : *comp); + return stbi__errpf("unknown image type", "Image not of any known type, or corrupt"); +} + +STBIDEF float *stbi_loadf_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__loadf_main(&s,x,y,comp,req_comp); +} + +STBIDEF float *stbi_loadf_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp) +{ + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user); + return stbi__loadf_main(&s,x,y,comp,req_comp); +} + +#ifndef STBI_NO_STDIO +STBIDEF float *stbi_loadf(char const *filename, int *x, int *y, int *comp, int req_comp) +{ + float *result; + FILE *f = stbi__fopen(filename, "rb"); + if (!f) return stbi__errpf("can't fopen", "Unable to open file"); + result = stbi_loadf_from_file(f,x,y,comp,req_comp); + fclose(f); + return result; +} + +STBIDEF float *stbi_loadf_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + stbi__context s; + stbi__start_file(&s,f); + return stbi__loadf_main(&s,x,y,comp,req_comp); +} +#endif // !STBI_NO_STDIO + +#endif // !STBI_NO_LINEAR + +// these is-hdr-or-not is defined independent of whether STBI_NO_LINEAR is +// defined, for API simplicity; if STBI_NO_LINEAR is defined, it always +// reports false! + +STBIDEF int stbi_is_hdr_from_memory(stbi_uc const *buffer, int len) +{ + #ifndef STBI_NO_HDR + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__hdr_test(&s); + #else + STBI_NOTUSED(buffer); + STBI_NOTUSED(len); + return 0; + #endif +} + +#ifndef STBI_NO_STDIO +STBIDEF int stbi_is_hdr (char const *filename) +{ + FILE *f = stbi__fopen(filename, "rb"); + int result=0; + if (f) { + result = stbi_is_hdr_from_file(f); + fclose(f); + } + return result; +} + +STBIDEF int stbi_is_hdr_from_file(FILE *f) +{ + #ifndef STBI_NO_HDR + stbi__context s; + stbi__start_file(&s,f); + return stbi__hdr_test(&s); + #else + STBI_NOTUSED(f); + return 0; + #endif +} +#endif // !STBI_NO_STDIO + +STBIDEF int stbi_is_hdr_from_callbacks(stbi_io_callbacks const *clbk, void *user) +{ + #ifndef STBI_NO_HDR + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user); + return stbi__hdr_test(&s); + #else + STBI_NOTUSED(clbk); + STBI_NOTUSED(user); + return 0; + #endif +} + +#ifndef STBI_NO_LINEAR +static float stbi__l2h_gamma=2.2f, stbi__l2h_scale=1.0f; + +STBIDEF void stbi_ldr_to_hdr_gamma(float gamma) { stbi__l2h_gamma = gamma; } +STBIDEF void stbi_ldr_to_hdr_scale(float scale) { stbi__l2h_scale = scale; } +#endif + +static float stbi__h2l_gamma_i=1.0f/2.2f, stbi__h2l_scale_i=1.0f; + +STBIDEF void stbi_hdr_to_ldr_gamma(float gamma) { stbi__h2l_gamma_i = 1/gamma; } +STBIDEF void stbi_hdr_to_ldr_scale(float scale) { stbi__h2l_scale_i = 1/scale; } + + +////////////////////////////////////////////////////////////////////////////// +// +// Common code used by all image loaders +// + +enum +{ + STBI__SCAN_load=0, + STBI__SCAN_type, + STBI__SCAN_header +}; + +static void stbi__refill_buffer(stbi__context *s) +{ + int n = (s->io.read)(s->io_user_data,(char*)s->buffer_start,s->buflen); + if (n == 0) { + // at end of file, treat same as if from memory, but need to handle case + // where s->img_buffer isn't pointing to safe memory, e.g. 0-byte file + s->read_from_callbacks = 0; + s->img_buffer = s->buffer_start; + s->img_buffer_end = s->buffer_start+1; + *s->img_buffer = 0; + } else { + s->img_buffer = s->buffer_start; + s->img_buffer_end = s->buffer_start + n; + } +} + +stbi_inline static stbi_uc stbi__get8(stbi__context *s) +{ + if (s->img_buffer < s->img_buffer_end) + return *s->img_buffer++; + if (s->read_from_callbacks) { + stbi__refill_buffer(s); + return *s->img_buffer++; + } + return 0; +} + +stbi_inline static int stbi__at_eof(stbi__context *s) +{ + if (s->io.read) { + if (!(s->io.eof)(s->io_user_data)) return 0; + // if feof() is true, check if buffer = end + // special case: we've only got the special 0 character at the end + if (s->read_from_callbacks == 0) return 1; + } + + return s->img_buffer >= s->img_buffer_end; +} + +static void stbi__skip(stbi__context *s, int n) +{ + if (n < 0) { + s->img_buffer = s->img_buffer_end; + return; + } + if (s->io.read) { + int blen = (int) (s->img_buffer_end - s->img_buffer); + if (blen < n) { + s->img_buffer = s->img_buffer_end; + (s->io.skip)(s->io_user_data, n - blen); + return; + } + } + s->img_buffer += n; +} + +static int stbi__getn(stbi__context *s, stbi_uc *buffer, int n) +{ + if (s->io.read) { + int blen = (int) (s->img_buffer_end - s->img_buffer); + if (blen < n) { + int res, count; + + memcpy(buffer, s->img_buffer, blen); + + count = (s->io.read)(s->io_user_data, (char*) buffer + blen, n - blen); + res = (count == (n-blen)); + s->img_buffer = s->img_buffer_end; + return res; + } + } + + if (s->img_buffer+n <= s->img_buffer_end) { + memcpy(buffer, s->img_buffer, n); + s->img_buffer += n; + return 1; + } else + return 0; +} + +static int stbi__get16be(stbi__context *s) +{ + int z = stbi__get8(s); + return (z << 8) + stbi__get8(s); +} + +static stbi__uint32 stbi__get32be(stbi__context *s) +{ + stbi__uint32 z = stbi__get16be(s); + return (z << 16) + stbi__get16be(s); +} + +#if defined(STBI_NO_BMP) && defined(STBI_NO_TGA) && defined(STBI_NO_GIF) +// nothing +#else +static int stbi__get16le(stbi__context *s) +{ + int z = stbi__get8(s); + return z + (stbi__get8(s) << 8); +} +#endif + +#ifndef STBI_NO_BMP +static stbi__uint32 stbi__get32le(stbi__context *s) +{ + stbi__uint32 z = stbi__get16le(s); + return z + (stbi__get16le(s) << 16); +} +#endif + +#define STBI__BYTECAST(x) ((stbi_uc) ((x) & 255)) // truncate int to byte without warnings + + +////////////////////////////////////////////////////////////////////////////// +// +// generic converter from built-in img_n to req_comp +// individual types do this automatically as much as possible (e.g. jpeg +// does all cases internally since it needs to colorspace convert anyway, +// and it never has alpha, so very few cases ). png can automatically +// interleave an alpha=255 channel, but falls back to this for other cases +// +// assume data buffer is malloced, so malloc a new one and free that one +// only failure mode is malloc failing + +static stbi_uc stbi__compute_y(int r, int g, int b) +{ + return (stbi_uc) (((r*77) + (g*150) + (29*b)) >> 8); +} + +static unsigned char *stbi__convert_format(unsigned char *data, int img_n, int req_comp, unsigned int x, unsigned int y) +{ + int i,j; + unsigned char *good; + + if (req_comp == img_n) return data; + STBI_ASSERT(req_comp >= 1 && req_comp <= 4); + + good = (unsigned char *) stbi__malloc(req_comp * x * y); + if (good == NULL) { + STBI_FREE(data); + return stbi__errpuc("outofmem", "Out of memory"); + } + + for (j=0; j < (int) y; ++j) { + unsigned char *src = data + j * x * img_n ; + unsigned char *dest = good + j * x * req_comp; + + #define COMBO(a,b) ((a)*8+(b)) + #define CASE(a,b) case COMBO(a,b): for(i=x-1; i >= 0; --i, src += a, dest += b) + // convert source image with img_n components to one with req_comp components; + // avoid switch per pixel, so use switch per scanline and massive macros + switch (COMBO(img_n, req_comp)) { + CASE(1,2) dest[0]=src[0], dest[1]=255; break; + CASE(1,3) dest[0]=dest[1]=dest[2]=src[0]; break; + CASE(1,4) dest[0]=dest[1]=dest[2]=src[0], dest[3]=255; break; + CASE(2,1) dest[0]=src[0]; break; + CASE(2,3) dest[0]=dest[1]=dest[2]=src[0]; break; + CASE(2,4) dest[0]=dest[1]=dest[2]=src[0], dest[3]=src[1]; break; + CASE(3,4) dest[0]=src[0],dest[1]=src[1],dest[2]=src[2],dest[3]=255; break; + CASE(3,1) dest[0]=stbi__compute_y(src[0],src[1],src[2]); break; + CASE(3,2) dest[0]=stbi__compute_y(src[0],src[1],src[2]), dest[1] = 255; break; + CASE(4,1) dest[0]=stbi__compute_y(src[0],src[1],src[2]); break; + CASE(4,2) dest[0]=stbi__compute_y(src[0],src[1],src[2]), dest[1] = src[3]; break; + CASE(4,3) dest[0]=src[0],dest[1]=src[1],dest[2]=src[2]; break; + default: STBI_ASSERT(0); + } + #undef CASE + } + + STBI_FREE(data); + return good; +} + +#ifndef STBI_NO_LINEAR +static float *stbi__ldr_to_hdr(stbi_uc *data, int x, int y, int comp) +{ + int i,k,n; + float *output = (float *) stbi__malloc(x * y * comp * sizeof(float)); + if (output == NULL) { STBI_FREE(data); return stbi__errpf("outofmem", "Out of memory"); } + // compute number of non-alpha components + if (comp & 1) n = comp; else n = comp-1; + for (i=0; i < x*y; ++i) { + for (k=0; k < n; ++k) { + output[i*comp + k] = (float) (pow(data[i*comp+k]/255.0f, stbi__l2h_gamma) * stbi__l2h_scale); + } + if (k < comp) output[i*comp + k] = data[i*comp+k]/255.0f; + } + STBI_FREE(data); + return output; +} +#endif + +#ifndef STBI_NO_HDR +#define stbi__float2int(x) ((int) (x)) +static stbi_uc *stbi__hdr_to_ldr(float *data, int x, int y, int comp) +{ + int i,k,n; + stbi_uc *output = (stbi_uc *) stbi__malloc(x * y * comp); + if (output == NULL) { STBI_FREE(data); return stbi__errpuc("outofmem", "Out of memory"); } + // compute number of non-alpha components + if (comp & 1) n = comp; else n = comp-1; + for (i=0; i < x*y; ++i) { + for (k=0; k < n; ++k) { + float z = (float) pow(data[i*comp+k]*stbi__h2l_scale_i, stbi__h2l_gamma_i) * 255 + 0.5f; + if (z < 0) z = 0; + if (z > 255) z = 255; + output[i*comp + k] = (stbi_uc) stbi__float2int(z); + } + if (k < comp) { + float z = data[i*comp+k] * 255 + 0.5f; + if (z < 0) z = 0; + if (z > 255) z = 255; + output[i*comp + k] = (stbi_uc) stbi__float2int(z); + } + } + STBI_FREE(data); + return output; +} +#endif + +////////////////////////////////////////////////////////////////////////////// +// +// "baseline" JPEG/JFIF decoder +// +// simple implementation +// - doesn't support delayed output of y-dimension +// - simple interface (only one output format: 8-bit interleaved RGB) +// - doesn't try to recover corrupt jpegs +// - doesn't allow partial loading, loading multiple at once +// - still fast on x86 (copying globals into locals doesn't help x86) +// - allocates lots of intermediate memory (full size of all components) +// - non-interleaved case requires this anyway +// - allows good upsampling (see next) +// high-quality +// - upsampled channels are bilinearly interpolated, even across blocks +// - quality integer IDCT derived from IJG's 'slow' +// performance +// - fast huffman; reasonable integer IDCT +// - some SIMD kernels for common paths on targets with SSE2/NEON +// - uses a lot of intermediate memory, could cache poorly + +#ifndef STBI_NO_JPEG + +// huffman decoding acceleration +#define FAST_BITS 9 // larger handles more cases; smaller stomps less cache + +typedef struct +{ + stbi_uc fast[1 << FAST_BITS]; + // weirdly, repacking this into AoS is a 10% speed loss, instead of a win + stbi__uint16 code[256]; + stbi_uc values[256]; + stbi_uc size[257]; + unsigned int maxcode[18]; + int delta[17]; // old 'firstsymbol' - old 'firstcode' +} stbi__huffman; + +typedef struct +{ + stbi__context *s; + stbi__huffman huff_dc[4]; + stbi__huffman huff_ac[4]; + stbi_uc dequant[4][64]; + stbi__int16 fast_ac[4][1 << FAST_BITS]; + +// sizes for components, interleaved MCUs + int img_h_max, img_v_max; + int img_mcu_x, img_mcu_y; + int img_mcu_w, img_mcu_h; + +// definition of jpeg image component + struct + { + int id; + int h,v; + int tq; + int hd,ha; + int dc_pred; + + int x,y,w2,h2; + stbi_uc *data; + void *raw_data, *raw_coeff; + stbi_uc *linebuf; + short *coeff; // progressive only + int coeff_w, coeff_h; // number of 8x8 coefficient blocks + } img_comp[4]; + + stbi__uint32 code_buffer; // jpeg entropy-coded buffer + int code_bits; // number of valid bits + unsigned char marker; // marker seen while filling entropy buffer + int nomore; // flag if we saw a marker so must stop + + int progressive; + int spec_start; + int spec_end; + int succ_high; + int succ_low; + int eob_run; + + int scan_n, order[4]; + int restart_interval, todo; + +// kernels + void (*idct_block_kernel)(stbi_uc *out, int out_stride, short data[64]); + void (*YCbCr_to_RGB_kernel)(stbi_uc *out, const stbi_uc *y, const stbi_uc *pcb, const stbi_uc *pcr, int count, int step); + stbi_uc *(*resample_row_hv_2_kernel)(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs); +} stbi__jpeg; + +static int stbi__build_huffman(stbi__huffman *h, int *count) +{ + int i,j,k=0,code; + // build size list for each symbol (from JPEG spec) + for (i=0; i < 16; ++i) + for (j=0; j < count[i]; ++j) + h->size[k++] = (stbi_uc) (i+1); + h->size[k] = 0; + + // compute actual symbols (from jpeg spec) + code = 0; + k = 0; + for(j=1; j <= 16; ++j) { + // compute delta to add to code to compute symbol id + h->delta[j] = k - code; + if (h->size[k] == j) { + while (h->size[k] == j) + h->code[k++] = (stbi__uint16) (code++); + if (code-1 >= (1 << j)) return stbi__err("bad code lengths","Corrupt JPEG"); + } + // compute largest code + 1 for this size, preshifted as needed later + h->maxcode[j] = code << (16-j); + code <<= 1; + } + h->maxcode[j] = 0xffffffff; + + // build non-spec acceleration table; 255 is flag for not-accelerated + memset(h->fast, 255, 1 << FAST_BITS); + for (i=0; i < k; ++i) { + int s = h->size[i]; + if (s <= FAST_BITS) { + int c = h->code[i] << (FAST_BITS-s); + int m = 1 << (FAST_BITS-s); + for (j=0; j < m; ++j) { + h->fast[c+j] = (stbi_uc) i; + } + } + } + return 1; +} + +// build a table that decodes both magnitude and value of small ACs in +// one go. +static void stbi__build_fast_ac(stbi__int16 *fast_ac, stbi__huffman *h) +{ + int i; + for (i=0; i < (1 << FAST_BITS); ++i) { + stbi_uc fast = h->fast[i]; + fast_ac[i] = 0; + if (fast < 255) { + int rs = h->values[fast]; + int run = (rs >> 4) & 15; + int magbits = rs & 15; + int len = h->size[fast]; + + if (magbits && len + magbits <= FAST_BITS) { + // magnitude code followed by receive_extend code + int k = ((i << len) & ((1 << FAST_BITS) - 1)) >> (FAST_BITS - magbits); + int m = 1 << (magbits - 1); + if (k < m) k += (-1 << magbits) + 1; + // if the result is small enough, we can fit it in fast_ac table + if (k >= -128 && k <= 127) + fast_ac[i] = (stbi__int16) ((k << 8) + (run << 4) + (len + magbits)); + } + } + } +} + +static void stbi__grow_buffer_unsafe(stbi__jpeg *j) +{ + do { + int b = j->nomore ? 0 : stbi__get8(j->s); + if (b == 0xff) { + int c = stbi__get8(j->s); + if (c != 0) { + j->marker = (unsigned char) c; + j->nomore = 1; + return; + } + } + j->code_buffer |= b << (24 - j->code_bits); + j->code_bits += 8; + } while (j->code_bits <= 24); +} + +// (1 << n) - 1 +static stbi__uint32 stbi__bmask[17]={0,1,3,7,15,31,63,127,255,511,1023,2047,4095,8191,16383,32767,65535}; + +// decode a jpeg huffman value from the bitstream +stbi_inline static int stbi__jpeg_huff_decode(stbi__jpeg *j, stbi__huffman *h) +{ + unsigned int temp; + int c,k; + + if (j->code_bits < 16) stbi__grow_buffer_unsafe(j); + + // look at the top FAST_BITS and determine what symbol ID it is, + // if the code is <= FAST_BITS + c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1); + k = h->fast[c]; + if (k < 255) { + int s = h->size[k]; + if (s > j->code_bits) + return -1; + j->code_buffer <<= s; + j->code_bits -= s; + return h->values[k]; + } + + // naive test is to shift the code_buffer down so k bits are + // valid, then test against maxcode. To speed this up, we've + // preshifted maxcode left so that it has (16-k) 0s at the + // end; in other words, regardless of the number of bits, it + // wants to be compared against something shifted to have 16; + // that way we don't need to shift inside the loop. + temp = j->code_buffer >> 16; + for (k=FAST_BITS+1 ; ; ++k) + if (temp < h->maxcode[k]) + break; + if (k == 17) { + // error! code not found + j->code_bits -= 16; + return -1; + } + + if (k > j->code_bits) + return -1; + + // convert the huffman code to the symbol id + c = ((j->code_buffer >> (32 - k)) & stbi__bmask[k]) + h->delta[k]; + STBI_ASSERT((((j->code_buffer) >> (32 - h->size[c])) & stbi__bmask[h->size[c]]) == h->code[c]); + + // convert the id to a symbol + j->code_bits -= k; + j->code_buffer <<= k; + return h->values[c]; +} + +// bias[n] = (-1<code_bits < n) stbi__grow_buffer_unsafe(j); + + sgn = (stbi__int32)j->code_buffer >> 31; // sign bit is always in MSB + k = stbi_lrot(j->code_buffer, n); + STBI_ASSERT(n >= 0 && n < (int) (sizeof(stbi__bmask)/sizeof(*stbi__bmask))); + j->code_buffer = k & ~stbi__bmask[n]; + k &= stbi__bmask[n]; + j->code_bits -= n; + return k + (stbi__jbias[n] & ~sgn); +} + +// get some unsigned bits +stbi_inline static int stbi__jpeg_get_bits(stbi__jpeg *j, int n) +{ + unsigned int k; + if (j->code_bits < n) stbi__grow_buffer_unsafe(j); + k = stbi_lrot(j->code_buffer, n); + j->code_buffer = k & ~stbi__bmask[n]; + k &= stbi__bmask[n]; + j->code_bits -= n; + return k; +} + +stbi_inline static int stbi__jpeg_get_bit(stbi__jpeg *j) +{ + unsigned int k; + if (j->code_bits < 1) stbi__grow_buffer_unsafe(j); + k = j->code_buffer; + j->code_buffer <<= 1; + --j->code_bits; + return k & 0x80000000; +} + +// given a value that's at position X in the zigzag stream, +// where does it appear in the 8x8 matrix coded as row-major? +static stbi_uc stbi__jpeg_dezigzag[64+15] = +{ + 0, 1, 8, 16, 9, 2, 3, 10, + 17, 24, 32, 25, 18, 11, 4, 5, + 12, 19, 26, 33, 40, 48, 41, 34, + 27, 20, 13, 6, 7, 14, 21, 28, + 35, 42, 49, 56, 57, 50, 43, 36, + 29, 22, 15, 23, 30, 37, 44, 51, + 58, 59, 52, 45, 38, 31, 39, 46, + 53, 60, 61, 54, 47, 55, 62, 63, + // let corrupt input sample past end + 63, 63, 63, 63, 63, 63, 63, 63, + 63, 63, 63, 63, 63, 63, 63 +}; + +// decode one 64-entry block-- +static int stbi__jpeg_decode_block(stbi__jpeg *j, short data[64], stbi__huffman *hdc, stbi__huffman *hac, stbi__int16 *fac, int b, stbi_uc *dequant) +{ + int diff,dc,k; + int t; + + if (j->code_bits < 16) stbi__grow_buffer_unsafe(j); + t = stbi__jpeg_huff_decode(j, hdc); + if (t < 0) return stbi__err("bad huffman code","Corrupt JPEG"); + + // 0 all the ac values now so we can do it 32-bits at a time + memset(data,0,64*sizeof(data[0])); + + diff = t ? stbi__extend_receive(j, t) : 0; + dc = j->img_comp[b].dc_pred + diff; + j->img_comp[b].dc_pred = dc; + data[0] = (short) (dc * dequant[0]); + + // decode AC components, see JPEG spec + k = 1; + do { + unsigned int zig; + int c,r,s; + if (j->code_bits < 16) stbi__grow_buffer_unsafe(j); + c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1); + r = fac[c]; + if (r) { // fast-AC path + k += (r >> 4) & 15; // run + s = r & 15; // combined length + j->code_buffer <<= s; + j->code_bits -= s; + // decode into unzigzag'd location + zig = stbi__jpeg_dezigzag[k++]; + data[zig] = (short) ((r >> 8) * dequant[zig]); + } else { + int rs = stbi__jpeg_huff_decode(j, hac); + if (rs < 0) return stbi__err("bad huffman code","Corrupt JPEG"); + s = rs & 15; + r = rs >> 4; + if (s == 0) { + if (rs != 0xf0) break; // end block + k += 16; + } else { + k += r; + // decode into unzigzag'd location + zig = stbi__jpeg_dezigzag[k++]; + data[zig] = (short) (stbi__extend_receive(j,s) * dequant[zig]); + } + } + } while (k < 64); + return 1; +} + +static int stbi__jpeg_decode_block_prog_dc(stbi__jpeg *j, short data[64], stbi__huffman *hdc, int b) +{ + int diff,dc; + int t; + if (j->spec_end != 0) return stbi__err("can't merge dc and ac", "Corrupt JPEG"); + + if (j->code_bits < 16) stbi__grow_buffer_unsafe(j); + + if (j->succ_high == 0) { + // first scan for DC coefficient, must be first + memset(data,0,64*sizeof(data[0])); // 0 all the ac values now + t = stbi__jpeg_huff_decode(j, hdc); + diff = t ? stbi__extend_receive(j, t) : 0; + + dc = j->img_comp[b].dc_pred + diff; + j->img_comp[b].dc_pred = dc; + data[0] = (short) (dc << j->succ_low); + } else { + // refinement scan for DC coefficient + if (stbi__jpeg_get_bit(j)) + data[0] += (short) (1 << j->succ_low); + } + return 1; +} + +// @OPTIMIZE: store non-zigzagged during the decode passes, +// and only de-zigzag when dequantizing +static int stbi__jpeg_decode_block_prog_ac(stbi__jpeg *j, short data[64], stbi__huffman *hac, stbi__int16 *fac) +{ + int k; + if (j->spec_start == 0) return stbi__err("can't merge dc and ac", "Corrupt JPEG"); + + if (j->succ_high == 0) { + int shift = j->succ_low; + + if (j->eob_run) { + --j->eob_run; + return 1; + } + + k = j->spec_start; + do { + unsigned int zig; + int c,r,s; + if (j->code_bits < 16) stbi__grow_buffer_unsafe(j); + c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1); + r = fac[c]; + if (r) { // fast-AC path + k += (r >> 4) & 15; // run + s = r & 15; // combined length + j->code_buffer <<= s; + j->code_bits -= s; + zig = stbi__jpeg_dezigzag[k++]; + data[zig] = (short) ((r >> 8) << shift); + } else { + int rs = stbi__jpeg_huff_decode(j, hac); + if (rs < 0) return stbi__err("bad huffman code","Corrupt JPEG"); + s = rs & 15; + r = rs >> 4; + if (s == 0) { + if (r < 15) { + j->eob_run = (1 << r); + if (r) + j->eob_run += stbi__jpeg_get_bits(j, r); + --j->eob_run; + break; + } + k += 16; + } else { + k += r; + zig = stbi__jpeg_dezigzag[k++]; + data[zig] = (short) (stbi__extend_receive(j,s) << shift); + } + } + } while (k <= j->spec_end); + } else { + // refinement scan for these AC coefficients + + short bit = (short) (1 << j->succ_low); + + if (j->eob_run) { + --j->eob_run; + for (k = j->spec_start; k <= j->spec_end; ++k) { + short *p = &data[stbi__jpeg_dezigzag[k]]; + if (*p != 0) + if (stbi__jpeg_get_bit(j)) + if ((*p & bit)==0) { + if (*p > 0) + *p += bit; + else + *p -= bit; + } + } + } else { + k = j->spec_start; + do { + int r,s; + int rs = stbi__jpeg_huff_decode(j, hac); // @OPTIMIZE see if we can use the fast path here, advance-by-r is so slow, eh + if (rs < 0) return stbi__err("bad huffman code","Corrupt JPEG"); + s = rs & 15; + r = rs >> 4; + if (s == 0) { + if (r < 15) { + j->eob_run = (1 << r) - 1; + if (r) + j->eob_run += stbi__jpeg_get_bits(j, r); + r = 64; // force end of block + } else { + // r=15 s=0 should write 16 0s, so we just do + // a run of 15 0s and then write s (which is 0), + // so we don't have to do anything special here + } + } else { + if (s != 1) return stbi__err("bad huffman code", "Corrupt JPEG"); + // sign bit + if (stbi__jpeg_get_bit(j)) + s = bit; + else + s = -bit; + } + + // advance by r + while (k <= j->spec_end) { + short *p = &data[stbi__jpeg_dezigzag[k++]]; + if (*p != 0) { + if (stbi__jpeg_get_bit(j)) + if ((*p & bit)==0) { + if (*p > 0) + *p += bit; + else + *p -= bit; + } + } else { + if (r == 0) { + *p = (short) s; + break; + } + --r; + } + } + } while (k <= j->spec_end); + } + } + return 1; +} + +// take a -128..127 value and stbi__clamp it and convert to 0..255 +stbi_inline static stbi_uc stbi__clamp(int x) +{ + // trick to use a single test to catch both cases + if ((unsigned int) x > 255) { + if (x < 0) return 0; + if (x > 255) return 255; + } + return (stbi_uc) x; +} + +#define stbi__f2f(x) ((int) (((x) * 4096 + 0.5))) +#define stbi__fsh(x) ((x) << 12) + +// derived from jidctint -- DCT_ISLOW +#define STBI__IDCT_1D(s0,s1,s2,s3,s4,s5,s6,s7) \ + int t0,t1,t2,t3,p1,p2,p3,p4,p5,x0,x1,x2,x3; \ + p2 = s2; \ + p3 = s6; \ + p1 = (p2+p3) * stbi__f2f(0.5411961f); \ + t2 = p1 + p3*stbi__f2f(-1.847759065f); \ + t3 = p1 + p2*stbi__f2f( 0.765366865f); \ + p2 = s0; \ + p3 = s4; \ + t0 = stbi__fsh(p2+p3); \ + t1 = stbi__fsh(p2-p3); \ + x0 = t0+t3; \ + x3 = t0-t3; \ + x1 = t1+t2; \ + x2 = t1-t2; \ + t0 = s7; \ + t1 = s5; \ + t2 = s3; \ + t3 = s1; \ + p3 = t0+t2; \ + p4 = t1+t3; \ + p1 = t0+t3; \ + p2 = t1+t2; \ + p5 = (p3+p4)*stbi__f2f( 1.175875602f); \ + t0 = t0*stbi__f2f( 0.298631336f); \ + t1 = t1*stbi__f2f( 2.053119869f); \ + t2 = t2*stbi__f2f( 3.072711026f); \ + t3 = t3*stbi__f2f( 1.501321110f); \ + p1 = p5 + p1*stbi__f2f(-0.899976223f); \ + p2 = p5 + p2*stbi__f2f(-2.562915447f); \ + p3 = p3*stbi__f2f(-1.961570560f); \ + p4 = p4*stbi__f2f(-0.390180644f); \ + t3 += p1+p4; \ + t2 += p2+p3; \ + t1 += p2+p4; \ + t0 += p1+p3; + +static void stbi__idct_block(stbi_uc *out, int out_stride, short data[64]) +{ + int i,val[64],*v=val; + stbi_uc *o; + short *d = data; + + // columns + for (i=0; i < 8; ++i,++d, ++v) { + // if all zeroes, shortcut -- this avoids dequantizing 0s and IDCTing + if (d[ 8]==0 && d[16]==0 && d[24]==0 && d[32]==0 + && d[40]==0 && d[48]==0 && d[56]==0) { + // no shortcut 0 seconds + // (1|2|3|4|5|6|7)==0 0 seconds + // all separate -0.047 seconds + // 1 && 2|3 && 4|5 && 6|7: -0.047 seconds + int dcterm = d[0] << 2; + v[0] = v[8] = v[16] = v[24] = v[32] = v[40] = v[48] = v[56] = dcterm; + } else { + STBI__IDCT_1D(d[ 0],d[ 8],d[16],d[24],d[32],d[40],d[48],d[56]) + // constants scaled things up by 1<<12; let's bring them back + // down, but keep 2 extra bits of precision + x0 += 512; x1 += 512; x2 += 512; x3 += 512; + v[ 0] = (x0+t3) >> 10; + v[56] = (x0-t3) >> 10; + v[ 8] = (x1+t2) >> 10; + v[48] = (x1-t2) >> 10; + v[16] = (x2+t1) >> 10; + v[40] = (x2-t1) >> 10; + v[24] = (x3+t0) >> 10; + v[32] = (x3-t0) >> 10; + } + } + + for (i=0, v=val, o=out; i < 8; ++i,v+=8,o+=out_stride) { + // no fast case since the first 1D IDCT spread components out + STBI__IDCT_1D(v[0],v[1],v[2],v[3],v[4],v[5],v[6],v[7]) + // constants scaled things up by 1<<12, plus we had 1<<2 from first + // loop, plus horizontal and vertical each scale by sqrt(8) so together + // we've got an extra 1<<3, so 1<<17 total we need to remove. + // so we want to round that, which means adding 0.5 * 1<<17, + // aka 65536. Also, we'll end up with -128 to 127 that we want + // to encode as 0..255 by adding 128, so we'll add that before the shift + x0 += 65536 + (128<<17); + x1 += 65536 + (128<<17); + x2 += 65536 + (128<<17); + x3 += 65536 + (128<<17); + // tried computing the shifts into temps, or'ing the temps to see + // if any were out of range, but that was slower + o[0] = stbi__clamp((x0+t3) >> 17); + o[7] = stbi__clamp((x0-t3) >> 17); + o[1] = stbi__clamp((x1+t2) >> 17); + o[6] = stbi__clamp((x1-t2) >> 17); + o[2] = stbi__clamp((x2+t1) >> 17); + o[5] = stbi__clamp((x2-t1) >> 17); + o[3] = stbi__clamp((x3+t0) >> 17); + o[4] = stbi__clamp((x3-t0) >> 17); + } +} + +#ifdef STBI_SSE2 +// sse2 integer IDCT. not the fastest possible implementation but it +// produces bit-identical results to the generic C version so it's +// fully "transparent". +static void stbi__idct_simd(stbi_uc *out, int out_stride, short data[64]) +{ + // This is constructed to match our regular (generic) integer IDCT exactly. + __m128i row0, row1, row2, row3, row4, row5, row6, row7; + __m128i tmp; + + // dot product constant: even elems=x, odd elems=y + #define dct_const(x,y) _mm_setr_epi16((x),(y),(x),(y),(x),(y),(x),(y)) + + // out(0) = c0[even]*x + c0[odd]*y (c0, x, y 16-bit, out 32-bit) + // out(1) = c1[even]*x + c1[odd]*y + #define dct_rot(out0,out1, x,y,c0,c1) \ + __m128i c0##lo = _mm_unpacklo_epi16((x),(y)); \ + __m128i c0##hi = _mm_unpackhi_epi16((x),(y)); \ + __m128i out0##_l = _mm_madd_epi16(c0##lo, c0); \ + __m128i out0##_h = _mm_madd_epi16(c0##hi, c0); \ + __m128i out1##_l = _mm_madd_epi16(c0##lo, c1); \ + __m128i out1##_h = _mm_madd_epi16(c0##hi, c1) + + // out = in << 12 (in 16-bit, out 32-bit) + #define dct_widen(out, in) \ + __m128i out##_l = _mm_srai_epi32(_mm_unpacklo_epi16(_mm_setzero_si128(), (in)), 4); \ + __m128i out##_h = _mm_srai_epi32(_mm_unpackhi_epi16(_mm_setzero_si128(), (in)), 4) + + // wide add + #define dct_wadd(out, a, b) \ + __m128i out##_l = _mm_add_epi32(a##_l, b##_l); \ + __m128i out##_h = _mm_add_epi32(a##_h, b##_h) + + // wide sub + #define dct_wsub(out, a, b) \ + __m128i out##_l = _mm_sub_epi32(a##_l, b##_l); \ + __m128i out##_h = _mm_sub_epi32(a##_h, b##_h) + + // butterfly a/b, add bias, then shift by "s" and pack + #define dct_bfly32o(out0, out1, a,b,bias,s) \ + { \ + __m128i abiased_l = _mm_add_epi32(a##_l, bias); \ + __m128i abiased_h = _mm_add_epi32(a##_h, bias); \ + dct_wadd(sum, abiased, b); \ + dct_wsub(dif, abiased, b); \ + out0 = _mm_packs_epi32(_mm_srai_epi32(sum_l, s), _mm_srai_epi32(sum_h, s)); \ + out1 = _mm_packs_epi32(_mm_srai_epi32(dif_l, s), _mm_srai_epi32(dif_h, s)); \ + } + + // 8-bit interleave step (for transposes) + #define dct_interleave8(a, b) \ + tmp = a; \ + a = _mm_unpacklo_epi8(a, b); \ + b = _mm_unpackhi_epi8(tmp, b) + + // 16-bit interleave step (for transposes) + #define dct_interleave16(a, b) \ + tmp = a; \ + a = _mm_unpacklo_epi16(a, b); \ + b = _mm_unpackhi_epi16(tmp, b) + + #define dct_pass(bias,shift) \ + { \ + /* even part */ \ + dct_rot(t2e,t3e, row2,row6, rot0_0,rot0_1); \ + __m128i sum04 = _mm_add_epi16(row0, row4); \ + __m128i dif04 = _mm_sub_epi16(row0, row4); \ + dct_widen(t0e, sum04); \ + dct_widen(t1e, dif04); \ + dct_wadd(x0, t0e, t3e); \ + dct_wsub(x3, t0e, t3e); \ + dct_wadd(x1, t1e, t2e); \ + dct_wsub(x2, t1e, t2e); \ + /* odd part */ \ + dct_rot(y0o,y2o, row7,row3, rot2_0,rot2_1); \ + dct_rot(y1o,y3o, row5,row1, rot3_0,rot3_1); \ + __m128i sum17 = _mm_add_epi16(row1, row7); \ + __m128i sum35 = _mm_add_epi16(row3, row5); \ + dct_rot(y4o,y5o, sum17,sum35, rot1_0,rot1_1); \ + dct_wadd(x4, y0o, y4o); \ + dct_wadd(x5, y1o, y5o); \ + dct_wadd(x6, y2o, y5o); \ + dct_wadd(x7, y3o, y4o); \ + dct_bfly32o(row0,row7, x0,x7,bias,shift); \ + dct_bfly32o(row1,row6, x1,x6,bias,shift); \ + dct_bfly32o(row2,row5, x2,x5,bias,shift); \ + dct_bfly32o(row3,row4, x3,x4,bias,shift); \ + } + + __m128i rot0_0 = dct_const(stbi__f2f(0.5411961f), stbi__f2f(0.5411961f) + stbi__f2f(-1.847759065f)); + __m128i rot0_1 = dct_const(stbi__f2f(0.5411961f) + stbi__f2f( 0.765366865f), stbi__f2f(0.5411961f)); + __m128i rot1_0 = dct_const(stbi__f2f(1.175875602f) + stbi__f2f(-0.899976223f), stbi__f2f(1.175875602f)); + __m128i rot1_1 = dct_const(stbi__f2f(1.175875602f), stbi__f2f(1.175875602f) + stbi__f2f(-2.562915447f)); + __m128i rot2_0 = dct_const(stbi__f2f(-1.961570560f) + stbi__f2f( 0.298631336f), stbi__f2f(-1.961570560f)); + __m128i rot2_1 = dct_const(stbi__f2f(-1.961570560f), stbi__f2f(-1.961570560f) + stbi__f2f( 3.072711026f)); + __m128i rot3_0 = dct_const(stbi__f2f(-0.390180644f) + stbi__f2f( 2.053119869f), stbi__f2f(-0.390180644f)); + __m128i rot3_1 = dct_const(stbi__f2f(-0.390180644f), stbi__f2f(-0.390180644f) + stbi__f2f( 1.501321110f)); + + // rounding biases in column/row passes, see stbi__idct_block for explanation. + __m128i bias_0 = _mm_set1_epi32(512); + __m128i bias_1 = _mm_set1_epi32(65536 + (128<<17)); + + // load + row0 = _mm_load_si128((const __m128i *) (data + 0*8)); + row1 = _mm_load_si128((const __m128i *) (data + 1*8)); + row2 = _mm_load_si128((const __m128i *) (data + 2*8)); + row3 = _mm_load_si128((const __m128i *) (data + 3*8)); + row4 = _mm_load_si128((const __m128i *) (data + 4*8)); + row5 = _mm_load_si128((const __m128i *) (data + 5*8)); + row6 = _mm_load_si128((const __m128i *) (data + 6*8)); + row7 = _mm_load_si128((const __m128i *) (data + 7*8)); + + // column pass + dct_pass(bias_0, 10); + + { + // 16bit 8x8 transpose pass 1 + dct_interleave16(row0, row4); + dct_interleave16(row1, row5); + dct_interleave16(row2, row6); + dct_interleave16(row3, row7); + + // transpose pass 2 + dct_interleave16(row0, row2); + dct_interleave16(row1, row3); + dct_interleave16(row4, row6); + dct_interleave16(row5, row7); + + // transpose pass 3 + dct_interleave16(row0, row1); + dct_interleave16(row2, row3); + dct_interleave16(row4, row5); + dct_interleave16(row6, row7); + } + + // row pass + dct_pass(bias_1, 17); + + { + // pack + __m128i p0 = _mm_packus_epi16(row0, row1); // a0a1a2a3...a7b0b1b2b3...b7 + __m128i p1 = _mm_packus_epi16(row2, row3); + __m128i p2 = _mm_packus_epi16(row4, row5); + __m128i p3 = _mm_packus_epi16(row6, row7); + + // 8bit 8x8 transpose pass 1 + dct_interleave8(p0, p2); // a0e0a1e1... + dct_interleave8(p1, p3); // c0g0c1g1... + + // transpose pass 2 + dct_interleave8(p0, p1); // a0c0e0g0... + dct_interleave8(p2, p3); // b0d0f0h0... + + // transpose pass 3 + dct_interleave8(p0, p2); // a0b0c0d0... + dct_interleave8(p1, p3); // a4b4c4d4... + + // store + _mm_storel_epi64((__m128i *) out, p0); out += out_stride; + _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p0, 0x4e)); out += out_stride; + _mm_storel_epi64((__m128i *) out, p2); out += out_stride; + _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p2, 0x4e)); out += out_stride; + _mm_storel_epi64((__m128i *) out, p1); out += out_stride; + _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p1, 0x4e)); out += out_stride; + _mm_storel_epi64((__m128i *) out, p3); out += out_stride; + _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p3, 0x4e)); + } + +#undef dct_const +#undef dct_rot +#undef dct_widen +#undef dct_wadd +#undef dct_wsub +#undef dct_bfly32o +#undef dct_interleave8 +#undef dct_interleave16 +#undef dct_pass +} + +#endif // STBI_SSE2 + +#ifdef STBI_NEON + +// NEON integer IDCT. should produce bit-identical +// results to the generic C version. +static void stbi__idct_simd(stbi_uc *out, int out_stride, short data[64]) +{ + int16x8_t row0, row1, row2, row3, row4, row5, row6, row7; + + int16x4_t rot0_0 = vdup_n_s16(stbi__f2f(0.5411961f)); + int16x4_t rot0_1 = vdup_n_s16(stbi__f2f(-1.847759065f)); + int16x4_t rot0_2 = vdup_n_s16(stbi__f2f( 0.765366865f)); + int16x4_t rot1_0 = vdup_n_s16(stbi__f2f( 1.175875602f)); + int16x4_t rot1_1 = vdup_n_s16(stbi__f2f(-0.899976223f)); + int16x4_t rot1_2 = vdup_n_s16(stbi__f2f(-2.562915447f)); + int16x4_t rot2_0 = vdup_n_s16(stbi__f2f(-1.961570560f)); + int16x4_t rot2_1 = vdup_n_s16(stbi__f2f(-0.390180644f)); + int16x4_t rot3_0 = vdup_n_s16(stbi__f2f( 0.298631336f)); + int16x4_t rot3_1 = vdup_n_s16(stbi__f2f( 2.053119869f)); + int16x4_t rot3_2 = vdup_n_s16(stbi__f2f( 3.072711026f)); + int16x4_t rot3_3 = vdup_n_s16(stbi__f2f( 1.501321110f)); + +#define dct_long_mul(out, inq, coeff) \ + int32x4_t out##_l = vmull_s16(vget_low_s16(inq), coeff); \ + int32x4_t out##_h = vmull_s16(vget_high_s16(inq), coeff) + +#define dct_long_mac(out, acc, inq, coeff) \ + int32x4_t out##_l = vmlal_s16(acc##_l, vget_low_s16(inq), coeff); \ + int32x4_t out##_h = vmlal_s16(acc##_h, vget_high_s16(inq), coeff) + +#define dct_widen(out, inq) \ + int32x4_t out##_l = vshll_n_s16(vget_low_s16(inq), 12); \ + int32x4_t out##_h = vshll_n_s16(vget_high_s16(inq), 12) + +// wide add +#define dct_wadd(out, a, b) \ + int32x4_t out##_l = vaddq_s32(a##_l, b##_l); \ + int32x4_t out##_h = vaddq_s32(a##_h, b##_h) + +// wide sub +#define dct_wsub(out, a, b) \ + int32x4_t out##_l = vsubq_s32(a##_l, b##_l); \ + int32x4_t out##_h = vsubq_s32(a##_h, b##_h) + +// butterfly a/b, then shift using "shiftop" by "s" and pack +#define dct_bfly32o(out0,out1, a,b,shiftop,s) \ + { \ + dct_wadd(sum, a, b); \ + dct_wsub(dif, a, b); \ + out0 = vcombine_s16(shiftop(sum_l, s), shiftop(sum_h, s)); \ + out1 = vcombine_s16(shiftop(dif_l, s), shiftop(dif_h, s)); \ + } + +#define dct_pass(shiftop, shift) \ + { \ + /* even part */ \ + int16x8_t sum26 = vaddq_s16(row2, row6); \ + dct_long_mul(p1e, sum26, rot0_0); \ + dct_long_mac(t2e, p1e, row6, rot0_1); \ + dct_long_mac(t3e, p1e, row2, rot0_2); \ + int16x8_t sum04 = vaddq_s16(row0, row4); \ + int16x8_t dif04 = vsubq_s16(row0, row4); \ + dct_widen(t0e, sum04); \ + dct_widen(t1e, dif04); \ + dct_wadd(x0, t0e, t3e); \ + dct_wsub(x3, t0e, t3e); \ + dct_wadd(x1, t1e, t2e); \ + dct_wsub(x2, t1e, t2e); \ + /* odd part */ \ + int16x8_t sum15 = vaddq_s16(row1, row5); \ + int16x8_t sum17 = vaddq_s16(row1, row7); \ + int16x8_t sum35 = vaddq_s16(row3, row5); \ + int16x8_t sum37 = vaddq_s16(row3, row7); \ + int16x8_t sumodd = vaddq_s16(sum17, sum35); \ + dct_long_mul(p5o, sumodd, rot1_0); \ + dct_long_mac(p1o, p5o, sum17, rot1_1); \ + dct_long_mac(p2o, p5o, sum35, rot1_2); \ + dct_long_mul(p3o, sum37, rot2_0); \ + dct_long_mul(p4o, sum15, rot2_1); \ + dct_wadd(sump13o, p1o, p3o); \ + dct_wadd(sump24o, p2o, p4o); \ + dct_wadd(sump23o, p2o, p3o); \ + dct_wadd(sump14o, p1o, p4o); \ + dct_long_mac(x4, sump13o, row7, rot3_0); \ + dct_long_mac(x5, sump24o, row5, rot3_1); \ + dct_long_mac(x6, sump23o, row3, rot3_2); \ + dct_long_mac(x7, sump14o, row1, rot3_3); \ + dct_bfly32o(row0,row7, x0,x7,shiftop,shift); \ + dct_bfly32o(row1,row6, x1,x6,shiftop,shift); \ + dct_bfly32o(row2,row5, x2,x5,shiftop,shift); \ + dct_bfly32o(row3,row4, x3,x4,shiftop,shift); \ + } + + // load + row0 = vld1q_s16(data + 0*8); + row1 = vld1q_s16(data + 1*8); + row2 = vld1q_s16(data + 2*8); + row3 = vld1q_s16(data + 3*8); + row4 = vld1q_s16(data + 4*8); + row5 = vld1q_s16(data + 5*8); + row6 = vld1q_s16(data + 6*8); + row7 = vld1q_s16(data + 7*8); + + // add DC bias + row0 = vaddq_s16(row0, vsetq_lane_s16(1024, vdupq_n_s16(0), 0)); + + // column pass + dct_pass(vrshrn_n_s32, 10); + + // 16bit 8x8 transpose + { +// these three map to a single VTRN.16, VTRN.32, and VSWP, respectively. +// whether compilers actually get this is another story, sadly. +#define dct_trn16(x, y) { int16x8x2_t t = vtrnq_s16(x, y); x = t.val[0]; y = t.val[1]; } +#define dct_trn32(x, y) { int32x4x2_t t = vtrnq_s32(vreinterpretq_s32_s16(x), vreinterpretq_s32_s16(y)); x = vreinterpretq_s16_s32(t.val[0]); y = vreinterpretq_s16_s32(t.val[1]); } +#define dct_trn64(x, y) { int16x8_t x0 = x; int16x8_t y0 = y; x = vcombine_s16(vget_low_s16(x0), vget_low_s16(y0)); y = vcombine_s16(vget_high_s16(x0), vget_high_s16(y0)); } + + // pass 1 + dct_trn16(row0, row1); // a0b0a2b2a4b4a6b6 + dct_trn16(row2, row3); + dct_trn16(row4, row5); + dct_trn16(row6, row7); + + // pass 2 + dct_trn32(row0, row2); // a0b0c0d0a4b4c4d4 + dct_trn32(row1, row3); + dct_trn32(row4, row6); + dct_trn32(row5, row7); + + // pass 3 + dct_trn64(row0, row4); // a0b0c0d0e0f0g0h0 + dct_trn64(row1, row5); + dct_trn64(row2, row6); + dct_trn64(row3, row7); + +#undef dct_trn16 +#undef dct_trn32 +#undef dct_trn64 + } + + // row pass + // vrshrn_n_s32 only supports shifts up to 16, we need + // 17. so do a non-rounding shift of 16 first then follow + // up with a rounding shift by 1. + dct_pass(vshrn_n_s32, 16); + + { + // pack and round + uint8x8_t p0 = vqrshrun_n_s16(row0, 1); + uint8x8_t p1 = vqrshrun_n_s16(row1, 1); + uint8x8_t p2 = vqrshrun_n_s16(row2, 1); + uint8x8_t p3 = vqrshrun_n_s16(row3, 1); + uint8x8_t p4 = vqrshrun_n_s16(row4, 1); + uint8x8_t p5 = vqrshrun_n_s16(row5, 1); + uint8x8_t p6 = vqrshrun_n_s16(row6, 1); + uint8x8_t p7 = vqrshrun_n_s16(row7, 1); + + // again, these can translate into one instruction, but often don't. +#define dct_trn8_8(x, y) { uint8x8x2_t t = vtrn_u8(x, y); x = t.val[0]; y = t.val[1]; } +#define dct_trn8_16(x, y) { uint16x4x2_t t = vtrn_u16(vreinterpret_u16_u8(x), vreinterpret_u16_u8(y)); x = vreinterpret_u8_u16(t.val[0]); y = vreinterpret_u8_u16(t.val[1]); } +#define dct_trn8_32(x, y) { uint32x2x2_t t = vtrn_u32(vreinterpret_u32_u8(x), vreinterpret_u32_u8(y)); x = vreinterpret_u8_u32(t.val[0]); y = vreinterpret_u8_u32(t.val[1]); } + + // sadly can't use interleaved stores here since we only write + // 8 bytes to each scan line! + + // 8x8 8-bit transpose pass 1 + dct_trn8_8(p0, p1); + dct_trn8_8(p2, p3); + dct_trn8_8(p4, p5); + dct_trn8_8(p6, p7); + + // pass 2 + dct_trn8_16(p0, p2); + dct_trn8_16(p1, p3); + dct_trn8_16(p4, p6); + dct_trn8_16(p5, p7); + + // pass 3 + dct_trn8_32(p0, p4); + dct_trn8_32(p1, p5); + dct_trn8_32(p2, p6); + dct_trn8_32(p3, p7); + + // store + vst1_u8(out, p0); out += out_stride; + vst1_u8(out, p1); out += out_stride; + vst1_u8(out, p2); out += out_stride; + vst1_u8(out, p3); out += out_stride; + vst1_u8(out, p4); out += out_stride; + vst1_u8(out, p5); out += out_stride; + vst1_u8(out, p6); out += out_stride; + vst1_u8(out, p7); + +#undef dct_trn8_8 +#undef dct_trn8_16 +#undef dct_trn8_32 + } + +#undef dct_long_mul +#undef dct_long_mac +#undef dct_widen +#undef dct_wadd +#undef dct_wsub +#undef dct_bfly32o +#undef dct_pass +} + +#endif // STBI_NEON + +#define STBI__MARKER_none 0xff +// if there's a pending marker from the entropy stream, return that +// otherwise, fetch from the stream and get a marker. if there's no +// marker, return 0xff, which is never a valid marker value +static stbi_uc stbi__get_marker(stbi__jpeg *j) +{ + stbi_uc x; + if (j->marker != STBI__MARKER_none) { x = j->marker; j->marker = STBI__MARKER_none; return x; } + x = stbi__get8(j->s); + if (x != 0xff) return STBI__MARKER_none; + while (x == 0xff) + x = stbi__get8(j->s); + return x; +} + +// in each scan, we'll have scan_n components, and the order +// of the components is specified by order[] +#define STBI__RESTART(x) ((x) >= 0xd0 && (x) <= 0xd7) + +// after a restart interval, stbi__jpeg_reset the entropy decoder and +// the dc prediction +static void stbi__jpeg_reset(stbi__jpeg *j) +{ + j->code_bits = 0; + j->code_buffer = 0; + j->nomore = 0; + j->img_comp[0].dc_pred = j->img_comp[1].dc_pred = j->img_comp[2].dc_pred = 0; + j->marker = STBI__MARKER_none; + j->todo = j->restart_interval ? j->restart_interval : 0x7fffffff; + j->eob_run = 0; + // no more than 1<<31 MCUs if no restart_interal? that's plenty safe, + // since we don't even allow 1<<30 pixels +} + +static int stbi__parse_entropy_coded_data(stbi__jpeg *z) +{ + stbi__jpeg_reset(z); + if (!z->progressive) { + if (z->scan_n == 1) { + int i,j; + STBI_SIMD_ALIGN(short, data[64]); + int n = z->order[0]; + // non-interleaved data, we just need to process one block at a time, + // in trivial scanline order + // number of blocks to do just depends on how many actual "pixels" this + // component has, independent of interleaved MCU blocking and such + int w = (z->img_comp[n].x+7) >> 3; + int h = (z->img_comp[n].y+7) >> 3; + for (j=0; j < h; ++j) { + for (i=0; i < w; ++i) { + int ha = z->img_comp[n].ha; + if (!stbi__jpeg_decode_block(z, data, z->huff_dc+z->img_comp[n].hd, z->huff_ac+ha, z->fast_ac[ha], n, z->dequant[z->img_comp[n].tq])) return 0; + z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*j*8+i*8, z->img_comp[n].w2, data); + // every data block is an MCU, so countdown the restart interval + if (--z->todo <= 0) { + if (z->code_bits < 24) stbi__grow_buffer_unsafe(z); + // if it's NOT a restart, then just bail, so we get corrupt data + // rather than no data + if (!STBI__RESTART(z->marker)) return 1; + stbi__jpeg_reset(z); + } + } + } + return 1; + } else { // interleaved + int i,j,k,x,y; + STBI_SIMD_ALIGN(short, data[64]); + for (j=0; j < z->img_mcu_y; ++j) { + for (i=0; i < z->img_mcu_x; ++i) { + // scan an interleaved mcu... process scan_n components in order + for (k=0; k < z->scan_n; ++k) { + int n = z->order[k]; + // scan out an mcu's worth of this component; that's just determined + // by the basic H and V specified for the component + for (y=0; y < z->img_comp[n].v; ++y) { + for (x=0; x < z->img_comp[n].h; ++x) { + int x2 = (i*z->img_comp[n].h + x)*8; + int y2 = (j*z->img_comp[n].v + y)*8; + int ha = z->img_comp[n].ha; + if (!stbi__jpeg_decode_block(z, data, z->huff_dc+z->img_comp[n].hd, z->huff_ac+ha, z->fast_ac[ha], n, z->dequant[z->img_comp[n].tq])) return 0; + z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*y2+x2, z->img_comp[n].w2, data); + } + } + } + // after all interleaved components, that's an interleaved MCU, + // so now count down the restart interval + if (--z->todo <= 0) { + if (z->code_bits < 24) stbi__grow_buffer_unsafe(z); + if (!STBI__RESTART(z->marker)) return 1; + stbi__jpeg_reset(z); + } + } + } + return 1; + } + } else { + if (z->scan_n == 1) { + int i,j; + int n = z->order[0]; + // non-interleaved data, we just need to process one block at a time, + // in trivial scanline order + // number of blocks to do just depends on how many actual "pixels" this + // component has, independent of interleaved MCU blocking and such + int w = (z->img_comp[n].x+7) >> 3; + int h = (z->img_comp[n].y+7) >> 3; + for (j=0; j < h; ++j) { + for (i=0; i < w; ++i) { + short *data = z->img_comp[n].coeff + 64 * (i + j * z->img_comp[n].coeff_w); + if (z->spec_start == 0) { + if (!stbi__jpeg_decode_block_prog_dc(z, data, &z->huff_dc[z->img_comp[n].hd], n)) + return 0; + } else { + int ha = z->img_comp[n].ha; + if (!stbi__jpeg_decode_block_prog_ac(z, data, &z->huff_ac[ha], z->fast_ac[ha])) + return 0; + } + // every data block is an MCU, so countdown the restart interval + if (--z->todo <= 0) { + if (z->code_bits < 24) stbi__grow_buffer_unsafe(z); + if (!STBI__RESTART(z->marker)) return 1; + stbi__jpeg_reset(z); + } + } + } + return 1; + } else { // interleaved + int i,j,k,x,y; + for (j=0; j < z->img_mcu_y; ++j) { + for (i=0; i < z->img_mcu_x; ++i) { + // scan an interleaved mcu... process scan_n components in order + for (k=0; k < z->scan_n; ++k) { + int n = z->order[k]; + // scan out an mcu's worth of this component; that's just determined + // by the basic H and V specified for the component + for (y=0; y < z->img_comp[n].v; ++y) { + for (x=0; x < z->img_comp[n].h; ++x) { + int x2 = (i*z->img_comp[n].h + x); + int y2 = (j*z->img_comp[n].v + y); + short *data = z->img_comp[n].coeff + 64 * (x2 + y2 * z->img_comp[n].coeff_w); + if (!stbi__jpeg_decode_block_prog_dc(z, data, &z->huff_dc[z->img_comp[n].hd], n)) + return 0; + } + } + } + // after all interleaved components, that's an interleaved MCU, + // so now count down the restart interval + if (--z->todo <= 0) { + if (z->code_bits < 24) stbi__grow_buffer_unsafe(z); + if (!STBI__RESTART(z->marker)) return 1; + stbi__jpeg_reset(z); + } + } + } + return 1; + } + } +} + +static void stbi__jpeg_dequantize(short *data, stbi_uc *dequant) +{ + int i; + for (i=0; i < 64; ++i) + data[i] *= dequant[i]; +} + +static void stbi__jpeg_finish(stbi__jpeg *z) +{ + if (z->progressive) { + // dequantize and idct the data + int i,j,n; + for (n=0; n < z->s->img_n; ++n) { + int w = (z->img_comp[n].x+7) >> 3; + int h = (z->img_comp[n].y+7) >> 3; + for (j=0; j < h; ++j) { + for (i=0; i < w; ++i) { + short *data = z->img_comp[n].coeff + 64 * (i + j * z->img_comp[n].coeff_w); + stbi__jpeg_dequantize(data, z->dequant[z->img_comp[n].tq]); + z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*j*8+i*8, z->img_comp[n].w2, data); + } + } + } + } +} + +static int stbi__process_marker(stbi__jpeg *z, int m) +{ + int L; + switch (m) { + case STBI__MARKER_none: // no marker found + return stbi__err("expected marker","Corrupt JPEG"); + + case 0xDD: // DRI - specify restart interval + if (stbi__get16be(z->s) != 4) return stbi__err("bad DRI len","Corrupt JPEG"); + z->restart_interval = stbi__get16be(z->s); + return 1; + + case 0xDB: // DQT - define quantization table + L = stbi__get16be(z->s)-2; + while (L > 0) { + int q = stbi__get8(z->s); + int p = q >> 4; + int t = q & 15,i; + if (p != 0) return stbi__err("bad DQT type","Corrupt JPEG"); + if (t > 3) return stbi__err("bad DQT table","Corrupt JPEG"); + for (i=0; i < 64; ++i) + z->dequant[t][stbi__jpeg_dezigzag[i]] = stbi__get8(z->s); + L -= 65; + } + return L==0; + + case 0xC4: // DHT - define huffman table + L = stbi__get16be(z->s)-2; + while (L > 0) { + stbi_uc *v; + int sizes[16],i,n=0; + int q = stbi__get8(z->s); + int tc = q >> 4; + int th = q & 15; + if (tc > 1 || th > 3) return stbi__err("bad DHT header","Corrupt JPEG"); + for (i=0; i < 16; ++i) { + sizes[i] = stbi__get8(z->s); + n += sizes[i]; + } + L -= 17; + if (tc == 0) { + if (!stbi__build_huffman(z->huff_dc+th, sizes)) return 0; + v = z->huff_dc[th].values; + } else { + if (!stbi__build_huffman(z->huff_ac+th, sizes)) return 0; + v = z->huff_ac[th].values; + } + for (i=0; i < n; ++i) + v[i] = stbi__get8(z->s); + if (tc != 0) + stbi__build_fast_ac(z->fast_ac[th], z->huff_ac + th); + L -= n; + } + return L==0; + } + // check for comment block or APP blocks + if ((m >= 0xE0 && m <= 0xEF) || m == 0xFE) { + stbi__skip(z->s, stbi__get16be(z->s)-2); + return 1; + } + return 0; +} + +// after we see SOS +static int stbi__process_scan_header(stbi__jpeg *z) +{ + int i; + int Ls = stbi__get16be(z->s); + z->scan_n = stbi__get8(z->s); + if (z->scan_n < 1 || z->scan_n > 4 || z->scan_n > (int) z->s->img_n) return stbi__err("bad SOS component count","Corrupt JPEG"); + if (Ls != 6+2*z->scan_n) return stbi__err("bad SOS len","Corrupt JPEG"); + for (i=0; i < z->scan_n; ++i) { + int id = stbi__get8(z->s), which; + int q = stbi__get8(z->s); + for (which = 0; which < z->s->img_n; ++which) + if (z->img_comp[which].id == id) + break; + if (which == z->s->img_n) return 0; // no match + z->img_comp[which].hd = q >> 4; if (z->img_comp[which].hd > 3) return stbi__err("bad DC huff","Corrupt JPEG"); + z->img_comp[which].ha = q & 15; if (z->img_comp[which].ha > 3) return stbi__err("bad AC huff","Corrupt JPEG"); + z->order[i] = which; + } + + { + int aa; + z->spec_start = stbi__get8(z->s); + z->spec_end = stbi__get8(z->s); // should be 63, but might be 0 + aa = stbi__get8(z->s); + z->succ_high = (aa >> 4); + z->succ_low = (aa & 15); + if (z->progressive) { + if (z->spec_start > 63 || z->spec_end > 63 || z->spec_start > z->spec_end || z->succ_high > 13 || z->succ_low > 13) + return stbi__err("bad SOS", "Corrupt JPEG"); + } else { + if (z->spec_start != 0) return stbi__err("bad SOS","Corrupt JPEG"); + if (z->succ_high != 0 || z->succ_low != 0) return stbi__err("bad SOS","Corrupt JPEG"); + z->spec_end = 63; + } + } + + return 1; +} + +static int stbi__process_frame_header(stbi__jpeg *z, int scan) +{ + stbi__context *s = z->s; + int Lf,p,i,q, h_max=1,v_max=1,c; + Lf = stbi__get16be(s); if (Lf < 11) return stbi__err("bad SOF len","Corrupt JPEG"); // JPEG + p = stbi__get8(s); if (p != 8) return stbi__err("only 8-bit","JPEG format not supported: 8-bit only"); // JPEG baseline + s->img_y = stbi__get16be(s); if (s->img_y == 0) return stbi__err("no header height", "JPEG format not supported: delayed height"); // Legal, but we don't handle it--but neither does IJG + s->img_x = stbi__get16be(s); if (s->img_x == 0) return stbi__err("0 width","Corrupt JPEG"); // JPEG requires + c = stbi__get8(s); + if (c != 3 && c != 1) return stbi__err("bad component count","Corrupt JPEG"); // JFIF requires + s->img_n = c; + for (i=0; i < c; ++i) { + z->img_comp[i].data = NULL; + z->img_comp[i].linebuf = NULL; + } + + if (Lf != 8+3*s->img_n) return stbi__err("bad SOF len","Corrupt JPEG"); + + for (i=0; i < s->img_n; ++i) { + z->img_comp[i].id = stbi__get8(s); + if (z->img_comp[i].id != i+1) // JFIF requires + if (z->img_comp[i].id != i) // some version of jpegtran outputs non-JFIF-compliant files! + return stbi__err("bad component ID","Corrupt JPEG"); + q = stbi__get8(s); + z->img_comp[i].h = (q >> 4); if (!z->img_comp[i].h || z->img_comp[i].h > 4) return stbi__err("bad H","Corrupt JPEG"); + z->img_comp[i].v = q & 15; if (!z->img_comp[i].v || z->img_comp[i].v > 4) return stbi__err("bad V","Corrupt JPEG"); + z->img_comp[i].tq = stbi__get8(s); if (z->img_comp[i].tq > 3) return stbi__err("bad TQ","Corrupt JPEG"); + } + + if (scan != STBI__SCAN_load) return 1; + + if ((1 << 30) / s->img_x / s->img_n < s->img_y) return stbi__err("too large", "Image too large to decode"); + + for (i=0; i < s->img_n; ++i) { + if (z->img_comp[i].h > h_max) h_max = z->img_comp[i].h; + if (z->img_comp[i].v > v_max) v_max = z->img_comp[i].v; + } + + // compute interleaved mcu info + z->img_h_max = h_max; + z->img_v_max = v_max; + z->img_mcu_w = h_max * 8; + z->img_mcu_h = v_max * 8; + z->img_mcu_x = (s->img_x + z->img_mcu_w-1) / z->img_mcu_w; + z->img_mcu_y = (s->img_y + z->img_mcu_h-1) / z->img_mcu_h; + + for (i=0; i < s->img_n; ++i) { + // number of effective pixels (e.g. for non-interleaved MCU) + z->img_comp[i].x = (s->img_x * z->img_comp[i].h + h_max-1) / h_max; + z->img_comp[i].y = (s->img_y * z->img_comp[i].v + v_max-1) / v_max; + // to simplify generation, we'll allocate enough memory to decode + // the bogus oversized data from using interleaved MCUs and their + // big blocks (e.g. a 16x16 iMCU on an image of width 33); we won't + // discard the extra data until colorspace conversion + z->img_comp[i].w2 = z->img_mcu_x * z->img_comp[i].h * 8; + z->img_comp[i].h2 = z->img_mcu_y * z->img_comp[i].v * 8; + z->img_comp[i].raw_data = stbi__malloc(z->img_comp[i].w2 * z->img_comp[i].h2+15); + + if (z->img_comp[i].raw_data == NULL) { + for(--i; i >= 0; --i) { + STBI_FREE(z->img_comp[i].raw_data); + z->img_comp[i].raw_data = NULL; + } + return stbi__err("outofmem", "Out of memory"); + } + // align blocks for idct using mmx/sse + z->img_comp[i].data = (stbi_uc*) (((size_t) z->img_comp[i].raw_data + 15) & ~15); + z->img_comp[i].linebuf = NULL; + if (z->progressive) { + z->img_comp[i].coeff_w = (z->img_comp[i].w2 + 7) >> 3; + z->img_comp[i].coeff_h = (z->img_comp[i].h2 + 7) >> 3; + z->img_comp[i].raw_coeff = STBI_MALLOC(z->img_comp[i].coeff_w * z->img_comp[i].coeff_h * 64 * sizeof(short) + 15); + z->img_comp[i].coeff = (short*) (((size_t) z->img_comp[i].raw_coeff + 15) & ~15); + } else { + z->img_comp[i].coeff = 0; + z->img_comp[i].raw_coeff = 0; + } + } + + return 1; +} + +// use comparisons since in some cases we handle more than one case (e.g. SOF) +#define stbi__DNL(x) ((x) == 0xdc) +#define stbi__SOI(x) ((x) == 0xd8) +#define stbi__EOI(x) ((x) == 0xd9) +#define stbi__SOF(x) ((x) == 0xc0 || (x) == 0xc1 || (x) == 0xc2) +#define stbi__SOS(x) ((x) == 0xda) + +#define stbi__SOF_progressive(x) ((x) == 0xc2) + +static int stbi__decode_jpeg_header(stbi__jpeg *z, int scan) +{ + int m; + z->marker = STBI__MARKER_none; // initialize cached marker to empty + m = stbi__get_marker(z); + if (!stbi__SOI(m)) return stbi__err("no SOI","Corrupt JPEG"); + if (scan == STBI__SCAN_type) return 1; + m = stbi__get_marker(z); + while (!stbi__SOF(m)) { + if (!stbi__process_marker(z,m)) return 0; + m = stbi__get_marker(z); + while (m == STBI__MARKER_none) { + // some files have extra padding after their blocks, so ok, we'll scan + if (stbi__at_eof(z->s)) return stbi__err("no SOF", "Corrupt JPEG"); + m = stbi__get_marker(z); + } + } + z->progressive = stbi__SOF_progressive(m); + if (!stbi__process_frame_header(z, scan)) return 0; + return 1; +} + +// decode image to YCbCr format +static int stbi__decode_jpeg_image(stbi__jpeg *j) +{ + int m; + for (m = 0; m < 4; m++) { + j->img_comp[m].raw_data = NULL; + j->img_comp[m].raw_coeff = NULL; + } + j->restart_interval = 0; + if (!stbi__decode_jpeg_header(j, STBI__SCAN_load)) return 0; + m = stbi__get_marker(j); + while (!stbi__EOI(m)) { + if (stbi__SOS(m)) { + if (!stbi__process_scan_header(j)) return 0; + if (!stbi__parse_entropy_coded_data(j)) return 0; + if (j->marker == STBI__MARKER_none ) { + // handle 0s at the end of image data from IP Kamera 9060 + while (!stbi__at_eof(j->s)) { + int x = stbi__get8(j->s); + if (x == 255) { + j->marker = stbi__get8(j->s); + break; + } else if (x != 0) { + return stbi__err("junk before marker", "Corrupt JPEG"); + } + } + // if we reach eof without hitting a marker, stbi__get_marker() below will fail and we'll eventually return 0 + } + } else { + if (!stbi__process_marker(j, m)) return 0; + } + m = stbi__get_marker(j); + } + if (j->progressive) + stbi__jpeg_finish(j); + return 1; +} + +// static jfif-centered resampling (across block boundaries) + +typedef stbi_uc *(*resample_row_func)(stbi_uc *out, stbi_uc *in0, stbi_uc *in1, + int w, int hs); + +#define stbi__div4(x) ((stbi_uc) ((x) >> 2)) + +static stbi_uc *resample_row_1(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + STBI_NOTUSED(out); + STBI_NOTUSED(in_far); + STBI_NOTUSED(w); + STBI_NOTUSED(hs); + return in_near; +} + +static stbi_uc* stbi__resample_row_v_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + // need to generate two samples vertically for every one in input + int i; + STBI_NOTUSED(hs); + for (i=0; i < w; ++i) + out[i] = stbi__div4(3*in_near[i] + in_far[i] + 2); + return out; +} + +static stbi_uc* stbi__resample_row_h_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + // need to generate two samples horizontally for every one in input + int i; + stbi_uc *input = in_near; + + if (w == 1) { + // if only one sample, can't do any interpolation + out[0] = out[1] = input[0]; + return out; + } + + out[0] = input[0]; + out[1] = stbi__div4(input[0]*3 + input[1] + 2); + for (i=1; i < w-1; ++i) { + int n = 3*input[i]+2; + out[i*2+0] = stbi__div4(n+input[i-1]); + out[i*2+1] = stbi__div4(n+input[i+1]); + } + out[i*2+0] = stbi__div4(input[w-2]*3 + input[w-1] + 2); + out[i*2+1] = input[w-1]; + + STBI_NOTUSED(in_far); + STBI_NOTUSED(hs); + + return out; +} + +#define stbi__div16(x) ((stbi_uc) ((x) >> 4)) + +static stbi_uc *stbi__resample_row_hv_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + // need to generate 2x2 samples for every one in input + int i,t0,t1; + if (w == 1) { + out[0] = out[1] = stbi__div4(3*in_near[0] + in_far[0] + 2); + return out; + } + + t1 = 3*in_near[0] + in_far[0]; + out[0] = stbi__div4(t1+2); + for (i=1; i < w; ++i) { + t0 = t1; + t1 = 3*in_near[i]+in_far[i]; + out[i*2-1] = stbi__div16(3*t0 + t1 + 8); + out[i*2 ] = stbi__div16(3*t1 + t0 + 8); + } + out[w*2-1] = stbi__div4(t1+2); + + STBI_NOTUSED(hs); + + return out; +} + +#if defined(STBI_SSE2) || defined(STBI_NEON) +static stbi_uc *stbi__resample_row_hv_2_simd(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + // need to generate 2x2 samples for every one in input + int i=0,t0,t1; + + if (w == 1) { + out[0] = out[1] = stbi__div4(3*in_near[0] + in_far[0] + 2); + return out; + } + + t1 = 3*in_near[0] + in_far[0]; + // process groups of 8 pixels for as long as we can. + // note we can't handle the last pixel in a row in this loop + // because we need to handle the filter boundary conditions. + for (; i < ((w-1) & ~7); i += 8) { +#if defined(STBI_SSE2) + // load and perform the vertical filtering pass + // this uses 3*x + y = 4*x + (y - x) + __m128i zero = _mm_setzero_si128(); + __m128i farb = _mm_loadl_epi64((__m128i *) (in_far + i)); + __m128i nearb = _mm_loadl_epi64((__m128i *) (in_near + i)); + __m128i farw = _mm_unpacklo_epi8(farb, zero); + __m128i nearw = _mm_unpacklo_epi8(nearb, zero); + __m128i diff = _mm_sub_epi16(farw, nearw); + __m128i nears = _mm_slli_epi16(nearw, 2); + __m128i curr = _mm_add_epi16(nears, diff); // current row + + // horizontal filter works the same based on shifted vers of current + // row. "prev" is current row shifted right by 1 pixel; we need to + // insert the previous pixel value (from t1). + // "next" is current row shifted left by 1 pixel, with first pixel + // of next block of 8 pixels added in. + __m128i prv0 = _mm_slli_si128(curr, 2); + __m128i nxt0 = _mm_srli_si128(curr, 2); + __m128i prev = _mm_insert_epi16(prv0, t1, 0); + __m128i next = _mm_insert_epi16(nxt0, 3*in_near[i+8] + in_far[i+8], 7); + + // horizontal filter, polyphase implementation since it's convenient: + // even pixels = 3*cur + prev = cur*4 + (prev - cur) + // odd pixels = 3*cur + next = cur*4 + (next - cur) + // note the shared term. + __m128i bias = _mm_set1_epi16(8); + __m128i curs = _mm_slli_epi16(curr, 2); + __m128i prvd = _mm_sub_epi16(prev, curr); + __m128i nxtd = _mm_sub_epi16(next, curr); + __m128i curb = _mm_add_epi16(curs, bias); + __m128i even = _mm_add_epi16(prvd, curb); + __m128i odd = _mm_add_epi16(nxtd, curb); + + // interleave even and odd pixels, then undo scaling. + __m128i int0 = _mm_unpacklo_epi16(even, odd); + __m128i int1 = _mm_unpackhi_epi16(even, odd); + __m128i de0 = _mm_srli_epi16(int0, 4); + __m128i de1 = _mm_srli_epi16(int1, 4); + + // pack and write output + __m128i outv = _mm_packus_epi16(de0, de1); + _mm_storeu_si128((__m128i *) (out + i*2), outv); +#elif defined(STBI_NEON) + // load and perform the vertical filtering pass + // this uses 3*x + y = 4*x + (y - x) + uint8x8_t farb = vld1_u8(in_far + i); + uint8x8_t nearb = vld1_u8(in_near + i); + int16x8_t diff = vreinterpretq_s16_u16(vsubl_u8(farb, nearb)); + int16x8_t nears = vreinterpretq_s16_u16(vshll_n_u8(nearb, 2)); + int16x8_t curr = vaddq_s16(nears, diff); // current row + + // horizontal filter works the same based on shifted vers of current + // row. "prev" is current row shifted right by 1 pixel; we need to + // insert the previous pixel value (from t1). + // "next" is current row shifted left by 1 pixel, with first pixel + // of next block of 8 pixels added in. + int16x8_t prv0 = vextq_s16(curr, curr, 7); + int16x8_t nxt0 = vextq_s16(curr, curr, 1); + int16x8_t prev = vsetq_lane_s16(t1, prv0, 0); + int16x8_t next = vsetq_lane_s16(3*in_near[i+8] + in_far[i+8], nxt0, 7); + + // horizontal filter, polyphase implementation since it's convenient: + // even pixels = 3*cur + prev = cur*4 + (prev - cur) + // odd pixels = 3*cur + next = cur*4 + (next - cur) + // note the shared term. + int16x8_t curs = vshlq_n_s16(curr, 2); + int16x8_t prvd = vsubq_s16(prev, curr); + int16x8_t nxtd = vsubq_s16(next, curr); + int16x8_t even = vaddq_s16(curs, prvd); + int16x8_t odd = vaddq_s16(curs, nxtd); + + // undo scaling and round, then store with even/odd phases interleaved + uint8x8x2_t o; + o.val[0] = vqrshrun_n_s16(even, 4); + o.val[1] = vqrshrun_n_s16(odd, 4); + vst2_u8(out + i*2, o); +#endif + + // "previous" value for next iter + t1 = 3*in_near[i+7] + in_far[i+7]; + } + + t0 = t1; + t1 = 3*in_near[i] + in_far[i]; + out[i*2] = stbi__div16(3*t1 + t0 + 8); + + for (++i; i < w; ++i) { + t0 = t1; + t1 = 3*in_near[i]+in_far[i]; + out[i*2-1] = stbi__div16(3*t0 + t1 + 8); + out[i*2 ] = stbi__div16(3*t1 + t0 + 8); + } + out[w*2-1] = stbi__div4(t1+2); + + STBI_NOTUSED(hs); + + return out; +} +#endif + +static stbi_uc *stbi__resample_row_generic(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + // resample with nearest-neighbor + int i,j; + STBI_NOTUSED(in_far); + for (i=0; i < w; ++i) + for (j=0; j < hs; ++j) + out[i*hs+j] = in_near[i]; + return out; +} + +#ifdef STBI_JPEG_OLD +// this is the same YCbCr-to-RGB calculation that stb_image has used +// historically before the algorithm changes in 1.49 +#define float2fixed(x) ((int) ((x) * 65536 + 0.5)) +static void stbi__YCbCr_to_RGB_row(stbi_uc *out, const stbi_uc *y, const stbi_uc *pcb, const stbi_uc *pcr, int count, int step) +{ + int i; + for (i=0; i < count; ++i) { + int y_fixed = (y[i] << 16) + 32768; // rounding + int r,g,b; + int cr = pcr[i] - 128; + int cb = pcb[i] - 128; + r = y_fixed + cr*float2fixed(1.40200f); + g = y_fixed - cr*float2fixed(0.71414f) - cb*float2fixed(0.34414f); + b = y_fixed + cb*float2fixed(1.77200f); + r >>= 16; + g >>= 16; + b >>= 16; + if ((unsigned) r > 255) { if (r < 0) r = 0; else r = 255; } + if ((unsigned) g > 255) { if (g < 0) g = 0; else g = 255; } + if ((unsigned) b > 255) { if (b < 0) b = 0; else b = 255; } + out[0] = (stbi_uc)r; + out[1] = (stbi_uc)g; + out[2] = (stbi_uc)b; + out[3] = 255; + out += step; + } +} +#else +// this is a reduced-precision calculation of YCbCr-to-RGB introduced +// to make sure the code produces the same results in both SIMD and scalar +#define float2fixed(x) (((int) ((x) * 4096.0f + 0.5f)) << 8) +static void stbi__YCbCr_to_RGB_row(stbi_uc *out, const stbi_uc *y, const stbi_uc *pcb, const stbi_uc *pcr, int count, int step) +{ + int i; + for (i=0; i < count; ++i) { + int y_fixed = (y[i] << 20) + (1<<19); // rounding + int r,g,b; + int cr = pcr[i] - 128; + int cb = pcb[i] - 128; + r = y_fixed + cr* float2fixed(1.40200f); + g = y_fixed + (cr*-float2fixed(0.71414f)) + ((cb*-float2fixed(0.34414f)) & 0xffff0000); + b = y_fixed + cb* float2fixed(1.77200f); + r >>= 20; + g >>= 20; + b >>= 20; + if ((unsigned) r > 255) { if (r < 0) r = 0; else r = 255; } + if ((unsigned) g > 255) { if (g < 0) g = 0; else g = 255; } + if ((unsigned) b > 255) { if (b < 0) b = 0; else b = 255; } + out[0] = (stbi_uc)r; + out[1] = (stbi_uc)g; + out[2] = (stbi_uc)b; + out[3] = 255; + out += step; + } +} +#endif + +#if defined(STBI_SSE2) || defined(STBI_NEON) +static void stbi__YCbCr_to_RGB_simd(stbi_uc *out, stbi_uc const *y, stbi_uc const *pcb, stbi_uc const *pcr, int count, int step) +{ + int i = 0; + +#ifdef STBI_SSE2 + // step == 3 is pretty ugly on the final interleave, and i'm not convinced + // it's useful in practice (you wouldn't use it for textures, for example). + // so just accelerate step == 4 case. + if (step == 4) { + // this is a fairly straightforward implementation and not super-optimized. + __m128i signflip = _mm_set1_epi8(-0x80); + __m128i cr_const0 = _mm_set1_epi16( (short) ( 1.40200f*4096.0f+0.5f)); + __m128i cr_const1 = _mm_set1_epi16( - (short) ( 0.71414f*4096.0f+0.5f)); + __m128i cb_const0 = _mm_set1_epi16( - (short) ( 0.34414f*4096.0f+0.5f)); + __m128i cb_const1 = _mm_set1_epi16( (short) ( 1.77200f*4096.0f+0.5f)); + __m128i y_bias = _mm_set1_epi8((char) (unsigned char) 128); + __m128i xw = _mm_set1_epi16(255); // alpha channel + + for (; i+7 < count; i += 8) { + // load + __m128i y_bytes = _mm_loadl_epi64((__m128i *) (y+i)); + __m128i cr_bytes = _mm_loadl_epi64((__m128i *) (pcr+i)); + __m128i cb_bytes = _mm_loadl_epi64((__m128i *) (pcb+i)); + __m128i cr_biased = _mm_xor_si128(cr_bytes, signflip); // -128 + __m128i cb_biased = _mm_xor_si128(cb_bytes, signflip); // -128 + + // unpack to short (and left-shift cr, cb by 8) + __m128i yw = _mm_unpacklo_epi8(y_bias, y_bytes); + __m128i crw = _mm_unpacklo_epi8(_mm_setzero_si128(), cr_biased); + __m128i cbw = _mm_unpacklo_epi8(_mm_setzero_si128(), cb_biased); + + // color transform + __m128i yws = _mm_srli_epi16(yw, 4); + __m128i cr0 = _mm_mulhi_epi16(cr_const0, crw); + __m128i cb0 = _mm_mulhi_epi16(cb_const0, cbw); + __m128i cb1 = _mm_mulhi_epi16(cbw, cb_const1); + __m128i cr1 = _mm_mulhi_epi16(crw, cr_const1); + __m128i rws = _mm_add_epi16(cr0, yws); + __m128i gwt = _mm_add_epi16(cb0, yws); + __m128i bws = _mm_add_epi16(yws, cb1); + __m128i gws = _mm_add_epi16(gwt, cr1); + + // descale + __m128i rw = _mm_srai_epi16(rws, 4); + __m128i bw = _mm_srai_epi16(bws, 4); + __m128i gw = _mm_srai_epi16(gws, 4); + + // back to byte, set up for transpose + __m128i brb = _mm_packus_epi16(rw, bw); + __m128i gxb = _mm_packus_epi16(gw, xw); + + // transpose to interleave channels + __m128i t0 = _mm_unpacklo_epi8(brb, gxb); + __m128i t1 = _mm_unpackhi_epi8(brb, gxb); + __m128i o0 = _mm_unpacklo_epi16(t0, t1); + __m128i o1 = _mm_unpackhi_epi16(t0, t1); + + // store + _mm_storeu_si128((__m128i *) (out + 0), o0); + _mm_storeu_si128((__m128i *) (out + 16), o1); + out += 32; + } + } +#endif + +#ifdef STBI_NEON + // in this version, step=3 support would be easy to add. but is there demand? + if (step == 4) { + // this is a fairly straightforward implementation and not super-optimized. + uint8x8_t signflip = vdup_n_u8(0x80); + int16x8_t cr_const0 = vdupq_n_s16( (short) ( 1.40200f*4096.0f+0.5f)); + int16x8_t cr_const1 = vdupq_n_s16( - (short) ( 0.71414f*4096.0f+0.5f)); + int16x8_t cb_const0 = vdupq_n_s16( - (short) ( 0.34414f*4096.0f+0.5f)); + int16x8_t cb_const1 = vdupq_n_s16( (short) ( 1.77200f*4096.0f+0.5f)); + + for (; i+7 < count; i += 8) { + // load + uint8x8_t y_bytes = vld1_u8(y + i); + uint8x8_t cr_bytes = vld1_u8(pcr + i); + uint8x8_t cb_bytes = vld1_u8(pcb + i); + int8x8_t cr_biased = vreinterpret_s8_u8(vsub_u8(cr_bytes, signflip)); + int8x8_t cb_biased = vreinterpret_s8_u8(vsub_u8(cb_bytes, signflip)); + + // expand to s16 + int16x8_t yws = vreinterpretq_s16_u16(vshll_n_u8(y_bytes, 4)); + int16x8_t crw = vshll_n_s8(cr_biased, 7); + int16x8_t cbw = vshll_n_s8(cb_biased, 7); + + // color transform + int16x8_t cr0 = vqdmulhq_s16(crw, cr_const0); + int16x8_t cb0 = vqdmulhq_s16(cbw, cb_const0); + int16x8_t cr1 = vqdmulhq_s16(crw, cr_const1); + int16x8_t cb1 = vqdmulhq_s16(cbw, cb_const1); + int16x8_t rws = vaddq_s16(yws, cr0); + int16x8_t gws = vaddq_s16(vaddq_s16(yws, cb0), cr1); + int16x8_t bws = vaddq_s16(yws, cb1); + + // undo scaling, round, convert to byte + uint8x8x4_t o; + o.val[0] = vqrshrun_n_s16(rws, 4); + o.val[1] = vqrshrun_n_s16(gws, 4); + o.val[2] = vqrshrun_n_s16(bws, 4); + o.val[3] = vdup_n_u8(255); + + // store, interleaving r/g/b/a + vst4_u8(out, o); + out += 8*4; + } + } +#endif + + for (; i < count; ++i) { + int y_fixed = (y[i] << 20) + (1<<19); // rounding + int r,g,b; + int cr = pcr[i] - 128; + int cb = pcb[i] - 128; + r = y_fixed + cr* float2fixed(1.40200f); + g = y_fixed + cr*-float2fixed(0.71414f) + ((cb*-float2fixed(0.34414f)) & 0xffff0000); + b = y_fixed + cb* float2fixed(1.77200f); + r >>= 20; + g >>= 20; + b >>= 20; + if ((unsigned) r > 255) { if (r < 0) r = 0; else r = 255; } + if ((unsigned) g > 255) { if (g < 0) g = 0; else g = 255; } + if ((unsigned) b > 255) { if (b < 0) b = 0; else b = 255; } + out[0] = (stbi_uc)r; + out[1] = (stbi_uc)g; + out[2] = (stbi_uc)b; + out[3] = 255; + out += step; + } +} +#endif + +// set up the kernels +static void stbi__setup_jpeg(stbi__jpeg *j) +{ + j->idct_block_kernel = stbi__idct_block; + j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_row; + j->resample_row_hv_2_kernel = stbi__resample_row_hv_2; + +#ifdef STBI_SSE2 + if (stbi__sse2_available()) { + j->idct_block_kernel = stbi__idct_simd; + #ifndef STBI_JPEG_OLD + j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_simd; + #endif + j->resample_row_hv_2_kernel = stbi__resample_row_hv_2_simd; + } +#endif + +#ifdef STBI_NEON + j->idct_block_kernel = stbi__idct_simd; + #ifndef STBI_JPEG_OLD + j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_simd; + #endif + j->resample_row_hv_2_kernel = stbi__resample_row_hv_2_simd; +#endif +} + +// clean up the temporary component buffers +static void stbi__cleanup_jpeg(stbi__jpeg *j) +{ + int i; + for (i=0; i < j->s->img_n; ++i) { + if (j->img_comp[i].raw_data) { + STBI_FREE(j->img_comp[i].raw_data); + j->img_comp[i].raw_data = NULL; + j->img_comp[i].data = NULL; + } + if (j->img_comp[i].raw_coeff) { + STBI_FREE(j->img_comp[i].raw_coeff); + j->img_comp[i].raw_coeff = 0; + j->img_comp[i].coeff = 0; + } + if (j->img_comp[i].linebuf) { + STBI_FREE(j->img_comp[i].linebuf); + j->img_comp[i].linebuf = NULL; + } + } +} + +typedef struct +{ + resample_row_func resample; + stbi_uc *line0,*line1; + int hs,vs; // expansion factor in each axis + int w_lores; // horizontal pixels pre-expansion + int ystep; // how far through vertical expansion we are + int ypos; // which pre-expansion row we're on +} stbi__resample; + +static stbi_uc *load_jpeg_image(stbi__jpeg *z, int *out_x, int *out_y, int *comp, int req_comp) +{ + int n, decode_n; + z->s->img_n = 0; // make stbi__cleanup_jpeg safe + + // validate req_comp + if (req_comp < 0 || req_comp > 4) return stbi__errpuc("bad req_comp", "Internal error"); + + // load a jpeg image from whichever source, but leave in YCbCr format + if (!stbi__decode_jpeg_image(z)) { stbi__cleanup_jpeg(z); return NULL; } + + // determine actual number of components to generate + n = req_comp ? req_comp : z->s->img_n; + + if (z->s->img_n == 3 && n < 3) + decode_n = 1; + else + decode_n = z->s->img_n; + + // resample and color-convert + { + int k; + unsigned int i,j; + stbi_uc *output; + stbi_uc *coutput[4]; + + stbi__resample res_comp[4]; + + for (k=0; k < decode_n; ++k) { + stbi__resample *r = &res_comp[k]; + + // allocate line buffer big enough for upsampling off the edges + // with upsample factor of 4 + z->img_comp[k].linebuf = (stbi_uc *) stbi__malloc(z->s->img_x + 3); + if (!z->img_comp[k].linebuf) { stbi__cleanup_jpeg(z); return stbi__errpuc("outofmem", "Out of memory"); } + + r->hs = z->img_h_max / z->img_comp[k].h; + r->vs = z->img_v_max / z->img_comp[k].v; + r->ystep = r->vs >> 1; + r->w_lores = (z->s->img_x + r->hs-1) / r->hs; + r->ypos = 0; + r->line0 = r->line1 = z->img_comp[k].data; + + if (r->hs == 1 && r->vs == 1) r->resample = resample_row_1; + else if (r->hs == 1 && r->vs == 2) r->resample = stbi__resample_row_v_2; + else if (r->hs == 2 && r->vs == 1) r->resample = stbi__resample_row_h_2; + else if (r->hs == 2 && r->vs == 2) r->resample = z->resample_row_hv_2_kernel; + else r->resample = stbi__resample_row_generic; + } + + // can't error after this so, this is safe + output = (stbi_uc *) stbi__malloc(n * z->s->img_x * z->s->img_y + 1); + if (!output) { stbi__cleanup_jpeg(z); return stbi__errpuc("outofmem", "Out of memory"); } + + // now go ahead and resample + for (j=0; j < z->s->img_y; ++j) { + stbi_uc *out = output + n * z->s->img_x * j; + for (k=0; k < decode_n; ++k) { + stbi__resample *r = &res_comp[k]; + int y_bot = r->ystep >= (r->vs >> 1); + coutput[k] = r->resample(z->img_comp[k].linebuf, + y_bot ? r->line1 : r->line0, + y_bot ? r->line0 : r->line1, + r->w_lores, r->hs); + if (++r->ystep >= r->vs) { + r->ystep = 0; + r->line0 = r->line1; + if (++r->ypos < z->img_comp[k].y) + r->line1 += z->img_comp[k].w2; + } + } + if (n >= 3) { + stbi_uc *y = coutput[0]; + if (z->s->img_n == 3) { + z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n); + } else + for (i=0; i < z->s->img_x; ++i) { + out[0] = out[1] = out[2] = y[i]; + out[3] = 255; // not used if n==3 + out += n; + } + } else { + stbi_uc *y = coutput[0]; + if (n == 1) + for (i=0; i < z->s->img_x; ++i) out[i] = y[i]; + else + for (i=0; i < z->s->img_x; ++i) *out++ = y[i], *out++ = 255; + } + } + stbi__cleanup_jpeg(z); + *out_x = z->s->img_x; + *out_y = z->s->img_y; + if (comp) *comp = z->s->img_n; // report original components, not output + return output; + } +} + +static unsigned char *stbi__jpeg_load(stbi__context *s, int *x, int *y, int *comp, int req_comp) +{ + stbi__jpeg j; + j.s = s; + stbi__setup_jpeg(&j); + return load_jpeg_image(&j, x,y,comp,req_comp); +} + +static int stbi__jpeg_test(stbi__context *s) +{ + int r; + stbi__jpeg j; + j.s = s; + stbi__setup_jpeg(&j); + r = stbi__decode_jpeg_header(&j, STBI__SCAN_type); + stbi__rewind(s); + return r; +} + +static int stbi__jpeg_info_raw(stbi__jpeg *j, int *x, int *y, int *comp) +{ + if (!stbi__decode_jpeg_header(j, STBI__SCAN_header)) { + stbi__rewind( j->s ); + return 0; + } + if (x) *x = j->s->img_x; + if (y) *y = j->s->img_y; + if (comp) *comp = j->s->img_n; + return 1; +} + +static int stbi__jpeg_info(stbi__context *s, int *x, int *y, int *comp) +{ + stbi__jpeg j; + j.s = s; + return stbi__jpeg_info_raw(&j, x, y, comp); +} +#endif + +// public domain zlib decode v0.2 Sean Barrett 2006-11-18 +// simple implementation +// - all input must be provided in an upfront buffer +// - all output is written to a single output buffer (can malloc/realloc) +// performance +// - fast huffman + +#ifndef STBI_NO_ZLIB + +// fast-way is faster to check than jpeg huffman, but slow way is slower +#define STBI__ZFAST_BITS 9 // accelerate all cases in default tables +#define STBI__ZFAST_MASK ((1 << STBI__ZFAST_BITS) - 1) + +// zlib-style huffman encoding +// (jpegs packs from left, zlib from right, so can't share code) +typedef struct +{ + stbi__uint16 fast[1 << STBI__ZFAST_BITS]; + stbi__uint16 firstcode[16]; + int maxcode[17]; + stbi__uint16 firstsymbol[16]; + stbi_uc size[288]; + stbi__uint16 value[288]; +} stbi__zhuffman; + +stbi_inline static int stbi__bitreverse16(int n) +{ + n = ((n & 0xAAAA) >> 1) | ((n & 0x5555) << 1); + n = ((n & 0xCCCC) >> 2) | ((n & 0x3333) << 2); + n = ((n & 0xF0F0) >> 4) | ((n & 0x0F0F) << 4); + n = ((n & 0xFF00) >> 8) | ((n & 0x00FF) << 8); + return n; +} + +stbi_inline static int stbi__bit_reverse(int v, int bits) +{ + STBI_ASSERT(bits <= 16); + // to bit reverse n bits, reverse 16 and shift + // e.g. 11 bits, bit reverse and shift away 5 + return stbi__bitreverse16(v) >> (16-bits); +} + +static int stbi__zbuild_huffman(stbi__zhuffman *z, stbi_uc *sizelist, int num) +{ + int i,k=0; + int code, next_code[16], sizes[17]; + + // DEFLATE spec for generating codes + memset(sizes, 0, sizeof(sizes)); + memset(z->fast, 0, sizeof(z->fast)); + for (i=0; i < num; ++i) + ++sizes[sizelist[i]]; + sizes[0] = 0; + for (i=1; i < 16; ++i) + if (sizes[i] > (1 << i)) + return stbi__err("bad sizes", "Corrupt PNG"); + code = 0; + for (i=1; i < 16; ++i) { + next_code[i] = code; + z->firstcode[i] = (stbi__uint16) code; + z->firstsymbol[i] = (stbi__uint16) k; + code = (code + sizes[i]); + if (sizes[i]) + if (code-1 >= (1 << i)) return stbi__err("bad codelengths","Corrupt PNG"); + z->maxcode[i] = code << (16-i); // preshift for inner loop + code <<= 1; + k += sizes[i]; + } + z->maxcode[16] = 0x10000; // sentinel + for (i=0; i < num; ++i) { + int s = sizelist[i]; + if (s) { + int c = next_code[s] - z->firstcode[s] + z->firstsymbol[s]; + stbi__uint16 fastv = (stbi__uint16) ((s << 9) | i); + z->size [c] = (stbi_uc ) s; + z->value[c] = (stbi__uint16) i; + if (s <= STBI__ZFAST_BITS) { + int j = stbi__bit_reverse(next_code[s],s); + while (j < (1 << STBI__ZFAST_BITS)) { + z->fast[j] = fastv; + j += (1 << s); + } + } + ++next_code[s]; + } + } + return 1; +} + +// zlib-from-memory implementation for PNG reading +// because PNG allows splitting the zlib stream arbitrarily, +// and it's annoying structurally to have PNG call ZLIB call PNG, +// we require PNG read all the IDATs and combine them into a single +// memory buffer + +typedef struct +{ + stbi_uc *zbuffer, *zbuffer_end; + int num_bits; + stbi__uint32 code_buffer; + + char *zout; + char *zout_start; + char *zout_end; + int z_expandable; + + stbi__zhuffman z_length, z_distance; +} stbi__zbuf; + +stbi_inline static stbi_uc stbi__zget8(stbi__zbuf *z) +{ + if (z->zbuffer >= z->zbuffer_end) return 0; + return *z->zbuffer++; +} + +static void stbi__fill_bits(stbi__zbuf *z) +{ + do { + STBI_ASSERT(z->code_buffer < (1U << z->num_bits)); + z->code_buffer |= (unsigned int) stbi__zget8(z) << z->num_bits; + z->num_bits += 8; + } while (z->num_bits <= 24); +} + +stbi_inline static unsigned int stbi__zreceive(stbi__zbuf *z, int n) +{ + unsigned int k; + if (z->num_bits < n) stbi__fill_bits(z); + k = z->code_buffer & ((1 << n) - 1); + z->code_buffer >>= n; + z->num_bits -= n; + return k; +} + +static int stbi__zhuffman_decode_slowpath(stbi__zbuf *a, stbi__zhuffman *z) +{ + int b,s,k; + // not resolved by fast table, so compute it the slow way + // use jpeg approach, which requires MSbits at top + k = stbi__bit_reverse(a->code_buffer, 16); + for (s=STBI__ZFAST_BITS+1; ; ++s) + if (k < z->maxcode[s]) + break; + if (s == 16) return -1; // invalid code! + // code size is s, so: + b = (k >> (16-s)) - z->firstcode[s] + z->firstsymbol[s]; + STBI_ASSERT(z->size[b] == s); + a->code_buffer >>= s; + a->num_bits -= s; + return z->value[b]; +} + +stbi_inline static int stbi__zhuffman_decode(stbi__zbuf *a, stbi__zhuffman *z) +{ + int b,s; + if (a->num_bits < 16) stbi__fill_bits(a); + b = z->fast[a->code_buffer & STBI__ZFAST_MASK]; + if (b) { + s = b >> 9; + a->code_buffer >>= s; + a->num_bits -= s; + return b & 511; + } + return stbi__zhuffman_decode_slowpath(a, z); +} + +static int stbi__zexpand(stbi__zbuf *z, char *zout, int n) // need to make room for n bytes +{ + char *q; + int cur, limit, old_limit; + z->zout = zout; + if (!z->z_expandable) return stbi__err("output buffer limit","Corrupt PNG"); + cur = (int) (z->zout - z->zout_start); + limit = old_limit = (int) (z->zout_end - z->zout_start); + while (cur + n > limit) + limit *= 2; + q = (char *) STBI_REALLOC_SIZED(z->zout_start, old_limit, limit); + STBI_NOTUSED(old_limit); + if (q == NULL) return stbi__err("outofmem", "Out of memory"); + z->zout_start = q; + z->zout = q + cur; + z->zout_end = q + limit; + return 1; +} + +static int stbi__zlength_base[31] = { + 3,4,5,6,7,8,9,10,11,13, + 15,17,19,23,27,31,35,43,51,59, + 67,83,99,115,131,163,195,227,258,0,0 }; + +static int stbi__zlength_extra[31]= +{ 0,0,0,0,0,0,0,0,1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,4,5,5,5,5,0,0,0 }; + +static int stbi__zdist_base[32] = { 1,2,3,4,5,7,9,13,17,25,33,49,65,97,129,193, +257,385,513,769,1025,1537,2049,3073,4097,6145,8193,12289,16385,24577,0,0}; + +static int stbi__zdist_extra[32] = +{ 0,0,0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,13,13}; + +static int stbi__parse_huffman_block(stbi__zbuf *a) +{ + char *zout = a->zout; + for(;;) { + int z = stbi__zhuffman_decode(a, &a->z_length); + if (z < 256) { + if (z < 0) return stbi__err("bad huffman code","Corrupt PNG"); // error in huffman codes + if (zout >= a->zout_end) { + if (!stbi__zexpand(a, zout, 1)) return 0; + zout = a->zout; + } + *zout++ = (char) z; + } else { + stbi_uc *p; + int len,dist; + if (z == 256) { + a->zout = zout; + return 1; + } + z -= 257; + len = stbi__zlength_base[z]; + if (stbi__zlength_extra[z]) len += stbi__zreceive(a, stbi__zlength_extra[z]); + z = stbi__zhuffman_decode(a, &a->z_distance); + if (z < 0) return stbi__err("bad huffman code","Corrupt PNG"); + dist = stbi__zdist_base[z]; + if (stbi__zdist_extra[z]) dist += stbi__zreceive(a, stbi__zdist_extra[z]); + if (zout - a->zout_start < dist) return stbi__err("bad dist","Corrupt PNG"); + if (zout + len > a->zout_end) { + if (!stbi__zexpand(a, zout, len)) return 0; + zout = a->zout; + } + p = (stbi_uc *) (zout - dist); + if (dist == 1) { // run of one byte; common in images. + stbi_uc v = *p; + if (len) { do *zout++ = v; while (--len); } + } else { + if (len) { do *zout++ = *p++; while (--len); } + } + } + } +} + +static int stbi__compute_huffman_codes(stbi__zbuf *a) +{ + static stbi_uc length_dezigzag[19] = { 16,17,18,0,8,7,9,6,10,5,11,4,12,3,13,2,14,1,15 }; + stbi__zhuffman z_codelength; + stbi_uc lencodes[286+32+137];//padding for maximum single op + stbi_uc codelength_sizes[19]; + int i,n; + + int hlit = stbi__zreceive(a,5) + 257; + int hdist = stbi__zreceive(a,5) + 1; + int hclen = stbi__zreceive(a,4) + 4; + + memset(codelength_sizes, 0, sizeof(codelength_sizes)); + for (i=0; i < hclen; ++i) { + int s = stbi__zreceive(a,3); + codelength_sizes[length_dezigzag[i]] = (stbi_uc) s; + } + if (!stbi__zbuild_huffman(&z_codelength, codelength_sizes, 19)) return 0; + + n = 0; + while (n < hlit + hdist) { + int c = stbi__zhuffman_decode(a, &z_codelength); + if (c < 0 || c >= 19) return stbi__err("bad codelengths", "Corrupt PNG"); + if (c < 16) + lencodes[n++] = (stbi_uc) c; + else if (c == 16) { + c = stbi__zreceive(a,2)+3; + memset(lencodes+n, lencodes[n-1], c); + n += c; + } else if (c == 17) { + c = stbi__zreceive(a,3)+3; + memset(lencodes+n, 0, c); + n += c; + } else { + STBI_ASSERT(c == 18); + c = stbi__zreceive(a,7)+11; + memset(lencodes+n, 0, c); + n += c; + } + } + if (n != hlit+hdist) return stbi__err("bad codelengths","Corrupt PNG"); + if (!stbi__zbuild_huffman(&a->z_length, lencodes, hlit)) return 0; + if (!stbi__zbuild_huffman(&a->z_distance, lencodes+hlit, hdist)) return 0; + return 1; +} + +static int stbi__parse_uncomperssed_block(stbi__zbuf *a) +{ + stbi_uc header[4]; + int len,nlen,k; + if (a->num_bits & 7) + stbi__zreceive(a, a->num_bits & 7); // discard + // drain the bit-packed data into header + k = 0; + while (a->num_bits > 0) { + header[k++] = (stbi_uc) (a->code_buffer & 255); // suppress MSVC run-time check + a->code_buffer >>= 8; + a->num_bits -= 8; + } + STBI_ASSERT(a->num_bits == 0); + // now fill header the normal way + while (k < 4) + header[k++] = stbi__zget8(a); + len = header[1] * 256 + header[0]; + nlen = header[3] * 256 + header[2]; + if (nlen != (len ^ 0xffff)) return stbi__err("zlib corrupt","Corrupt PNG"); + if (a->zbuffer + len > a->zbuffer_end) return stbi__err("read past buffer","Corrupt PNG"); + if (a->zout + len > a->zout_end) + if (!stbi__zexpand(a, a->zout, len)) return 0; + memcpy(a->zout, a->zbuffer, len); + a->zbuffer += len; + a->zout += len; + return 1; +} + +static int stbi__parse_zlib_header(stbi__zbuf *a) +{ + int cmf = stbi__zget8(a); + int cm = cmf & 15; + /* int cinfo = cmf >> 4; */ + int flg = stbi__zget8(a); + if ((cmf*256+flg) % 31 != 0) return stbi__err("bad zlib header","Corrupt PNG"); // zlib spec + if (flg & 32) return stbi__err("no preset dict","Corrupt PNG"); // preset dictionary not allowed in png + if (cm != 8) return stbi__err("bad compression","Corrupt PNG"); // DEFLATE required for png + // window = 1 << (8 + cinfo)... but who cares, we fully buffer output + return 1; +} + +// @TODO: should statically initialize these for optimal thread safety +static stbi_uc stbi__zdefault_length[288], stbi__zdefault_distance[32]; +static void stbi__init_zdefaults(void) +{ + int i; // use <= to match clearly with spec + for (i=0; i <= 143; ++i) stbi__zdefault_length[i] = 8; + for ( ; i <= 255; ++i) stbi__zdefault_length[i] = 9; + for ( ; i <= 279; ++i) stbi__zdefault_length[i] = 7; + for ( ; i <= 287; ++i) stbi__zdefault_length[i] = 8; + + for (i=0; i <= 31; ++i) stbi__zdefault_distance[i] = 5; +} + +static int stbi__parse_zlib(stbi__zbuf *a, int parse_header) +{ + int final, type; + if (parse_header) + if (!stbi__parse_zlib_header(a)) return 0; + a->num_bits = 0; + a->code_buffer = 0; + do { + final = stbi__zreceive(a,1); + type = stbi__zreceive(a,2); + if (type == 0) { + if (!stbi__parse_uncomperssed_block(a)) return 0; + } else if (type == 3) { + return 0; + } else { + if (type == 1) { + // use fixed code lengths + if (!stbi__zdefault_distance[31]) stbi__init_zdefaults(); + if (!stbi__zbuild_huffman(&a->z_length , stbi__zdefault_length , 288)) return 0; + if (!stbi__zbuild_huffman(&a->z_distance, stbi__zdefault_distance, 32)) return 0; + } else { + if (!stbi__compute_huffman_codes(a)) return 0; + } + if (!stbi__parse_huffman_block(a)) return 0; + } + } while (!final); + return 1; +} + +static int stbi__do_zlib(stbi__zbuf *a, char *obuf, int olen, int exp, int parse_header) +{ + a->zout_start = obuf; + a->zout = obuf; + a->zout_end = obuf + olen; + a->z_expandable = exp; + + return stbi__parse_zlib(a, parse_header); +} + +STBIDEF char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen) +{ + stbi__zbuf a; + char *p = (char *) stbi__malloc(initial_size); + if (p == NULL) return NULL; + a.zbuffer = (stbi_uc *) buffer; + a.zbuffer_end = (stbi_uc *) buffer + len; + if (stbi__do_zlib(&a, p, initial_size, 1, 1)) { + if (outlen) *outlen = (int) (a.zout - a.zout_start); + return a.zout_start; + } else { + STBI_FREE(a.zout_start); + return NULL; + } +} + +STBIDEF char *stbi_zlib_decode_malloc(char const *buffer, int len, int *outlen) +{ + return stbi_zlib_decode_malloc_guesssize(buffer, len, 16384, outlen); +} + +STBIDEF char *stbi_zlib_decode_malloc_guesssize_headerflag(const char *buffer, int len, int initial_size, int *outlen, int parse_header) +{ + stbi__zbuf a; + char *p = (char *) stbi__malloc(initial_size); + if (p == NULL) return NULL; + a.zbuffer = (stbi_uc *) buffer; + a.zbuffer_end = (stbi_uc *) buffer + len; + if (stbi__do_zlib(&a, p, initial_size, 1, parse_header)) { + if (outlen) *outlen = (int) (a.zout - a.zout_start); + return a.zout_start; + } else { + STBI_FREE(a.zout_start); + return NULL; + } +} + +STBIDEF int stbi_zlib_decode_buffer(char *obuffer, int olen, char const *ibuffer, int ilen) +{ + stbi__zbuf a; + a.zbuffer = (stbi_uc *) ibuffer; + a.zbuffer_end = (stbi_uc *) ibuffer + ilen; + if (stbi__do_zlib(&a, obuffer, olen, 0, 1)) + return (int) (a.zout - a.zout_start); + else + return -1; +} + +STBIDEF char *stbi_zlib_decode_noheader_malloc(char const *buffer, int len, int *outlen) +{ + stbi__zbuf a; + char *p = (char *) stbi__malloc(16384); + if (p == NULL) return NULL; + a.zbuffer = (stbi_uc *) buffer; + a.zbuffer_end = (stbi_uc *) buffer+len; + if (stbi__do_zlib(&a, p, 16384, 1, 0)) { + if (outlen) *outlen = (int) (a.zout - a.zout_start); + return a.zout_start; + } else { + STBI_FREE(a.zout_start); + return NULL; + } +} + +STBIDEF int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen) +{ + stbi__zbuf a; + a.zbuffer = (stbi_uc *) ibuffer; + a.zbuffer_end = (stbi_uc *) ibuffer + ilen; + if (stbi__do_zlib(&a, obuffer, olen, 0, 0)) + return (int) (a.zout - a.zout_start); + else + return -1; +} +#endif + +// public domain "baseline" PNG decoder v0.10 Sean Barrett 2006-11-18 +// simple implementation +// - only 8-bit samples +// - no CRC checking +// - allocates lots of intermediate memory +// - avoids problem of streaming data between subsystems +// - avoids explicit window management +// performance +// - uses stb_zlib, a PD zlib implementation with fast huffman decoding + +#ifndef STBI_NO_PNG +typedef struct +{ + stbi__uint32 length; + stbi__uint32 type; +} stbi__pngchunk; + +static stbi__pngchunk stbi__get_chunk_header(stbi__context *s) +{ + stbi__pngchunk c; + c.length = stbi__get32be(s); + c.type = stbi__get32be(s); + return c; +} + +static int stbi__check_png_header(stbi__context *s) +{ + static stbi_uc png_sig[8] = { 137,80,78,71,13,10,26,10 }; + int i; + for (i=0; i < 8; ++i) + if (stbi__get8(s) != png_sig[i]) return stbi__err("bad png sig","Not a PNG"); + return 1; +} + +typedef struct +{ + stbi__context *s; + stbi_uc *idata, *expanded, *out; +} stbi__png; + + +enum { + STBI__F_none=0, + STBI__F_sub=1, + STBI__F_up=2, + STBI__F_avg=3, + STBI__F_paeth=4, + // synthetic filters used for first scanline to avoid needing a dummy row of 0s + STBI__F_avg_first, + STBI__F_paeth_first +}; + +static stbi_uc first_row_filter[5] = +{ + STBI__F_none, + STBI__F_sub, + STBI__F_none, + STBI__F_avg_first, + STBI__F_paeth_first +}; + +static int stbi__paeth(int a, int b, int c) +{ + int p = a + b - c; + int pa = abs(p-a); + int pb = abs(p-b); + int pc = abs(p-c); + if (pa <= pb && pa <= pc) return a; + if (pb <= pc) return b; + return c; +} + +static stbi_uc stbi__depth_scale_table[9] = { 0, 0xff, 0x55, 0, 0x11, 0,0,0, 0x01 }; + +// create the png data from post-deflated data +static int stbi__create_png_image_raw(stbi__png *a, stbi_uc *raw, stbi__uint32 raw_len, int out_n, stbi__uint32 x, stbi__uint32 y, int depth, int color) +{ + stbi__context *s = a->s; + stbi__uint32 i,j,stride = x*out_n; + stbi__uint32 img_len, img_width_bytes; + int k; + int img_n = s->img_n; // copy it into a local for later + + STBI_ASSERT(out_n == s->img_n || out_n == s->img_n+1); + a->out = (stbi_uc *) stbi__malloc(x * y * out_n); // extra bytes to write off the end into + if (!a->out) return stbi__err("outofmem", "Out of memory"); + + img_width_bytes = (((img_n * x * depth) + 7) >> 3); + img_len = (img_width_bytes + 1) * y; + if (s->img_x == x && s->img_y == y) { + if (raw_len != img_len) return stbi__err("not enough pixels","Corrupt PNG"); + } else { // interlaced: + if (raw_len < img_len) return stbi__err("not enough pixels","Corrupt PNG"); + } + + for (j=0; j < y; ++j) { + stbi_uc *cur = a->out + stride*j; + stbi_uc *prior = cur - stride; + int filter = *raw++; + int filter_bytes = img_n; + int width = x; + if (filter > 4) + return stbi__err("invalid filter","Corrupt PNG"); + + if (depth < 8) { + STBI_ASSERT(img_width_bytes <= x); + cur += x*out_n - img_width_bytes; // store output to the rightmost img_len bytes, so we can decode in place + filter_bytes = 1; + width = img_width_bytes; + } + + // if first row, use special filter that doesn't sample previous row + if (j == 0) filter = first_row_filter[filter]; + + // handle first byte explicitly + for (k=0; k < filter_bytes; ++k) { + switch (filter) { + case STBI__F_none : cur[k] = raw[k]; break; + case STBI__F_sub : cur[k] = raw[k]; break; + case STBI__F_up : cur[k] = STBI__BYTECAST(raw[k] + prior[k]); break; + case STBI__F_avg : cur[k] = STBI__BYTECAST(raw[k] + (prior[k]>>1)); break; + case STBI__F_paeth : cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(0,prior[k],0)); break; + case STBI__F_avg_first : cur[k] = raw[k]; break; + case STBI__F_paeth_first: cur[k] = raw[k]; break; + } + } + + if (depth == 8) { + if (img_n != out_n) + cur[img_n] = 255; // first pixel + raw += img_n; + cur += out_n; + prior += out_n; + } else { + raw += 1; + cur += 1; + prior += 1; + } + + // this is a little gross, so that we don't switch per-pixel or per-component + if (depth < 8 || img_n == out_n) { + int nk = (width - 1)*img_n; + #define CASE(f) \ + case f: \ + for (k=0; k < nk; ++k) + switch (filter) { + // "none" filter turns into a memcpy here; make that explicit. + case STBI__F_none: memcpy(cur, raw, nk); break; + CASE(STBI__F_sub) cur[k] = STBI__BYTECAST(raw[k] + cur[k-filter_bytes]); break; + CASE(STBI__F_up) cur[k] = STBI__BYTECAST(raw[k] + prior[k]); break; + CASE(STBI__F_avg) cur[k] = STBI__BYTECAST(raw[k] + ((prior[k] + cur[k-filter_bytes])>>1)); break; + CASE(STBI__F_paeth) cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-filter_bytes],prior[k],prior[k-filter_bytes])); break; + CASE(STBI__F_avg_first) cur[k] = STBI__BYTECAST(raw[k] + (cur[k-filter_bytes] >> 1)); break; + CASE(STBI__F_paeth_first) cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-filter_bytes],0,0)); break; + } + #undef CASE + raw += nk; + } else { + STBI_ASSERT(img_n+1 == out_n); + #define CASE(f) \ + case f: \ + for (i=x-1; i >= 1; --i, cur[img_n]=255,raw+=img_n,cur+=out_n,prior+=out_n) \ + for (k=0; k < img_n; ++k) + switch (filter) { + CASE(STBI__F_none) cur[k] = raw[k]; break; + CASE(STBI__F_sub) cur[k] = STBI__BYTECAST(raw[k] + cur[k-out_n]); break; + CASE(STBI__F_up) cur[k] = STBI__BYTECAST(raw[k] + prior[k]); break; + CASE(STBI__F_avg) cur[k] = STBI__BYTECAST(raw[k] + ((prior[k] + cur[k-out_n])>>1)); break; + CASE(STBI__F_paeth) cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-out_n],prior[k],prior[k-out_n])); break; + CASE(STBI__F_avg_first) cur[k] = STBI__BYTECAST(raw[k] + (cur[k-out_n] >> 1)); break; + CASE(STBI__F_paeth_first) cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-out_n],0,0)); break; + } + #undef CASE + } + } + + // we make a separate pass to expand bits to pixels; for performance, + // this could run two scanlines behind the above code, so it won't + // intefere with filtering but will still be in the cache. + if (depth < 8) { + for (j=0; j < y; ++j) { + stbi_uc *cur = a->out + stride*j; + stbi_uc *in = a->out + stride*j + x*out_n - img_width_bytes; + // unpack 1/2/4-bit into a 8-bit buffer. allows us to keep the common 8-bit path optimal at minimal cost for 1/2/4-bit + // png guarante byte alignment, if width is not multiple of 8/4/2 we'll decode dummy trailing data that will be skipped in the later loop + stbi_uc scale = (color == 0) ? stbi__depth_scale_table[depth] : 1; // scale grayscale values to 0..255 range + + // note that the final byte might overshoot and write more data than desired. + // we can allocate enough data that this never writes out of memory, but it + // could also overwrite the next scanline. can it overwrite non-empty data + // on the next scanline? yes, consider 1-pixel-wide scanlines with 1-bit-per-pixel. + // so we need to explicitly clamp the final ones + + if (depth == 4) { + for (k=x*img_n; k >= 2; k-=2, ++in) { + *cur++ = scale * ((*in >> 4) ); + *cur++ = scale * ((*in ) & 0x0f); + } + if (k > 0) *cur++ = scale * ((*in >> 4) ); + } else if (depth == 2) { + for (k=x*img_n; k >= 4; k-=4, ++in) { + *cur++ = scale * ((*in >> 6) ); + *cur++ = scale * ((*in >> 4) & 0x03); + *cur++ = scale * ((*in >> 2) & 0x03); + *cur++ = scale * ((*in ) & 0x03); + } + if (k > 0) *cur++ = scale * ((*in >> 6) ); + if (k > 1) *cur++ = scale * ((*in >> 4) & 0x03); + if (k > 2) *cur++ = scale * ((*in >> 2) & 0x03); + } else if (depth == 1) { + for (k=x*img_n; k >= 8; k-=8, ++in) { + *cur++ = scale * ((*in >> 7) ); + *cur++ = scale * ((*in >> 6) & 0x01); + *cur++ = scale * ((*in >> 5) & 0x01); + *cur++ = scale * ((*in >> 4) & 0x01); + *cur++ = scale * ((*in >> 3) & 0x01); + *cur++ = scale * ((*in >> 2) & 0x01); + *cur++ = scale * ((*in >> 1) & 0x01); + *cur++ = scale * ((*in ) & 0x01); + } + if (k > 0) *cur++ = scale * ((*in >> 7) ); + if (k > 1) *cur++ = scale * ((*in >> 6) & 0x01); + if (k > 2) *cur++ = scale * ((*in >> 5) & 0x01); + if (k > 3) *cur++ = scale * ((*in >> 4) & 0x01); + if (k > 4) *cur++ = scale * ((*in >> 3) & 0x01); + if (k > 5) *cur++ = scale * ((*in >> 2) & 0x01); + if (k > 6) *cur++ = scale * ((*in >> 1) & 0x01); + } + if (img_n != out_n) { + int q; + // insert alpha = 255 + cur = a->out + stride*j; + if (img_n == 1) { + for (q=x-1; q >= 0; --q) { + cur[q*2+1] = 255; + cur[q*2+0] = cur[q]; + } + } else { + STBI_ASSERT(img_n == 3); + for (q=x-1; q >= 0; --q) { + cur[q*4+3] = 255; + cur[q*4+2] = cur[q*3+2]; + cur[q*4+1] = cur[q*3+1]; + cur[q*4+0] = cur[q*3+0]; + } + } + } + } + } + + return 1; +} + +static int stbi__create_png_image(stbi__png *a, stbi_uc *image_data, stbi__uint32 image_data_len, int out_n, int depth, int color, int interlaced) +{ + stbi_uc *final; + int p; + if (!interlaced) + return stbi__create_png_image_raw(a, image_data, image_data_len, out_n, a->s->img_x, a->s->img_y, depth, color); + + // de-interlacing + final = (stbi_uc *) stbi__malloc(a->s->img_x * a->s->img_y * out_n); + for (p=0; p < 7; ++p) { + int xorig[] = { 0,4,0,2,0,1,0 }; + int yorig[] = { 0,0,4,0,2,0,1 }; + int xspc[] = { 8,8,4,4,2,2,1 }; + int yspc[] = { 8,8,8,4,4,2,2 }; + int i,j,x,y; + // pass1_x[4] = 0, pass1_x[5] = 1, pass1_x[12] = 1 + x = (a->s->img_x - xorig[p] + xspc[p]-1) / xspc[p]; + y = (a->s->img_y - yorig[p] + yspc[p]-1) / yspc[p]; + if (x && y) { + stbi__uint32 img_len = ((((a->s->img_n * x * depth) + 7) >> 3) + 1) * y; + if (!stbi__create_png_image_raw(a, image_data, image_data_len, out_n, x, y, depth, color)) { + STBI_FREE(final); + return 0; + } + for (j=0; j < y; ++j) { + for (i=0; i < x; ++i) { + int out_y = j*yspc[p]+yorig[p]; + int out_x = i*xspc[p]+xorig[p]; + memcpy(final + out_y*a->s->img_x*out_n + out_x*out_n, + a->out + (j*x+i)*out_n, out_n); + } + } + STBI_FREE(a->out); + image_data += img_len; + image_data_len -= img_len; + } + } + a->out = final; + + return 1; +} + +static int stbi__compute_transparency(stbi__png *z, stbi_uc tc[3], int out_n) +{ + stbi__context *s = z->s; + stbi__uint32 i, pixel_count = s->img_x * s->img_y; + stbi_uc *p = z->out; + + // compute color-based transparency, assuming we've + // already got 255 as the alpha value in the output + STBI_ASSERT(out_n == 2 || out_n == 4); + + if (out_n == 2) { + for (i=0; i < pixel_count; ++i) { + p[1] = (p[0] == tc[0] ? 0 : 255); + p += 2; + } + } else { + for (i=0; i < pixel_count; ++i) { + if (p[0] == tc[0] && p[1] == tc[1] && p[2] == tc[2]) + p[3] = 0; + p += 4; + } + } + return 1; +} + +static int stbi__expand_png_palette(stbi__png *a, stbi_uc *palette, int len, int pal_img_n) +{ + stbi__uint32 i, pixel_count = a->s->img_x * a->s->img_y; + stbi_uc *p, *temp_out, *orig = a->out; + + p = (stbi_uc *) stbi__malloc(pixel_count * pal_img_n); + if (p == NULL) return stbi__err("outofmem", "Out of memory"); + + // between here and free(out) below, exitting would leak + temp_out = p; + + if (pal_img_n == 3) { + for (i=0; i < pixel_count; ++i) { + int n = orig[i]*4; + p[0] = palette[n ]; + p[1] = palette[n+1]; + p[2] = palette[n+2]; + p += 3; + } + } else { + for (i=0; i < pixel_count; ++i) { + int n = orig[i]*4; + p[0] = palette[n ]; + p[1] = palette[n+1]; + p[2] = palette[n+2]; + p[3] = palette[n+3]; + p += 4; + } + } + STBI_FREE(a->out); + a->out = temp_out; + + STBI_NOTUSED(len); + + return 1; +} + +static int stbi__unpremultiply_on_load = 0; +static int stbi__de_iphone_flag = 0; + +STBIDEF void stbi_set_unpremultiply_on_load(int flag_true_if_should_unpremultiply) +{ + stbi__unpremultiply_on_load = flag_true_if_should_unpremultiply; +} + +STBIDEF void stbi_convert_iphone_png_to_rgb(int flag_true_if_should_convert) +{ + stbi__de_iphone_flag = flag_true_if_should_convert; +} + +static void stbi__de_iphone(stbi__png *z) +{ + stbi__context *s = z->s; + stbi__uint32 i, pixel_count = s->img_x * s->img_y; + stbi_uc *p = z->out; + + if (s->img_out_n == 3) { // convert bgr to rgb + for (i=0; i < pixel_count; ++i) { + stbi_uc t = p[0]; + p[0] = p[2]; + p[2] = t; + p += 3; + } + } else { + STBI_ASSERT(s->img_out_n == 4); + if (stbi__unpremultiply_on_load) { + // convert bgr to rgb and unpremultiply + for (i=0; i < pixel_count; ++i) { + stbi_uc a = p[3]; + stbi_uc t = p[0]; + if (a) { + p[0] = p[2] * 255 / a; + p[1] = p[1] * 255 / a; + p[2] = t * 255 / a; + } else { + p[0] = p[2]; + p[2] = t; + } + p += 4; + } + } else { + // convert bgr to rgb + for (i=0; i < pixel_count; ++i) { + stbi_uc t = p[0]; + p[0] = p[2]; + p[2] = t; + p += 4; + } + } + } +} + +#define STBI__PNG_TYPE(a,b,c,d) (((a) << 24) + ((b) << 16) + ((c) << 8) + (d)) + +static int stbi__parse_png_file(stbi__png *z, int scan, int req_comp) +{ + stbi_uc palette[1024], pal_img_n=0; + stbi_uc has_trans=0, tc[3]; + stbi__uint32 ioff=0, idata_limit=0, i, pal_len=0; + int first=1,k,interlace=0, color=0, depth=0, is_iphone=0; + stbi__context *s = z->s; + + z->expanded = NULL; + z->idata = NULL; + z->out = NULL; + + if (!stbi__check_png_header(s)) return 0; + + if (scan == STBI__SCAN_type) return 1; + + for (;;) { + stbi__pngchunk c = stbi__get_chunk_header(s); + switch (c.type) { + case STBI__PNG_TYPE('C','g','B','I'): + is_iphone = 1; + stbi__skip(s, c.length); + break; + case STBI__PNG_TYPE('I','H','D','R'): { + int comp,filter; + if (!first) return stbi__err("multiple IHDR","Corrupt PNG"); + first = 0; + if (c.length != 13) return stbi__err("bad IHDR len","Corrupt PNG"); + s->img_x = stbi__get32be(s); if (s->img_x > (1 << 24)) return stbi__err("too large","Very large image (corrupt?)"); + s->img_y = stbi__get32be(s); if (s->img_y > (1 << 24)) return stbi__err("too large","Very large image (corrupt?)"); + depth = stbi__get8(s); if (depth != 1 && depth != 2 && depth != 4 && depth != 8) return stbi__err("1/2/4/8-bit only","PNG not supported: 1/2/4/8-bit only"); + color = stbi__get8(s); if (color > 6) return stbi__err("bad ctype","Corrupt PNG"); + if (color == 3) pal_img_n = 3; else if (color & 1) return stbi__err("bad ctype","Corrupt PNG"); + comp = stbi__get8(s); if (comp) return stbi__err("bad comp method","Corrupt PNG"); + filter= stbi__get8(s); if (filter) return stbi__err("bad filter method","Corrupt PNG"); + interlace = stbi__get8(s); if (interlace>1) return stbi__err("bad interlace method","Corrupt PNG"); + if (!s->img_x || !s->img_y) return stbi__err("0-pixel image","Corrupt PNG"); + if (!pal_img_n) { + s->img_n = (color & 2 ? 3 : 1) + (color & 4 ? 1 : 0); + if ((1 << 30) / s->img_x / s->img_n < s->img_y) return stbi__err("too large", "Image too large to decode"); + if (scan == STBI__SCAN_header) return 1; + } else { + // if paletted, then pal_n is our final components, and + // img_n is # components to decompress/filter. + s->img_n = 1; + if ((1 << 30) / s->img_x / 4 < s->img_y) return stbi__err("too large","Corrupt PNG"); + // if SCAN_header, have to scan to see if we have a tRNS + } + break; + } + + case STBI__PNG_TYPE('P','L','T','E'): { + if (first) return stbi__err("first not IHDR", "Corrupt PNG"); + if (c.length > 256*3) return stbi__err("invalid PLTE","Corrupt PNG"); + pal_len = c.length / 3; + if (pal_len * 3 != c.length) return stbi__err("invalid PLTE","Corrupt PNG"); + for (i=0; i < pal_len; ++i) { + palette[i*4+0] = stbi__get8(s); + palette[i*4+1] = stbi__get8(s); + palette[i*4+2] = stbi__get8(s); + palette[i*4+3] = 255; + } + break; + } + + case STBI__PNG_TYPE('t','R','N','S'): { + if (first) return stbi__err("first not IHDR", "Corrupt PNG"); + if (z->idata) return stbi__err("tRNS after IDAT","Corrupt PNG"); + if (pal_img_n) { + if (scan == STBI__SCAN_header) { s->img_n = 4; return 1; } + if (pal_len == 0) return stbi__err("tRNS before PLTE","Corrupt PNG"); + if (c.length > pal_len) return stbi__err("bad tRNS len","Corrupt PNG"); + pal_img_n = 4; + for (i=0; i < c.length; ++i) + palette[i*4+3] = stbi__get8(s); + } else { + if (!(s->img_n & 1)) return stbi__err("tRNS with alpha","Corrupt PNG"); + if (c.length != (stbi__uint32) s->img_n*2) return stbi__err("bad tRNS len","Corrupt PNG"); + has_trans = 1; + for (k=0; k < s->img_n; ++k) + tc[k] = (stbi_uc) (stbi__get16be(s) & 255) * stbi__depth_scale_table[depth]; // non 8-bit images will be larger + } + break; + } + + case STBI__PNG_TYPE('I','D','A','T'): { + if (first) return stbi__err("first not IHDR", "Corrupt PNG"); + if (pal_img_n && !pal_len) return stbi__err("no PLTE","Corrupt PNG"); + if (scan == STBI__SCAN_header) { s->img_n = pal_img_n; return 1; } + if ((int)(ioff + c.length) < (int)ioff) return 0; + if (ioff + c.length > idata_limit) { + stbi__uint32 idata_limit_old = idata_limit; + stbi_uc *p; + if (idata_limit == 0) idata_limit = c.length > 4096 ? c.length : 4096; + while (ioff + c.length > idata_limit) + idata_limit *= 2; + STBI_NOTUSED(idata_limit_old); + p = (stbi_uc *) STBI_REALLOC_SIZED(z->idata, idata_limit_old, idata_limit); if (p == NULL) return stbi__err("outofmem", "Out of memory"); + z->idata = p; + } + if (!stbi__getn(s, z->idata+ioff,c.length)) return stbi__err("outofdata","Corrupt PNG"); + ioff += c.length; + break; + } + + case STBI__PNG_TYPE('I','E','N','D'): { + stbi__uint32 raw_len, bpl; + if (first) return stbi__err("first not IHDR", "Corrupt PNG"); + if (scan != STBI__SCAN_load) return 1; + if (z->idata == NULL) return stbi__err("no IDAT","Corrupt PNG"); + // initial guess for decoded data size to avoid unnecessary reallocs + bpl = (s->img_x * depth + 7) / 8; // bytes per line, per component + raw_len = bpl * s->img_y * s->img_n /* pixels */ + s->img_y /* filter mode per row */; + z->expanded = (stbi_uc *) stbi_zlib_decode_malloc_guesssize_headerflag((char *) z->idata, ioff, raw_len, (int *) &raw_len, !is_iphone); + if (z->expanded == NULL) return 0; // zlib should set error + STBI_FREE(z->idata); z->idata = NULL; + if ((req_comp == s->img_n+1 && req_comp != 3 && !pal_img_n) || has_trans) + s->img_out_n = s->img_n+1; + else + s->img_out_n = s->img_n; + if (!stbi__create_png_image(z, z->expanded, raw_len, s->img_out_n, depth, color, interlace)) return 0; + if (has_trans) + if (!stbi__compute_transparency(z, tc, s->img_out_n)) return 0; + if (is_iphone && stbi__de_iphone_flag && s->img_out_n > 2) + stbi__de_iphone(z); + if (pal_img_n) { + // pal_img_n == 3 or 4 + s->img_n = pal_img_n; // record the actual colors we had + s->img_out_n = pal_img_n; + if (req_comp >= 3) s->img_out_n = req_comp; + if (!stbi__expand_png_palette(z, palette, pal_len, s->img_out_n)) + return 0; + } + STBI_FREE(z->expanded); z->expanded = NULL; + return 1; + } + + default: + // if critical, fail + if (first) return stbi__err("first not IHDR", "Corrupt PNG"); + if ((c.type & (1 << 29)) == 0) { + #ifndef STBI_NO_FAILURE_STRINGS + // not threadsafe + static char invalid_chunk[] = "XXXX PNG chunk not known"; + invalid_chunk[0] = STBI__BYTECAST(c.type >> 24); + invalid_chunk[1] = STBI__BYTECAST(c.type >> 16); + invalid_chunk[2] = STBI__BYTECAST(c.type >> 8); + invalid_chunk[3] = STBI__BYTECAST(c.type >> 0); + #endif + return stbi__err(invalid_chunk, "PNG not supported: unknown PNG chunk type"); + } + stbi__skip(s, c.length); + break; + } + // end of PNG chunk, read and skip CRC + stbi__get32be(s); + } +} + +static unsigned char *stbi__do_png(stbi__png *p, int *x, int *y, int *n, int req_comp) +{ + unsigned char *result=NULL; + if (req_comp < 0 || req_comp > 4) return stbi__errpuc("bad req_comp", "Internal error"); + if (stbi__parse_png_file(p, STBI__SCAN_load, req_comp)) { + result = p->out; + p->out = NULL; + if (req_comp && req_comp != p->s->img_out_n) { + result = stbi__convert_format(result, p->s->img_out_n, req_comp, p->s->img_x, p->s->img_y); + p->s->img_out_n = req_comp; + if (result == NULL) return result; + } + *x = p->s->img_x; + *y = p->s->img_y; + if (n) *n = p->s->img_out_n; + } + STBI_FREE(p->out); p->out = NULL; + STBI_FREE(p->expanded); p->expanded = NULL; + STBI_FREE(p->idata); p->idata = NULL; + + return result; +} + +static unsigned char *stbi__png_load(stbi__context *s, int *x, int *y, int *comp, int req_comp) +{ + stbi__png p; + p.s = s; + return stbi__do_png(&p, x,y,comp,req_comp); +} + +static int stbi__png_test(stbi__context *s) +{ + int r; + r = stbi__check_png_header(s); + stbi__rewind(s); + return r; +} + +static int stbi__png_info_raw(stbi__png *p, int *x, int *y, int *comp) +{ + if (!stbi__parse_png_file(p, STBI__SCAN_header, 0)) { + stbi__rewind( p->s ); + return 0; + } + if (x) *x = p->s->img_x; + if (y) *y = p->s->img_y; + if (comp) *comp = p->s->img_n; + return 1; +} + +static int stbi__png_info(stbi__context *s, int *x, int *y, int *comp) +{ + stbi__png p; + p.s = s; + return stbi__png_info_raw(&p, x, y, comp); +} +#endif + +// Microsoft/Windows BMP image + +#ifndef STBI_NO_BMP +static int stbi__bmp_test_raw(stbi__context *s) +{ + int r; + int sz; + if (stbi__get8(s) != 'B') return 0; + if (stbi__get8(s) != 'M') return 0; + stbi__get32le(s); // discard filesize + stbi__get16le(s); // discard reserved + stbi__get16le(s); // discard reserved + stbi__get32le(s); // discard data offset + sz = stbi__get32le(s); + r = (sz == 12 || sz == 40 || sz == 56 || sz == 108 || sz == 124); + return r; +} + +static int stbi__bmp_test(stbi__context *s) +{ + int r = stbi__bmp_test_raw(s); + stbi__rewind(s); + return r; +} + + +// returns 0..31 for the highest set bit +static int stbi__high_bit(unsigned int z) +{ + int n=0; + if (z == 0) return -1; + if (z >= 0x10000) n += 16, z >>= 16; + if (z >= 0x00100) n += 8, z >>= 8; + if (z >= 0x00010) n += 4, z >>= 4; + if (z >= 0x00004) n += 2, z >>= 2; + if (z >= 0x00002) n += 1, z >>= 1; + return n; +} + +static int stbi__bitcount(unsigned int a) +{ + a = (a & 0x55555555) + ((a >> 1) & 0x55555555); // max 2 + a = (a & 0x33333333) + ((a >> 2) & 0x33333333); // max 4 + a = (a + (a >> 4)) & 0x0f0f0f0f; // max 8 per 4, now 8 bits + a = (a + (a >> 8)); // max 16 per 8 bits + a = (a + (a >> 16)); // max 32 per 8 bits + return a & 0xff; +} + +static int stbi__shiftsigned(int v, int shift, int bits) +{ + int result; + int z=0; + + if (shift < 0) v <<= -shift; + else v >>= shift; + result = v; + + z = bits; + while (z < 8) { + result += v >> z; + z += bits; + } + return result; +} + +typedef struct +{ + int bpp, offset, hsz; + unsigned int mr,mg,mb,ma, all_a; +} stbi__bmp_data; + +static void *stbi__bmp_parse_header(stbi__context *s, stbi__bmp_data *info) +{ + int hsz; + if (stbi__get8(s) != 'B' || stbi__get8(s) != 'M') return stbi__errpuc("not BMP", "Corrupt BMP"); + stbi__get32le(s); // discard filesize + stbi__get16le(s); // discard reserved + stbi__get16le(s); // discard reserved + info->offset = stbi__get32le(s); + info->hsz = hsz = stbi__get32le(s); + + if (hsz != 12 && hsz != 40 && hsz != 56 && hsz != 108 && hsz != 124) return stbi__errpuc("unknown BMP", "BMP type not supported: unknown"); + if (hsz == 12) { + s->img_x = stbi__get16le(s); + s->img_y = stbi__get16le(s); + } else { + s->img_x = stbi__get32le(s); + s->img_y = stbi__get32le(s); + } + if (stbi__get16le(s) != 1) return stbi__errpuc("bad BMP", "bad BMP"); + info->bpp = stbi__get16le(s); + if (info->bpp == 1) return stbi__errpuc("monochrome", "BMP type not supported: 1-bit"); + if (hsz != 12) { + int compress = stbi__get32le(s); + if (compress == 1 || compress == 2) return stbi__errpuc("BMP RLE", "BMP type not supported: RLE"); + stbi__get32le(s); // discard sizeof + stbi__get32le(s); // discard hres + stbi__get32le(s); // discard vres + stbi__get32le(s); // discard colorsused + stbi__get32le(s); // discard max important + if (hsz == 40 || hsz == 56) { + if (hsz == 56) { + stbi__get32le(s); + stbi__get32le(s); + stbi__get32le(s); + stbi__get32le(s); + } + if (info->bpp == 16 || info->bpp == 32) { + info->mr = info->mg = info->mb = 0; + if (compress == 0) { + if (info->bpp == 32) { + info->mr = 0xffu << 16; + info->mg = 0xffu << 8; + info->mb = 0xffu << 0; + info->ma = 0xffu << 24; + info->all_a = 0; // if all_a is 0 at end, then we loaded alpha channel but it was all 0 + } else { + info->mr = 31u << 10; + info->mg = 31u << 5; + info->mb = 31u << 0; + } + } else if (compress == 3) { + info->mr = stbi__get32le(s); + info->mg = stbi__get32le(s); + info->mb = stbi__get32le(s); + // not documented, but generated by photoshop and handled by mspaint + if (info->mr == info->mg && info->mg == info->mb) { + // ?!?!? + return stbi__errpuc("bad BMP", "bad BMP"); + } + } else + return stbi__errpuc("bad BMP", "bad BMP"); + } + } else { + int i; + if (hsz != 108 && hsz != 124) + return stbi__errpuc("bad BMP", "bad BMP"); + info->mr = stbi__get32le(s); + info->mg = stbi__get32le(s); + info->mb = stbi__get32le(s); + info->ma = stbi__get32le(s); + stbi__get32le(s); // discard color space + for (i=0; i < 12; ++i) + stbi__get32le(s); // discard color space parameters + if (hsz == 124) { + stbi__get32le(s); // discard rendering intent + stbi__get32le(s); // discard offset of profile data + stbi__get32le(s); // discard size of profile data + stbi__get32le(s); // discard reserved + } + } + } + return (void *) 1; +} + + +static stbi_uc *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int req_comp) +{ + stbi_uc *out; + unsigned int mr=0,mg=0,mb=0,ma=0, all_a; + stbi_uc pal[256][4]; + int psize=0,i,j,width; + int flip_vertically, pad, target; + stbi__bmp_data info; + + info.all_a = 255; + if (stbi__bmp_parse_header(s, &info) == NULL) + return NULL; // error code already set + + flip_vertically = ((int) s->img_y) > 0; + s->img_y = abs((int) s->img_y); + + mr = info.mr; + mg = info.mg; + mb = info.mb; + ma = info.ma; + all_a = info.all_a; + + if (info.hsz == 12) { + if (info.bpp < 24) + psize = (info.offset - 14 - 24) / 3; + } else { + if (info.bpp < 16) + psize = (info.offset - 14 - info.hsz) >> 2; + } + + s->img_n = ma ? 4 : 3; + if (req_comp && req_comp >= 3) // we can directly decode 3 or 4 + target = req_comp; + else + target = s->img_n; // if they want monochrome, we'll post-convert + + out = (stbi_uc *) stbi__malloc(target * s->img_x * s->img_y); + if (!out) return stbi__errpuc("outofmem", "Out of memory"); + if (info.bpp < 16) { + int z=0; + if (psize == 0 || psize > 256) { STBI_FREE(out); return stbi__errpuc("invalid", "Corrupt BMP"); } + for (i=0; i < psize; ++i) { + pal[i][2] = stbi__get8(s); + pal[i][1] = stbi__get8(s); + pal[i][0] = stbi__get8(s); + if (info.hsz != 12) stbi__get8(s); + pal[i][3] = 255; + } + stbi__skip(s, info.offset - 14 - info.hsz - psize * (info.hsz == 12 ? 3 : 4)); + if (info.bpp == 4) width = (s->img_x + 1) >> 1; + else if (info.bpp == 8) width = s->img_x; + else { STBI_FREE(out); return stbi__errpuc("bad bpp", "Corrupt BMP"); } + pad = (-width)&3; + for (j=0; j < (int) s->img_y; ++j) { + for (i=0; i < (int) s->img_x; i += 2) { + int v=stbi__get8(s),v2=0; + if (info.bpp == 4) { + v2 = v & 15; + v >>= 4; + } + out[z++] = pal[v][0]; + out[z++] = pal[v][1]; + out[z++] = pal[v][2]; + if (target == 4) out[z++] = 255; + if (i+1 == (int) s->img_x) break; + v = (info.bpp == 8) ? stbi__get8(s) : v2; + out[z++] = pal[v][0]; + out[z++] = pal[v][1]; + out[z++] = pal[v][2]; + if (target == 4) out[z++] = 255; + } + stbi__skip(s, pad); + } + } else { + int rshift=0,gshift=0,bshift=0,ashift=0,rcount=0,gcount=0,bcount=0,acount=0; + int z = 0; + int easy=0; + stbi__skip(s, info.offset - 14 - info.hsz); + if (info.bpp == 24) width = 3 * s->img_x; + else if (info.bpp == 16) width = 2*s->img_x; + else /* bpp = 32 and pad = 0 */ width=0; + pad = (-width) & 3; + if (info.bpp == 24) { + easy = 1; + } else if (info.bpp == 32) { + if (mb == 0xff && mg == 0xff00 && mr == 0x00ff0000 && ma == 0xff000000) + easy = 2; + } + if (!easy) { + if (!mr || !mg || !mb) { STBI_FREE(out); return stbi__errpuc("bad masks", "Corrupt BMP"); } + // right shift amt to put high bit in position #7 + rshift = stbi__high_bit(mr)-7; rcount = stbi__bitcount(mr); + gshift = stbi__high_bit(mg)-7; gcount = stbi__bitcount(mg); + bshift = stbi__high_bit(mb)-7; bcount = stbi__bitcount(mb); + ashift = stbi__high_bit(ma)-7; acount = stbi__bitcount(ma); + } + for (j=0; j < (int) s->img_y; ++j) { + if (easy) { + for (i=0; i < (int) s->img_x; ++i) { + unsigned char a; + out[z+2] = stbi__get8(s); + out[z+1] = stbi__get8(s); + out[z+0] = stbi__get8(s); + z += 3; + a = (easy == 2 ? stbi__get8(s) : 255); + all_a |= a; + if (target == 4) out[z++] = a; + } + } else { + int bpp = info.bpp; + for (i=0; i < (int) s->img_x; ++i) { + stbi__uint32 v = (bpp == 16 ? (stbi__uint32) stbi__get16le(s) : stbi__get32le(s)); + int a; + out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mr, rshift, rcount)); + out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mg, gshift, gcount)); + out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mb, bshift, bcount)); + a = (ma ? stbi__shiftsigned(v & ma, ashift, acount) : 255); + all_a |= a; + if (target == 4) out[z++] = STBI__BYTECAST(a); + } + } + stbi__skip(s, pad); + } + } + + // if alpha channel is all 0s, replace with all 255s + if (target == 4 && all_a == 0) + for (i=4*s->img_x*s->img_y-1; i >= 0; i -= 4) + out[i] = 255; + + if (flip_vertically) { + stbi_uc t; + for (j=0; j < (int) s->img_y>>1; ++j) { + stbi_uc *p1 = out + j *s->img_x*target; + stbi_uc *p2 = out + (s->img_y-1-j)*s->img_x*target; + for (i=0; i < (int) s->img_x*target; ++i) { + t = p1[i], p1[i] = p2[i], p2[i] = t; + } + } + } + + if (req_comp && req_comp != target) { + out = stbi__convert_format(out, target, req_comp, s->img_x, s->img_y); + if (out == NULL) return out; // stbi__convert_format frees input on failure + } + + *x = s->img_x; + *y = s->img_y; + if (comp) *comp = s->img_n; + return out; +} +#endif + +// Targa Truevision - TGA +// by Jonathan Dummer +#ifndef STBI_NO_TGA +// returns STBI_rgb or whatever, 0 on error +static int stbi__tga_get_comp(int bits_per_pixel, int is_grey, int* is_rgb16) +{ + // only RGB or RGBA (incl. 16bit) or grey allowed + if(is_rgb16) *is_rgb16 = 0; + switch(bits_per_pixel) { + case 8: return STBI_grey; + case 16: if(is_grey) return STBI_grey_alpha; + // else: fall-through + case 15: if(is_rgb16) *is_rgb16 = 1; + return STBI_rgb; + case 24: // fall-through + case 32: return bits_per_pixel/8; + default: return 0; + } +} + +static int stbi__tga_info(stbi__context *s, int *x, int *y, int *comp) +{ + int tga_w, tga_h, tga_comp, tga_image_type, tga_bits_per_pixel, tga_colormap_bpp; + int sz, tga_colormap_type; + stbi__get8(s); // discard Offset + tga_colormap_type = stbi__get8(s); // colormap type + if( tga_colormap_type > 1 ) { + stbi__rewind(s); + return 0; // only RGB or indexed allowed + } + tga_image_type = stbi__get8(s); // image type + if ( tga_colormap_type == 1 ) { // colormapped (paletted) image + if (tga_image_type != 1 && tga_image_type != 9) { + stbi__rewind(s); + return 0; + } + stbi__skip(s,4); // skip index of first colormap entry and number of entries + sz = stbi__get8(s); // check bits per palette color entry + if ( (sz != 8) && (sz != 15) && (sz != 16) && (sz != 24) && (sz != 32) ) { + stbi__rewind(s); + return 0; + } + stbi__skip(s,4); // skip image x and y origin + tga_colormap_bpp = sz; + } else { // "normal" image w/o colormap - only RGB or grey allowed, +/- RLE + if ( (tga_image_type != 2) && (tga_image_type != 3) && (tga_image_type != 10) && (tga_image_type != 11) ) { + stbi__rewind(s); + return 0; // only RGB or grey allowed, +/- RLE + } + stbi__skip(s,9); // skip colormap specification and image x/y origin + tga_colormap_bpp = 0; + } + tga_w = stbi__get16le(s); + if( tga_w < 1 ) { + stbi__rewind(s); + return 0; // test width + } + tga_h = stbi__get16le(s); + if( tga_h < 1 ) { + stbi__rewind(s); + return 0; // test height + } + tga_bits_per_pixel = stbi__get8(s); // bits per pixel + stbi__get8(s); // ignore alpha bits + if (tga_colormap_bpp != 0) { + if((tga_bits_per_pixel != 8) && (tga_bits_per_pixel != 16)) { + // when using a colormap, tga_bits_per_pixel is the size of the indexes + // I don't think anything but 8 or 16bit indexes makes sense + stbi__rewind(s); + return 0; + } + tga_comp = stbi__tga_get_comp(tga_colormap_bpp, 0, NULL); + } else { + tga_comp = stbi__tga_get_comp(tga_bits_per_pixel, (tga_image_type == 3) || (tga_image_type == 11), NULL); + } + if(!tga_comp) { + stbi__rewind(s); + return 0; + } + if (x) *x = tga_w; + if (y) *y = tga_h; + if (comp) *comp = tga_comp; + return 1; // seems to have passed everything +} + +static int stbi__tga_test(stbi__context *s) +{ + int res = 0; + int sz, tga_color_type; + stbi__get8(s); // discard Offset + tga_color_type = stbi__get8(s); // color type + if ( tga_color_type > 1 ) goto errorEnd; // only RGB or indexed allowed + sz = stbi__get8(s); // image type + if ( tga_color_type == 1 ) { // colormapped (paletted) image + if (sz != 1 && sz != 9) goto errorEnd; // colortype 1 demands image type 1 or 9 + stbi__skip(s,4); // skip index of first colormap entry and number of entries + sz = stbi__get8(s); // check bits per palette color entry + if ( (sz != 8) && (sz != 15) && (sz != 16) && (sz != 24) && (sz != 32) ) goto errorEnd; + stbi__skip(s,4); // skip image x and y origin + } else { // "normal" image w/o colormap + if ( (sz != 2) && (sz != 3) && (sz != 10) && (sz != 11) ) goto errorEnd; // only RGB or grey allowed, +/- RLE + stbi__skip(s,9); // skip colormap specification and image x/y origin + } + if ( stbi__get16le(s) < 1 ) goto errorEnd; // test width + if ( stbi__get16le(s) < 1 ) goto errorEnd; // test height + sz = stbi__get8(s); // bits per pixel + if ( (tga_color_type == 1) && (sz != 8) && (sz != 16) ) goto errorEnd; // for colormapped images, bpp is size of an index + if ( (sz != 8) && (sz != 15) && (sz != 16) && (sz != 24) && (sz != 32) ) goto errorEnd; + + res = 1; // if we got this far, everything's good and we can return 1 instead of 0 + +errorEnd: + stbi__rewind(s); + return res; +} + +// read 16bit value and convert to 24bit RGB +void stbi__tga_read_rgb16(stbi__context *s, stbi_uc* out) +{ + stbi__uint16 px = stbi__get16le(s); + stbi__uint16 fiveBitMask = 31; + // we have 3 channels with 5bits each + int r = (px >> 10) & fiveBitMask; + int g = (px >> 5) & fiveBitMask; + int b = px & fiveBitMask; + // Note that this saves the data in RGB(A) order, so it doesn't need to be swapped later + out[0] = (r * 255)/31; + out[1] = (g * 255)/31; + out[2] = (b * 255)/31; + + // some people claim that the most significant bit might be used for alpha + // (possibly if an alpha-bit is set in the "image descriptor byte") + // but that only made 16bit test images completely translucent.. + // so let's treat all 15 and 16bit TGAs as RGB with no alpha. +} + +static stbi_uc *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int req_comp) +{ + // read in the TGA header stuff + int tga_offset = stbi__get8(s); + int tga_indexed = stbi__get8(s); + int tga_image_type = stbi__get8(s); + int tga_is_RLE = 0; + int tga_palette_start = stbi__get16le(s); + int tga_palette_len = stbi__get16le(s); + int tga_palette_bits = stbi__get8(s); + int tga_x_origin = stbi__get16le(s); + int tga_y_origin = stbi__get16le(s); + int tga_width = stbi__get16le(s); + int tga_height = stbi__get16le(s); + int tga_bits_per_pixel = stbi__get8(s); + int tga_comp, tga_rgb16=0; + int tga_inverted = stbi__get8(s); + // int tga_alpha_bits = tga_inverted & 15; // the 4 lowest bits - unused (useless?) + // image data + unsigned char *tga_data; + unsigned char *tga_palette = NULL; + int i, j; + unsigned char raw_data[4]; + int RLE_count = 0; + int RLE_repeating = 0; + int read_next_pixel = 1; + + // do a tiny bit of precessing + if ( tga_image_type >= 8 ) + { + tga_image_type -= 8; + tga_is_RLE = 1; + } + tga_inverted = 1 - ((tga_inverted >> 5) & 1); + + // If I'm paletted, then I'll use the number of bits from the palette + if ( tga_indexed ) tga_comp = stbi__tga_get_comp(tga_palette_bits, 0, &tga_rgb16); + else tga_comp = stbi__tga_get_comp(tga_bits_per_pixel, (tga_image_type == 3), &tga_rgb16); + + if(!tga_comp) // shouldn't really happen, stbi__tga_test() should have ensured basic consistency + return stbi__errpuc("bad format", "Can't find out TGA pixelformat"); + + // tga info + *x = tga_width; + *y = tga_height; + if (comp) *comp = tga_comp; + + tga_data = (unsigned char*)stbi__malloc( (size_t)tga_width * tga_height * tga_comp ); + if (!tga_data) return stbi__errpuc("outofmem", "Out of memory"); + + // skip to the data's starting position (offset usually = 0) + stbi__skip(s, tga_offset ); + + if ( !tga_indexed && !tga_is_RLE && !tga_rgb16 ) { + for (i=0; i < tga_height; ++i) { + int row = tga_inverted ? tga_height -i - 1 : i; + stbi_uc *tga_row = tga_data + row*tga_width*tga_comp; + stbi__getn(s, tga_row, tga_width * tga_comp); + } + } else { + // do I need to load a palette? + if ( tga_indexed) + { + // any data to skip? (offset usually = 0) + stbi__skip(s, tga_palette_start ); + // load the palette + tga_palette = (unsigned char*)stbi__malloc( tga_palette_len * tga_comp ); + if (!tga_palette) { + STBI_FREE(tga_data); + return stbi__errpuc("outofmem", "Out of memory"); + } + if (tga_rgb16) { + stbi_uc *pal_entry = tga_palette; + STBI_ASSERT(tga_comp == STBI_rgb); + for (i=0; i < tga_palette_len; ++i) { + stbi__tga_read_rgb16(s, pal_entry); + pal_entry += tga_comp; + } + } else if (!stbi__getn(s, tga_palette, tga_palette_len * tga_comp)) { + STBI_FREE(tga_data); + STBI_FREE(tga_palette); + return stbi__errpuc("bad palette", "Corrupt TGA"); + } + } + // load the data + for (i=0; i < tga_width * tga_height; ++i) + { + // if I'm in RLE mode, do I need to get a RLE stbi__pngchunk? + if ( tga_is_RLE ) + { + if ( RLE_count == 0 ) + { + // yep, get the next byte as a RLE command + int RLE_cmd = stbi__get8(s); + RLE_count = 1 + (RLE_cmd & 127); + RLE_repeating = RLE_cmd >> 7; + read_next_pixel = 1; + } else if ( !RLE_repeating ) + { + read_next_pixel = 1; + } + } else + { + read_next_pixel = 1; + } + // OK, if I need to read a pixel, do it now + if ( read_next_pixel ) + { + // load however much data we did have + if ( tga_indexed ) + { + // read in index, then perform the lookup + int pal_idx = (tga_bits_per_pixel == 8) ? stbi__get8(s) : stbi__get16le(s); + if ( pal_idx >= tga_palette_len ) { + // invalid index + pal_idx = 0; + } + pal_idx *= tga_comp; + for (j = 0; j < tga_comp; ++j) { + raw_data[j] = tga_palette[pal_idx+j]; + } + } else if(tga_rgb16) { + STBI_ASSERT(tga_comp == STBI_rgb); + stbi__tga_read_rgb16(s, raw_data); + } else { + // read in the data raw + for (j = 0; j < tga_comp; ++j) { + raw_data[j] = stbi__get8(s); + } + } + // clear the reading flag for the next pixel + read_next_pixel = 0; + } // end of reading a pixel + + // copy data + for (j = 0; j < tga_comp; ++j) + tga_data[i*tga_comp+j] = raw_data[j]; + + // in case we're in RLE mode, keep counting down + --RLE_count; + } + // do I need to invert the image? + if ( tga_inverted ) + { + for (j = 0; j*2 < tga_height; ++j) + { + int index1 = j * tga_width * tga_comp; + int index2 = (tga_height - 1 - j) * tga_width * tga_comp; + for (i = tga_width * tga_comp; i > 0; --i) + { + unsigned char temp = tga_data[index1]; + tga_data[index1] = tga_data[index2]; + tga_data[index2] = temp; + ++index1; + ++index2; + } + } + } + // clear my palette, if I had one + if ( tga_palette != NULL ) + { + STBI_FREE( tga_palette ); + } + } + + // swap RGB - if the source data was RGB16, it already is in the right order + if (tga_comp >= 3 && !tga_rgb16) + { + unsigned char* tga_pixel = tga_data; + for (i=0; i < tga_width * tga_height; ++i) + { + unsigned char temp = tga_pixel[0]; + tga_pixel[0] = tga_pixel[2]; + tga_pixel[2] = temp; + tga_pixel += tga_comp; + } + } + + // convert to target component count + if (req_comp && req_comp != tga_comp) + tga_data = stbi__convert_format(tga_data, tga_comp, req_comp, tga_width, tga_height); + + // the things I do to get rid of an error message, and yet keep + // Microsoft's C compilers happy... [8^( + tga_palette_start = tga_palette_len = tga_palette_bits = + tga_x_origin = tga_y_origin = 0; + // OK, done + return tga_data; +} +#endif + +// ************************************************************************************************* +// Photoshop PSD loader -- PD by Thatcher Ulrich, integration by Nicolas Schulz, tweaked by STB + +#ifndef STBI_NO_PSD +static int stbi__psd_test(stbi__context *s) +{ + int r = (stbi__get32be(s) == 0x38425053); + stbi__rewind(s); + return r; +} + +static stbi_uc *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int req_comp) +{ + int pixelCount; + int channelCount, compression; + int channel, i, count, len; + int bitdepth; + int w,h; + stbi_uc *out; + + // Check identifier + if (stbi__get32be(s) != 0x38425053) // "8BPS" + return stbi__errpuc("not PSD", "Corrupt PSD image"); + + // Check file type version. + if (stbi__get16be(s) != 1) + return stbi__errpuc("wrong version", "Unsupported version of PSD image"); + + // Skip 6 reserved bytes. + stbi__skip(s, 6 ); + + // Read the number of channels (R, G, B, A, etc). + channelCount = stbi__get16be(s); + if (channelCount < 0 || channelCount > 16) + return stbi__errpuc("wrong channel count", "Unsupported number of channels in PSD image"); + + // Read the rows and columns of the image. + h = stbi__get32be(s); + w = stbi__get32be(s); + + // Make sure the depth is 8 bits. + bitdepth = stbi__get16be(s); + if (bitdepth != 8 && bitdepth != 16) + return stbi__errpuc("unsupported bit depth", "PSD bit depth is not 8 or 16 bit"); + + // Make sure the color mode is RGB. + // Valid options are: + // 0: Bitmap + // 1: Grayscale + // 2: Indexed color + // 3: RGB color + // 4: CMYK color + // 7: Multichannel + // 8: Duotone + // 9: Lab color + if (stbi__get16be(s) != 3) + return stbi__errpuc("wrong color format", "PSD is not in RGB color format"); + + // Skip the Mode Data. (It's the palette for indexed color; other info for other modes.) + stbi__skip(s,stbi__get32be(s) ); + + // Skip the image resources. (resolution, pen tool paths, etc) + stbi__skip(s, stbi__get32be(s) ); + + // Skip the reserved data. + stbi__skip(s, stbi__get32be(s) ); + + // Find out if the data is compressed. + // Known values: + // 0: no compression + // 1: RLE compressed + compression = stbi__get16be(s); + if (compression > 1) + return stbi__errpuc("bad compression", "PSD has an unknown compression format"); + + // Create the destination image. + out = (stbi_uc *) stbi__malloc(4 * w*h); + if (!out) return stbi__errpuc("outofmem", "Out of memory"); + pixelCount = w*h; + + // Initialize the data to zero. + //memset( out, 0, pixelCount * 4 ); + + // Finally, the image data. + if (compression) { + // RLE as used by .PSD and .TIFF + // Loop until you get the number of unpacked bytes you are expecting: + // Read the next source byte into n. + // If n is between 0 and 127 inclusive, copy the next n+1 bytes literally. + // Else if n is between -127 and -1 inclusive, copy the next byte -n+1 times. + // Else if n is 128, noop. + // Endloop + + // The RLE-compressed data is preceeded by a 2-byte data count for each row in the data, + // which we're going to just skip. + stbi__skip(s, h * channelCount * 2 ); + + // Read the RLE data by channel. + for (channel = 0; channel < 4; channel++) { + stbi_uc *p; + + p = out+channel; + if (channel >= channelCount) { + // Fill this channel with default data. + for (i = 0; i < pixelCount; i++, p += 4) + *p = (channel == 3 ? 255 : 0); + } else { + // Read the RLE data. + count = 0; + while (count < pixelCount) { + len = stbi__get8(s); + if (len == 128) { + // No-op. + } else if (len < 128) { + // Copy next len+1 bytes literally. + len++; + count += len; + while (len) { + *p = stbi__get8(s); + p += 4; + len--; + } + } else if (len > 128) { + stbi_uc val; + // Next -len+1 bytes in the dest are replicated from next source byte. + // (Interpret len as a negative 8-bit int.) + len ^= 0x0FF; + len += 2; + val = stbi__get8(s); + count += len; + while (len) { + *p = val; + p += 4; + len--; + } + } + } + } + } + + } else { + // We're at the raw image data. It's each channel in order (Red, Green, Blue, Alpha, ...) + // where each channel consists of an 8-bit value for each pixel in the image. + + // Read the data by channel. + for (channel = 0; channel < 4; channel++) { + stbi_uc *p; + + p = out + channel; + if (channel >= channelCount) { + // Fill this channel with default data. + stbi_uc val = channel == 3 ? 255 : 0; + for (i = 0; i < pixelCount; i++, p += 4) + *p = val; + } else { + // Read the data. + if (bitdepth == 16) { + for (i = 0; i < pixelCount; i++, p += 4) + *p = (stbi_uc) (stbi__get16be(s) >> 8); + } else { + for (i = 0; i < pixelCount; i++, p += 4) + *p = stbi__get8(s); + } + } + } + } + + if (req_comp && req_comp != 4) { + out = stbi__convert_format(out, 4, req_comp, w, h); + if (out == NULL) return out; // stbi__convert_format frees input on failure + } + + if (comp) *comp = 4; + *y = h; + *x = w; + + return out; +} +#endif + +// ************************************************************************************************* +// Softimage PIC loader +// by Tom Seddon +// +// See http://softimage.wiki.softimage.com/index.php/INFO:_PIC_file_format +// See http://ozviz.wasp.uwa.edu.au/~pbourke/dataformats/softimagepic/ + +#ifndef STBI_NO_PIC +static int stbi__pic_is4(stbi__context *s,const char *str) +{ + int i; + for (i=0; i<4; ++i) + if (stbi__get8(s) != (stbi_uc)str[i]) + return 0; + + return 1; +} + +static int stbi__pic_test_core(stbi__context *s) +{ + int i; + + if (!stbi__pic_is4(s,"\x53\x80\xF6\x34")) + return 0; + + for(i=0;i<84;++i) + stbi__get8(s); + + if (!stbi__pic_is4(s,"PICT")) + return 0; + + return 1; +} + +typedef struct +{ + stbi_uc size,type,channel; +} stbi__pic_packet; + +static stbi_uc *stbi__readval(stbi__context *s, int channel, stbi_uc *dest) +{ + int mask=0x80, i; + + for (i=0; i<4; ++i, mask>>=1) { + if (channel & mask) { + if (stbi__at_eof(s)) return stbi__errpuc("bad file","PIC file too short"); + dest[i]=stbi__get8(s); + } + } + + return dest; +} + +static void stbi__copyval(int channel,stbi_uc *dest,const stbi_uc *src) +{ + int mask=0x80,i; + + for (i=0;i<4; ++i, mask>>=1) + if (channel&mask) + dest[i]=src[i]; +} + +static stbi_uc *stbi__pic_load_core(stbi__context *s,int width,int height,int *comp, stbi_uc *result) +{ + int act_comp=0,num_packets=0,y,chained; + stbi__pic_packet packets[10]; + + // this will (should...) cater for even some bizarre stuff like having data + // for the same channel in multiple packets. + do { + stbi__pic_packet *packet; + + if (num_packets==sizeof(packets)/sizeof(packets[0])) + return stbi__errpuc("bad format","too many packets"); + + packet = &packets[num_packets++]; + + chained = stbi__get8(s); + packet->size = stbi__get8(s); + packet->type = stbi__get8(s); + packet->channel = stbi__get8(s); + + act_comp |= packet->channel; + + if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (reading packets)"); + if (packet->size != 8) return stbi__errpuc("bad format","packet isn't 8bpp"); + } while (chained); + + *comp = (act_comp & 0x10 ? 4 : 3); // has alpha channel? + + for(y=0; ytype) { + default: + return stbi__errpuc("bad format","packet has bad compression type"); + + case 0: {//uncompressed + int x; + + for(x=0;xchannel,dest)) + return 0; + break; + } + + case 1://Pure RLE + { + int left=width, i; + + while (left>0) { + stbi_uc count,value[4]; + + count=stbi__get8(s); + if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (pure read count)"); + + if (count > left) + count = (stbi_uc) left; + + if (!stbi__readval(s,packet->channel,value)) return 0; + + for(i=0; ichannel,dest,value); + left -= count; + } + } + break; + + case 2: {//Mixed RLE + int left=width; + while (left>0) { + int count = stbi__get8(s), i; + if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (mixed read count)"); + + if (count >= 128) { // Repeated + stbi_uc value[4]; + + if (count==128) + count = stbi__get16be(s); + else + count -= 127; + if (count > left) + return stbi__errpuc("bad file","scanline overrun"); + + if (!stbi__readval(s,packet->channel,value)) + return 0; + + for(i=0;ichannel,dest,value); + } else { // Raw + ++count; + if (count>left) return stbi__errpuc("bad file","scanline overrun"); + + for(i=0;ichannel,dest)) + return 0; + } + left-=count; + } + break; + } + } + } + } + + return result; +} + +static stbi_uc *stbi__pic_load(stbi__context *s,int *px,int *py,int *comp,int req_comp) +{ + stbi_uc *result; + int i, x,y; + + for (i=0; i<92; ++i) + stbi__get8(s); + + x = stbi__get16be(s); + y = stbi__get16be(s); + if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (pic header)"); + if ((1 << 28) / x < y) return stbi__errpuc("too large", "Image too large to decode"); + + stbi__get32be(s); //skip `ratio' + stbi__get16be(s); //skip `fields' + stbi__get16be(s); //skip `pad' + + // intermediate buffer is RGBA + result = (stbi_uc *) stbi__malloc(x*y*4); + memset(result, 0xff, x*y*4); + + if (!stbi__pic_load_core(s,x,y,comp, result)) { + STBI_FREE(result); + result=0; + } + *px = x; + *py = y; + if (req_comp == 0) req_comp = *comp; + result=stbi__convert_format(result,4,req_comp,x,y); + + return result; +} + +static int stbi__pic_test(stbi__context *s) +{ + int r = stbi__pic_test_core(s); + stbi__rewind(s); + return r; +} +#endif + +// ************************************************************************************************* +// GIF loader -- public domain by Jean-Marc Lienher -- simplified/shrunk by stb + +#ifndef STBI_NO_GIF +typedef struct +{ + stbi__int16 prefix; + stbi_uc first; + stbi_uc suffix; +} stbi__gif_lzw; + +typedef struct +{ + int w,h; + stbi_uc *out, *old_out; // output buffer (always 4 components) + int flags, bgindex, ratio, transparent, eflags, delay; + stbi_uc pal[256][4]; + stbi_uc lpal[256][4]; + stbi__gif_lzw codes[4096]; + stbi_uc *color_table; + int parse, step; + int lflags; + int start_x, start_y; + int max_x, max_y; + int cur_x, cur_y; + int line_size; +} stbi__gif; + +static int stbi__gif_test_raw(stbi__context *s) +{ + int sz; + if (stbi__get8(s) != 'G' || stbi__get8(s) != 'I' || stbi__get8(s) != 'F' || stbi__get8(s) != '8') return 0; + sz = stbi__get8(s); + if (sz != '9' && sz != '7') return 0; + if (stbi__get8(s) != 'a') return 0; + return 1; +} + +static int stbi__gif_test(stbi__context *s) +{ + int r = stbi__gif_test_raw(s); + stbi__rewind(s); + return r; +} + +static void stbi__gif_parse_colortable(stbi__context *s, stbi_uc pal[256][4], int num_entries, int transp) +{ + int i; + for (i=0; i < num_entries; ++i) { + pal[i][2] = stbi__get8(s); + pal[i][1] = stbi__get8(s); + pal[i][0] = stbi__get8(s); + pal[i][3] = transp == i ? 0 : 255; + } +} + +static int stbi__gif_header(stbi__context *s, stbi__gif *g, int *comp, int is_info) +{ + stbi_uc version; + if (stbi__get8(s) != 'G' || stbi__get8(s) != 'I' || stbi__get8(s) != 'F' || stbi__get8(s) != '8') + return stbi__err("not GIF", "Corrupt GIF"); + + version = stbi__get8(s); + if (version != '7' && version != '9') return stbi__err("not GIF", "Corrupt GIF"); + if (stbi__get8(s) != 'a') return stbi__err("not GIF", "Corrupt GIF"); + + stbi__g_failure_reason = ""; + g->w = stbi__get16le(s); + g->h = stbi__get16le(s); + g->flags = stbi__get8(s); + g->bgindex = stbi__get8(s); + g->ratio = stbi__get8(s); + g->transparent = -1; + + if (comp != 0) *comp = 4; // can't actually tell whether it's 3 or 4 until we parse the comments + + if (is_info) return 1; + + if (g->flags & 0x80) + stbi__gif_parse_colortable(s,g->pal, 2 << (g->flags & 7), -1); + + return 1; +} + +static int stbi__gif_info_raw(stbi__context *s, int *x, int *y, int *comp) +{ + stbi__gif g; + if (!stbi__gif_header(s, &g, comp, 1)) { + stbi__rewind( s ); + return 0; + } + if (x) *x = g.w; + if (y) *y = g.h; + return 1; +} + +static void stbi__out_gif_code(stbi__gif *g, stbi__uint16 code) +{ + stbi_uc *p, *c; + + // recurse to decode the prefixes, since the linked-list is backwards, + // and working backwards through an interleaved image would be nasty + if (g->codes[code].prefix >= 0) + stbi__out_gif_code(g, g->codes[code].prefix); + + if (g->cur_y >= g->max_y) return; + + p = &g->out[g->cur_x + g->cur_y]; + c = &g->color_table[g->codes[code].suffix * 4]; + + if (c[3] >= 128) { + p[0] = c[2]; + p[1] = c[1]; + p[2] = c[0]; + p[3] = c[3]; + } + g->cur_x += 4; + + if (g->cur_x >= g->max_x) { + g->cur_x = g->start_x; + g->cur_y += g->step; + + while (g->cur_y >= g->max_y && g->parse > 0) { + g->step = (1 << g->parse) * g->line_size; + g->cur_y = g->start_y + (g->step >> 1); + --g->parse; + } + } +} + +static stbi_uc *stbi__process_gif_raster(stbi__context *s, stbi__gif *g) +{ + stbi_uc lzw_cs; + stbi__int32 len, init_code; + stbi__uint32 first; + stbi__int32 codesize, codemask, avail, oldcode, bits, valid_bits, clear; + stbi__gif_lzw *p; + + lzw_cs = stbi__get8(s); + if (lzw_cs > 12) return NULL; + clear = 1 << lzw_cs; + first = 1; + codesize = lzw_cs + 1; + codemask = (1 << codesize) - 1; + bits = 0; + valid_bits = 0; + for (init_code = 0; init_code < clear; init_code++) { + g->codes[init_code].prefix = -1; + g->codes[init_code].first = (stbi_uc) init_code; + g->codes[init_code].suffix = (stbi_uc) init_code; + } + + // support no starting clear code + avail = clear+2; + oldcode = -1; + + len = 0; + for(;;) { + if (valid_bits < codesize) { + if (len == 0) { + len = stbi__get8(s); // start new block + if (len == 0) + return g->out; + } + --len; + bits |= (stbi__int32) stbi__get8(s) << valid_bits; + valid_bits += 8; + } else { + stbi__int32 code = bits & codemask; + bits >>= codesize; + valid_bits -= codesize; + // @OPTIMIZE: is there some way we can accelerate the non-clear path? + if (code == clear) { // clear code + codesize = lzw_cs + 1; + codemask = (1 << codesize) - 1; + avail = clear + 2; + oldcode = -1; + first = 0; + } else if (code == clear + 1) { // end of stream code + stbi__skip(s, len); + while ((len = stbi__get8(s)) > 0) + stbi__skip(s,len); + return g->out; + } else if (code <= avail) { + if (first) return stbi__errpuc("no clear code", "Corrupt GIF"); + + if (oldcode >= 0) { + p = &g->codes[avail++]; + if (avail > 4096) return stbi__errpuc("too many codes", "Corrupt GIF"); + p->prefix = (stbi__int16) oldcode; + p->first = g->codes[oldcode].first; + p->suffix = (code == avail) ? p->first : g->codes[code].first; + } else if (code == avail) + return stbi__errpuc("illegal code in raster", "Corrupt GIF"); + + stbi__out_gif_code(g, (stbi__uint16) code); + + if ((avail & codemask) == 0 && avail <= 0x0FFF) { + codesize++; + codemask = (1 << codesize) - 1; + } + + oldcode = code; + } else { + return stbi__errpuc("illegal code in raster", "Corrupt GIF"); + } + } + } +} + +static void stbi__fill_gif_background(stbi__gif *g, int x0, int y0, int x1, int y1) +{ + int x, y; + stbi_uc *c = g->pal[g->bgindex]; + for (y = y0; y < y1; y += 4 * g->w) { + for (x = x0; x < x1; x += 4) { + stbi_uc *p = &g->out[y + x]; + p[0] = c[2]; + p[1] = c[1]; + p[2] = c[0]; + p[3] = 0; + } + } +} + +// this function is designed to support animated gifs, although stb_image doesn't support it +static stbi_uc *stbi__gif_load_next(stbi__context *s, stbi__gif *g, int *comp, int req_comp) +{ + int i; + stbi_uc *prev_out = 0; + + if (g->out == 0 && !stbi__gif_header(s, g, comp,0)) + return 0; // stbi__g_failure_reason set by stbi__gif_header + + prev_out = g->out; + g->out = (stbi_uc *) stbi__malloc(4 * g->w * g->h); + if (g->out == 0) return stbi__errpuc("outofmem", "Out of memory"); + + switch ((g->eflags & 0x1C) >> 2) { + case 0: // unspecified (also always used on 1st frame) + stbi__fill_gif_background(g, 0, 0, 4 * g->w, 4 * g->w * g->h); + break; + case 1: // do not dispose + if (prev_out) memcpy(g->out, prev_out, 4 * g->w * g->h); + g->old_out = prev_out; + break; + case 2: // dispose to background + if (prev_out) memcpy(g->out, prev_out, 4 * g->w * g->h); + stbi__fill_gif_background(g, g->start_x, g->start_y, g->max_x, g->max_y); + break; + case 3: // dispose to previous + if (g->old_out) { + for (i = g->start_y; i < g->max_y; i += 4 * g->w) + memcpy(&g->out[i + g->start_x], &g->old_out[i + g->start_x], g->max_x - g->start_x); + } + break; + } + + for (;;) { + switch (stbi__get8(s)) { + case 0x2C: /* Image Descriptor */ + { + int prev_trans = -1; + stbi__int32 x, y, w, h; + stbi_uc *o; + + x = stbi__get16le(s); + y = stbi__get16le(s); + w = stbi__get16le(s); + h = stbi__get16le(s); + if (((x + w) > (g->w)) || ((y + h) > (g->h))) + return stbi__errpuc("bad Image Descriptor", "Corrupt GIF"); + + g->line_size = g->w * 4; + g->start_x = x * 4; + g->start_y = y * g->line_size; + g->max_x = g->start_x + w * 4; + g->max_y = g->start_y + h * g->line_size; + g->cur_x = g->start_x; + g->cur_y = g->start_y; + + g->lflags = stbi__get8(s); + + if (g->lflags & 0x40) { + g->step = 8 * g->line_size; // first interlaced spacing + g->parse = 3; + } else { + g->step = g->line_size; + g->parse = 0; + } + + if (g->lflags & 0x80) { + stbi__gif_parse_colortable(s,g->lpal, 2 << (g->lflags & 7), g->eflags & 0x01 ? g->transparent : -1); + g->color_table = (stbi_uc *) g->lpal; + } else if (g->flags & 0x80) { + if (g->transparent >= 0 && (g->eflags & 0x01)) { + prev_trans = g->pal[g->transparent][3]; + g->pal[g->transparent][3] = 0; + } + g->color_table = (stbi_uc *) g->pal; + } else + return stbi__errpuc("missing color table", "Corrupt GIF"); + + o = stbi__process_gif_raster(s, g); + if (o == NULL) return NULL; + + if (prev_trans != -1) + g->pal[g->transparent][3] = (stbi_uc) prev_trans; + + return o; + } + + case 0x21: // Comment Extension. + { + int len; + if (stbi__get8(s) == 0xF9) { // Graphic Control Extension. + len = stbi__get8(s); + if (len == 4) { + g->eflags = stbi__get8(s); + g->delay = stbi__get16le(s); + g->transparent = stbi__get8(s); + } else { + stbi__skip(s, len); + break; + } + } + while ((len = stbi__get8(s)) != 0) + stbi__skip(s, len); + break; + } + + case 0x3B: // gif stream termination code + return (stbi_uc *) s; // using '1' causes warning on some compilers + + default: + return stbi__errpuc("unknown code", "Corrupt GIF"); + } + } + + STBI_NOTUSED(req_comp); +} + +static stbi_uc *stbi__gif_load(stbi__context *s, int *x, int *y, int *comp, int req_comp) +{ + stbi_uc *u = 0; + stbi__gif g; + memset(&g, 0, sizeof(g)); + + u = stbi__gif_load_next(s, &g, comp, req_comp); + if (u == (stbi_uc *) s) u = 0; // end of animated gif marker + if (u) { + *x = g.w; + *y = g.h; + if (req_comp && req_comp != 4) + u = stbi__convert_format(u, 4, req_comp, g.w, g.h); + } + else if (g.out) + STBI_FREE(g.out); + + return u; +} + +static int stbi__gif_info(stbi__context *s, int *x, int *y, int *comp) +{ + return stbi__gif_info_raw(s,x,y,comp); +} +#endif + +// ************************************************************************************************* +// Radiance RGBE HDR loader +// originally by Nicolas Schulz +#ifndef STBI_NO_HDR +static int stbi__hdr_test_core(stbi__context *s) +{ + const char *signature = "#?RADIANCE\n"; + int i; + for (i=0; signature[i]; ++i) + if (stbi__get8(s) != signature[i]) + return 0; + return 1; +} + +static int stbi__hdr_test(stbi__context* s) +{ + int r = stbi__hdr_test_core(s); + stbi__rewind(s); + return r; +} + +#define STBI__HDR_BUFLEN 1024 +static char *stbi__hdr_gettoken(stbi__context *z, char *buffer) +{ + int len=0; + char c = '\0'; + + c = (char) stbi__get8(z); + + while (!stbi__at_eof(z) && c != '\n') { + buffer[len++] = c; + if (len == STBI__HDR_BUFLEN-1) { + // flush to end of line + while (!stbi__at_eof(z) && stbi__get8(z) != '\n') + ; + break; + } + c = (char) stbi__get8(z); + } + + buffer[len] = 0; + return buffer; +} + +static void stbi__hdr_convert(float *output, stbi_uc *input, int req_comp) +{ + if ( input[3] != 0 ) { + float f1; + // Exponent + f1 = (float) ldexp(1.0f, input[3] - (int)(128 + 8)); + if (req_comp <= 2) + output[0] = (input[0] + input[1] + input[2]) * f1 / 3; + else { + output[0] = input[0] * f1; + output[1] = input[1] * f1; + output[2] = input[2] * f1; + } + if (req_comp == 2) output[1] = 1; + if (req_comp == 4) output[3] = 1; + } else { + switch (req_comp) { + case 4: output[3] = 1; /* fallthrough */ + case 3: output[0] = output[1] = output[2] = 0; + break; + case 2: output[1] = 1; /* fallthrough */ + case 1: output[0] = 0; + break; + } + } +} + +static float *stbi__hdr_load(stbi__context *s, int *x, int *y, int *comp, int req_comp) +{ + char buffer[STBI__HDR_BUFLEN]; + char *token; + int valid = 0; + int width, height; + stbi_uc *scanline; + float *hdr_data; + int len; + unsigned char count, value; + int i, j, k, c1,c2, z; + + + // Check identifier + if (strcmp(stbi__hdr_gettoken(s,buffer), "#?RADIANCE") != 0) + return stbi__errpf("not HDR", "Corrupt HDR image"); + + // Parse header + for(;;) { + token = stbi__hdr_gettoken(s,buffer); + if (token[0] == 0) break; + if (strcmp(token, "FORMAT=32-bit_rle_rgbe") == 0) valid = 1; + } + + if (!valid) return stbi__errpf("unsupported format", "Unsupported HDR format"); + + // Parse width and height + // can't use sscanf() if we're not using stdio! + token = stbi__hdr_gettoken(s,buffer); + if (strncmp(token, "-Y ", 3)) return stbi__errpf("unsupported data layout", "Unsupported HDR format"); + token += 3; + height = (int) strtol(token, &token, 10); + while (*token == ' ') ++token; + if (strncmp(token, "+X ", 3)) return stbi__errpf("unsupported data layout", "Unsupported HDR format"); + token += 3; + width = (int) strtol(token, NULL, 10); + + *x = width; + *y = height; + + if (comp) *comp = 3; + if (req_comp == 0) req_comp = 3; + + // Read data + hdr_data = (float *) stbi__malloc(height * width * req_comp * sizeof(float)); + + // Load image data + // image data is stored as some number of sca + if ( width < 8 || width >= 32768) { + // Read flat data + for (j=0; j < height; ++j) { + for (i=0; i < width; ++i) { + stbi_uc rgbe[4]; + main_decode_loop: + stbi__getn(s, rgbe, 4); + stbi__hdr_convert(hdr_data + j * width * req_comp + i * req_comp, rgbe, req_comp); + } + } + } else { + // Read RLE-encoded data + scanline = NULL; + + for (j = 0; j < height; ++j) { + c1 = stbi__get8(s); + c2 = stbi__get8(s); + len = stbi__get8(s); + if (c1 != 2 || c2 != 2 || (len & 0x80)) { + // not run-length encoded, so we have to actually use THIS data as a decoded + // pixel (note this can't be a valid pixel--one of RGB must be >= 128) + stbi_uc rgbe[4]; + rgbe[0] = (stbi_uc) c1; + rgbe[1] = (stbi_uc) c2; + rgbe[2] = (stbi_uc) len; + rgbe[3] = (stbi_uc) stbi__get8(s); + stbi__hdr_convert(hdr_data, rgbe, req_comp); + i = 1; + j = 0; + STBI_FREE(scanline); + goto main_decode_loop; // yes, this makes no sense + } + len <<= 8; + len |= stbi__get8(s); + if (len != width) { STBI_FREE(hdr_data); STBI_FREE(scanline); return stbi__errpf("invalid decoded scanline length", "corrupt HDR"); } + if (scanline == NULL) scanline = (stbi_uc *) stbi__malloc(width * 4); + + for (k = 0; k < 4; ++k) { + i = 0; + while (i < width) { + count = stbi__get8(s); + if (count > 128) { + // Run + value = stbi__get8(s); + count -= 128; + for (z = 0; z < count; ++z) + scanline[i++ * 4 + k] = value; + } else { + // Dump + for (z = 0; z < count; ++z) + scanline[i++ * 4 + k] = stbi__get8(s); + } + } + } + for (i=0; i < width; ++i) + stbi__hdr_convert(hdr_data+(j*width + i)*req_comp, scanline + i*4, req_comp); + } + STBI_FREE(scanline); + } + + return hdr_data; +} + +static int stbi__hdr_info(stbi__context *s, int *x, int *y, int *comp) +{ + char buffer[STBI__HDR_BUFLEN]; + char *token; + int valid = 0; + + if (stbi__hdr_test(s) == 0) { + stbi__rewind( s ); + return 0; + } + + for(;;) { + token = stbi__hdr_gettoken(s,buffer); + if (token[0] == 0) break; + if (strcmp(token, "FORMAT=32-bit_rle_rgbe") == 0) valid = 1; + } + + if (!valid) { + stbi__rewind( s ); + return 0; + } + token = stbi__hdr_gettoken(s,buffer); + if (strncmp(token, "-Y ", 3)) { + stbi__rewind( s ); + return 0; + } + token += 3; + *y = (int) strtol(token, &token, 10); + while (*token == ' ') ++token; + if (strncmp(token, "+X ", 3)) { + stbi__rewind( s ); + return 0; + } + token += 3; + *x = (int) strtol(token, NULL, 10); + *comp = 3; + return 1; +} +#endif // STBI_NO_HDR + +#ifndef STBI_NO_BMP +static int stbi__bmp_info(stbi__context *s, int *x, int *y, int *comp) +{ + void *p; + stbi__bmp_data info; + + info.all_a = 255; + p = stbi__bmp_parse_header(s, &info); + stbi__rewind( s ); + if (p == NULL) + return 0; + *x = s->img_x; + *y = s->img_y; + *comp = info.ma ? 4 : 3; + return 1; +} +#endif + +#ifndef STBI_NO_PSD +static int stbi__psd_info(stbi__context *s, int *x, int *y, int *comp) +{ + int channelCount; + if (stbi__get32be(s) != 0x38425053) { + stbi__rewind( s ); + return 0; + } + if (stbi__get16be(s) != 1) { + stbi__rewind( s ); + return 0; + } + stbi__skip(s, 6); + channelCount = stbi__get16be(s); + if (channelCount < 0 || channelCount > 16) { + stbi__rewind( s ); + return 0; + } + *y = stbi__get32be(s); + *x = stbi__get32be(s); + if (stbi__get16be(s) != 8) { + stbi__rewind( s ); + return 0; + } + if (stbi__get16be(s) != 3) { + stbi__rewind( s ); + return 0; + } + *comp = 4; + return 1; +} +#endif + +#ifndef STBI_NO_PIC +static int stbi__pic_info(stbi__context *s, int *x, int *y, int *comp) +{ + int act_comp=0,num_packets=0,chained; + stbi__pic_packet packets[10]; + + if (!stbi__pic_is4(s,"\x53\x80\xF6\x34")) { + stbi__rewind(s); + return 0; + } + + stbi__skip(s, 88); + + *x = stbi__get16be(s); + *y = stbi__get16be(s); + if (stbi__at_eof(s)) { + stbi__rewind( s); + return 0; + } + if ( (*x) != 0 && (1 << 28) / (*x) < (*y)) { + stbi__rewind( s ); + return 0; + } + + stbi__skip(s, 8); + + do { + stbi__pic_packet *packet; + + if (num_packets==sizeof(packets)/sizeof(packets[0])) + return 0; + + packet = &packets[num_packets++]; + chained = stbi__get8(s); + packet->size = stbi__get8(s); + packet->type = stbi__get8(s); + packet->channel = stbi__get8(s); + act_comp |= packet->channel; + + if (stbi__at_eof(s)) { + stbi__rewind( s ); + return 0; + } + if (packet->size != 8) { + stbi__rewind( s ); + return 0; + } + } while (chained); + + *comp = (act_comp & 0x10 ? 4 : 3); + + return 1; +} +#endif + +// ************************************************************************************************* +// Portable Gray Map and Portable Pixel Map loader +// by Ken Miller +// +// PGM: http://netpbm.sourceforge.net/doc/pgm.html +// PPM: http://netpbm.sourceforge.net/doc/ppm.html +// +// Known limitations: +// Does not support comments in the header section +// Does not support ASCII image data (formats P2 and P3) +// Does not support 16-bit-per-channel + +#ifndef STBI_NO_PNM + +static int stbi__pnm_test(stbi__context *s) +{ + char p, t; + p = (char) stbi__get8(s); + t = (char) stbi__get8(s); + if (p != 'P' || (t != '5' && t != '6')) { + stbi__rewind( s ); + return 0; + } + return 1; +} + +static stbi_uc *stbi__pnm_load(stbi__context *s, int *x, int *y, int *comp, int req_comp) +{ + stbi_uc *out; + if (!stbi__pnm_info(s, (int *)&s->img_x, (int *)&s->img_y, (int *)&s->img_n)) + return 0; + *x = s->img_x; + *y = s->img_y; + *comp = s->img_n; + + out = (stbi_uc *) stbi__malloc(s->img_n * s->img_x * s->img_y); + if (!out) return stbi__errpuc("outofmem", "Out of memory"); + stbi__getn(s, out, s->img_n * s->img_x * s->img_y); + + if (req_comp && req_comp != s->img_n) { + out = stbi__convert_format(out, s->img_n, req_comp, s->img_x, s->img_y); + if (out == NULL) return out; // stbi__convert_format frees input on failure + } + return out; +} + +static int stbi__pnm_isspace(char c) +{ + return c == ' ' || c == '\t' || c == '\n' || c == '\v' || c == '\f' || c == '\r'; +} + +static void stbi__pnm_skip_whitespace(stbi__context *s, char *c) +{ + for (;;) { + while (!stbi__at_eof(s) && stbi__pnm_isspace(*c)) + *c = (char) stbi__get8(s); + + if (stbi__at_eof(s) || *c != '#') + break; + + while (!stbi__at_eof(s) && *c != '\n' && *c != '\r' ) + *c = (char) stbi__get8(s); + } +} + +static int stbi__pnm_isdigit(char c) +{ + return c >= '0' && c <= '9'; +} + +static int stbi__pnm_getinteger(stbi__context *s, char *c) +{ + int value = 0; + + while (!stbi__at_eof(s) && stbi__pnm_isdigit(*c)) { + value = value*10 + (*c - '0'); + *c = (char) stbi__get8(s); + } + + return value; +} + +static int stbi__pnm_info(stbi__context *s, int *x, int *y, int *comp) +{ + int maxv; + char c, p, t; + + stbi__rewind( s ); + + // Get identifier + p = (char) stbi__get8(s); + t = (char) stbi__get8(s); + if (p != 'P' || (t != '5' && t != '6')) { + stbi__rewind( s ); + return 0; + } + + *comp = (t == '6') ? 3 : 1; // '5' is 1-component .pgm; '6' is 3-component .ppm + + c = (char) stbi__get8(s); + stbi__pnm_skip_whitespace(s, &c); + + *x = stbi__pnm_getinteger(s, &c); // read width + stbi__pnm_skip_whitespace(s, &c); + + *y = stbi__pnm_getinteger(s, &c); // read height + stbi__pnm_skip_whitespace(s, &c); + + maxv = stbi__pnm_getinteger(s, &c); // read max value + + if (maxv > 255) + return stbi__err("max value > 255", "PPM image not 8-bit"); + else + return 1; +} +#endif + +static int stbi__info_main(stbi__context *s, int *x, int *y, int *comp) +{ + #ifndef STBI_NO_JPEG + if (stbi__jpeg_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_PNG + if (stbi__png_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_GIF + if (stbi__gif_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_BMP + if (stbi__bmp_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_PSD + if (stbi__psd_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_PIC + if (stbi__pic_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_PNM + if (stbi__pnm_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_HDR + if (stbi__hdr_info(s, x, y, comp)) return 1; + #endif + + // test tga last because it's a crappy test! + #ifndef STBI_NO_TGA + if (stbi__tga_info(s, x, y, comp)) + return 1; + #endif + return stbi__err("unknown image type", "Image not of any known type, or corrupt"); +} + +#ifndef STBI_NO_STDIO +STBIDEF int stbi_info(char const *filename, int *x, int *y, int *comp) +{ + FILE *f = stbi__fopen(filename, "rb"); + int result; + if (!f) return stbi__err("can't fopen", "Unable to open file"); + result = stbi_info_from_file(f, x, y, comp); + fclose(f); + return result; +} + +STBIDEF int stbi_info_from_file(FILE *f, int *x, int *y, int *comp) +{ + int r; + stbi__context s; + long pos = ftell(f); + stbi__start_file(&s, f); + r = stbi__info_main(&s,x,y,comp); + fseek(f,pos,SEEK_SET); + return r; +} +#endif // !STBI_NO_STDIO + +STBIDEF int stbi_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp) +{ + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__info_main(&s,x,y,comp); +} + +STBIDEF int stbi_info_from_callbacks(stbi_io_callbacks const *c, void *user, int *x, int *y, int *comp) +{ + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *) c, user); + return stbi__info_main(&s,x,y,comp); +} + +#endif // STB_IMAGE_IMPLEMENTATION + +/* + revision history: + 2.10 (2016-01-22) avoid warning introduced in 2.09 by STBI_REALLOC_SIZED + 2.09 (2016-01-16) allow comments in PNM files + 16-bit-per-pixel TGA (not bit-per-component) + info() for TGA could break due to .hdr handling + info() for BMP to shares code instead of sloppy parse + can use STBI_REALLOC_SIZED if allocator doesn't support realloc + code cleanup + 2.08 (2015-09-13) fix to 2.07 cleanup, reading RGB PSD as RGBA + 2.07 (2015-09-13) fix compiler warnings + partial animated GIF support + limited 16-bpc PSD support + #ifdef unused functions + bug with < 92 byte PIC,PNM,HDR,TGA + 2.06 (2015-04-19) fix bug where PSD returns wrong '*comp' value + 2.05 (2015-04-19) fix bug in progressive JPEG handling, fix warning + 2.04 (2015-04-15) try to re-enable SIMD on MinGW 64-bit + 2.03 (2015-04-12) extra corruption checking (mmozeiko) + stbi_set_flip_vertically_on_load (nguillemot) + fix NEON support; fix mingw support + 2.02 (2015-01-19) fix incorrect assert, fix warning + 2.01 (2015-01-17) fix various warnings; suppress SIMD on gcc 32-bit without -msse2 + 2.00b (2014-12-25) fix STBI_MALLOC in progressive JPEG + 2.00 (2014-12-25) optimize JPG, including x86 SSE2 & NEON SIMD (ryg) + progressive JPEG (stb) + PGM/PPM support (Ken Miller) + STBI_MALLOC,STBI_REALLOC,STBI_FREE + GIF bugfix -- seemingly never worked + STBI_NO_*, STBI_ONLY_* + 1.48 (2014-12-14) fix incorrectly-named assert() + 1.47 (2014-12-14) 1/2/4-bit PNG support, both direct and paletted (Omar Cornut & stb) + optimize PNG (ryg) + fix bug in interlaced PNG with user-specified channel count (stb) + 1.46 (2014-08-26) + fix broken tRNS chunk (colorkey-style transparency) in non-paletted PNG + 1.45 (2014-08-16) + fix MSVC-ARM internal compiler error by wrapping malloc + 1.44 (2014-08-07) + various warning fixes from Ronny Chevalier + 1.43 (2014-07-15) + fix MSVC-only compiler problem in code changed in 1.42 + 1.42 (2014-07-09) + don't define _CRT_SECURE_NO_WARNINGS (affects user code) + fixes to stbi__cleanup_jpeg path + added STBI_ASSERT to avoid requiring assert.h + 1.41 (2014-06-25) + fix search&replace from 1.36 that messed up comments/error messages + 1.40 (2014-06-22) + fix gcc struct-initialization warning + 1.39 (2014-06-15) + fix to TGA optimization when req_comp != number of components in TGA; + fix to GIF loading because BMP wasn't rewinding (whoops, no GIFs in my test suite) + add support for BMP version 5 (more ignored fields) + 1.38 (2014-06-06) + suppress MSVC warnings on integer casts truncating values + fix accidental rename of 'skip' field of I/O + 1.37 (2014-06-04) + remove duplicate typedef + 1.36 (2014-06-03) + convert to header file single-file library + if de-iphone isn't set, load iphone images color-swapped instead of returning NULL + 1.35 (2014-05-27) + various warnings + fix broken STBI_SIMD path + fix bug where stbi_load_from_file no longer left file pointer in correct place + fix broken non-easy path for 32-bit BMP (possibly never used) + TGA optimization by Arseny Kapoulkine + 1.34 (unknown) + use STBI_NOTUSED in stbi__resample_row_generic(), fix one more leak in tga failure case + 1.33 (2011-07-14) + make stbi_is_hdr work in STBI_NO_HDR (as specified), minor compiler-friendly improvements + 1.32 (2011-07-13) + support for "info" function for all supported filetypes (SpartanJ) + 1.31 (2011-06-20) + a few more leak fixes, bug in PNG handling (SpartanJ) + 1.30 (2011-06-11) + added ability to load files via callbacks to accomidate custom input streams (Ben Wenger) + removed deprecated format-specific test/load functions + removed support for installable file formats (stbi_loader) -- would have been broken for IO callbacks anyway + error cases in bmp and tga give messages and don't leak (Raymond Barbiero, grisha) + fix inefficiency in decoding 32-bit BMP (David Woo) + 1.29 (2010-08-16) + various warning fixes from Aurelien Pocheville + 1.28 (2010-08-01) + fix bug in GIF palette transparency (SpartanJ) + 1.27 (2010-08-01) + cast-to-stbi_uc to fix warnings + 1.26 (2010-07-24) + fix bug in file buffering for PNG reported by SpartanJ + 1.25 (2010-07-17) + refix trans_data warning (Won Chun) + 1.24 (2010-07-12) + perf improvements reading from files on platforms with lock-heavy fgetc() + minor perf improvements for jpeg + deprecated type-specific functions so we'll get feedback if they're needed + attempt to fix trans_data warning (Won Chun) + 1.23 fixed bug in iPhone support + 1.22 (2010-07-10) + removed image *writing* support + stbi_info support from Jetro Lauha + GIF support from Jean-Marc Lienher + iPhone PNG-extensions from James Brown + warning-fixes from Nicolas Schulz and Janez Zemva (i.stbi__err. Janez (U+017D)emva) + 1.21 fix use of 'stbi_uc' in header (reported by jon blow) + 1.20 added support for Softimage PIC, by Tom Seddon + 1.19 bug in interlaced PNG corruption check (found by ryg) + 1.18 (2008-08-02) + fix a threading bug (local mutable static) + 1.17 support interlaced PNG + 1.16 major bugfix - stbi__convert_format converted one too many pixels + 1.15 initialize some fields for thread safety + 1.14 fix threadsafe conversion bug + header-file-only version (#define STBI_HEADER_FILE_ONLY before including) + 1.13 threadsafe + 1.12 const qualifiers in the API + 1.11 Support installable IDCT, colorspace conversion routines + 1.10 Fixes for 64-bit (don't use "unsigned long") + optimized upsampling by Fabian "ryg" Giesen + 1.09 Fix format-conversion for PSD code (bad global variables!) + 1.08 Thatcher Ulrich's PSD code integrated by Nicolas Schulz + 1.07 attempt to fix C++ warning/errors again + 1.06 attempt to fix C++ warning/errors again + 1.05 fix TGA loading to return correct *comp and use good luminance calc + 1.04 default float alpha is 1, not 255; use 'void *' for stbi_image_free + 1.03 bugfixes to STBI_NO_STDIO, STBI_NO_HDR + 1.02 support for (subset of) HDR files, float interface for preferred access to them + 1.01 fix bug: possible bug in handling right-side up bmps... not sure + fix bug: the stbi__bmp_load() and stbi__tga_load() functions didn't work at all + 1.00 interface to zlib that skips zlib header + 0.99 correct handling of alpha in palette + 0.98 TGA loader by lonesock; dynamically add loaders (untested) + 0.97 jpeg errors on too large a file; also catch another malloc failure + 0.96 fix detection of invalid v value - particleman@mollyrocket forum + 0.95 during header scan, seek to markers in case of padding + 0.94 STBI_NO_STDIO to disable stdio usage; rename all #defines the same + 0.93 handle jpegtran output; verbose errors + 0.92 read 4,8,16,24,32-bit BMP files of several formats + 0.91 output 24-bit Windows 3.0 BMP files + 0.90 fix a few more warnings; bump version number to approach 1.0 + 0.61 bugfixes due to Marc LeBlanc, Christopher Lloyd + 0.60 fix compiling as c++ + 0.59 fix warnings: merge Dave Moore's -Wall fixes + 0.58 fix bug: zlib uncompressed mode len/nlen was wrong endian + 0.57 fix bug: jpg last huffman symbol before marker was >9 bits but less than 16 available + 0.56 fix bug: zlib uncompressed mode len vs. nlen + 0.55 fix bug: restart_interval not initialized to 0 + 0.54 allow NULL for 'int *comp' + 0.53 fix bug in png 3->4; speedup png decoding + 0.52 png handles req_comp=3,4 directly; minor cleanup; jpeg comments + 0.51 obey req_comp requests, 1-component jpegs return as 1-component, + on 'test' only check type, not whether we support this variant + 0.50 (2006-11-19) + first released version +*/ diff --git a/source/pc/stb_vorbis.c b/source/pc/stb_vorbis.c new file mode 100644 index 0000000..d54087f --- /dev/null +++ b/source/pc/stb_vorbis.c @@ -0,0 +1,5462 @@ +// Ogg Vorbis audio decoder - v1.07 - public domain +// http://nothings.org/stb_vorbis/ +// +// Original version written by Sean Barrett in 2007. +// +// Originally sponsored by RAD Game Tools. Seeking sponsored +// by Phillip Bennefall, Marc Andersen, Aaron Baker, Elias Software, +// Aras Pranckevicius, and Sean Barrett. +// +// LICENSE +// +// This software is in the public domain. Where that dedication is not +// recognized, you are granted a perpetual, irrevocable license to copy, +// distribute, and modify this file as you see fit. +// +// No warranty for any purpose is expressed or implied by the author (nor +// by RAD Game Tools). Report bugs and send enhancements to the author. +// +// Limitations: +// +// - floor 0 not supported (used in old ogg vorbis files pre-2004) +// - lossless sample-truncation at beginning ignored +// - cannot concatenate multiple vorbis streams +// - sample positions are 32-bit, limiting seekable 192Khz +// files to around 6 hours (Ogg supports 64-bit) +// +// Feature contributors: +// Dougall Johnson (sample-exact seeking) +// +// Bugfix/warning contributors: +// Terje Mathisen Niklas Frykholm Andy Hill +// Casey Muratori John Bolton Gargaj +// Laurent Gomila Marc LeBlanc Ronny Chevalier +// Bernhard Wodo Evan Balster "alxprd"@github +// Tom Beaumont Ingo Leitgeb Nicolas Guillemot +// Phillip Bennefall Rohit +// +// Partial history: +// 1.07 - 2015/01/16 - fixes for crashes on invalid files; warning fixes; const +// 1.06 - 2015/08/31 - full, correct support for seeking API (Dougall Johnson) +// some crash fixes when out of memory or with corrupt files +// fix some inappropriately signed shifts +// 1.05 - 2015/04/19 - don't define __forceinline if it's redundant +// 1.04 - 2014/08/27 - fix missing const-correct case in API +// 1.03 - 2014/08/07 - warning fixes +// 1.02 - 2014/07/09 - declare qsort comparison as explicitly _cdecl in Windows +// 1.01 - 2014/06/18 - fix stb_vorbis_get_samples_float (interleaved was correct) +// 1.0 - 2014/05/26 - fix memory leaks; fix warnings; fix bugs in >2-channel; +// (API change) report sample rate for decode-full-file funcs +// +// See end of file for full version history. + + +////////////////////////////////////////////////////////////////////////////// +// +// HEADER BEGINS HERE +// + +#ifndef STB_VORBIS_INCLUDE_STB_VORBIS_H +#define STB_VORBIS_INCLUDE_STB_VORBIS_H + +#if defined(STB_VORBIS_NO_CRT) && !defined(STB_VORBIS_NO_STDIO) +#define STB_VORBIS_NO_STDIO 1 +#endif + +#ifndef STB_VORBIS_NO_STDIO +#include +#endif + +#ifdef __cplusplus +extern "C" { +#endif + +/////////// THREAD SAFETY + +// Individual stb_vorbis* handles are not thread-safe; you cannot decode from +// them from multiple threads at the same time. However, you can have multiple +// stb_vorbis* handles and decode from them independently in multiple thrads. + + +/////////// MEMORY ALLOCATION + +// normally stb_vorbis uses malloc() to allocate memory at startup, +// and alloca() to allocate temporary memory during a frame on the +// stack. (Memory consumption will depend on the amount of setup +// data in the file and how you set the compile flags for speed +// vs. size. In my test files the maximal-size usage is ~150KB.) +// +// You can modify the wrapper functions in the source (setup_malloc, +// setup_temp_malloc, temp_malloc) to change this behavior, or you +// can use a simpler allocation model: you pass in a buffer from +// which stb_vorbis will allocate _all_ its memory (including the +// temp memory). "open" may fail with a VORBIS_outofmem if you +// do not pass in enough data; there is no way to determine how +// much you do need except to succeed (at which point you can +// query get_info to find the exact amount required. yes I know +// this is lame). +// +// If you pass in a non-NULL buffer of the type below, allocation +// will occur from it as described above. Otherwise just pass NULL +// to use malloc()/alloca() + +typedef struct +{ + char *alloc_buffer; + int alloc_buffer_length_in_bytes; +} stb_vorbis_alloc; + + +/////////// FUNCTIONS USEABLE WITH ALL INPUT MODES + +typedef struct stb_vorbis stb_vorbis; + +typedef struct +{ + unsigned int sample_rate; + int channels; + + unsigned int setup_memory_required; + unsigned int setup_temp_memory_required; + unsigned int temp_memory_required; + + int max_frame_size; +} stb_vorbis_info; + +// get general information about the file +extern stb_vorbis_info stb_vorbis_get_info(stb_vorbis *f); + +// get the last error detected (clears it, too) +extern int stb_vorbis_get_error(stb_vorbis *f); + +// close an ogg vorbis file and free all memory in use +extern void stb_vorbis_close(stb_vorbis *f); + +// this function returns the offset (in samples) from the beginning of the +// file that will be returned by the next decode, if it is known, or -1 +// otherwise. after a flush_pushdata() call, this may take a while before +// it becomes valid again. +// NOT WORKING YET after a seek with PULLDATA API +extern int stb_vorbis_get_sample_offset(stb_vorbis *f); + +// returns the current seek point within the file, or offset from the beginning +// of the memory buffer. In pushdata mode it returns 0. +extern unsigned int stb_vorbis_get_file_offset(stb_vorbis *f); + +/////////// PUSHDATA API + +#ifndef STB_VORBIS_NO_PUSHDATA_API + +// this API allows you to get blocks of data from any source and hand +// them to stb_vorbis. you have to buffer them; stb_vorbis will tell +// you how much it used, and you have to give it the rest next time; +// and stb_vorbis may not have enough data to work with and you will +// need to give it the same data again PLUS more. Note that the Vorbis +// specification does not bound the size of an individual frame. + +extern stb_vorbis *stb_vorbis_open_pushdata( + const unsigned char * datablock, int datablock_length_in_bytes, + int *datablock_memory_consumed_in_bytes, + int *error, + const stb_vorbis_alloc *alloc_buffer); +// create a vorbis decoder by passing in the initial data block containing +// the ogg&vorbis headers (you don't need to do parse them, just provide +// the first N bytes of the file--you're told if it's not enough, see below) +// on success, returns an stb_vorbis *, does not set error, returns the amount of +// data parsed/consumed on this call in *datablock_memory_consumed_in_bytes; +// on failure, returns NULL on error and sets *error, does not change *datablock_memory_consumed +// if returns NULL and *error is VORBIS_need_more_data, then the input block was +// incomplete and you need to pass in a larger block from the start of the file + +extern int stb_vorbis_decode_frame_pushdata( + stb_vorbis *f, + const unsigned char *datablock, int datablock_length_in_bytes, + int *channels, // place to write number of float * buffers + float ***output, // place to write float ** array of float * buffers + int *samples // place to write number of output samples + ); +// decode a frame of audio sample data if possible from the passed-in data block +// +// return value: number of bytes we used from datablock +// +// possible cases: +// 0 bytes used, 0 samples output (need more data) +// N bytes used, 0 samples output (resynching the stream, keep going) +// N bytes used, M samples output (one frame of data) +// note that after opening a file, you will ALWAYS get one N-bytes,0-sample +// frame, because Vorbis always "discards" the first frame. +// +// Note that on resynch, stb_vorbis will rarely consume all of the buffer, +// instead only datablock_length_in_bytes-3 or less. This is because it wants +// to avoid missing parts of a page header if they cross a datablock boundary, +// without writing state-machiney code to record a partial detection. +// +// The number of channels returned are stored in *channels (which can be +// NULL--it is always the same as the number of channels reported by +// get_info). *output will contain an array of float* buffers, one per +// channel. In other words, (*output)[0][0] contains the first sample from +// the first channel, and (*output)[1][0] contains the first sample from +// the second channel. + +extern void stb_vorbis_flush_pushdata(stb_vorbis *f); +// inform stb_vorbis that your next datablock will not be contiguous with +// previous ones (e.g. you've seeked in the data); future attempts to decode +// frames will cause stb_vorbis to resynchronize (as noted above), and +// once it sees a valid Ogg page (typically 4-8KB, as large as 64KB), it +// will begin decoding the _next_ frame. +// +// if you want to seek using pushdata, you need to seek in your file, then +// call stb_vorbis_flush_pushdata(), then start calling decoding, then once +// decoding is returning you data, call stb_vorbis_get_sample_offset, and +// if you don't like the result, seek your file again and repeat. +#endif + + +////////// PULLING INPUT API + +#ifndef STB_VORBIS_NO_PULLDATA_API +// This API assumes stb_vorbis is allowed to pull data from a source-- +// either a block of memory containing the _entire_ vorbis stream, or a +// FILE * that you or it create, or possibly some other reading mechanism +// if you go modify the source to replace the FILE * case with some kind +// of callback to your code. (But if you don't support seeking, you may +// just want to go ahead and use pushdata.) + +#if !defined(STB_VORBIS_NO_STDIO) && !defined(STB_VORBIS_NO_INTEGER_CONVERSION) +extern int stb_vorbis_decode_filename(const char *filename, int *channels, int *sample_rate, short **output); +#endif +#if !defined(STB_VORBIS_NO_INTEGER_CONVERSION) +extern int stb_vorbis_decode_memory(const unsigned char *mem, int len, int *channels, int *sample_rate, short **output); +#endif +// decode an entire file and output the data interleaved into a malloc()ed +// buffer stored in *output. The return value is the number of samples +// decoded, or -1 if the file could not be opened or was not an ogg vorbis file. +// When you're done with it, just free() the pointer returned in *output. + +extern stb_vorbis * stb_vorbis_open_memory(const unsigned char *data, int len, + int *error, const stb_vorbis_alloc *alloc_buffer); +// create an ogg vorbis decoder from an ogg vorbis stream in memory (note +// this must be the entire stream!). on failure, returns NULL and sets *error + +#ifndef STB_VORBIS_NO_STDIO +extern stb_vorbis * stb_vorbis_open_filename(const char *filename, + int *error, const stb_vorbis_alloc *alloc_buffer); +// create an ogg vorbis decoder from a filename via fopen(). on failure, +// returns NULL and sets *error (possibly to VORBIS_file_open_failure). + +extern stb_vorbis * stb_vorbis_open_file(FILE *f, int close_handle_on_close, + int *error, const stb_vorbis_alloc *alloc_buffer); +// create an ogg vorbis decoder from an open FILE *, looking for a stream at +// the _current_ seek point (ftell). on failure, returns NULL and sets *error. +// note that stb_vorbis must "own" this stream; if you seek it in between +// calls to stb_vorbis, it will become confused. Morever, if you attempt to +// perform stb_vorbis_seek_*() operations on this file, it will assume it +// owns the _entire_ rest of the file after the start point. Use the next +// function, stb_vorbis_open_file_section(), to limit it. + +extern stb_vorbis * stb_vorbis_open_file_section(FILE *f, int close_handle_on_close, + int *error, const stb_vorbis_alloc *alloc_buffer, unsigned int len); +// create an ogg vorbis decoder from an open FILE *, looking for a stream at +// the _current_ seek point (ftell); the stream will be of length 'len' bytes. +// on failure, returns NULL and sets *error. note that stb_vorbis must "own" +// this stream; if you seek it in between calls to stb_vorbis, it will become +// confused. +#endif + +extern int stb_vorbis_seek_frame(stb_vorbis *f, unsigned int sample_number); +extern int stb_vorbis_seek(stb_vorbis *f, unsigned int sample_number); +// these functions seek in the Vorbis file to (approximately) 'sample_number'. +// after calling seek_frame(), the next call to get_frame_*() will include +// the specified sample. after calling stb_vorbis_seek(), the next call to +// stb_vorbis_get_samples_* will start with the specified sample. If you +// do not need to seek to EXACTLY the target sample when using get_samples_*, +// you can also use seek_frame(). + +extern void stb_vorbis_seek_start(stb_vorbis *f); +// this function is equivalent to stb_vorbis_seek(f,0) + +extern unsigned int stb_vorbis_stream_length_in_samples(stb_vorbis *f); +extern float stb_vorbis_stream_length_in_seconds(stb_vorbis *f); +// these functions return the total length of the vorbis stream + +extern int stb_vorbis_get_frame_float(stb_vorbis *f, int *channels, float ***output); +// decode the next frame and return the number of samples. the number of +// channels returned are stored in *channels (which can be NULL--it is always +// the same as the number of channels reported by get_info). *output will +// contain an array of float* buffers, one per channel. These outputs will +// be overwritten on the next call to stb_vorbis_get_frame_*. +// +// You generally should not intermix calls to stb_vorbis_get_frame_*() +// and stb_vorbis_get_samples_*(), since the latter calls the former. + +#ifndef STB_VORBIS_NO_INTEGER_CONVERSION +extern int stb_vorbis_get_frame_short_interleaved(stb_vorbis *f, int num_c, short *buffer, int num_shorts); +extern int stb_vorbis_get_frame_short (stb_vorbis *f, int num_c, short **buffer, int num_samples); +#endif +// decode the next frame and return the number of *samples* per channel. +// Note that for interleaved data, you pass in the number of shorts (the +// size of your array), but the return value is the number of samples per +// channel, not the total number of samples. +// +// The data is coerced to the number of channels you request according to the +// channel coercion rules (see below). You must pass in the size of your +// buffer(s) so that stb_vorbis will not overwrite the end of the buffer. +// The maximum buffer size needed can be gotten from get_info(); however, +// the Vorbis I specification implies an absolute maximum of 4096 samples +// per channel. + +// Channel coercion rules: +// Let M be the number of channels requested, and N the number of channels present, +// and Cn be the nth channel; let stereo L be the sum of all L and center channels, +// and stereo R be the sum of all R and center channels (channel assignment from the +// vorbis spec). +// M N output +// 1 k sum(Ck) for all k +// 2 * stereo L, stereo R +// k l k > l, the first l channels, then 0s +// k l k <= l, the first k channels +// Note that this is not _good_ surround etc. mixing at all! It's just so +// you get something useful. + +extern int stb_vorbis_get_samples_float_interleaved(stb_vorbis *f, int channels, float *buffer, int num_floats); +extern int stb_vorbis_get_samples_float(stb_vorbis *f, int channels, float **buffer, int num_samples); +// gets num_samples samples, not necessarily on a frame boundary--this requires +// buffering so you have to supply the buffers. DOES NOT APPLY THE COERCION RULES. +// Returns the number of samples stored per channel; it may be less than requested +// at the end of the file. If there are no more samples in the file, returns 0. + +#ifndef STB_VORBIS_NO_INTEGER_CONVERSION +extern int stb_vorbis_get_samples_short_interleaved(stb_vorbis *f, int channels, short *buffer, int num_shorts); +extern int stb_vorbis_get_samples_short(stb_vorbis *f, int channels, short **buffer, int num_samples); +#endif +// gets num_samples samples, not necessarily on a frame boundary--this requires +// buffering so you have to supply the buffers. Applies the coercion rules above +// to produce 'channels' channels. Returns the number of samples stored per channel; +// it may be less than requested at the end of the file. If there are no more +// samples in the file, returns 0. + +#endif + +//////// ERROR CODES + +enum STBVorbisError +{ + VORBIS__no_error, + + VORBIS_need_more_data=1, // not a real error + + VORBIS_invalid_api_mixing, // can't mix API modes + VORBIS_outofmem, // not enough memory + VORBIS_feature_not_supported, // uses floor 0 + VORBIS_too_many_channels, // STB_VORBIS_MAX_CHANNELS is too small + VORBIS_file_open_failure, // fopen() failed + VORBIS_seek_without_length, // can't seek in unknown-length file + + VORBIS_unexpected_eof=10, // file is truncated? + VORBIS_seek_invalid, // seek past EOF + + // decoding errors (corrupt/invalid stream) -- you probably + // don't care about the exact details of these + + // vorbis errors: + VORBIS_invalid_setup=20, + VORBIS_invalid_stream, + + // ogg errors: + VORBIS_missing_capture_pattern=30, + VORBIS_invalid_stream_structure_version, + VORBIS_continued_packet_flag_invalid, + VORBIS_incorrect_stream_serial_number, + VORBIS_invalid_first_page, + VORBIS_bad_packet_type, + VORBIS_cant_find_last_page, + VORBIS_seek_failed +}; + + +#ifdef __cplusplus +} +#endif + +#endif // STB_VORBIS_INCLUDE_STB_VORBIS_H +// +// HEADER ENDS HERE +// +////////////////////////////////////////////////////////////////////////////// + +#ifndef STB_VORBIS_HEADER_ONLY + +// global configuration settings (e.g. set these in the project/makefile), +// or just set them in this file at the top (although ideally the first few +// should be visible when the header file is compiled too, although it's not +// crucial) + +// STB_VORBIS_NO_PUSHDATA_API +// does not compile the code for the various stb_vorbis_*_pushdata() +// functions +// #define STB_VORBIS_NO_PUSHDATA_API + +// STB_VORBIS_NO_PULLDATA_API +// does not compile the code for the non-pushdata APIs +// #define STB_VORBIS_NO_PULLDATA_API + +// STB_VORBIS_NO_STDIO +// does not compile the code for the APIs that use FILE *s internally +// or externally (implied by STB_VORBIS_NO_PULLDATA_API) +// #define STB_VORBIS_NO_STDIO + +// STB_VORBIS_NO_INTEGER_CONVERSION +// does not compile the code for converting audio sample data from +// float to integer (implied by STB_VORBIS_NO_PULLDATA_API) +// #define STB_VORBIS_NO_INTEGER_CONVERSION + +// STB_VORBIS_NO_FAST_SCALED_FLOAT +// does not use a fast float-to-int trick to accelerate float-to-int on +// most platforms which requires endianness be defined correctly. +//#define STB_VORBIS_NO_FAST_SCALED_FLOAT + + +// STB_VORBIS_MAX_CHANNELS [number] +// globally define this to the maximum number of channels you need. +// The spec does not put a restriction on channels except that +// the count is stored in a byte, so 255 is the hard limit. +// Reducing this saves about 16 bytes per value, so using 16 saves +// (255-16)*16 or around 4KB. Plus anything other memory usage +// I forgot to account for. Can probably go as low as 8 (7.1 audio), +// 6 (5.1 audio), or 2 (stereo only). +#ifndef STB_VORBIS_MAX_CHANNELS +#define STB_VORBIS_MAX_CHANNELS 16 // enough for anyone? +#endif + +// STB_VORBIS_PUSHDATA_CRC_COUNT [number] +// after a flush_pushdata(), stb_vorbis begins scanning for the +// next valid page, without backtracking. when it finds something +// that looks like a page, it streams through it and verifies its +// CRC32. Should that validation fail, it keeps scanning. But it's +// possible that _while_ streaming through to check the CRC32 of +// one candidate page, it sees another candidate page. This #define +// determines how many "overlapping" candidate pages it can search +// at once. Note that "real" pages are typically ~4KB to ~8KB, whereas +// garbage pages could be as big as 64KB, but probably average ~16KB. +// So don't hose ourselves by scanning an apparent 64KB page and +// missing a ton of real ones in the interim; so minimum of 2 +#ifndef STB_VORBIS_PUSHDATA_CRC_COUNT +#define STB_VORBIS_PUSHDATA_CRC_COUNT 4 +#endif + +// STB_VORBIS_FAST_HUFFMAN_LENGTH [number] +// sets the log size of the huffman-acceleration table. Maximum +// supported value is 24. with larger numbers, more decodings are O(1), +// but the table size is larger so worse cache missing, so you'll have +// to probe (and try multiple ogg vorbis files) to find the sweet spot. +#ifndef STB_VORBIS_FAST_HUFFMAN_LENGTH +#define STB_VORBIS_FAST_HUFFMAN_LENGTH 10 +#endif + +// STB_VORBIS_FAST_BINARY_LENGTH [number] +// sets the log size of the binary-search acceleration table. this +// is used in similar fashion to the fast-huffman size to set initial +// parameters for the binary search + +// STB_VORBIS_FAST_HUFFMAN_INT +// The fast huffman tables are much more efficient if they can be +// stored as 16-bit results instead of 32-bit results. This restricts +// the codebooks to having only 65535 possible outcomes, though. +// (At least, accelerated by the huffman table.) +#ifndef STB_VORBIS_FAST_HUFFMAN_INT +#define STB_VORBIS_FAST_HUFFMAN_SHORT +#endif + +// STB_VORBIS_NO_HUFFMAN_BINARY_SEARCH +// If the 'fast huffman' search doesn't succeed, then stb_vorbis falls +// back on binary searching for the correct one. This requires storing +// extra tables with the huffman codes in sorted order. Defining this +// symbol trades off space for speed by forcing a linear search in the +// non-fast case, except for "sparse" codebooks. +// #define STB_VORBIS_NO_HUFFMAN_BINARY_SEARCH + +// STB_VORBIS_DIVIDES_IN_RESIDUE +// stb_vorbis precomputes the result of the scalar residue decoding +// that would otherwise require a divide per chunk. you can trade off +// space for time by defining this symbol. +// #define STB_VORBIS_DIVIDES_IN_RESIDUE + +// STB_VORBIS_DIVIDES_IN_CODEBOOK +// vorbis VQ codebooks can be encoded two ways: with every case explicitly +// stored, or with all elements being chosen from a small range of values, +// and all values possible in all elements. By default, stb_vorbis expands +// this latter kind out to look like the former kind for ease of decoding, +// because otherwise an integer divide-per-vector-element is required to +// unpack the index. If you define STB_VORBIS_DIVIDES_IN_CODEBOOK, you can +// trade off storage for speed. +//#define STB_VORBIS_DIVIDES_IN_CODEBOOK + +#ifdef STB_VORBIS_CODEBOOK_SHORTS +#error "STB_VORBIS_CODEBOOK_SHORTS is no longer supported as it produced incorrect results for some input formats" +#endif + +// STB_VORBIS_DIVIDE_TABLE +// this replaces small integer divides in the floor decode loop with +// table lookups. made less than 1% difference, so disabled by default. + +// STB_VORBIS_NO_INLINE_DECODE +// disables the inlining of the scalar codebook fast-huffman decode. +// might save a little codespace; useful for debugging +// #define STB_VORBIS_NO_INLINE_DECODE + +// STB_VORBIS_NO_DEFER_FLOOR +// Normally we only decode the floor without synthesizing the actual +// full curve. We can instead synthesize the curve immediately. This +// requires more memory and is very likely slower, so I don't think +// you'd ever want to do it except for debugging. +// #define STB_VORBIS_NO_DEFER_FLOOR + + + + +////////////////////////////////////////////////////////////////////////////// + +#ifdef STB_VORBIS_NO_PULLDATA_API + #define STB_VORBIS_NO_INTEGER_CONVERSION + #define STB_VORBIS_NO_STDIO +#endif + +#if defined(STB_VORBIS_NO_CRT) && !defined(STB_VORBIS_NO_STDIO) + #define STB_VORBIS_NO_STDIO 1 +#endif + +#ifndef STB_VORBIS_NO_INTEGER_CONVERSION +#ifndef STB_VORBIS_NO_FAST_SCALED_FLOAT + + // only need endianness for fast-float-to-int, which we don't + // use for pushdata + + #ifndef STB_VORBIS_BIG_ENDIAN + #define STB_VORBIS_ENDIAN 0 + #else + #define STB_VORBIS_ENDIAN 1 + #endif + +#endif +#endif + + +#ifndef STB_VORBIS_NO_STDIO +#include +#endif + +#ifndef STB_VORBIS_NO_CRT +#include +#include +#include +#include +#if !(defined(__APPLE__) || defined(MACOSX) || defined(macintosh) || defined(Macintosh)) +#include +#endif +#else // STB_VORBIS_NO_CRT +#define NULL 0 +#define malloc(s) 0 +#define free(s) ((void) 0) +#define realloc(s) 0 +#endif // STB_VORBIS_NO_CRT + +#include + +#ifdef __MINGW32__ + // eff you mingw: + // "fixed": + // http://sourceforge.net/p/mingw-w64/mailman/message/32882927/ + // "no that broke the build, reverted, who cares about C": + // http://sourceforge.net/p/mingw-w64/mailman/message/32890381/ + #ifdef __forceinline + #undef __forceinline + #endif + #define __forceinline +#elif !defined(_MSC_VER) + #if __GNUC__ + #define __forceinline inline + #else + #define __forceinline + #endif +#endif + +#if STB_VORBIS_MAX_CHANNELS > 256 +#error "Value of STB_VORBIS_MAX_CHANNELS outside of allowed range" +#endif + +#if STB_VORBIS_FAST_HUFFMAN_LENGTH > 24 +#error "Value of STB_VORBIS_FAST_HUFFMAN_LENGTH outside of allowed range" +#endif + + +#if 0 +#include +#define CHECK(f) _CrtIsValidHeapPointer(f->channel_buffers[1]) +#else +#define CHECK(f) ((void) 0) +#endif + +#define MAX_BLOCKSIZE_LOG 13 // from specification +#define MAX_BLOCKSIZE (1 << MAX_BLOCKSIZE_LOG) + + +typedef unsigned char uint8; +typedef signed char int8; +typedef unsigned short uint16; +typedef signed short int16; +typedef unsigned int uint32; +typedef signed int int32; + +#ifndef TRUE +#define TRUE 1 +#define FALSE 0 +#endif + +typedef float codetype; + +// @NOTE +// +// Some arrays below are tagged "//varies", which means it's actually +// a variable-sized piece of data, but rather than malloc I assume it's +// small enough it's better to just allocate it all together with the +// main thing +// +// Most of the variables are specified with the smallest size I could pack +// them into. It might give better performance to make them all full-sized +// integers. It should be safe to freely rearrange the structures or change +// the sizes larger--nothing relies on silently truncating etc., nor the +// order of variables. + +#define FAST_HUFFMAN_TABLE_SIZE (1 << STB_VORBIS_FAST_HUFFMAN_LENGTH) +#define FAST_HUFFMAN_TABLE_MASK (FAST_HUFFMAN_TABLE_SIZE - 1) + +typedef struct +{ + int dimensions, entries; + uint8 *codeword_lengths; + float minimum_value; + float delta_value; + uint8 value_bits; + uint8 lookup_type; + uint8 sequence_p; + uint8 sparse; + uint32 lookup_values; + codetype *multiplicands; + uint32 *codewords; + #ifdef STB_VORBIS_FAST_HUFFMAN_SHORT + int16 fast_huffman[FAST_HUFFMAN_TABLE_SIZE]; + #else + int32 fast_huffman[FAST_HUFFMAN_TABLE_SIZE]; + #endif + uint32 *sorted_codewords; + int *sorted_values; + int sorted_entries; +} Codebook; + +typedef struct +{ + uint8 order; + uint16 rate; + uint16 bark_map_size; + uint8 amplitude_bits; + uint8 amplitude_offset; + uint8 number_of_books; + uint8 book_list[16]; // varies +} Floor0; + +typedef struct +{ + uint8 partitions; + uint8 partition_class_list[32]; // varies + uint8 class_dimensions[16]; // varies + uint8 class_subclasses[16]; // varies + uint8 class_masterbooks[16]; // varies + int16 subclass_books[16][8]; // varies + uint16 Xlist[31*8+2]; // varies + uint8 sorted_order[31*8+2]; + uint8 neighbors[31*8+2][2]; + uint8 floor1_multiplier; + uint8 rangebits; + int values; +} Floor1; + +typedef union +{ + Floor0 floor0; + Floor1 floor1; +} Floor; + +typedef struct +{ + uint32 begin, end; + uint32 part_size; + uint8 classifications; + uint8 classbook; + uint8 **classdata; + int16 (*residue_books)[8]; +} Residue; + +typedef struct +{ + uint8 magnitude; + uint8 angle; + uint8 mux; +} MappingChannel; + +typedef struct +{ + uint16 coupling_steps; + MappingChannel *chan; + uint8 submaps; + uint8 submap_floor[15]; // varies + uint8 submap_residue[15]; // varies +} Mapping; + +typedef struct +{ + uint8 blockflag; + uint8 mapping; + uint16 windowtype; + uint16 transformtype; +} Mode; + +typedef struct +{ + uint32 goal_crc; // expected crc if match + int bytes_left; // bytes left in packet + uint32 crc_so_far; // running crc + int bytes_done; // bytes processed in _current_ chunk + uint32 sample_loc; // granule pos encoded in page +} CRCscan; + +typedef struct +{ + uint32 page_start, page_end; + uint32 last_decoded_sample; +} ProbedPage; + +struct stb_vorbis +{ + // user-accessible info + unsigned int sample_rate; + int channels; + + unsigned int setup_memory_required; + unsigned int temp_memory_required; + unsigned int setup_temp_memory_required; + + // input config +#ifndef STB_VORBIS_NO_STDIO + FILE *f; + uint32 f_start; + int close_on_free; +#endif + + uint8 *stream; + uint8 *stream_start; + uint8 *stream_end; + + uint32 stream_len; + + uint8 push_mode; + + uint32 first_audio_page_offset; + + ProbedPage p_first, p_last; + + // memory management + stb_vorbis_alloc alloc; + int setup_offset; + int temp_offset; + + // run-time results + int eof; + enum STBVorbisError error; + + // user-useful data + + // header info + int blocksize[2]; + int blocksize_0, blocksize_1; + int codebook_count; + Codebook *codebooks; + int floor_count; + uint16 floor_types[64]; // varies + Floor *floor_config; + int residue_count; + uint16 residue_types[64]; // varies + Residue *residue_config; + int mapping_count; + Mapping *mapping; + int mode_count; + Mode mode_config[64]; // varies + + uint32 total_samples; + + // decode buffer + float *channel_buffers[STB_VORBIS_MAX_CHANNELS]; + float *outputs [STB_VORBIS_MAX_CHANNELS]; + + float *previous_window[STB_VORBIS_MAX_CHANNELS]; + int previous_length; + + #ifndef STB_VORBIS_NO_DEFER_FLOOR + int16 *finalY[STB_VORBIS_MAX_CHANNELS]; + #else + float *floor_buffers[STB_VORBIS_MAX_CHANNELS]; + #endif + + uint32 current_loc; // sample location of next frame to decode + int current_loc_valid; + + // per-blocksize precomputed data + + // twiddle factors + float *A[2],*B[2],*C[2]; + float *window[2]; + uint16 *bit_reverse[2]; + + // current page/packet/segment streaming info + uint32 serial; // stream serial number for verification + int last_page; + int segment_count; + uint8 segments[255]; + uint8 page_flag; + uint8 bytes_in_seg; + uint8 first_decode; + int next_seg; + int last_seg; // flag that we're on the last segment + int last_seg_which; // what was the segment number of the last seg? + uint32 acc; + int valid_bits; + int packet_bytes; + int end_seg_with_known_loc; + uint32 known_loc_for_packet; + int discard_samples_deferred; + uint32 samples_output; + + // push mode scanning + int page_crc_tests; // only in push_mode: number of tests active; -1 if not searching +#ifndef STB_VORBIS_NO_PUSHDATA_API + CRCscan scan[STB_VORBIS_PUSHDATA_CRC_COUNT]; +#endif + + // sample-access + int channel_buffer_start; + int channel_buffer_end; +}; + +#if defined(STB_VORBIS_NO_PUSHDATA_API) + #define IS_PUSH_MODE(f) FALSE +#elif defined(STB_VORBIS_NO_PULLDATA_API) + #define IS_PUSH_MODE(f) TRUE +#else + #define IS_PUSH_MODE(f) ((f)->push_mode) +#endif + +typedef struct stb_vorbis vorb; + +static int error(vorb *f, enum STBVorbisError e) +{ + f->error = e; + if (!f->eof && e != VORBIS_need_more_data) { + f->error=e; // breakpoint for debugging + } + return 0; +} + + +// these functions are used for allocating temporary memory +// while decoding. if you can afford the stack space, use +// alloca(); otherwise, provide a temp buffer and it will +// allocate out of those. + +#define array_size_required(count,size) (count*(sizeof(void *)+(size))) + +#define temp_alloc(f,size) (f->alloc.alloc_buffer ? setup_temp_malloc(f,size) : alloca(size)) +#ifdef dealloca +#define temp_free(f,p) (f->alloc.alloc_buffer ? 0 : dealloca(size)) +#else +#define temp_free(f,p) 0 +#endif +#define temp_alloc_save(f) ((f)->temp_offset) +#define temp_alloc_restore(f,p) ((f)->temp_offset = (p)) + +#define temp_block_array(f,count,size) make_block_array(temp_alloc(f,array_size_required(count,size)), count, size) + +// given a sufficiently large block of memory, make an array of pointers to subblocks of it +static void *make_block_array(void *mem, int count, int size) +{ + int i; + void ** p = (void **) mem; + char *q = (char *) (p + count); + for (i=0; i < count; ++i) { + p[i] = q; + q += size; + } + return p; +} + +static void *setup_malloc(vorb *f, int sz) +{ + sz = (sz+3) & ~3; + f->setup_memory_required += sz; + if (f->alloc.alloc_buffer) { + void *p = (char *) f->alloc.alloc_buffer + f->setup_offset; + if (f->setup_offset + sz > f->temp_offset) return NULL; + f->setup_offset += sz; + return p; + } + return sz ? malloc(sz) : NULL; +} + +static void setup_free(vorb *f, void *p) +{ + if (f->alloc.alloc_buffer) return; // do nothing; setup mem is a stack + free(p); +} + +static void *setup_temp_malloc(vorb *f, int sz) +{ + sz = (sz+3) & ~3; + if (f->alloc.alloc_buffer) { + if (f->temp_offset - sz < f->setup_offset) return NULL; + f->temp_offset -= sz; + return (char *) f->alloc.alloc_buffer + f->temp_offset; + } + return malloc(sz); +} + +static void setup_temp_free(vorb *f, void *p, int sz) +{ + if (f->alloc.alloc_buffer) { + f->temp_offset += (sz+3)&~3; + return; + } + free(p); +} + +#define CRC32_POLY 0x04c11db7 // from spec + +static uint32 crc_table[256]; +static void crc32_init(void) +{ + int i,j; + uint32 s; + for(i=0; i < 256; i++) { + for (s=(uint32) i << 24, j=0; j < 8; ++j) + s = (s << 1) ^ (s >= (1U<<31) ? CRC32_POLY : 0); + crc_table[i] = s; + } +} + +static __forceinline uint32 crc32_update(uint32 crc, uint8 byte) +{ + return (crc << 8) ^ crc_table[byte ^ (crc >> 24)]; +} + + +// used in setup, and for huffman that doesn't go fast path +static unsigned int bit_reverse(unsigned int n) +{ + n = ((n & 0xAAAAAAAA) >> 1) | ((n & 0x55555555) << 1); + n = ((n & 0xCCCCCCCC) >> 2) | ((n & 0x33333333) << 2); + n = ((n & 0xF0F0F0F0) >> 4) | ((n & 0x0F0F0F0F) << 4); + n = ((n & 0xFF00FF00) >> 8) | ((n & 0x00FF00FF) << 8); + return (n >> 16) | (n << 16); +} + +static float square(float x) +{ + return x*x; +} + +// this is a weird definition of log2() for which log2(1) = 1, log2(2) = 2, log2(4) = 3 +// as required by the specification. fast(?) implementation from stb.h +// @OPTIMIZE: called multiple times per-packet with "constants"; move to setup +static int ilog(int32 n) +{ + static signed char log2_4[16] = { 0,1,2,2,3,3,3,3,4,4,4,4,4,4,4,4 }; + + // 2 compares if n < 16, 3 compares otherwise (4 if signed or n > 1<<29) + if (n < (1 << 14)) + if (n < (1 << 4)) return 0 + log2_4[n ]; + else if (n < (1 << 9)) return 5 + log2_4[n >> 5]; + else return 10 + log2_4[n >> 10]; + else if (n < (1 << 24)) + if (n < (1 << 19)) return 15 + log2_4[n >> 15]; + else return 20 + log2_4[n >> 20]; + else if (n < (1 << 29)) return 25 + log2_4[n >> 25]; + else if (n < (1 << 31)) return 30 + log2_4[n >> 30]; + else return 0; // signed n returns 0 +} + +#ifndef M_PI + #define M_PI 3.14159265358979323846264f // from CRC +#endif + +// code length assigned to a value with no huffman encoding +#define NO_CODE 255 + +/////////////////////// LEAF SETUP FUNCTIONS ////////////////////////// +// +// these functions are only called at setup, and only a few times +// per file + +static float float32_unpack(uint32 x) +{ + // from the specification + uint32 mantissa = x & 0x1fffff; + uint32 sign = x & 0x80000000; + uint32 exp = (x & 0x7fe00000) >> 21; + double res = sign ? -(double)mantissa : (double)mantissa; + return (float) ldexp((float)res, exp-788); +} + + +// zlib & jpeg huffman tables assume that the output symbols +// can either be arbitrarily arranged, or have monotonically +// increasing frequencies--they rely on the lengths being sorted; +// this makes for a very simple generation algorithm. +// vorbis allows a huffman table with non-sorted lengths. This +// requires a more sophisticated construction, since symbols in +// order do not map to huffman codes "in order". +static void add_entry(Codebook *c, uint32 huff_code, int symbol, int count, int len, uint32 *values) +{ + if (!c->sparse) { + c->codewords [symbol] = huff_code; + } else { + c->codewords [count] = huff_code; + c->codeword_lengths[count] = len; + values [count] = symbol; + } +} + +static int compute_codewords(Codebook *c, uint8 *len, int n, uint32 *values) +{ + int i,k,m=0; + uint32 available[32]; + + memset(available, 0, sizeof(available)); + // find the first entry + for (k=0; k < n; ++k) if (len[k] < NO_CODE) break; + if (k == n) { assert(c->sorted_entries == 0); return TRUE; } + // add to the list + add_entry(c, 0, k, m++, len[k], values); + // add all available leaves + for (i=1; i <= len[k]; ++i) + available[i] = 1U << (32-i); + // note that the above code treats the first case specially, + // but it's really the same as the following code, so they + // could probably be combined (except the initial code is 0, + // and I use 0 in available[] to mean 'empty') + for (i=k+1; i < n; ++i) { + uint32 res; + int z = len[i], y; + if (z == NO_CODE) continue; + // find lowest available leaf (should always be earliest, + // which is what the specification calls for) + // note that this property, and the fact we can never have + // more than one free leaf at a given level, isn't totally + // trivial to prove, but it seems true and the assert never + // fires, so! + while (z > 0 && !available[z]) --z; + if (z == 0) { return FALSE; } + res = available[z]; + assert(z >= 0 && z < 32); + available[z] = 0; + add_entry(c, bit_reverse(res), i, m++, len[i], values); + // propogate availability up the tree + if (z != len[i]) { + assert(len[i] >= 0 && len[i] < 32); + for (y=len[i]; y > z; --y) { + assert(available[y] == 0); + available[y] = res + (1 << (32-y)); + } + } + } + return TRUE; +} + +// accelerated huffman table allows fast O(1) match of all symbols +// of length <= STB_VORBIS_FAST_HUFFMAN_LENGTH +static void compute_accelerated_huffman(Codebook *c) +{ + int i, len; + for (i=0; i < FAST_HUFFMAN_TABLE_SIZE; ++i) + c->fast_huffman[i] = -1; + + len = c->sparse ? c->sorted_entries : c->entries; + #ifdef STB_VORBIS_FAST_HUFFMAN_SHORT + if (len > 32767) len = 32767; // largest possible value we can encode! + #endif + for (i=0; i < len; ++i) { + if (c->codeword_lengths[i] <= STB_VORBIS_FAST_HUFFMAN_LENGTH) { + uint32 z = c->sparse ? bit_reverse(c->sorted_codewords[i]) : c->codewords[i]; + // set table entries for all bit combinations in the higher bits + while (z < FAST_HUFFMAN_TABLE_SIZE) { + c->fast_huffman[z] = i; + z += 1 << c->codeword_lengths[i]; + } + } + } +} + +#ifdef _MSC_VER +#define STBV_CDECL __cdecl +#else +#define STBV_CDECL +#endif + +static int STBV_CDECL uint32_compare(const void *p, const void *q) +{ + uint32 x = * (uint32 *) p; + uint32 y = * (uint32 *) q; + return x < y ? -1 : x > y; +} + +static int include_in_sort(Codebook *c, uint8 len) +{ + if (c->sparse) { assert(len != NO_CODE); return TRUE; } + if (len == NO_CODE) return FALSE; + if (len > STB_VORBIS_FAST_HUFFMAN_LENGTH) return TRUE; + return FALSE; +} + +// if the fast table above doesn't work, we want to binary +// search them... need to reverse the bits +static void compute_sorted_huffman(Codebook *c, uint8 *lengths, uint32 *values) +{ + int i, len; + // build a list of all the entries + // OPTIMIZATION: don't include the short ones, since they'll be caught by FAST_HUFFMAN. + // this is kind of a frivolous optimization--I don't see any performance improvement, + // but it's like 4 extra lines of code, so. + if (!c->sparse) { + int k = 0; + for (i=0; i < c->entries; ++i) + if (include_in_sort(c, lengths[i])) + c->sorted_codewords[k++] = bit_reverse(c->codewords[i]); + assert(k == c->sorted_entries); + } else { + for (i=0; i < c->sorted_entries; ++i) + c->sorted_codewords[i] = bit_reverse(c->codewords[i]); + } + + qsort(c->sorted_codewords, c->sorted_entries, sizeof(c->sorted_codewords[0]), uint32_compare); + c->sorted_codewords[c->sorted_entries] = 0xffffffff; + + len = c->sparse ? c->sorted_entries : c->entries; + // now we need to indicate how they correspond; we could either + // #1: sort a different data structure that says who they correspond to + // #2: for each sorted entry, search the original list to find who corresponds + // #3: for each original entry, find the sorted entry + // #1 requires extra storage, #2 is slow, #3 can use binary search! + for (i=0; i < len; ++i) { + int huff_len = c->sparse ? lengths[values[i]] : lengths[i]; + if (include_in_sort(c,huff_len)) { + uint32 code = bit_reverse(c->codewords[i]); + int x=0, n=c->sorted_entries; + while (n > 1) { + // invariant: sc[x] <= code < sc[x+n] + int m = x + (n >> 1); + if (c->sorted_codewords[m] <= code) { + x = m; + n -= (n>>1); + } else { + n >>= 1; + } + } + assert(c->sorted_codewords[x] == code); + if (c->sparse) { + c->sorted_values[x] = values[i]; + c->codeword_lengths[x] = huff_len; + } else { + c->sorted_values[x] = i; + } + } + } +} + +// only run while parsing the header (3 times) +static int vorbis_validate(uint8 *data) +{ + static uint8 vorbis[6] = { 'v', 'o', 'r', 'b', 'i', 's' }; + return memcmp(data, vorbis, 6) == 0; +} + +// called from setup only, once per code book +// (formula implied by specification) +static int lookup1_values(int entries, int dim) +{ + int r = (int) floor(exp((float) log((float) entries) / dim)); + if ((int) floor(pow((float) r+1, dim)) <= entries) // (int) cast for MinGW warning; + ++r; // floor() to avoid _ftol() when non-CRT + assert(pow((float) r+1, dim) > entries); + assert((int) floor(pow((float) r, dim)) <= entries); // (int),floor() as above + return r; +} + +// called twice per file +static void compute_twiddle_factors(int n, float *A, float *B, float *C) +{ + int n4 = n >> 2, n8 = n >> 3; + int k,k2; + + for (k=k2=0; k < n4; ++k,k2+=2) { + A[k2 ] = (float) cos(4*k*M_PI/n); + A[k2+1] = (float) -sin(4*k*M_PI/n); + B[k2 ] = (float) cos((k2+1)*M_PI/n/2) * 0.5f; + B[k2+1] = (float) sin((k2+1)*M_PI/n/2) * 0.5f; + } + for (k=k2=0; k < n8; ++k,k2+=2) { + C[k2 ] = (float) cos(2*(k2+1)*M_PI/n); + C[k2+1] = (float) -sin(2*(k2+1)*M_PI/n); + } +} + +static void compute_window(int n, float *window) +{ + int n2 = n >> 1, i; + for (i=0; i < n2; ++i) + window[i] = (float) sin(0.5 * M_PI * square((float) sin((i - 0 + 0.5) / n2 * 0.5 * M_PI))); +} + +static void compute_bitreverse(int n, uint16 *rev) +{ + int ld = ilog(n) - 1; // ilog is off-by-one from normal definitions + int i, n8 = n >> 3; + for (i=0; i < n8; ++i) + rev[i] = (bit_reverse(i) >> (32-ld+3)) << 2; +} + +static int init_blocksize(vorb *f, int b, int n) +{ + int n2 = n >> 1, n4 = n >> 2, n8 = n >> 3; + f->A[b] = (float *) setup_malloc(f, sizeof(float) * n2); + f->B[b] = (float *) setup_malloc(f, sizeof(float) * n2); + f->C[b] = (float *) setup_malloc(f, sizeof(float) * n4); + if (!f->A[b] || !f->B[b] || !f->C[b]) return error(f, VORBIS_outofmem); + compute_twiddle_factors(n, f->A[b], f->B[b], f->C[b]); + f->window[b] = (float *) setup_malloc(f, sizeof(float) * n2); + if (!f->window[b]) return error(f, VORBIS_outofmem); + compute_window(n, f->window[b]); + f->bit_reverse[b] = (uint16 *) setup_malloc(f, sizeof(uint16) * n8); + if (!f->bit_reverse[b]) return error(f, VORBIS_outofmem); + compute_bitreverse(n, f->bit_reverse[b]); + return TRUE; +} + +static void neighbors(uint16 *x, int n, int *plow, int *phigh) +{ + int low = -1; + int high = 65536; + int i; + for (i=0; i < n; ++i) { + if (x[i] > low && x[i] < x[n]) { *plow = i; low = x[i]; } + if (x[i] < high && x[i] > x[n]) { *phigh = i; high = x[i]; } + } +} + +// this has been repurposed so y is now the original index instead of y +typedef struct +{ + uint16 x,y; +} Point; + +static int STBV_CDECL point_compare(const void *p, const void *q) +{ + Point *a = (Point *) p; + Point *b = (Point *) q; + return a->x < b->x ? -1 : a->x > b->x; +} + +// +/////////////////////// END LEAF SETUP FUNCTIONS ////////////////////////// + + +#if defined(STB_VORBIS_NO_STDIO) + #define USE_MEMORY(z) TRUE +#else + #define USE_MEMORY(z) ((z)->stream) +#endif + +static uint8 get8(vorb *z) +{ + if (USE_MEMORY(z)) { + if (z->stream >= z->stream_end) { z->eof = TRUE; return 0; } + return *z->stream++; + } + + #ifndef STB_VORBIS_NO_STDIO + { + int c = fgetc(z->f); + if (c == EOF) { z->eof = TRUE; return 0; } + return c; + } + #endif +} + +static uint32 get32(vorb *f) +{ + uint32 x; + x = get8(f); + x += get8(f) << 8; + x += get8(f) << 16; + x += (uint32) get8(f) << 24; + return x; +} + +static int getn(vorb *z, uint8 *data, int n) +{ + if (USE_MEMORY(z)) { + if (z->stream+n > z->stream_end) { z->eof = 1; return 0; } + memcpy(data, z->stream, n); + z->stream += n; + return 1; + } + + #ifndef STB_VORBIS_NO_STDIO + if (fread(data, n, 1, z->f) == 1) + return 1; + else { + z->eof = 1; + return 0; + } + #endif +} + +static void skip(vorb *z, int n) +{ + if (USE_MEMORY(z)) { + z->stream += n; + if (z->stream >= z->stream_end) z->eof = 1; + return; + } + #ifndef STB_VORBIS_NO_STDIO + { + long x = ftell(z->f); + fseek(z->f, x+n, SEEK_SET); + } + #endif +} + +static int set_file_offset(stb_vorbis *f, unsigned int loc) +{ + #ifndef STB_VORBIS_NO_PUSHDATA_API + if (f->push_mode) return 0; + #endif + f->eof = 0; + if (USE_MEMORY(f)) { + if (f->stream_start + loc >= f->stream_end || f->stream_start + loc < f->stream_start) { + f->stream = f->stream_end; + f->eof = 1; + return 0; + } else { + f->stream = f->stream_start + loc; + return 1; + } + } + #ifndef STB_VORBIS_NO_STDIO + if (loc + f->f_start < loc || loc >= 0x80000000) { + loc = 0x7fffffff; + f->eof = 1; + } else { + loc += f->f_start; + } + if (!fseek(f->f, loc, SEEK_SET)) + return 1; + f->eof = 1; + fseek(f->f, f->f_start, SEEK_END); + return 0; + #endif +} + + +static uint8 ogg_page_header[4] = { 0x4f, 0x67, 0x67, 0x53 }; + +static int capture_pattern(vorb *f) +{ + if (0x4f != get8(f)) return FALSE; + if (0x67 != get8(f)) return FALSE; + if (0x67 != get8(f)) return FALSE; + if (0x53 != get8(f)) return FALSE; + return TRUE; +} + +#define PAGEFLAG_continued_packet 1 +#define PAGEFLAG_first_page 2 +#define PAGEFLAG_last_page 4 + +static int start_page_no_capturepattern(vorb *f) +{ + uint32 loc0,loc1,n; + // stream structure version + if (0 != get8(f)) return error(f, VORBIS_invalid_stream_structure_version); + // header flag + f->page_flag = get8(f); + // absolute granule position + loc0 = get32(f); + loc1 = get32(f); + // @TODO: validate loc0,loc1 as valid positions? + // stream serial number -- vorbis doesn't interleave, so discard + get32(f); + //if (f->serial != get32(f)) return error(f, VORBIS_incorrect_stream_serial_number); + // page sequence number + n = get32(f); + f->last_page = n; + // CRC32 + get32(f); + // page_segments + f->segment_count = get8(f); + if (!getn(f, f->segments, f->segment_count)) + return error(f, VORBIS_unexpected_eof); + // assume we _don't_ know any the sample position of any segments + f->end_seg_with_known_loc = -2; + if (loc0 != ~0U || loc1 != ~0U) { + int i; + // determine which packet is the last one that will complete + for (i=f->segment_count-1; i >= 0; --i) + if (f->segments[i] < 255) + break; + // 'i' is now the index of the _last_ segment of a packet that ends + if (i >= 0) { + f->end_seg_with_known_loc = i; + f->known_loc_for_packet = loc0; + } + } + if (f->first_decode) { + int i,len; + ProbedPage p; + len = 0; + for (i=0; i < f->segment_count; ++i) + len += f->segments[i]; + len += 27 + f->segment_count; + p.page_start = f->first_audio_page_offset; + p.page_end = p.page_start + len; + p.last_decoded_sample = loc0; + f->p_first = p; + } + f->next_seg = 0; + return TRUE; +} + +static int start_page(vorb *f) +{ + if (!capture_pattern(f)) return error(f, VORBIS_missing_capture_pattern); + return start_page_no_capturepattern(f); +} + +static int start_packet(vorb *f) +{ + while (f->next_seg == -1) { + if (!start_page(f)) return FALSE; + if (f->page_flag & PAGEFLAG_continued_packet) + return error(f, VORBIS_continued_packet_flag_invalid); + } + f->last_seg = FALSE; + f->valid_bits = 0; + f->packet_bytes = 0; + f->bytes_in_seg = 0; + // f->next_seg is now valid + return TRUE; +} + +static int maybe_start_packet(vorb *f) +{ + if (f->next_seg == -1) { + int x = get8(f); + if (f->eof) return FALSE; // EOF at page boundary is not an error! + if (0x4f != x ) return error(f, VORBIS_missing_capture_pattern); + if (0x67 != get8(f)) return error(f, VORBIS_missing_capture_pattern); + if (0x67 != get8(f)) return error(f, VORBIS_missing_capture_pattern); + if (0x53 != get8(f)) return error(f, VORBIS_missing_capture_pattern); + if (!start_page_no_capturepattern(f)) return FALSE; + if (f->page_flag & PAGEFLAG_continued_packet) { + // set up enough state that we can read this packet if we want, + // e.g. during recovery + f->last_seg = FALSE; + f->bytes_in_seg = 0; + return error(f, VORBIS_continued_packet_flag_invalid); + } + } + return start_packet(f); +} + +static int next_segment(vorb *f) +{ + int len; + if (f->last_seg) return 0; + if (f->next_seg == -1) { + f->last_seg_which = f->segment_count-1; // in case start_page fails + if (!start_page(f)) { f->last_seg = 1; return 0; } + if (!(f->page_flag & PAGEFLAG_continued_packet)) return error(f, VORBIS_continued_packet_flag_invalid); + } + len = f->segments[f->next_seg++]; + if (len < 255) { + f->last_seg = TRUE; + f->last_seg_which = f->next_seg-1; + } + if (f->next_seg >= f->segment_count) + f->next_seg = -1; + assert(f->bytes_in_seg == 0); + f->bytes_in_seg = len; + return len; +} + +#define EOP (-1) +#define INVALID_BITS (-1) + +static int get8_packet_raw(vorb *f) +{ + if (!f->bytes_in_seg) { // CLANG! + if (f->last_seg) return EOP; + else if (!next_segment(f)) return EOP; + } + assert(f->bytes_in_seg > 0); + --f->bytes_in_seg; + ++f->packet_bytes; + return get8(f); +} + +static int get8_packet(vorb *f) +{ + int x = get8_packet_raw(f); + f->valid_bits = 0; + return x; +} + +static void flush_packet(vorb *f) +{ + while (get8_packet_raw(f) != EOP); +} + +// @OPTIMIZE: this is the secondary bit decoder, so it's probably not as important +// as the huffman decoder? +static uint32 get_bits(vorb *f, int n) +{ + uint32 z; + + if (f->valid_bits < 0) return 0; + if (f->valid_bits < n) { + if (n > 24) { + // the accumulator technique below would not work correctly in this case + z = get_bits(f, 24); + z += get_bits(f, n-24) << 24; + return z; + } + if (f->valid_bits == 0) f->acc = 0; + while (f->valid_bits < n) { + int z = get8_packet_raw(f); + if (z == EOP) { + f->valid_bits = INVALID_BITS; + return 0; + } + f->acc += z << f->valid_bits; + f->valid_bits += 8; + } + } + if (f->valid_bits < 0) return 0; + z = f->acc & ((1 << n)-1); + f->acc >>= n; + f->valid_bits -= n; + return z; +} + +// @OPTIMIZE: primary accumulator for huffman +// expand the buffer to as many bits as possible without reading off end of packet +// it might be nice to allow f->valid_bits and f->acc to be stored in registers, +// e.g. cache them locally and decode locally +static __forceinline void prep_huffman(vorb *f) +{ + if (f->valid_bits <= 24) { + if (f->valid_bits == 0) f->acc = 0; + do { + int z; + if (f->last_seg && !f->bytes_in_seg) return; + z = get8_packet_raw(f); + if (z == EOP) return; + f->acc += (unsigned) z << f->valid_bits; + f->valid_bits += 8; + } while (f->valid_bits <= 24); + } +} + +enum +{ + VORBIS_packet_id = 1, + VORBIS_packet_comment = 3, + VORBIS_packet_setup = 5 +}; + +static int codebook_decode_scalar_raw(vorb *f, Codebook *c) +{ + int i; + prep_huffman(f); + + if (c->codewords == NULL && c->sorted_codewords == NULL) + return -1; + + // cases to use binary search: sorted_codewords && !c->codewords + // sorted_codewords && c->entries > 8 + if (c->entries > 8 ? c->sorted_codewords!=NULL : !c->codewords) { + // binary search + uint32 code = bit_reverse(f->acc); + int x=0, n=c->sorted_entries, len; + + while (n > 1) { + // invariant: sc[x] <= code < sc[x+n] + int m = x + (n >> 1); + if (c->sorted_codewords[m] <= code) { + x = m; + n -= (n>>1); + } else { + n >>= 1; + } + } + // x is now the sorted index + if (!c->sparse) x = c->sorted_values[x]; + // x is now sorted index if sparse, or symbol otherwise + len = c->codeword_lengths[x]; + if (f->valid_bits >= len) { + f->acc >>= len; + f->valid_bits -= len; + return x; + } + + f->valid_bits = 0; + return -1; + } + + // if small, linear search + assert(!c->sparse); + for (i=0; i < c->entries; ++i) { + if (c->codeword_lengths[i] == NO_CODE) continue; + if (c->codewords[i] == (f->acc & ((1 << c->codeword_lengths[i])-1))) { + if (f->valid_bits >= c->codeword_lengths[i]) { + f->acc >>= c->codeword_lengths[i]; + f->valid_bits -= c->codeword_lengths[i]; + return i; + } + f->valid_bits = 0; + return -1; + } + } + + error(f, VORBIS_invalid_stream); + f->valid_bits = 0; + return -1; +} + +#ifndef STB_VORBIS_NO_INLINE_DECODE + +#define DECODE_RAW(var, f,c) \ + if (f->valid_bits < STB_VORBIS_FAST_HUFFMAN_LENGTH) \ + prep_huffman(f); \ + var = f->acc & FAST_HUFFMAN_TABLE_MASK; \ + var = c->fast_huffman[var]; \ + if (var >= 0) { \ + int n = c->codeword_lengths[var]; \ + f->acc >>= n; \ + f->valid_bits -= n; \ + if (f->valid_bits < 0) { f->valid_bits = 0; var = -1; } \ + } else { \ + var = codebook_decode_scalar_raw(f,c); \ + } + +#else + +static int codebook_decode_scalar(vorb *f, Codebook *c) +{ + int i; + if (f->valid_bits < STB_VORBIS_FAST_HUFFMAN_LENGTH) + prep_huffman(f); + // fast huffman table lookup + i = f->acc & FAST_HUFFMAN_TABLE_MASK; + i = c->fast_huffman[i]; + if (i >= 0) { + f->acc >>= c->codeword_lengths[i]; + f->valid_bits -= c->codeword_lengths[i]; + if (f->valid_bits < 0) { f->valid_bits = 0; return -1; } + return i; + } + return codebook_decode_scalar_raw(f,c); +} + +#define DECODE_RAW(var,f,c) var = codebook_decode_scalar(f,c); + +#endif + +#define DECODE(var,f,c) \ + DECODE_RAW(var,f,c) \ + if (c->sparse) var = c->sorted_values[var]; + +#ifndef STB_VORBIS_DIVIDES_IN_CODEBOOK + #define DECODE_VQ(var,f,c) DECODE_RAW(var,f,c) +#else + #define DECODE_VQ(var,f,c) DECODE(var,f,c) +#endif + + + + + + +// CODEBOOK_ELEMENT_FAST is an optimization for the CODEBOOK_FLOATS case +// where we avoid one addition +#define CODEBOOK_ELEMENT(c,off) (c->multiplicands[off]) +#define CODEBOOK_ELEMENT_FAST(c,off) (c->multiplicands[off]) +#define CODEBOOK_ELEMENT_BASE(c) (0) + +static int codebook_decode_start(vorb *f, Codebook *c) +{ + int z = -1; + + // type 0 is only legal in a scalar context + if (c->lookup_type == 0) + error(f, VORBIS_invalid_stream); + else { + DECODE_VQ(z,f,c); + if (c->sparse) assert(z < c->sorted_entries); + if (z < 0) { // check for EOP + if (!f->bytes_in_seg) + if (f->last_seg) + return z; + error(f, VORBIS_invalid_stream); + } + } + return z; +} + +static int codebook_decode(vorb *f, Codebook *c, float *output, int len) +{ + int i,z = codebook_decode_start(f,c); + if (z < 0) return FALSE; + if (len > c->dimensions) len = c->dimensions; + +#ifdef STB_VORBIS_DIVIDES_IN_CODEBOOK + if (c->lookup_type == 1) { + float last = CODEBOOK_ELEMENT_BASE(c); + int div = 1; + for (i=0; i < len; ++i) { + int off = (z / div) % c->lookup_values; + float val = CODEBOOK_ELEMENT_FAST(c,off) + last; + output[i] += val; + if (c->sequence_p) last = val + c->minimum_value; + div *= c->lookup_values; + } + return TRUE; + } +#endif + + z *= c->dimensions; + if (c->sequence_p) { + float last = CODEBOOK_ELEMENT_BASE(c); + for (i=0; i < len; ++i) { + float val = CODEBOOK_ELEMENT_FAST(c,z+i) + last; + output[i] += val; + last = val + c->minimum_value; + } + } else { + float last = CODEBOOK_ELEMENT_BASE(c); + for (i=0; i < len; ++i) { + output[i] += CODEBOOK_ELEMENT_FAST(c,z+i) + last; + } + } + + return TRUE; +} + +static int codebook_decode_step(vorb *f, Codebook *c, float *output, int len, int step) +{ + int i,z = codebook_decode_start(f,c); + float last = CODEBOOK_ELEMENT_BASE(c); + if (z < 0) return FALSE; + if (len > c->dimensions) len = c->dimensions; + +#ifdef STB_VORBIS_DIVIDES_IN_CODEBOOK + if (c->lookup_type == 1) { + int div = 1; + for (i=0; i < len; ++i) { + int off = (z / div) % c->lookup_values; + float val = CODEBOOK_ELEMENT_FAST(c,off) + last; + output[i*step] += val; + if (c->sequence_p) last = val; + div *= c->lookup_values; + } + return TRUE; + } +#endif + + z *= c->dimensions; + for (i=0; i < len; ++i) { + float val = CODEBOOK_ELEMENT_FAST(c,z+i) + last; + output[i*step] += val; + if (c->sequence_p) last = val; + } + + return TRUE; +} + +static int codebook_decode_deinterleave_repeat(vorb *f, Codebook *c, float **outputs, int ch, int *c_inter_p, int *p_inter_p, int len, int total_decode) +{ + int c_inter = *c_inter_p; + int p_inter = *p_inter_p; + int i,z, effective = c->dimensions; + + // type 0 is only legal in a scalar context + if (c->lookup_type == 0) return error(f, VORBIS_invalid_stream); + + while (total_decode > 0) { + float last = CODEBOOK_ELEMENT_BASE(c); + DECODE_VQ(z,f,c); + #ifndef STB_VORBIS_DIVIDES_IN_CODEBOOK + assert(!c->sparse || z < c->sorted_entries); + #endif + if (z < 0) { + if (!f->bytes_in_seg) + if (f->last_seg) return FALSE; + return error(f, VORBIS_invalid_stream); + } + + // if this will take us off the end of the buffers, stop short! + // we check by computing the length of the virtual interleaved + // buffer (len*ch), our current offset within it (p_inter*ch)+(c_inter), + // and the length we'll be using (effective) + if (c_inter + p_inter*ch + effective > len * ch) { + effective = len*ch - (p_inter*ch - c_inter); + } + + #ifdef STB_VORBIS_DIVIDES_IN_CODEBOOK + if (c->lookup_type == 1) { + int div = 1; + for (i=0; i < effective; ++i) { + int off = (z / div) % c->lookup_values; + float val = CODEBOOK_ELEMENT_FAST(c,off) + last; + if (outputs[c_inter]) + outputs[c_inter][p_inter] += val; + if (++c_inter == ch) { c_inter = 0; ++p_inter; } + if (c->sequence_p) last = val; + div *= c->lookup_values; + } + } else + #endif + { + z *= c->dimensions; + if (c->sequence_p) { + for (i=0; i < effective; ++i) { + float val = CODEBOOK_ELEMENT_FAST(c,z+i) + last; + if (outputs[c_inter]) + outputs[c_inter][p_inter] += val; + if (++c_inter == ch) { c_inter = 0; ++p_inter; } + last = val; + } + } else { + for (i=0; i < effective; ++i) { + float val = CODEBOOK_ELEMENT_FAST(c,z+i) + last; + if (outputs[c_inter]) + outputs[c_inter][p_inter] += val; + if (++c_inter == ch) { c_inter = 0; ++p_inter; } + } + } + } + + total_decode -= effective; + } + *c_inter_p = c_inter; + *p_inter_p = p_inter; + return TRUE; +} + +#ifndef STB_VORBIS_DIVIDES_IN_CODEBOOK +static int codebook_decode_deinterleave_repeat_2(vorb *f, Codebook *c, float **outputs, int *c_inter_p, int *p_inter_p, int len, int total_decode) +{ + int c_inter = *c_inter_p; + int p_inter = *p_inter_p; + int i,z, effective = c->dimensions; + + // type 0 is only legal in a scalar context + if (c->lookup_type == 0) return error(f, VORBIS_invalid_stream); + + while (total_decode > 0) { + float last = CODEBOOK_ELEMENT_BASE(c); + DECODE_VQ(z,f,c); + + if (z < 0) { + if (!f->bytes_in_seg) + if (f->last_seg) return FALSE; + return error(f, VORBIS_invalid_stream); + } + + // if this will take us off the end of the buffers, stop short! + // we check by computing the length of the virtual interleaved + // buffer (len*ch), our current offset within it (p_inter*ch)+(c_inter), + // and the length we'll be using (effective) + if (c_inter + p_inter*2 + effective > len * 2) { + effective = len*2 - (p_inter*2 - c_inter); + } + + { + z *= c->dimensions; + if (c->sequence_p) { + // haven't optimized this case because I don't have any examples + for (i=0; i < effective; ++i) { + float val = CODEBOOK_ELEMENT_FAST(c,z+i) + last; + if (outputs[c_inter]) + outputs[c_inter][p_inter] += val; + if (++c_inter == 2) { c_inter = 0; ++p_inter; } + last = val; + } + } else { + i=0; + if (c_inter == 1 && i < effective) { + float val = CODEBOOK_ELEMENT_FAST(c,z+i) + last; + if (outputs[c_inter]) + outputs[c_inter][p_inter] += val; + c_inter = 0; ++p_inter; + ++i; + } + { + float *z0 = outputs[0]; + float *z1 = outputs[1]; + for (; i+1 < effective;) { + float v0 = CODEBOOK_ELEMENT_FAST(c,z+i) + last; + float v1 = CODEBOOK_ELEMENT_FAST(c,z+i+1) + last; + if (z0) + z0[p_inter] += v0; + if (z1) + z1[p_inter] += v1; + ++p_inter; + i += 2; + } + } + if (i < effective) { + float val = CODEBOOK_ELEMENT_FAST(c,z+i) + last; + if (outputs[c_inter]) + outputs[c_inter][p_inter] += val; + if (++c_inter == 2) { c_inter = 0; ++p_inter; } + } + } + } + + total_decode -= effective; + } + *c_inter_p = c_inter; + *p_inter_p = p_inter; + return TRUE; +} +#endif + +static int predict_point(int x, int x0, int x1, int y0, int y1) +{ + int dy = y1 - y0; + int adx = x1 - x0; + // @OPTIMIZE: force int division to round in the right direction... is this necessary on x86? + int err = abs(dy) * (x - x0); + int off = err / adx; + return dy < 0 ? y0 - off : y0 + off; +} + +// the following table is block-copied from the specification +static float inverse_db_table[256] = +{ + 1.0649863e-07f, 1.1341951e-07f, 1.2079015e-07f, 1.2863978e-07f, + 1.3699951e-07f, 1.4590251e-07f, 1.5538408e-07f, 1.6548181e-07f, + 1.7623575e-07f, 1.8768855e-07f, 1.9988561e-07f, 2.1287530e-07f, + 2.2670913e-07f, 2.4144197e-07f, 2.5713223e-07f, 2.7384213e-07f, + 2.9163793e-07f, 3.1059021e-07f, 3.3077411e-07f, 3.5226968e-07f, + 3.7516214e-07f, 3.9954229e-07f, 4.2550680e-07f, 4.5315863e-07f, + 4.8260743e-07f, 5.1396998e-07f, 5.4737065e-07f, 5.8294187e-07f, + 6.2082472e-07f, 6.6116941e-07f, 7.0413592e-07f, 7.4989464e-07f, + 7.9862701e-07f, 8.5052630e-07f, 9.0579828e-07f, 9.6466216e-07f, + 1.0273513e-06f, 1.0941144e-06f, 1.1652161e-06f, 1.2409384e-06f, + 1.3215816e-06f, 1.4074654e-06f, 1.4989305e-06f, 1.5963394e-06f, + 1.7000785e-06f, 1.8105592e-06f, 1.9282195e-06f, 2.0535261e-06f, + 2.1869758e-06f, 2.3290978e-06f, 2.4804557e-06f, 2.6416497e-06f, + 2.8133190e-06f, 2.9961443e-06f, 3.1908506e-06f, 3.3982101e-06f, + 3.6190449e-06f, 3.8542308e-06f, 4.1047004e-06f, 4.3714470e-06f, + 4.6555282e-06f, 4.9580707e-06f, 5.2802740e-06f, 5.6234160e-06f, + 5.9888572e-06f, 6.3780469e-06f, 6.7925283e-06f, 7.2339451e-06f, + 7.7040476e-06f, 8.2047000e-06f, 8.7378876e-06f, 9.3057248e-06f, + 9.9104632e-06f, 1.0554501e-05f, 1.1240392e-05f, 1.1970856e-05f, + 1.2748789e-05f, 1.3577278e-05f, 1.4459606e-05f, 1.5399272e-05f, + 1.6400004e-05f, 1.7465768e-05f, 1.8600792e-05f, 1.9809576e-05f, + 2.1096914e-05f, 2.2467911e-05f, 2.3928002e-05f, 2.5482978e-05f, + 2.7139006e-05f, 2.8902651e-05f, 3.0780908e-05f, 3.2781225e-05f, + 3.4911534e-05f, 3.7180282e-05f, 3.9596466e-05f, 4.2169667e-05f, + 4.4910090e-05f, 4.7828601e-05f, 5.0936773e-05f, 5.4246931e-05f, + 5.7772202e-05f, 6.1526565e-05f, 6.5524908e-05f, 6.9783085e-05f, + 7.4317983e-05f, 7.9147585e-05f, 8.4291040e-05f, 8.9768747e-05f, + 9.5602426e-05f, 0.00010181521f, 0.00010843174f, 0.00011547824f, + 0.00012298267f, 0.00013097477f, 0.00013948625f, 0.00014855085f, + 0.00015820453f, 0.00016848555f, 0.00017943469f, 0.00019109536f, + 0.00020351382f, 0.00021673929f, 0.00023082423f, 0.00024582449f, + 0.00026179955f, 0.00027881276f, 0.00029693158f, 0.00031622787f, + 0.00033677814f, 0.00035866388f, 0.00038197188f, 0.00040679456f, + 0.00043323036f, 0.00046138411f, 0.00049136745f, 0.00052329927f, + 0.00055730621f, 0.00059352311f, 0.00063209358f, 0.00067317058f, + 0.00071691700f, 0.00076350630f, 0.00081312324f, 0.00086596457f, + 0.00092223983f, 0.00098217216f, 0.0010459992f, 0.0011139742f, + 0.0011863665f, 0.0012634633f, 0.0013455702f, 0.0014330129f, + 0.0015261382f, 0.0016253153f, 0.0017309374f, 0.0018434235f, + 0.0019632195f, 0.0020908006f, 0.0022266726f, 0.0023713743f, + 0.0025254795f, 0.0026895994f, 0.0028643847f, 0.0030505286f, + 0.0032487691f, 0.0034598925f, 0.0036847358f, 0.0039241906f, + 0.0041792066f, 0.0044507950f, 0.0047400328f, 0.0050480668f, + 0.0053761186f, 0.0057254891f, 0.0060975636f, 0.0064938176f, + 0.0069158225f, 0.0073652516f, 0.0078438871f, 0.0083536271f, + 0.0088964928f, 0.009474637f, 0.010090352f, 0.010746080f, + 0.011444421f, 0.012188144f, 0.012980198f, 0.013823725f, + 0.014722068f, 0.015678791f, 0.016697687f, 0.017782797f, + 0.018938423f, 0.020169149f, 0.021479854f, 0.022875735f, + 0.024362330f, 0.025945531f, 0.027631618f, 0.029427276f, + 0.031339626f, 0.033376252f, 0.035545228f, 0.037855157f, + 0.040315199f, 0.042935108f, 0.045725273f, 0.048696758f, + 0.051861348f, 0.055231591f, 0.058820850f, 0.062643361f, + 0.066714279f, 0.071049749f, 0.075666962f, 0.080584227f, + 0.085821044f, 0.091398179f, 0.097337747f, 0.10366330f, + 0.11039993f, 0.11757434f, 0.12521498f, 0.13335215f, + 0.14201813f, 0.15124727f, 0.16107617f, 0.17154380f, + 0.18269168f, 0.19456402f, 0.20720788f, 0.22067342f, + 0.23501402f, 0.25028656f, 0.26655159f, 0.28387361f, + 0.30232132f, 0.32196786f, 0.34289114f, 0.36517414f, + 0.38890521f, 0.41417847f, 0.44109412f, 0.46975890f, + 0.50028648f, 0.53279791f, 0.56742212f, 0.60429640f, + 0.64356699f, 0.68538959f, 0.72993007f, 0.77736504f, + 0.82788260f, 0.88168307f, 0.9389798f, 1.0f +}; + + +// @OPTIMIZE: if you want to replace this bresenham line-drawing routine, +// note that you must produce bit-identical output to decode correctly; +// this specific sequence of operations is specified in the spec (it's +// drawing integer-quantized frequency-space lines that the encoder +// expects to be exactly the same) +// ... also, isn't the whole point of Bresenham's algorithm to NOT +// have to divide in the setup? sigh. +#ifndef STB_VORBIS_NO_DEFER_FLOOR +#define LINE_OP(a,b) a *= b +#else +#define LINE_OP(a,b) a = b +#endif + +#ifdef STB_VORBIS_DIVIDE_TABLE +#define DIVTAB_NUMER 32 +#define DIVTAB_DENOM 64 +int8 integer_divide_table[DIVTAB_NUMER][DIVTAB_DENOM]; // 2KB +#endif + +static __forceinline void draw_line(float *output, int x0, int y0, int x1, int y1, int n) +{ + int dy = y1 - y0; + int adx = x1 - x0; + int ady = abs(dy); + int base; + int x=x0,y=y0; + int err = 0; + int sy; + +#ifdef STB_VORBIS_DIVIDE_TABLE + if (adx < DIVTAB_DENOM && ady < DIVTAB_NUMER) { + if (dy < 0) { + base = -integer_divide_table[ady][adx]; + sy = base-1; + } else { + base = integer_divide_table[ady][adx]; + sy = base+1; + } + } else { + base = dy / adx; + if (dy < 0) + sy = base - 1; + else + sy = base+1; + } +#else + base = dy / adx; + if (dy < 0) + sy = base - 1; + else + sy = base+1; +#endif + ady -= abs(base) * adx; + if (x1 > n) x1 = n; + if (x < x1) { + LINE_OP(output[x], inverse_db_table[y]); + for (++x; x < x1; ++x) { + err += ady; + if (err >= adx) { + err -= adx; + y += sy; + } else + y += base; + LINE_OP(output[x], inverse_db_table[y]); + } + } +} + +static int residue_decode(vorb *f, Codebook *book, float *target, int offset, int n, int rtype) +{ + int k; + if (rtype == 0) { + int step = n / book->dimensions; + for (k=0; k < step; ++k) + if (!codebook_decode_step(f, book, target+offset+k, n-offset-k, step)) + return FALSE; + } else { + for (k=0; k < n; ) { + if (!codebook_decode(f, book, target+offset, n-k)) + return FALSE; + k += book->dimensions; + offset += book->dimensions; + } + } + return TRUE; +} + +static void decode_residue(vorb *f, float *residue_buffers[], int ch, int n, int rn, uint8 *do_not_decode) +{ + int i,j,pass; + Residue *r = f->residue_config + rn; + int rtype = f->residue_types[rn]; + int c = r->classbook; + int classwords = f->codebooks[c].dimensions; + int n_read = r->end - r->begin; + int part_read = n_read / r->part_size; + int temp_alloc_point = temp_alloc_save(f); + #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + uint8 ***part_classdata = (uint8 ***) temp_block_array(f,f->channels, part_read * sizeof(**part_classdata)); + #else + int **classifications = (int **) temp_block_array(f,f->channels, part_read * sizeof(**classifications)); + #endif + + CHECK(f); + + for (i=0; i < ch; ++i) + if (!do_not_decode[i]) + memset(residue_buffers[i], 0, sizeof(float) * n); + + if (rtype == 2 && ch != 1) { + for (j=0; j < ch; ++j) + if (!do_not_decode[j]) + break; + if (j == ch) + goto done; + + for (pass=0; pass < 8; ++pass) { + int pcount = 0, class_set = 0; + if (ch == 2) { + while (pcount < part_read) { + int z = r->begin + pcount*r->part_size; + int c_inter = (z & 1), p_inter = z>>1; + if (pass == 0) { + Codebook *c = f->codebooks+r->classbook; + int q; + DECODE(q,f,c); + if (q == EOP) goto done; + #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + part_classdata[0][class_set] = r->classdata[q]; + #else + for (i=classwords-1; i >= 0; --i) { + classifications[0][i+pcount] = q % r->classifications; + q /= r->classifications; + } + #endif + } + for (i=0; i < classwords && pcount < part_read; ++i, ++pcount) { + int z = r->begin + pcount*r->part_size; + #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + int c = part_classdata[0][class_set][i]; + #else + int c = classifications[0][pcount]; + #endif + int b = r->residue_books[c][pass]; + if (b >= 0) { + Codebook *book = f->codebooks + b; + #ifdef STB_VORBIS_DIVIDES_IN_CODEBOOK + if (!codebook_decode_deinterleave_repeat(f, book, residue_buffers, ch, &c_inter, &p_inter, n, r->part_size)) + goto done; + #else + // saves 1% + if (!codebook_decode_deinterleave_repeat(f, book, residue_buffers, ch, &c_inter, &p_inter, n, r->part_size)) + goto done; + #endif + } else { + z += r->part_size; + c_inter = z & 1; + p_inter = z >> 1; + } + } + #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + ++class_set; + #endif + } + } else if (ch == 1) { + while (pcount < part_read) { + int z = r->begin + pcount*r->part_size; + int c_inter = 0, p_inter = z; + if (pass == 0) { + Codebook *c = f->codebooks+r->classbook; + int q; + DECODE(q,f,c); + if (q == EOP) goto done; + #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + part_classdata[0][class_set] = r->classdata[q]; + #else + for (i=classwords-1; i >= 0; --i) { + classifications[0][i+pcount] = q % r->classifications; + q /= r->classifications; + } + #endif + } + for (i=0; i < classwords && pcount < part_read; ++i, ++pcount) { + int z = r->begin + pcount*r->part_size; + #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + int c = part_classdata[0][class_set][i]; + #else + int c = classifications[0][pcount]; + #endif + int b = r->residue_books[c][pass]; + if (b >= 0) { + Codebook *book = f->codebooks + b; + if (!codebook_decode_deinterleave_repeat(f, book, residue_buffers, ch, &c_inter, &p_inter, n, r->part_size)) + goto done; + } else { + z += r->part_size; + c_inter = 0; + p_inter = z; + } + } + #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + ++class_set; + #endif + } + } else { + while (pcount < part_read) { + int z = r->begin + pcount*r->part_size; + int c_inter = z % ch, p_inter = z/ch; + if (pass == 0) { + Codebook *c = f->codebooks+r->classbook; + int q; + DECODE(q,f,c); + if (q == EOP) goto done; + #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + part_classdata[0][class_set] = r->classdata[q]; + #else + for (i=classwords-1; i >= 0; --i) { + classifications[0][i+pcount] = q % r->classifications; + q /= r->classifications; + } + #endif + } + for (i=0; i < classwords && pcount < part_read; ++i, ++pcount) { + int z = r->begin + pcount*r->part_size; + #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + int c = part_classdata[0][class_set][i]; + #else + int c = classifications[0][pcount]; + #endif + int b = r->residue_books[c][pass]; + if (b >= 0) { + Codebook *book = f->codebooks + b; + if (!codebook_decode_deinterleave_repeat(f, book, residue_buffers, ch, &c_inter, &p_inter, n, r->part_size)) + goto done; + } else { + z += r->part_size; + c_inter = z % ch; + p_inter = z / ch; + } + } + #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + ++class_set; + #endif + } + } + } + goto done; + } + CHECK(f); + + for (pass=0; pass < 8; ++pass) { + int pcount = 0, class_set=0; + while (pcount < part_read) { + if (pass == 0) { + for (j=0; j < ch; ++j) { + if (!do_not_decode[j]) { + Codebook *c = f->codebooks+r->classbook; + int temp; + DECODE(temp,f,c); + if (temp == EOP) goto done; + #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + part_classdata[j][class_set] = r->classdata[temp]; + #else + for (i=classwords-1; i >= 0; --i) { + classifications[j][i+pcount] = temp % r->classifications; + temp /= r->classifications; + } + #endif + } + } + } + for (i=0; i < classwords && pcount < part_read; ++i, ++pcount) { + for (j=0; j < ch; ++j) { + if (!do_not_decode[j]) { + #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + int c = part_classdata[j][class_set][i]; + #else + int c = classifications[j][pcount]; + #endif + int b = r->residue_books[c][pass]; + if (b >= 0) { + float *target = residue_buffers[j]; + int offset = r->begin + pcount * r->part_size; + int n = r->part_size; + Codebook *book = f->codebooks + b; + if (!residue_decode(f, book, target, offset, n, rtype)) + goto done; + } + } + } + } + #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + ++class_set; + #endif + } + } + done: + CHECK(f); + temp_alloc_restore(f,temp_alloc_point); +} + + +#if 0 +// slow way for debugging +void inverse_mdct_slow(float *buffer, int n) +{ + int i,j; + int n2 = n >> 1; + float *x = (float *) malloc(sizeof(*x) * n2); + memcpy(x, buffer, sizeof(*x) * n2); + for (i=0; i < n; ++i) { + float acc = 0; + for (j=0; j < n2; ++j) + // formula from paper: + //acc += n/4.0f * x[j] * (float) cos(M_PI / 2 / n * (2 * i + 1 + n/2.0)*(2*j+1)); + // formula from wikipedia + //acc += 2.0f / n2 * x[j] * (float) cos(M_PI/n2 * (i + 0.5 + n2/2)*(j + 0.5)); + // these are equivalent, except the formula from the paper inverts the multiplier! + // however, what actually works is NO MULTIPLIER!?! + //acc += 64 * 2.0f / n2 * x[j] * (float) cos(M_PI/n2 * (i + 0.5 + n2/2)*(j + 0.5)); + acc += x[j] * (float) cos(M_PI / 2 / n * (2 * i + 1 + n/2.0)*(2*j+1)); + buffer[i] = acc; + } + free(x); +} +#elif 0 +// same as above, but just barely able to run in real time on modern machines +void inverse_mdct_slow(float *buffer, int n, vorb *f, int blocktype) +{ + float mcos[16384]; + int i,j; + int n2 = n >> 1, nmask = (n << 2) -1; + float *x = (float *) malloc(sizeof(*x) * n2); + memcpy(x, buffer, sizeof(*x) * n2); + for (i=0; i < 4*n; ++i) + mcos[i] = (float) cos(M_PI / 2 * i / n); + + for (i=0; i < n; ++i) { + float acc = 0; + for (j=0; j < n2; ++j) + acc += x[j] * mcos[(2 * i + 1 + n2)*(2*j+1) & nmask]; + buffer[i] = acc; + } + free(x); +} +#elif 0 +// transform to use a slow dct-iv; this is STILL basically trivial, +// but only requires half as many ops +void dct_iv_slow(float *buffer, int n) +{ + float mcos[16384]; + float x[2048]; + int i,j; + int n2 = n >> 1, nmask = (n << 3) - 1; + memcpy(x, buffer, sizeof(*x) * n); + for (i=0; i < 8*n; ++i) + mcos[i] = (float) cos(M_PI / 4 * i / n); + for (i=0; i < n; ++i) { + float acc = 0; + for (j=0; j < n; ++j) + acc += x[j] * mcos[((2 * i + 1)*(2*j+1)) & nmask]; + buffer[i] = acc; + } +} + +void inverse_mdct_slow(float *buffer, int n, vorb *f, int blocktype) +{ + int i, n4 = n >> 2, n2 = n >> 1, n3_4 = n - n4; + float temp[4096]; + + memcpy(temp, buffer, n2 * sizeof(float)); + dct_iv_slow(temp, n2); // returns -c'-d, a-b' + + for (i=0; i < n4 ; ++i) buffer[i] = temp[i+n4]; // a-b' + for ( ; i < n3_4; ++i) buffer[i] = -temp[n3_4 - i - 1]; // b-a', c+d' + for ( ; i < n ; ++i) buffer[i] = -temp[i - n3_4]; // c'+d +} +#endif + +#ifndef LIBVORBIS_MDCT +#define LIBVORBIS_MDCT 0 +#endif + +#if LIBVORBIS_MDCT +// directly call the vorbis MDCT using an interface documented +// by Jeff Roberts... useful for performance comparison +typedef struct +{ + int n; + int log2n; + + float *trig; + int *bitrev; + + float scale; +} mdct_lookup; + +extern void mdct_init(mdct_lookup *lookup, int n); +extern void mdct_clear(mdct_lookup *l); +extern void mdct_backward(mdct_lookup *init, float *in, float *out); + +mdct_lookup M1,M2; + +void inverse_mdct(float *buffer, int n, vorb *f, int blocktype) +{ + mdct_lookup *M; + if (M1.n == n) M = &M1; + else if (M2.n == n) M = &M2; + else if (M1.n == 0) { mdct_init(&M1, n); M = &M1; } + else { + if (M2.n) __asm int 3; + mdct_init(&M2, n); + M = &M2; + } + + mdct_backward(M, buffer, buffer); +} +#endif + + +// the following were split out into separate functions while optimizing; +// they could be pushed back up but eh. __forceinline showed no change; +// they're probably already being inlined. +static void imdct_step3_iter0_loop(int n, float *e, int i_off, int k_off, float *A) +{ + float *ee0 = e + i_off; + float *ee2 = ee0 + k_off; + int i; + + assert((n & 3) == 0); + for (i=(n>>2); i > 0; --i) { + float k00_20, k01_21; + k00_20 = ee0[ 0] - ee2[ 0]; + k01_21 = ee0[-1] - ee2[-1]; + ee0[ 0] += ee2[ 0];//ee0[ 0] = ee0[ 0] + ee2[ 0]; + ee0[-1] += ee2[-1];//ee0[-1] = ee0[-1] + ee2[-1]; + ee2[ 0] = k00_20 * A[0] - k01_21 * A[1]; + ee2[-1] = k01_21 * A[0] + k00_20 * A[1]; + A += 8; + + k00_20 = ee0[-2] - ee2[-2]; + k01_21 = ee0[-3] - ee2[-3]; + ee0[-2] += ee2[-2];//ee0[-2] = ee0[-2] + ee2[-2]; + ee0[-3] += ee2[-3];//ee0[-3] = ee0[-3] + ee2[-3]; + ee2[-2] = k00_20 * A[0] - k01_21 * A[1]; + ee2[-3] = k01_21 * A[0] + k00_20 * A[1]; + A += 8; + + k00_20 = ee0[-4] - ee2[-4]; + k01_21 = ee0[-5] - ee2[-5]; + ee0[-4] += ee2[-4];//ee0[-4] = ee0[-4] + ee2[-4]; + ee0[-5] += ee2[-5];//ee0[-5] = ee0[-5] + ee2[-5]; + ee2[-4] = k00_20 * A[0] - k01_21 * A[1]; + ee2[-5] = k01_21 * A[0] + k00_20 * A[1]; + A += 8; + + k00_20 = ee0[-6] - ee2[-6]; + k01_21 = ee0[-7] - ee2[-7]; + ee0[-6] += ee2[-6];//ee0[-6] = ee0[-6] + ee2[-6]; + ee0[-7] += ee2[-7];//ee0[-7] = ee0[-7] + ee2[-7]; + ee2[-6] = k00_20 * A[0] - k01_21 * A[1]; + ee2[-7] = k01_21 * A[0] + k00_20 * A[1]; + A += 8; + ee0 -= 8; + ee2 -= 8; + } +} + +static void imdct_step3_inner_r_loop(int lim, float *e, int d0, int k_off, float *A, int k1) +{ + int i; + float k00_20, k01_21; + + float *e0 = e + d0; + float *e2 = e0 + k_off; + + for (i=lim >> 2; i > 0; --i) { + k00_20 = e0[-0] - e2[-0]; + k01_21 = e0[-1] - e2[-1]; + e0[-0] += e2[-0];//e0[-0] = e0[-0] + e2[-0]; + e0[-1] += e2[-1];//e0[-1] = e0[-1] + e2[-1]; + e2[-0] = (k00_20)*A[0] - (k01_21) * A[1]; + e2[-1] = (k01_21)*A[0] + (k00_20) * A[1]; + + A += k1; + + k00_20 = e0[-2] - e2[-2]; + k01_21 = e0[-3] - e2[-3]; + e0[-2] += e2[-2];//e0[-2] = e0[-2] + e2[-2]; + e0[-3] += e2[-3];//e0[-3] = e0[-3] + e2[-3]; + e2[-2] = (k00_20)*A[0] - (k01_21) * A[1]; + e2[-3] = (k01_21)*A[0] + (k00_20) * A[1]; + + A += k1; + + k00_20 = e0[-4] - e2[-4]; + k01_21 = e0[-5] - e2[-5]; + e0[-4] += e2[-4];//e0[-4] = e0[-4] + e2[-4]; + e0[-5] += e2[-5];//e0[-5] = e0[-5] + e2[-5]; + e2[-4] = (k00_20)*A[0] - (k01_21) * A[1]; + e2[-5] = (k01_21)*A[0] + (k00_20) * A[1]; + + A += k1; + + k00_20 = e0[-6] - e2[-6]; + k01_21 = e0[-7] - e2[-7]; + e0[-6] += e2[-6];//e0[-6] = e0[-6] + e2[-6]; + e0[-7] += e2[-7];//e0[-7] = e0[-7] + e2[-7]; + e2[-6] = (k00_20)*A[0] - (k01_21) * A[1]; + e2[-7] = (k01_21)*A[0] + (k00_20) * A[1]; + + e0 -= 8; + e2 -= 8; + + A += k1; + } +} + +static void imdct_step3_inner_s_loop(int n, float *e, int i_off, int k_off, float *A, int a_off, int k0) +{ + int i; + float A0 = A[0]; + float A1 = A[0+1]; + float A2 = A[0+a_off]; + float A3 = A[0+a_off+1]; + float A4 = A[0+a_off*2+0]; + float A5 = A[0+a_off*2+1]; + float A6 = A[0+a_off*3+0]; + float A7 = A[0+a_off*3+1]; + + float k00,k11; + + float *ee0 = e +i_off; + float *ee2 = ee0+k_off; + + for (i=n; i > 0; --i) { + k00 = ee0[ 0] - ee2[ 0]; + k11 = ee0[-1] - ee2[-1]; + ee0[ 0] = ee0[ 0] + ee2[ 0]; + ee0[-1] = ee0[-1] + ee2[-1]; + ee2[ 0] = (k00) * A0 - (k11) * A1; + ee2[-1] = (k11) * A0 + (k00) * A1; + + k00 = ee0[-2] - ee2[-2]; + k11 = ee0[-3] - ee2[-3]; + ee0[-2] = ee0[-2] + ee2[-2]; + ee0[-3] = ee0[-3] + ee2[-3]; + ee2[-2] = (k00) * A2 - (k11) * A3; + ee2[-3] = (k11) * A2 + (k00) * A3; + + k00 = ee0[-4] - ee2[-4]; + k11 = ee0[-5] - ee2[-5]; + ee0[-4] = ee0[-4] + ee2[-4]; + ee0[-5] = ee0[-5] + ee2[-5]; + ee2[-4] = (k00) * A4 - (k11) * A5; + ee2[-5] = (k11) * A4 + (k00) * A5; + + k00 = ee0[-6] - ee2[-6]; + k11 = ee0[-7] - ee2[-7]; + ee0[-6] = ee0[-6] + ee2[-6]; + ee0[-7] = ee0[-7] + ee2[-7]; + ee2[-6] = (k00) * A6 - (k11) * A7; + ee2[-7] = (k11) * A6 + (k00) * A7; + + ee0 -= k0; + ee2 -= k0; + } +} + +static __forceinline void iter_54(float *z) +{ + float k00,k11,k22,k33; + float y0,y1,y2,y3; + + k00 = z[ 0] - z[-4]; + y0 = z[ 0] + z[-4]; + y2 = z[-2] + z[-6]; + k22 = z[-2] - z[-6]; + + z[-0] = y0 + y2; // z0 + z4 + z2 + z6 + z[-2] = y0 - y2; // z0 + z4 - z2 - z6 + + // done with y0,y2 + + k33 = z[-3] - z[-7]; + + z[-4] = k00 + k33; // z0 - z4 + z3 - z7 + z[-6] = k00 - k33; // z0 - z4 - z3 + z7 + + // done with k33 + + k11 = z[-1] - z[-5]; + y1 = z[-1] + z[-5]; + y3 = z[-3] + z[-7]; + + z[-1] = y1 + y3; // z1 + z5 + z3 + z7 + z[-3] = y1 - y3; // z1 + z5 - z3 - z7 + z[-5] = k11 - k22; // z1 - z5 + z2 - z6 + z[-7] = k11 + k22; // z1 - z5 - z2 + z6 +} + +static void imdct_step3_inner_s_loop_ld654(int n, float *e, int i_off, float *A, int base_n) +{ + int a_off = base_n >> 3; + float A2 = A[0+a_off]; + float *z = e + i_off; + float *base = z - 16 * n; + + while (z > base) { + float k00,k11; + + k00 = z[-0] - z[-8]; + k11 = z[-1] - z[-9]; + z[-0] = z[-0] + z[-8]; + z[-1] = z[-1] + z[-9]; + z[-8] = k00; + z[-9] = k11 ; + + k00 = z[ -2] - z[-10]; + k11 = z[ -3] - z[-11]; + z[ -2] = z[ -2] + z[-10]; + z[ -3] = z[ -3] + z[-11]; + z[-10] = (k00+k11) * A2; + z[-11] = (k11-k00) * A2; + + k00 = z[-12] - z[ -4]; // reverse to avoid a unary negation + k11 = z[ -5] - z[-13]; + z[ -4] = z[ -4] + z[-12]; + z[ -5] = z[ -5] + z[-13]; + z[-12] = k11; + z[-13] = k00; + + k00 = z[-14] - z[ -6]; // reverse to avoid a unary negation + k11 = z[ -7] - z[-15]; + z[ -6] = z[ -6] + z[-14]; + z[ -7] = z[ -7] + z[-15]; + z[-14] = (k00+k11) * A2; + z[-15] = (k00-k11) * A2; + + iter_54(z); + iter_54(z-8); + z -= 16; + } +} + +static void inverse_mdct(float *buffer, int n, vorb *f, int blocktype) +{ + int n2 = n >> 1, n4 = n >> 2, n8 = n >> 3, l; + int ld; + // @OPTIMIZE: reduce register pressure by using fewer variables? + int save_point = temp_alloc_save(f); + float *buf2 = (float *) temp_alloc(f, n2 * sizeof(*buf2)); + float *u=NULL,*v=NULL; + // twiddle factors + float *A = f->A[blocktype]; + + // IMDCT algorithm from "The use of multirate filter banks for coding of high quality digital audio" + // See notes about bugs in that paper in less-optimal implementation 'inverse_mdct_old' after this function. + + // kernel from paper + + + // merged: + // copy and reflect spectral data + // step 0 + + // note that it turns out that the items added together during + // this step are, in fact, being added to themselves (as reflected + // by step 0). inexplicable inefficiency! this became obvious + // once I combined the passes. + + // so there's a missing 'times 2' here (for adding X to itself). + // this propogates through linearly to the end, where the numbers + // are 1/2 too small, and need to be compensated for. + + { + float *d,*e, *AA, *e_stop; + d = &buf2[n2-2]; + AA = A; + e = &buffer[0]; + e_stop = &buffer[n2]; + while (e != e_stop) { + d[1] = (e[0] * AA[0] - e[2]*AA[1]); + d[0] = (e[0] * AA[1] + e[2]*AA[0]); + d -= 2; + AA += 2; + e += 4; + } + + e = &buffer[n2-3]; + while (d >= buf2) { + d[1] = (-e[2] * AA[0] - -e[0]*AA[1]); + d[0] = (-e[2] * AA[1] + -e[0]*AA[0]); + d -= 2; + AA += 2; + e -= 4; + } + } + + // now we use symbolic names for these, so that we can + // possibly swap their meaning as we change which operations + // are in place + + u = buffer; + v = buf2; + + // step 2 (paper output is w, now u) + // this could be in place, but the data ends up in the wrong + // place... _somebody_'s got to swap it, so this is nominated + { + float *AA = &A[n2-8]; + float *d0,*d1, *e0, *e1; + + e0 = &v[n4]; + e1 = &v[0]; + + d0 = &u[n4]; + d1 = &u[0]; + + while (AA >= A) { + float v40_20, v41_21; + + v41_21 = e0[1] - e1[1]; + v40_20 = e0[0] - e1[0]; + d0[1] = e0[1] + e1[1]; + d0[0] = e0[0] + e1[0]; + d1[1] = v41_21*AA[4] - v40_20*AA[5]; + d1[0] = v40_20*AA[4] + v41_21*AA[5]; + + v41_21 = e0[3] - e1[3]; + v40_20 = e0[2] - e1[2]; + d0[3] = e0[3] + e1[3]; + d0[2] = e0[2] + e1[2]; + d1[3] = v41_21*AA[0] - v40_20*AA[1]; + d1[2] = v40_20*AA[0] + v41_21*AA[1]; + + AA -= 8; + + d0 += 4; + d1 += 4; + e0 += 4; + e1 += 4; + } + } + + // step 3 + ld = ilog(n) - 1; // ilog is off-by-one from normal definitions + + // optimized step 3: + + // the original step3 loop can be nested r inside s or s inside r; + // it's written originally as s inside r, but this is dumb when r + // iterates many times, and s few. So I have two copies of it and + // switch between them halfway. + + // this is iteration 0 of step 3 + imdct_step3_iter0_loop(n >> 4, u, n2-1-n4*0, -(n >> 3), A); + imdct_step3_iter0_loop(n >> 4, u, n2-1-n4*1, -(n >> 3), A); + + // this is iteration 1 of step 3 + imdct_step3_inner_r_loop(n >> 5, u, n2-1 - n8*0, -(n >> 4), A, 16); + imdct_step3_inner_r_loop(n >> 5, u, n2-1 - n8*1, -(n >> 4), A, 16); + imdct_step3_inner_r_loop(n >> 5, u, n2-1 - n8*2, -(n >> 4), A, 16); + imdct_step3_inner_r_loop(n >> 5, u, n2-1 - n8*3, -(n >> 4), A, 16); + + l=2; + for (; l < (ld-3)>>1; ++l) { + int k0 = n >> (l+2), k0_2 = k0>>1; + int lim = 1 << (l+1); + int i; + for (i=0; i < lim; ++i) + imdct_step3_inner_r_loop(n >> (l+4), u, n2-1 - k0*i, -k0_2, A, 1 << (l+3)); + } + + for (; l < ld-6; ++l) { + int k0 = n >> (l+2), k1 = 1 << (l+3), k0_2 = k0>>1; + int rlim = n >> (l+6), r; + int lim = 1 << (l+1); + int i_off; + float *A0 = A; + i_off = n2-1; + for (r=rlim; r > 0; --r) { + imdct_step3_inner_s_loop(lim, u, i_off, -k0_2, A0, k1, k0); + A0 += k1*4; + i_off -= 8; + } + } + + // iterations with count: + // ld-6,-5,-4 all interleaved together + // the big win comes from getting rid of needless flops + // due to the constants on pass 5 & 4 being all 1 and 0; + // combining them to be simultaneous to improve cache made little difference + imdct_step3_inner_s_loop_ld654(n >> 5, u, n2-1, A, n); + + // output is u + + // step 4, 5, and 6 + // cannot be in-place because of step 5 + { + uint16 *bitrev = f->bit_reverse[blocktype]; + // weirdly, I'd have thought reading sequentially and writing + // erratically would have been better than vice-versa, but in + // fact that's not what my testing showed. (That is, with + // j = bitreverse(i), do you read i and write j, or read j and write i.) + + float *d0 = &v[n4-4]; + float *d1 = &v[n2-4]; + while (d0 >= v) { + int k4; + + k4 = bitrev[0]; + d1[3] = u[k4+0]; + d1[2] = u[k4+1]; + d0[3] = u[k4+2]; + d0[2] = u[k4+3]; + + k4 = bitrev[1]; + d1[1] = u[k4+0]; + d1[0] = u[k4+1]; + d0[1] = u[k4+2]; + d0[0] = u[k4+3]; + + d0 -= 4; + d1 -= 4; + bitrev += 2; + } + } + // (paper output is u, now v) + + + // data must be in buf2 + assert(v == buf2); + + // step 7 (paper output is v, now v) + // this is now in place + { + float *C = f->C[blocktype]; + float *d, *e; + + d = v; + e = v + n2 - 4; + + while (d < e) { + float a02,a11,b0,b1,b2,b3; + + a02 = d[0] - e[2]; + a11 = d[1] + e[3]; + + b0 = C[1]*a02 + C[0]*a11; + b1 = C[1]*a11 - C[0]*a02; + + b2 = d[0] + e[ 2]; + b3 = d[1] - e[ 3]; + + d[0] = b2 + b0; + d[1] = b3 + b1; + e[2] = b2 - b0; + e[3] = b1 - b3; + + a02 = d[2] - e[0]; + a11 = d[3] + e[1]; + + b0 = C[3]*a02 + C[2]*a11; + b1 = C[3]*a11 - C[2]*a02; + + b2 = d[2] + e[ 0]; + b3 = d[3] - e[ 1]; + + d[2] = b2 + b0; + d[3] = b3 + b1; + e[0] = b2 - b0; + e[1] = b1 - b3; + + C += 4; + d += 4; + e -= 4; + } + } + + // data must be in buf2 + + + // step 8+decode (paper output is X, now buffer) + // this generates pairs of data a la 8 and pushes them directly through + // the decode kernel (pushing rather than pulling) to avoid having + // to make another pass later + + // this cannot POSSIBLY be in place, so we refer to the buffers directly + + { + float *d0,*d1,*d2,*d3; + + float *B = f->B[blocktype] + n2 - 8; + float *e = buf2 + n2 - 8; + d0 = &buffer[0]; + d1 = &buffer[n2-4]; + d2 = &buffer[n2]; + d3 = &buffer[n-4]; + while (e >= v) { + float p0,p1,p2,p3; + + p3 = e[6]*B[7] - e[7]*B[6]; + p2 = -e[6]*B[6] - e[7]*B[7]; + + d0[0] = p3; + d1[3] = - p3; + d2[0] = p2; + d3[3] = p2; + + p1 = e[4]*B[5] - e[5]*B[4]; + p0 = -e[4]*B[4] - e[5]*B[5]; + + d0[1] = p1; + d1[2] = - p1; + d2[1] = p0; + d3[2] = p0; + + p3 = e[2]*B[3] - e[3]*B[2]; + p2 = -e[2]*B[2] - e[3]*B[3]; + + d0[2] = p3; + d1[1] = - p3; + d2[2] = p2; + d3[1] = p2; + + p1 = e[0]*B[1] - e[1]*B[0]; + p0 = -e[0]*B[0] - e[1]*B[1]; + + d0[3] = p1; + d1[0] = - p1; + d2[3] = p0; + d3[0] = p0; + + B -= 8; + e -= 8; + d0 += 4; + d2 += 4; + d1 -= 4; + d3 -= 4; + } + } + + temp_alloc_restore(f,save_point); +} + +#if 0 +// this is the original version of the above code, if you want to optimize it from scratch +void inverse_mdct_naive(float *buffer, int n) +{ + float s; + float A[1 << 12], B[1 << 12], C[1 << 11]; + int i,k,k2,k4, n2 = n >> 1, n4 = n >> 2, n8 = n >> 3, l; + int n3_4 = n - n4, ld; + // how can they claim this only uses N words?! + // oh, because they're only used sparsely, whoops + float u[1 << 13], X[1 << 13], v[1 << 13], w[1 << 13]; + // set up twiddle factors + + for (k=k2=0; k < n4; ++k,k2+=2) { + A[k2 ] = (float) cos(4*k*M_PI/n); + A[k2+1] = (float) -sin(4*k*M_PI/n); + B[k2 ] = (float) cos((k2+1)*M_PI/n/2); + B[k2+1] = (float) sin((k2+1)*M_PI/n/2); + } + for (k=k2=0; k < n8; ++k,k2+=2) { + C[k2 ] = (float) cos(2*(k2+1)*M_PI/n); + C[k2+1] = (float) -sin(2*(k2+1)*M_PI/n); + } + + // IMDCT algorithm from "The use of multirate filter banks for coding of high quality digital audio" + // Note there are bugs in that pseudocode, presumably due to them attempting + // to rename the arrays nicely rather than representing the way their actual + // implementation bounces buffers back and forth. As a result, even in the + // "some formulars corrected" version, a direct implementation fails. These + // are noted below as "paper bug". + + // copy and reflect spectral data + for (k=0; k < n2; ++k) u[k] = buffer[k]; + for ( ; k < n ; ++k) u[k] = -buffer[n - k - 1]; + // kernel from paper + // step 1 + for (k=k2=k4=0; k < n4; k+=1, k2+=2, k4+=4) { + v[n-k4-1] = (u[k4] - u[n-k4-1]) * A[k2] - (u[k4+2] - u[n-k4-3])*A[k2+1]; + v[n-k4-3] = (u[k4] - u[n-k4-1]) * A[k2+1] + (u[k4+2] - u[n-k4-3])*A[k2]; + } + // step 2 + for (k=k4=0; k < n8; k+=1, k4+=4) { + w[n2+3+k4] = v[n2+3+k4] + v[k4+3]; + w[n2+1+k4] = v[n2+1+k4] + v[k4+1]; + w[k4+3] = (v[n2+3+k4] - v[k4+3])*A[n2-4-k4] - (v[n2+1+k4]-v[k4+1])*A[n2-3-k4]; + w[k4+1] = (v[n2+1+k4] - v[k4+1])*A[n2-4-k4] + (v[n2+3+k4]-v[k4+3])*A[n2-3-k4]; + } + // step 3 + ld = ilog(n) - 1; // ilog is off-by-one from normal definitions + for (l=0; l < ld-3; ++l) { + int k0 = n >> (l+2), k1 = 1 << (l+3); + int rlim = n >> (l+4), r4, r; + int s2lim = 1 << (l+2), s2; + for (r=r4=0; r < rlim; r4+=4,++r) { + for (s2=0; s2 < s2lim; s2+=2) { + u[n-1-k0*s2-r4] = w[n-1-k0*s2-r4] + w[n-1-k0*(s2+1)-r4]; + u[n-3-k0*s2-r4] = w[n-3-k0*s2-r4] + w[n-3-k0*(s2+1)-r4]; + u[n-1-k0*(s2+1)-r4] = (w[n-1-k0*s2-r4] - w[n-1-k0*(s2+1)-r4]) * A[r*k1] + - (w[n-3-k0*s2-r4] - w[n-3-k0*(s2+1)-r4]) * A[r*k1+1]; + u[n-3-k0*(s2+1)-r4] = (w[n-3-k0*s2-r4] - w[n-3-k0*(s2+1)-r4]) * A[r*k1] + + (w[n-1-k0*s2-r4] - w[n-1-k0*(s2+1)-r4]) * A[r*k1+1]; + } + } + if (l+1 < ld-3) { + // paper bug: ping-ponging of u&w here is omitted + memcpy(w, u, sizeof(u)); + } + } + + // step 4 + for (i=0; i < n8; ++i) { + int j = bit_reverse(i) >> (32-ld+3); + assert(j < n8); + if (i == j) { + // paper bug: original code probably swapped in place; if copying, + // need to directly copy in this case + int i8 = i << 3; + v[i8+1] = u[i8+1]; + v[i8+3] = u[i8+3]; + v[i8+5] = u[i8+5]; + v[i8+7] = u[i8+7]; + } else if (i < j) { + int i8 = i << 3, j8 = j << 3; + v[j8+1] = u[i8+1], v[i8+1] = u[j8 + 1]; + v[j8+3] = u[i8+3], v[i8+3] = u[j8 + 3]; + v[j8+5] = u[i8+5], v[i8+5] = u[j8 + 5]; + v[j8+7] = u[i8+7], v[i8+7] = u[j8 + 7]; + } + } + // step 5 + for (k=0; k < n2; ++k) { + w[k] = v[k*2+1]; + } + // step 6 + for (k=k2=k4=0; k < n8; ++k, k2 += 2, k4 += 4) { + u[n-1-k2] = w[k4]; + u[n-2-k2] = w[k4+1]; + u[n3_4 - 1 - k2] = w[k4+2]; + u[n3_4 - 2 - k2] = w[k4+3]; + } + // step 7 + for (k=k2=0; k < n8; ++k, k2 += 2) { + v[n2 + k2 ] = ( u[n2 + k2] + u[n-2-k2] + C[k2+1]*(u[n2+k2]-u[n-2-k2]) + C[k2]*(u[n2+k2+1]+u[n-2-k2+1]))/2; + v[n-2 - k2] = ( u[n2 + k2] + u[n-2-k2] - C[k2+1]*(u[n2+k2]-u[n-2-k2]) - C[k2]*(u[n2+k2+1]+u[n-2-k2+1]))/2; + v[n2+1+ k2] = ( u[n2+1+k2] - u[n-1-k2] + C[k2+1]*(u[n2+1+k2]+u[n-1-k2]) - C[k2]*(u[n2+k2]-u[n-2-k2]))/2; + v[n-1 - k2] = (-u[n2+1+k2] + u[n-1-k2] + C[k2+1]*(u[n2+1+k2]+u[n-1-k2]) - C[k2]*(u[n2+k2]-u[n-2-k2]))/2; + } + // step 8 + for (k=k2=0; k < n4; ++k,k2 += 2) { + X[k] = v[k2+n2]*B[k2 ] + v[k2+1+n2]*B[k2+1]; + X[n2-1-k] = v[k2+n2]*B[k2+1] - v[k2+1+n2]*B[k2 ]; + } + + // decode kernel to output + // determined the following value experimentally + // (by first figuring out what made inverse_mdct_slow work); then matching that here + // (probably vorbis encoder premultiplies by n or n/2, to save it on the decoder?) + s = 0.5; // theoretically would be n4 + + // [[[ note! the s value of 0.5 is compensated for by the B[] in the current code, + // so it needs to use the "old" B values to behave correctly, or else + // set s to 1.0 ]]] + for (i=0; i < n4 ; ++i) buffer[i] = s * X[i+n4]; + for ( ; i < n3_4; ++i) buffer[i] = -s * X[n3_4 - i - 1]; + for ( ; i < n ; ++i) buffer[i] = -s * X[i - n3_4]; +} +#endif + +static float *get_window(vorb *f, int len) +{ + len <<= 1; + if (len == f->blocksize_0) return f->window[0]; + if (len == f->blocksize_1) return f->window[1]; + assert(0); + return NULL; +} + +#ifndef STB_VORBIS_NO_DEFER_FLOOR +typedef int16 YTYPE; +#else +typedef int YTYPE; +#endif +static int do_floor(vorb *f, Mapping *map, int i, int n, float *target, YTYPE *finalY, uint8 *step2_flag) +{ + int n2 = n >> 1; + int s = map->chan[i].mux, floor; + floor = map->submap_floor[s]; + if (f->floor_types[floor] == 0) { + return error(f, VORBIS_invalid_stream); + } else { + Floor1 *g = &f->floor_config[floor].floor1; + int j,q; + int lx = 0, ly = finalY[0] * g->floor1_multiplier; + for (q=1; q < g->values; ++q) { + j = g->sorted_order[q]; + #ifndef STB_VORBIS_NO_DEFER_FLOOR + if (finalY[j] >= 0) + #else + if (step2_flag[j]) + #endif + { + int hy = finalY[j] * g->floor1_multiplier; + int hx = g->Xlist[j]; + if (lx != hx) + draw_line(target, lx,ly, hx,hy, n2); + CHECK(f); + lx = hx, ly = hy; + } + } + if (lx < n2) { + // optimization of: draw_line(target, lx,ly, n,ly, n2); + for (j=lx; j < n2; ++j) + LINE_OP(target[j], inverse_db_table[ly]); + CHECK(f); + } + } + return TRUE; +} + +// The meaning of "left" and "right" +// +// For a given frame: +// we compute samples from 0..n +// window_center is n/2 +// we'll window and mix the samples from left_start to left_end with data from the previous frame +// all of the samples from left_end to right_start can be output without mixing; however, +// this interval is 0-length except when transitioning between short and long frames +// all of the samples from right_start to right_end need to be mixed with the next frame, +// which we don't have, so those get saved in a buffer +// frame N's right_end-right_start, the number of samples to mix with the next frame, +// has to be the same as frame N+1's left_end-left_start (which they are by +// construction) + +static int vorbis_decode_initial(vorb *f, int *p_left_start, int *p_left_end, int *p_right_start, int *p_right_end, int *mode) +{ + Mode *m; + int i, n, prev, next, window_center; + f->channel_buffer_start = f->channel_buffer_end = 0; + + retry: + if (f->eof) return FALSE; + if (!maybe_start_packet(f)) + return FALSE; + // check packet type + if (get_bits(f,1) != 0) { + if (IS_PUSH_MODE(f)) + return error(f,VORBIS_bad_packet_type); + while (EOP != get8_packet(f)); + goto retry; + } + + if (f->alloc.alloc_buffer) + assert(f->alloc.alloc_buffer_length_in_bytes == f->temp_offset); + + i = get_bits(f, ilog(f->mode_count-1)); + if (i == EOP) return FALSE; + if (i >= f->mode_count) return FALSE; + *mode = i; + m = f->mode_config + i; + if (m->blockflag) { + n = f->blocksize_1; + prev = get_bits(f,1); + next = get_bits(f,1); + } else { + prev = next = 0; + n = f->blocksize_0; + } + +// WINDOWING + + window_center = n >> 1; + if (m->blockflag && !prev) { + *p_left_start = (n - f->blocksize_0) >> 2; + *p_left_end = (n + f->blocksize_0) >> 2; + } else { + *p_left_start = 0; + *p_left_end = window_center; + } + if (m->blockflag && !next) { + *p_right_start = (n*3 - f->blocksize_0) >> 2; + *p_right_end = (n*3 + f->blocksize_0) >> 2; + } else { + *p_right_start = window_center; + *p_right_end = n; + } + + return TRUE; +} + +static int vorbis_decode_packet_rest(vorb *f, int *len, Mode *m, int left_start, int left_end, int right_start, int right_end, int *p_left) +{ + Mapping *map; + int i,j,k,n,n2; + int zero_channel[256]; + int really_zero_channel[256]; + +// WINDOWING + + n = f->blocksize[m->blockflag]; + map = &f->mapping[m->mapping]; + +// FLOORS + n2 = n >> 1; + + CHECK(f); + + for (i=0; i < f->channels; ++i) { + int s = map->chan[i].mux, floor; + zero_channel[i] = FALSE; + floor = map->submap_floor[s]; + if (f->floor_types[floor] == 0) { + return error(f, VORBIS_invalid_stream); + } else { + Floor1 *g = &f->floor_config[floor].floor1; + if (get_bits(f, 1)) { + short *finalY; + uint8 step2_flag[256]; + static int range_list[4] = { 256, 128, 86, 64 }; + int range = range_list[g->floor1_multiplier-1]; + int offset = 2; + finalY = f->finalY[i]; + finalY[0] = get_bits(f, ilog(range)-1); + finalY[1] = get_bits(f, ilog(range)-1); + for (j=0; j < g->partitions; ++j) { + int pclass = g->partition_class_list[j]; + int cdim = g->class_dimensions[pclass]; + int cbits = g->class_subclasses[pclass]; + int csub = (1 << cbits)-1; + int cval = 0; + if (cbits) { + Codebook *c = f->codebooks + g->class_masterbooks[pclass]; + DECODE(cval,f,c); + } + for (k=0; k < cdim; ++k) { + int book = g->subclass_books[pclass][cval & csub]; + cval = cval >> cbits; + if (book >= 0) { + int temp; + Codebook *c = f->codebooks + book; + DECODE(temp,f,c); + finalY[offset++] = temp; + } else + finalY[offset++] = 0; + } + } + if (f->valid_bits == INVALID_BITS) goto error; // behavior according to spec + step2_flag[0] = step2_flag[1] = 1; + for (j=2; j < g->values; ++j) { + int low, high, pred, highroom, lowroom, room, val; + low = g->neighbors[j][0]; + high = g->neighbors[j][1]; + //neighbors(g->Xlist, j, &low, &high); + pred = predict_point(g->Xlist[j], g->Xlist[low], g->Xlist[high], finalY[low], finalY[high]); + val = finalY[j]; + highroom = range - pred; + lowroom = pred; + if (highroom < lowroom) + room = highroom * 2; + else + room = lowroom * 2; + if (val) { + step2_flag[low] = step2_flag[high] = 1; + step2_flag[j] = 1; + if (val >= room) + if (highroom > lowroom) + finalY[j] = val - lowroom + pred; + else + finalY[j] = pred - val + highroom - 1; + else + if (val & 1) + finalY[j] = pred - ((val+1)>>1); + else + finalY[j] = pred + (val>>1); + } else { + step2_flag[j] = 0; + finalY[j] = pred; + } + } + +#ifdef STB_VORBIS_NO_DEFER_FLOOR + do_floor(f, map, i, n, f->floor_buffers[i], finalY, step2_flag); +#else + // defer final floor computation until _after_ residue + for (j=0; j < g->values; ++j) { + if (!step2_flag[j]) + finalY[j] = -1; + } +#endif + } else { + error: + zero_channel[i] = TRUE; + } + // So we just defer everything else to later + + // at this point we've decoded the floor into buffer + } + } + CHECK(f); + // at this point we've decoded all floors + + if (f->alloc.alloc_buffer) + assert(f->alloc.alloc_buffer_length_in_bytes == f->temp_offset); + + // re-enable coupled channels if necessary + memcpy(really_zero_channel, zero_channel, sizeof(really_zero_channel[0]) * f->channels); + for (i=0; i < map->coupling_steps; ++i) + if (!zero_channel[map->chan[i].magnitude] || !zero_channel[map->chan[i].angle]) { + zero_channel[map->chan[i].magnitude] = zero_channel[map->chan[i].angle] = FALSE; + } + + CHECK(f); +// RESIDUE DECODE + for (i=0; i < map->submaps; ++i) { + float *residue_buffers[STB_VORBIS_MAX_CHANNELS]; + int r; + uint8 do_not_decode[256]; + int ch = 0; + for (j=0; j < f->channels; ++j) { + if (map->chan[j].mux == i) { + if (zero_channel[j]) { + do_not_decode[ch] = TRUE; + residue_buffers[ch] = NULL; + } else { + do_not_decode[ch] = FALSE; + residue_buffers[ch] = f->channel_buffers[j]; + } + ++ch; + } + } + r = map->submap_residue[i]; + decode_residue(f, residue_buffers, ch, n2, r, do_not_decode); + } + + if (f->alloc.alloc_buffer) + assert(f->alloc.alloc_buffer_length_in_bytes == f->temp_offset); + CHECK(f); + +// INVERSE COUPLING + for (i = map->coupling_steps-1; i >= 0; --i) { + int n2 = n >> 1; + float *m = f->channel_buffers[map->chan[i].magnitude]; + float *a = f->channel_buffers[map->chan[i].angle ]; + for (j=0; j < n2; ++j) { + float a2,m2; + if (m[j] > 0) + if (a[j] > 0) + m2 = m[j], a2 = m[j] - a[j]; + else + a2 = m[j], m2 = m[j] + a[j]; + else + if (a[j] > 0) + m2 = m[j], a2 = m[j] + a[j]; + else + a2 = m[j], m2 = m[j] - a[j]; + m[j] = m2; + a[j] = a2; + } + } + CHECK(f); + + // finish decoding the floors +#ifndef STB_VORBIS_NO_DEFER_FLOOR + for (i=0; i < f->channels; ++i) { + if (really_zero_channel[i]) { + memset(f->channel_buffers[i], 0, sizeof(*f->channel_buffers[i]) * n2); + } else { + do_floor(f, map, i, n, f->channel_buffers[i], f->finalY[i], NULL); + } + } +#else + for (i=0; i < f->channels; ++i) { + if (really_zero_channel[i]) { + memset(f->channel_buffers[i], 0, sizeof(*f->channel_buffers[i]) * n2); + } else { + for (j=0; j < n2; ++j) + f->channel_buffers[i][j] *= f->floor_buffers[i][j]; + } + } +#endif + +// INVERSE MDCT + CHECK(f); + for (i=0; i < f->channels; ++i) + inverse_mdct(f->channel_buffers[i], n, f, m->blockflag); + CHECK(f); + + // this shouldn't be necessary, unless we exited on an error + // and want to flush to get to the next packet + flush_packet(f); + + if (f->first_decode) { + // assume we start so first non-discarded sample is sample 0 + // this isn't to spec, but spec would require us to read ahead + // and decode the size of all current frames--could be done, + // but presumably it's not a commonly used feature + f->current_loc = -n2; // start of first frame is positioned for discard + // we might have to discard samples "from" the next frame too, + // if we're lapping a large block then a small at the start? + f->discard_samples_deferred = n - right_end; + f->current_loc_valid = TRUE; + f->first_decode = FALSE; + } else if (f->discard_samples_deferred) { + if (f->discard_samples_deferred >= right_start - left_start) { + f->discard_samples_deferred -= (right_start - left_start); + left_start = right_start; + *p_left = left_start; + } else { + left_start += f->discard_samples_deferred; + *p_left = left_start; + f->discard_samples_deferred = 0; + } + } else if (f->previous_length == 0 && f->current_loc_valid) { + // we're recovering from a seek... that means we're going to discard + // the samples from this packet even though we know our position from + // the last page header, so we need to update the position based on + // the discarded samples here + // but wait, the code below is going to add this in itself even + // on a discard, so we don't need to do it here... + } + + // check if we have ogg information about the sample # for this packet + if (f->last_seg_which == f->end_seg_with_known_loc) { + // if we have a valid current loc, and this is final: + if (f->current_loc_valid && (f->page_flag & PAGEFLAG_last_page)) { + uint32 current_end = f->known_loc_for_packet - (n-right_end); + // then let's infer the size of the (probably) short final frame + if (current_end < f->current_loc + (right_end-left_start)) { + if (current_end < f->current_loc) { + // negative truncation, that's impossible! + *len = 0; + } else { + *len = current_end - f->current_loc; + } + *len += left_start; + if (*len > right_end) *len = right_end; // this should never happen + f->current_loc += *len; + return TRUE; + } + } + // otherwise, just set our sample loc + // guess that the ogg granule pos refers to the _middle_ of the + // last frame? + // set f->current_loc to the position of left_start + f->current_loc = f->known_loc_for_packet - (n2-left_start); + f->current_loc_valid = TRUE; + } + if (f->current_loc_valid) + f->current_loc += (right_start - left_start); + + if (f->alloc.alloc_buffer) + assert(f->alloc.alloc_buffer_length_in_bytes == f->temp_offset); + *len = right_end; // ignore samples after the window goes to 0 + CHECK(f); + + return TRUE; +} + +static int vorbis_decode_packet(vorb *f, int *len, int *p_left, int *p_right) +{ + int mode, left_end, right_end; + if (!vorbis_decode_initial(f, p_left, &left_end, p_right, &right_end, &mode)) return 0; + return vorbis_decode_packet_rest(f, len, f->mode_config + mode, *p_left, left_end, *p_right, right_end, p_left); +} + +static int vorbis_finish_frame(stb_vorbis *f, int len, int left, int right) +{ + int prev,i,j; + // we use right&left (the start of the right- and left-window sin()-regions) + // to determine how much to return, rather than inferring from the rules + // (same result, clearer code); 'left' indicates where our sin() window + // starts, therefore where the previous window's right edge starts, and + // therefore where to start mixing from the previous buffer. 'right' + // indicates where our sin() ending-window starts, therefore that's where + // we start saving, and where our returned-data ends. + + // mixin from previous window + if (f->previous_length) { + int i,j, n = f->previous_length; + float *w = get_window(f, n); + for (i=0; i < f->channels; ++i) { + for (j=0; j < n; ++j) + f->channel_buffers[i][left+j] = + f->channel_buffers[i][left+j]*w[ j] + + f->previous_window[i][ j]*w[n-1-j]; + } + } + + prev = f->previous_length; + + // last half of this data becomes previous window + f->previous_length = len - right; + + // @OPTIMIZE: could avoid this copy by double-buffering the + // output (flipping previous_window with channel_buffers), but + // then previous_window would have to be 2x as large, and + // channel_buffers couldn't be temp mem (although they're NOT + // currently temp mem, they could be (unless we want to level + // performance by spreading out the computation)) + for (i=0; i < f->channels; ++i) + for (j=0; right+j < len; ++j) + f->previous_window[i][j] = f->channel_buffers[i][right+j]; + + if (!prev) + // there was no previous packet, so this data isn't valid... + // this isn't entirely true, only the would-have-overlapped data + // isn't valid, but this seems to be what the spec requires + return 0; + + // truncate a short frame + if (len < right) right = len; + + f->samples_output += right-left; + + return right - left; +} + +static void vorbis_pump_first_frame(stb_vorbis *f) +{ + int len, right, left; + if (vorbis_decode_packet(f, &len, &left, &right)) + vorbis_finish_frame(f, len, left, right); +} + +#ifndef STB_VORBIS_NO_PUSHDATA_API +static int is_whole_packet_present(stb_vorbis *f, int end_page) +{ + // make sure that we have the packet available before continuing... + // this requires a full ogg parse, but we know we can fetch from f->stream + + // instead of coding this out explicitly, we could save the current read state, + // read the next packet with get8() until end-of-packet, check f->eof, then + // reset the state? but that would be slower, esp. since we'd have over 256 bytes + // of state to restore (primarily the page segment table) + + int s = f->next_seg, first = TRUE; + uint8 *p = f->stream; + + if (s != -1) { // if we're not starting the packet with a 'continue on next page' flag + for (; s < f->segment_count; ++s) { + p += f->segments[s]; + if (f->segments[s] < 255) // stop at first short segment + break; + } + // either this continues, or it ends it... + if (end_page) + if (s < f->segment_count-1) return error(f, VORBIS_invalid_stream); + if (s == f->segment_count) + s = -1; // set 'crosses page' flag + if (p > f->stream_end) return error(f, VORBIS_need_more_data); + first = FALSE; + } + for (; s == -1;) { + uint8 *q; + int n; + + // check that we have the page header ready + if (p + 26 >= f->stream_end) return error(f, VORBIS_need_more_data); + // validate the page + if (memcmp(p, ogg_page_header, 4)) return error(f, VORBIS_invalid_stream); + if (p[4] != 0) return error(f, VORBIS_invalid_stream); + if (first) { // the first segment must NOT have 'continued_packet', later ones MUST + if (f->previous_length) + if ((p[5] & PAGEFLAG_continued_packet)) return error(f, VORBIS_invalid_stream); + // if no previous length, we're resynching, so we can come in on a continued-packet, + // which we'll just drop + } else { + if (!(p[5] & PAGEFLAG_continued_packet)) return error(f, VORBIS_invalid_stream); + } + n = p[26]; // segment counts + q = p+27; // q points to segment table + p = q + n; // advance past header + // make sure we've read the segment table + if (p > f->stream_end) return error(f, VORBIS_need_more_data); + for (s=0; s < n; ++s) { + p += q[s]; + if (q[s] < 255) + break; + } + if (end_page) + if (s < n-1) return error(f, VORBIS_invalid_stream); + if (s == n) + s = -1; // set 'crosses page' flag + if (p > f->stream_end) return error(f, VORBIS_need_more_data); + first = FALSE; + } + return TRUE; +} +#endif // !STB_VORBIS_NO_PUSHDATA_API + +static int start_decoder(vorb *f) +{ + uint8 header[6], x,y; + int len,i,j,k, max_submaps = 0; + int longest_floorlist=0; + + // first page, first packet + + if (!start_page(f)) return FALSE; + // validate page flag + if (!(f->page_flag & PAGEFLAG_first_page)) return error(f, VORBIS_invalid_first_page); + if (f->page_flag & PAGEFLAG_last_page) return error(f, VORBIS_invalid_first_page); + if (f->page_flag & PAGEFLAG_continued_packet) return error(f, VORBIS_invalid_first_page); + // check for expected packet length + if (f->segment_count != 1) return error(f, VORBIS_invalid_first_page); + if (f->segments[0] != 30) return error(f, VORBIS_invalid_first_page); + // read packet + // check packet header + if (get8(f) != VORBIS_packet_id) return error(f, VORBIS_invalid_first_page); + if (!getn(f, header, 6)) return error(f, VORBIS_unexpected_eof); + if (!vorbis_validate(header)) return error(f, VORBIS_invalid_first_page); + // vorbis_version + if (get32(f) != 0) return error(f, VORBIS_invalid_first_page); + f->channels = get8(f); if (!f->channels) return error(f, VORBIS_invalid_first_page); + if (f->channels > STB_VORBIS_MAX_CHANNELS) return error(f, VORBIS_too_many_channels); + f->sample_rate = get32(f); if (!f->sample_rate) return error(f, VORBIS_invalid_first_page); + get32(f); // bitrate_maximum + get32(f); // bitrate_nominal + get32(f); // bitrate_minimum + x = get8(f); + { + int log0,log1; + log0 = x & 15; + log1 = x >> 4; + f->blocksize_0 = 1 << log0; + f->blocksize_1 = 1 << log1; + if (log0 < 6 || log0 > 13) return error(f, VORBIS_invalid_setup); + if (log1 < 6 || log1 > 13) return error(f, VORBIS_invalid_setup); + if (log0 > log1) return error(f, VORBIS_invalid_setup); + } + + // framing_flag + x = get8(f); + if (!(x & 1)) return error(f, VORBIS_invalid_first_page); + + // second packet! + if (!start_page(f)) return FALSE; + + if (!start_packet(f)) return FALSE; + do { + len = next_segment(f); + skip(f, len); + f->bytes_in_seg = 0; + } while (len); + + // third packet! + if (!start_packet(f)) return FALSE; + + #ifndef STB_VORBIS_NO_PUSHDATA_API + if (IS_PUSH_MODE(f)) { + if (!is_whole_packet_present(f, TRUE)) { + // convert error in ogg header to write type + if (f->error == VORBIS_invalid_stream) + f->error = VORBIS_invalid_setup; + return FALSE; + } + } + #endif + + crc32_init(); // always init it, to avoid multithread race conditions + + if (get8_packet(f) != VORBIS_packet_setup) return error(f, VORBIS_invalid_setup); + for (i=0; i < 6; ++i) header[i] = get8_packet(f); + if (!vorbis_validate(header)) return error(f, VORBIS_invalid_setup); + + // codebooks + + f->codebook_count = get_bits(f,8) + 1; + f->codebooks = (Codebook *) setup_malloc(f, sizeof(*f->codebooks) * f->codebook_count); + if (f->codebooks == NULL) return error(f, VORBIS_outofmem); + memset(f->codebooks, 0, sizeof(*f->codebooks) * f->codebook_count); + for (i=0; i < f->codebook_count; ++i) { + uint32 *values; + int ordered, sorted_count; + int total=0; + uint8 *lengths; + Codebook *c = f->codebooks+i; + CHECK(f); + x = get_bits(f, 8); if (x != 0x42) return error(f, VORBIS_invalid_setup); + x = get_bits(f, 8); if (x != 0x43) return error(f, VORBIS_invalid_setup); + x = get_bits(f, 8); if (x != 0x56) return error(f, VORBIS_invalid_setup); + x = get_bits(f, 8); + c->dimensions = (get_bits(f, 8)<<8) + x; + x = get_bits(f, 8); + y = get_bits(f, 8); + c->entries = (get_bits(f, 8)<<16) + (y<<8) + x; + ordered = get_bits(f,1); + c->sparse = ordered ? 0 : get_bits(f,1); + + if (c->dimensions == 0 && c->entries != 0) return error(f, VORBIS_invalid_setup); + + if (c->sparse) + lengths = (uint8 *) setup_temp_malloc(f, c->entries); + else + lengths = c->codeword_lengths = (uint8 *) setup_malloc(f, c->entries); + + if (!lengths) return error(f, VORBIS_outofmem); + + if (ordered) { + int current_entry = 0; + int current_length = get_bits(f,5) + 1; + while (current_entry < c->entries) { + int limit = c->entries - current_entry; + int n = get_bits(f, ilog(limit)); + if (current_entry + n > (int) c->entries) { return error(f, VORBIS_invalid_setup); } + memset(lengths + current_entry, current_length, n); + current_entry += n; + ++current_length; + } + } else { + for (j=0; j < c->entries; ++j) { + int present = c->sparse ? get_bits(f,1) : 1; + if (present) { + lengths[j] = get_bits(f, 5) + 1; + ++total; + if (lengths[j] == 32) + return error(f, VORBIS_invalid_setup); + } else { + lengths[j] = NO_CODE; + } + } + } + + if (c->sparse && total >= c->entries >> 2) { + // convert sparse items to non-sparse! + if (c->entries > (int) f->setup_temp_memory_required) + f->setup_temp_memory_required = c->entries; + + c->codeword_lengths = (uint8 *) setup_malloc(f, c->entries); + if (c->codeword_lengths == NULL) return error(f, VORBIS_outofmem); + memcpy(c->codeword_lengths, lengths, c->entries); + setup_temp_free(f, lengths, c->entries); // note this is only safe if there have been no intervening temp mallocs! + lengths = c->codeword_lengths; + c->sparse = 0; + } + + // compute the size of the sorted tables + if (c->sparse) { + sorted_count = total; + } else { + sorted_count = 0; + #ifndef STB_VORBIS_NO_HUFFMAN_BINARY_SEARCH + for (j=0; j < c->entries; ++j) + if (lengths[j] > STB_VORBIS_FAST_HUFFMAN_LENGTH && lengths[j] != NO_CODE) + ++sorted_count; + #endif + } + + c->sorted_entries = sorted_count; + values = NULL; + + CHECK(f); + if (!c->sparse) { + c->codewords = (uint32 *) setup_malloc(f, sizeof(c->codewords[0]) * c->entries); + if (!c->codewords) return error(f, VORBIS_outofmem); + } else { + unsigned int size; + if (c->sorted_entries) { + c->codeword_lengths = (uint8 *) setup_malloc(f, c->sorted_entries); + if (!c->codeword_lengths) return error(f, VORBIS_outofmem); + c->codewords = (uint32 *) setup_temp_malloc(f, sizeof(*c->codewords) * c->sorted_entries); + if (!c->codewords) return error(f, VORBIS_outofmem); + values = (uint32 *) setup_temp_malloc(f, sizeof(*values) * c->sorted_entries); + if (!values) return error(f, VORBIS_outofmem); + } + size = c->entries + (sizeof(*c->codewords) + sizeof(*values)) * c->sorted_entries; + if (size > f->setup_temp_memory_required) + f->setup_temp_memory_required = size; + } + + if (!compute_codewords(c, lengths, c->entries, values)) { + if (c->sparse) setup_temp_free(f, values, 0); + return error(f, VORBIS_invalid_setup); + } + + if (c->sorted_entries) { + // allocate an extra slot for sentinels + c->sorted_codewords = (uint32 *) setup_malloc(f, sizeof(*c->sorted_codewords) * (c->sorted_entries+1)); + if (c->sorted_codewords == NULL) return error(f, VORBIS_outofmem); + // allocate an extra slot at the front so that c->sorted_values[-1] is defined + // so that we can catch that case without an extra if + c->sorted_values = ( int *) setup_malloc(f, sizeof(*c->sorted_values ) * (c->sorted_entries+1)); + if (c->sorted_values == NULL) return error(f, VORBIS_outofmem); + ++c->sorted_values; + c->sorted_values[-1] = -1; + compute_sorted_huffman(c, lengths, values); + } + + if (c->sparse) { + setup_temp_free(f, values, sizeof(*values)*c->sorted_entries); + setup_temp_free(f, c->codewords, sizeof(*c->codewords)*c->sorted_entries); + setup_temp_free(f, lengths, c->entries); + c->codewords = NULL; + } + + compute_accelerated_huffman(c); + + CHECK(f); + c->lookup_type = get_bits(f, 4); + if (c->lookup_type > 2) return error(f, VORBIS_invalid_setup); + if (c->lookup_type > 0) { + uint16 *mults; + c->minimum_value = float32_unpack(get_bits(f, 32)); + c->delta_value = float32_unpack(get_bits(f, 32)); + c->value_bits = get_bits(f, 4)+1; + c->sequence_p = get_bits(f,1); + if (c->lookup_type == 1) { + c->lookup_values = lookup1_values(c->entries, c->dimensions); + } else { + c->lookup_values = c->entries * c->dimensions; + } + if (c->lookup_values == 0) return error(f, VORBIS_invalid_setup); + mults = (uint16 *) setup_temp_malloc(f, sizeof(mults[0]) * c->lookup_values); + if (mults == NULL) return error(f, VORBIS_outofmem); + for (j=0; j < (int) c->lookup_values; ++j) { + int q = get_bits(f, c->value_bits); + if (q == EOP) { setup_temp_free(f,mults,sizeof(mults[0])*c->lookup_values); return error(f, VORBIS_invalid_setup); } + mults[j] = q; + } + +#ifndef STB_VORBIS_DIVIDES_IN_CODEBOOK + if (c->lookup_type == 1) { + int len, sparse = c->sparse; + float last=0; + // pre-expand the lookup1-style multiplicands, to avoid a divide in the inner loop + if (sparse) { + if (c->sorted_entries == 0) goto skip; + c->multiplicands = (codetype *) setup_malloc(f, sizeof(c->multiplicands[0]) * c->sorted_entries * c->dimensions); + } else + c->multiplicands = (codetype *) setup_malloc(f, sizeof(c->multiplicands[0]) * c->entries * c->dimensions); + if (c->multiplicands == NULL) { setup_temp_free(f,mults,sizeof(mults[0])*c->lookup_values); return error(f, VORBIS_outofmem); } + len = sparse ? c->sorted_entries : c->entries; + for (j=0; j < len; ++j) { + unsigned int z = sparse ? c->sorted_values[j] : j; + unsigned int div=1; + for (k=0; k < c->dimensions; ++k) { + int off = (z / div) % c->lookup_values; + float val = mults[off]; + val = mults[off]*c->delta_value + c->minimum_value + last; + c->multiplicands[j*c->dimensions + k] = val; + if (c->sequence_p) + last = val; + if (k+1 < c->dimensions) { + if (div > UINT_MAX / (unsigned int) c->lookup_values) { + setup_temp_free(f, mults,sizeof(mults[0])*c->lookup_values); + return error(f, VORBIS_invalid_setup); + } + div *= c->lookup_values; + } + } + } + setup_temp_free(f, mults,sizeof(mults[0])*c->lookup_values); + c->lookup_type = 2; + } + else +#endif + { + float last=0; + CHECK(f); + c->multiplicands = (codetype *) setup_malloc(f, sizeof(c->multiplicands[0]) * c->lookup_values); + if (c->multiplicands == NULL) { setup_temp_free(f, mults,sizeof(mults[0])*c->lookup_values); return error(f, VORBIS_outofmem); } + for (j=0; j < (int) c->lookup_values; ++j) { + float val = mults[j] * c->delta_value + c->minimum_value + last; + c->multiplicands[j] = val; + if (c->sequence_p) + last = val; + } + setup_temp_free(f, mults,sizeof(mults[0])*c->lookup_values); + } +#ifndef STB_VORBIS_DIVIDES_IN_CODEBOOK + skip:; +#endif + + CHECK(f); + } + CHECK(f); + } + + // time domain transfers (notused) + + x = get_bits(f, 6) + 1; + for (i=0; i < x; ++i) { + uint32 z = get_bits(f, 16); + if (z != 0) return error(f, VORBIS_invalid_setup); + } + + // Floors + f->floor_count = get_bits(f, 6)+1; + f->floor_config = (Floor *) setup_malloc(f, f->floor_count * sizeof(*f->floor_config)); + if (f->floor_config == NULL) return error(f, VORBIS_outofmem); + for (i=0; i < f->floor_count; ++i) { + f->floor_types[i] = get_bits(f, 16); + if (f->floor_types[i] > 1) return error(f, VORBIS_invalid_setup); + if (f->floor_types[i] == 0) { + Floor0 *g = &f->floor_config[i].floor0; + g->order = get_bits(f,8); + g->rate = get_bits(f,16); + g->bark_map_size = get_bits(f,16); + g->amplitude_bits = get_bits(f,6); + g->amplitude_offset = get_bits(f,8); + g->number_of_books = get_bits(f,4) + 1; + for (j=0; j < g->number_of_books; ++j) + g->book_list[j] = get_bits(f,8); + return error(f, VORBIS_feature_not_supported); + } else { + Point p[31*8+2]; + Floor1 *g = &f->floor_config[i].floor1; + int max_class = -1; + g->partitions = get_bits(f, 5); + for (j=0; j < g->partitions; ++j) { + g->partition_class_list[j] = get_bits(f, 4); + if (g->partition_class_list[j] > max_class) + max_class = g->partition_class_list[j]; + } + for (j=0; j <= max_class; ++j) { + g->class_dimensions[j] = get_bits(f, 3)+1; + g->class_subclasses[j] = get_bits(f, 2); + if (g->class_subclasses[j]) { + g->class_masterbooks[j] = get_bits(f, 8); + if (g->class_masterbooks[j] >= f->codebook_count) return error(f, VORBIS_invalid_setup); + } + for (k=0; k < 1 << g->class_subclasses[j]; ++k) { + g->subclass_books[j][k] = get_bits(f,8)-1; + if (g->subclass_books[j][k] >= f->codebook_count) return error(f, VORBIS_invalid_setup); + } + } + g->floor1_multiplier = get_bits(f,2)+1; + g->rangebits = get_bits(f,4); + g->Xlist[0] = 0; + g->Xlist[1] = 1 << g->rangebits; + g->values = 2; + for (j=0; j < g->partitions; ++j) { + int c = g->partition_class_list[j]; + for (k=0; k < g->class_dimensions[c]; ++k) { + g->Xlist[g->values] = get_bits(f, g->rangebits); + ++g->values; + } + } + // precompute the sorting + for (j=0; j < g->values; ++j) { + p[j].x = g->Xlist[j]; + p[j].y = j; + } + qsort(p, g->values, sizeof(p[0]), point_compare); + for (j=0; j < g->values; ++j) + g->sorted_order[j] = (uint8) p[j].y; + // precompute the neighbors + for (j=2; j < g->values; ++j) { + int low,hi; + neighbors(g->Xlist, j, &low,&hi); + g->neighbors[j][0] = low; + g->neighbors[j][1] = hi; + } + + if (g->values > longest_floorlist) + longest_floorlist = g->values; + } + } + + // Residue + f->residue_count = get_bits(f, 6)+1; + f->residue_config = (Residue *) setup_malloc(f, f->residue_count * sizeof(f->residue_config[0])); + if (f->residue_config == NULL) return error(f, VORBIS_outofmem); + memset(f->residue_config, 0, f->residue_count * sizeof(f->residue_config[0])); + for (i=0; i < f->residue_count; ++i) { + uint8 residue_cascade[64]; + Residue *r = f->residue_config+i; + f->residue_types[i] = get_bits(f, 16); + if (f->residue_types[i] > 2) return error(f, VORBIS_invalid_setup); + r->begin = get_bits(f, 24); + r->end = get_bits(f, 24); + if (r->end < r->begin) return error(f, VORBIS_invalid_setup); + r->part_size = get_bits(f,24)+1; + r->classifications = get_bits(f,6)+1; + r->classbook = get_bits(f,8); + if (r->classbook >= f->codebook_count) return error(f, VORBIS_invalid_setup); + for (j=0; j < r->classifications; ++j) { + uint8 high_bits=0; + uint8 low_bits=get_bits(f,3); + if (get_bits(f,1)) + high_bits = get_bits(f,5); + residue_cascade[j] = high_bits*8 + low_bits; + } + r->residue_books = (short (*)[8]) setup_malloc(f, sizeof(r->residue_books[0]) * r->classifications); + if (r->residue_books == NULL) return error(f, VORBIS_outofmem); + for (j=0; j < r->classifications; ++j) { + for (k=0; k < 8; ++k) { + if (residue_cascade[j] & (1 << k)) { + r->residue_books[j][k] = get_bits(f, 8); + if (r->residue_books[j][k] >= f->codebook_count) return error(f, VORBIS_invalid_setup); + } else { + r->residue_books[j][k] = -1; + } + } + } + // precompute the classifications[] array to avoid inner-loop mod/divide + // call it 'classdata' since we already have r->classifications + r->classdata = (uint8 **) setup_malloc(f, sizeof(*r->classdata) * f->codebooks[r->classbook].entries); + if (!r->classdata) return error(f, VORBIS_outofmem); + memset(r->classdata, 0, sizeof(*r->classdata) * f->codebooks[r->classbook].entries); + for (j=0; j < f->codebooks[r->classbook].entries; ++j) { + int classwords = f->codebooks[r->classbook].dimensions; + int temp = j; + r->classdata[j] = (uint8 *) setup_malloc(f, sizeof(r->classdata[j][0]) * classwords); + if (r->classdata[j] == NULL) return error(f, VORBIS_outofmem); + for (k=classwords-1; k >= 0; --k) { + r->classdata[j][k] = temp % r->classifications; + temp /= r->classifications; + } + } + } + + f->mapping_count = get_bits(f,6)+1; + f->mapping = (Mapping *) setup_malloc(f, f->mapping_count * sizeof(*f->mapping)); + if (f->mapping == NULL) return error(f, VORBIS_outofmem); + memset(f->mapping, 0, f->mapping_count * sizeof(*f->mapping)); + for (i=0; i < f->mapping_count; ++i) { + Mapping *m = f->mapping + i; + int mapping_type = get_bits(f,16); + if (mapping_type != 0) return error(f, VORBIS_invalid_setup); + m->chan = (MappingChannel *) setup_malloc(f, f->channels * sizeof(*m->chan)); + if (m->chan == NULL) return error(f, VORBIS_outofmem); + if (get_bits(f,1)) + m->submaps = get_bits(f,4)+1; + else + m->submaps = 1; + if (m->submaps > max_submaps) + max_submaps = m->submaps; + if (get_bits(f,1)) { + m->coupling_steps = get_bits(f,8)+1; + for (k=0; k < m->coupling_steps; ++k) { + m->chan[k].magnitude = get_bits(f, ilog(f->channels-1)); + m->chan[k].angle = get_bits(f, ilog(f->channels-1)); + if (m->chan[k].magnitude >= f->channels) return error(f, VORBIS_invalid_setup); + if (m->chan[k].angle >= f->channels) return error(f, VORBIS_invalid_setup); + if (m->chan[k].magnitude == m->chan[k].angle) return error(f, VORBIS_invalid_setup); + } + } else + m->coupling_steps = 0; + + // reserved field + if (get_bits(f,2)) return error(f, VORBIS_invalid_setup); + if (m->submaps > 1) { + for (j=0; j < f->channels; ++j) { + m->chan[j].mux = get_bits(f, 4); + if (m->chan[j].mux >= m->submaps) return error(f, VORBIS_invalid_setup); + } + } else + // @SPECIFICATION: this case is missing from the spec + for (j=0; j < f->channels; ++j) + m->chan[j].mux = 0; + + for (j=0; j < m->submaps; ++j) { + get_bits(f,8); // discard + m->submap_floor[j] = get_bits(f,8); + m->submap_residue[j] = get_bits(f,8); + if (m->submap_floor[j] >= f->floor_count) return error(f, VORBIS_invalid_setup); + if (m->submap_residue[j] >= f->residue_count) return error(f, VORBIS_invalid_setup); + } + } + + // Modes + f->mode_count = get_bits(f, 6)+1; + for (i=0; i < f->mode_count; ++i) { + Mode *m = f->mode_config+i; + m->blockflag = get_bits(f,1); + m->windowtype = get_bits(f,16); + m->transformtype = get_bits(f,16); + m->mapping = get_bits(f,8); + if (m->windowtype != 0) return error(f, VORBIS_invalid_setup); + if (m->transformtype != 0) return error(f, VORBIS_invalid_setup); + if (m->mapping >= f->mapping_count) return error(f, VORBIS_invalid_setup); + } + + flush_packet(f); + + f->previous_length = 0; + + for (i=0; i < f->channels; ++i) { + f->channel_buffers[i] = (float *) setup_malloc(f, sizeof(float) * f->blocksize_1); + f->previous_window[i] = (float *) setup_malloc(f, sizeof(float) * f->blocksize_1/2); + f->finalY[i] = (int16 *) setup_malloc(f, sizeof(int16) * longest_floorlist); + if (f->channel_buffers[i] == NULL || f->previous_window[i] == NULL || f->finalY[i] == NULL) return error(f, VORBIS_outofmem); + #ifdef STB_VORBIS_NO_DEFER_FLOOR + f->floor_buffers[i] = (float *) setup_malloc(f, sizeof(float) * f->blocksize_1/2); + if (f->floor_buffers[i] == NULL) return error(f, VORBIS_outofmem); + #endif + } + + if (!init_blocksize(f, 0, f->blocksize_0)) return FALSE; + if (!init_blocksize(f, 1, f->blocksize_1)) return FALSE; + f->blocksize[0] = f->blocksize_0; + f->blocksize[1] = f->blocksize_1; + +#ifdef STB_VORBIS_DIVIDE_TABLE + if (integer_divide_table[1][1]==0) + for (i=0; i < DIVTAB_NUMER; ++i) + for (j=1; j < DIVTAB_DENOM; ++j) + integer_divide_table[i][j] = i / j; +#endif + + // compute how much temporary memory is needed + + // 1. + { + uint32 imdct_mem = (f->blocksize_1 * sizeof(float) >> 1); + uint32 classify_mem; + int i,max_part_read=0; + for (i=0; i < f->residue_count; ++i) { + Residue *r = f->residue_config + i; + int n_read = r->end - r->begin; + int part_read = n_read / r->part_size; + if (part_read > max_part_read) + max_part_read = part_read; + } + #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + classify_mem = f->channels * (sizeof(void*) + max_part_read * sizeof(uint8 *)); + #else + classify_mem = f->channels * (sizeof(void*) + max_part_read * sizeof(int *)); + #endif + + f->temp_memory_required = classify_mem; + if (imdct_mem > f->temp_memory_required) + f->temp_memory_required = imdct_mem; + } + + f->first_decode = TRUE; + + if (f->alloc.alloc_buffer) { + assert(f->temp_offset == f->alloc.alloc_buffer_length_in_bytes); + // check if there's enough temp memory so we don't error later + if (f->setup_offset + sizeof(*f) + f->temp_memory_required > (unsigned) f->temp_offset) + return error(f, VORBIS_outofmem); + } + + f->first_audio_page_offset = stb_vorbis_get_file_offset(f); + + return TRUE; +} + +static void vorbis_deinit(stb_vorbis *p) +{ + int i,j; + if (p->residue_config) { + for (i=0; i < p->residue_count; ++i) { + Residue *r = p->residue_config+i; + if (r->classdata) { + for (j=0; j < p->codebooks[r->classbook].entries; ++j) + setup_free(p, r->classdata[j]); + setup_free(p, r->classdata); + } + setup_free(p, r->residue_books); + } + } + + if (p->codebooks) { + CHECK(p); + for (i=0; i < p->codebook_count; ++i) { + Codebook *c = p->codebooks + i; + setup_free(p, c->codeword_lengths); + setup_free(p, c->multiplicands); + setup_free(p, c->codewords); + setup_free(p, c->sorted_codewords); + // c->sorted_values[-1] is the first entry in the array + setup_free(p, c->sorted_values ? c->sorted_values-1 : NULL); + } + setup_free(p, p->codebooks); + } + setup_free(p, p->floor_config); + setup_free(p, p->residue_config); + if (p->mapping) { + for (i=0; i < p->mapping_count; ++i) + setup_free(p, p->mapping[i].chan); + setup_free(p, p->mapping); + } + CHECK(p); + for (i=0; i < p->channels && i < STB_VORBIS_MAX_CHANNELS; ++i) { + setup_free(p, p->channel_buffers[i]); + setup_free(p, p->previous_window[i]); + #ifdef STB_VORBIS_NO_DEFER_FLOOR + setup_free(p, p->floor_buffers[i]); + #endif + setup_free(p, p->finalY[i]); + } + for (i=0; i < 2; ++i) { + setup_free(p, p->A[i]); + setup_free(p, p->B[i]); + setup_free(p, p->C[i]); + setup_free(p, p->window[i]); + setup_free(p, p->bit_reverse[i]); + } + #ifndef STB_VORBIS_NO_STDIO + if (p->close_on_free) fclose(p->f); + #endif +} + +void stb_vorbis_close(stb_vorbis *p) +{ + if (p == NULL) return; + vorbis_deinit(p); + setup_free(p,p); +} + +static void vorbis_init(stb_vorbis *p, const stb_vorbis_alloc *z) +{ + memset(p, 0, sizeof(*p)); // NULL out all malloc'd pointers to start + if (z) { + p->alloc = *z; + p->alloc.alloc_buffer_length_in_bytes = (p->alloc.alloc_buffer_length_in_bytes+3) & ~3; + p->temp_offset = p->alloc.alloc_buffer_length_in_bytes; + } + p->eof = 0; + p->error = VORBIS__no_error; + p->stream = NULL; + p->codebooks = NULL; + p->page_crc_tests = -1; + #ifndef STB_VORBIS_NO_STDIO + p->close_on_free = FALSE; + p->f = NULL; + #endif +} + +int stb_vorbis_get_sample_offset(stb_vorbis *f) +{ + if (f->current_loc_valid) + return f->current_loc; + else + return -1; +} + +stb_vorbis_info stb_vorbis_get_info(stb_vorbis *f) +{ + stb_vorbis_info d; + d.channels = f->channels; + d.sample_rate = f->sample_rate; + d.setup_memory_required = f->setup_memory_required; + d.setup_temp_memory_required = f->setup_temp_memory_required; + d.temp_memory_required = f->temp_memory_required; + d.max_frame_size = f->blocksize_1 >> 1; + return d; +} + +int stb_vorbis_get_error(stb_vorbis *f) +{ + int e = f->error; + f->error = VORBIS__no_error; + return e; +} + +static stb_vorbis * vorbis_alloc(stb_vorbis *f) +{ + stb_vorbis *p = (stb_vorbis *) setup_malloc(f, sizeof(*p)); + return p; +} + +#ifndef STB_VORBIS_NO_PUSHDATA_API + +void stb_vorbis_flush_pushdata(stb_vorbis *f) +{ + f->previous_length = 0; + f->page_crc_tests = 0; + f->discard_samples_deferred = 0; + f->current_loc_valid = FALSE; + f->first_decode = FALSE; + f->samples_output = 0; + f->channel_buffer_start = 0; + f->channel_buffer_end = 0; +} + +static int vorbis_search_for_page_pushdata(vorb *f, uint8 *data, int data_len) +{ + int i,n; + for (i=0; i < f->page_crc_tests; ++i) + f->scan[i].bytes_done = 0; + + // if we have room for more scans, search for them first, because + // they may cause us to stop early if their header is incomplete + if (f->page_crc_tests < STB_VORBIS_PUSHDATA_CRC_COUNT) { + if (data_len < 4) return 0; + data_len -= 3; // need to look for 4-byte sequence, so don't miss + // one that straddles a boundary + for (i=0; i < data_len; ++i) { + if (data[i] == 0x4f) { + if (0==memcmp(data+i, ogg_page_header, 4)) { + int j,len; + uint32 crc; + // make sure we have the whole page header + if (i+26 >= data_len || i+27+data[i+26] >= data_len) { + // only read up to this page start, so hopefully we'll + // have the whole page header start next time + data_len = i; + break; + } + // ok, we have it all; compute the length of the page + len = 27 + data[i+26]; + for (j=0; j < data[i+26]; ++j) + len += data[i+27+j]; + // scan everything up to the embedded crc (which we must 0) + crc = 0; + for (j=0; j < 22; ++j) + crc = crc32_update(crc, data[i+j]); + // now process 4 0-bytes + for ( ; j < 26; ++j) + crc = crc32_update(crc, 0); + // len is the total number of bytes we need to scan + n = f->page_crc_tests++; + f->scan[n].bytes_left = len-j; + f->scan[n].crc_so_far = crc; + f->scan[n].goal_crc = data[i+22] + (data[i+23] << 8) + (data[i+24]<<16) + (data[i+25]<<24); + // if the last frame on a page is continued to the next, then + // we can't recover the sample_loc immediately + if (data[i+27+data[i+26]-1] == 255) + f->scan[n].sample_loc = ~0; + else + f->scan[n].sample_loc = data[i+6] + (data[i+7] << 8) + (data[i+ 8]<<16) + (data[i+ 9]<<24); + f->scan[n].bytes_done = i+j; + if (f->page_crc_tests == STB_VORBIS_PUSHDATA_CRC_COUNT) + break; + // keep going if we still have room for more + } + } + } + } + + for (i=0; i < f->page_crc_tests;) { + uint32 crc; + int j; + int n = f->scan[i].bytes_done; + int m = f->scan[i].bytes_left; + if (m > data_len - n) m = data_len - n; + // m is the bytes to scan in the current chunk + crc = f->scan[i].crc_so_far; + for (j=0; j < m; ++j) + crc = crc32_update(crc, data[n+j]); + f->scan[i].bytes_left -= m; + f->scan[i].crc_so_far = crc; + if (f->scan[i].bytes_left == 0) { + // does it match? + if (f->scan[i].crc_so_far == f->scan[i].goal_crc) { + // Houston, we have page + data_len = n+m; // consumption amount is wherever that scan ended + f->page_crc_tests = -1; // drop out of page scan mode + f->previous_length = 0; // decode-but-don't-output one frame + f->next_seg = -1; // start a new page + f->current_loc = f->scan[i].sample_loc; // set the current sample location + // to the amount we'd have decoded had we decoded this page + f->current_loc_valid = f->current_loc != ~0U; + return data_len; + } + // delete entry + f->scan[i] = f->scan[--f->page_crc_tests]; + } else { + ++i; + } + } + + return data_len; +} + +// return value: number of bytes we used +int stb_vorbis_decode_frame_pushdata( + stb_vorbis *f, // the file we're decoding + const uint8 *data, int data_len, // the memory available for decoding + int *channels, // place to write number of float * buffers + float ***output, // place to write float ** array of float * buffers + int *samples // place to write number of output samples + ) +{ + int i; + int len,right,left; + + if (!IS_PUSH_MODE(f)) return error(f, VORBIS_invalid_api_mixing); + + if (f->page_crc_tests >= 0) { + *samples = 0; + return vorbis_search_for_page_pushdata(f, (uint8 *) data, data_len); + } + + f->stream = (uint8 *) data; + f->stream_end = (uint8 *) data + data_len; + f->error = VORBIS__no_error; + + // check that we have the entire packet in memory + if (!is_whole_packet_present(f, FALSE)) { + *samples = 0; + return 0; + } + + if (!vorbis_decode_packet(f, &len, &left, &right)) { + // save the actual error we encountered + enum STBVorbisError error = f->error; + if (error == VORBIS_bad_packet_type) { + // flush and resynch + f->error = VORBIS__no_error; + while (get8_packet(f) != EOP) + if (f->eof) break; + *samples = 0; + return f->stream - data; + } + if (error == VORBIS_continued_packet_flag_invalid) { + if (f->previous_length == 0) { + // we may be resynching, in which case it's ok to hit one + // of these; just discard the packet + f->error = VORBIS__no_error; + while (get8_packet(f) != EOP) + if (f->eof) break; + *samples = 0; + return f->stream - data; + } + } + // if we get an error while parsing, what to do? + // well, it DEFINITELY won't work to continue from where we are! + stb_vorbis_flush_pushdata(f); + // restore the error that actually made us bail + f->error = error; + *samples = 0; + return 1; + } + + // success! + len = vorbis_finish_frame(f, len, left, right); + for (i=0; i < f->channels; ++i) + f->outputs[i] = f->channel_buffers[i] + left; + + if (channels) *channels = f->channels; + *samples = len; + *output = f->outputs; + return f->stream - data; +} + +stb_vorbis *stb_vorbis_open_pushdata( + const unsigned char *data, int data_len, // the memory available for decoding + int *data_used, // only defined if result is not NULL + int *error, const stb_vorbis_alloc *alloc) +{ + stb_vorbis *f, p; + vorbis_init(&p, alloc); + p.stream = (uint8 *) data; + p.stream_end = (uint8 *) data + data_len; + p.push_mode = TRUE; + if (!start_decoder(&p)) { + if (p.eof) + *error = VORBIS_need_more_data; + else + *error = p.error; + return NULL; + } + f = vorbis_alloc(&p); + if (f) { + *f = p; + *data_used = f->stream - data; + *error = 0; + return f; + } else { + vorbis_deinit(&p); + return NULL; + } +} +#endif // STB_VORBIS_NO_PUSHDATA_API + +unsigned int stb_vorbis_get_file_offset(stb_vorbis *f) +{ + #ifndef STB_VORBIS_NO_PUSHDATA_API + if (f->push_mode) return 0; + #endif + if (USE_MEMORY(f)) return f->stream - f->stream_start; + #ifndef STB_VORBIS_NO_STDIO + return ftell(f->f) - f->f_start; + #endif +} + +#ifndef STB_VORBIS_NO_PULLDATA_API +// +// DATA-PULLING API +// + +static uint32 vorbis_find_page(stb_vorbis *f, uint32 *end, uint32 *last) +{ + for(;;) { + int n; + if (f->eof) return 0; + n = get8(f); + if (n == 0x4f) { // page header candidate + unsigned int retry_loc = stb_vorbis_get_file_offset(f); + int i; + // check if we're off the end of a file_section stream + if (retry_loc - 25 > f->stream_len) + return 0; + // check the rest of the header + for (i=1; i < 4; ++i) + if (get8(f) != ogg_page_header[i]) + break; + if (f->eof) return 0; + if (i == 4) { + uint8 header[27]; + uint32 i, crc, goal, len; + for (i=0; i < 4; ++i) + header[i] = ogg_page_header[i]; + for (; i < 27; ++i) + header[i] = get8(f); + if (f->eof) return 0; + if (header[4] != 0) goto invalid; + goal = header[22] + (header[23] << 8) + (header[24]<<16) + (header[25]<<24); + for (i=22; i < 26; ++i) + header[i] = 0; + crc = 0; + for (i=0; i < 27; ++i) + crc = crc32_update(crc, header[i]); + len = 0; + for (i=0; i < header[26]; ++i) { + int s = get8(f); + crc = crc32_update(crc, s); + len += s; + } + if (len && f->eof) return 0; + for (i=0; i < len; ++i) + crc = crc32_update(crc, get8(f)); + // finished parsing probable page + if (crc == goal) { + // we could now check that it's either got the last + // page flag set, OR it's followed by the capture + // pattern, but I guess TECHNICALLY you could have + // a file with garbage between each ogg page and recover + // from it automatically? So even though that paranoia + // might decrease the chance of an invalid decode by + // another 2^32, not worth it since it would hose those + // invalid-but-useful files? + if (end) + *end = stb_vorbis_get_file_offset(f); + if (last) { + if (header[5] & 0x04) + *last = 1; + else + *last = 0; + } + set_file_offset(f, retry_loc-1); + return 1; + } + } + invalid: + // not a valid page, so rewind and look for next one + set_file_offset(f, retry_loc); + } + } +} + + +#define SAMPLE_unknown 0xffffffff + +// seeking is implemented with a binary search, which narrows down the range to +// 64K, before using a linear search (because finding the synchronization +// pattern can be expensive, and the chance we'd find the end page again is +// relatively high for small ranges) +// +// two initial interpolation-style probes are used at the start of the search +// to try to bound either side of the binary search sensibly, while still +// working in O(log n) time if they fail. + +static int get_seek_page_info(stb_vorbis *f, ProbedPage *z) +{ + uint8 header[27], lacing[255]; + int i,len; + + // record where the page starts + z->page_start = stb_vorbis_get_file_offset(f); + + // parse the header + getn(f, header, 27); + if (header[0] != 'O' || header[1] != 'g' || header[2] != 'g' || header[3] != 'S') + return 0; + getn(f, lacing, header[26]); + + // determine the length of the payload + len = 0; + for (i=0; i < header[26]; ++i) + len += lacing[i]; + + // this implies where the page ends + z->page_end = z->page_start + 27 + header[26] + len; + + // read the last-decoded sample out of the data + z->last_decoded_sample = header[6] + (header[7] << 8) + (header[8] << 16) + (header[9] << 24); + + // restore file state to where we were + set_file_offset(f, z->page_start); + return 1; +} + +// rarely used function to seek back to the preceeding page while finding the +// start of a packet +static int go_to_page_before(stb_vorbis *f, unsigned int limit_offset) +{ + unsigned int previous_safe, end; + + // now we want to seek back 64K from the limit + if (limit_offset >= 65536 && limit_offset-65536 >= f->first_audio_page_offset) + previous_safe = limit_offset - 65536; + else + previous_safe = f->first_audio_page_offset; + + set_file_offset(f, previous_safe); + + while (vorbis_find_page(f, &end, NULL)) { + if (end >= limit_offset && stb_vorbis_get_file_offset(f) < limit_offset) + return 1; + set_file_offset(f, end); + } + + return 0; +} + +// implements the search logic for finding a page and starting decoding. if +// the function succeeds, current_loc_valid will be true and current_loc will +// be less than or equal to the provided sample number (the closer the +// better). +static int seek_to_sample_coarse(stb_vorbis *f, uint32 sample_number) +{ + ProbedPage left, right, mid; + int i, start_seg_with_known_loc, end_pos, page_start; + uint32 delta, stream_length, padding; + double offset, bytes_per_sample; + int probe = 0; + + // find the last page and validate the target sample + stream_length = stb_vorbis_stream_length_in_samples(f); + if (stream_length == 0) return error(f, VORBIS_seek_without_length); + if (sample_number > stream_length) return error(f, VORBIS_seek_invalid); + + // this is the maximum difference between the window-center (which is the + // actual granule position value), and the right-start (which the spec + // indicates should be the granule position (give or take one)). + padding = ((f->blocksize_1 - f->blocksize_0) >> 2); + if (sample_number < padding) + sample_number = 0; + else + sample_number -= padding; + + left = f->p_first; + while (left.last_decoded_sample == ~0U) { + // (untested) the first page does not have a 'last_decoded_sample' + set_file_offset(f, left.page_end); + if (!get_seek_page_info(f, &left)) goto error; + } + + right = f->p_last; + assert(right.last_decoded_sample != ~0U); + + // starting from the start is handled differently + if (sample_number <= left.last_decoded_sample) { + stb_vorbis_seek_start(f); + return 1; + } + + while (left.page_end != right.page_start) { + assert(left.page_end < right.page_start); + // search range in bytes + delta = right.page_start - left.page_end; + if (delta <= 65536) { + // there's only 64K left to search - handle it linearly + set_file_offset(f, left.page_end); + } else { + if (probe < 2) { + if (probe == 0) { + // first probe (interpolate) + double data_bytes = right.page_end - left.page_start; + bytes_per_sample = data_bytes / right.last_decoded_sample; + offset = left.page_start + bytes_per_sample * (sample_number - left.last_decoded_sample); + } else { + // second probe (try to bound the other side) + double error = ((double) sample_number - mid.last_decoded_sample) * bytes_per_sample; + if (error >= 0 && error < 8000) error = 8000; + if (error < 0 && error > -8000) error = -8000; + offset += error * 2; + } + + // ensure the offset is valid + if (offset < left.page_end) + offset = left.page_end; + if (offset > right.page_start - 65536) + offset = right.page_start - 65536; + + set_file_offset(f, (unsigned int) offset); + } else { + // binary search for large ranges (offset by 32K to ensure + // we don't hit the right page) + set_file_offset(f, left.page_end + (delta / 2) - 32768); + } + + if (!vorbis_find_page(f, NULL, NULL)) goto error; + } + + for (;;) { + if (!get_seek_page_info(f, &mid)) goto error; + if (mid.last_decoded_sample != ~0U) break; + // (untested) no frames end on this page + set_file_offset(f, mid.page_end); + assert(mid.page_start < right.page_start); + } + + // if we've just found the last page again then we're in a tricky file, + // and we're close enough. + if (mid.page_start == right.page_start) + break; + + if (sample_number < mid.last_decoded_sample) + right = mid; + else + left = mid; + + ++probe; + } + + // seek back to start of the last packet + page_start = left.page_start; + set_file_offset(f, page_start); + if (!start_page(f)) return error(f, VORBIS_seek_failed); + end_pos = f->end_seg_with_known_loc; + assert(end_pos >= 0); + + for (;;) { + for (i = end_pos; i > 0; --i) + if (f->segments[i-1] != 255) + break; + + start_seg_with_known_loc = i; + + if (start_seg_with_known_loc > 0 || !(f->page_flag & PAGEFLAG_continued_packet)) + break; + + // (untested) the final packet begins on an earlier page + if (!go_to_page_before(f, page_start)) + goto error; + + page_start = stb_vorbis_get_file_offset(f); + if (!start_page(f)) goto error; + end_pos = f->segment_count - 1; + } + + // prepare to start decoding + f->current_loc_valid = FALSE; + f->last_seg = FALSE; + f->valid_bits = 0; + f->packet_bytes = 0; + f->bytes_in_seg = 0; + f->previous_length = 0; + f->next_seg = start_seg_with_known_loc; + + for (i = 0; i < start_seg_with_known_loc; i++) + skip(f, f->segments[i]); + + // start decoding (optimizable - this frame is generally discarded) + vorbis_pump_first_frame(f); + return 1; + +error: + // try to restore the file to a valid state + stb_vorbis_seek_start(f); + return error(f, VORBIS_seek_failed); +} + +// the same as vorbis_decode_initial, but without advancing +static int peek_decode_initial(vorb *f, int *p_left_start, int *p_left_end, int *p_right_start, int *p_right_end, int *mode) +{ + int bits_read, bytes_read; + + if (!vorbis_decode_initial(f, p_left_start, p_left_end, p_right_start, p_right_end, mode)) + return 0; + + // either 1 or 2 bytes were read, figure out which so we can rewind + bits_read = 1 + ilog(f->mode_count-1); + if (f->mode_config[*mode].blockflag) + bits_read += 2; + bytes_read = (bits_read + 7) / 8; + + f->bytes_in_seg += bytes_read; + f->packet_bytes -= bytes_read; + skip(f, -bytes_read); + if (f->next_seg == -1) + f->next_seg = f->segment_count - 1; + else + f->next_seg--; + f->valid_bits = 0; + + return 1; +} + +int stb_vorbis_seek_frame(stb_vorbis *f, unsigned int sample_number) +{ + uint32 max_frame_samples; + + if (IS_PUSH_MODE(f)) return error(f, VORBIS_invalid_api_mixing); + + // fast page-level search + if (!seek_to_sample_coarse(f, sample_number)) + return 0; + + assert(f->current_loc_valid); + assert(f->current_loc <= sample_number); + + // linear search for the relevant packet + max_frame_samples = (f->blocksize_1*3 - f->blocksize_0) >> 2; + while (f->current_loc < sample_number) { + int left_start, left_end, right_start, right_end, mode, frame_samples; + if (!peek_decode_initial(f, &left_start, &left_end, &right_start, &right_end, &mode)) + return error(f, VORBIS_seek_failed); + // calculate the number of samples returned by the next frame + frame_samples = right_start - left_start; + if (f->current_loc + frame_samples > sample_number) { + return 1; // the next frame will contain the sample + } else if (f->current_loc + frame_samples + max_frame_samples > sample_number) { + // there's a chance the frame after this could contain the sample + vorbis_pump_first_frame(f); + } else { + // this frame is too early to be relevant + f->current_loc += frame_samples; + f->previous_length = 0; + maybe_start_packet(f); + flush_packet(f); + } + } + // the next frame will start with the sample + assert(f->current_loc == sample_number); + return 1; +} + +int stb_vorbis_seek(stb_vorbis *f, unsigned int sample_number) +{ + if (!stb_vorbis_seek_frame(f, sample_number)) + return 0; + + if (sample_number != f->current_loc) { + int n; + uint32 frame_start = f->current_loc; + stb_vorbis_get_frame_float(f, &n, NULL); + assert(sample_number > frame_start); + assert(f->channel_buffer_start + (int) (sample_number-frame_start) <= f->channel_buffer_end); + f->channel_buffer_start += (sample_number - frame_start); + } + + return 1; +} + +void stb_vorbis_seek_start(stb_vorbis *f) +{ + if (IS_PUSH_MODE(f)) { error(f, VORBIS_invalid_api_mixing); return; } + set_file_offset(f, f->first_audio_page_offset); + f->previous_length = 0; + f->first_decode = TRUE; + f->next_seg = -1; + vorbis_pump_first_frame(f); +} + +unsigned int stb_vorbis_stream_length_in_samples(stb_vorbis *f) +{ + unsigned int restore_offset, previous_safe; + unsigned int end, last_page_loc; + + if (IS_PUSH_MODE(f)) return error(f, VORBIS_invalid_api_mixing); + if (!f->total_samples) { + unsigned int last; + uint32 lo,hi; + char header[6]; + + // first, store the current decode position so we can restore it + restore_offset = stb_vorbis_get_file_offset(f); + + // now we want to seek back 64K from the end (the last page must + // be at most a little less than 64K, but let's allow a little slop) + if (f->stream_len >= 65536 && f->stream_len-65536 >= f->first_audio_page_offset) + previous_safe = f->stream_len - 65536; + else + previous_safe = f->first_audio_page_offset; + + set_file_offset(f, previous_safe); + // previous_safe is now our candidate 'earliest known place that seeking + // to will lead to the final page' + + if (!vorbis_find_page(f, &end, &last)) { + // if we can't find a page, we're hosed! + f->error = VORBIS_cant_find_last_page; + f->total_samples = 0xffffffff; + goto done; + } + + // check if there are more pages + last_page_loc = stb_vorbis_get_file_offset(f); + + // stop when the last_page flag is set, not when we reach eof; + // this allows us to stop short of a 'file_section' end without + // explicitly checking the length of the section + while (!last) { + set_file_offset(f, end); + if (!vorbis_find_page(f, &end, &last)) { + // the last page we found didn't have the 'last page' flag + // set. whoops! + break; + } + previous_safe = last_page_loc+1; + last_page_loc = stb_vorbis_get_file_offset(f); + } + + set_file_offset(f, last_page_loc); + + // parse the header + getn(f, (unsigned char *)header, 6); + // extract the absolute granule position + lo = get32(f); + hi = get32(f); + if (lo == 0xffffffff && hi == 0xffffffff) { + f->error = VORBIS_cant_find_last_page; + f->total_samples = SAMPLE_unknown; + goto done; + } + if (hi) + lo = 0xfffffffe; // saturate + f->total_samples = lo; + + f->p_last.page_start = last_page_loc; + f->p_last.page_end = end; + f->p_last.last_decoded_sample = lo; + + done: + set_file_offset(f, restore_offset); + } + return f->total_samples == SAMPLE_unknown ? 0 : f->total_samples; +} + +float stb_vorbis_stream_length_in_seconds(stb_vorbis *f) +{ + return stb_vorbis_stream_length_in_samples(f) / (float) f->sample_rate; +} + + + +int stb_vorbis_get_frame_float(stb_vorbis *f, int *channels, float ***output) +{ + int len, right,left,i; + if (IS_PUSH_MODE(f)) return error(f, VORBIS_invalid_api_mixing); + + if (!vorbis_decode_packet(f, &len, &left, &right)) { + f->channel_buffer_start = f->channel_buffer_end = 0; + return 0; + } + + len = vorbis_finish_frame(f, len, left, right); + for (i=0; i < f->channels; ++i) + f->outputs[i] = f->channel_buffers[i] + left; + + f->channel_buffer_start = left; + f->channel_buffer_end = left+len; + + if (channels) *channels = f->channels; + if (output) *output = f->outputs; + return len; +} + +#ifndef STB_VORBIS_NO_STDIO + +stb_vorbis * stb_vorbis_open_file_section(FILE *file, int close_on_free, int *error, const stb_vorbis_alloc *alloc, unsigned int length) +{ + stb_vorbis *f, p; + vorbis_init(&p, alloc); + p.f = file; + p.f_start = ftell(file); + p.stream_len = length; + p.close_on_free = close_on_free; + if (start_decoder(&p)) { + f = vorbis_alloc(&p); + if (f) { + *f = p; + vorbis_pump_first_frame(f); + return f; + } + } + if (error) *error = p.error; + vorbis_deinit(&p); + return NULL; +} + +stb_vorbis * stb_vorbis_open_file(FILE *file, int close_on_free, int *error, const stb_vorbis_alloc *alloc) +{ + unsigned int len, start; + start = ftell(file); + fseek(file, 0, SEEK_END); + len = ftell(file) - start; + fseek(file, start, SEEK_SET); + return stb_vorbis_open_file_section(file, close_on_free, error, alloc, len); +} + +stb_vorbis * stb_vorbis_open_filename(const char *filename, int *error, const stb_vorbis_alloc *alloc) +{ + FILE *f = fopen(filename, "rb"); + if (f) + return stb_vorbis_open_file(f, TRUE, error, alloc); + if (error) *error = VORBIS_file_open_failure; + return NULL; +} +#endif // STB_VORBIS_NO_STDIO + +stb_vorbis * stb_vorbis_open_memory(const unsigned char *data, int len, int *error, const stb_vorbis_alloc *alloc) +{ + stb_vorbis *f, p; + if (data == NULL) return NULL; + vorbis_init(&p, alloc); + p.stream = (uint8 *) data; + p.stream_end = (uint8 *) data + len; + p.stream_start = (uint8 *) p.stream; + p.stream_len = len; + p.push_mode = FALSE; + if (start_decoder(&p)) { + f = vorbis_alloc(&p); + if (f) { + *f = p; + vorbis_pump_first_frame(f); + return f; + } + } + if (error) *error = p.error; + vorbis_deinit(&p); + return NULL; +} + +#ifndef STB_VORBIS_NO_INTEGER_CONVERSION +#define PLAYBACK_MONO 1 +#define PLAYBACK_LEFT 2 +#define PLAYBACK_RIGHT 4 + +#define L (PLAYBACK_LEFT | PLAYBACK_MONO) +#define C (PLAYBACK_LEFT | PLAYBACK_RIGHT | PLAYBACK_MONO) +#define R (PLAYBACK_RIGHT | PLAYBACK_MONO) + +static int8 channel_position[7][6] = +{ + { 0 }, + { C }, + { L, R }, + { L, C, R }, + { L, R, L, R }, + { L, C, R, L, R }, + { L, C, R, L, R, C }, +}; + + +#ifndef STB_VORBIS_NO_FAST_SCALED_FLOAT + typedef union { + float f; + int i; + } float_conv; + typedef char stb_vorbis_float_size_test[sizeof(float)==4 && sizeof(int) == 4]; + #define FASTDEF(x) float_conv x + // add (1<<23) to convert to int, then divide by 2^SHIFT, then add 0.5/2^SHIFT to round + #define MAGIC(SHIFT) (1.5f * (1 << (23-SHIFT)) + 0.5f/(1 << SHIFT)) + #define ADDEND(SHIFT) (((150-SHIFT) << 23) + (1 << 22)) + #define FAST_SCALED_FLOAT_TO_INT(temp,x,s) (temp.f = (x) + MAGIC(s), temp.i - ADDEND(s)) + #define check_endianness() +#else + #define FAST_SCALED_FLOAT_TO_INT(temp,x,s) ((int) ((x) * (1 << (s)))) + #define check_endianness() + #define FASTDEF(x) +#endif + +static void copy_samples(short *dest, float *src, int len) +{ + int i; + check_endianness(); + for (i=0; i < len; ++i) { + FASTDEF(temp); + int v = FAST_SCALED_FLOAT_TO_INT(temp, src[i],15); + if ((unsigned int) (v + 32768) > 65535) + v = v < 0 ? -32768 : 32767; + dest[i] = v; + } +} + +static void compute_samples(int mask, short *output, int num_c, float **data, int d_offset, int len) +{ + #define BUFFER_SIZE 32 + float buffer[BUFFER_SIZE]; + int i,j,o,n = BUFFER_SIZE; + check_endianness(); + for (o = 0; o < len; o += BUFFER_SIZE) { + memset(buffer, 0, sizeof(buffer)); + if (o + n > len) n = len - o; + for (j=0; j < num_c; ++j) { + if (channel_position[num_c][j] & mask) { + for (i=0; i < n; ++i) + buffer[i] += data[j][d_offset+o+i]; + } + } + for (i=0; i < n; ++i) { + FASTDEF(temp); + int v = FAST_SCALED_FLOAT_TO_INT(temp,buffer[i],15); + if ((unsigned int) (v + 32768) > 65535) + v = v < 0 ? -32768 : 32767; + output[o+i] = v; + } + } +} + +static void compute_stereo_samples(short *output, int num_c, float **data, int d_offset, int len) +{ + #define BUFFER_SIZE 32 + float buffer[BUFFER_SIZE]; + int i,j,o,n = BUFFER_SIZE >> 1; + // o is the offset in the source data + check_endianness(); + for (o = 0; o < len; o += BUFFER_SIZE >> 1) { + // o2 is the offset in the output data + int o2 = o << 1; + memset(buffer, 0, sizeof(buffer)); + if (o + n > len) n = len - o; + for (j=0; j < num_c; ++j) { + int m = channel_position[num_c][j] & (PLAYBACK_LEFT | PLAYBACK_RIGHT); + if (m == (PLAYBACK_LEFT | PLAYBACK_RIGHT)) { + for (i=0; i < n; ++i) { + buffer[i*2+0] += data[j][d_offset+o+i]; + buffer[i*2+1] += data[j][d_offset+o+i]; + } + } else if (m == PLAYBACK_LEFT) { + for (i=0; i < n; ++i) { + buffer[i*2+0] += data[j][d_offset+o+i]; + } + } else if (m == PLAYBACK_RIGHT) { + for (i=0; i < n; ++i) { + buffer[i*2+1] += data[j][d_offset+o+i]; + } + } + } + for (i=0; i < (n<<1); ++i) { + FASTDEF(temp); + int v = FAST_SCALED_FLOAT_TO_INT(temp,buffer[i],15); + if ((unsigned int) (v + 32768) > 65535) + v = v < 0 ? -32768 : 32767; + output[o2+i] = v; + } + } +} + +static void convert_samples_short(int buf_c, short **buffer, int b_offset, int data_c, float **data, int d_offset, int samples) +{ + int i; + if (buf_c != data_c && buf_c <= 2 && data_c <= 6) { + static int channel_selector[3][2] = { {0}, {PLAYBACK_MONO}, {PLAYBACK_LEFT, PLAYBACK_RIGHT} }; + for (i=0; i < buf_c; ++i) + compute_samples(channel_selector[buf_c][i], buffer[i]+b_offset, data_c, data, d_offset, samples); + } else { + int limit = buf_c < data_c ? buf_c : data_c; + for (i=0; i < limit; ++i) + copy_samples(buffer[i]+b_offset, data[i]+d_offset, samples); + for ( ; i < buf_c; ++i) + memset(buffer[i]+b_offset, 0, sizeof(short) * samples); + } +} + +int stb_vorbis_get_frame_short(stb_vorbis *f, int num_c, short **buffer, int num_samples) +{ + float **output; + int len = stb_vorbis_get_frame_float(f, NULL, &output); + if (len > num_samples) len = num_samples; + if (len) + convert_samples_short(num_c, buffer, 0, f->channels, output, 0, len); + return len; +} + +static void convert_channels_short_interleaved(int buf_c, short *buffer, int data_c, float **data, int d_offset, int len) +{ + int i; + check_endianness(); + if (buf_c != data_c && buf_c <= 2 && data_c <= 6) { + assert(buf_c == 2); + for (i=0; i < buf_c; ++i) + compute_stereo_samples(buffer, data_c, data, d_offset, len); + } else { + int limit = buf_c < data_c ? buf_c : data_c; + int j; + for (j=0; j < len; ++j) { + for (i=0; i < limit; ++i) { + FASTDEF(temp); + float f = data[i][d_offset+j]; + int v = FAST_SCALED_FLOAT_TO_INT(temp, f,15);//data[i][d_offset+j],15); + if ((unsigned int) (v + 32768) > 65535) + v = v < 0 ? -32768 : 32767; + *buffer++ = v; + } + for ( ; i < buf_c; ++i) + *buffer++ = 0; + } + } +} + +int stb_vorbis_get_frame_short_interleaved(stb_vorbis *f, int num_c, short *buffer, int num_shorts) +{ + float **output; + int len; + if (num_c == 1) return stb_vorbis_get_frame_short(f,num_c,&buffer, num_shorts); + len = stb_vorbis_get_frame_float(f, NULL, &output); + if (len) { + if (len*num_c > num_shorts) len = num_shorts / num_c; + convert_channels_short_interleaved(num_c, buffer, f->channels, output, 0, len); + } + return len; +} + +int stb_vorbis_get_samples_short_interleaved(stb_vorbis *f, int channels, short *buffer, int num_shorts) +{ + float **outputs; + int len = num_shorts / channels; + int n=0; + int z = f->channels; + if (z > channels) z = channels; + while (n < len) { + int k = f->channel_buffer_end - f->channel_buffer_start; + if (n+k >= len) k = len - n; + if (k) + convert_channels_short_interleaved(channels, buffer, f->channels, f->channel_buffers, f->channel_buffer_start, k); + buffer += k*channels; + n += k; + f->channel_buffer_start += k; + if (n == len) break; + if (!stb_vorbis_get_frame_float(f, NULL, &outputs)) break; + } + return n; +} + +int stb_vorbis_get_samples_short(stb_vorbis *f, int channels, short **buffer, int len) +{ + float **outputs; + int n=0; + int z = f->channels; + if (z > channels) z = channels; + while (n < len) { + int k = f->channel_buffer_end - f->channel_buffer_start; + if (n+k >= len) k = len - n; + if (k) + convert_samples_short(channels, buffer, n, f->channels, f->channel_buffers, f->channel_buffer_start, k); + n += k; + f->channel_buffer_start += k; + if (n == len) break; + if (!stb_vorbis_get_frame_float(f, NULL, &outputs)) break; + } + return n; +} + +#ifndef STB_VORBIS_NO_STDIO +int stb_vorbis_decode_filename(const char *filename, int *channels, int *sample_rate, short **output) +{ + int data_len, offset, total, limit, error; + short *data; + stb_vorbis *v = stb_vorbis_open_filename(filename, &error, NULL); + if (v == NULL) return -1; + limit = v->channels * 4096; + *channels = v->channels; + if (sample_rate) + *sample_rate = v->sample_rate; + offset = data_len = 0; + total = limit; + data = (short *) malloc(total * sizeof(*data)); + if (data == NULL) { + stb_vorbis_close(v); + return -2; + } + for (;;) { + int n = stb_vorbis_get_frame_short_interleaved(v, v->channels, data+offset, total-offset); + if (n == 0) break; + data_len += n; + offset += n * v->channels; + if (offset + limit > total) { + short *data2; + total *= 2; + data2 = (short *) realloc(data, total * sizeof(*data)); + if (data2 == NULL) { + free(data); + stb_vorbis_close(v); + return -2; + } + data = data2; + } + } + *output = data; + stb_vorbis_close(v); + return data_len; +} +#endif // NO_STDIO + +int stb_vorbis_decode_memory(const uint8 *mem, int len, int *channels, int *sample_rate, short **output) +{ + int data_len, offset, total, limit, error; + short *data; + stb_vorbis *v = stb_vorbis_open_memory(mem, len, &error, NULL); + if (v == NULL) return -1; + limit = v->channels * 4096; + *channels = v->channels; + if (sample_rate) + *sample_rate = v->sample_rate; + offset = data_len = 0; + total = limit; + data = (short *) malloc(total * sizeof(*data)); + if (data == NULL) { + stb_vorbis_close(v); + return -2; + } + for (;;) { + int n = stb_vorbis_get_frame_short_interleaved(v, v->channels, data+offset, total-offset); + if (n == 0) break; + data_len += n; + offset += n * v->channels; + if (offset + limit > total) { + short *data2; + total *= 2; + data2 = (short *) realloc(data, total * sizeof(*data)); + if (data2 == NULL) { + free(data); + stb_vorbis_close(v); + return -2; + } + data = data2; + } + } + *output = data; + stb_vorbis_close(v); + return data_len; +} +#endif // STB_VORBIS_NO_INTEGER_CONVERSION + +int stb_vorbis_get_samples_float_interleaved(stb_vorbis *f, int channels, float *buffer, int num_floats) +{ + float **outputs; + int len = num_floats / channels; + int n=0; + int z = f->channels; + if (z > channels) z = channels; + while (n < len) { + int i,j; + int k = f->channel_buffer_end - f->channel_buffer_start; + if (n+k >= len) k = len - n; + for (j=0; j < k; ++j) { + for (i=0; i < z; ++i) + *buffer++ = f->channel_buffers[i][f->channel_buffer_start+j]; + for ( ; i < channels; ++i) + *buffer++ = 0; + } + n += k; + f->channel_buffer_start += k; + if (n == len) + break; + if (!stb_vorbis_get_frame_float(f, NULL, &outputs)) + break; + } + return n; +} + +int stb_vorbis_get_samples_float(stb_vorbis *f, int channels, float **buffer, int num_samples) +{ + float **outputs; + int n=0; + int z = f->channels; + if (z > channels) z = channels; + while (n < num_samples) { + int i; + int k = f->channel_buffer_end - f->channel_buffer_start; + if (n+k >= num_samples) k = num_samples - n; + if (k) { + for (i=0; i < z; ++i) + memcpy(buffer[i]+n, f->channel_buffers[i]+f->channel_buffer_start, sizeof(float)*k); + for ( ; i < channels; ++i) + memset(buffer[i]+n, 0, sizeof(float) * k); + } + n += k; + f->channel_buffer_start += k; + if (n == num_samples) + break; + if (!stb_vorbis_get_frame_float(f, NULL, &outputs)) + break; + } + return n; +} +#endif // STB_VORBIS_NO_PULLDATA_API + +/* Version history + 1.07 - 2015/01/16 - fixed some warnings, fix mingw, const-correct API + some more crash fixes when out of memory or with corrupt files + 1.06 - 2015/08/31 - full, correct support for seeking API (Dougall Johnson) + some crash fixes when out of memory or with corrupt files + 1.05 - 2015/04/19 - don't define __forceinline if it's redundant + 1.04 - 2014/08/27 - fix missing const-correct case in API + 1.03 - 2014/08/07 - Warning fixes + 1.02 - 2014/07/09 - Declare qsort compare function _cdecl on windows + 1.01 - 2014/06/18 - fix stb_vorbis_get_samples_float + 1.0 - 2014/05/26 - fix memory leaks; fix warnings; fix bugs in multichannel + (API change) report sample rate for decode-full-file funcs + 0.99996 - bracket #include for macintosh compilation by Laurent Gomila + 0.99995 - use union instead of pointer-cast for fast-float-to-int to avoid alias-optimization problem + 0.99994 - change fast-float-to-int to work in single-precision FPU mode, remove endian-dependence + 0.99993 - remove assert that fired on legal files with empty tables + 0.99992 - rewind-to-start + 0.99991 - bugfix to stb_vorbis_get_samples_short by Bernhard Wodo + 0.9999 - (should have been 0.99990) fix no-CRT support, compiling as C++ + 0.9998 - add a full-decode function with a memory source + 0.9997 - fix a bug in the read-from-FILE case in 0.9996 addition + 0.9996 - query length of vorbis stream in samples/seconds + 0.9995 - bugfix to another optimization that only happened in certain files + 0.9994 - bugfix to one of the optimizations that caused significant (but inaudible?) errors + 0.9993 - performance improvements; runs in 99% to 104% of time of reference implementation + 0.9992 - performance improvement of IMDCT; now performs close to reference implementation + 0.9991 - performance improvement of IMDCT + 0.999 - (should have been 0.9990) performance improvement of IMDCT + 0.998 - no-CRT support from Casey Muratori + 0.997 - bugfixes for bugs found by Terje Mathisen + 0.996 - bugfix: fast-huffman decode initialized incorrectly for sparse codebooks; fixing gives 10% speedup - found by Terje Mathisen + 0.995 - bugfix: fix to 'effective' overrun detection - found by Terje Mathisen + 0.994 - bugfix: garbage decode on final VQ symbol of a non-multiple - found by Terje Mathisen + 0.993 - bugfix: pushdata API required 1 extra byte for empty page (failed to consume final page if empty) - found by Terje Mathisen + 0.992 - fixes for MinGW warning + 0.991 - turn fast-float-conversion on by default + 0.990 - fix push-mode seek recovery if you seek into the headers + 0.98b - fix to bad release of 0.98 + 0.98 - fix push-mode seek recovery; robustify float-to-int and support non-fast mode + 0.97 - builds under c++ (typecasting, don't use 'class' keyword) + 0.96 - somehow MY 0.95 was right, but the web one was wrong, so here's my 0.95 rereleased as 0.96, fixes a typo in the clamping code + 0.95 - clamping code for 16-bit functions + 0.94 - not publically released + 0.93 - fixed all-zero-floor case (was decoding garbage) + 0.92 - fixed a memory leak + 0.91 - conditional compiles to omit parts of the API and the infrastructure to support them: STB_VORBIS_NO_PULLDATA_API, STB_VORBIS_NO_PUSHDATA_API, STB_VORBIS_NO_STDIO, STB_VORBIS_NO_INTEGER_CONVERSION + 0.90 - first public release +*/ + +#endif // STB_VORBIS_HEADER_ONLY diff --git a/source/pc/stb_vorbis.h b/source/pc/stb_vorbis.h new file mode 100644 index 0000000..90a1570 --- /dev/null +++ b/source/pc/stb_vorbis.h @@ -0,0 +1,2 @@ +#define STB_VORBIS_HEADER_ONLY +#include "stb_vorbis.c" \ No newline at end of file diff --git a/source/pc/wav.cpp b/source/pc/wav.cpp index 04a2a19..6ab7861 100644 --- a/source/pc/wav.cpp +++ b/source/pc/wav.cpp @@ -1,6 +1,5 @@ #include "wav.h" -#include #include #include #include @@ -18,8 +17,7 @@ bool wav_find_chunk(FILE* fd, const char* magic) { return true; } -WAV* wav_read(const char* file) { - FILE* fd = fopen(file, "r"); +WAV* wav_read(FILE* fd) { if(!fd) { printf("ERROR: Could not open WAV file: %s\n", strerror(errno)); return NULL; @@ -52,8 +50,6 @@ WAV* wav_read(const char* file) { data.data = (u8*) malloc(data.chunkSize); fread(data.data, 1, data.chunkSize, fd); - fclose(fd); - WAV* wav = (WAV*) malloc(sizeof(WAV)); wav->riff = riff; wav->format = format; diff --git a/source/pc/wav.h b/source/pc/wav.h index b82531c..ef5adae 100644 --- a/source/pc/wav.h +++ b/source/pc/wav.h @@ -1,6 +1,8 @@ #ifndef __WAV_H__ #define __WAV_H__ +#include + #include "../types.h" typedef struct { @@ -32,7 +34,7 @@ typedef struct { Data data; } WAV; -WAV* wav_read(const char* file); +WAV* wav_read(FILE* fd); void wav_free(WAV* wav); #endif \ No newline at end of file